Citation
Effect of Prosthodontic Services on Self-Rated Oral Health Outcomes

Material Information

Title:
Effect of Prosthodontic Services on Self-Rated Oral Health Outcomes
Creator:
MENG, XIAOXIAN ( Author, Primary )
Copyright Date:
2008

Subjects

Subjects / Keywords:
Dentistry ( jstor )
Dentures ( jstor )
Mastication ( jstor )
Modeling ( jstor )
Oral health ( jstor )
P values ( jstor )
Prosthodontics ( jstor )
Selection bias ( jstor )
Teeth ( jstor )
Tooth loss ( jstor )

Record Information

Source Institution:
University of Florida
Holding Location:
University of Florida
Rights Management:
Copyright Xiaoxian Meng. Permission granted to the University of Florida to digitize, archive and distribute this item for non-profit research and educational purposes. Any reuse of this item in excess of fair use or other copyright exemptions requires permission of the copyright holder.
Embargo Date:
8/31/2006

Downloads

This item is only available as the following downloads:


Full Text

PAGE 1

EFFECT OF PROSTHODONTIC SERVICES ON SELF-RATED ORAL HEALTH OUTCOMES By XIAOXIAN MENG A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2005

PAGE 2

Copyright 2005 by Xiaoxian Meng

PAGE 3

This work is dedicated to my family, who has never gone a day without supporting me.

PAGE 4

iv ACKNOWLEDGMENTS Over the past three years, I have been graced with the support and understanding of many outstanding people. I will be eternally gr ateful. I would like to thank Dr. R. Paul Duncan, my mentor and dissertation chair, for his complete support, continuing commitment, and encouragement with this dissertation and throughout my doctoral studies. I also thank Dr. Marc W. Heft for his insights, understanding, and caring during the process of working with him. I would like to thank Dr. Gregg H. Gilbert for his intelligent guidance and valuable discussions during the dissertation process. I wish to acknowledge Dr. Jeffrey S. Harman for sharing his knowledge in econometrics, especially his expertise with selection bias issues, and Dr. Neale R. Chumbler for his helpful advice in the area of access and utilization. I thank Dr. Brent J. Shelton and Dr. Mark Litaker for the statistical consults they provided to me. I would also like to acknowledge Dr. Sonia Makhija for editing my dissertation. I am indebted to my dear friend, Britta Neugaard, for her support, caring, and compassion throughout this journey. I would not have been able to complete this endeavor without my familyÂ’s support. I would like to thank my parents, Yingsong and Guangrong, who have been extremely patient, generous, and supportive in the pursuit of my dreams. Special thanks go to my husband Qing, who has been giving me unconditional love and support since the day we met. I also thank my brother, Yongdong; and my sister-in-law, Jiahong, for taking care of our parents and families when I am pursuing my dreams in the United States.

PAGE 5

v TABLE OF CONTENTS page ACKNOWLEDGMENTS.................................................................................................iv LIST OF TABLES.............................................................................................................ix LIST OF FIGURES...........................................................................................................xi ABSTRACT......................................................................................................................x ii CHAPTER 1 INTRODUCTION........................................................................................................1 Rationale for the Study.................................................................................................2 Compelling Need to Link Prosthodontic Services with Oral Health Outcomes...2 Compelling Need to Conduc t a Longitudinal Study.............................................4 Compelling Need to Use Multiple Meas ures of Self-Rated Oral Health..............5 Overall Purpose of the Study........................................................................................7 Objectives of the Study.................................................................................................8 Objective 1.............................................................................................................8 Objective 2.............................................................................................................8 Objective 3.............................................................................................................9 Objective 4...........................................................................................................10 2 LITERATURE REVIEW...........................................................................................12 Impact of Tooth Loss on Oral H ealth-Related Quality of Life..................................12 Factors Associated with Pros thodontic Services Utilization......................................14 Impact of Prosthodontic Services Use on Oral Health-Related Quality of Life.........16 Impact on Chewing Performance........................................................................16 Impact on Appearance.........................................................................................19 Impact on Speaking.............................................................................................20 Impact on Dietary Intake and Nutrition...............................................................20 Negative Impact of Prosthodontic Services................................................................21 3 THEORETICAL AND CONCEPTUAL MODELS..................................................23 Behavioral Model of Health Services Use..................................................................23 Multidimensional Conceptual Model of Oral Health.................................................30

PAGE 6

vi 4 RESEARCH DESIGN AND METHODS..................................................................34 Sample and Sampling Procedures..............................................................................34 Data Collection Methods............................................................................................36 Description of Variables.............................................................................................39 Dependent Variables...........................................................................................39 Three original measures under the do main of self-rated oral health............39 Dynamic changes in self-rated oral health...................................................40 Independent Variables.........................................................................................41 Original measures of prosthodontic services use.........................................41 Recoded prosthodontic services use.............................................................43 Control Variables.................................................................................................45 Predisposing factors.....................................................................................45 Enabling factors............................................................................................46 Need factors..................................................................................................46 Statistical Methods......................................................................................................51 Analytical Methods for Objective 1....................................................................51 Analytical Methods for Objective 2....................................................................52 Cross-sectional bivariate analys es for research question 2a........................52 Cross-sectional multivariate logistic regression for research question 2b...52 Longitudinal logistic regression with correction of selection bias for research question 2c................................................................................53 Longitudinal logistic regression with correction of selection bias for research question 2d................................................................................65 Analytical Methods for Obj ective 3 and Objective 4..........................................65 Sensitivity Tests...................................................................................................65 5 RESULTS...................................................................................................................67 Results for Objective 1...............................................................................................67 Incidence of Prosthodontic Se rvices Use during 24 Months...............................67 Prevalence of Self-Rated Oral Health Outcomes................................................67 Patterns of Changes in Se lf-Rated Oral Health...................................................68 Results for Objective 2...............................................................................................72 Cross-Sectional Bivariate Associations at the Baseline Interview (Research Question 2a).....................................................................................................72 Cross-Sectional Multivariate Associations at the Baseline Interview (Research Question 2b)....................................................................................73 Longitudinal Logistic Regression of Satisfaction with Chewing Ability with Correction of Selection Bias (Research Question 2c)......................................78 Results of the firststage models for re search question 2c...........................79 Results of the second-stage model for research question 2c........................84 Longitudinal Logistic Regression of Changes in Satisfaction with Chewing Ability with Correction of Selec tion Bias (Research Question 2d).................85 Results of the firststage models for re search question 2d...........................86 Results of the second-stage models for research question 2d......................87

PAGE 7

vii Results for Objective 3...............................................................................................89 Cross-Sectional Bivariate Associations at the Baseline Interview (Research Question 3a).....................................................................................................89 Cross-Sectional Multivariate Associations at the Baseline Interview (Research Question 3b)....................................................................................90 Longitudinal Logistic Regression of Satisfaction with Dental Appearance with Correction of Selection Bias (Research Question 3c).............................97 Results of the firststage models for re search question 3c...........................97 Results of the second-stage model for research question 3c........................97 Longitudinal Logistic Regression of Changes in Satisfaction with Dental Appearance with Correction of Sel ection Bias (Research Question 3d).........98 Results of the firststage models for re search question 3d...........................98 Results of the second-stage models for research question 3d......................99 Results for Objective 4.............................................................................................102 Cross-Sectional Bivariate Associations at the Baseline Interview (Research Question 4a)...................................................................................................102 Cross-sectional Multivariate Associations at the Baseline Interview (Research Question 4b)...................................................................................................105 Longitudinal Logistic Regression of Self-Rated Overall Oral Health with Correction of Selection Bias (Research Question 4c)....................................108 Results of the firststage models for re search question 4c.........................108 Results of the second-stage model for research question 4c......................108 Longitudinal Logistic Regression of Changes in Self-Rated Overall Oral Health with Correction of Selec tion Bias (Research Question 4d)................109 Results of the firststage models for re search question 4d.........................110 Results of the second-stage models for research question 4d....................110 6 DISSCUSSION.........................................................................................................113 Discussion of the Results for Objective 1.................................................................113 Discussion of the Results for Objective 2.................................................................115 Discussion of the Results for Objective 3.................................................................118 Discussion of the Results for Objective 4.................................................................122 7 CONCLUSIONS AND LIMITATIONS..................................................................124 Conclusions...............................................................................................................124 Limitations................................................................................................................125 APPENDIX A CORRESPONDING QUESTIONNAIRE ITEMS...................................................128 B THE SAS PROGRAM FOR INSTRU MENTAL VARIABLE APPROACH.........132 C SENSITIVITY TESTS.............................................................................................143

PAGE 8

viii LIST OF REFERENCES.................................................................................................152 BIOGRAPHICAL SKETCH...........................................................................................166

PAGE 9

ix LIST OF TABLES Table page 4-1. Number of Participants be tween Baseline and 24 Months.......................................39 4-2. All Variables with Response Categories..................................................................48 4-3. Observed Versus Predicated Finished Fixed Prosthodontic Treatment Obtained from the First-Stage Model.......................................................................................62 4-4. Observed Versus Predicated Fini shed Removable Prosthodontic Treatment Obtained from the First-Stage Model.......................................................................62 4-5. Observed Versus Predicated Unfi nished Fixed Prosthodontic Treatment Obtained from the First-Stage Model.......................................................................62 4-6. Observed Versus Predicated Unfini shed Removable Prosthodontic Treatment Obtained from the First-Stage Model.......................................................................63 5-1. Incidence of Prosthodontic Services Use during 24 Months...................................69 5-2. Prevalence of Self-Rated Oral Health Outcomes during 24 Months.......................70 5-3. Dynamic Changes in Self-Rated Oral Health during 24 Months.............................71 5-4. Satisfaction with Chewing Ability at Baseline for the Sample Overall, by Baseline Prosthodontic Services Use and by Predisposing, Enabling, and Need Factors (PEN Factors)...............................................................................................74 5-5. Cross-Sectional Logistic Regression of Satisfaction with Chewing Ability at Baseline.....................................................................................................................77 5-6. Instrumental Variables for Research Question 2c....................................................81 5-7. The First-Stage Model to Calculate Predicated Probabilities of Finished Fixed Prosthodontic Treatment...........................................................................................81 5-8. The First-Stage Model to Calculate Predicated Probabilities of Finished Removable Prosthodontic Treatment........................................................................83 5-9. Parameter Estimates from the Second-Stage Model for Research Question 2c.......85

PAGE 10

x 5-10. Instrumental Variables for Research Question 2d....................................................86 5-11. Parameter Estimates from the Second-Stage Model for Research Question 2d (Improvement of Satisfaction with Chewing Ability)..............................................88 5-12. Parameter Estimates from the Second-Stage Model for Research Question 2d (Deterioration of Satisfacti on with Chewing Ability)..............................................89 5-13. Satisfaction with Dental Appearance at Baseline for the Sample Overall, by Baseline Prosthodontic Services Use and by Predisposing, Enabling, and Need Factors (PEN Factors)...............................................................................................92 5-14. Cross-Sectional Logistic Regression of Satisfaction with Dental Appearance at Baseline....................................................................................................................95 5-15. Instrumental Variables for Research Question 3c....................................................97 5-16. Parameter Estimates from the Second-Stage Model for Research Question 3c......98 5-17. Instrumental Variables for Research Question 3d....................................................99 5-18. Parameter Estimates from the Second-Stage Model for Research Question 3d (Improvement of Satisfaction with Dental Appearance)........................................100 5-19. Parameter Estimates from the Second-Stage Model for Research Question 3d (Deterioration of Satisfaction with Dental Appearance)........................................101 5-20. Self-Rated Overall Oral Health at Baseline for the Sample Overall, by Baseline Prosthodontic Services Use and by Predisposing, Enabling, and Need factors (PEN Factors).........................................................................................................103 5-21. Cross-Sectional Logistic Regression of Self-Rated Overall Oral Health at Baseline..................................................................................................................106 5-22. Instrumental Variables for Research Question 4c..................................................108 5-23. Parameter Estimates from the Second-Stage Model for Research Question 4c....109 5-24. Instrumental Variables for Research Question 4d..................................................110 5-25. Parameter Estimates from the Second-Stage Model for Research Question 4d (Improvement of Self-Rated Overall Oral Health).................................................111 5-26. Parameter Estimates from the Second-Stage Model for Research Question 4d (Deterioration of Self-Rated Overall Oral Health).................................................112

PAGE 11

xi LIST OF FIGURES Figure page 3-1. Theoretical Framework............................................................................................30 3-2. Multidimensional Conceptual Model of Oral Health...............................................31 4-1. Coding of Dynamic Dependent Variables...............................................................42 4-2. Coding of Prosthodontic Services............................................................................44 4-3. Possible Ways of Biased Estimated Tr eatment Effects Due to Selection Bias........56 4-4. Illustration of the Instru mental Variable Approach.................................................58

PAGE 12

xii Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy EFFECT OF PROSTHODONTIC SERVICES ON SELF-RATED ORAL HEALTH OUTCOMES By Xiaoxian Meng August 2005 Chair: R. Paul Duncan Major Department: Health Services Research, Management and Policy This dissertation investigated a fundamental question that had not been addressed in the literature to date: “What is the dynamic impact of prosthodontic services on multiple measures of oral health outcomes?” To accomplish this overall purpose, we used data obtained from the Florida Dent al Care Study (FDCS), which is a populationbased longitudinal cohort study of self-reported oral health and dental care. The sample included 873 persons who had at least 1 tooth and were 45 years old or older. The Andersen Behavioral Model of Health Services Use and the Multidimensional Conceptual Model of Oral Health framed our analyses. Objectives of our study were (1) to quantify the patterns of prosthodontic services use and changes in self-rated oral health outcomes; (2) to quantify the impact of prosthodontic services use on satisfaction with chewing ability; (3) to quantify the impact of prosthodontic services use on satisfaction with dental appearance; and (4) to quantify the impact of prosthodontic services use on self-rated overall oral health. Descriptive

PAGE 13

xiii statistics, cross-sectional ordinal logistic regression, and longitudinal binomial logistic regression with correction for selection bias using the instrumental variable approach were used to answer the research questions. Results showed that fixed prosthodontic services were more common than removable prosthodontic services. Most people were satisfied with their chewing ability and dental appearance, and rated their overall oral health positively. The finished removable prosthodontic treatment was associated with an improved satisfaction with chewing ability and satisfaction with dental appearance. Our results did not show any beneficial effect of fi xed prosthodontic treatment. Our findings document the therapeutic effectiveness of removable prosthodontic treatment but not fixed prosthodontic treatment. These results suggest switching to the less expensive removable prosthodontic treatment from the more expensive fixed prosthodontic treatment will result in a noticeable saving in health care expenditure without compromising health outcomes.

PAGE 14

1 CHAPTER 1 INTRODUCTION The concept has been widely accepted that oral health plays an essential and integral role in maintaining general health and well-being throughout life. The significance of oral health is well depict ed by a popular quote: “You are not healthy without good oral health,” made by Dr. C. Everett Koop, former Surgeon General of the United States, 1981 to 1989. Previous studies have clearly indicated that normal physical, social, and psychological functi ons such as chewing, speaking, laughing, appearance, self-image, and social contact can be compromised by dental disorders (Brodeur et al., 1993; Chierici and Laws on, 1973; Nowjack-Raymer and Sheiham, 2003; Ofstehage, 1987). While overall oral health status in the United States has improved during the past decades, oral diseases and conditions still exert a heavy burden on people’s quality of life. The ultimate goal of any health care service is to maintain or improve people’s health status. The outcomes of health services use have been of increasing concern to the general public, policy makers, and health care planners with the efforts to improve access and increase utilization. Efforts to improve access and utilization are valuable only if such efforts result in positive health outcomes. During the past decades, the whole health care system in the United States has come under great economic and political pressure to reduce cost, increase accessibility, and improve quality and effectiveness of care. As an important component of health care, dental care is responsible for maintaining or improving oral health and for preventing or ameliorating declines in oral health.

PAGE 15

2 Similarly, dental care is expected to be delivered in an effective, safe, but less costly manner. Although a focus on oral health outcomes is one of the recommendations made by the Institute of Medicine (IOM) (Grem bowski, 1997; Institute of Medicine, 1995), there is a lack of strong and consistent supporting evidence to demonstrate the costeffectiveness of various types of dental treat ment. Further systematic investigations of the mechanisms underlying oral health outcomes and the effectiveness of dental treatment are essential in meeting this recommendation. Rationale for the Study Compelling Need to Link Prosthodontic S ervices with Oral Health Outcomes Tooth loss is the eventual consequence of prevalent dental diseases, in particular dental caries and periodontal diseases. To prevent negative consequences or ameliorate burdens associated with tooth loss, a treatment consensus among clinicians is to replace missing teeth with various forms of prosthodontic appliances. Because tooth loss has historically been quite prevalent, using prosthodontics (tooth-replacing appliances) of some type is a way of life for many people, especially among older adults. Prosthodontic treatment is of special interest in oral health outcomes research because it has a substantial impact on successful aging, with a growing need for this type of dental services among the elderly who are generally without a complete natural dentition (Douglass et al., 1988). The ideal outcome of prosthodontic treatment is to functionally and socio-psychologically rehabilitate compromised quality of life. However, there is a lack of agreement in the literature with regard to the value of prosthodontic treatment in affecting the quality of life for people with tooth loss. Some studies have shown that prosthodontic treatment improves chewing ability and esthetics (Gilbert et al., 2004; Van Waas et al., 1994), and contributes to oral comfort (Witter et

PAGE 16

3 al., 1989). Other studies have indicated that prosthodontic treatment is not always as effective as anticipated (Käyser et al., 1990; Ow et al., 1997; Witter et al., 1999). Sometimes, it even induces negative consequences (Nevalainen et al., 2004; Rossi et al., 1995). Throughout much of human history, tooth loss has been considered a natural consequence of the aging process. Due to the introduction of preventive dentistry and improved dental awareness in the general public, the overall prevalence of tooth loss and complete edentulism has been steadily declining over the past decades in the United States, with more people retaining at least some of their natural teeth for life (Marcus et al., 1996). Tooth loss is no longer viewed as an inevitable phenomenon associated with aging. Consequently, the need for prosthodontic services could be significantly affected by these changes in the prevalence of tooth loss. Despite such progress, tooth loss and the subsequent use of prostheses remain highly prevalent among adults. A nationwide survey, the Third National Health and Nutrition Examination Survey (NHANES III), has documented the prevalence of tooth loss in the United States (Marcus et al., 1996). Data showed that the percentage of persons without intact dentition (which means at least some teeth are missing) increases by age, ranging from 33.6% among people aged 18 to 24, to 98% among people aged 75 or older. The rate of complete edentulism (which means no teeth are retained) also increases by age, ranging from 0% among people aged 18 to 24, to 43.9% among people aged 75 or older. Information on denture use in the United States has been documented by the same survey (Redford et al., 1996). Results showed that older adults are more likely than younger adults to wear dentures. While only 3.1% of people aged 18 to 34

PAGE 17

4 wear a removable prosthesis of some type, wearing a denture is a way of life for more than 50% of Americans age 55 or older (Redford et al., 1996). After the steady decline in prevalence of tooth loss, one would expect a decline in the need for prosthodontic services. However, a series of studies (Douglass et al., 1988; Douglass et al., 2002; Douglass and Watson, 2002) showed the opposite trend. In the United States, the need for prosthodontic services is projected to increase because of the increases in the proportion of old people as well as the overall population (Douglass et al., 1988). Although the percentage of edentulous individuals is decreasing, the effective demand for complete dentures continues to increase (Douglass et al., 2002). Compared to an estimated dental workforce that will actively provide prosthodontic dental care consonant with the projected need for prosthodontic services, unmet needs are increasing for fixed partial dentures (FPD), removable partial dentures (RPD) and complete dentures (CD). The growing need is projected to greatly exceed the available supply in the years 2005, 2010, and 2020. In terms of provider time, the total unmet need is estimated to be 488 million hours in 2005, 517 million hours in 2010, and 560 million hours in 2020 (Douglass and Watson, 2002). Douglass and Watson (2002) argued that the increasing need for prosthodontic rehabilitation might be due to population growth and extended life expectancy. Because more people retain some of their natural teeth for most of their lives, most of the population will be partially edentulous, thus the need for fixed or removable partial prostheses will increase (Joshi et al., 1996). Compelling Need to Conduct a Longitudinal Study Relatively little is known about the ways that oral health changes after an individual with tooth loss receives a prosthesis. Such knowledge is essential to advance our understanding of the effectiveness of the prosthodontic treatment. Yet, cross-

PAGE 18

5 sectional assessment remains the norm in prosthodontic services research. Most of the previous studies regarding the effect of prosthodontic services use are limited by cross-sectional designs comparing oral health outcomes between prosthodontic service users and non-users. The cross-sectional comparisons could not separate the treatment effect of prosthodontic services from other changes over time. Dental services researchers have recently begun to conceptualize oral health as the product of a complex and dynamic process. However, few studies have used longitudinal designs to quantify the dynamic impact of prosthodontic services on changes in oral health (Gilbert et al., 2004). Thus, there is a compelling need for up-to-date information about the impact of prosthodontic services on oral health outcomes through longitudinal analysis. Compelling Need to Use Multiple Mea sures of Self-Rated Oral Health Just as the definition of health is not merely the lack of diseases, the concept of oral health also encompasses physical, psychological, behavioral, and social components. Measures of oral health or evaluations of dental treatment should include degrees of dysfunction and patientsÂ’ perceptions. However, patientsÂ’ concerns about their oral and dental status have not received sufficient a ttention. Because clinical examinations are limited to providing information on oral diseases or tissue damage at the organic level, the quality and effectiveness of dental care cannot be comprehensively evaluated without valid socio-psychological measures (Bader and Shugars, 1995). Moreover, professional evaluation gathered by direct clinical examin ations does not necessarily reflect a personÂ’s own assessment. In general, patients are more optimistic regarding their oral health status than dentists. Reisine and Bailit (1980) noted that of those people with the lowest levels of oral health status (15 missing t eeth or deep periodontal pocket), 40% evaluated

PAGE 19

6 their oral health as “good” or “excellent.” Thus, they suggested that professional criteria might not reflect an individual’s assessment of his or her own oral health. Barenthin (1977) also found that there were no significant differences in satisfaction with general dental condition in subjects with good versus poor clinically evaluated dental health. Self-assessments of oral health have gained interest among dental clinicians and dental services researchers. It has been recognized that self-reported multiple measure of oral health are reliable, sensitive, and informative tools in evaluating the effectiveness of dental care. Self-assessments help us understand what aspects of oral health matter most to patients; and what oral diseases, conditions, and dental impacts affect a person’s assessment of his or her own oral health. Several instruments measuring multiple dimensions of oral health have been proposed in the past decades, and some of them have wide implications. For example, the self-reported measures used by Fiske and colleagues (1990) include a socio-dental measure of oral handicap. The “Geriatric Oral Health Assessment Index” (GOHAI) (Atchison and Dolan, 1990) includes a broad array of self-reported oral health measures specifically designed for the older population. This instrument has been used longitudinally (Dolan et al., 1998). The “Dental Impact Profile” proposed by Strauss and Hunt (1993) uses a battery approach to assess dental impact and subjective perception about the value of oral health. The “Dental Impact on Daily Living” (Leao and Sheiham, 1995) includes 36 items that cover the measures of “appearance,” “pain,” “comfort,” “general performance,” and “eating restric tion.” The “Oral Health Impact Profile” (OHIP) (Slade and Spencer, 1994) is a 49-item self-report questionnaire designed to evaluate patients’ oral pain, discomfort, f unctional limitation, disability, and handicap. It

PAGE 20

7 generates a single score for each individual through a weighted index approach. Gilbert et al. (1998a) conceptualized and validated a multidimensional conceptual model of oral health (details in Chapter 3). Their model comprises five domains of oral health and oral health-related quality of life: (1) oral disease and tissue damage, (2) oral pain and discomfort, (3) oral functional limitation, (4) oral disadvantage, and (5) self-rated oral health. Although the treatment effect of various types of prosthodontic services has been widely investigated in recent decades, many studies were conducted in a laboratory environment using sophisticated technologies. Outcomes obtained through objective methods do not necessarily reflect what patie nts subjectively perceive, and thus do not provide a complete picture of the treatme nt effect of prosthodontic services. Furthermore, most of the studies only focus on a particular aspect of oral health. Relatively few studies have simultaneously examined the impact of prosthodontic treatment on multiple measures of oral health outcomes. Therefore, a study using subjectively rated multiple oral health outcomes offers valuable evidence from the patientÂ’s perspective. Overall Purpose of the Study Although more than $ 66 billion are spent each year on dental care in the U.S. (Levit et al., 2003), little research has been conducted on the effectiveness of specific dental services using longitudinal data. Little is known about what constructs needed and effective prosthodontic services, or about what constructs overtreatment and inappropriate use of prosthodontic services. This dissertation studies a fundamental question that has not been addressed in the literature to date: What is the dynamic impact of prosthodontic services on multiple measures of oral health outcomes? To accomplish

PAGE 21

8 this overall purpose, this dissertation uses data obtained from the Florida Dental Care Study (FDCS), a population-based longitudinal cohort study of self-reported oral health and dental care. Multiple measures of self-reported oral health are outcomes that determine the effectiveness of various types of prosthodontic services. This dissertation has the potential to provide new evidence to understand the value of prosthodontic services, as well as to broaden the literature regarding how prosthodontic services, along with other factors, affect oral health outcomes over time. This dissertation could also provide valuable evidence to guide the allocation of the limited dental care resources. Objectives of the Study This dissertation is aimed to achieve four objectives. Each objective is associated with several specific research questions. Objective 1 To quantify the incidence of prosthodontic se rvices use, the prevalence of self-rated oral health outcomes, and the patterns of changes in self-rated oral health outcomes during 24 months follow-up in a diverse sample of dentate middle-aged and older adults. Objective 2 To quantify the impact of prosthodontic services use, along with other predisposing, enabling, and need factors, on satisfaction with chewing ability. Research question 2a: What are the bivariate associations between baseline satisfaction with chewing ability and other relevant factors, specifically, baseline use of prostheses and predisposing, enabling, and need factors? Hypothesis 2a: There are significant bivariate associations between baseline satisfaction with chewing ability and other factors (baseline use of prostheses and predisposing, enabling, and need factors). Research question 2b: What are the multivariate associations between baseline satisfaction with chewing ability and baseline use of prostheses controlling for predisposing, enabling, and need factors?

PAGE 22

9 Hypothesis 2b: Controlling for predisposing, enabling, and need factors, the use of certain types of prostheses is asso ciated with a greater satisfaction with chewing ability at baseline. Research question 2c: Controlling for predisposing, enabling, and need factors, what is the impact of prosthodontic services use on satisfaction with chewing ability at each follow-up interview during a 24-month period? Hypothesis 2c : With predisposing, enabling, and need factors taken into account, receiving certain types of prosthodontic treatment is associated with a greater satisfaction with chewing ability during a 24-month period. Research question 2d: With predisposing, enabling, and need factors taken into account, what are the impacts of prosthodontic services use on the dynamic changes in satisfaction with chewing ability (improvement /deterioration) in each 6-month interval during a 24-month period? Hypothesis 2d: With predisposing, enabling, and need factors taken into account, prosthodontic services use is associated with a higher likelihood of positive change (improvement) and a lower likelihood of negative change (deterioration) in satisfaction with chewing ability. Objective 3 To quantify the impact of prosthodontic services use, along with other predisposing, enabling, and need factors, on satisfaction with dental appearance. Research question 3a: What are the bivariate associations between baseline satisfaction with dental appearance and other relevant factors, specifically, baseline use of prostheses and predisposing, enabling, and need factors? Hypothesis 3a: There are significant bivariate associations between baseline satisfaction with dental appearance and other factors (baseline use of prostheses and predisposing, enabling, and need factors). Research question 3b: What are the multivariate associations between baseline satisfaction with dental appearance and baseline use of prostheses controlling for predisposing, enabling, and need factors? Hypothesis 3b: Controlling for predisposing, enabling, and need factors, the use of certain types of prostheses is asso ciated with a greater satisfaction with dental appearance at baseline.

PAGE 23

10 Research question 3c: Controlling for predisposing, enabling, and need factors, what is the impact of prosthodontic services use on satisfaction with dental appearance at each follow-up interview during a 24-month period? Hypothesis 3c: With predisposing, enabling, and need factors taken into account, receiving certain types of prosthodontic treatment is associated with a greater satisfaction with dental appearance during a 24-month period. Research question 3d: With predisposing, enabling, and need factors taken into account, what are the impacts of prosthodontic services use on the dynamic changes in satisfaction with dental appearance (improvement /deterioration) in each 6-month interval during a 24-month period? Hypothesis 3d: With predisposing, enabling, and need factors taken into account, prosthodontic services use is associated with a higher likelihood of positive change (improvement) and a lower likelihood of negative change (deterioration) in satisfaction with dental appearance. Objective 4 To quantify the impact of prosthodontic services use, along with other predisposing, enabling, and need factors on self-rated overall oral health. Research question 4a: What are the bivariate associations between baseline ratings of overall oral health and other relevant factors, specifically, baseline use of prostheses and predisposing, enabling, and need factors? Hypothesis 4a: There are significant bivariate associations between baseline ratings of overall oral health and other factors (baseline use of prostheses and predisposing, enabling, and need factors). Research question 4b: What are the multivariate associations between baseline ratings of overall oral health and baseline use of prostheses controlling for predisposing, enabling, and need factors? Hypothesis 4b : Controlling for predisposing, enabling, and need factors, the use of certain types of prostheses is associated with better ratings of overall oral health at baseline. Research question 4c: Controlling for predisposing, enabling, and need factors, what is the impact of prosthodontic services use on self-rated overall oral health at each follow-up interview during a 24-month period? Hypothesis 4c: With predisposing, enabling, and need factors taken into account, receiving certain types of prosthodontic treatment is associated with better ratings of overall oral health during a 24-month period.

PAGE 24

11 Research question 4d: With predisposing, enabling, and need factors taken into account, what are the impacts of prosthodontic services use on the dynamic changes in self-rated overall oral health (improvement /deterioration) in each 6month interval during a 24-month period? Hypothesis 4d: With predisposing, enabling, and need factors taken into account, prosthodontic services use is associated with a higher likelihood of positive change (improvement) and a lower likelihood of negative change (deterioration) of self-rated overall oral health.

PAGE 25

12 CHAPTER 2 LITERATURE REVIEW Impact of Tooth Loss on Oral Health-Related Quality of Life To understand the value of prosthodontic services, it is necessary to review the impact of tooth loss on a person’s quality of life. The primary function of teeth is to enable us to chew food. At the same time, teeth play important roles in facilitating speaking and laughing; teeth enhance facial appearance, help to establish positive selfimage, and assist normal social interaction. Strauss and Hunt (1993) observed that appearance, eating and chewing, being free of pain and discomfort, and speech are the most important effects of having adequate dentition. Thus, tooth loss can substantially influence many aspects of a person’s quality of life. Consensus among clinicians is the larger the number of teeth retained, the better the chewing ability maintained, which has been supported by research (Oesterberg and Steen, 1982; Rosenoer and Sheiham, 1995). However, some researchers (Käyser et al., 1990; Witter et al., 1999) have posited that, in contemporary society, complete integrity of the dental arch is no longer needed to fully chew modern diets. Ten occluding pairs of teeth (one occluding pair consists of a tooth in the upper arch and the corresponding tooth in the lower arch that it bites against or occlude s), for a total of 20 properly distributed teeth, may be adequate for optimal chewing ability (Käyser et al., 1990; Witter et al., 1999). Previous studies (Nowjack-Raymer and Sh eiham, 2003; Sheiham et al., 1999) have shown that fewer remaining teeth, edentulism, and compromised chewing ability are

PAGE 26

13 associated with altered dietary selection and consequent improper nutrient intake. Tooth loss also may be related to gastrointestinal disorders (Brodeur et al., 1993). The features most commonly associated with facial attraction are the eyes and mouth (Baldwin, 1980). The mouth and teeth pl ay an important role when an individual speaks or approaches another person. Ugly or missing teeth, for example, are noticed by others, and those who have missing or otherw ise problematic teeth in visible locations express concern about their appearance. Previous studies have shown that overall selfimage was significantly related to missing teet h, tooth displacement, tooth discoloration, and bad tooth shape (Newmann et al., 1989) . Although Newmann and coworkers (1989) suggested that, for older people, dental appearance may no longer be an important priority in relation to other health needs and concerns, Goldstein (1986) has found that appearance and physical attractiveness does not appear to decline with age and the maintenance of oral esthetics was a major criterion of successful aging. Tooth loss is related to compromised dental appearance, and consequently results in negative judgments that may cause lower self-esteem and social withdrawal (Ofstehage, 1987). Missing teeth, especially missing anterior teeth, is unacceptable to the individual, public, and dentists (Oosterhaven et al., 1989). Missing teeth usually result in less positive feelings and more negative feelings, and significantly affect daily life activities (Oosterhaven et al., 1989). In a study involving 60 community-dwelling older people who were asked to rank the attractiveness of faces, the faces with missing teeth tended to be ranked the least attractive (York and Holtzman, 1999). Tooth loss can also cause speech defects. There is agreement in the literature (Chierici and Lawson, 1973; Palmer, 1974) that the anterior teeth serve as the obstruction

PAGE 27

14 against which the tongue apex might deliver a stream of air to create the friction sound, as in /S/ sound production. Thus, missing anterior teeth will cause omission of sounds. However, opinions differ on the impact of missing posterior teeth on speech. Palmer (1974) indicated that posterior teeth play an important role in normal sound production. In a normally speaking individual with full dentition, it is thought that the molar teeth act either as orienting surfaces or as fulcrums against which the tongue operates (Palmer 1974). On the contrary, Chierici and Lawson (1973) claimed that accommodations to missing posterior teeth usually cause minimal or no speech distortion. Just as various forms of loss of bodily organs exert substantial effects on a personÂ’s psychological well-being (Blomberg and Lindqui st, 1983), tooth loss also has important psychological consequences. Bergendal (1989) reported that subjects recognized tooth loss as a serious life event and weighted complete tooth loss more important than marriage, retirement, or changing jobs. A qualitative study reported that to those subjects who were interviewed, tooth loss reflected impending loss of youth and virility, decreased attractiveness, and body deterioration (Allen and McMillan, 2003). Fiske et al. (1998) explored the emotional effects of comp lete edentulism (total tooth loss) and found the reactions to total tooth loss included bereavement, lowered self-confidence, altered self-image, dislike of appearance, altered behaviors in socializing, and even a sense of mortality. Factors Associated with Prosthodontic Services Utilization Earlier work using the FDCS data (Gilbert et al., 2002a) found that prosthodontic treatment was associated with certain soci o-demographic factors: African-Americans, people with lower levels of education, and people with less financial resources were less likely to receive both fixed and removable prosthodontic services. However, the

PAGE 28

15 associations between socio-demographic factors and prosthodontic use must be interpreted cautiously because the actual use of prosthodontic services reflects both a prior history of tooth loss and the pattern of prosthodontic utilization. Findings of previous studies seem to conflict because they reflect different aspects of the same problem. On one hand, the elderly, people with less education, and people from low social economic background are more likely to experience tooth loss than their respective counterparts (Marcus et al., 1994). Fewer remaining functioning natural teeth make people more likely to use prosthodontic services. For example, a study conducted in Swedish adults showed that less education was associated with a higher likelihood of wearing a prosthesis (Hjern et al., 2001). On the other hand, after controlling for the number of lost teeth and other sociodemographic and service-related factors, younger people and people from a high social class are more likely to obtain prosthodontic rehabilitation (McGrath and Bedi, 2001). Although complete edentulism is significantly more prevalent in people with a low income (Marcus et al., 1994), Maupomé and MacEntee (1998) reported that the use of complete dentures was more common among people in the high-income bracket. Some previous studies have found that financial factors do not play a decisive role in using or not using dental services, but do affect the amount of services finally received (Locker, 1997; Österberg et al., 1995). Denture wearers have significantly lower rate s of using professional dental services in contrast to older adults with natural teeth. They seem to have little interest in regularly attending dental appointments (Mojon and M acEntee, 1992; Strayer et al., 1988; Weyant et al., 2004). So, who receives previous treatme nt remains unclear, as does the impact of that treatment on future dental care.

PAGE 29

16 Impact of Prosthodontic Services Use on Oral Health-Related Quality of Life In terms of the impact on overall oral health-related quality of life, McGrath and Bedi (2001) applied an OHQoL-UK© instrument to measure overall scores of oral health-related quality of life (OHQoL). Afte r controlling for the number of lost teeth, persons who wore removable partial or complete dentures were more than twice as likely to enjoy an enhanced overall quality of life compared to persons who had no recourse to removable dentures. However, Leake and others (1994) found that, compared to patients with the same number of natural functional units but no denture (a functional unit was defined as a pair of teeth in the upper and lower arches which bite against each other), denture wearers did not gain improvement in any of the six functional and sociopsychological measures. In another study (John et al., 2004), the German Version of the Oral Health Impact Profile (OHIP-G) was used to measure OHQoL. Wearing removable partial dentures or complete dentures was found to be a strong predictor for compromised OHQoL. Weyant and colleagues (2004) al so reported that denture wearing was significantly associated with higher rates of oral functional limitations, higher levels of depression, greater use of prescription medica tion, and poorer self-rated general health. However, interpretation of the results from both studies (John et al., 2004; Weyant et al., 2004) should be made with caution because these studies did not control for the number of lost teeth. Thus, their findings might not reflect the pure impact of prosthodontic treatment on quality of life for those people w ho experienced tooth loss, with or without having recourse to prosthodontic services. Impact on Chewing Performance The most important physiological function of teeth is to chew. Thus, much attention has been paid to the effect of prosthodontic treatment on the mechanical

PAGE 30

17 performance. However, there is a lack of consensus in the literature with regard to the chewing benefit of replacing missing teeth by a certain type of prosthesis. Some researchers believe that as long as a person retains ten occluding pairs of teeth, for a total of 20 well-distributed teeth (“shortened de ntal arch”), no obvious advantage will be gained by replacing missing teeth with removable dentures (Käyser et al., 1990; Witter et al., 1999). Ow and others (1997) found that there were no significant differences in chewing ability between people who were not wearing dentures and those who were wearing any type of dentures. Carlsson (1984) reported that severely compromised oral function in edentulous persons could not be substantially improved with complete dentures of good quality. On the other hand, some researchers observed positive effects of prosthodontic treatment. Van der Bilt et al. (1994) recorded the objective and subjective masticatory function of a group with an average of 5.8 missing posterior teeth, before and after the receipt of removable or fixed prosthodontic treatment. They found that both objective and subjective masticatory function improved as a result of the treatment. After the prosthodontic treatment, the average score on positive feelings about chewing ability significantly increased, which was comparable to that of a control group with a complete dentition (Van der Bilt et al., 1994). Agerberg and Carlsson (1981) reported that complete denture wearers reported better chewing ability than non-denture wearers who had few remaining natural teeth. Van Waas and colleagues (1994) reported that although, in general, denture wearers were less satisfied with their dental state than nondenture wearers, seventy-seven percent of de nture wearers indicated that their chewing ability was better than before they had a denture. A randomized clinical trial (Kapur et

PAGE 31

18 al., 1997) showed that wearing a removable partial denture enhanced masticatory efficiency with obvious improvement being observed within 16 weeks after insertion. Another study (Miyaura et al., 2000) found similar results; significant increases of biting ability were observed 2 months after insertion for all fixed and removable partial dentures and complete dentures. A longitudinal study (Gilbert et al., 2004) found that removable prosthodontic treatment (removable partial dentures and complete dentures) was associated with a decreased probability of chewing difficulty onset for all people who experienced loss of teeth, while fixed prosthodontic treatment (cro wns, bridges, dental implants, etc.) was not. Their findings added to the support for the rehabilitative value of removable dentures in diverse groups of dentate adults. The results also suggested that the primary benefits of fixed prosthodontic restoration mi ght not lie in restoring mechanical function but in preventing the decline in social and psychological function such as aesthetics and self-image. This suggestion was consistent with the findings of a qualitative study conducted in Sweden to explore the process of socio-psychological change, starting with the deteriorated dentition, continuing with living and coping with a removable denture, and ending with being replaced with a fixed prosthesis (Trulsson et al., 2002). A onehour taped interview was carried out with each of the 18 patients in a conversational style. The interviews were then transcribed verbatim and analyzed based on grounded theory. The results illustrated that poor dental status was associated with altered selfimage, unsocial behavior, and deviating personality. Wearing a removable denture was connected with feelings of uncertainty and insecurity, decreased attraction, and restricted social interactions. The interviewees indicated they were afraid their removable dentures

PAGE 32

19 would be loose or even fall out of the mouth, thus they always tried to avoid eating or laughing with other people. When the removable dentures were replaced with the fixed prosthodontic appliances, the interviewees felt that they were able to recapture social security, to regain attractiveness, and to re-establish a positive self-image (Trulsson et al., 2002). Their results suggested the major reasons for patients to replace removable dentures with fixed prosthodontic appliances were socio-psychological motivations. A five-year longitudinal study revealed that fi xed prostheses also provided better oral comfort and a healthier oral environment than removable prostheses (Budtz-Jorgensen and Isidor, 1990). Impact on Appearance Teeth physically support the lower part of the face; thus, play an important role in maintaining the normal form of the face. Tooth loss could have a substantial impact on facial appearance. Esthetic reasons are a strong motivation for the replacement of teeth (Conny et al., 1985). Some studies have shown that aesthetic consideration is one of the main reasons for prosthodontic treatment, especially among individuals with missing anterior teeth (Schuurs et al., 1990; Tervone n, 1988). Zarb et al. (1978) noted that to enhance facial appearance was the primary motivation expressed by partially edentulous patients for seeking prosthodontic treatment, followed by masticatory reasons. However, the aesthetic effect of prosthodontic treatment has been poorly understood, especially from the patientsÂ’ perspective. A study c onducted by Van Vaas and others (1994) did not find a significant esthetic value of removable partial dentures, even among those who had dentures for anterior missing teeth. A pa st study showed un-replaced posterior missing teeth are somewhat tolerated due to a lack of aesthetic sensitivity (Spartley, 1988). Liedberg et al. (1991) observed a high preval ence of un-replaced missing teeth even in

PAGE 33

20 the anesthetically sensitive anterior region among Swedish elderly men. Razak et al. (1990) also found many Malaysians did not have their missing anterior and posterior teeth replaced by prostheses. Their results suggest this tolerance of the aesthetically sensitive dental space may be due to a cultural acceptance. Impact on Speaking A national survey conducted in Great Britain (Sheiham et al., 2001) revealed that, second to eating, speaking was the most common function affected by dental health status. Hypothetically, prosthodontic appliances may either improve or hinder speech, based on the quality of the prosthesis and the location of missing teeth. However, the effect of prosthodontic restoration on speaki ng has not been thoroughly investigated. A few existing studies (Frank et al., 2000; St eele et al., 1997) found that a substantial percent of patients who received prosthodontic treatment were dissatisfied with their speaking. Impact on Dietary Intake and Nutrition The impact on dietary intake and nutrition is also conflicting. Krall and colleagues (1998) indicated that wearing removable partial dentures helped to maintain a healthy diet and nutritional intake. On the contrar y, most of the studies found that placement of removable partial dentures or fixed partial dentures did not result in an observable dietary improvement (Garrett et al., 1997; Moynihan et al., 2000; Lamy et al., 1999). A study showed that patients wearing dentures had d ecreased intake of protein, fiber, calcium, vitamin A, ascorbic acid, riboflavin, and other nutrients (Greksa et al., 1995; Papas et al., 1998).

PAGE 34

21 Negative Impact of Prosthodontic Services The outcome of prosthodontic treatment is not always positive. Some studies have identified negative views toward removable dentures among some persons. For example, Frank and others (Frank et al., 1998) have revi ewed that 3% to 40% of removable partial dentures, with an average of approximately 25%, were considered unsuccessful because patients were either dissatisfied with them or unable to wear them. Their own results from a sample of lower removable denture w earers revealed that about 26% of denture wearers were slightly satisfied, or slightly, moderately, or completely dissatisfied with their prostheses. Chewing, speech, and appearance are the most frequently quoted reasons for dissatisfaction (Frank et al., 1998). In the Florida Dental Care Study, Dolan et al. (2001) observed various non-ideal fre quencies of wearing removable prostheses, depending on the types of the removable prostheses. Many of the FDCS subjects who ever had a certain type of the removable prosthesis reported that they never wore the denture. Prosthesis-related soreness and broken prostheses were the most commonly reported reasons associated with not wearing prostheses. For example, 43% of subjects with prosthesis-related soreness reported never wearing their mandibular partial denture, compared to 24% of those without soreness (Dolan et al., 2001). Rosenoer and Sheiham (1995) reported that approximately one third of their subjects had a denture but had not worn it for more than one year, and more than half of the subjects rarely wore their partial lower dentures. Blomberg and Lindquist (1983) found that 44% of patients wearing dentures perceived the dentures caused a deterioration in their way of life. Elias and Sheiham (1999) found that persons with open de ntal spaces but without removable partial dentures were more satisfied with their mouth than persons with removable partial dentures.

PAGE 35

22 Trulsson and others (2002) reported that w earing removable dentures could trigger a personÂ’s sense of uncertainty and insecurity, thus decreasing their social contact with others. Furthermore, it has been demonstrated that removable dentures can create a suitable environment for facilitating the growth of yeast and other oral organisms by providing a good attachment surface (Budtz-Jorg ensen et al., 1983; Nevalainen et al., 2004). Thus, removable dentures were considered a risk factor for several oral diseases such as stomatitis and other oral soft-tissu e diseases (Budtz-Jorgensen et al., 1983; Rossi et al., 1995; Theilade and Budtz-Jorgensen, 1988). Ramfjord (1974) reported that the replacement of missing molars was a common source of iatrogenic periodontal diseases. Nevalainen and others (2004) also reported that a presence of removable partial dentures and complete dentures was associated with an increased rate of root caries and diminished periodontal health.

PAGE 36

23 CHAPTER 3 THEORETICAL AND CONCEPTUAL MODELS Behavioral Model of Health Services Use In recent decades, many theories and models have been proposed to better conceptualize and operationalize the study of hea lth services research. Some theories and models portray a broader picture of access: why do people use or do not use health care services of various types, and what is the impact of the services on health outcomes? Models of this type include AndersenÂ’s behavioral model of health care use (Andersen, 1995), and a model of access to personal health care services presented by the Institute of Medicine (IOM) (IOM, 1993). Some other theories and models focus more on explaining personal health behaviors through socio-psychological approaches. These theories and models include, but are not limited to, the health belief model (Becker, 1974), MechanicÂ’s illness behavior model (Mechanic, 1982; McHugh and Vallis, 1985), SuchmanÂ’s social psychological theory (1965a, 1965b), ParsonÂ’s sick role theory (Parsons, 1951; Parsons, 1975), PescosolidoÂ’s social network theory (Pescosolido and Kronenfeld, 1995), and a public health model of the dental care process (Grembowski et al., 1989). The model of access to personal health care services presented by the IOM (IOM, 1993) has four constructs: barriers to access, utilization of health services, mediator factors, and health outcomes. Three of the four constructs (barriers, utilization, and outcome) are of particular interest to the IOM Access Monitoring Project committee. These three constructs overlap with the major constructs of AndersenÂ’s behavioral model

PAGE 37

24 of health services use. But Andersen’s behavioral model covers more predictive factors than the IOM model. The health belief model is also a well-applied and widely accepted model. It was developed in the 1950s in an effort to explain a widespread failure of public health programs in disease prevention and early detection (Becker, 1974). Over the years, many investigators have applied this model to diverse populations to address preventive behaviors, such as breast cancer screening (Fulton et al., 1991), HIV testing (de Paoli et al., 2004), oral cancer screening (Tan et al., 2001), and preventive dental care (Kegeles, 1963). This model includes five socio-psyc hological constructs: perceived severity, perceived susceptibility, perceived benefits, perceived barriers, and cues to action to mainly address the behavior of individuals who are not currently ill (Becker, 1974). Mechanic’s illness behavior model (Mechanic, 1982; McHugh and Vallis, 1985) is oriented to explain the individual “help seeking” activity undertaken by a person who perceives himself or herself to be ill. It integrates the biological, psychological, and social dimensions that a person experiences. Illness behavior is a dynamic process in which the individual monitors his/her body, defi nes his/her state of health, discovers a suitable remedy, and uses various forms of help and professional health care. This model has been broadly applied in, but not limited to, pain studies (Pilowsky and Spence, 1975), depression (Rief et al., 2003), coping resear ch (Lazarus and Folkman, 1984; Mok et al., 2004), and somatic diseases (de Rosa et al., 2004; Walsh et al., 2004). However, it has been rarely used in the empirical dental behavioral research. Talcott Parsons conceptualized the “sick role” in 1951 (Parsons, 1951; Parsons, 1975). Compared to Mechanic’s illness behavior model, the sick role explores people’s

PAGE 38

25 health seeking or illness avoidance behavior from a set of defined social roles. Application of sick role has been used in intensive care (Zussman, 1992), social welfare (Cole and Lejeune, 1972) and elderly and poor patients (Arluke et al., 1979). However, along with other theories or models, such as SuchmanÂ’s social psychological theory and PescosolidoÂ’s social network theory, the sick role has not been found to be applied in dental literature. Grembowski and colleagues (Grembowski et al. 1989) have developed a public health model of the dental care process that is particularly aimed at explaining the dental care process. Although the relationships used in the model were heavily based on the findings from previous studies, there is a lack of new empirical studies conducted to test the validity of the model. The Andersen behavioral model of health services use proposed by Andersen and colleagues has been widely used in medical and dental services research. The model was originally developed by Ronald Andersen almost four decades ago in an attempt to better understand familiesÂ’ use of professional health services (Andersen, 1968). The model was re-directed to explain an individualÂ’s u tilization behaviors because of the differences among family members. The original model, also the most parsimonious model, suggests that a personÂ’s use of health services is a function of a set of predisposing, enabling, and need factors. Over the past decades, the behavioral model of health services use has undergone several revisions and expansions along with the improvement of knowledge in health services research. Aday and Andersen (1974) revised the model in the 1970s by adding two domains in the model: health care system factors and consumer satisfaction. Since Andersen and colleagues realized that the ultimate goal of any given health care services

PAGE 39

26 is to maintain and improve peopleÂ’s health status, the model underwent the second phase of revision in the 1980s and early 1990s (Anders en et al., 1994) by incl uding health status outcomes. The model at that phase also emphasized the influence of external environment on health care use, and the importance of personal health practice. Given that access and utilization are not stationary, and health status can be either a cause or a consequence of health care utilization, the most recent version of this model captures the dynamic and feedback relationships among four constructs: environment characteristics, population characteristics, health behavior, and health outcomes (Andersen, 1995). Environmental characteristics can be summarized by two components: external environment and health care system. External environment refers to those physical and social environmental factors that directly impact an individualÂ’s health behavior and health outcomes. These factors include, but are not limited to, the local economic climate, relative wealth, community politics, level of violence, and prevailing social norms, etc. Health care system encompasses those factors within the health care system that influence the availability, accessibility, acceptability, and accommodation of health care (Andersen, 1995). Examples of these factors include provider to population ratio, location and type of providers, and number of hospitals. The relevant population characteristics encompass three groups: predisposing, enabling, and need factors. Predisposing factors are those biological (demographic characteristics), sociological (social structure) and psychological (health beliefs) attributes that exist before the disease. Demographic characteristics include gender, race, and age, etc. Social structure measures fact ors, such as education, occupation, and social interactions, which influence a personÂ’s status in a society. Health beliefs refer to an

PAGE 40

27 individualÂ’s values and attitudes toward the benefit of health care and knowledge of health care information. Predisposing characteristics predict how likely an individual is prone to use health care services. Some of these factors are immutable such as age, gender, and race, and some are mutable such as health beliefs and attitudes. Enabling characteristics measure an individualÂ’s ability to realize the use of health care. Enabling resources, which facilitate or impede the use of services, are not only from individuals and families, such as income and health insurance, but also from organizations and communities, such as provider supply and hospital beds. So far, organizational resources have not received enough attention in health services research (Andersen, 1995). Need factors include both self-perceived imperatives and the professional assessments about the need for care. Andersen (1995) indicated that selfperceived need is, by and large, a social phenomenon, while professionally evaluated need is more biologically oriented. A substantial discrepancy between perceived need and evaluated need for dental care has been well documented. A consistent finding is that, when evaluated need is the standard for comparison, many persons are unaware of their dental care needs (Gilbert et al., 1994; Otchere et al., 1990; Tennstedt et al., 1994; Tervonen and Knuuttila, 1988). Health behavior includes two aspects: personal health practice and the use of professional health care. Both aspects are important to maintain or improve a personÂ’s health status. Personal health practice refers to a wide range of self-care behaviors, such as self-treatment, exercises, appropriate diet and nutrition, enough sleep, safe sex, etc. It has been reported that personal health practice is so common that many health-ralated symptoms are actually treated by the use of over-the-counter medications, homemade

PAGE 41

28 remedies, or changes in activity or diet (Eisenberg et al. 1993; Kronenfeld and Wasner, 1982; Stoller et al., 1993). Gilbert et al. (2002b) reported that dental self-care could lead to an extreme: fourteen percent of tooth loss occurred outside a health care facility over a 72-month period among a dentate middle-aged and older population. A main purpose of the behavioral model is to predict an individualÂ’s behaviors in the use of professional health care. This model has been extensively used to investigate the utilization behaviors in various types of health services, which include, but are not limited to, emergency services (Huang et al., 2003), ambulatory care (Brown et al., 2004), long-term care (Bradley et al., 2002; Bradley et al., 2004), co mmunity care (Van Achterbert et al., 1996), prenatal care (LaVeist, et al., 1995), and dental care (Atchison and Andersen, 2000; Dobalian et al., 2003; Gilbert et al., 1998b). An expanded version of the behavioral model framed the serial analyses for the WHO International Collaborative Study of Oral Health Outcomes at USA research sites (ICS-II) (Andersen and Davidson, 1997; Andersen et al., 1997; Atchison et al., 1997; Atchison and Andersen, 2000; Davidson and Andersen, 1997). The behavioral model has also been used in other longitudinal oral health studies (Gilbert et al., 1990; Gilbert et al., 1998b; Lo and Schwarz, 1998). In addition to the general population, the behavioral model has also been applied to some specific populations, such as vulne rable populations (Gelberg et al., 2000), homeless people (Gallagher et al., 1997; Lim et al., 2002), old people (Bradley et al., 2002; Bradley et al., 2004), people with HI V (Dobalian et al., 2003; Dobalian et al., 2004), people with mental health problems (Albizu-Garcia et al., 2001), and low income populations (Davidson et al., 2004). Andersen (1995) suggested that predisposing, enabling, and need factors have differential abilities to predict or explain utilization

PAGE 42

29 behaviors depending on the types of health care services. For example, demographic characteristics and need factors largely account for variation in the use of services in response to serious diseases or conditions such as hospital in patient care, while social structure, health attitudes, and enabling factors are more important to predict behaviors in the use of discretionary services such as dental care. This assumption has been demonstrated by empirical evidence in prosthetic dentistry. A longitudinal study (Gilbert et al., 2002a) thoroughly investigated the factor s influencing the use of various types of dental services and posited that the use of fixed and removable prosthodontic services was discretionary, depending on certain enabling factors. Health outcome is one of the major constructs in the behavioral model of health services use. Outcomes of health services are measured by perceived health status that is assessed by individuals, evaluated health status that is judged by the health professional, and consumer satisfaction that is sensed by individuals about the quality and convenience of the services. The behavioral model suggests that both population characteristics and health utilization behavior can have a direct impact on health outcome. Population characteristics can also impact health outcome through health utilization behavior. Although this model has been extensively used in health services research, most studies are limited to addressing access and utilization problems. Less often are they expanded to investigate the effect of population characteristics and utilization behavior on health outcomes. Among the few outcome studies using the Andersen behavioral model as the conceptual framework, the WHO International Collaborative Study of Oral Health Outcomes at USA research sites (ICS-II) is a good example. The ICS-II is a multicountry and multi-city project examining the impact of environmental, health care

PAGE 43

30 system, and personal characteristics on dental care utilization and oral health outcomes (Andersen and Davidson, 1997; Andersen et al., 1997; Atchison et al., 1997; Atchison and Andersen, 2000; Davidson and Andersen, 1997). This study is aimed at investigating how selected population characteristics (predisposing, enabling, and need factors) and prosthodontic utilization behavior dynamically impact oral health outcomes. Figure 3-1 provides a schematic of the theoretical framework used in this dissertation, which is an application of the Andersen behavioral model. Figure 3-1. Theoretical Framework Multidimensional Conceptual Model of Oral Health A variety of outcomes exist for any speci fic dental treatment. Understanding how prosthodontic services may influence oral health outcomes first requires an understanding of the multiple dimensions of oral health. In this study, outcomes are assessed by multiple measures of self-rated oral health, which is a dimension of oral health suggested by a multidimensional conceptual model of oral health (Figure 3-2). The model was conceptualized in the Florida Dental Care Study (FDCS) (Gilbert et al., 1998a), by Population characteristics Health behavior Health outcomes Predisposing Enabling Need factors factors factors Various types of prosthodontic services Oral health outcomes

PAGE 44

31 adapting, with revision, the work of two groups of investigators: (1) Locker (1988), who adapted to the oral health context a model by the World Health Organization (1980), and (2) Johnson and Wolinsky (1993), who modified a model by Nagi (1976), predominantly by adding self-rated health as a construct. The construct validity of the multidimensional model of oral health has been tested by previous studies (Gilbert et al. 1998a). Self-rated oral health Figure 3-2. Multidimensional Conceptual Model of Oral Health. Reprinted with permission from Gilbert GH, Duncan RP, Heft MW, Dolan TA, Vogal WB (1998). Multidimensionality of oral health in dentate adults. Med Care 36: 988-1001 (Figure 1, Page 989). This multidimensional model posits a sequential causal process that involves specific antecedents and consequences, and parallels the biomedical conception of the natural history of oral disease. This model co mprises five domains of oral health and oral health-related quality of life: (1) Oral disease and tissue damage connote disorders at the organic level, such as dental decay, periodontal diseases, or tooth loss. This domain is measured by both clinical examination and self-assessment. Previous work (Gilbert et Oral disease & tissue damage Oral pain & discomfort Oral functional limitation Oral disadvantage Self-rated oral health

PAGE 45

32 al., 2002c) documented a high agreement between clinically derived tooth loss and selfreported tooth loss. (2) Oral pain and discomfort refer to painful or uncomfortable experiences as a response to oral disease and tissue damage, such as toothache pain or denture soreness. (3) Oral functional limitation refers to the compromised physiological or psychological function caused by oral dis ease and tissue damage, or oral pain and discomfort. Examples include chewing di fficulty or difficulty speaking or pronouncing words. (4) Oral disadvantage is a more socially involved dimension. It connotes that normal social functions such as laughing, sleeping, or eating with others, are compromised. Oral disadvantage may be a consequence of oral disease and tissue damage, oral pain and discomfort, or oral functional limitation. (5) Self-rated oral health is the global assessment of oral health. Oral disease and tissue damage, oral pain and discomfort, oral functional limitation, or oral disadvantage may have different influences on this dimension. Examples in this domain include satisfaction with chewing ability, satisfaction with dental appearance, and self-rated overall oral health. Oral functional limitation and oral disadvantage reflect whet her or not specific behaviors have been exhibited, such as whether persons have actually avoided chewing certain foods or avoided laughing. However, self-rated oral health is entirely subjective because it is an unobserved assessment. As depicted in Figure 3-2, the five dimensi ons of oral health are organized based on the proximal-distal continuum proposed by Bre nner et al. (1995). “Proximal” dimensions (oral disease and tissue damage, and oral pain and discomfort) refer to specific signs and symptoms linked most directly with a pathologic origin or an anatomic structure. In contrast, “distal” dimensions (oral functional limitation, oral disadvantage, and self-rated

PAGE 46

33 oral health) encompass broader functional, social, and psychological sequelae of oral diseases or conditions (Peek et al., 1999). Gilbert et al. (1998a) suggested that the changeability of oral health varies depending on the dimensions. The “distal” dimensions of oral health may be less likely to change over time than “proximal” dimensions because the “distal” dimensions are influenced by more factors (Foerster et al., 1998; Gilbert et al., 1997a; Gilbert et al., 1997b). Previous analysis using the FDCS data (Peek et al., 1999) has demonstrated such patterns: dimensions of oral disease and tissue damage, and oral pain and discomfort exhibited more variability than the dimensions of oral functional limitation, oral disadvantage, and self-rated oral health. However, the dimension of selfrated oral health was considerably more changeable than suggested by the “distal” position in the model. Another longitudinal study also showed that the effect of dental treatment varied across dimensions (Fiske et al., 1990). Problems in the “proximal” dimensions were more responsive to dental treatment than those in the “distal” dimensions (Fiske et al., 1990). This is not surprising given that the “proximal” dimensions of oral health represent disorders at the organic level and their immediate painful or uncomfortable responses. Thus, the treatment is direct and more likely to precipitate observable or perceivable changes. In this study, the dimension of self-rated oral health is the outcome of interest. It is the most “distal” dimension suggested by this model and is totally subjective. Three measures under this dimension are included in this study: satisfaction with chewing ability, satisfaction with dental appearance, and self-rated overall oral health.

PAGE 47

34 CHAPTER 4 RESEARCH DESIGN AND METHODS This chapter describes the methodology that is used in this dissertation. Included are descriptions of sample and sampling procedures, data collection methods, variables, and statistical methods. Sample and Sampling Procedures This dissertation uses data from a popul ation-based longitudinal study, the Florida Dental Care Study (FDCS), to address the research questions. The overall purpose of the FDCS was to develop a risk assessment model of longitudinal oral health outcomes in middle-aged and older adults (Gilbert et al., 1998a; Gilbert et al., 1998b; Peek et al., 1999). The aims of the FDCS included (1) describing the long-term changes in specific oral health outcomes; (2) describing the demand for specific dental services over time; (3) determining the predisposing, enabling, and need factors for specific dental services; and (4) determining the dental behavioral, attitudinal, and socioeconomic predictors of longitudinal oral health outcomes. Thus, the sampling strategy of the FDCS was to identify and recruit a large number of persons at hypothesized increased risk for oral health decrements, namely, low income persons, blacks, residents of rural areas, and those 45 years old or older (Gilbert et al., 1997c). Four counties in north Florida were selected: one metropolitan county (Duval County), and three non-metropolitan counties (Hamilton, Jefferson, and Madison Counties). These four counties are geographically close to each other and located near to

PAGE 48

35 the University of Florida, where the study was originally administrated. The four counties also provide a rural-urban contrast, and have large proportions of blacks, older adults, and low-income persons (Gilbert et al., 1997c). Since the metropolitan county (Duval County) also contains sparsely populated areas, only persons who resided in one of the 21 urbanized zip-code areas in this county were eligible for the study. Three nonmetropolitan counties were considered as “rural” counties. Although the classification was slightly different than the definition used by the US Bureau of the Census (US Bureau of the Census, 1992), the total number and density of the county populations make them more “rural” than “urban” (Gilbert et al., 1997c). Low-income persons were then defined using the poverty level provided by the 1990 US Bureau of the Census (US Bureau of the Census, 1992). The homeless and persons who resided in nursing homes, adults congregate living facilities, adult foster homes, hospices, military installations, or correctional facilities were excluded from the population of interest. Based on the information provided by the Census Bureau specifically for this project (US Bureau of the Census, 1994), the exclusion accounted for 4 percent of the 45-year-old and older population in the targeted sampling areas. Since tooth loss was one of the major oral heath outcomes of interest in FDCS study, only persons who had at least one natural tooth remaining were included for the study. The final list of eligible persons included those who (1) had at least one natural tooth; (2) lived in either Duval, Hamilton, Jefferson, or Madison Counties; (3) if lived in Duval County, lived in one of 21 specified zip-code areas; (4) had a telephone in their household; (5) were 45 years old or older; (6) did not live in an institutional setting; (7)

PAGE 49

36 spoke English; (8) were either Black or non-Hispanic White; and (9) were capable of engaging in a coherent telephone conversation. A telephone-screening phase began in May 1993 and lasted nearly 3 months. A dual sampling frame telephone screening methodology was designed: both random-digitdialing (RDD) method and directory-listed sampling were employed to achieve both greater coverage and greater cost efficienc y. A computer-assisted telephone interviewing (CATI) was used to facilitate the screening. Ultimately, a pool of 5,254 persons met the age, residence, race, and income-related criteria. Of these 5,254 persons, 3,998 persons met the additional requirement for dentate status. After the eligible population was identified, a stratified random sample of 1800 subjects was selected for a baseline in-person interview and clinical dental examination. Of them, 707 subjects were contacted but refused, 125 were unreachable, 95 had died or were judged ineligible, and, finally, 873 subjects were successfully recruited for baseline data collection. The final participants resulted in a representative sample of the population of interest. At baseline, this sample had similar dental visit behaviors and corresponding sociodemographic determinants compared to the National Health Interview Survey (NHIS) data (Gilbert et al., 1997c). The informed consent of all human subjects who participated in the investiga tion was obtained after a protocol approved by the Institutional Review Board of the University of Florida and the University of Alabama at Birmingham. Data Collection Methods The field phase of baseline data collection from the 873 participants began in August 1993 and ended in April 1994. Trained interviewers administrated the in-person interview, which typically lasted about 30 minutes. Test and retest reliability of the

PAGE 50

37 interview questions were estimated afterwards and judged to be satisfactory. Immediately following the in-person interview, a clinical dental examination was conducted. Previous studies (Dolan et al., 1997; Gilbert et al., 1996; Ringelberg et al., 1996) have described in detail the examination protocol, clinical diagnostic criteria, and quantified interexaminer reliability ( ranges from 0.82 to 0.91, depending on different types of examinations), all of which were reported to be satisfactory. Information from the clinical dental examination, with the exception of the number of remaining teeth, is not of major interest in this dissertation. Briefly, the examination recorded the presence and location of remaining teeth, root fragment s (missing more than three-fourths of the clinical crown), bulk restoration fractures (m issing, partly missing, or fractured fillings), fractured teeth involving the dental cusp and/or incisal edges, severe root defects, teeth that were severely mobile (non-physiologic occluso-apical movement or more than 2 mm bucco-lingual movement), and worst site per tooth based on the periodontal attachment level relative to the cemento-enamel junction. A follow-up telephone interview occurred every 6 months after the baseline interview. Every 24 months, post baseline, an in-person interview was conducted and was followed immediately by a clinical examination identical to the one conducted at baseline. The active data collection lasted for 72 months. In summary, subjects participated in an in-person interview and clinical dental examination at baseline, 24 months, 48 months, and 72 months. These were followed by only telephone interviews at 6, 12, 18, 30, 36, 42, 54, 60, and 66 months following baseline. In this dissertation the first 24 months of data (baseline, 6-month, 12-month, 18-month, and 24-month data) are used because certain measures are not available at some of the subsequent follow-ups.

PAGE 51

38 By the end of the 24-month data collection period, 764 (unweighted) persons remained in the study, of whom 723 (unweighted) participated in a clinical examination. Loss of follow-up included 35 (unweighted) indi viduals who refused to participate, 29 (unweighted) deceased individuals, 10 (unweight ed) individuals who were not medically able to participate, and 35 (unweighted) indi viduals who could not be located. Possible bias due to subject attrition has been evaluated previously by comparing characteristics of those who participated at the 24-month interview with those who did not (Gilbert et al., 1998b; Peek et al., 1999). Briefly, persons who participated were more likely to have been well-educated (92% of high school graduates participated through 24 months, compared to 84% of non-graduates; 2 test, p<0.05), regular dental attenders (92% of regular dental attenders participated, compared to 88% for problem-oriented attenders; 2 test, p< 0.05), above the 100% poverty threshold (92% of the non-poor, compared to 84% of the poor; 2 test, p<0.05), in better self-rated general health (91% of those who rated their health as excellent participated, compared to 83% for those who rated their health as poor; Mantel-Haenszel 2 trend test, p<0.05), White (92% of whites, 85% of blacks; 2 test, p<0.05), and had more teeth present at baseline (the mean number of teeth present at baseline ± standard deviation was 21.0 ±7.2 for the 723 24-month clinical examination participants; for the nonparticipants it was 20.5 ± 8.0). No differences were observed with respect to age group, sex, area of residence, ability to pay an unexpected $500 dental bill, or present financial situation. In all previous investigations, data were weighted using the sampling proportions to reflect the population in the counties studied. Table 41 shows the number of participants (unweighted and weighted) at 6-month intervals between baseline and 24 months (Gilbert et al., 2002b).

PAGE 52

39 Table 4-1. Number of Participants between Baseline and 24 Months Number of participants Months after baseline Interview methods Unweighted number Weighted number 0 (Baseline) In-person Interview & Clinical Exam 873 873 6 months Telephone Interview 849 856 12 months Telephone Interview 819 829 18 months Telephone Interview 805 817 24 months In-person Interview & Clinical Exam 764 788 Description of Variables Variables included in this dissertation are framed according to the model of health services use (Andersen, 1995). Dependent variables, independent variables, and control variables are described in this section. Dependent Variables In this dissertation, oral health outcomes are assessed by self-rated oral health measures suggested by the multidimensional conceptual model of oral health, which is graphically depicted in Figure 3-2. The construct validity of this model has been documented previously (Gilbert et al., 1998a). This model suggests five dimensions of oral health; however, only the self-rated oral health domain is of interest in this dissertation, and thus is included in the analysis. Under the domain of self-rated oral health, three measures are included. Three original measures under the domain of self-rated oral health Under the domain of self-rated oral health, three original measures are selected in this dissertation : satisfaction with chewing ability, satisfaction with dental appearance, and self-rated overall oral health. These three domains are the patientÂ’s overall assessment of his or her chewing ability, dental appearance, and global oral health. Satisfaction with chewing ability and satisfaction with dental appearance were rated as

PAGE 53

40 “1 = very satisfied,” “2 = satisfied,” “3 = dissatisfied,” and “4 = very dissatisfied.” Selfrated overall oral health was measured by asking participants to rate their overall oral health or present condition of teeth as “1 = excellent,” “2 = very good,” “3 = good,” “4 = fair,” or “5 = poor.” What should be noticed here is that the higher response scales stand for lower levels of satisfaction with chewing ability, lower levels of satisfaction with dental appearance, and lower levels of overall oral health. Dynamic changes in self-rated oral health To improve and maintain oral health are nearly universally accepted as the objectives of dental care. The previous anal yses from the FDCS suggest that dental care acts in both a curative and preventive manner. The transition probability of positive change, which measures the improvement of already deteriorated oral health, suggests a curative manner. The transition probability of negative change, which measures the maintenance of oral function from deterioration, suggests a preventive manner. In addition, the transition probability of negative change can capture the potential harmful consequences that prosthodontic services may exert. In order to assess the dynamic changes in the three original measures, six new dependent variables are created in this dissertation based on the coding systems depicted in Figure 4-1. Improvement of self-rated oral health. Improvement of self-rated oral health is recoded as a dichotomous variable including two categories: “1 = improved” and “0 = other.” “Improvement” is defined as reporting better self-rated oral health measures at the TX interview than those measures at the TX-1 interview (Figure 4-1). Since the lower response scales stand for higher levels of satisfaction with chewing ability, higher levels of satisfaction with dental appearance, and hi gher levels of overall oral health, if lower response scales were reported in TX than TX-1 (SX < SX-1), it is considered as having

PAGE 54

41 improvement, and coded as “1.” Otherwise, it is coded as “0.” Since there are three measures of self-rated oral health, three new improvement variables are created: improvement of satisfaction with chewing ability , improvement of satisfaction with dental appearance , and improvement of self-rated overall oral health . Deterioration of self-rated oral health. Deterioration of self-rated oral health is also recoded as a dichotomous variable to include two categories: “1 = deteriorated” and “0 = other.” “Deterioration” is defined as reporting worse self-rated oral health measures at the TX interview than those measures at the TX-1 interview (Figure 4-1). Because the higher response scales stand for lower levels of satisfaction with chewing ability, lower levels of satisfaction with dental appearance, and lower levels of overall oral health, if higher response scales were reported in TX than TX-1 (SX > SX-1), it is considered as having deterioration, and coded as “1.” Otherwise, it is coded as “0.” Thus the variables included in this domain consist of deterioration of satisfaction with chewing ability , deterioration of satisfaction with dental appearance , and deterioration of self-rated overall oral health . Independent Variables Original measures of prosthodontic services use The original measures of prosthodontic services use in the FDCS were various types of prosthodontic treatment. Prost hodontic treatment was measured by asking subjects whether or not they had the following dental procedures performed in the previous 6 months: a dental crown, cap, fixed bridge, implant, removable partial denture, or full denture made or repaired.

PAGE 55

42 Figure 4-1. Coding of Dynamic Dependent Variables Current interval Interview time TX-1 TX a Improvement of self-rated oral health b (coded as “1”) SX-1 > SX c (A lower scale of rating was reported in TX than TX-1) Other (coded as “0”) SX-1 SX (A higher or equal scale of rating was reported in TX) Deterioration of self-rated oral health (coded as “1”) SX-1 < SX (A higher scale of rating was reported in TX than TX-1) Other (coded as “0”) SX-1 SX (A lower or equal scale of rating was reported in TX) a TX-1 and TX refer to the time of the interviews. For example, if TX is the 12-month interview point, TX-1 connotes the 6-month interview point. b Self-rated oral health includes satisfaction with chewing ability, satisfaction with dental app earance, and self-rated overall oral health. c SX-1 connotes ratings of self-rated oral health reported at TX-1, and SX connotes ratings reported at TX.

PAGE 56

43 Recoded prosthodontic services use For the purpose of this dissertation, a dental crown, cap, fixed bridge, and implant are categorized as fixed prosthodontic treatment, while a removable partial denture and full denture are categorized as removable pr osthodontic treatment. Figure 4-2 graphically depicts the coding scheme for the prosthodontic services. Given that some of the prosthodontic services could take more than one 6-month interval to finish, services in both the current interval (the interval between the TX-1 and TX interviews) and the interval immediately after it (TX to TX+1) are measured (Figure 4-2). If a prosthodontic service was reported in the current interval but no service was reported in the interval immediately after it, the service is classified as having been finished with the current interval. If a prosthodontic service was reported in the current interval and the interval immediately after, it is considered as not having been finished within the current interval and a follow-up service was needed. Since both finished and unfinished prosthodontic services are of interest in this dissertation, four dichotomous variables are created based on the subjects’ prosthodontic service status. Finished fixed prosthodontic treatment. Fixed prosthodontic treatment finished within the current interval is coded as “1.” Otherwise, it is coded as “0.” Unfinished fixed prosthodontic treatment. Unfinished fixed prosthodontic treatment that required completion in the next interview interval is coded as “1.” Otherwise, it is coded as “0.” Finished removable prosthodontic treatment. Removable prosthodontic treatment finished within the current interval is coded as “1.” Otherwise, it is coded as “0.” Unfinished removable prosthodontic treatment. Unfinished removable prosthodontic treatment that required completion in the next interview interval is coded as “1.” Otherwise, it is coded as “0.”

PAGE 57

44 Figure 4-2. Coding of Prosthodontic Services Current interval Interval immediately following Interview time TX-1 TX TX+1 a Fixed prosthodontic services (f) Finished fixed prosthodontic treatment (coded as “1”) 1(f) b 0(f) c (Received a fixed prosthodontic service during the TX-1 to TX interval, and the service was finished within this current interval) Unfinished fixed prosthodontic treatment (coded as “1”) 1(f) 1(f) (Unfinished fixed prosthodontic treatment continued into the following interval: TX to TX+1 interval) Removable prosthodontic services(r) Finished removable prosthodontic treatment (coded as “1”) 1(r) d 0(r) e (Received a removable prosthodontic service during the TX-1 to TX interval, and the service was finished within this current interval) Unfinished removable prosthodontic treatment (coded as “1”) 1(r) 1(r) (Unfinished removable prosthodontic treatment continued into the following interval: TX to TX+1 interval) a TX-1, TX and TX+1 refer to the time of the interviews. For example, if TX is the 12-month interview point, TX-1, and TX+1 connote the 6-month and the 18-month interviews, respectively. b 1(f) connotes having a fixed prosthodontic service during the 6-month interval. c 0(f) connotes not having a fixed prosthodontic service during the 6-month interval. d 1(r) connotes having a removable prosthodontic service during the 6-month interval. e 0(r) connotes not having a removable prosthodontic service during the 6-month interval.

PAGE 58

45 Control Variables Andersen’s behavioral model of health services use indicates that population characteristics not only impact health outcomes through health services utilization, but also have a direct effect on health outcomes. In general, population characteristics consist of three domains: predisposing, enabling, and need factors (PEN factors). In this dissertation, the relationship of main interest is the effect of prosthodontic services use on self-rated oral health outcomes. Although the direct and indirect relationships between PEN factors and oral health outcomes are not of primary interest, their effects should be controlled for. Some control variables are recoded from their original forms because the preliminary analyses indicate that the original response levels of some control variables cause convergence problems in the longitudinal regression analyses. Control variables in each domain are described in this section. Predisposing factors For the purpose of this dissertation, predisposing variables include the following demographic and social structural variables: ag e, gender, area of residence, race, level of formal education, self-rated general health, approach to dental care, frustration with past care, and whether the dentist whom the patient frequently sees is one that is nearest to the home residence. All predisposing characteristics were queried at the baseline in-person interview, with the exception of self-rated general health, which was measured during the telephone screening interview. In this dissertation, age is recoded from a continuous variable into a dichotomous variable: 45 to 64 years old and 65 years old or older. The original measure of education contains eight le vels. In this dissertation, it is reclassified into two levels: completed at least high school and did not complete high school. The original response of self-rated general health is in a five-point scale: “excellent,” “very

PAGE 59

46 good,” “good,” “fair” or “poor.” The variable is recoded into a dichotomous variable as “good” and “not good.” The original question of approach to dental care contains four ordinal responses: (1) “I never go to dentist”; (2) “I go to dentist when I have a problem or when I know that I need to get something fixed”; (3) “I go to a dentist occasionally, whether or not I have a problem”; and (4) “I go to dentist regularly.” For the purpose of this dissertation, persons who responded #1 or #2 are classified as “problem-oriented attenders” and those who responded #3 or #4 are classified as “regular attenders.” Frustration with past care was originally measured by asking subjects whether they received dental treatment that had not worked, or had not lasted as long as they thought it should have. The original five ordinal scales are collapsed into two: (1) frustrated and (2) not frustrated. Enabling factors Enabling factors were also captured at the baseline in-person interview. These variables include poverty status relative to 150% thresholds, ability to pay an unexpected $500 dental bill, and insurance status. Poverty status relative to 150% thresholds was defined by criteria provided by the 1990 U.S. Bureau of the Census (US Bureau of the Census, 1992). Need factors Data on need variables, which are of interest in this dissertation, were gathered by selected self-reported items at baseline and at the follow-up interviews. Variables that were gathered only at the baseline in-person interview include the number of occluding pairs of teeth, having a sore denture, havi ng a broken denture, and having a dry mouth. Some researchers believe that as long as a person retains ten occluding pairs of teeth, for a total of 20 well-distributed teeth, no obvious advantage will be gained by replacing

PAGE 60

47 missing teeth with a prosthodontic appliance. Ho wever, not every dentist agrees with this opinion. A previous study did show the benefi cial effect of prosthodontic services for people with more than 10 pairs of teeth (Gilbert et al., 2004). Thus, in this dissertation, the number of occluding pairs of teeth is calculated based on the record from the baseline clinical examination to divide subjects into two groups: subjects with ten or more than ten occluding pairs of teeth and subjects with fewer than ten occluding pairs of teeth. Need variables, which were gathered only at follow-up interviews, include perceived need for dental care, problems with current prostheses, and tooth loss during the past 6 months. Perceived need for dental care is recoded into a binary variable from the original measure with five response categories: (1) “yes, for a routine check up”; (2) “yes, for a dental problem”; (3) “no, the dental problem can wait”; (4) “no, the mouth is in a good shape now or do not have a dental problem”; and (5) “no, do not need to see a dentist.” The newly recoded perceived need includes two response categories: “1 = not at this time,” and “0 = other.” Problem with existing prosthodontic appliance is constructed from three self-reported items. These items queried whether the subjects had the problems of (1) “loose cap or bridge,” (2) “sore denture,” or (3) “denture broken,” during the prior 6-month interval. Subjects who answered affirmatively to ANY of the three questions are considered as having a problem with an existing prosthodontic appliance (coded as “1”). Subjects who answered negatively to ALL of the three questions are classified as not having a problem with an existing prosthodontic appliance (coded as “0”). Tooth loss was measured by asking subjects whether they had lost any teeth or had any teeth removed during the prior 6-month interval. Agreement between self-reported and clinically examined tooth loss was high (Gilbert et al., 2002c).

PAGE 61

48 Need variables gathered at both the baseline in-person interview and the follow-up interviews include various self-reported dent al problems/oral conditions. Some dental problems/oral conditions, such as having a l oose cap or bridge, or having a broken tooth or cap, constitute a direct need for prosthodon tic treatment. Other dental problems/oral conditions, such as cavities or abscessed teeth, are not initially treated by prosthodontic care. However, these oral conditions represent the need for other types of dental care services, and can have a profound impact on subjectively perceived oral and dental health. Thus, in this dissertation, measures of other oral conditions are also included in the analysis to control for the impact of these conditions on self-rated oral health outcomes. These need variables include having a loose cap or bridge, having a broken tooth or cap, having a broken filling, having a toothache, having a sensitive tooth, having a cavity, having an abscessed tooth, having inf ected or sore gums, having bleeding gums, having a loose tooth, having stained teeth, and having a problem with bad breath. All of the above described variables, with response categories, are summarized in Table 4-2. The corresponding original questionnaire items are listed in Appendix A. Table 4-2. All Variables with Response Categories Variables Response levels Dependent Variables Original measures of self-rated oral health Satisfaction with chewing ability 1 = very satisfied 2 = satisfied 3 = dissatisfied 4 = very dissatisfied Satisfaction with dental appearance 1 = very satisfied 2 = satisfied 3 = dissatisfied 4 = very dissatisfied Self-rated overall oral health 1 = excellent 2 = very good 3 = good 4 = fair 5 = poor

PAGE 62

49 Table 4-2. Continued Variables Response levels Improvement of self-rated oral health Improvement of satisfaction with chewing ability 1 = improved 0 = other Improvement of satisfaction with dental appearance 1 = improved 0 = other Improvement of self-rated overall oral health 1 = improved 0 = other Deterioration of self-rated oral health Deterioration of satisfaction with chewing ability 1 = deteriorated 0 = other Deterioration of satisfaction with dental appearance 1 = deteriorated 0 = other Deterioration of self-rated overall oral health 1 = deteriorated 0 = other Independent Variables Baseline prosthodontic services use Wearing a full denture 1 = yes 0 = no Wearing a partial denture 1 = yes 0 = no Longitudinal prosthodontic services use Finished fixed prosthodontic treatment 1 = finished fixed prosthodontic treatment 0 = otherwise Unfinished fixed prosthodontic treatment 1 = unfinished fixed prosthodontic treatment 0 = otherwise Finished removable prosthodontic treatment 1 = finished removable prosthodontic treatment 0 = otherwise Unfinished removable prosthodontic treatment 1 = unfinished removable prosthodontic treatment 0 = otherwise Control Variables Predisposing factors Age 1 = 65 years old or older 0 = 45 to 64 years old Gender 1 = male 2 = female Area of residence 1 = urban 0 = rural Race 1 = White 2 = Black Level of formal education 1 = completed at least high school 0 = did not complete high school Self-rated general health 1 = good 2 = not good Approach to dental care 1 = regular attender 0 = problem attender

PAGE 63

50 Table 4-2. Continued Variables Response levels Frustration with past dental care 1 = frustrated 0 = not frustrated Whether the dentist whom the patient frequently sees is one who is nearest 1 = yes 2 = no Enabling factors Poverty Level 1 = below 150% poverty level 0 = not below 150% poverty level Able to pay an unexpected $500 dental bill 1 = able to pay comfortably 2 = able to pay but with difficulty 3 = not able to pay Dental insurance status 1 = has dental insurance 2 = no dental insurance Need factors Need factors measured only at baseline Number of occluding pairs of teeth 1 = 10 occluding pairs of teeth or more 0 = fewer than 10 occluding pairs of teeth Has a sore denture 1 = yes 2 = no Has a broken denture 1 = yes 2 = no Has a dry mouth 1 = yes 2 = no Need factors measured only at follow-up intervals Perceived need for dental care 1 = not at this time 0 = other Problems with current prostheses 1 = yes 0 = no Tooth loss in the past six months 1 = yes 2 = no Need factors measured at both baseline and follow-up interviews Has a loose cap or bridge 1 = yes 2 = no Has a broken tooth or cap 1 = yes 2 = no Has a broken filling 1 = yes 2 = no Has a toothache 1 = yes 2 = no Has a sensitive tooth 1 = yes 2 = no Has a cavity 1 = yes 2 = no Has an abscessed tooth 1 = yes 2 = no Has infected or sore gums 1 = yes 2 = no

PAGE 64

51 Table 4-2. Continued Variables Response levels Has bleeding gums 1 = yes 2 = no Has a loose tooth 1 = yes 2 = no Has stained teeth 1 = yes 2 = no Has a problem with bad breath 1 = yes 2 = no Statistical Methods In this dissertation, data are weighted using the sampling proportions to reflect the population in the counties studied (Gilbert et al.1997c). Except where specified to the contrary, numbers and percentages reported in this dissertation are weighted values. All comments about statistical significance in this dissertation refer to probabilities of less than 0.05. With the exception of prosthodontic services, this dissertation focuses on the first 24 months data (baseline, 6-month, 12-month, 18-month, and 24-month data). With reference to prosthodontic services, the 30-month interview is used because such information is needed to classify whether or not the service had been finished within the 24-month interview point (as depicted in Figure 4-2). Analytical Methods for Objective 1 Descriptive analyses are conducted to quantify the incidence of prosthodontic services use and the prevalence of self-rated oral health outcomes. Similar analyses are also performed to quantify the incidences of dynamic changes in self-rated oral health outcomes during a 24-month period.

PAGE 65

52 Analytical Methods for Objective 2 Cross-sectional bivariate analyses for research question 2a Baseline satisfaction with chewing ability is dichotomized as “1 =very satisfied/satisfied,” and “2 = very dissatisfied/dissatisfied.” Since all variables included in the bivariate analyses are nominal variables, chi-square tests are used to quantify the bivariate associations between baseline satisfaction with chewing ability and other relevant factors, namely, baseline use of prostheses and PEN factors. Cross-sectional multivariate logistic regression for research question 2b Logistic regression analysis is conducted to explore multivariate associations between baseline satisfaction with chewing ability and baseline use of prostheses controlling for PEN factors. Since previous studies (Käyser et al., 1990; Witter et al., 1999) have argued that prosthodontic services are not needed as long as a person retains ten occluding pairs of teeth for a total of 20 well-distributed teeth (“shortened dental arch”), the associations between baseline satisfaction with chewing ability and the number of occluding pairs of teeth, with othe r factors taken into account, is also of interest in this analysis. Since the measure of satisfaction with chewing ability is on a four-point ordinal scale, an ordered logistic regression fitted with a proportional odds model is used to conduct the analysis. A proportional odds model is an appropriate statistical modeling to test the relationship between an ordinal outcome variable and a set of explanatory variables (Stokes et al., 2001). PROC LOGISTIC in SAS fits the proportional odds model whenever the response variable has more than two levels. A basic assumption of the proportional odds model (“proportional odds assumption”) is that the slope parameter

PAGE 66

53 is the same across all levels of an outcome variable (Allison, 1999). The proportional odds assumption is tested to ensure that using this model is appropriate. Longitudinal logistic regression with correction of selection bias for research question 2c A longitudinal logistic regression with corr ection of selection bias is conducted to predict the effect of prosthodontic services use on satisfaction with chewing ability at each follow-up interview point during a 24-month period. Compared to the baseline cross-sectional logistic regression for research question 2b, the method used to answer research question 2c has two features: both the issue of repeated measures for the same individual throughout a 24-month period and the issue of treatment selection bias are taken into consideration. One common problem encountered in any longitudinal data analysis is the observations are not independent. In other words, a person’s responses at multiple points in time are correlated. If we ignore the correlation and treat the observations as though they are independent, we will usually end up with underestimated standard errors, overestimated statistics, and inefficient coefficient estimates (Allison, 1999). The generalized estimating equation (GEE) approach (Liang and Zeger, 1986; Liang and Zeger, 1993), which is an extension of generalized linear models, is a relatively new method for longitudinal data analysis. Briefly, in the GEE approach, the maximum likelihood estimates can be acquired by a GEE algorithm called “iterative generalized least squares.” The off-diagonal elements in the weight matrix W, which is in the matrix formulation of generalized least squares, are functions of the correlations among the observations. These correlations are re-estimated at each iteration. By doing this, efficient estimates of the coefficients and improved standard error estimates can be

PAGE 67

54 obtained (Stokes et al., 2001). In SAS, GEE estimation is invoked with the REPEATED statement in the GENMOD procedure (SAS Institute, Inc., Cary, NC). In the GENMOD procedure, the TYPE option is used to specify the correlation structure among an individualÂ’s observations within each time period. Liang and Zeger (1986) suggested several forms of working correlation structures, but the GEE approach does not require strong assumptions about the actual correlation structure to provide consistent estimators of the regression coefficients and their variances. UN stands for an unstructured correlation matrix when the correlation matrix is completely unspecified. It provides the most efficient estimator for and is appropriate when the number of subjects is large relative to the number of time points. EXCH stands for an exchangeable working correlation structure when the correlations between all time points are assumed to be equal. But the assumption of constant correlation between any two observations at the different time points may not be held in longitudinal data. AR stands for a lag-1 autoregressive correlation structure (Allison, 1999; Stokes et al., 2001). For the purpose of this dissertation, autoregressive correlation structure (AR) is chosen because, as indicated in Figure 3-1, the outcomes (Y) have a feedback impact on the prosthodontic treatment (T). Since the longitudinal data is used in this dissertation, the feedback impact is not simultaneous. Instead, the previous outcomes (Yt-1) will affect the current utilization behavior (Tt). Because the current outcomes (Yt) are affected by the current utilization behavior (Tt), the current outcomes (Yt) are associated with the previous outcomes (Yt-1) through the current utilization behaviors (Tt). Thus this model has an autoregressive problem. Autoregressive corre lation structure (AR) can closely reflect the actual correlation structure of this model.

PAGE 68

55 Because the measure of satisfaction with chewing ability is on a four-point ordinal scale, theoretically, a proportional odds model fitted with GEE should be used to conduct the analysis. As mentioned previously, when a proportional odds model is used, the proportional odds assumption should be tested to ensure this model is correctly applied. However, there is no mature method to test the proportional odds assumption fitted with GEE. Although Stiger and colleagues (1999) have reported a SAS macro to test the proportional odds assumption fitted with GEE, the validity of the SAS macro has not been justified by other studies. Because of the above methodological limitation, satisfaction with chewing ability is dichotomized as “1 = very satisfied/satisfied,” and “2 = very dissatisfied/dissatisfied.” Thus, a binomial logistic regression fitted with GEE is conducted. When investigating the true effectiveness of treatment, a randomized controlled trial (RCT) is considered as the “gold standard.” However, RCT may not be feasible due to ethical considerations of withholding treat ment or other restrictions. Under such circumstances, a properly designed and implemented observational study is a good alternative. Although it has been demonstrated in empirical research that observational studies can provide a valid and invaluable source of information, selection bias is a generic drawback that arises when investigators have no control over the treatment assignment. In other words, subjects self-select into a treatment group instead of being randomly assigned to it. Therefore, large differences on observed and unobserved covariates in the treatment and non-treatment groups may exist, and these differences could lead to biased inferences of treatment effects. For example, if some unobserved characteristics make some subjects more likely to self-select into a treatment group and

PAGE 69

56 these unobserved characteristics are related to the outcome, the estimated effect of the treatment on the outcome will be biased. The sign and the strength of the correlations between the unobserved characteristics and the treatment variable, as well as the outcome variable, affect the bias. The possible ways of biased estimated effects are depicted in Figure 4-3. Figure 4-3. Possible Ways of Biased Estimated Treatment Effects Due to Selection Bias Several statistical methods have been proposed to adjust for selection bias due to observed and unobserved confounders. The propensity score, which is defined as the conditional probability of being treated given the individualÂ’s covariates, has been used to reduce bias in observational studies (DÂ’Agostino, 1998; Posner et al., 2001; + + will be too positive will be too negative + + will be too negative will be too negative Unobserved Characteristics Treatment Outcome Unobserved Characteristics Unobserved Characteristics Unobserved Characteristics Treatment Treatment Treatment Outcome Outcome Outcome

PAGE 70

57 Rosenbaum, 1995). However, one weakness of propensity score matching is that it only accounts for the bias caused by the observed covariates, but does not correct for the bias caused by the unobserved covariates. Heckman two-stage estimation (Crown et al., 1998; Heckman, 1979; Kennedy, 2003) is another option that has been widely used in addressing selection bias. An earlier investigation using the FDCS data found that accounting for selection bias using Heckman’s two-stage estimation made a major difference in the substantive conclusions concerning the outcome of interest (Shelton et al., 2003). This model is conducted in two stages. The first-stage model is aimed at modeling the probabilities of “selecting” into the treatment groups of interest. The predicted probabilities from this model are then used to create a new covariate, which is called an Inverse Mill’s Ratio (IMR). The IMR is entered into a second-stage model along with the treatment variable to predict outcomes. Successful use of this method usually requires that at least one variable in the first-stage model is not included in the second-stage model (Winship and Mare, 1992). A problem with the Heckman’s estimator is that it is sensitive to model specification. If the underlying assumptions of the models are violated, the Heckman procedure yields inconsistent estimates (Winship and Mare, 1992). In addition, the Heckman’s estimates tend to fluctuate depending on which variables are included in the first-stage model (Winship and Mare, 1992). The instrumental variable approach is another widely used alternative to address unmeasured confounders (McClellan and Newhouse, 1997; McClellan et al., 1994; Newhouse and McClellan, 1998; Posner et al., 2001) . This approach is also conducted in two stages. In the first stage of the analysis, an instrumental variable or a set of

PAGE 71

58 instrumental variables are used as independent variables to predict exposure status (i.e., prosthodontic treatment status in this dissertation). In the second stage of the analysis, the differences in outcomes are examined as a function of differences in predicted exposure. Those instrumental variables included in the first stage of the analysis are excluded from the second stage. A common problem for the instrumental variable approach is to find a valid instrument. For a variable, or group of variables, to be considered a valid instrument, it should be neither associated with the outcome beyond its effect on exposure, nor associated with unmeasured confounders. However, the lack of association with unmeasured confounders can be only conceptually credible and cannot be empirically tested (Greenland, 2000). Another problem of the instrumental variable approach is that when the instrument variables are not highly predictive of exposure, this usually leads to a reduction in precision (Posner et al., 2001). Figure 4-4 illustrates the instrumental variable approach used in this dissertation. Figure 4-4. Illustration of the Instrumental Variable Approach Z has the following properties: (1) Z is independent of U; (2) Z is associated with T; (3) Z is independent of Y given T and U Z (Instrumental variable) T (Prosthodontic treatment) Y (Oral health outcomes) U (Unobserved variable)

PAGE 72

59 Since each of the above methods is not without its shortcomings, for the purpose of this dissertation, both Heckman’s two-stage estimation approach and the instrumental variable approach were originally attempted. To use the Heckman’s two-stage estimation approach, a two-stage probit model using GEE (Shelton et al., 2003) was applied. It is well known that an underlying assumption of the probit model is that the error term is normally distributed. Nevertheless, preliminary analyses indicated that the normality assumption was not satisfied at both stages. Because Heckman’s two-stage estimation approach is sensitive to the model misspecification, the estimates obtained from this approach using the FDCS data would be inconsistent. Thus, Heckman’s two-stage estimation approach is not used in the final analyses. In order to use the instrumental variable approach to adjust for treatment selection bias, the first step is to decide which variable or set of variables can be used as valid instruments. As described earlier, two important criteria of being a valid instrument are the association between the instrument and the treatment as well as the lack of association between the instrument and the outcome after controlling for the treatment. These two criteria can be empirically tested. Another important condition that a valid instrument has to meet is it cannot be correlated with unobserved confounders. However, the lack of association between instruments and unmeasured covariates can only be conceptually credible rather than being empirically verified (Posner et al., 2001). In this dissertation, a set of variables is identified as “theoretically feasible” to be instrumental variables. These potential instruments include perceived need, approach to dental care, dental insurance, frustration with past dental care, distance to dental clinic, and ability to pay an unexpected $500 dental bill. Previous studies have shown that these variables are

PAGE 73

60 important to explain an individual’s dental car e-seeking or utilization behavior (Gilbert et al., 2002a; Maupomé and MacEntee, 1998). A logical expectation is that once people get into the dental care system, the treatment effect will be the same after controlling for other variables. To exam whether these variables constitute valid instruments, a binominal logistic regression using GEE is conducted to test the associations between these variables and the use of prosthodontic services. A feature of this analysis is that there are four different types of prosthodon tic services of interest (finished fixed prosthodontic treatment, unfinished fixed pr osthodontic treatment, finished removable prosthodontic treatment, and unfinished removable prosthodontic treatment). Thus, a binominal logistic regression analysis using GEE is conducted for each type of prosthodontic service, and, in total, four regression analyses are conducted. Another binominal logistic regression using GEE is conducted to test the lack of associations between those potential instrumental variables and the outcome (satisfaction with chewing ability) after controlling for th e use of prosthodontic services and other relevant factors. Only those variables which are theoretically feasible, significantly associated with the respective prosthodontic service use, and not significantly associated with satisfaction with chewing ability, are considered as valid instruments. Some variables meet the above criteria for the finished fixed prosthodontic treatment (FINFIXED) and the finished removable prosthodontic treatment (FINDENTUR) and thus are considered as valid instruments for these two types of prosthodontic services. These valid instruments are reported in Chapter 5. However, no valid instrument is found for the unfinished fixed prosthodontic treatment (BOTHFIXED) and the unfinished removable prosthodontic treatment (BOTHDENTUR).

PAGE 74

61 In the first stage, covariates plus the valid instruments are used to predict treatment status. Because certain valid instruments are found for the finished fixed prosthodontic treatment (FINFIXED) and the finish ed removable prosthodontic treatment (FINDENTUR), the predicted probabilities of having received finished fixed prosthodontic treatment (FINFIXEDHAT) and the predicated probability of having received finished removable prosthodontic treatment (FINDENTURHAT) are calculated from two individual first-stage models. Since both the finished fixed prosthodontic treatment and the finished removable prosthodontic treatment are binary variables, a binominal logistic regression fitted with GEE is used to calculate the predicated probability for each variable. As mentioned earlier, one problem of the instrumental variable approach is that when the instrument variables are not highly predictive of the treatment, this usually leads to a reduction in precision (Posner et al., 2001). Therefore, goodness of fit tests are conducted to ensure that the first-stage models predict the finished fixed prosthodontic treatment and the finished removable prosthodontic treatment well. Since there is no readily defined analogs to the fit statistics in the GEE method (Stokes et al., 2001), two alternative methods are used to approximately test the goodness of fit. First, a “pseudo” goodness of fit test is conducted using the PROC LOGIST procedure by treating correlated observations as independent. Second, DEL is calculated by comparing the predicated probability of treatment and the observed treatment. Although no instrumental variables are found for the unfinished fixed prosthodontic treatment and the unfinished removable prosthodontic treatment, the goodness of fit tests are still conducted for these two treatment variables. The “pseudo” goodness of fit tests indicate that the respective

PAGE 75

62 first-stage model is fit for the finished fixed prosthodontic treatment (p = 1.000) and the finished removable prosthodontic treatment (p = 0.937), but is not fit for the unfinished fixed prosthodontic treatment (p = 0.0038) and the unfinished removable prosthodontic treatment (p < .001). Table 4-3 to Table 46 show individual DEL for each of the firststage models. Table 4-3. Observed Versus Pr edicated Finished Fixed Pros thodontic Treatment Obtained from the First-Stage Model Predicated treatment Observed treatment No (0) Yes (1) Total No (0) 2637 15 2652 Yes (1) 96 31 127 Total 2733 46 2779 Rule K errors = 15 + 96 = 111 P (Y=0) = 2652 / 2779 = 0.95 P (Y=1) = 127 / 2779 = 0.05 Rule U errors = 46 * 0.95 + 2733 * 0.05 = 43 + 137 = 180 DEL = 1 (111 / 180) = 1 0.61 = 0.39 Table 4-4. Observed Versus Predicated Fi nished Removable Prosthodontic Treatment Obtained from the First-Stage Model Predicated treatment Observed treatment No (0) Yes (1) Total No (0) 2657 24 2681 Yes (1) 72 26 98 Total 2729 50 2779 Rule K errors = 24 + 72 = 96 P (Y=0) = 2681 / 2779 = 0.96 P (Y=1) = 98 / 2779 = 0.04 Rule U errors = 50 * 0.96 + 2729 * 0.04 = 48 + 109 = 157 DEL = 1 (96 / 157) = 1 0.61 = 0.39 Table 4-5. Observed Versus Predicated Unfinished Fixed Prosthodontic Treatment Obtained from the First-Stage Model Predicated treatment Observed treatment No (0) Yes (1) Total No (0) 2737 5 2742 Yes (1) 34 3 37 Total 2771 8 2779

PAGE 76

63 Rule K errors = 5 + 34 = 39 P (Y=0) = 2742 / 2779 = 0.99 P (Y=1) = 37 / 2779 = 0.01 Rule U errors = 8 * 0.99 + 2771 * 0.01 = 8 + 28 = 36 DEL = 1 (39 / 36) = 1 1.08 = -0.08 Table 4-6. Observed Versus Predicated Un finished Removable Prosthodontic Treatment Obtained from the First-Stage Model Predicated treatment Observed treatment No (0) Yes (1) Total No (0) 2748 2 2750 Yes (1) 29 0 29 Total 2777 2 2779 Rule K errors = 2 + 29 = 31 P (Y=0) = 2750 / 2779 = 0.99 P (Y=1) = 29 / 2779 = 0.01 Rule U errors = 2 * 0.99 + 2777 * 0.01 = 2 + 28 = 30 DEL = 1 (31 / 30) = 1 1.03 = -0.03 DEL calculated from the first-stage model for the finished fixed prosthodontic treatment and DEL calculated from the first-stage model for the finished removable prosthodontic treatment are in the acceptable range. DEL calculated from the first-stage model for the unfinished fixed prosthodontic tr eatment and DEL calculated from the firststage-model for the unfinished removable prosthodontic treatment are negative, which indicate that these two models are not predictive of these two treatments. Since there is no valid instrumental variables for these two treatments, and the first-stage models are not fit to predict such treatments, the unfinished fixed prosthodontic treatment and the unfinished removable prosthodontic treatment are excluded from the second-stage model. The predicted probability of finished fixed prosthodontic treatment (FINFIXEDHAT) and the predicated probability of finished removable prosthodontic treatment (FINDENTURHAT) are then used in lieu of treatment status as independent variables in the second stage to predict the outcome (satisfaction with chewing ability),

PAGE 77

64 along with other measured covariates. However, the variables, which are used as instruments in the first stage, are excluded from the second stage since they are assumed not to be directly associated with the outcome. Because the outcome variable for the second-stage model (satisfaction with chewing ability) is recoded as a dichotomous variable, a binominal logistic regression fitted with GEE is used. In summary, the firststage and the second-stage models can be written as Equations 4-1 to 4-3. The first-stage models: g[Pr(FINFIXED)t] = 1 + 1X1t + 1Z1t + ut (4-1) g[Pr(FINDENTUR)t] = 2 + 2X2t + 2Z2t + vt (4-2) The second-stage model: g[Pr(Yt)] = 3 + 3X3t + 3(FINFIXEDHAT)t + 4(FINDENTURHAT)t + t (4-3) Where g(·) is an appropriate link function which reveals the relationship between the dependent variable and the set of explanatory variables. In Equation 4-1 to Equation 4-3, the links are the binominal logistic regression fitted with GEE. (FINFIXED)t is the finished fixed prosthodontic treatme nt at time “t,” and (FINDENTUR)t is the finished removable prosthodontic treatment at time “t.” (FINFIXEDHAT)t is the predicated probability of finished fixed prosthodontic treatment at time “t,” and (FINDENTURHAT)t is the predicated probability of finished removable prosthodontic treatment at time “t.” Yt is the outcome of satisfaction with chewing ability at time “t.” X1t is a vector of covariates that are though to be related to the finished fixed prosthodontic treatment at time “t.” X2t is a vector of covariates that are though to be related to the finished removable prosthodontic treatment at time “t.” X3t is a vector of covariates that are though to be related to the outcome at time “t.” Zit is a vector of the

PAGE 78

65 instrumental variables for the finished fixed prosthodontic treatment at time “t.” Z2t is a vector of the instrumental variables for the finished removable prosthodontic treatment at time “t.” Longitudinal logistic regression with correction of selection bias for research question 2d All analytical methods are very similar to the methods used to answer research question 2c except (1) the outcome variables are replaced with the dynamic changes in satisfaction with chewing ability: improvement of satisfaction with chewing ability and deterioration of satisfaction with chewing ability, and (2) satisfaction with chewing ability reported at the preceding interview point (satisfaction with chewing ability at TX-1 interview point) is included as a covariate in the model. Since both outcome variables are binary outcome variables, a binominal logistic regression fitted with GEE is used at the second-stage model for each outcome variable. Analytical Methods for Objective 3 and Objective 4 All analytical methods are identical to the methods used for objective 2. But the outcome variables for objective 3 are replaced with satisfaction with dental appearance, improvement of satisfaction with dental appear ance, and deterioration of satisfaction with dental appearance. Similarly, the outcome variables for objective 4 are replaced with self-rated overall oral health, improvement of self-rated overall oral health, and deterioration of self-rated overall oral health. Sensitivity Tests There is no established method reported in the literature that can simultaneously account for selection bias for multiple treatments. For the purpose of this dissertation, selection bias is corrected for both the fi nished fixed prosthodontic treatment and the

PAGE 79

66 finished removable prosthodontic treatment. In order to compare the results from the different possible approaches, sensitivity tests are conducted for each outcome variable: (1) correcting selection bias only for the finished fixed prosthodontic treatment; (2) correcting selection bias only for the finished removable prosthodontic treatment; and (3) not correcting selection bias at all. Appendix B lists the SAS code for all analyses (including sensitivity tests) using the instrumental variable approach. The results of the sensitivity tests are reported in Appendix C.

PAGE 80

67 CHAPTER 5 RESULTS Results for Objective 1 Incidence of Prosthodontic Services Use during 24 Months Table 5-1 presents the use of prosthodontic services during 24 months. Fixed prosthodontics (crowns and bridges) were the most common form of prosthodontic treatment received. Approximately 5% to 6% of study participants received fixed prosthodontic treatment in each interval, and such treatment was finished within that interval. Approximately 1% to 2% of study participants received fixed prosthodontic treatment in each interval, but the treatment was not finished within the current interval and continued to the next interval. Fewer subjects received removable prosthodontic treatment than fixed prosthodontic treatment . Approximately 2% to 3% of subjects received removable prosthodontic treatment which was finished within the current interval. However, only 0.4% to 0.8% of subjects received unfinished removable prosthodontic treatment which had to continue to the next interval. Prevalence of Self-Rated Oral Health Outcomes Table 5-2 presents the prevalence of three original measures of self-rated oral health during 24 months: satisfaction with chewing ability, satisfaction with dental appearance, and self-rated overall oral health. Approximately 16% of subjects reported they were “dissatisfied” or “very dissatisfied” with their chewing ability at baseline, and approximately 13% to 17% reported so at the follow-up interviews. A higher percentage of subjects reported “dissatisfied” or “very dissatisfied” with their dental appearance than

PAGE 81

68 with their chewing ability at each interview point. Approximately 24% of subjects reported they were “dissatisfied” or “very dissatisfied” with their dental appearance at baseline, and approximately 19% to 22% reported so at the follow-up interview points. The prevalence of self-rated overall oral health is slightly lower than that of satisfaction with dental appearance. At baseline, approximately 21% of subjects rated their overall oral health as “fair” or “poor,” and approximately 16% to 17% rated so at the follow-up interviews. Patterns of Changes in Self-Rated Oral Health Table 5-3 shows the incidence of dynamic changes in self-rated oral health during 24 months: improvement of satisfaction with chewing ability, deterioration of satisfaction with chewing ability, improvement of satisfaction with dental appearance, deterioration of satisfaction with dental appearance, impr ovement of self-rated overall oral health, and deterioration of self-rated overall oral health. The incidence rates of improvement of satisfaction with chewing ability ranged from 11.3% to 21.5%, depending on the interval. The incidence rates of deterioration of satisfaction with chewing ability ranged from 11.8% to 18.4%, depending on the interval. The incidence rates of improvement of satisfaction with dental appearance and deterioration of satisfaction with dental appearance showed the similar trend: 15.9% to 21.1% of study participants experienced improved satisfaction with dental appearance by the end of the interval, while 15.9% to 19.2% of subjects experienced deteriorated satisfaction with dental appearance. The in cidence of improvement of self-rated overall oral health in each interval varies from 19.7% to 30.9%, and the incidence of deterioration of self-rated overall oral health in each interval varies from 19.6% to 26.7%.

PAGE 82

69 Table 5-1. Incidence of Prosthodontic Services Use during 24 Months Interval 1 a (n = 856) Interval 2 b (n = 829) Interval 3 c (n = 817) Interval 4 d (n = 788) Prosthodontic services use Weighted n (%) Fixed prosthodontic treatment finished within the current interval 1 = yes 0 = no 52 (6.2) 778 (93.8) 50 (6.1) 767 (93.9) 38 (4.9) 746 (95.1) 44 (5.7) 727 (94.3) Fixed prosthodontic treatment continued to the following interval 1 = yes 0 = no 18 (2.1) 812 (97.9) 11 (1.3) 807 (98.7) 14 (1.7) 771 (98.3) 7 (0.9) 764 (99.1) Removable prosthodontic treatment finished within the current interval 1 = yes 0 = no 15 (1.8) 814 (98.2) 22 (2.7) 795 (97.3) 27 (3.5) 757 (96.5) 23 (3.0) 749 (97.0) Removable prosthodontic treatment continued to the following interval 1 = yes 0 = no 6 (0.7) 823 (99.3) 6 (0.8) 811 (99.2) 6 (0.7) 779 (99.3) 3 (0.4) 768 (99.6) a Interval 1 is the 6 months interval betw een baseline interview and 6-month interview b Interval 2 is the 6 months interval be tween 6-month interview and 12-month interview c Interval 3 is the 6 months interval be tween 12-month interview and 18-month interview d Interval 4 is the 6 months interval be tween 18-month interview and 24-month interview

PAGE 83

70 Table 5-2. Prevalence of Self-Rated Oral Health Outcomes during 24 Months Baseline interview (n = 873) 6-month interview (n = 856) 12-month interview (n = 829) 18-month interview (n = 817) 24-month interview (n = 788) Measures of self-rated oral health Weighted n (%) Satisfaction with chewing ability 1 = very satisfied 2 = satisfied 3 = dissatisfied 4 = very dissatisfied 397 (45.6) 336 (38.6) 100 (11.4) 38 (4.4) 441 (51.7) 298 (34.9) 93 (10.9) 21 (2.5) 394 (47.6) 296 (35.8) 116 (14.0) 22 (2.7) 342 (42.0) 357 (43.8) 79 (9.7) 37 (4.5) 361 (45.8) 315 (40.0) 85 (10.8) 26 (3.3) Satisfaction with dental appearance 1 = very satisfied 2 = satisfied 3 = dissatisfied 4 = very dissatisfied 208 (24.0) 451 (52.1) 158 (18.2) 50 (5.7) 207 (24.3) 457 (53.4) 155 (18.2) 35 (4.1) 198 (24.0) 444 (53.8) 155 (18.9) 28 (3.4) 182 (22.4) 471 (57.8) 129 (15.8) 33 (4.0) 186 (23.6) 454 (57.8) 114 (14.5) 33 (4.2) Self-rated overall oral health 1 = excellent 2 = very good 3 = good 4 = fair 5 = poor 172 (19.8) 265 (30.5) 252 (29.0) 109 (12.6) 70 (8.1) 221 (26.0) 250 (29.4) 236 (27.7) 102 (11.9) 42 (5.0) 188 (22.8) 271 (32.8) 224 (27.2) 99 (12.1) 42 (5.1) 179 (22.0) 286 (35.2) 220 (27.0) 89 (11.0) 39 (4.8) 178 (22.9) 230 (29.6) 249 (32.0) 78 (10.1) 42 (5.4)

PAGE 84

71 Table 5-3. Dynamic Changes in Self-Rated Oral Health during 24 Months 6-month interview (n = 856) 12-month interview (n = 829) 18-month interview (n = 817) 24-month interview (n = 788) Changes of self-rated oral health Weighted n (%) Improvement of satisfaction with chewing ability 1 = yes 0 = no 183 (21.5) 668 (78.5) 93 (11.3) 733 (88.7) 101 (12.4) 713 (87.6) 117 (15.0) 666 (85.0) Deterioration of satisfaction with chewing ability 1 = yes 0 = no 120 (14.1) 731 (85.9) 150 (18.3) 675 (81.7) 150 (18.4) 664 (81.6) 92 (11.8) 690 (88.2) Improvement of satisfaction with dental appearance 1 = yes 0 = no 179 (21.1) 668 (78.9) 142 (17.2) 683 (82.8) 129 (15.9) 683 (84.1) 148 (19.0) 631 (81.0) Deterioration of satisfaction with dental appearance 1 = yes 0 = no 163 (19.2) 685 (80.8) 131 (15.9) 694 (84.1) 138 (17.0) 673 (83.0) 146 (18.7) 633 (81.3) Improvement of self-rated overall oral health 1 = yes 0 = no 262 (30.9) 586 (69.1) 166 (20.2) 656 (79.8) 184 (22.7) 625 (77.3) 152 (19.7) 620 (80.3) Deterioration of self-rated overall oral health 1 = yes 0 = no 166 (19.6) 681 (80.4) 179 (21.8) 644 (78.2) 176 (21.8) 632 (78.2) 206 (26.7) 566 (73.3)

PAGE 85

72 Results for Objective 2 Cross-Sectional Bivariate Associations at the Baseline Interview (Research Question 2a) The bivariate associations between satisfaction with chewing ability at baseline and the baseline use of prosthodontic services as well as selected PEN factors are shown in Table 5-4. Whether or not wearing a full dent ure or a partial denture at baseline was not significantly associated with satisfaction with chewing ability. Among those predisposing factors, age, gender, and area of residence were not significantly associated with satisfaction with chewing ability. White s were more satisfied with their chewing ability than blacks, and people who had finished at least high school were more satisfied with their chewing ability than those who had not. People who reported their general health as “good,” who were regular dental serv ices attenders, and who were not frustrated with past dental care were more satisfied w ith their chewing ability than their respective counterparts. Satisfaction with chewing ability was significantly associated with all three enabling factors. Subjects whose income was not below 150% poverty level, who were able to pay an unexpected $500 dental bill either comfortably or with difficulty, and who had a certain type of dental insurance were more satisfied with their chewing ability than those whose income was below 150% poverty level, who were unable to pay an unexpected $500 dental bill, and who had no dental insurance. Satisfaction with chewing ability was more common in subjects who had at least 10 occluding pairs of teeth or more. Among those self-reported measures of oral disease/tissue damage, all measures with the exception of a broken denture and dry mouth were significantly associated with satisfaction with chewing ability. Subjects who

PAGE 86

73 reported that they had a sore denture, a loose cap or bridge, a broken tooth or cap, a broken filling, a toothache, a sensitive tooth, a cavity, an abscessed tooth, infected or sore gums, bleeding gums, a loose tooth, stained t eeth, and a problem with bad breath were less satisfied with their chewing ability than those who did not report these problems. Cross-Sectional Multivariate Associatio ns at the Baseline Interview (Research Question 2b) An ordinal logistic regression using a proportional odds model is conducted to identify the associations between baseline satisfaction with chewing ability and a set of independent measures, which include baseline prosthodontic use, predisposing, enabling, and need factors. The proportional odds assumption is tested and satisfied in this model. Table 5-5 lists the results from the ordinal logistic regression of baseline satisfaction with chewing ability. In the SAS program, the default setup of the ordinal logistic regression model is that the explanatory variables predict the probability of being in a lower category of the dependent variable rather than in a higher category (Allison, 1999). Since the lower categories represent being more satisfied with chewing ability (“1 = very satisfied,” “2 = satisfied,” “3 = dissatisfied,” and “4 = very dissatisfied”), the odds ratios in this model are interpreted as an increase in the odds of moving down to a lower category, which means moving up one level of sa tisfaction (e.g. from “2 = satisfied” to “1 = very satisfied,” or from “3 = dissatisfied” to “2 = satisfied”) with an increase in one level of the explanatory covariates. After controlling for the number of pairs of teeth and other PEN factors, wearing a partial denture was significantly associated with a greater satisfaction with chewing ability (OR = 2.055, 95% CI = [1.058 3.993], p = .03). Compared to being a problem oriented dental attender, being a regular dental attender increased the odds of reporting a lower level of category (a higher level of satisfaction

PAGE 87

74 with chewing ability) by approximately 1.8 times (OR = 2.808). After the other conditions were taken into account, reporting a sore denture, a toothache and/or abscessed tooth, a loose tooth, and dry mout h were significantly associated with less satisfaction with chewing ability. Neither ability to pay a $500 dental bill nor dental insurance status was significantly associated with satisfaction with chewing ability. Table 5-4. Satisfaction with Chewing Ability at Baseline for the Sample Overall, by Baseline Prosthodontic Services Use a nd by Predisposing, Enabling, and Need Factors (PEN Factors) PEN factors (Weighted n) % Subjects who were satisfied or very satisfied with chewing ability % Subjects who were dissatisfied or very dissatisfied with chewing ability p-value Prosthodontic Services Has and wears a full denture 1 = yes (81) 0 = no (791) 79.3 84.7 20.7 15.3 0.2094 Has and wears a partial denture 1 = yes (169) 0 = no (703) 82.2 84.6 17.8 15.4 0.4413 Predisposing Factors Age 1 = 65 years old or older (361) 0 = 45 to 64 years old (511) 84.8 83.7 15.2 16.3 0.6773 Gender 1 = male (383) 2 = female (489) 84.3 84.1 15.7 15.9 0.9255 Area of residence 1 = urban (436) 0 = rural (435) 84.9 83.5 15.1 16.5 0.5744 Race 1 = White (628) 2 = Black (241) 88.0 75.0 12.0 25.0 <.0001 Level of formal education 1 = completed at least high school (689) 0 = did not complete high school (183) 87.7 71.0 12.3 29.0 <.0001 Self-rated general health 1 = good (641) 2 = not good (224) 87.8 73.6 12.2 26.4 <.0001 Approach to dental care 1 = regular attender (473) 0 = problem attender (399) 93.7 72.8 6.3 27.2 <.0001

PAGE 88

75 Table 5-4. Continued PEN factors (Weighted n) % Subjects who were satisfied or very satisfied with chewing ability % Subjects who were dissatisfied or very dissatisfied with chewing ability p-value Frustration with past dental care 1 = frustrated (133) 0 = not frustrated (737) 75.8 85.6 24.2 14.4 0.0042 Enabling Factors Poverty Level 1 = below150% poverty level (272) 0 = not below 150% poverty level (521) 75.4 89.5 24.6 10.5 <.0001 Able to pay an unexpected $500 dental bill 1 = able to pay comfortably (406) 2 = able to pay but with difficulty (342) 3 = not able to pay (122) 93.5 81.5 60.4 6.5 18.5 39.6 <.0001 Dental insurance status 1 = has certain type of dental insurance (293) 2 = no dental insurance (579) 89.4 81.5 10.7 18.5 0.0029 Need Factors Number of occluding pairs of teeth 1 = 10 occluding pairs of teeth or more (512) 0 = fewer than 10 occluding pairs of teeth (356) 93.6 70.5 6.4 29.6 <.0001 Has a sore denture 1 = yes (57) 2 = no (232) 55.3 85.3 44.7 14.7 <.0001 Has a broken denture 1 = yes (38) 2 = no (251) 74.6 80.2 25.4 19.8 0.4275 Has a loose cap or bridge 1 = yes (10) 2 = no (861) 48.6 84.6 51.4 15.4 0.0016 Has a broken tooth or cap 1 = yes (178) 2 = no (684) 66.3 88.7 33.7 11.3 <.0001 Has a broken filling 1 = yes (130) 2 = no (722) 71.2 86.4 28.8 13.7 <.0001 Has a toothache 1 = yes (100) 2 = no (771) 61.7 87.1 38.3 12.9 <.0001

PAGE 89

76 Table 5-4. Continued PEN factors (Weighted n) % Subjects who were satisfied or very satisfied with chewing ability % Subjects who were dissatisfied or very dissatisfied with chewing ability p-value Has a sensitive tooth 1 = yes (261) 2 = no (610) 74.3 88.5 25.7 11.6 <.0001 Has a cavity 1 = yes (168) 2 = no (640) 64.3 90.9 35.7 9.1 <.0001 Has an abscessed tooth 1 = yes (22) 2 = no (835) 39.6 85.9 60.4 14.1 <.0001 Has infected or sore gums 1 = yes (102) 2 = no (765) 64.4 87.0 35.6 13.0 <.0001 Has bleeding gums 1 = yes (117) 2 = no (754) 67.2 86.9 32.8 13.1 <.0001 Has a loose tooth 1 = yes (115) 2 = no (747) 62.4 87.6 37.6 12.4 <.0001 Has stained teeth 1 = yes (341) 2 = no (511) 72.9 92.4 27.1 7.6 <.0001 Has a problem with bad breath 1 = yes (154) 2 = no (674) 70.6 87.7 29.4 12.3 <.0001 Has a dry mouth 1 = yes (190) 2 = no (678) 79.9 85.5 20.1 14.5 0.0605

PAGE 90

77 Table 5-5. Cross-Sectional Logistic Regression of Satisfaction with Chewing Ability at Baseline PEN factors (Weighted n) Estimate Odds ratio (95% Confidence interval) p-value Prosthodontic Services Has and wears a full denture 1 = yes 0 = no 0.5685 -1.766 (0.851 3.664) -0.1270 Has and wears a partial denture 1 = yes 0 = no 0.7202 -2.055 (1.058 3.993) -0.0336 Predisposing Factors Age 1 = 65 years old or older 0 = 45 to 64 years old -0.1125 -0.894 (0.492 1.624) -0.7120 Gender 1 = male 2 = female -0.2505 -0.606 (0.336 1.094) -0.0964 Race 1 = White 2 = Black 0.2781 -1.744 (0.902 3.372) -0.0982 Approach to dental care 1 = regular attender 0 = problem attender 1.0326 -2.808 (1.410 5.595) -0.0033 Frustration with past care 1 = frustrated 0 = not frustrated -0.5827 -0.558 (0.247 1.261) -0.1609 Enabling Factors Able to pay an unexpected $500 dental bill 1 = able to pay comfortably 2 = able to pay but with difficulty 3 = not able to pay 0.2538 -0.0164 -1.634 (0.630 4.242) 1.247 (0.548 2.837) -0.3186 0.9380 Dental insurance status 1 = has certain type of dental insurance 2 = no dental insurance 0.2367 -1.605 (0.816 3.158) -0.1703 Need Factors Number of occluding pairs of teeth 1 = 10 occluding pairs of teeth or more 0 = fewer than 10 occluding pairs of teeth 0.9439 -2.570 (0.945 6.992) -0.0646 Has a sore denture 1 = yes 2 = no -0.7637 -0.217 (0.103 0.458) -<.0001 Has a broken denture 1 = yes 2 = no -0.2188 -0.646 (0.270 1.544) -0.3252 Has a loose cap or bridge 1 = yes 2 = no 0.0864 -1.189 (0.018 78.092) -0.9355

PAGE 91

78 Table 5-5. Continued PEN factors (Weighted n) Estimate Odds ratio (95% Confidence interval) p-value Has a broken tooth or cap 1 = yes 2 = no 0.1471 -1.342 (0.553 3.255) -0.5152 Has a broken filling 1 = yes 2 = no -0.3459 -0.501 (0.180 1.396) -0.1860 Has a toothache and /or abscessed tooth 1 = yes 2 = no -0.5250 -0.350 (0.150 0.819) -0.0155 Has a sensitive tooth 1 = yes 2 = no -0.2770 -0.575 (0.290 1.140) -0.1131 Has a cavity 1 = yes 2 = no -0.2532 -0.603 (0.268 1.356) -0.2212 Has an abscessed tooth 1 = yes 2 = no -0.7431 -0.226 (0.032 1.587) -0.1348 Has infected or sore gums 1 = yes 2 = no 0.1661 -1.394 (0.525 3.699) -0.5048 Has bleeding gums 1 = yes 2 = no -0.0086 -0.983 (0.377 2.561) -0.9718 Has a loose tooth 1 = yes 2 = no -0.5652 -0.323 (0.155 0.674) -0.0026 Has stained teeth 1 = yes 2 = no -0.1138 -0.796 (0.419 1.512) -0.4866 Has a problem with bad breath 1 = yes 2 = no -0.0671 -0.874 (0.408 1.873) -0.7298 Has a dry mouth 1 = yes 2 = no -0.4219 -0.430 (0.210 0.880) -0.0208 Longitudinal Logistic Regression of Satisfaction with Chewing Ability with Correction of Selection Bias (Research Question 2c) A longitudinal logistic regression with corr ection of selection bias is conducted to predict the effect of prosthodontic services use on satisfaction with chewing ability at each follow-up interview point during a 24-month period.

PAGE 92

79 In this dissertation, several variables such as perceived need, approach to dental care, dental insurance status, frustration with past dental care, distance to dental clinic, and ability to pay an unexpected $500 dental bill, etc., are considered as potential instruments because of their conceptual credib ility. The relationships between this set of variables and the finished fixed prosthodontic treatment, as well as the finished removable prosthodontic treatment, are tested through a binominal logistic regression fitted with GEE for each treatment variable. Four variables are significantly associated with the finished fixed prosthodontic treatment: perceived need for dental care (p = 0.0210), whether the dentist or clinic that a subject usually sees is one that is nearest (p = 0.0032), approach to dental care (p = 0.0076), and ability to pay an unexpected $500 dental bill comfortably (p = 0.0089) and with difficulty (p = 0.0481). One variable is significantly associated with the finished removable prosthodontic treatment: frustration with past dental care (p = 0.0178). These five variables together with those two finished prosthodontic treatment variables (fixed and removable) are entered in a set of binominal logistic regressions fitted with GEE to test the lack of association between these variables and each outcome variable. Only those variables which are not associated with outcomes after the treatment variables are controlled for (p>0.05) are considered as valid instruments and used in the first-stage model of instrumental variable approach. Table 5-6 lists the valid instruments used in the first-stage model for research question 2c. Results of the first-stage models for research question 2c In the SAS GENMOD procedure, the default setup of the binomial logistic regression model is the explanatory variables predict the probability of being in the lower level of the response variable rather than in the higher level (Allison, 1999). In the firststage analyses, the DESCENDING option is used to reverse the default order. The

PAGE 93

80 models in the first-stage analyses predict the probability of being in the higher category of the dependent variables (“finished fixed prosthodontic treatment” and “finished removable prosthodontic treatment”), which means the probability of “1 = yes, had received finished fixed prosthodontic treatment” and the probability of “1 = yes, had received finished removable prosthodontic treatment”). Table 5-7 and Table 5-8 show the results of the first-stage models fitted with GEE for research question 2c. The main purpose of the first-stage models is to calculate the predicted probabilities in each interval for each participant of having received finished fixed prosthodontic treatment or having r eceived finished removable prosthodontic treatment. The predicted probabilities of having received these two types of treatments are then carried over to a second stage model for each outcome variable to adjust for the treatment selection bias. The first-stage analyses also have identified the independent determinants of having received finished fixed prosthodon tic treatment or finished removable prosthodontic treatment. As shown in Table 5-7, whites were significantly more likely to have received finished fixed prosthodontic treatment than blacks (OR = 3.930, 95% CI = [2.072 7.452], p < .0001). Regular dental attenders were more likely than problemoriented attenders to have received fini shed fixed prosthodontic treatment (OR = 3.249, 95% CI = [1.600 6.596], p = 0.0011). However, persons whose dentist was the nearest one were less likely to have received such treatment. Persons who had certain types of dental insurance and who had problems with a current prosthesis were more likely to have received the treatment. People who did not have a perceived need for dental care and who had lost teeth in the past six months were less likely to have received the

PAGE 94

81 finished fixed prosthodontic treatment, wh ile people who had experienced a broken filling, tooth fracture, cavities, and a loose toot h in the past six months were more likely to have received such treatment. As shown in Table 5-8, people who had problems with a current prosthesis and who had lost teeth in the past six months were more likely to have received finished removable prosthodontic treatment. Table 5-6. Instrumental Variables for Research Question 2c Outcome variable Treatment variable Valid instrumental variables p-value a p-value b Satisfaction with chewing ability Finished fixed prosthodontic treatment Perceived need for dental care 1 = not at this time 0 = other 0.0210 0.6140 Whether the dentist or clinic that a subject usually sees is one that is nearest 1 = yes 2 = no 0.0032 0.9784 Able to pay an unexpected $500 dental bill 2 = able to pay, but with difficulty 3 = not able to pay 0.0481 0.0836 Finished removable prosthodontic treatment Frustration with past dental care 1 = yes 0 = no 0.0178 0.0728 a Associations between the instrument al variables and the treatment variables b Lack of a direct association between the in strumental variables and the outcome variable Table 5-7. The First-Stage Model to Calculat e Predicated Probabilities of Finished Fixed Prosthodontic Treatment Parameter Estimate Odds ratio (95% Confidence interval) p-value Predisposing Factors Age 1 = 45 to 64 years old 0 = 65 years old or older 0.0886 -1.093 (0.707 1.688) -0.6900 Gender 1 = male 2 = female -0.1932 -0.824 (0.526 1.292) -0.3994 Race 1 = White 2 = Black 1.3686 -3.930 (2.072 7.452) -<.0001 Approach to dental care 1 = regular attender 0 = problem attender 1.1783 -3.249 (1.600 6.596) -0.0011

PAGE 95

82 Table 5-7. Continued Parameter Estimate Odds ratio (95% Confidence interval) p-value Frustration with past care 1 = frustrated 0 = not frustrated 0.4713 -1.602 (0.892 2.877) -0.1146 Whether the dentist that the patient frequently goes to see is the one that is nearest 1 = yes 2 = no -1.1558 -0.315 (0.190 0.521) -<.0001 Enabling Factors Able to pay an unexpected $500 dental bill 1 = able to pay comfortably 2 = able to pay but with difficulty 3 = not able to pay 0.8747 0.4299 -2.398 (0.927 6.206) 1.537 (0.619 3.815) -0.0714 0.3540 Insurance status 1 = has insurance 2 = no insurance 0.7334 -2.082 (1.340 3.236) -0.0011 Need Factors Perceived need for dental care 1 = not at this time 0 = other -2.3103 -0.099 (0.024 0.405) -0.0013 Problems with current prosthesis 1 = yes 0 = no 2.8735 -17.699 (9.408 33.295) -<.0001 Tooth loss in the past six months 1 = yes 2 = no -0.9128 -0.401 (0.165 0.978) -0.0447 Broken filling 1 = yes 0 = no 0.8677 -2.381 (1.173 4.834) -0.0163 Tooth fracture 1 = yes 0 = no 2.2092 -9.108 (5.318 15.599) -<.0001 Cavities 1 = yes 0 = no 0.8950 -2.447 (1.194 5.016) -0.0145 Loose tooth 1 = yes 0 = no 0.9489 -2.583 (1.567 4.257) -0.0002 Teeth stained or look bad 1 = yes 0 = no -0.7074 -0.493 (0.217 1.120) -0.0912

PAGE 96

83 Table 5-8. The First-Stage Model to Calculate Predicated Probabilities of Finished Removable Prosthodontic Treatment Parameter Estimate Odds ratio (95% Confidence interval) p-value Predisposing Factors Age 1 = 45 to 64 years old 0 = 65 years old or older 0.2345 -1.264 (0.660 2.422) -0.4795 Gender 1 = male 2 = female 0.3871 -1.473 (0.775 2.799) -0.2375 Race 1 = White 2 = Black -0.5338 -0.586 (0.298 1.154) -0.1223 Approach to dental care 1 = regular attender 0 = problem attender 0.1439 -1.155 (0.575 2.320) -0.6860 Frustration with past care 1 = frustrated 0 = not frustrated 0.1330 -1.142 (0.447 2.917) -0.7809 Whether the dentist that the patient frequently goes to see is the one that is nearest 1 = yes 2 = no -0.0229 -0.977 (0.479 1.994) -0.9497 Enabling Factors Able to pay an unexpected $500 dental bill 1 = able to pay comfortably 2 = able to pay but with difficulty 3 = not able to pay -0.5428 0.8153 -0.581 (0.151 2.234) 2.260 (0.814 6.271) -0.4295 0.1174 Insurance status 1 = has insurance 2 = no insurance 0.3496 -1.419 (0.700 2.873) -0.3317 Need Factors Perceived need for dental care 1 = not at this time 0 = other -0.2107 -0.810 (0.199 3.298) -0.7687 Problems with current prosthesis 1 = yes 0 = no 4.5262 -92.407 (36.551 233.614) -<.0001 Tooth loss in the past six months 1 = yes 2 = no 2.9728 -19.547 (7.236 52.800) -<.0001 Broken filling 1 = yes 0 = no -0.4294 -0.651 (0.248 1.709) -0.3832 Tooth fracture 1 = yes 0 = no -0.5587 -0.572 (0.255 1.281) -0.1745

PAGE 97

84 Table 5-8. Continued Parameter Estimate Odds ratio (95% Confidence interval) p-value Cavities 1 = yes 0 = no 0.2230 -1.250 (0.638 2.449) -0.5161 Loose tooth 1 = yes 0 = no -0.1017 -0.903 (0.376 2.173) -0.8203 Teeth stained or look bad 1 = yes 0 = no -0.4347 -0.6478 (0.297 1.411) -0.2739 Results of the second-stage model for research question 2c Table 5-9 lists the results of the second-stage models fitted with GEE for the dichotomized satisfaction with chewing ability over a 24-month period. In the secondstage model, satisfaction with chewing ability is a function of the covariates for instrumental variable adjustment (“predicte d probability of having r eceived finished fixed prosthodontic treatment” and “predicated probability of having received finished removable prosthodontic treatment” as calculated from the first-stage models) and other relevant covariates. The instrumental variables listed in Table 5-6 are excluded from the second-stage models since their effects on having received finished fixed prosthodontic treatment and having received finished rem ovable prosthodontic treatment have been accounted for through the first-stage model, and they do not have direct association with satisfaction with chewing ability after controlling for the two types of treatments. Similar to what has been mentioned in the first-stage models, DESCENDING option is used in the second-stage model. Thus, the second-stage model predicts the probability of being in the higher category of the satisfaction with chewing ability. Since the measurement of satisfaction with chewing ability in this analysis has been

PAGE 98

85 dichotomized, the second-stage model predicts the probability of “2 = dissatisfied/ very dissatisfied.” None of the two types of prosthodontic tr eatment was significantly associated with satisfaction with chewing ability. Compared to blacks, whites were less likely to be dissatisfied with their chewing ability (OR = 0.506, 95% CI = [0.344 0.745], p = 0.0005). Persons who had at least 10 occluding pairs of teeth were less likely than those who had fewer than 10 occluding pairs of teeth to have reported dissatisfaction with chewing ability (OR = 0.224, 95% CI = [0.142 0.354], p < 0.0001). Table 5-9. Parameter Estimates from the Second-Stage Model for Research Question 2c Parameter Estimate Odds ratio (95% Confidence interval) p-value Predicted probability of finished fixed prosthodontic treatment within the current interval -0.5819 0.559 (0.152 2.057) 0.3814 Predicted probability of finished removable prosthodontic treatment within the current interval -0.0476 0.954 (0.208 4.378) 0.9512 Age 1 = 45 to 64 years old 0 = 65 years old or older -0.0907 -0.913 (0.624 1.338) -0.6415 Gender 1 = male 2 = female 0.1735 -1.189 (0.797 1.776) -0.3964 Race 1 = White 2 = Black -0.6803 -0.506 (0.344 0.745) -0.0005 Tooth loss in the past six months 1 = yes 2 = no 0.2736 -1.315 (0.781 2.213) -0.3035 Number of pairs of teeth 1 = 10 occluding pairs of teeth or more 0 = fewer than 10 occluding pairs of teeth -1.4946 -0.224 (0.142 0.354) -<.0001 Longitudinal Logistic Regression of Chan ges in Satisfaction with Chewing Ability with Correction of Selection Bias (Research Question 2d) Longitudinal logistic regression with correction of selection bias are conducted to predict the effect of prosthodontic services use on changes in satisfaction with chewing

PAGE 99

86 ability, namely, improvement of satisfaction with chewing ability and deterioration of satisfaction with chewing ability, at each follow-up interview point during a 24-month period. Table 5-10 lists the valid instruments used in the first-stage models for research question 2d. Table 5-10. Instrumental Variables for Research Question 2d Outcome variable Treatment variable Valid instrumental variables p-value a p-value b Improvement of satisfaction with chewing ability Finished fixed prosthodontic treatment Perceived need for dental care 1 = not at this time 0 = other 0.0210 0.6081 Whether the dentist or clinic that a subject usually sees is one that is nearest 1 = yes 2 = no 0.0032 0.3398 Finished removable prosthodontic treatment Frustration with past dental care 1 = yes 0 = no 0.0178 0.2233 Deterioration of satisfaction with chewing ability Finished fixed prosthodontic treatment Perceived need for dental care 1 = not at this time 0 = other 0.0210 0.9164 Whether the dentist or clinic that a subject usually sees is one that is nearest 1 = yes 2 = no 0.0032 0.0643 Able to pay an unexpected $500 dental bill 2 = able to pay, but with difficulty 3 = not able to pay 0.0481 0.7251 Finished removable prosthodontic treatment Frustration with past dental care 1 = yes 0 = no 0.0178 0.4678 a Associations between the instrument al variables and the treatment variables b Lack of a direct association between the in strumental variables and the outcome variable Results of the first-stage models for research question 2d The explanatory variables in the first-stage models are the same for all longitudinal analyses in this dissertation. Thus, the results of the first-stage models are the same as what are listed in Table 5-7 and Table 5-8 for each outcome variable of interest. What

PAGE 100

87 should be noticed here is that although perceived need for dental care, whether the dentist or clinic that a subject usually sees is the nearest one, approach to dental care, ability to pay an unexpected $500 dental bill, and frustration with the past dental care are all included in the first-stage models for each outcome variable, only the variables listed in the corresponding tables, which show instrumental variables for an particular outcome variable, are considered as valid instruments for that outcome. Variables other than listed in the corresponding tables are entered in the first-stage models as conventional covariates. In the current analysis, only the variables listed in Table 5-10 are considered as instruments. Results of the second-stage models for research question 2d Table 5-11 lists the results from the second-stage model fitted with GEE for the improvement of satisfaction with chewing ability over a 24-month period. Subjects who had received finished removable prosthodontic treatment were more likely to report improved satisfaction with chewing ability (OR = 1.584, 95% CI = [1.090 2.303], p < 0.0159). Persons who were more satisfied with their chewing ability at the preceding interval were less likely to report improved chewing ability at the current interval (OR = 0.170, 95% CI = [0.122 0.238], p < .0001). Having received finished fixed prosthodontic treatment, age, gender, tooth loss in the past six months, and number of pairs of teeth at baseline were not significantly associated with the improvement of satisfaction with chewing ability. Table 5-12 shows the results from the second-stage model fitted with GEE for the deterioration of satisfaction with chewing ability over a 24-month period. None of the two types of prosthodontic treatment was signi ficantly associated with deterioration of satisfaction with chewing ability. Persons who were more satisfied with their chewing

PAGE 101

88 ability at the preceding interval were more likely to report deteriorated chewing ability at the current interval (OR = 2.806, 95% CI = [2.035 3.870], p < .0001). Compared to blacks, whites were less likely to report that their chewing abilities were deteriorated during the past interval (OR = 0.659, 95% CI = [0.516 0.840], p = 0.0008). Persons who had experienced tooth loss in the past 6 months were more likely to report deteriorated satisfaction with chewing ability (OR = 1.699, 95% CI = [1.062 2.720], p = 0.0271). Persons who had at least 10 occluding pairs of teeth were less likely than those who had fewer than 10 occluding pairs of teeth to report deteriorated chewing abilities (OR = 0.548, 95% CI = [0.429 0.699], p < .0001). Table 5-11. Parameter Estimates from the Second-Stage Model for Research Question 2d (Improvement of Satisfaction with Chewing Ability) Parameter Estimate Odds ratio (95% Confidence interval) p-value Predicted probability of finished fixed prosthodontic treatment within the current interval 0.0512 1.053 (0.274 4.043) 0.9405 Predicted probability of finished removable prosthodontic treatment within the current interval 0.4602 1.584 (1.090 2.303) 0.0159 Satisfaction with chewing ability at the preceding interview (TX-1) 1 = very satisfied/satisfied 2 = very dissatisfied/dissatisfied -1.7693 -0.170 (0.122 0.238) -<.0001 Age 1 = 45 to 64 years old 0 = 65 years old or older -0.0077 -0.992 (0.776 1.269) -0.9511 Gender 1 = male 2 = female -0.2472 -0.781 (0.597 1.022) -0.0713 Race 1 = White 2 = Black -0.0980 -0.907 (0.695 1.182) -0.4696 Tooth loss in the past six months 1 = yes 2 = no -0.0772 -0.926 (0.608 1.410) -0.7191 Number of pairs of teeth 1 = 10 occluding pairs of teeth or more 0 = fewer than 10 occluding pairs of teeth 0.0671 -1.069 (0.812 1.409) -0.6333

PAGE 102

89 Table 5-12. Parameter Estimates from the Second-Stage Model for Research Question 2d (Deterioration of Satisfaction with Chewing Ability) Parameter Estimate Odds ratio (95% Confidence interval) p-value Predicted probability of finished fixed prosthodontic treatment within the current interval -0.1795 0.836 (0.263 2.659) 0.7612 Predicted probability of finished removable prosthodontic treatment within the current interval -0.0826 0.921 (0.216 3.917) 0.9110 Satisfaction with chewing ability at the preceding interview (TX-1) 1 = very satisfied/satisfied 2 = very dissatisfied/dissatisfied 1.0317 -2.806 (2.035 3.870) -<.0001 Age 1 = 45 to 64 years old 0 = 65 years old or older 0.1266 -1.135 (0.904 1.425) -0.2754 Gender 1 = male 2 = female -0.1568 -0.855 (0.668 1.094) -0.2134 Race 1 = White 2 = Black -0.4177 -0.659 (0.516 0.840) -0.0008 Tooth loss in the past six months 1 = yes 2 = no 0.5303 -1.699 (1.062 2.720) -0.0271 Number of pairs of teeth 1 = 10 occluding pairs of teeth or more 0 = fewer than 10 occluding pairs of teeth -0.6023 -0.548 (0.429 0.699) -<.0001 Results for Objective 3 Cross-Sectional Bivariate Associations at the Baseline Interview (Research Question 3a) Table 5-13 shows the bivariate associations between satisfaction with dental appearance at baseline and the baseline prosthodontic services use as well as the selected PEN factors. The bivariate comparisons show that there were no statistically significant associations between satisfaction with dental appearance and wearing a full denture or a partial denture. Subjects who were 65 years old or older, male, whites, and those who had finished at least high school were more satisfied with dental appearance than their counterparts. Subjects who reported “good” general health, who were regular dental

PAGE 103

90 attenders, and who were not frustrated with their past dental care were more satisfied with their dental appearance. The area of residence was not significantly associated with satisfaction with dental appearance. Subjects whose income was not below 150% poverty level, who were able to pay an unexpected $500 dental bill either comfortably or with difficulty were more satisfied with their chewing ability than those whose income was below 150% poverty level and who were unable to pay an unexpected $500 dental bill. There was no statistically significant association between satisfaction with dental appearance and having a certain type of dental insurance. Satisfaction with dental appearance was significantly associated with all need factors except reporting a broken denture. People who had fewer than 10 occluding pairs of teeth and who reported that they had a sore denture, a loose cap or bridge, a broken tooth or cap, a broken filling, a toothache, a sensitive tooth, a cavity, an abscessed tooth, infected or sore gums, bleeding gums, a l oose tooth, stained teeth, a problem with bad breath, and a dry mouth were less satisfied with their dental appearance than those who did not report these problems. Cross-Sectional Multivariate Associatio ns at the Baseline Interview (Research Question 3b) A similar cross-sectional ordered logistic regression using proportional odds model is conducted to identify independent determinants of satisfaction with dental appearance at baseline. The original response of the outcome of interest, satisfaction with dental appearance, is on a four-point ordinal scale: “1 = very satisfied,” “2 = satisfied,” “3 = dissatisfied,” and “4 = very dissatisfied.” However, preliminary logistic regression analyses indicated that the proportional odds assumption was violated when the original

PAGE 104

91 measure of satisfaction with dental appearance is used. When subjects who reported “dissatisfied” are pooled with those who reported “very dissatisfied,” the proportional odds assumption was satisfied. Thus, a recoded outcome variable which is on a threepoint ordinal scare (“1 = very satisfied,” “2 = satisfied,” and “3 = dissatisfied/very dissatisfied”), is used in the final model. Table 5-14 shows the results from the ordinal logistic regression of recoded satisfaction with dental appearance. Similar to what has been described in Table 5-5, the odds ratios in this model are also interpreted as the increase in the odds of moving down a lower category, which means moving up one level of satisfaction with dental appearance (e.g. from “2 = satisfied” to “1 = very satisfied,” or from “3 = dissatisfied/very dissatisfied” to “2 = satisfied”) with an increase in one level of the explanatory covariates. With other factors taken into account, subjects wearing a full denture were more satisfied with their dental appearance than those who were not. Subjects wearing a partial denture also felt more satisfied with their dental appearance than those who were not. Subjects aged 65 or above were less satisfied with their dental appearance and regular dental attenders were more satisfied, compared to their counterparts. Compared to the odds for persons who were unable to pay an unexpected $500 dental bill, the odds of being in a lower category (a higher level of satisfaction with dental appearance) for those who were able to pay the bill comfortably increased nearly 2 times (OR = 3.185). The odds of being in a higher level of satisfaction for persons who had at least 10 occluding pairs of teeth were nearly 6 times the odds for persons who had fewer than 10 occluding pairs of teeth (OR = 5.944). After other conditions were controlled for, having a sore denture, a toothache, bleeding

PAGE 105

92 gums, a loose tooth, or stained teeth were si gnificantly associated with less satisfaction with dental appearance. Table 5-13. Satisfaction with Dental Appearance at Baseline for the Sample Overall, by Baseline Prosthodontic Services Use a nd by Predisposing, Enabling, and Need Factors (PEN Factors) PEN factors (Weighted n) % Subjects who were very satisfied and satisfied with dental appearance % Subjects who were dissatisfied and very dissatisfied with dental appearance p-value Prosthodontic Services Wearing a full denture 1 = yes (81) 0 = no (786) 79.8 75.7 20.2 24.3 0.4157 Wearing a partial denture 1 = yes (170) 0 = no (697) 80.8 74.9 19.2 25.1 0.1066 Predisposing Factors Age 1 = 65 years old or older (360) 0 = 45 to 64 years old (507) 81.5 72.2 18.5 27.8 0.0015 Gender 1 = male (378) 2 = female (489) 80.0 73.1 20.0 26.9 0.0185 Area of residence 1 = urban (432) 0 = rural (435) 75.2 77.0 24.8 23.0 0.5260 Race 1 = White (623) 2 = Black (241) 80.7 65.0 19.3 35.0 <.0001 Level of formal education 1 = completed at least high school (683) 0 = did not complete high school (184) 79.7 62.7 20.3 37.3 <.0001 Self-rated general health 1 = good (639) 2 = not good (222) 81.9 59.7 18.1 40.3 <.0001 Approach to dental care 1 = regular attender (470) 0 = problem attender (396) 88.6 61.2 11.4 38.8 <.0001 Frustration with past care 1 = frustrated (133) 0 = not frustrated (732) 64.3 78.4 35.7 21.6 0.0004 Enabling Factors Poverty Level 1 = below150% poverty level (270) 0 = not below 150% poverty level (518) 65.4 82.9 34.6 17.1 <.0001

PAGE 106

93 Table 5-13. Continued PEN factors (Weighted n) % Subjects who were very satisfied and satisfied with dental appearance % Subjects who were dissatisfied and very dissatisfied with dental appearance p-value Able to pay an unexpected $500 dental bill 1 = able to pay comfortably (403) 2 = able to pay but with difficulty (342) 3 = not able to pay (121) 90.6 67.8 51.0 9.4 32.2 49.0 <.0001 Dental insurance status 1 = has certain type of dental insurance (290) 2 = no dental insurance (577) 79.9 74.2 20.1 25.8 0.0643 Need Factors Number of occluding pairs of teeth 1 = 10 occluding pairs of teeth or more (509) 0 = fewer than 10 occluding pairs of teeth (353) 83.3 65.9 16.7 34.1 <.0001 A sore denture 1 = yes (56) 2 = no (233) 50.1 83.4 49.9 16.7 <.0001 A broken denture 1 = yes (38) 2 = no (251) 67.5 78.3 32.5 21.7 0.1396 A loose cap or bridge 1 = yes (10) 2 = no (856) 48.6 76.4 51.4 23.6 0.0369 A broken tooth or cap 1 = yes (176) 2 = no (681) 48.8 82.9 51.2 17.1 <.0001 A broken filling 1 = yes (127) 2 = no (722) 56.9 79.2 43.1 20.8 <.0001 A toothache 1 = yes (100) 2 = no (766) 51.5 79.3 48.5 20.7 <.0001 A sensitive tooth 1 = yes (261) 2 = no (605) 61.7 82.3 38.3 17.7 <.0001 A cavity 1 = yes (164) 2 = no (640) 50.5 84.6 49.5 15.4 <.0001 An abscessed tooth 1 = yes (22) 2 = no (835) 50.1 76.7 50.0 23.3 0.0043 A infected or sore gums 1 = yes (102) 2 = no (760) 54.2 79.1 45.8 20.9 <.0001

PAGE 107

94 Table 5-13. Continued PEN factors (Weighted n) % Subjects who were very satisfied and satisfied with dental appearance % Subjects who were dissatisfied and very dissatisfied with dental appearance p-value Bleeding gums 1 = yes (116) 2 = no (750) 48.1 80.5 51.9 19.5 <.0001 A loose tooth 1 = yes (115) 2 = no (743) 49.5 80.8 50.5 19.2 <.0001 Stained teeth 1 = yes (341) 2 = no (508) 51.7 92.4 48.3 7.6 <.0001 A problem with bad breath 1 = yes (154) 2 = no (670) 58.9 81.0 41.1 19.0 <.0001 A dry mouth 1 = yes (190) 2 = no (675) 70.4 77.7 29.6 22.3 0.0363

PAGE 108

95 Table 5-14. Cross-Sectional Logistic Regressi on of Satisfaction with Dental Appearance at Baseline PEN factors (Weighted n) Estimate Odds ratio (95% Confidence interval) p-value Prosthodontic Services Has and wears a full denture 1 = yes 0 = no 1.3797 -3.974 (1.726 9.146) -0.0012 Has and wears a partial denture 1 = yes 0 = no 0.8999 -2.459 (1.147 5.275) -0.0208 Predisposing Factors Age 1 = 65 years old or older 0 = 45 to 64 years old -1.2146 -0.297 (0.150 0.589) -0.0005 Gender 1 = male 2 = female 0.1283 -1.292 (0.675 2.476) -0.4393 Race 1 = White 2 = Black -0.2513 -0.605 (0.293 1.251) -0.1750 Approach to dental care 1 = regular attender 0 = problem attender 1.3329 -3.792 (1.736 8.284) -0.0008 Frustration with past care 1 = frustrated 0 = not frustrated -0.4210 -0.656 (0.248 1.736) -0.3961 Enabling Factors Able to pay an unexpected $500 dental bill 1 = able to pay comfortably 2 = able to pay but with difficulty 3 = not able to pay 0.6952 -0.2321 -3.185 (1.071 9.472) 1.260 (0.479 3.316) -0.0128 0.3274 Dental insurance status 1 = has certain type of dental insurance 2 = no dental insurance -0.0518 -0.902 (0.437 1.860) -0.7792 Need Factors Number of occluding pairs of teeth 1 = 10 occluding pairs of teeth or more 0 = fewer than 10 occluding pairs of teeth 1.7824 -5.944 (2.056 17.183) -0.0010 Has a sore denture 1 = yes 2 = no -0.5448 -0.336 (0.139 0.813) -0.0155 Has a broken denture 1 = yes 2 = no -0.4228 -0.429 (0.156 1.182) -0.1018 Has a loose cap or bridge 1 = yes 2 = no 0.5347 -2.913 (0.007 >999.99) -0.7257

PAGE 109

96 Table 5-14. Continued PEN factors (Weighted n) Estimate Odds ratio (95% Confidence interval) p-value Has a broken tooth or cap 1 = yes 2 = no -0.2606 -0.594 (0.201 1.758) -0.3467 Has a broken filling 1 = yes 2 = no -0.4896 -0.376 (0.108 1.307) -0.1239 Has a toothache 1 = yes 2 = no -0.6879 -0.253 (0.091 0.700) -0.0081 Has a sensitive tooth 1 = yes 2 = no -0.2842 -0.566 (0.268 1.199) -0.1375 Has a cavity 1 = yes 2 = no -0.2678 -0.585 (0.219 1.566) -0.2860 Has an abscessed tooth 1 = yes 2 = no 0.7761 -4.722 (0.478 46.685) -0.1843 Has infected or sore gums 1 = yes 2 = no 0.1417 -1.328 (0.417 4.226) -0.6313 Has bleeding gums 1 = yes 2 = no -0.9346 -0.154 (0.047 0.510) -0.0022 Has a loose tooth 1 = yes 2 = no -0.8374 -0.187 (0.077 0.455) -0.0002 Has stained teeth 1 = yes 2 = no -0.6390 -0.279 (0.133 0.582) -0.0007 Has a problem with bad breath 1 = yes 2 = no -0.1113 -0.800 (0.330 1.939) -0.6219 Has a dry mouth 1 = yes 2 = no 0.2007 -1.494 (0.667 3.346) -0.3294

PAGE 110

97 Longitudinal Logistic Regression of Sati sfaction with Dental Appearance with Correction of Selection Bias (Research Question 3c) A longitudinal logistic regression with corr ection of selection bias is conducted to predict the effect of prosthodontic services use on satisfaction with dental appearance at each follow-up interview point during a 24-month period. Table 5-15 lists the valid instruments used in the first-stage models for research question 3c. Table 5-15. Instrumental Variables for Research Question 3c Outcome variable Treatment variable Valid instrumental variables p-value a p-value b Satisfaction with dental appearance Finished fixed prosthodontic treatment Perceived need for dental care 1 = not at this time 0 = other 0.0210 0.6762 whether the dentist or clinic that a subject usually sees is one that is nearest 1 = yes 2 = no 0.0032 0.3164 Finished removable prosthodontic treatment Frustration with past dental care 1 = yes 0 = no 0.0178 0.0707 a Associations between the instrument al variables and the treatment variables b Lack of a direct association between the in strumental variables and the outcome variable Results of the first-stage models for research question 3c As listed in Table 5-7 and Table 5-8. Results of the second-stage model for research question 3c Table 5-16 lists the results from the second-stage model fitted with GEE for the dichotomized satisfaction with dental appearance over a 24-month period. None of the two types of prosthodontic treatment was signi ficantly associated with satisfaction with dental appearance. Compared to blacks, whites were less likely to be dissatisfied with their dental appearance (OR = 0.530, 95% CI = [0.381 0.738], p = 0.0002). Persons who had at least 10 occluding pairs of teeth were less likely than those who had fewer than 10

PAGE 111

98 occluding pairs of teeth to have reported dissatisfaction with their dental appearance (OR = 0.365, 95% CI = [0.255 0.523], p < .0001). Table 5-16. Parameter Estimates from the Second-Stage Model for Research Question 3c Parameter Estimate Odds ratio (95% Confidence interval) p-value Predicted probability of finished fixed prosthodontic treatment within the current interval -0.4317 0.649 (0.207 2.039) 0.4596 Predicted probability of finished removable prosthodontic treatment within the current interval -0.8515 0.427 (0.075 2.416) 0.3358 Age 1 = 45 to 64 years old 0 = 65 years old or older -0.3053 -0.737 (0.527 1.030) -0.0737 Gender 1 = male 2 = female -0.3135 -0.731 (0.519 1.028) 0.0720 Race 1 = White 2 = Black -0.6349 -0.530 (0.381 0.738) -0.0002 Tooth loss in the past six months 1 = yes 2 = no 0.0235 -1.024 (0.653 1.606) -0.9185 Number of pairs of teeth 1 = 10 occluding pairs of teeth or more 0 = fewer than 10 occluding pairs of teeth -1.0068 -0.365 (0.255 0.523) -<.0001 Longitudinal Logistic Regression of Ch anges in Satisfaction with Dental Appearance with Correction of Selection Bias (Research Question 3d) Longitudinal logistic regressions with correction of selection bias are conducted to predict the effect of prosthodontic services use on changes in satisfaction with dental appearance, namely, improvement of satisfaction with dental appearance and deterioration of satisfaction with dental appearance, at each follow-up interview point during a 24-month period. Table 5-17 lists the valid instruments used in the first-stage models for research question 3d. Results of the first-stage models for research question 3d As listed in Table 5-7 and Table 5-8.

PAGE 112

99 Results of the second-stage models for research question 3d Table 5-18 shows the results from the second-stage model fitted with GEE for the improvement of satisfaction with dental appearance over a 24-month period. Subjects who had received finished removable prosthodon tic treatment were 4 times more likely to report improved satisfaction with dental appearance (OR = 4.945). Persons who were more satisfied with their dental appearance at the preceding interval were less likely to report improved dental appearance at the current interval (OR = 0.157, 95% CI = [0.121 0.204], p < .0001). Persons who had at least 10 occluding pairs of teeth were more likely than those who had fewer than 10 occluding pairs of teeth to report improved dental appearance (OR = 1.529, 95% CI = [1.204 1.942], p = 0.0005). Table 5-17. Instrumental Variables for Research Question 3d Outcome variable Treatment variable Valid instrumental variables p-value a p-value b Improvement of satisfaction with dental appearance Finished fixed prosthodontic treatment Perceived need for dental care 1 = not at this time 0 = other 0.0210 0.5581 Whether the dentist or clinic that a subject usually sees is one that is nearest 1 = yes 2 = no 0.0032 0.6476 Approach to dental care 1 = regular attender 0 = problem attender 0.0076 0.0696 Able to pay an unexpected $500 dental bill 1 = able to pay comfortably 2 = able to pay, but with difficulty 3 = not able to pay 0.0089 0.0481 0.8236 0.6577 Finished removable prosthodontic treatment Frustration with past dental care 1 = yes 0 = no 0.0178 0.6078 Deterioration of satisfaction with dental appearance Finished fixed prosthodontic treatment Whether the dentist or clinic that a subject usually sees is one that is nearest 1 = yes 2 = no 0.0032 0.8591

PAGE 113

100 Table 5-17 Continued Outcome variable Treatment variable Valid instrumental variables p-value a p-value b Approach to dental care 1 = regular attender 0 = problem attender 0.0076 0.9594 Able to pay an unexpected $500 dental bill 1 = able to pay comfortably 2 = able to pay, but with difficulty 3 = not able to pay 0.0089 0.0481 0.3767 0.8502 Finished removable prosthodontic treatment Frustration with past dental care 1 = yes 0 = no 0.0178 0.3841 a Associations between the instrument al variables and the treatment variables b Lack of a direct association between the in strumental variables and the outcome variable Table 5-18. Parameter Estimates from the Second-Stage Model for Research Question 3d (Improvement of Satisfaction with Dental Appearance) Parameter Estimate Odds ratio (95% Confidence interval) p-value Predicted probability of finished fixed prosthodontic treatment within the current interval -0.7116 0.491 (0.151 1.598) 0.2374 Predicted probability of finished removable prosthodontic treatment within the current interval 1.5984 4.945 (1.221 20.030) 0.0251 Satisfaction with dental appearance at the preceding interview (TX-1) 1 = very satisfied/satisfied 2 = very dissatisfied/dissatisfied -1.8504 -0.157 (0.121 0.204) -<.0001 Age 1 = 45 to 64 years old 0 = 65 years old or older 0.0291 -1.030 (0.833 1.272) -0.7873 Gender 1 = male 2 = female -0.0400 -0.961 (0.772 1.195) -0.7190 Race 1 = White 2 = Black 0.1670 -1.182 (0.939 1.488) -0.1555 Tooth loss in the past six months 1 = yes 2 = no 0.2264 -1.254 (0.860 1.829) -0.2391 Number of pairs of teeth 1 = 10 occluding pairs of teeth or more 0 = fewer than 10 occluding pairs of teeth 0.4244 -1.529 (1.204 1.942) -0.0005

PAGE 114

101 Table 5-19 shows the results from the second-stage model fitted with GEE for the deterioration of satisfaction with dental appearance over a 24-month period. Having received either type of prosthodontic treatment was not associated with deterioration of satisfaction with dental appearance. Persons who were more satisfied with their dental appearance at the preceding interval were more likely to report deteriorated dental appearance at the current interval (OR = 3.441, 95% CI = [2.558 4.630], p < .0001). Whites and persons who had at least 10 occluding pairs of teeth were less likely to report their satisfaction with dental appearance had deteriorated during the past interval, comparing to their respective counterpart. Table 5-19. Parameter Estimates from the Second-Stage Model for Research Question 3d (Deterioration of Satisfaction with Dental Appearance) Parameter Estimate Odds ratio (95% Confidence interval) p-value Predicted probability of finished fixed prosthodontic treatment within the current interval -0.6833 0.505 (0.159 1.605) 0.2468 Predicted probability of finished removable prosthodontic treatment within the current interval 0.3718 1.450 (0.362 5.806) 0.5994 Satisfaction with dental appearance at the preceding interview (TX-1) 1 = very satisfied/satisfied 2 = very dissatisfied/dissatisfied 1.2359 -3.441 (2.558 4.630) -<.0001 Age 1 = 45 to 64 years old 0 = 65 years old or older 0.0815 -1.085 (0.873 1.348) -0.4621 Gender 1 = male 2 = female -0.0893 -0.915 (0.732 1.143) -0.4327 Race 1 = White 2 = Black -0.2589 -0.772 (0.613 0.972) -0.0276 Tooth loss in the past six months 1 = yes 2 = no -0.1329 -0.876 (0.546 1.404) -0.5812 Number of pairs of teeth 1 = 10 occluding pairs of teeth or more 0 = fewer than 10 occluding pairs of teeth -0.2389 -0.787 (0.623 0.995) -0.0457

PAGE 115

102 Results for Objective 4 Cross-Sectional Bivariate Associations at the Baseline Interview (Research Question 4a) Table 5-20 shows the bivariate associations between self-rated overall oral health at baseline and the baseline prosthodontic services use as well as the selected PEN factors. The bivariate comparisons show no statistically significant association between wearing a full denture and self-rated overall oral health. More subjects who were not wearing a partial denture rated their overall oral health as “excellent,” “very good,” or “good” compared to those who were wearing a partial denture. Subjects who were 65 years old or older, male, White, and who had finished at least high school rated their overall oral health better than their counterparts. The ratings of overall oral health as “excellent,” “very good” or “good” were more common among subjects who reported “good” general health and subjects who were regular dental attenders. Rural/urban area of residence and frustration with past care were not significantly associated with satisfaction with dental appearance. Ratings of overall oral health were significantly associated with all three enabling factors. Subjects whose income was not below 150% poverty level, who were able to pay an unexpected $500 dental bill either comfortably or with difficulty, and who had a certain type of dental insurance rated their overall oral health better than their counterparts. Self-rated overall oral health was significantly associated with all need factors except reporting a broken denture. The ratings of the overall oral health were worse among people with fewer than 10 occluding pairs of teeth and who reported that they had a sore denture, a loose cap or bridge, a br oken tooth or cap, a broken filling, a toothache,

PAGE 116

103 a sensitive tooth, a cavity, an abscessed tooth, infected or sore gums, bleeding gums, a loose tooth, stained teeth, a problem with bad breath, and a dry mouth. Table 5-20. Self-Rated Overall Oral Health at Baseline for the Sample Overall, by Baseline Prosthodontic Services Use a nd by Predisposing, Enabling, and Need factors (PEN Factors) PEN factors (Weighted n) % Subjects who rated their overall oral health as excellent, very good, or good % Subjects who rated their overall oral health as fair or poor p-value Prosthodontic Services Has and wears a full denture 1 = yes (81) 0 = no (788) 72.3 80.1 27.7 19.9 0.1024 Has and wears a partial denture 1 = yes (170) 0 = no (699) 73.0 80.9 27.0 19.1 0.0226 Predisposing Factors Age 1 = 65 years old or older (361) 0 = 45 to 64 years old (508) 87.2 73.7 12.8 26.3 <.0001 Gender 1 = male (379) 2 = female (490) 83.6 76.1 16.4 23.9 0.0069 Area of residence 1 = urban (434) 0 = rural (435) 79.9 78.8 20.1 21.2 0.6856 Race 1 = White (624) 2 = Black (242) 83.6 69.0 16.4 31.0 <.0001 Level of formal education 1 = completed at least high school (685) 0 = did not complete high school (184) 83.3 64.7 16.7 35.3 <.0001 Self-rated general health 1 = good (639) 2 = not good (224) 85.4 62.5 14.6 37.5 <.0001 Approach to dental care 1 = regular attender (470) 0 = problem attender (399) 92.2 64.2 7.8 35.8 <.0001 Frustration with past care 1 = frustrated (133) 0 = not frustrated (734) 75.9 80.1 24.1 19.9 0.2684 Enabling Factors Poverty Level 1 = below150% poverty level (272) 0 = not below 150% poverty level (518) 66.7 85.6 33.3 14.4 <.0001

PAGE 117

104 Table 5-20. Continued PEN factors (Weighted n) % Subjects who rated their overall oral health as excellent, very good, or good % Subjects who rated their overall oral health as fair or poor p-value Able to pay an unexpected $500 dental bill 1 = able to pay comfortably (403) 2 = able to pay but with difficulty (342) 3 = not able to pay (122) 89.3 77.2 52.1 10.7 22.8 47.9 <.0001 Dental insurance status 1 = has certain type of dental insurance (290) 2 = no dental insurance (579) 83.5 77.2 16.5 22.8 0.0314 Need Factors Number of occluding pairs of teeth 1 = 10 occluding pairs of teeth or more (509) 0 = fewer than 10 occluding pairs of teeth (356) 89.8 64.5 10.2 35.5 <.0001 Has a sore denture 1 = yes (57) 2 = no (233) 54.1 77.6 45.9 22.4 0.0004 Has a broken denture 1 = yes (38) 2 = no (251) 60.2 75.0 39.8 25.0 0.0557 Has a loose cap or bridge 1 = yes (10) 2 = no (858) 54.7 79.7 45.3 20.3 0.0482 Has a broken tooth or cap 1 = yes (178) 2 = no (681) 54.7 85.7 45.3 14.3 <.0001 Has a broken filling 1 = yes (127) 2 = no (723) 61.5 82.4 38.5 17.6 <.0001 Has a toothache 1 = yes (100) 2 = no (769) 62.0 81.6 38.0 18.4 <.0001 Has a sensitive tooth 1 = yes (261) 2 = no (607) 69.8 83.5 30.2 16.5 <.0001 Has a cavity 1 = yes (165) 2 = no (640) 49.7 88.3 50.3 11.7 <.0001 Has an abscessed tooth 1 = yes (22) 2 = no (836) 43.2 80.4 56.8 19.6 <.0001 Has infected or sore gums 1 = yes (102) 2 = no (763) 59.5 82.1 40.5 17.9 <.0001

PAGE 118

105 Table 5-20. Continued PEN factors (Weighted n) % Subjects who rated their overall oral health as excellent, very good, or good % Subjects who rated their overall oral health as fair or poor p-value Has bleeding gums 1 = yes (117) 2 = no (751) 54.5 83.3 45.5 16.7 <.0001 Has a loose tooth 1 = yes (115) 2 = no (744) 53.5 83.7 46.5 16.3 <.0001 Has stained teeth 1 = yes (341) 2 = no (509) 61.1 91.9 38.9 8.1 <.0001 Has a problem with bad breath 1 = yes (154) 2 = no (671) 59.6 85.3 40.4 14.7 <.0001 Has a dry mouth 1 = yes (190) 2 = no (676) 69.7 82.2 30.3 17.8 0.0002 Cross-sectional Multivariate Associatio ns at the Baseline Interview (Research Question 4b) A similar cross-sectional ordered logistic regression using proportional odds model is conducted to evaluate the multivariate associations between self-rated overall oral health and a set of independent measures which include baseline prosthodontic use, predisposing, enabling, and need factors. Proportional odds assumption is satisfied in this model. Table 5-21 shows the results from the ordinal logistic regression of the baseline self-rated overall oral health. Neither wearing a full denture nor wearing a partial denture was associated with baseline self-rated overall oral health. Except approach to dental care, none of the remaining predisposing and enabling factors was significant predicators of self-rated overall oral health. Regular dental care attenders rated their overall oral health better than those problem-oriented attenders. The odds of being in a higher level of ratings for persons who had at least 10 occluding pairs of teeth were nearly 4 times the odds for

PAGE 119

106 persons who had fewer than 10 occluding pair s of teeth (OR = 3.807). Reporting a sore denture, a broken filling, a toothache, bleedi ng gums, stained teeth, and a problem with bad breath were significantly associated with lower ratings of overall oral health. Table 5-21. Cross-Sectional Logistic Regressi on of Self-Rated Overall Oral Health at Baseline PEN factors (Weighted n) Estimate Odds ratio (95% Confidence interval) p-value Prosthodontic Services Has and wears a full denture 1 = yes 0 = no -0.0211 -0.495 (0.507 1.890) -0.9499 Has and wears a partial denture 1 = yes 0 = no 0.3392 -0.495 (0.768 2.565) -0.2700 Predisposing Factors Age 1 = 65 years old or older 0 = 45 to 64 years old 0.5041 -0.495 (0.969 2.828) -0.0650 Gender 1 = male 2 = female -0.0994 -0.495 (0.484 1.387) -0.4585 Race 1 = White 2 = Black 0.2404 -0.495 (0.884 2.960) -0.1189 Approach to dental care 1 = regular attender 0 = problem attender 0.8530 -0.495 (1.244 4.428) -0.0085 Frustration with past care 1 = frustrated 0 = not frustrated 0.6235 -0.495 (0.872 3.993) -0.1083 Enabling Factors Able to pay an unexpected $500 dental bill 1 = able to pay comfortably 2 = able to pay but with difficulty 3 = not able to pay -0.2230 -0.0936 -0.495 (0.241 1.408) 0.495 (0.305 1.444) -0.3291 0.6279 Dental insurance status 1 = has certain type of dental insurance 2 = no dental insurance 0.1283 -0.495 (0.717 2.330) -0.3934 Need Factors Number of occluding pairs of teeth 1 = 10 occluding pairs of teeth or more 0 = fewer than 10 occluding pairs of teeth 1.3368 -0.495 (1.616 8.971) -0.0022 Has a sore denture 1 = yes 2 = no -0.3514 -0.495 (0.249 0.983) -0.0445

PAGE 120

107 Table 5-21. Continued PEN factors (Weighted n) Estimate Odds ratio (95% Confidence interval) p-value Has a broken denture 1 = yes 2 = no 0.0156 -1.032 (0.464 2.293) -0.9389 Has a loose cap or bridge 1 = yes 2 = no -0.8928 -0.168 (0.001 26.381) -0.4890 Has a broken tooth or cap 1 = yes 2 = no 0.0558 -1.118 (0.481 2.599) -0.7955 Has a broken filling 1 = yes 2 = no -0.6577 -0.268 (0.101 0.715) -0.0085 Has a toothache 1 = yes 2 = no -0.4139 -0.437 (0.198 0.965) -0.0404 Has a sensitive tooth 1 = yes 2 = no -0.2754 -0.577 (0.309 1.074) -0.0829 Has a cavity 1 = yes 2 = no -0.3672 -0.480 (0.221 1.040) -0.0628 Has an abscessed tooth 1 = yes 2 = no 0.0800 -1.173 (0.188 7.322) -0.8641 Has infected or sore gums 1 = yes 2 = no -0.0889 -0.837 (0.333 2.102) -0.7052 Has bleeding gums 1 = yes 2 = no -0.5124 -0.359 (0.145 0.890) -0.0270 Has a loose tooth 1 = yes 2 = no -0.3085 -0.540 (0.271 1.074) -0.0791 Has stained teeth 1 = yes 2 = no -0.4483 -0.408 (0.224 0.743) -0.0034 Has a problem with bad breath 1 = yes 2 = no -0.3685 -0.479 (0.236 0.969) -0.0405 Has a dry mouth 1 = yes 2 = no -0.2044 -0.664 (0.342 1.289) -0.2268

PAGE 121

108 Longitudinal Logistic Regression of Self-R ated Overall Oral Health with Correction of Selection Bias (Research Question 4c) A longitudinal logistic regression with corr ection of selection bias is conducted to predict the effect of prosthodontic services use on self-rated overall oral health at each follow-up interview point during a 24-month period. Table 5-22 lists the valid instruments used in the first-stage model for research question 4c. Table 5-22. Instrumental Variables for Research Question 4c Outcome variable Treatment variable Valid instrumental variables p-value a p-value b Self-rated overall oral health Finished fixed prosthodontic treatment Perceived need for dental care 1 = not at this time 0 = other 0.0210 0.5038 Whether the dentist or clinic that a subject usually sees is one that is nearest 1 = yes 2 = no 0.0032 0.0792 Finished removable prosthodontic treatment Frustration with past dental care 1 = yes 0 = no 0.0178 0.2847 a Associations between the instrument al variables and the treatment variables b Lack of a direct association between the in strumental variables and the outcome variable Results of the first-stage models for research question 4c As listed in Table 5-7 and Table 5-8. Results of the second-stage model for research question 4c Table 5-23 shows the results from the second-stage model fitted with GEE for the dichotomized self-rated overall oral health over a 24-month period. None of the two types of prosthodontic treatment was significan tly associated with self-rated overall oral health. People who were aged 45 to 64 years old were less likely to rate their overall oral health negatively (“fair/poor”) (OR = 0.577, 95% CI = [0.404 0.822], p = 0.0024). Compared to blacks, whites were less likely to rate their overall oral health as “fair” or “poor” (OR = 0.440, 95% CI = [0.301 0.642], p < .0001). Persons who had at least 10

PAGE 122

109 occluding pairs of teeth and who had r eceived finished removable prosthodontic treatment were less likely to rate their overall oral health as “fair” or “poor” (OR = 0.178, 95% CI = [0.116 0.272], p < .0001). Table 5-23. Parameter Estimates from the Second-Stage Model for Research Question 4c Parameter Estimate Odds ratio (95% Confidence interval) p-value Predicted probability of finished fixed prosthodontic treatment within the current interval 0.0811 1.084 (0.331 3.548) 0.8933 Predicted probability of finished removable prosthodontic treatment within the current interval -0.8047 0.447 (0.117 1.704) 0.2383 Age 1 = 45 to 64 years old 0 = 65 years old or older -0.5505 -0.577 (0.404 0.822) -0.0024 Gender 1 = male 2 = female -0.2119 -0.809 (0.544 1.204) -0.2961 Race 1 = White 2 = Black -0.8220 -0.440 (0.301 0.642) -<.0001 Tooth loss in the past six months 1 = yes 2 = no 0.3620 -1.436 (0.963 2.141) -0.0757 Number of pairs of teeth 1 = 10 occluding pairs of teeth or more 0 = fewer than 10 occluding pairs of teeth -1.7270 -0.178 (0.116 0.272) -<.0001 Longitudinal Logistic Regression of Change s in Self-Rated Overall Oral Health with Correction of Selection Bias (Research Question 4d) Longitudinal logistic regressions with correction of selection bias are conducted to predict the effect of prosthodontic services use on changes of self-rated overall oral health, namely, improvement of self-rated overall oral health and deterioration of selfrated overall oral health, at each follow-up interview point during a 24-month period. Table 5-24 lists the valid instruments used in the first-stage model for research question 4d.

PAGE 123

110 Table 5-24. Instrumental Variables for Research Question 4d Outcome variable Treatment variable Valid instrumental variables p-value a p-value b Improvement of self-rated overall oral health Finished fixed prosthodontic treatment Perceived need for dental care 1 = not at this time 0 = other 0.0210 0.0961 Whether the dentist or clinic that a subject usually sees is one that is nearest 1 = yes 2 = no 0.0032 0.8692 Approach to dental care 1 = regular attender 0 = problem attender 0.0076 0.0569 Able to pay an unexpected $500 dental bill 1 = able to pay comfortably 2 = able to pay, but with difficulty 3 = not able to pay 0.0089 0.0481 0.8516 0.6000 Finished removable prosthodontic treatment Frustration with past dental care 1 = yes 0 = no 0.0178 0.0566 Deterioration of self-rated overall oral health Finished fixed prosthodontic treatment Perceived need for dental care 1 = not at this time 0 = other 0.0210 0.4457 Approach to dental care 1 = regular attender 0 = problem attender 0.0076 0.4843 Finished removable prosthodontic treatment Frustration with past dental care 1 = yes 0 = no 0.0178 0.7330 a Associations between the instrument al variables and the treatment variables b Lack of a direct association between the in strumental variables and the outcome variable Results of the first-stage models for research question 4d As listed in Table 5-7 and Table 5-8. Results of the second-stage models for research question 4d Table 5-25 shows the results from the second-stage model fitted with GEE for the improvement of self-rated overall oral health over a 24-month period. None of the two types of prosthodontic treatment was significan tly associated with improvement of selfrated overall oral health. Persons who rated their overall oral health as “excellent/very

PAGE 124

111 good/good” at the preceding interval were less likely to have reported improved overall oral health at the current interval (OR = 0.320, 95% CI = [0.238 0.431], p < .0001). Table 5-25. Parameter Estimates from the Second-Stage Model for Research Question 4d (Improvement of Self-Rated Overall Oral Health) Parameter Estimate Odds ratio (95% Confidence interval) p-value Predicted probability of finished fixed prosthodontic treatment within the current interval -0.7264 0.484 (0.151 1.550) 0.2214 Predicted probability of finished removable prosthodontic treatment within the current interval 0.6502 1.916 (0.652 5.633) 0.2374 Self-rated overall oral health at the preceding interview (TX-1) 1 = excellent/very good/good 2 = fair/poor -1.1380 -0.320 (0.238 0.431) -<.0001 Age 1 = 45 to 64 years old 0 = 65 years old or older 0.0826 -1.086 (0.894 1.320) -0.4060 Gender 1 = male 2 = female -0.0251 -0.975 (0.792 1.200) -0.8125 Race 1 = White 2 = Black 0.0803 -1.084 (0.878 1.338) -0.4550 Tooth loss in the past six months 1 = yes 2 = no 0.1970 -1.218 (0.823 1.801) -0.3239 Number of pairs of teeth 1 = 10 occluding pairs of teeth or more 0 = fewer than 10 occluding pairs of teeth 0.0218 -1.022 (0.821 1.273) -0.8455 Table 5-26 shows the results from the second-stage model fitted with GEE for the deterioration of self-rated overall oral health over a 24-month period. None of the two types of prosthodontic treatment was significantly associated with deterioration of selfrated overall oral health. Persons who rated their overall oral health as “excellent/very good/good” at the preceding interval were nearly 3 times more likely (OR = 4.280) to have reported deteriorated overall oral health at the current interval. Compared to blacks, whites were less likely to have reported deteriorated overall oral health during 6-month

PAGE 125

112 interval (OR = 0.716, 95% CI = [0.582 0.880], p = 0.0015). Persons who had at least 10 occluding pairs of teeth were less likely than those who had fewer than 10 occluding pairs of teeth to have reported deteriorated overall oral health (OR = 0.708, 95% CI = [0.573 0.876], p = 0.0014). Table 5-26. Parameter Estimates from the Second-Stage Model for Research Question 4d (Deterioration of Self-Rated Overall Oral Health) Parameter Estimate Odds ratio (95% Confidence interval) p-value Predicted probability of finished fixed prosthodontic treatment within the current interval -0.2066 0.813 (0.288 2.298) 0.6967 Predicted probability of finished removable prosthodontic treatment within the current interval 0.0856 1.089 (0.304 3.902) 0.8954 Self-rated overall oral health at the preceding interview (TX-1) 1 = excellent/very good/good 2 = fair/poor 1.4540 -4.280 (3.119 5.874) -<.0001 Age 1 = 45 to 64 years old 0 = 65 years old or older 0.1090 -1.115 (0.921 1.350) -0.2634 Gender 1 = male 2 = female -0.0589 -0.943 (0.774 1.148) -0.5578 Race 1 = White 2 = Black -0.3345 -0.716 (0.582 0.880) -0.0015 Tooth loss in the past six months 1 = yes 2 = no 0.3150 -1.370 (0.893 2.102) -0.1489 Number of pairs of teeth 1 = 10 occluding pairs of teeth or more 0 = fewer than 10 occluding pairs of teeth -0.3450 -0.708 (0.573 0.876) -0.0014

PAGE 126

113 CHAPTER 6 DISCUSSION Discussion of the Results for Objective 1 The results suggest that fixed prosthodontic services are more commonly received than removable prosthodontic services. Finished fixed prosthodontic services were the most common form of prosthodontic services received. Incidence rates of unfinished fixed prosthodontics and unfinished removable prosthodontics were very low, which suggest that most prosthodontic treatments can be finished within 6 months. Subjects tended to rate their chewing ability positively. In this study, approximately 84% of subjects reported “satisfied” or “very satisfied” with their chewing ability at the baseline interview. The follow-up interviews show a similar trend. Over 24 months, less than 17% of the study subjects (varies from 13.4% to 16.7%, depending on the interval) reported “dissatisfied” or “very dissatisfied” with their chewing ability. Fewer subjects were “satisfied” or “very satisfied” with their dental appearance than chewing ability. Approximately 76% of subjects reported to be “satisfied” or “very satisfied” with their dental appearance. Over 24 months, 18.7% to 22.3% of the subjects were “dissatisfied” or “very dissatisfied” with their dental appearance. The prevalence of dissatisfaction with dental appearance in the current study is similar to the findings obtained from the Oral Health Impact Prof ile (OHIP) (Slade and Spencer, 1994) and the Geriatric Oral Health Assessment Index (GOHAI) (Atchison and Dolan, 1990). The prevalence rate was 26.9% in the study using the OHIP (Slade and Spencer, 1994) and 27.7% in the study using the GOHAI (Atchison and Dolan, 1990). The slight variations

PAGE 127

114 observed in this dissertation and the other two studies may be an artifact of differences in the question wording and sampling frame across the studies. For example, the FDCS queries the current satisfaction with dental appearance while the OHIP questions are based on the problems experienced in the year before administration (Slade and Spencer, 1994). The subjects in the study using the GOHAI ( 65 years) (Atchison and Dolan, 1990) are older than those in this dissertation ( 45 years). The prevalence of self-rated overall oral hea lth shows a similar trend in satisfaction with chewing ability and satisfaction with dental appearance. Many more subjects rated their overall oral health positively (“excellent,” “very good,” or “good”) than negatively (“fair” or “poor”). Proportions reporting positive oral health are comparable to previous findings. Matthias et al. (1995) studied 550 older subjects with a mean age of 74.5 years and found that 75% of participants reported “excellent,” “very good,” or “good” oral health. This proportion seems higher in this dissertation than that of other studies (Berkey et al., 1985; Reinsine and Bailit, 1980). Reinsine and Bailit (1980) found that 65% of subjects (mean age = 34.6) reported “excellent” or “good” health. This number is 67% in Berkey’s sample (mean age = 86.8). Matthias et al. (1995) concluded that it is difficult to speculate on the variations in these self-ratings. Differences in sample size, age, and other demographic factors may contribute to the variations. It is observed that people’s self-rated oral health is not static. It changes over time. The incidence of improvement is similar to the incidence of deterioration across all three measures of oral health (satisfaction with chewing ability, satisfaction with dental appearance, and self-rated overall oral health).

PAGE 128

115 Discussion of the Results for Objective 2 The bivariate results indicate that wearing a full or partial denture at baseline did not significantly affect subjectsÂ’ satisfaction with chewing ability. However, some PEN variables were significantly associated with it. A higher prevalence of satisfaction was observed among whites, high school graduates, regular dental attenders, people who reported their general health positively, and people who were not frustrated with their past dental care. This dissertation also documents that all three enabling factors (150% poverty level, ability to pay an unexpected $500 dental bill, and dental insurance status) were significantly associated with satisfaction with chewing ability. Satisfaction with chewing ability was more common among subjects with at least 10 occluding pairs of teeth or more and those who did not repor t having problems of oral disease/tissue damage. The results of the cross-sectional ordinal logistic regression present the independent impact of covariates on baseline satisfaction with chewing ability. Contrary to the results obtained from the bivariate analysis, wearing a partial denture at baseline was significantly associated with a greater satisf action with chewing ability, with other factors taken into account. The levels of satisfaction with chewing ability were higher among people who were regular dental attenders, and those who reported a sore denture, a toothache and/or abscessed tooth, a loose toot h, or dry mouth. Neither ability to pay an unexpected $500 dental bill nor dental insuran ce status were a significant predictor of satisfaction with chewing ability. Van Waas et al. (1994) studied a sample of 320 dentate noninstitutionalized older subjects and found 77% of the removable partial denture wearers reported improved chewing ability than before they received the treatment. In this dissertation, the results of

PAGE 129

116 the second-stage longitudinal model for “satisfaction with chewing ability” indicate that having received either types of prosthodontic treatment was not significantly associated with satisfaction with chewing ability. However, the model for “improvement of satisfaction with chewing ability” shows th at subjects who had received finished removable prosthodontic treatment were more likely to report improved satisfaction with chewing ability. The findings suggest that finished removable prosthodontic treatment acts in a curative manner that improves already deteriorated satisfaction with chewing ability. The findings also suggest that the dynamic change in oral health must be integrated into the assessment of treatment effectiveness. To prevent or ameliorate decline in masticatory function for people with tooth loss, fixed or removable prosthodontic appliances are the clinical standard of care. Previous studies using objective measures of masticatory performance showed inconsistent results: some researchers concluded that prosthodontic appliances did not provide good functional replacement for natural dentition (Carlsson, 1984; Kapur and Soman, 2004; Mahmood et al., 1992), while others indicated that objective masticatory function was improved after prosthodontic treatment (Atk inson and Ralph, 1971; Carlson et al., 1992; Van der Bilt et al., 1994). Contrary to the conflicting results obtained from the objective evaluations, subjective studies generally showed many people were satisfied with their chewing ability even when they had few or no remaining natural teeth and had restricted objective biting force (Tsuga et al., 1998; Van der Bilt et al., 1994). They found that subjectively assessed chewing ability is weakly correlated with objective bite force (Tsuga et al., 1998) and it was unpredictable how much the satisfaction with chewing ability would increase as a function of prost hodontic treatment (Van der Bilt et al., 1994).

PAGE 130

117 Boretti et al. (1995) claimed that it was patients themselves who wore and functioned with the prosthodontic appliances, thus their perception and satisfaction with those appliances were very important. In the case of complete dentures, the subjective assessment might be more important than the objective tests (Boretti et al., 1995). The findings from this dissertation confirm the va lue of removable prosthodontic treatment in improving the satisfaction with chewing ability. The findings from the second-stage longitudinal models for the dynamic outcomes (“improvement of satisfaction with chewing ability” and “deterioration of satisfaction with chewing ability”) also show that being more satisfied with chewing ability at the preceding interval was significantly associated with a lower likelihood of reporting improved satisfaction with chewing ability at the current interval but a higher likelihood of reporting deteriorated satisfaction with chewing ability at the current interval. The results suggest that previous satisfaction with chewing ability does affect the likelihood of both positive (“improvement”) and negative (“deterioration”) change. Before making a conclusion that previous satisfaction influences the treatment effect, attention should be given regarding ceiling and floor effects. Ceiling effects exist when a score reaches a maximum extreme, while floor effects exist when a score reaches a minimum extreme. In both situations, there is only one direction in which subsequent measurements can change–to the middle. Therefore, for those people who report “very satisfied with chewing ability,” further improvement in satisfaction with chewing ability will not be detected. This results in an increased probability of detecting a decrement. The inverse is true for floor effects. Based on the ways “improvement” and “deterioration” are

PAGE 131

118 defined in this dissertation (Chapter 4), it is difficult to tell to what extent that previous satisfaction affects the changes, due to potential ceiling and floor effects. The results from the second-stage longitudinal models for “satisfaction with chewing ability” and “deterioration of satisfaction with chewing ability” indicate that whites were less likely to be dissatisfied with their chewing ability and whites were also less likely to report deteriorated satisfaction with chewing ability. Since “deterioration of satisfaction with chewing ability” is a dynamic measure which captures the change in satisfaction rather than satisfaction itself, the findings suggest that there was a racial difference in both satisfaction with chewing ability as well as the dynamic change in satisfaction with chewing ability over time af ter the prosthodontic services and other factors were taken into account. The levels of satisfaction with chewing ability among whites were higher and more stable over time than among blacks, thus, less likely to decline. A similar finding was also found for subjects with more than 10 occluding pairs of teeth. They were less likely than subjects with fewer than 10 occluding pairs of teeth to be dissatisfied with their chewing ability and were less likely to have reported deteriorated satisfaction with chewing ability. For the same abovementioned reason, for subjects with more than 10 occluding pairs of teeth, the levels of satisfaction with chewing ability was high and stable over time, thus less likely to worsen. Discussion of the Results for Objective 3 Dental appearance is important in social interaction (Oosterhaven et al., 1989). Previous studies have shown that individuals with missing teeth, especially with missing anterior teeth, are dissatisfied with their dental appearance (Oosterhaven et al., 1989; York and Holtzman, 1999). Turunen et al . (1993) found that housing conditions and

PAGE 132

119 education were significantly associated with satisfaction with dental appearance. In this dissertation, baseline bivariate analyses show whether or not wearing a full or partial denture was not significantly associated with satisfaction with dental appearance but most of predisposing, enabling, and need factors were significantly related to it. Demographic factors, such as age, gender, race, education, general health, approach to dental care, and experience with past dental care, affected s ubjects’ satisfaction with dental appearance. Persons who reported “good” general health, who were regular dental attenders, and who were not frustrated with their past dental care were more satisfied with their dental appearance. Enabling factors such as poverty level and ability to pay were significantly associated with satisfaction with dental a ppearance. With the exception of reporting a broken denture, the results of bivariate analysis show that satisfaction with dental appearance was significantly associated with all need factors. There is a lack of consistency in the litera ture regarding the impact of prosthodontic services on patient satisfaction with their dental appearance. Matthias et al. (1993) reported that dentures were associated with unattractive appearance, while Gordon and co-workers (1988) found that denture wearer s had higher appearance ratings than did those with natural teeth. In this dissertation, the independent determinants of satisfaction with dental appearance at baseline were identif ied from the results of the cross-sectional logistic regression. Contrary to the results obtained from the bivariate comparisons, subjects wearing a full or partial denture were more satisfied with their dental appearance than those who were not, after controlling for the number of occluding pairs of teeth and other PEN factors. Persons with at least 10 occluding pairs of teeth were more satisfied

PAGE 133

120 with their dental appearance than persons w ith fewer than 10 occluding pairs of teeth (OR = 5.944). Previous studies have shown that many fact ors, such as gender, race, discolored teeth, presence of restoration, personality factor s, and attitude affect patient satisfaction with dental appearance (Gordon et al., 1988; Neumann et al., 1989). The results from the current study confirm previous findings. This dissertation also shows that subjects who were 65 years or old, who were unable to pay an unexpected $500 dental bill, who had fewer than 10 occluding pairs of teeth, and who had reported a sore denture, a toothache, bleeding gums, a loose tooth, or stained teet h were significantly associated with less satisfaction with dental appearance. The findings from the second-stage longitudinal models of the dynamic outcomes (“improvement of satisfaction with dental appearance” and “deterioration of satisfaction with dental appearance”) show that being more satisfied with dental appearance at the preceding interval was associated with a lower likelihood of reporting improved satisfaction with dental appearance but a higher likelihood of reporting deteriorated satisfaction with dental appearance at the current interval. As mentioned previously, ceiling and flooring effects should be considered to interpret the results. Results from the second-stage longitudinal models for “satisfaction with dental appearance” and “deterioration of satisfaction with dental appearance” indicate that whites were more satisfied with their dental appearance and whites were also less likely to report deteriorated satisfaction with dental appearance over 24 months. These findings suggest that there was a racial difference in both satisfaction with dental appearance as well as the negative change of satisfaction with dental appearance over time after the

PAGE 134

121 prosthodontic treatments and other factors were taken into account. The levels of satisfaction with dental appearance among whites were higher and more stable over time than blacks. Results from the second-stage longitudinal models for “satisfaction with dental appearance,” “improvement of satisfaction with dental appearance,” and “deterioration of satisfaction with dental appearance” suggest that people who had at least 10 occluding pairs of teeth were more satisfied with their dental appearance, were more likely to report the improvement of dental appearance, and were less likely to report the deterioration of dental appearance. Results also show that having received finished removable prosthodontic treatment in the previous interval was significantly associated with an increased probability of reporting improved satisfaction with dental appearance in the current interval. This finding is consistent with what has been found in a study conducted by Gordon and colleagues (Gordon et al., 1988). Since fixed prosthodontics provide a more integral feeling and more natural appearance than removable prosthodontics, we hypothesized that fixed prosthodontic treatment was more valuable in terms of improving people’s satisfaction with dental appearance. However, results show that there were no significant associations. Fixed prosthodontic treatment may be very valuable in enhancing aesthetics in the anterior teeth area, but its impact on dental appearance in the posterior teeth area is not clear. Since information is not available in the FDCS dataset in terms of the number and position of the replaced teeth, the impact of fixed prosthodontic treatment could be mixed.

PAGE 135

122 Discussion of the Results for Objective 4 Using a single-item self-rating of oral health similar to the item used in the current study, Reinsine and Bailit (1980) found that age, gender, and the number of missing and decayed teeth were significantly associated with self-rated oral health. In this dissertation, bivariate cross-sectional analyses indicate there was no statistically significant association between wearing a full denture and self-rated overall oral health, while wearing a partial denture was associated with more negative ratings (“fair” or “poor”) of overall oral health. The prevalence of positive ratings (“excellent,” “very good,” or “good”) was significantly higher among older people, male, whites, high school graduates, people in good general health, and regular dental attenders. People whose family income was not below 150% poverty level, who were able to pay an unexpected $500 dental bill, and who had a certain type of dental insurance rated their overall oral health status more positively than their counterparts. People who had more than 10 occluding pairs of teeth and who did not repor t some problems of dental disease/tissue damage had a higher prevalence of positive ratings. However, after controlling for other factors in the cross-sectional logistic regression, neither wearing a full denture nor wearing a partial denture exerted a significant impact on the ratings of overall oral health. Age, gender, race, ability to pay, and dental insurance were no longer significantly associated with overall oral health as they were in the bivariate comparisons. Having more than 10 occluding pairs of teeth was a significant predictor of better ratings of oral health, and having a problem with a sore denture, a broken filling, a toothache, bleeding gums, stained teeth, and a problem with bad breath accounted for a higher prevalence in negative ratings of overall oral health.

PAGE 136

123 Results of the longitudinal analyses show that higher ratings of self-rated overall oral health at the preceding interval were associated with a lower likelihood of reporting improved overall oral health but a higher likelihood of reporting deteriorated overall oral health at the current interval. Again, as previously mentioned, ceiling and flooring effects should be considered to interpret the results. Results from the second-stage longitudinal models for “self-rated overall oral health” and “deterioration of self-rated overall oral health” indicate that whites rated their overall oral health higher than blacks and whites were also less likely to report deteriorated overall oral health. The findings suggest that there is a racial difference in both self-rated overall oral health as well as the negative change in self-rated oral health over time after the prosthodontic treatments and other factors were taken into account. The ratings of overall oral health among whites were higher and more stable over time than blacks. A similar finding was also found for people with more than 10 occluding pairs of teeth: they rated their overall oral health higher than people with fewer than 10 occluding pairs of teeth and they were less likely to report deteriorated overall oral health. For the same abovementioned reason, for people with more than 10 occluding pairs of teeth, the ratings of self-rated oral health were high and stable over time, thus were less likely to worsen.

PAGE 137

124 CHAPTER 7 CONCLUSIONS AND LIMITATIONS Conclusions There is a lack of consensus in the literature with regard to the value of prosthodontic treatment. Previous studies have shown both positive and negative consequences. A general conclusion that can be drawn from this dissertation is removable prosthodontic treatment acts in a curative manner on multiple aspects of oral health once this treatment is finished. The finished removable prosthodontic treatment improves an individualÂ’s satisfaction with chewing ability and satisfaction with dental appearance. The findings from this dissertation provide positive empirical evidence on the value of removable prosthodontic treatment. Some studies have shown that compared to removable prosthodontic treatment, fixed prosthodontic treatment provides better oral esthetics, better oral comfort and a healthier oral environment (Budtz-Jorgensen and Isidor, 1990; Trulsson et al., 2002). However, th e results of this dissertation do not show any beneficial effect of fixed prosthodontic treatment. The results of this dissertation also have shown that fixed prosthodontic treatment is the most common type of prosthodontic treatment. However, the findings of this dissertation have documented the therapeutic effectiveness of removable prosthodontic treatment but not fixed prosthodontic treatment . If a less expensive treatment can bring the same or even better results as a more expensive treatment, switching to the less expensive treatment will result in a noticeable saving in health care expenditure without compromising health outcomes. Many clinicians have negative opinions toward

PAGE 138

125 removable prosthodontic treatment in terms of functional efficiency, esthetics, comfort, retention, and oral hygiene (Frank et al., 2000). Weintraub and Weintraub (1997) also reported reduced removable prosthodontics curriculum hours in many dental schools, which may result in a decreased overall quality of removable prosthodontics. However, this type of treatment still has high therapeutic and cost-effective value as shown in this dissertation. Limitations Several methodological limitations of this dissertation warrant discussion. At baseline, 873 subjects participated the study. By the end of the 24-month data collection period, 764 (unweighted) persons remained in the study. As has mentioned in Chapter 4, subjects who remained in the study were more likely to have been well-educated, regular dental attenders, above the 100% poverty threshold, in better self-rated general health, White, and had more teeth present at baseline. Because information about the use of prosthodontic services and oral health outcomes is not available for non-participants, it is difficulty to evaluate the possible bias regarding the effectiveness of the prosthodontic treatment due to subject attrition. Another limitation is information about the number and location of the replaced teeth is not available from the Florida Dental Care Study and therefore is not included in analyses in this dissertation. A lack of such information may confound the conclusions drawn from this dissertation regarding the prosthodontic treatment effects. Owall (1986) indicated that the number, location and the relative importance of the missing teeth should be taken into consideration when a prosthodontic treatment is analyzed. For example, many researchers (Carlsson, 1984; Käyser et al., 1990; Witter et al., 1999) maintained the presence of incisors, canines, and premolars, for a total of 20 properly

PAGE 139

126 distributed teeth, were sufficient to fulfill the functional requirements. Previous studies showed that there was little social and functional gain by replacing missing teeth with prostheses unless a large number of teeth or all teeth were lost (Agerberg and Carlsson, 1981; Leake et al., 1994; Slagter et al., 1992). Spartely (1988) indicated that unreplaced posterior missing teeth are highly tolerated by patients because these teeth are not aesthetically sensitive (Spartley, 1988). Frank et al. (1998) also hypothesized that the replacement of a larger number of missing teeth and anterior teeth would be associated with higher satisfaction with the prostheses, but their results did not support their hypothesis. Therefore, the replacement of anterior teeth with a prosthesis may have different functional or socio-psychological impacts with the replacement of posterior teeth. In this dissertation, upper complete denture, lower complete denture, upper removable partial denture, and lower removable partial denture are pooled together as “removable prosthodontic treatment” due to the low incidence rate of each individual type. Previous studies have shown that patients’ satisfaction varies depending on the types of the removable prostheses. Some researchers (Bell, 1972; Loupe et al., 1988) indicated that even if the technical quality of a complete denture was satisfactory, people still felt frustrated in wearing it. However, elebi and Knexovi -Zlatari (2003) noted that complete denture wearers were significantly more satisfied with chewing, speech, and retention than removable partial denture wearers. They also found patients were more satisfied with the retention and the comfort of wearing lower removable partial dentures than wearing upper removable partial dentures ( elebi and Knexovi -Zlatari ,

PAGE 140

127 2003). Therefore, the effects of various type s of removable prosthodontic services could be mixed in this dissertation. Another limitation is the incidence rates of unfinished fixed prosthodontic treatment and unfinished removable prosthodontic treatment are very low, and no valid instrumental variables are found for these two types of prosthodontic treatment. Thus they are not included in the second-stage analyses. Therefore, information regarding the effect of unfinished versus finished pr osthodontic treatment is not available.

PAGE 141

128 APPENDIX A CORRESPONDING QUESTIONNAIRE ITEMS Dependent Variables Self-Rated Oral Health How satisfied are you with your ability to chew overall? 1. very satisfied 2. satisfied 3. dissatisfied 4. very dissatisfied 8. don't know How satisfied are you with the appearance of your teeth and/or dentures? 1. very satisfied 2. satisfied 3. dissatisfied 4. very dissatisfied 8. don't know Compared to others your age, how would you rate the health of your mouth overall? Would you say the health of your mouth overall is? 1. Excellent 2. Very Good 3. Good 4. Fair 5. Poor Independent Variables Prosthodontic Services Please tell me if you had any of these procedures done at that visit. Dental cap or implant made/fixed Partial denture made or repaired Full denture made or repaired

PAGE 142

129 Selected Population Characteristics Predisposing factors Age Gender Area of residence Race What is the highest level of formal schooling you have completed? 1. Less than eighth grade 2. Completed eighth grade 3. Some high school 4. High school graduate 5. Some college 6. Completed a 2-year college degree 7. Completed a 4-year college degree 8. Post graduate degree Compared to others your age, how would you rate your general health? Would you say your general health is? 1. Excellent 2. Very Good 3. Good 4. Fair 5. Poor Which of the following statements best describes your approach to dental care? 1. I never go to a dentist. 2. I go to a dentist when I have a problem or when I know that I need to get something fixed. 3. I go to a dentist occasionally, whether or not I have a problem. 4. I go to a dentist regularly. Have you ever had dental treatment that has not worked, or dental treatment that has not lasted as long as you thought it should have? 1. yes =====> 49a. How does this make you feel about dental care now? 2. no 1. extremely frustrated 2. very frustrated 3. moderately frustrated 4. a little frustrated 5. not at all frustrated

PAGE 143

130 Enabling factors If you were faced with an unexpected $500 de ntal bill, how would you best describe your situation? Would you be ... ? 1. Able to pay comfortably ====> 73a. Is that because you have dental insurance or because you would be able to pay comfortably even if you did not have dental insurance? 1. Because of dental insurance 2. Even if you did not have dental insurance 8. Don't know 9. Refused 2. Able to pay, but with difficulty 3. Not able to pay the bill Some people have dental insurance that pays for part of their dental bills, such as from an employer, Medicaid, or the VA. Are you covered by any such dental insurance program? 1. yes ====> IF YES: Is it covered by...? 2. no 1. Your employer 2. Medicaid 3. Department of Veterans Affairs (VA) 4. Other [SPECIFY]: ________________________________ ====> What does it cover? [CIRCLE ALL THAT APPLY] 1. pulling teeth 2. fillings 3. caps 4. root canals 5. dentures 6. other ________________ 8. DK Would you say that your household's total annual income before taxes is under or over $20,000? [IF INCOME IS EXACTLY AT A CATEGORY DIVISION, PLACE IN THE HIGHER CATEGORY] 1. Under $20,000 2. Over $20,000 [GO TO QUESTION #78] 9. Refused [GO TO QUESTION #79] How long does it usually take you to get to this dentist or clinic? 1. Under 10 minutes 2. 10-20 minutes 3. 21-30 minutes 4. 31-40 minutes 5. 41-50 minutes

PAGE 144

131 6. 51 minutes 1 hour 7. over 1 hour Is the dentist or clinic you mentioned the one that is nearest you? 1. Yes [IF YES, GO TO #10] 2. No 8. Don't know Need factors Do you think you need to see a dentist now or in the next couple of weeks? 1. yes =====> Is that... a. for a routine check-up b. for a dental problem ===> What problem? Specify _____________________________________ 2. no ======> Is that... c. because, although you have a dental problem, it can wait ===> What problem? specify __________________ d. because your mouth is in good shape now e. because you feel you don't ever need to see a dentist I need to know what the reason for the visit was. Please tell me if it was for any of these reasons. More than one answer is possible. Cap or bridge was loose Sore denture Denture broke Broken filling Broken tooth or cap Dental cavities Abscessed tooth Infected or sore gums Bleeding gums Tooth was loose Teeth looked bad Bad breath Have you lost any teeth or had any teet h removed since we visited you about 6 months ago? 1. Yes 2. No

PAGE 145

132 APPENDIX B THE SAS PROGRAM FOR INSTRUMENTAL VARIABLE APPROACH /******************************************************* ** ** ** INSTRUMENTAL VARIABLE APPROACH ** ** (UPDATED ON JUNE 28, 2005) ** ** ** *******************************************************/ /* INSTRUMENTAL VARIABLE APPROACH FOR XIAOXIAN'S DISSERTATION */ /* CORRECT SELECTION BIAS FOR TWO TREATMENTS -FINFIXED & FINDENTUR*/ /* EXCLUDE TWO TREATMENTS BECAUSE OF LOW INCIDENCE RATES -BOTHFIXED & BOTHDENTUR */ OPTIONS PS= 70 LS= 120 ; LIBNAME H 'C:\Documents and Settings\xmeng\My Documents\FDCS DATA\WORKING'; DATA ONE; SET H.DISSERTATION; RUN ; /****************************************************** ****** FIND VALID INSTRUMENTAL VARIABLES ******* *******************************************************/ /*TEST "ASSOCIATIONS" BETWEEN POTENTIAL IVs AND FINFIXED*/ PROC GENMOD DATA=ONE DESCENDING; WEIGHT NML_WT; CLASS NAME2 INTERVAL t1q8 ABLE500 T1Q74; MODEL FINFIXED=PNEED3 t1q8 APPROACH RR_FRUST ABLE500 T1Q74 /LINK = LOGIT D=BINOMIAL; REPEATED SUBJECT=NAME2/WITHIN=INTERVAL TYPE=AR CORRW; TITLE 'FIND IVs (I) -TESTING FOR ASSOCIATIONS BETWEEN IVs AND FINFIXED'; RUN ; /*TEST "ASSOCIATIONS" BETWEEN POTENTIAL IVs AND FINDENTUR*/ PROC GENMOD DATA=ONE DESCENDING; WEIGHT NML_WT; CLASS NAME2 INTERVAL t1q8 ABLE500 T1Q74; MODEL FINDENTUR=PNEED3 t1q8 APPROACH RR_FRUST ABLE500 T1Q74 /LINK = LOGIT D=BINOMIAL; REPEATED SUBJECT=NAME2/WITHIN=INTERVAL TYPE=AR CORRW; TITLE 'FIND IVs (I) -TESTING FOR ASSOCIATIONS BETWEEN IVs AND FINDENTUR'; RUN ;

PAGE 146

133 /*TEST "ASSOCIATIONS" BETWEEN POTENTIAL IVs AND BOTHFIXED*/ /* RESULTS SHOW NO GOOD IVs */ PROC GENMOD DATA=ONE DESCENDING; WEIGHT NML_WT; CLASS NAME2 INTERVAL t1q8 ABLE500 T1Q74; MODEL BOTHFIXED=PNEED3 t1q8 APPROACH RR_FRUST ABLE500 T1Q74 /LINK = LOGIT D=BINOMIAL; REPEATED SUBJECT=NAME2/WITHIN=INTERVAL TYPE=AR CORRW; TITLE 'FIND IVs (I) -TESTING FOR ASSOCIATIONS BETWEEN IVs AND BOTHFIXED'; RUN ; /*TEST "ASSOCIATIONS" BETWEEN POTENTIAL IVs AND BOTHDENTUR*/ /* RESULTS SHOW NO GOOD IVs */ PROC GENMOD DATA=ONE DESCENDING; WEIGHT NML_WT; CLASS NAME2 INTERVAL t1q8 ABLE500 T1Q74; MODEL BOTHDENTUR=PNEED3 t1q8 APPROACH RR_FRUST ABLE500 T1Q74 /LINK = LOGIT D=BINOMIAL; REPEATED SUBJECT=NAME2/WITHIN=INTERVAL TYPE=AR CORRW; TITLE 'FIND IVs (I) -TESTING FOR ASSOCIATIONS BETWEEN IVs AND BOTHDENTUR'; RUN ; /*TEST "LACK OF ASSOCIATIONS" BETWEEN POTENTIAL IVs AND OUTCOMES AFTER CONTROLLING FOR FINFIXED & FINDENTUR*/ %LET RESPS = IMPROSAT1 DETERSAT1 IMPROSAT2 DETERSAT2 IMPRORATE DETERRATE SATIS1_D SATIS2_D RATE_D; %MACRO SSM ; %DO M= 1 %TO 9 ; %LET RESPNS = %SCAN (&RESPS, &M); %LET CVAR=NAME2 INTERVAL t1q8 ABLE500; PROC GENMOD DATA = ONE DESCENDING; WEIGHT NML_WT; CLASS &CVAR; MODEL &RESPNS = PNEED3 t1q8 APPROACH ABLE500 RR_FRUST FINFIXED FINDENTUR / LINK = LOGIT DIST = BINOMIAL; REPEATED SUBJECT=NAME2/WITHIN=INTERVAL TYPE=AR CORRW; TITLE 'FIND IVs (II)-TESTING FOR NON-ASSOCIATIONS BETWEEN IVs AND OUTCOMES'; RUN; %END; %MEND ; % SSM ; /**************************************************************** ****** IV FIRST STAGE -CALCULATE THE PREDICTED VALUES ****** ****** OF TREATMENTS (FINFIXEDHAT, FINDENTUR, BOTHFIXED ****** ****** BOTHDEUTUR) ****** ****************************************************************/ /* CALCULATE PREDICATED VALUES */

PAGE 147

134 PROC GENMOD DATA=ONE DESCENDING; WEIGHT NML_WT; CLASS NAME2 INTERVAL LOST t1q8 GENDER RACE ABLE500 T1Q74 BROKFILL BROKTOOT CAVIT STAINED; MODEL FINFIXED=PNEED3 PROBPROS LOST t1q8 AGEGROUP GENDER RACE APPROACH RR_FRUST ABLE500 T1Q74 BROKFILL BROKTOOT CAVIT LOOSE2 STAINED /LINK = LOGIT D=BINOMIAL; REPEATED SUBJECT=NAME2/WITHIN=INTERVAL TYPE=AR CORRW; OUTPUT OUT =FIRSTFFOUT XBETA = ETA PRED = FINFIXEDHAT RESRAW = RESID1; TITLE 'IV STAGE ONE -USING LOGIT LINK TO GET THE PREDICTED VALUE OF FINFIEXED FINFIXEDHAT)'; RUN ; PROC GENMOD DATA=ONE DESCENDING; WEIGHT NML_WT; CLASS NAME2 INTERVAL LOST t1q8 GENDER RACE ABLE500 T1Q74 BROKFILL BROKTOOT CAVIT STAINED; MODEL FINDENTUR=PNEED3 PROBPROS LOST t1q8 AGEGROUP GENDER RACE APPROACH RR_FRUST ABLE500 T1Q74 BROKFILL BROKTOOT CAVIT LOOSE2 STAINED /LINK = LOGIT D=BINOMIAL; REPEATED SUBJECT=NAME2/WITHIN=INTERVAL TYPE=AR CORRW; OUTPUT OUT =FIRSTFDOUT XBETA = ETA PRED = FINDENTURHAT RESRAW = RESID2; TITLE 'IV STAGE ONE -USING LOGIT LINK TO GET THE PREDICTED VALUE OF FINDENTUR (FINDENTURHAT)'; RUN ; PROC GENMOD DATA=ONE DESCENDING; WEIGHT NML_WT; CLASS NAME2 INTERVAL LOST t1q8 GENDER RACE ABLE500 T1Q74 BROKFILL BROKTOOT CAVIT STAINED; MODEL BOTHFIXED=PNEED3 PROBPROS LOST t1q8 AGEGROUP GENDER RACE APPROACH RR_FRUST ABLE500 T1Q74 BROKFILL BROKTOOT CAVIT LOOSE2 STAINED /LINK = LOGIT D=BINOMIAL; REPEATED SUBJECT=NAME2/WITHIN=INTERVAL TYPE=AR CORRW; OUTPUT OUT =FIRSTBFOUT XBETA = ETA PRED = BOTHFIXEDHAT RESRAW = RESID3; TITLE 'IV STAGE ONE -USING LOGIT LINK TO GET THE PREDICTED VALUE OF BOTHFIEXED (BOTHFIXEDHAT)'; RUN ; PROC GENMOD DATA=ONE DESCENDING; WEIGHT NML_WT; CLASS NAME2 INTERVAL LOST t1q8 GENDER RACE ABLE500 T1Q74 BROKFILL BROKTOOT CAVIT STAINED; MODEL BOTHDENTUR=PNEED3 PROBPROS LOST t1q8 AGEGROUP GENDER RACE APPROACH RR_FRUST ABLE500 T1Q74 BROKFILL BROKTOOT CAVIT LOOSE2 STAINED /LINK = LOGIT D=BINOMIAL; REPEATED SUBJECT=NAME2/WITHIN=INTERVAL TYPE=AR CORRW; OUTPUT OUT =FIRSTBDOUT XBETA = ETA PRED = BOTHDENTURHAT RESRAW = RESID4; TITLE 'IV STAGE ONE -USING LOGIT LINK TO GET THE PREDICTED VALUE OF BOTHDENTUR (BOTHDENTURHAT)'; RUN ; PROC SORT DATA=ONE; BY NAME2 INTERVAL; RUN ;

PAGE 148

135 PROC SORT DATA=FIRSTFFOUT; BY NAME2 INTERVAL; RUN ; PROC SORT DATA=FIRSTFDOUT; BY NAME2 INTERVAL; RUN ; PROC SORT DATA=FIRSTBFOUT; BY NAME2 INTERVAL; RUN ; PROC SORT DATA=FIRSTBDOUT; BY NAME2 INTERVAL; RUN ; DATA TWO; MERGE ONE FIRSTFFOUT(KEEP=NAME2 INTERVAL FINFIXEDHAT RESID1) FIRSTFDOUT(KEEP=NAME2 INTERVAL FINDENTURHAT RESID2) FIRSTBFOUT(KEEP=NAME2 INTERVAL BOTHFIXEDHAT RESID3) FIRSTBDOUT(KEEP=NAME2 INTERVAL BOTHDENTURHAT RESID4); BY NAME2 INTERVAL; IF FINFIXEDHAT< 0.5 THEN FINFIXEDHAT_D = 0 ; IF FINFIXEDHAT>= 0.5 THEN FINFIXEDHAT_D = 1 ; IF FINFIXEDHAT= . THEN FINFIXEDHAT_D = . ; IF FINDENTURHAT< 0.5 THEN FINDENTURHAT_D = 0 ; IF FINDENTURHAT>= 0.5 THEN FINDENTURHAT_D = 1 ; IF FINDENTURHAT= . THEN FINDENTURHAT_D= . ; IF BOTHFIXEDHAT< 0.5 THEN BOTHFIXEDHAT_D = 0 ; IF BOTHFIXEDHAT>= 0.5 THEN BOTHFIXEDHAT_D = 1 ; IF BOTHFIXEDHAT= . THEN BOTHFIXEDHAT_D = . ; IF BOTHDENTURHAT< 0.5 THEN BOTHDENTURHAT_D = 0 ; IF BOTHDENTURHAT>= 0.5 THEN BOTHDENTURHAT_D = 1 ; IF BOTHDENTURHAT= . THEN BOTHDENTURHAT_D= . ; RUN ; DATA H.DISSERTATION; SET TWO; RUN ; /* CALCULATE DEL TO COMPARE PREDICATED VALUES AND TRUE VALUES */ LIBNAME M 'C:\Documents and Settings\xmeng\My Documents\FDCS DATA\WORKING'; DATA THREE; SET M.DISSERTATION; RUN ; PROC FREQ DATA=THREE; TABLES FINFIXED*FINFIXEDHAT_D FINDENTUR*FINDENTURHAT_D BOTHFIXED*BOTHFIXEDHAT_D BOTHDENTUR*BOTHDENTURHAT_D/NOROW NOPCT; RUN ;

PAGE 149

136 /***************************************************************** ****** IV SECOND STAGE -CORRECT SELECTION BIAS FOR BOTH ****** ****** FINFIXED & FINDENTUR (FINFIXEDHAT & FINDENTURHAT) ****** ******************************************************************/ LIBNAME M 'C:\Documents and Settings\xmeng\My Documents\FDCS DATA\WORKING'; DATA THREE; SET M.DISSERTATION; RUN ; /* IV STAGE TWO MODELS (DYNAMIC OUTCOME VARIABLES) */ PROC GENMOD DATA=THREE DESCENDING; WEIGHT NML_WT; CLASS NAME2 INTERVAL Tx_1SAT1_D GENDER RACE LOST; MODEL IMPROSAT1=Tx_1SAT1_D AGEGROUP GENDER RACE LOST DN_PRS10 FINFIXEDHAT FINDENTURHAT/LINK=LOGIT DIST=BINOMIAL; REPEATED SUBJECT=NAME2/WITHIN=INTERVAL TYPE=AR CORRW; RUN ; PROC GENMOD DATA=THREE DESCENDING; WEIGHT NML_WT; CLASS NAME2 INTERVAL Tx_1SAT1_D GENDER RACE LOST; MODEL DETERSAT1=Tx_1SAT1_D AGEGROUP GENDER RACE LOST DN_PRS10 FINFIXEDHAT FINDENTURHAT/LINK=LOGIT DIST=BINOMIAL; REPEATED SUBJECT=NAME2/WITHIN=INTERVAL TYPE=AR CORRW; RUN ; PROC GENMOD DATA=THREE DESCENDING; WEIGHT NML_WT; CLASS NAME2 INTERVAL Tx_1SAT2_D GENDER RACE LOST; MODEL IMPROSAT2=Tx_1SAT2_D AGEGROUP GENDER RACE LOST DN_PRS10 FINFIXEDHAT FINDENTURHAT/LINK=LOGIT DIST=BINOMIAL; REPEATED SUBJECT=NAME2/WITHIN=INTERVAL TYPE=AR CORRW; RUN ; PROC GENMOD DATA=THREE DESCENDING; WEIGHT NML_WT; CLASS NAME2 INTERVAL Tx_1SAT2_D GENDER RACE LOST; MODEL DETERSAT2=Tx_1SAT2_D AGEGROUP GENDER RACE LOST DN_PRS10 FINFIXEDHAT FINDENTURHAT/LINK=LOGIT DIST=BINOMIAL; REPEATED SUBJECT=NAME2/WITHIN=INTERVAL TYPE=AR CORRW; RUN ; PROC GENMOD DATA=THREE DESCENDING; WEIGHT NML_WT; CLASS NAME2 INTERVAL Tx_1RATE_D GENDER RACE LOST; MODEL IMPRORATE=Tx_1RATE_D AGEGROUP GENDER RACE LOST DN_PRS10 FINFIXEDHAT FINDENTURHAT/LINK=LOGIT DIST=BINOMIAL; REPEATED SUBJECT=NAME2/WITHIN=INTERVAL TYPE=AR CORRW; RUN ; PROC GENMOD DATA=THREE DESCENDING; WEIGHT NML_WT; CLASS NAME2 INTERVAL Tx_1RATE_D GENDER RACE LOST;

PAGE 150

137 MODEL DETERRATE=Tx_1RATE_D AGEGROUP GENDER RACE LOST DN_PRS10 FINFIXEDHAT FINDENTURHAT/LINK=LOGIT DIST=BINOMIAL; REPEATED SUBJECT=NAME2/WITHIN=INTERVAL TYPE=AR CORRW; RUN ; /* IV STAGE TWO MODELS (ORIGINAL OUTCOME VARIABLES BUT DICHOTOMIZED)*/ PROC GENMOD DATA=THREE DESCENDING; WEIGHT NML_WT; CLASS NAME2 INTERVAL GENDER RACE LOST; MODEL SATIS1_D=AGEGROUP GENDER RACE LOST DN_PRS10 FINFIXEDHAT FINDENTURHAT /LINK=LOGIT DIST=BINOMIAL; REPEATED SUBJECT=NAME2/WITHIN=INTERVAL TYPE=AR CORRW; RUN ; PROC GENMOD DATA=THREE DESCENDING; WEIGHT NML_WT; CLASS NAME2 INTERVAL GENDER RACE LOST; MODEL SATIS2_D=AGEGROUP GENDER RACE LOST DN_PRS10 FINFIXEDHAT FINDENTURHAT /LINK=LOGIT DIST=BINOMIAL; REPEATED SUBJECT=NAME2/WITHIN=INTERVAL TYPE=AR CORRW; RUN ; PROC GENMOD DATA=THREE DESCENDING; WEIGHT NML_WT; CLASS NAME2 INTERVAL GENDER RACE LOST; MODEL RATE_D=AGEGROUP GENDER RACE LOST DN_PRS10 FINFIXEDHAT FINDENTURHAT /LINK=LOGIT DIST=BINOMIAL; REPEATED SUBJECT=NAME2/WITHIN=INTERVAL TYPE=AR CORRW; RUN ; /***************************************************************** ****** SENSITIVITY TEST (I) -CORRECT SELECTION BIAS ****** ****** ONLY FOR FINFIXED (FINFIXEDHAT) ****** ******************************************************************/ LIBNAME M 'C:\Documents and Settings\xmeng\My Documents\FDCS DATA\WORKING'; DATA THREE; SET M.DISSERTATION; RUN ; /* IV STAGE TWO MODELS (DYNAMIC OUTCOME VARIABLES) */ PROC GENMOD DATA=THREE DESCENDING; WEIGHT NML_WT; CLASS NAME2 INTERVAL Tx_1SAT1_D GENDER RACE LOST; MODEL IMPROSAT1=Tx_1SAT1_D AGEGROUP GENDER RACE LOST DN_PRS10 FINFIXEDHAT FINDENTUR/LINK=LOGIT DIST=BINOMIAL; REPEATED SUBJECT=NAME2/WITHIN=INTERVAL TYPE=AR CORRW; RUN ; PROC GENMOD DATA=THREE DESCENDING; WEIGHT NML_WT; CLASS NAME2 INTERVAL Tx_1SAT1_D GENDER RACE LOST; MODEL DETERSAT1=Tx_1SAT1_D AGEGROUP GENDER RACE LOST DN_PRS10 FINFIXEDHAT FINDENTUR/LINK=LOGIT DIST=BINOMIAL;

PAGE 151

138 REPEATED SUBJECT=NAME2/WITHIN=INTERVAL TYPE=AR CORRW; RUN ; PROC GENMOD DATA=THREE DESCENDING; WEIGHT NML_WT; CLASS NAME2 INTERVAL Tx_1SAT2_D GENDER RACE LOST; MODEL IMPROSAT2=Tx_1SAT2_D AGEGROUP GENDER RACE LOST DN_PRS10 FINFIXEDHAT FINDENTUR/LINK=LOGIT DIST=BINOMIAL; REPEATED SUBJECT=NAME2/WITHIN=INTERVAL TYPE=AR CORRW; RUN ; PROC GENMOD DATA=THREE DESCENDING; WEIGHT NML_WT; CLASS NAME2 INTERVAL Tx_1SAT2_D GENDER RACE LOST; MODEL DETERSAT2=Tx_1SAT2_D AGEGROUP GENDER RACE LOST DN_PRS10 FINFIXEDHAT FINDENTUR/LINK=LOGIT DIST=BINOMIAL; REPEATED SUBJECT=NAME2/WITHIN=INTERVAL TYPE=AR CORRW; RUN ; PROC GENMOD DATA=THREE DESCENDING; WEIGHT NML_WT; CLASS NAME2 INTERVAL Tx_1RATE_D GENDER RACE LOST; MODEL IMPRORATE=Tx_1RATE_D AGEGROUP GENDER RACE LOST DN_PRS10 FINFIXEDHAT FINDENTUR/LINK=LOGIT DIST=BINOMIAL; REPEATED SUBJECT=NAME2/WITHIN=INTERVAL TYPE=AR CORRW; RUN ; PROC GENMOD DATA=THREE DESCENDING; WEIGHT NML_WT; CLASS NAME2 INTERVAL Tx_1RATE_D GENDER RACE LOST; MODEL DETERRATE=Tx_1RATE_D AGEGROUP GENDER RACE LOST DN_PRS10 FINFIXEDHAT FINDENTUR/LINK=LOGIT DIST=BINOMIAL; REPEATED SUBJECT=NAME2/WITHIN=INTERVAL TYPE=AR CORRW; RUN ; /* IV STAGE TWO MODELS (ORIGINAL OUTCOME VARIABLES BUT DICHOTOMIZED)*/ PROC GENMOD DATA=THREE DESCENDING; WEIGHT NML_WT; CLASS NAME2 INTERVAL GENDER RACE LOST; MODEL SATIS1_D=AGEGROUP GENDER RACE LOST DN_PRS10 FINFIXEDHAT FINDENTUR /LINK=LOGIT DIST=BINOMIAL; REPEATED SUBJECT=NAME2/WITHIN=INTERVAL TYPE=AR CORRW; RUN ; PROC GENMOD DATA=THREE DESCENDING; WEIGHT NML_WT; CLASS NAME2 INTERVAL GENDER RACE LOST; MODEL SATIS2_D=AGEGROUP GENDER RACE LOST DN_PRS10 FINFIXEDHAT FINDENTUR /LINK=LOGIT DIST=BINOMIAL; REPEATED SUBJECT=NAME2/WITHIN=INTERVAL TYPE=AR CORRW; RUN ; PROC GENMOD DATA=THREE DESCENDING; WEIGHT NML_WT; CLASS NAME2 INTERVAL GENDER RACE LOST; MODEL RATE_D=AGEGROUP GENDER RACE LOST DN_PRS10 FINFIXEDHAT FINDENTUR /LINK=LOGIT DIST=BINOMIAL;

PAGE 152

139 REPEATED SUBJECT=NAME2/WITHIN=INTERVAL TYPE=AR CORRW; RUN ; /***************************************************************** ****** SENSITIVITY TEST (II) -CORRECT SELECTION BIAS ****** ****** ONLY FOR FINDENTUR (FINDENTURHAT) ****** ******************************************************************/ LIBNAME M 'C:\Documents and Settings\xmeng\My Documents\FDCS DATA\WORKING'; DATA THREE; SET M.DISSERTATION; RUN ; /* IV STAGE TWO MODELS (DYNAMIC OUTCOME VARIABLES) */ PROC GENMOD DATA=THREE DESCENDING; WEIGHT NML_WT; CLASS NAME2 INTERVAL Tx_1SAT1_D GENDER RACE LOST; MODEL IMPROSAT1=Tx_1SAT1_D AGEGROUP GENDER RACE LOST DN_PRS10 FINFIXED FINDENTURHAT/LINK=LOGIT DIST=BINOMIAL; REPEATED SUBJECT=NAME2/WITHIN=INTERVAL TYPE=AR CORRW; RUN ; PROC GENMOD DATA=THREE DESCENDING; WEIGHT NML_WT; CLASS NAME2 INTERVAL Tx_1SAT1_D GENDER RACE LOST; MODEL DETERSAT1=Tx_1SAT1_D AGEGROUP GENDER RACE LOST DN_PRS10 FINFIXED FINDENTURHAT/LINK=LOGIT DIST=BINOMIAL; REPEATED SUBJECT=NAME2/WITHIN=INTERVAL TYPE=AR CORRW; RUN ; PROC GENMOD DATA=THREE DESCENDING; WEIGHT NML_WT; CLASS NAME2 INTERVAL Tx_1SAT2_D GENDER RACE LOST; MODEL IMPROSAT2=Tx_1SAT2_D AGEGROUP GENDER RACE LOST DN_PRS10 FINFIXED FINDENTURHAT/LINK=LOGIT DIST=BINOMIAL; REPEATED SUBJECT=NAME2/WITHIN=INTERVAL TYPE=AR CORRW; RUN ; PROC GENMOD DATA=THREE DESCENDING; WEIGHT NML_WT; CLASS NAME2 INTERVAL Tx_1SAT2_D GENDER RACE LOST; MODEL DETERSAT2=Tx_1SAT2_D AGEGROUP GENDER RACE LOST DN_PRS10 FINFIXED FINDENTURHAT/LINK=LOGIT DIST=BINOMIAL; REPEATED SUBJECT=NAME2/WITHIN=INTERVAL TYPE=AR CORRW; RUN ; PROC GENMOD DATA=THREE DESCENDING; WEIGHT NML_WT; CLASS NAME2 INTERVAL Tx_1RATE_D GENDER RACE LOST; MODEL IMPRORATE=Tx_1RATE_D AGEGROUP GENDER RACE LOST DN_PRS10 FINFIXED FINDENTURHAT/LINK=LOGIT DIST=BINOMIAL; REPEATED SUBJECT=NAME2/WITHIN=INTERVAL TYPE=AR CORRW; RUN ; PROC GENMOD DATA=THREE DESCENDING;

PAGE 153

140 WEIGHT NML_WT; CLASS NAME2 INTERVAL Tx_1RATE_D GENDER RACE LOST; MODEL DETERRATE=Tx_1RATE_D AGEGROUP GENDER RACE LOST DN_PRS10 FINFIXED FINDENTURHAT/LINK=LOGIT DIST=BINOMIAL; REPEATED SUBJECT=NAME2/WITHIN=INTERVAL TYPE=AR CORRW; RUN ; /* IV STAGE TWO MODELS (ORIGINAL OUTCOME VARIABLES BUT DICHOTOMIZED)*/ PROC GENMOD DATA=THREE DESCENDING; WEIGHT NML_WT; CLASS NAME2 INTERVAL GENDER RACE LOST; MODEL SATIS1_D=AGEGROUP GENDER RACE LOST DN_PRS10 FINFIXED FINDENTURHAT /LINK=LOGIT DIST=BINOMIAL; REPEATED SUBJECT=NAME2/WITHIN=INTERVAL TYPE=AR CORRW; RUN ; PROC GENMOD DATA=THREE DESCENDING; WEIGHT NML_WT; CLASS NAME2 INTERVAL GENDER RACE LOST; MODEL SATIS2_D=AGEGROUP GENDER RACE LOST DN_PRS10 FINFIXED FINDENTURHAT /LINK=LOGIT DIST=BINOMIAL; REPEATED SUBJECT=NAME2/WITHIN=INTERVAL TYPE=AR CORRW; RUN ; PROC GENMOD DATA=THREE DESCENDING; WEIGHT NML_WT; CLASS NAME2 INTERVAL GENDER RACE LOST; MODEL RATE_D=AGEGROUP GENDER RACE LOST DN_PRS10 FINFIXED FINDENTURHAT /LINK=LOGIT DIST=BINOMIAL; REPEATED SUBJECT=NAME2/WITHIN=INTERVAL TYPE=AR CORRW; RUN ; /************************************************************ ****** SENSITIVITY TEST (III) -WITHOUT CORRECTING ****** ****** SELECTION BIAS ****** *************************************************************/ LIBNAME M 'C:\Documents and Settings\xmeng\My Documents\FDCS DATA\WORKING'; DATA THREE; SET M.DISSERTATION; RUN ; /* IV STAGE TWO MODELS (DYNAMIC OUTCOME VARIABLES) */ PROC GENMOD DATA=THREE DESCENDING; WEIGHT NML_WT; CLASS NAME2 INTERVAL Tx_1SAT1_D GENDER RACE LOST; MODEL IMPROSAT1=Tx_1SAT1_D AGEGROUP GENDER RACE LOST DN_PRS10 FINFIXED FINDENTUR/LINK=LOGIT DIST=BINOMIAL; REPEATED SUBJECT=NAME2/WITHIN=INTERVAL TYPE=AR CORRW; RUN ; PROC GENMOD DATA=THREE DESCENDING; WEIGHT NML_WT; CLASS NAME2 INTERVAL Tx_1SAT1_D GENDER RACE LOST;

PAGE 154

141 MODEL DETERSAT1=Tx_1SAT1_D AGEGROUP GENDER RACE LOST DN_PRS10 FINFIXED FINDENTUR/LINK=LOGIT DIST=BINOMIAL; REPEATED SUBJECT=NAME2/WITHIN=INTERVAL TYPE=AR CORRW; RUN ; PROC GENMOD DATA=THREE DESCENDING; WEIGHT NML_WT; CLASS NAME2 INTERVAL Tx_1SAT2_D GENDER RACE LOST; MODEL IMPROSAT2=Tx_1SAT2_D AGEGROUP GENDER RACE LOST DN_PRS10 FINFIXED FINDENTUR/LINK=LOGIT DIST=BINOMIAL; REPEATED SUBJECT=NAME2/WITHIN=INTERVAL TYPE=AR CORRW; RUN ; PROC GENMOD DATA=THREE DESCENDING; WEIGHT NML_WT; CLASS NAME2 INTERVAL Tx_1SAT2_D GENDER RACE LOST; MODEL DETERSAT2=Tx_1SAT2_D AGEGROUP GENDER RACE LOST DN_PRS10 FINFIXED FINDENTUR/LINK=LOGIT DIST=BINOMIAL; REPEATED SUBJECT=NAME2/WITHIN=INTERVAL TYPE=AR CORRW; RUN ; PROC GENMOD DATA=THREE DESCENDING; WEIGHT NML_WT; CLASS NAME2 INTERVAL Tx_1RATE_D GENDER RACE LOST; MODEL IMPRORATE=Tx_1RATE_D AGEGROUP GENDER RACE LOST DN_PRS10 FINFIXED FINDENTUR/LINK=LOGIT DIST=BINOMIAL; REPEATED SUBJECT=NAME2/WITHIN=INTERVAL TYPE=AR CORRW; RUN ; PROC GENMOD DATA=THREE DESCENDING; WEIGHT NML_WT; CLASS NAME2 INTERVAL Tx_1RATE_D GENDER RACE LOST; MODEL DETERRATE=Tx_1RATE_D AGEGROUP GENDER RACE LOST DN_PRS10 FINFIXED FINDENTUR/LINK=LOGIT DIST=BINOMIAL; REPEATED SUBJECT=NAME2/WITHIN=INTERVAL TYPE=AR CORRW; RUN ; /* IV STAGE TWO MODELS (ORIGINAL OUTCOME VARIABLES BUT DICHOTOMIZED) */ PROC GENMOD DATA=THREE DESCENDING; WEIGHT NML_WT; CLASS NAME2 INTERVAL GENDER RACE LOST; MODEL SATIS1_D=AGEGROUP GENDER RACE LOST DN_PRS10 FINFIXED FINDENTUR /LINK=LOGIT DIST=BINOMIAL; REPEATED SUBJECT=NAME2/WITHIN=INTERVAL TYPE=AR CORRW; RUN ; PROC GENMOD DATA=THREE DESCENDING; WEIGHT NML_WT; CLASS NAME2 INTERVAL GENDER RACE LOST; MODEL SATIS2_D=AGEGROUP GENDER RACE LOST DN_PRS10 FINFIXED FINDENTUR /LINK=LOGIT DIST=BINOMIAL; REPEATED SUBJECT=NAME2/WITHIN=INTERVAL TYPE=AR CORRW; RUN ; PROC GENMOD DATA=THREE DESCENDING; WEIGHT NML_WT;

PAGE 155

142 CLASS NAME2 INTERVAL GENDER RACE LOST; MODEL RATE_D=AGEGROUP GENDER RACE LOST DN_PRS10 FINFIXED FINDENTUR /LINK=LOGIT DIST=BINOMIAL; REPEATED SUBJECT=NAME2/WITHIN=INTERVAL TYPE=AR CORRW; RUN ; /**************************************** **** Test "Pseudo" Goodness of Fit **** ****************************************/ PROC LOGIST ; MODEL FINFIXED=PNEED3 PROBPROS LOST t1q8 AGEGROUP GENDER RACE APPROACH RR_FRUST ABLE500 T1Q74 BROKFILL BROKTOOT CAVIT LOOSE2 STAINED/AGGREGATE SCALE=NONE; RUN ; PROC LOGIST ; MODEL FINDENTUR=PNEED3 PROBPROS LOST t1q8 AGEGROUP GENDER RACE APPROACH RR_FRUST ABLE500 T1Q74 BROKFILL BROKTOOT CAVIT LOOSE2 STAINED/AGGREGATE SCALE=NONE; RUN ; PROC LOGIST ; MODEL BOTHFIXED=PNEED3 PROBPROS LOST t1q8 AGEGROUP GENDER RACE APPROACH RR_FRUST ABLE500 T1Q74 BROKFILL BROKTOOT CAVIT LOOSE2 STAINED/AGGREGATE SCALE=NONE; RUN ; PROC LOGIST ; MODEL BOTHDENTUR=PNEED3 PROBPROS LOST t1q8 AGEGROUP GENDER RACE APPROACH RR_FRUST ABLE500 T1Q74 BROKFILL BROKTOOT CAVIT LOOSE2 STAINED/AGGREGATE SCALE=NONE; RUN ;

PAGE 156

143 APPENDIX C SENSITIVITY TESTS Table 1. Sensitivity tests for “satisfaction with chewing ability” Correct selection bias for “finished fixed prosthodontic treatment” only Correct selection bias for “finished removable prosthodontic treatment” only Without correction of selection bias Parameter Odds ratio p-value Odds ratio p-value Odds ratio p-value Finished fixed prosthodontic treatment within the current interval 1 = yes 0 = no 0.699 -0.5153 0.974 -0.9118 1.016 -0.9352 Finished removable prosthodontic treatment within the current interval 1 = yes 0 = no 0.689 -0.4145 0.758 -0.7199 1.532 -0.3275 Age 1 = 45 to 64 years old 0 = 65 years old or older 0.883 -0.5323 0.869 -0.4818 1.174 -0.3977 Gender 1 = male 2 = female 1.167 -0.4649 1.175 -0.4417 1.216 -0.3242 Race 1 = White 2 = Black 0.505 -0.0006 0.499 -0.0004 1.946 -0.0003 Tooth loss in the past six months 1 = yes 2 = no 1.406 -0.1865 1.398 -0.2221 1.466 -0.0995 Number of pairs of teeth 1 = 10 occluding pairs of teeth or more 0 = fewer than 10 occluding pairs of teeth 0.232 -<.0001 0.230 -<.0001 4.488 -<.0001

PAGE 157

144 Table 2. Sensitivity tests for “improvement of satisfaction with chewing ability” Correct selection bias for “finished fixed prosthodontic treatment” only Correct selection bias for “finished removable prosthodontic treatment” only Without correction of selection bias Parameter Odds ratio p-value Odds ratio p-value Odds ratio p-value Finished fixed prosthodontic treatment within the current interval 1 = yes 0 = no 1.336 -0.6173 1.379 -0.3173 1.475 -0.1773 Finished removable prosthodontic treatment within the current interval 1 = yes 0 = no 2.309 -0.0024 3.252 -0.1000 2.127 -0.0038 Satisfaction with chewing ability at the preceding interview (TX-1) 1 = very satisfied/satisfied 2 = very dissatisfied/dissatisfied 0.168 -<.0001 0.170 -<.0001 5.093 -<.0001 Age 1 = 45 to 64 years old 0 = 65 years old or older 0.999 -0.9944 0.992 -0.9515 1.017 -0.8872 Gender 1 = male 2 = female 0.770 -0.0629 0.775 -0.0681 1.262 -0.0675 Race 1 = White 2 = Black 0.893 -0.4023 0.899 -0.4216 1.160 -0.2114 Tooth loss in the past six months 1 = yes 2 = no 1.016 -0.9441 0.968 -0.8781 1.101 -0.6673 Number of pairs of teeth 1 = 10 occluding pairs of teeth or more 0 = fewer than 10 occluding pairs of teeth 1.134 -0.3870 1.075 -0.6148 1.126 -0.3711

PAGE 158

145 Table 3. Sensitivity tests for “deterioration of satisfaction with chewing ability” Correct selection bias for “finished fixed prosthodontic treatment” only Correct selection bias for “finished removable prosthodontic treatment” only Without correction of selection bias Parameter Odds ratio p-value Odds ratio p-value Odds ratio p-value Finished fixed prosthodontic treatment within the current interval 1 = yes 0 = no 1.074 -0.8957 0.834 -0.5660 1.190 -0.5639 Finished removable prosthodontic treatment within the current interval 1 = yes 0 = no 0.581 -0.2596 0.980 -0.9776 1.699 -0.2556 Satisfaction with chewing ability at the preceding interview (TX-1) 1 = very satisfied/satisfied 2 = very dissatisfied/dissatisfied 2.701 -<.0001 2.742 -<.0001 2.427 -<.0001 Age 1 = 45 to 64 years old 0 = 65 years old or older 1.086 -0.4873 1.088 -0.4764 1.069 -0.5469 Gender 1 = male 2 = female 0.877 -0.3059 0.875 -0.2972 1.111 -0.3767 Race 1 = White 2 = Black 0.647 -0.0006 0.651 -0.0005 1.526 -0.0002 Tooth loss in the past six months 1 = yes 2 = no 1.875 -0.0038 1.762 -0.0204 1.688 -0.0090 Number of pairs of teeth 1 = 10 occluding pairs of teeth or more 0 = fewer than 10 occluding pairs of teeth 0.533 -<.0001 0.550 -<.0001 1.759 -<.0001

PAGE 159

146 Table 4. Sensitivity tests for “satisfaction with dental appearance” Correct selection bias for “finished fixed prosthodontic treatment” only Correct selection bias for “finished removable prosthodontic treatment” only Without correction of selection bias Parameter Odds ratio p-value Odds ratio p-value Odds ratio p-value Finished fixed prosthodontic treatment within the current interval 1 = yes 0 = no 0.642 -0.4070 0.662 -0.0674 1.632 -0.0348 Finished removable prosthodontic treatment within the current interval 1 = yes 0 = no 0.493 -0.1021 0.454 -0.3411 2.224 -0.0514 Age 1 = 45 to 64 years old 0 = 65 years old or older 0.737 -0.0779 0.723 -0.0609 1.326 -0.0872 Gender 1 = male 2 = female 0.730 -0.0746 0.735 -0.0807 1.330 -0.0894 Race 1 = White 2 = Black 0.515 -0.0001 0.519 -0.0001 1.769 -0.0004 Tooth loss in the past six months 1 = yes 2 = no 0.969 -0.8907 0.998 -0.9916 1.011 -0.9561 Number of pairs of teeth 1 = 10 occluding pairs of teeth or more 0 = fewer than 10 occluding pairs of teeth 0.361 -<.0001 0.367 -<.0001 2.953 -<.0001

PAGE 160

147 Table 5. Sensitivity tests for “improvement of satisfaction with dental appearance” Correct selection bias for “finished fixed prosthodontic treatment” only Correct selection bias for “finished removable prosthodontic treatment” only Without correction of selection bias Parameter Odds ratio p-value Odds ratio p-value Odds ratio p-value Finished fixed prosthodontic treatment within the current interval 1 = yes 0 = no 0.659 -0.4489 1.553 -0.1335 1.596 -0.0661 Finished removable prosthodontic treatment within the current interval 1 = yes 0 = no 2.884 -0.0003 2.755 -0.1365 2.828 -<.0001 Satisfaction with dental appearance at the preceding interview (TX-1) 1 = very satisfied/satisfied 2 = very dissatisfied/dissatisfied 0.160 -<.0001 0.156 -<.0001 6.446 -<.0001 Age 1 = 45 to 64 years old 0 = 65 years old or older 1.031 -0.7754 1.014 -0.8989 1.054 -0.6138 Gender 1 = male 2 = female 0.956 -0.6905 0.965 -0.7538 1.111 -0.3385 Race 1 = White 2 = Black 1.151 -0.2395 1.115 -0.3514 1.078 -0.4914 Tooth loss in the past six months 1 = yes 2 = no 1.349 -0.1246 1.336 -0.1291 1.174 -0.4190 Number of pairs of teeth 1 = 10 occluding pairs of teeth or more 0 = fewer than 10 occluding pairs of teeth 1.583 -0.0003 1.471 -0.0019 1.597 -<.0001

PAGE 161

148 Table 6. Sensitivity tests for “deterioration of satisfaction with dental appearance” Correct selection bias for “finished fixed prosthodontic treatment” only Correct selection bias for “finished removable prosthodontic treatment” only Without correction of selection bias Parameter Odds ratio p-value Odds ratio p-value Odds ratio p-value Finished fixed prosthodontic treatment within the current interval 1 = yes 0 = no 0.543 -0.2762 0.554 -0.0547 1.593 -0.1165 Finished removable prosthodontic treatment within the current interval 1 = yes 0 = no 1.411 -0.4268 1.284 -0.7251 1.269 -0.5709 Satisfaction with dental appearance at the preceding interview (TX-1) 1 = very satisfied/satisfied 2 = very dissatisfied/dissatisfied 3.453 -<.0001 3.472 -<.0001 3.065 -<.0001 Age 1 = 45 to 64 years old 0 = 65 years old or older 1.088 -0.4542 1.087 -0.4558 1.105 -0.3459 Gender 1 = male 2 = female 0.908 -0.4040 0.909 -0.4080 1.115 -0.3218 Race 1 = White 2 = Black 0.773 -0.0283 0.768 -0.0208 1.208 -0.0734 Tooth loss in the past six months 1 = yes 2 = no 0.804 -0.3407 0.818 -0.4226 1.285 -0.2501 Number of pairs of teeth 1 = 10 occluding pairs of teeth or more 0 = fewer than 10 occluding pairs of teeth 0.795 -0.0610 0.792 -0.0555 1.286 -0.0302

PAGE 162

149 Table 7. Sensitivity tests for “self-rated overall oral health” Correct selection bias for “finished fixed prosthodontic treatment” only Correct selection bias for “finished removable prosthodontic treatment” only Without correction of selection bias Parameter Odds ratio p-value Odds ratio p-value Odds ratio p-value Finished fixed prosthodontic treatment within the current interval 1 = yes 0 = no 1.213 -0.7177 1.141 -0.4755 1.075 -0.7040 Finished removable prosthodontic treatment within the current interval 1 = yes 0 = no 0.409 -0.0019 0.451 -0.2052 2.520 -0.0026 Age 1 = 45 to 64 years old 0 = 65 years old or older 0.598 -0.0052 0.588 -0.0039 1.709 -0.0025 Gender 1 = male 2 = female 0.761 -0.1905 0.777 -0.2259 1.220 -0.3162 Race 1 = White 2 = Black 0.456 -<.0001 0.463 -<.0001 2.060 -<.0001 Tooth loss in the past six months 1 = yes 2 = no 1.525 -0.0269 1.482 -0.0542 1.552 -0.0140 Number of pairs of teeth 1 = 10 occluding pairs of teeth or more 0 = fewer than 10 occluding pairs of teeth 0.176 -<.0001 0.178 -<.0001 5.643 -<.0001

PAGE 163

150 Table 8. Sensitivity tests for “improvement of self-rated overall oral health” Correct selection bias for “finished fixed prosthodontic treatment” only Correct selection bias for “finished removable prosthodontic treatment” only Without correction of selection bias Parameter Odds ratio p-value Odds ratio p-value Odds ratio p-value Finished fixed prosthodontic treatment within the current interval 1 = yes 0 = no 0.548 -0.2787 1.371 -0.2028 1.359 -0.1868 Finished removable prosthodontic treatment within the current interval 1 = yes 0 = no 1.685 -0.0542 1.091 -0.8650 1.751 -0.0305 Self-rated overall oral health at the preceding interview (TX-1) 1 = excellent/very good/good 2 = fair/poor 0.310 -<.0001 0.306 -<.0001 3.310 -<.0001 Age 1 = 45 to 64 years old 0 = 65 years old or older 1.090 -0.3972 1.077 -0.4696 1.126 -0.2164 Gender 1 = male 2 = female 0.991 -0.9314 0.996 -0.9712 1.011 -0.9176 Race 1 = White 2 = Black 1.076 -0.5011 1.033 -0.7584 1.046 -0.6462 Tooth loss in the past six months 1 = yes 2 = no 1.152 -0.4839 1.215 -0.3439 1.067 -0.7561 Number of pairs of teeth 1 = 10 occluding pairs of teeth or more 0 = fewer than 10 occluding pairs of teeth 1.044 -0.7160 0.993 -0.9512 1.052 -0.6460

PAGE 164

151 Table 9. Sensitivity tests for “deterioration of self-rated overall oral health” Correct selection bias for “finished fixed prosthodontic treatment” only Correct selection bias for “finished removable prosthodontic treatment” only Without correction of selection bias Parameter Odds ratio p-value Odds ratio p-value Odds ratio p-value Finished fixed prosthodontic treatment within the current interval 1 = yes 0 = no 1.081 -0.8643 0.862 -0.5798 1.014 -0.9557 Finished removable prosthodontic treatment within the current interval 1 = yes 0 = no 0.453 -0.0286 1.170 -0.7929 2.062 -0.0374 Self-rated overall oral health at the preceding interview (TX-1) 1 = excellent/very good/good 2 = fair/poor 4.313 -<.0001 4.296 -<.0001 3.902 -<.0001 Age 1 = 45 to 64 years old 0 = 65 years old or older 1.117 -0.2616 1.118 -0.2590 1.105 -0.2838 Gender 1 = male 2 = female 0.943 -0.5645 0.941 -0.5554 1.037 -0.7083 Race 1 = White 2 = Black 0.704 -0.0008 0.708 -0.0008 1.429 -0.0002 Tooth loss in the past six months 1 = yes 2 = no 1.584 -0.0254 1.417 -0.1111 1.385 -0.0936 Number of pairs of teeth 1 = 10 occluding pairs of teeth or more 0 = fewer than 10 occluding pairs of teeth 0.688 -0.0007 0.720 -0.0031 1.387 -0.0017

PAGE 165

152 LIST OF REFERENCES Aday LA, Andersen RM (1974). A framework for the study of access to medical care, Health Serv Res 9: 208-20. Agerberg G, Carlsson GE (1981). Chewing ability in relation to dental and general health. Analysis of data obtained from a questionnaire. Acta Odontol Scand 39: 147-53. Albizu-Garcia CE, Alegria M, Freeman D, Vera M (2001). Gender and health services use for a mental health problem. Soc Sci Med 53: 865-78. Allen PF, McMillan AS (2003). A review of the functional and psychosocial outcomes of edentulousness treated with complete replacement dentures. J Can Dent Assoc 69: 662-62e. Allison PD. Logistic regression using the SAS system: Theory and application. Cary, N.C.: SAS institute, 1999. Andersen RM. A behavioral model of familiesÂ’ use of health services. Chicago, IL: Center for Health Administration Studies, University of Chicago, 1968. Andersen RM (1995). Revising the behavioral model and access to medical care: Does it matter? J Health Soc Behav 36: 1-10. Andersen RM, Davidson P, Ganz P (1994). Symbiotic relationships of quality of life, health services research and other health research. Qual Life Res 3: 365-71. Andersen RM, Davidson PL (1997). Ethnicit y, aging, and oral health outcomes: A conceptual framework. Adv Dent Res 11: 203-9. Andersen RM, Davidson PL, Nakazono TT (1997). Oral health policy and programmatic implications: Lesions from ICS-II. Adv Dent Res 11: 291-303. Arluke A, Kennedy L, Kessler R (1979). Reexamining the sick role concept: an empirical assessment, J Health Soc Behav r 20: 30-6. Atchison HF and Ralph WJ (1971). Tooth loss and biting force in man. J Dent Res 52: 225-8.

PAGE 166

153 Atchison KA, Andersen RM (2000). Demonstrating successful aging using the International Collaborative Study for oral health outcomes. J Public Health Dent 60: 282-8. Atchison KA, Davidson PL, Nakazono TT (1997). Predisposing, enabling, and need for dental treatment characteristics of ICS-II USA ethnically diverse groups. Adv Dent Res 11: 223-34. Atchison KA, Dolan TA (1990). Development of the geriatric oral health assessment index. J Dent Edu 54: 680-7. Bader JD, Shugars DA (1995). Variation, treatme nt outcomes, and practice guidelines in dental practice. J Dent Edu 59: 61-95. Baldwin DC (1980). Appearance and esthetics in oral health. Community Dent Oral Epidemiol 8: 224-56. Barenthin I (1977). Dental health status and dental satisfaction. Int J Epidemiol 6:73-9. Becker MH. The health belief model and personal health behavior. Thorofare, New Jersey: Charles B. Slack, Inc., 1974. Bell Jr DH (1972). Prosthodontic failures related to improper patient education and lack of patient acceptance. Dent Clin N Am 16: 109-18. Bergendal B (1989). The relative importance of tooth loss and denture wearing in Swedish adults. Community Dent Health 6: 103-11. Berkey DB, Call RL, Loupe MJ (1985). Oral health perceptions and self-esteem in noninstitutionalized older adults. Gerodontics 1: 213-6. Blomberg S, Lindquist LW (1983). Psychol ogical reactions to edentulousness and treatment with jawbone-anchored bridges. Acta Psychiatr Scand 68: 251-62. Boretti, G, Bickel M, Geering AH (1995). A re view of masticatory ability and efficiency. J Prosthet Dent 74: 400-3. Bradley EH, Curry LA, McGraw SA, Webster TR, Kasl SV, Andersen R (2004). Intended use of informal long-term care: The role of race and ethnicity. Ethn Health 9: 37-54. Bradley EH, McGraw SA, Curry L, Buckser A, King KL, Kasl SV, Andersen R (2002). Expanding the Andersen model: the role of psychosocial factors in long-term care use. Health Serv Res 37: 1221-42. Brenner MH, Curbow B, Legro MW (1995). Th e proximal-distal continuum of multiple health outcomes measures: The case of cataract surgery. Med Care 33: AS236-44.

PAGE 167

154 Brodeur J, Laurin D, Vallee R, Lachapelle D (1993). Nutrient intake and gastrointestinal disorders related to masticatory performance in the edentulous elderly. J Prosthet Dent 70: 468-73. Brown ER, Davidson PL, Yu H, Wyn R, Andersen RM, Becerra L, Razack N (2004). Effect of community factors on access to ambulatory care for lower-income adults in large urban communities. Inquiry 41: 39-56. Budtz-Jorgensen E, Isidor F (1990). A 5-y ear longitudinal study of cantilevered fixed partial dentures compared with removable partial dentures in a geriatric population. J Prosthet Dent 64: 42-7. Budtz-Jorgensen E, Theilade E, Theilade J (1983). Quantitative relationship between yeast and bacteria in denture-induced stomatitis. Scand J Dent Res 91: 134-42. Carlson BR, Carlsson GE, Helkimo E, Yont chev E (1992). Masticatory function in patients with extensive fixed cantilever prostheses. J Prosthet Dent 68: 918-23. Carlsson GE (1984). Masticatory efficiency: th e effect of age, the loss of teeth and prosthetic rehabilitation. Int Dent J 34: 93-7. elebi A, Knexovi -Zlatari D (2003). A comparison of patient’s satisfaction between complete and partial removable denture wearers. J Dent 31: 445-51. Chierici G, Lawson L (1973). Clinical speech considerations in prosthodontics: Perspectives of the prosthodontist and speech pathologist. J Prosthet Dent 29: 2939. Cole S and LeJeune R (1972). Illness and the legitimation of failure. Am Sociol Rev 37: 347–56. Conny DJ, Tedesco LA, Brewer JD, Albino JE (1985). Changes of attitude in fixed prosthodontic patients. J Prosthet Dent 53: 451-4. Crown WH, Obenchain RL, Englehart L, La ir T, Beusching DP, Croghan T (1998). The application sample selection models to outcomes research: The case of evaluating the effects of antidepressant therapy on resource utilization. Statist Med 17: 194358. D’Agostino RB (1998). Tutorial in biostatistics: Propensity score methods for bias reduction in the comparison of a treatment to a non-randomized control group. Statist Med 17: 2265-81. Davidson PL, Andersen RM (1997). Determinants of dental care utilization for diverse ethnic and age groups. Adv Dent Res 11: 254-62.

PAGE 168

155 Davidson PL, Andersen RM, Wyn R, Brow n ER (2004). A framework for evaluating safety-net an other community level factors on access for low-income populations. Inquiry 41: 21-38. Davis EK, Albino JE, Tedesco LA, Portenby BS, Ortman LF (1986). Expections and satisfaction of denture patients in a university clinic. J Prosthet Dent 55: 59-63. De Paoli MM, Manongi R, Klepp KI (2004). Factors influencing acceptability of voluntary counseling and HIV-testing among pregnant women in Northern Tanzania. AIDS Care 16: 411-25. De Rosa A, Troncone A, Vacca M, Ciacci C (2004). Characteristics and quality of illness behavior in celiac disease. Psychosomatics 45: 336-42. Dobalian A, Andersen RM, Stein JA, Hays RD, Cunningham WE, Marcus M (2003). The impact of HIV on oral health and subsequent use of dental services. J Public Health Dent 63: 78-85. Dobalian A, Tsao JCI, Duncan RP (2004). Pain and the use of outpatient services among persons with HIV: Results from a nationally representative survey. Med Care 42: 129-38. Dolan TA, Gilbert GH, Duncan RP, Foerster U (2001). Risk indicators of edentulism, partial tooth loss and prosthetic status among black and white middle-aged and older adults. Community Dent Oral Epidemiol 29: 329-40. Dolan TA, Gilbert GH, Ringelberg ML, Legler DW, Antonson DE, Foerster U, Heft MW (1997). Behavioral risk indicators of attachment loss in adult Floridians. J Clin Periodontol 24: 223-32. Dolan TA, Peek CW, Stuck AE, Beck JC (1998). Three-year changes in global oral health ratings by elder dentate adults. Commun Dent Oral Epidemiol 26: 62-69. Douglass CW, Gammon MD, Atwood DA (1988). Need and effective demand for prosthodontic treatment. J Prosthet Dent 59: 94-104. Douglass CW, Shih A., Ostry L (2002). Will there be a need for complete dentures in the United States in 2020? J Prosthet Dent 87: 5-8. Douglass CW, Watson AJ (2002). Future needs for fixed and removable partial dentures in the United States. J Prosthet Dent 87: 9-14. Elias AC, Sheiham A (1999). The relationship between satisfaction with mouth and number, position and condition of teeth: studies in Brazilian adults. J Oral Rehabil 26:53-71.

PAGE 169

156 Eisenberg DM, Kessler RC, Foster C, No rlock FE, Calkins DR, Delbanco TL (1993). Unconventional medicine in the United States: prevalence, costs, and patterns of use. New Engl J Med 328: 246-52. Fiske J, Davis DM, Frances C, Gelbier S (1998). The emotional effects of tooth loss in edentulous people. British Dent J 184: 90-3. Fiske J, Gelbier S, Watson RM (1990). The benefit of dental care to an elderly population assessed using a sociodental measure of oral handicap. Br Dent J 168: 153-6. Foerster U, Gilbert GH, Duncan RP ( 1998). Oral functional limitation among dentate adults. J Public Health Dent 58: 202-9. Frank RP, Brudvik JS, Leroux BG, Milgro m P, Hawkins NR (2000). Relationship between the standards of removable partial denture construction, clinical acceptability, and patient satisfaction. J Prothet Dent 83: 521-7. Frank RP, Milgrom P, Leroux BG, Hawkins NR (1998). Treatment outcomes with mandibular removable partial dentures: A population-based study of patient satisfaction. J Pros Dent 80: 36-45. Fulton JP, Buechner JS, Scott HD, DeBuono BA, Feldman JP, Smith RA, Kovenock D (1991). A study guided by the health belief model of the predictors of breast cancer screening of women ages 40 and older. Public Health Rep 106: 410-20. Gallagher TC, Andersen RM, Koegel P, Gelberg L (1997). Determinants of regular source of care among homeless adults in Los Angeles. Med Care 35: 81430. Garrett NR, Kapur KK, Hasse AL, Dent RJ (1997) Veterans Administration Cooperative Dental Implant Study–Comparisons between fixed partial dentures supported by blade-vent implants and removable partial dentures. Part V: Comparisons of pretreatment and posttreatment dietary intake. J Prosthet Dent 77: 153-61. Gelberg L, Andersen RM, Leake BD (2000). The behavioral model for vulnerable populations: Application to medical care use and outcomes for homeless people. Health Serv Res 34: 1273-302. Gilbert GH, Antonson DE, Mjör IA, Ringelber g ML, Dolan TA, Foerster U, Legler DW, Duncan RP, Heft MW (1996). Coronal caries, root fragments, and restoration and cusp fractures in U.S. adults. Caries Res 30: 101-11. Gilbert GH, Branch LG, Orav EJ (1990). Predictors of older adults’ longitudinal dental care use. Med Care 28: 1165-80. Gilbert GH, Chavers LS, Shelton BJ (2002c). Comparison of two methods of estimating 48-month tooth loss incidence. J Public Health Dent 62: 163-9.

PAGE 170

157 Gilbert GH, Duncan RP, Heft MW, Coward RT (1997b). Dental health attitudes among dentate black and white adults. Med Care 35: 255-71. Gilbert GH, Duncan RP, Heft MW, Dolan TA, Vogel WB (1997a). Oral disadvantage among dentate adults. Commun Dent Oral Epidemiol 25: 301-13. Gilbert GH, Duncan RP, Heft MW, Dolan TA, Vogel WB (1998a). Multidimensionality of oral health in dentate adults. Med Care 36: 988-1001. Gilbert GH, Duncan RP, Kulley AM, Coward RT, Heft MW (1997c). Evaluation of bias and logistics in a survey of adults at increased risk of oral health decrements. J Public Health Dent 57: 48-58. Gilbert GH, Duncan RP, Rose JS, Shelton BJ (2002a). Tooth loss occurring outside of a health care facility: 72-month incidence. J Den Res 81: 860-5. Gilbert GH, Duncan RP, Vogel WB (1998b). Determinants of dental care use in dentate adults: Six-monthly use during a 24-month period in the Florida Dental Care Study. Soc Sci Med 47: 727-37. Gilbert GH, Heft MW, Duncan RP, Ringelber g ML (1994). Perceived need for dental care in dentate older adults. Int Dent J 44: 145-52. Gilbert GH, Meng X, Duncan RP, Shelton BJ (2004). Incidence of tooth loss and prosthodontic dental care: Effect on chewi ng difficulty onset, a component of oral health-related quality of life. J Am Geriatr Soc 52: 880-5. Gilbert GH, Shelton BJ, Duncan RP (2002b). Us e of specific dental treatment procedures by dentate adults during a 24-month period. Community Dent Oral Epidemiol 30: 260-76. Goldstein R (1986). Geriatric patients are ideal candidates for esthetic dentistry. Dent Management 26: 37-42. Gordon SR, Fryer GE, Niessen L (1988). Patie nt satisfaction with current dental condition related to self-concept and dental status. J Prosthet Dent 59: 323-7. Greenland S (2000). An introduction to instrumental variable for epidemiologists. Int J Epedimiol 29: 722-9. Greksa LP, Parraga IM, Clark CA (1995). Th e dietary adequacy of edentulous older adults. J Prothet Dent 73: 142-5. Grembowski D (1997). The role of health services research in the renaissance of the dental profession. J Dent Edu 61: 10-5. Grembowski D, Andersen RM, Chen M (1989). A public health model of the dental care process. Med Care Rev 46: 439-96.

PAGE 171

158 Heckman J (1979). Sample selection as a specification error. Econometrica 47: 153-61. Hjern A, Grindefjord M, Sundberg H, Rosén M (2001). Social inequality in oral health and use of dental care in Sweden. Community Dent Oral Epidemiol 29: 167-74. Huang JA, Tsai WC, Chen YC, Hu WH, Yang DY (2003). Factors associated with frequent use of emergency services in a medical center. J Formos Med Assoc 102: 222-8. Institute of Medicine. Access to health care in American. Washington, D.C.: National Academy Press, 1993. Institute of Medicine. Dental education at the crossroads: Challenges and changes. Washington, D.C.: National Academy Press, 1995. John MT, Koepsell TD, Hujoel P, Miglioretti DL, LeResche L, Micheelis W (2004). Demographic factors, denture status and oral health-related quality of life. Community Dent Oral Epidemiol 32: 125-32. Johnson RJ, Wolinsky FD (1993). The structure of health status among older adults: disease, disability, functional limitation, and perceived health. J Health Soc Behav 34: 105-21. Joshi A, Douglass CW, Feldman H, Mitchell P, Jette A (1996). Consequences of success: Do more teeth translate into more disease and utilization? J Public Health Dent 56: 190-7. Kapur KK, Garrett NR, Dent RJ, Hasse AL ( 1997). A randomized clinical trial of two basic removable partial denture designs. Part II: Comparisons of masticatory scores. J Prosthet Dent 78: 15-21. Kapur KK, Soman SD (2004). Masticatory pe rformance and efficiency in denture wearers, 1964. J Prosthet Dent 92 : 107-11. Käyser AF, Meeuwissen R, Meeuwissen JH (1990). An occlusal concept for dentate geriatric patients. Community Dent Oral Epidemiol 18: 319. Kegeles SS (1963). Some motives for seeking preventive dental care. J Am Dent Assoc 67: 90-8. Kennedy P. A guide to econometrics. Cambridge, MA: The MIT Press, 2003. Krall E, Hayes C, Garcia RJ (1998) How den tition status and masticatory function affect nutrition intake. J Am Dent Assoc 129: 1261-9. Kronenfeld JJ, Wasner C (1982). The use of unorthodox therapies and marginal practitioners. Soc Sci Med 16: 1119-25.

PAGE 172

159 Lamy M, Mojon P, Kalykakis G, Legrand R, Butz-Jorgensen E(1999). Oral status and nutrition in the institutionalized elderly. J Dent 27: 443-8. LaVeist TA, Keith VM, Gutierrez ML(1995). Black/White differences in prenatal care utilization: An assessment of predisposing and enabling factors. Health Serv Res 30:43-58. Lazarus RS, Folkman S. Stress, appraisal, and coping. New York: Springer, 1984. Leake JL, Hawkins R, Locker D (1994). Social and functional impact of reduced posterior dental units in older adults. J Oral Rehabil 21: 1-10. Leao A, Sheiham A (1995). Relation between clinical dental status and subjective impacts on daily living . J Dent Res 74: 1408-13. Levit K, Smith C, Cowan C, Lazenby H, Sensenig A, Catlin A (2003). Trends in U.S. health care spending, 2001. Health Aff 22: 154-64. Liang KY, Zeger SL (1986). Longitudinal data analysis using generalized linear models. Biometrika 73: 13-22. Liang KY, Zeger SL (1993). Regression analysis for correlated data. Annu Rev Public Health 14: 43-68. Liedberg B, Norlen P, Owall B (1991). Teeth, tooth spaces, and prosthetic appliances in elderly men in Malmo, Sweden. Community Dent Oral Epidemiol 19: 164-8. Lim YW, Andersen R, Leake B, Cunningham W, Gelberg L (2002) How accessible is medical care for homeless women? Med Care 40: 510-20. Lo ECM, Schwarz E (1998). Determinants for dental visit behavior among Hong Kong Chinese in a longitudinal study. J Public Health Dent 58: 220-7. Locker D (1988). Measuring oral health: a conceptual framework. Community Dent Health 5: 3-18. Locker D (1997). Clinical correlates of changes in self-perceived oral health in older adults. Community Dent Oral Epidemiol 25: 199-203. Loupe MJ, Goodkind RJ, Smith BJ, Clay DJ, DiAngelis AJ (1988). Modifying the expections of denture patients. Gerodontics 4: 90-4. Mahmood WA, Watson CJ, Ogden AR, Hawkins RV (1992). Use of image analysis in determining masticatory efficiency in patients presenting for immediate dentures. Int J Prosthodont 5: 359-66. Marcus SE, Drury TF, Brown LJ, Zion GR ( 1996). Tooth retention and tooth loss in the permanent dentition of adults: United States, 1988-1991. J Dent Res 75: 684-95.

PAGE 173

160 Marcus SE, Kaste LM, Brown LJ (1994). Pr evalence and demographic correlations of tooth loss among the elderly in the United States. Special Care Dent 14: 123-7. Matthias RE, Atchison KA, Lubben JE, De Jong F, Schweitzer SO (1995). Factors affecting self-ratings or oral health. J Public Health Dent 55: 197-204. Matthias RE, Atchison KA, Schweitzer SO, Lubben JE, Mayer-Oakes A, De Jong F (1993). Comparisons between dentist ratings and self-ratings of dental appearance in an elderly population. Special Care Dent 13: 53-60. Maupomé G, MacEntee MI (1998). Prosthodontic profiles relating to economic status, social network, and social support in an elderly population living independently in Canada. J Prosthet Dent 80: 598-604. McClellan M, McNeil BJ, Newhouse JP (1994). Does more intensive treatment of acute myocardial infarction reduce mortality? JAMA 272: 859-66. McClellan M, Newhouse JP (1997). The marginal costs and benefits of medical technology: a panel instrumental-variables approach. J Econom 77: 39-64. McGrath C, Bedi R (2001). Can denture impr ove the quality of life of those who have experienced considerable tooth loss. J Den 29: 243-6. McHugh S, Vallis TM. Illness behavior: A multidimensional model. New York: Plenum Press, 1985. Mechanic D. Symptoms, illness behavior, and help-seeking. New Brunswick, New Jersey: Rutgers University Press, 1982. Miyaura K, Morita M, Matsuka Y, Yamashita A, Watanabe T (2000). Rehabilitation of biting abilities in patients with different types of dental prostheses. J Oral Rehabil 27: 1073-6. Mojon P, MacEntee MI. (1992). Discrepancy between need for prosthodontic treatment and complaints in an elderly edentulous population. Community Dent Oral Epidemiol 20: 48-52. Mok E, Lai C, Zhang ZX (2004). Coping with chronic renal failure in Hong Kong. Int J Nurs Stud 41: 205-13. Moynihan PJ, Butler TJ, Thomason JM, Jepson NJ (2000). Nutrient intake in partially dentate patients: The effect of prosthetic rehabilitation. J Dent 28: 557-63. Nagi SZ (1976). An epidemiology of disability among adults in the United States. Milbank Q 54: 439-67. Nevalainen MJ, Närhi TO, Ainamo A (2004). A 5-year follow-up study in the prosthetic rehabilitation of the elderly in Helsinki, Finland. J Oral Rehabil 31: 647-52.

PAGE 174

161 Newhouse JP, McClellan MB (1998). Econometrics in outcomes research: The use of instrumental variables. Annual Review Pub Health 19: 17-34. Newmann LM, Christensen C, Cavanaugh C ( 1989). Dental esthetic satisfaction in adults. J Am Dent Assoc 118: 565-70. Nowjack-Raymer RE, & Sheiham A (2003). Association of edentulism and diet and nutrition in U.S. adults. J Dent Res 82, 123-6. Oesterberg T, Steen B (1982). Relationship betw een dental state and dietary intake in 70 year-old males and females in Gotegorg, Sweden: a population study. J Oral Rehabil 9: 509-21. Ofstehage J (1987). The social and psyc hological importance of dental-facial attractiveness in the elderly. Vet Admin Newsletter 1: 31-3. Oosterhaven SP, Westert GP, Schaub RMH (1989) . Perception and significance of dental appearance: The case of missing teeth. Community Dent Oral Epidemiol 17: 123-6. Österberg T, Era P, Gause-Nilsson I, Steen B (1995). Dental state and functional capacity in 75-year-olds in three Nordic localities. J Oral Rehabil 22: 653-60. Otchere DF, Leake JL, Locker D (1990). Comparing older adults’ perceived need for dental care with a normative hierarchy of needs. Gerodontology 9: 111-7. Ow RKK, Loh T, Neo J, Khoo J (1997). Per ceived masticatory function among elderly people. J Oral Rehabil 24: 131-7. Owall BE (1986). Prosthetic epidemiology. Int Dent J 36: 230-4. Palmer JM (1974). Analysis of speech in prosthodontic practice. J Prosthet Dent 31: 60514. Papas AS, Palmer CA, Rounds MC, Russell RM (1998). The effects of denture status on nutrition. Special Care Dent 18: 17-25. Parsons T (1975). The sick role and role of the physician reconsidered. Milbank Mem Fund Q Health Soc 53: 257-78. Parsons T. The social system. New York: Free Press; 1951. Peek CW, Gilbert GH, Duncan RP, Heft MW, Henretta JC (1999). Patterns of change in self-reported oral health among dentate adults. Med Care 37: 1237-48. Pescosolido B & Kronenfeld JJ (1995). Health, illness, and healing in an uncertain era: Challenges from and for medical sociology. J Health Soc Behav Spec No: 5-33. Pilowsky I, Spence ND (1975). Patterns of illness behavior in patients with intractable pain. J Psychosom Res 23:117-30.

PAGE 175

162 Posner MA, Ash AS, Freund KM, Moskowitz MA, Shwartz M (2001). Comparing standard regression, propensity score matching, and instrumental variables methods for determining the influence of mammography on stage of diagnosis. Health Serv Outcomes Res Methodol 2: 279-90. Ramfjord DJ (1974). Periodontal aspects of restorative dentistry. J Oral Rehabil 1: 10726. Razak IA, Jaafar N, Jalalludin RI, Esa R ( 1990). PatientsÂ’ preference for exodontia versus preservation in Malaysia. Community Dent Oral Epidemiol 18: 131-2. Redford M, Drury TF, Kingman A, Brown LJ (1996). Denture use and the technical quality of dental prostheses among persons 18-74 years of age: United States, 1988-1991. J Dent Res 75: 714-25. Reinsine ST, Bailit L (1980). Clinical oral health status and adult perceptions of oral health. Soc Sci Med 14A: 597-605. Rief W, Ihle D, Pilger F (2003). A new approach to assess illness behaviour. J Psychosom Res 54: 405-14. Ringelberg ML, Gilbert GH, Antonson DE, Dola n TA, Legler DW, Foerster U, Heft MW (1996). Root caries and root defects in urban and rural adults. J Am Dent Assoc 127: 885-91. Rosenbaum PR. Observational studies. New York: Springer-Verlag, 1995. Rosenoer LM, Sheiham A (1995). Dental impacts on daily life and satisfaction with teeth in relation to dental status in adults. J Oral Rehabil 22: 469-80. Rossi T, Laine J, Eerola E, Kotilainen P, Peltonen R (1995). Denture carriage of methicillin-resistant Staphylococcus aureus . Lancet 345: 1577. Schuurs AH, Duivenvoorden HJ, Thoden van Velzen SK, Verhage F, & Makkes PC. (1990). Value of teeth. Community Dent Oral Epidemiol 18: 22-6. Sheiham A, Steele JG, Marcenes W, Finch S, Walls AWG (1999) The impact of oral health on stated ability to eat certain foods: findings from the National Diet and Nutrition Survey of Older People in Great Britain. Gerodontology 16: 11-20. Sheiham A, Steele JG, Marcenes W, Tsakos G, Finch S, & Walls AWG (2001). Prevalence of impacts of dental and or al disorders and their effects on eating among older people: a national survey in Great Britain. Community Dent Oral Epidemiol 29: 195-203. Shelton BJ, Gilbert GH, Lu Z, Bradshaw P, Chavers LS, Howard G (2003). Comparing longitudinal binary outcomes in an observational oral health study. Statist Med 22: 2057-70.

PAGE 176

163 Slade GD, Spencer AJ (1994). Development and evaluation of the oral health impact profile. Commun Dent Health 11: 3-11. Slagter AP, Olthoff LW, Bosman F, Steen WHA (1992). Masticatory ability, denture quality, and oral conditions in edentulous subjects. J Prosthet Dent 68: 299-307. Spartley MH (1988). Posterior edentulousness and the prescription of partial dentures. Aust Dent J 33: 43-6. Steele JG, Ayatollahi SM, Walls AW, & Murray JJ (1997). Clinical factors related to reported satisfaction with oral function amongst dentate older adults in England. Community Dent Oral Epidemiol 25: 143-9. Stiger TR, Barnhart HX, Williamson JM (1999). Testing proportionality in the proportional odds model fitted with GEE. Statist Med 18: 1419-33. Stokes ME, Davis CS, Koch GG. Categorical da ta analysis using the SAS system. Cary, N.C.: SAS institute, 2001. Stoller EP, Forster LE, Portugal S (1993). Self-care responses to symptoms by older people: a health diary study of illness behavior. Med Care 31: 24-42. Strauss RP, Hunt RJ (1993). Understanding the value of teeth to older adults: influences on the quality of life. J Am Dent Assoc 124: 105-10. Strayer MS, Branch LG, Jones JA, Adelson R (1988). Predictors of the use of dental services by older veterans. Spec Care Dent 10: 16-20. Suchman EA (1965a). Social patterns of illness and medical care. J Health Soc Behav 6: 2–16. Suchman EA (1965b). Stages of illness and medical care. J Health Soc Behav 6: 114-28. Tan BS, Ng KH, Esa R (2001). Health beliefs in oral cancer: Malaysian estate Indian scenario. Patient Educ Couns 42: 205-11. Tennstedt SL, Brambila DL, Jette AM, McGuire SM (1994). Understanding dental service use by older adults: sociobehavioral factors vs. need. J Public Health Dent 54: 211-9. Tervonen T (1988). Condition of prosthetic constructions and subjective needs for replacing missing teeth in a Finnish adult population. J Oral Rehabil 15: 505-13. Tervonen T, Knuuttila M (1988). Awareness of dental disorders and discrepancy between “objective” and “subjective” dental treatment need. Community Dent Oral Epidemiol 16: 345-48.

PAGE 177

164 Theilade E, Budtz-Jorgensen E. (1988). Predominant cultivatable microflora of plaque on removable dentures in patients with denture induced stomatitis. Oral Microbiol Immunol 3: 8-13. Trulsson U, Engstrand P, Berggren U, Nanm ark U, Branemak PI (2002). Edentulousness and oral rehabilitation experience from the patient perspective. Eur J Oral Sci 110: 417-24. Tsuga K, Carlsson GE, Ostergerg T, Karlsson S (1998). Self-assessed masticatory ability in relation to maximal bite force and dental state in 80-year-old subjects. J Oral Rehab 25: 117-24. Turunen S, Nyyssone V, Vesala H (1993). Perspectives on poor dental health and its determinants. Community Dent Health 10: 49-55. US Bureau of the Census. Census of popul ation and housing, 1990: Public use microdata samples. Washington, D.C.: US technical Documentation, 1992. US Bureau of the Census. Unpublished special tabulations for the University of Florida from the 1990 Census of population and housing for the US and four counties in north Florida, 1994. Van Achterbert T, Stevens FC, Crebolde r HF, Philipsen H (1996). Predictors of professional and non-professional community care for care-dependent adults. Health Policy 36: 83-98. Van der Bilt A, Olthoff LW, Bosman F, Oosterhaven SP (1994). Chewing performance before and after rehabilitation of post-canine teeth in man. J Dent Res 73: 1677-83. Van Waas MAJ, Meeuwissen JH, Meeuwissen R, Käyser AF, Kalk W, Van’t Hof MA (1994). Relationship between wearing a removable partial denture and satisfaction in the elderly. Community Dent Oral Epidemiol 2: 315-8. Walsh JC, Lynch M, Murphy AW, Daly K (2004). Factors influencing the decision to seek treatment for symptoms of acute myocardial infarction: an evaluation of the Self-Regulatory Model of illness behaviour. J Psychosom Res 56: 67-73. Weintraub AM, Weintraub GS (1997). The dent al student as technician: an 18-year follow-up of preclinical laboratory programs. J Prosthodont 6: 128-36. Weyant RJ, Pandav RS, Plowman JL, Ganguli M (2004). Medical and cognitive correlates of denture wearing in older community-dwelling adults. J Am Geriatr Soc 52: 596-600. Winship C, Mare RD (1992). Models for sample selection bias. Annual Rev Sociol 18: 327-50.

PAGE 178

165 Witter DJ, Van Elteren PH, Käyser AF, Van Rossum GMJM (1989). The effect of removable partial dentures on the oral function in shortened dental arches. J Oral Rehabil 16: 27-33. Witter DJ, Van Palenstein Helderman WH, Creugers NH, Käyser AF (1999). The shortened dental arch concept and its implications for oral health care. Community Dent Oral Epidemiol 27: 249-58. York J, Holtzman J. (1999). Facial attractiveness and the aged. Special Care Dent 19: 848. Zarb GA, Bergman B, Clayton JA, McKay HF. Prosthodontic treatment for partially edentulous patient. St. Louis: C.V. Mosby, 1978. Zussman, R. Intensive care: Medical ethi cs and the medical profession. Chicago: University of Chicago Press, 1992.

PAGE 179

166 BIOGRAPHICAL SKETCH Xiaoxian Meng is originally from Chengdu, China. She graduated from the West China University of Medical Sciences in 1999 with D.D.S. and Ms.D degrees. Upon completion of her degrees, Xiaoxian worked as a part-time dentist and clinical instructor in the Department of Endodontics at the West China University of Medical Sciences for one year. Xiaoxian came to the University of Florida in 2000 and completed a MPH program in community health education in 2002. In the fall of 2002, she entered the doctoral program in health services research at the University of Florida and received a University of Florida Alumni Fellowship. During the course of her doctoral studies, Xiaoxian participated in many research projects and presented at several international conferences and meetings. She is also an active member of professional organizations such as International Association for Dental Resear ch (IADR), and American Association for Dental Research (AADR).