![]() ![]() |
![]() |
UFDC Home | myUFDC Home | Help |
|
|
||||||||||||||||||||||||||||||||||||||||||||
Full Text | |||||||||||||||||||||||||||||||||||||||||||||
AS SESS ING AND TREATING CHRONIC PAIN IN COMMUNITY-DWELLNG OLDER ADULTS WITH DEMENTIA: A SINGLE-SUBJECT APPROACH By AMANDA FLOETKE ELLIOTT 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 2006 Copyright 2006 by Amanda Floetke Elliott ACKNOWLEDGMENTS I would like to express my profound appreciation to my dissertation chair and mentor, Dr. Ann Horgas, for her constant support and dedication to guiding me through the wonderful world of research. Her direction and faith in me throughout these past five years have made it possible to complete the journey through this program. I would also like to acknowledge Dr. Jennifer Elder, Dr. Michael Marsiske, and Dr. Meredeth Rowe and thank them for their time and support and for sharing with me their knowledge and expertise throughout the fruition of this research proj ect. Many thanks go to the families that participated in this study and who welcomed me into their homes each day. I have learned so much more than I ever expected too. I feel truly blessed to have had the privilege to know each of them. I would like to thank my family and friends for everything that they have done to see me through this program. I thank my parents and brother for constantly supporting me and encouraging me to succeed. I thank my friends for all of their many different kinds of support; I needed them all. Finally, I especially thank my husband for putting up with me these last five years and seeing me through this program, even though it meant a lot of time apart from each other. I truly could not and would not have made it to here without his love and support. TABLE OF CONTENTS page ACKNOWLEDGMENT S .........._.... .............._ iii.._... ..... LI ST OF T ABLE S .........._.... .............._ viii...__ .... LIST OF FIGURES .............. .................... ix AB STRAC T ................ .............. xi CHAPTER 1 INTRODUCTION ................. ...............1.......... ...... Background and Significance of the Problem ................. .............. ......... .....2 Rationale and Need for this Study ................. ......... ...............6..... Theoretical Framework.................. .................7 Statement of Problem and Study Specific Aims ................. .......... ................1 1 2 REVIEW OF THE LITERATURE ................. ...............14........... ... Pain Prevalence ................. ...............14................. Pain Assessment .............. ...............16.... Self-Report .............. ...............16... Observational Strategies ................. ............. ...............19....... Advantages of observational methods............... ...............20 Limitations of observational methods ................. ................ ......... .21 Pain Behaviors ................. ...............22........... .... Identifying Pain Behaviors ................. ...............23........... .... Testing Pre-identified Behaviors ............... .... ........... ........... ............2 Underassessment and Treatment of Pain in Persons with Dementia. ................... ......26 Caregiver Assessments ................ .... ....... ............2 Informal Caregivers versus Trained Ob servers ................. ....___ ..............28 Informal Caregivers versus Participant Self-Report ................. ... .. ............ .....29 Strategies to Strengthen the Validity and Reliability of Caregiver Report ................33 Excess Disabilities related to Dementia and Chronic Pain............... ..................3 Pain Management and Treatment Guidelines............... ...............3 Sum m ary ................. ...............41.......... ...... 3 M ETHODS .............. ...............42.... Study Design............... ...............42. Sample .............. ..... .. ...............45 Recruitment Procedures ................... ....... ...... ... ........ .. .. .. .......4 Inclusion and Exclusion Criteria for Participants with Dementia ................... ....46 Inclusion and Exclusion Criteria for Caregivers .............. ....................4 Participant Pair 1, Caregiver 1 (CG 1) and Participant 1 (P 1) .........................47 Participant Pair 2, Caregiver 2 (CG 2) and Participant 2 (P 2) ...........................47 Participant Pair 3, Caregiver 3 (CG 3) and Participant 3 (P 3) ...........................47 Participant Pairs Withdrawn from the Study ....._._._ ........___ ........._.....48 M measures ................ ......... ...... ......... .............4 Comprehensive Intake Assessment ................. ............ ........... ...... .........4 Caregiver Ratings of Pain Behaviors and Excess Disabilities ................... .........49 Trained Observer Ratings of Pain Behaviors ......._.__ .... ........ ..............52 Treatment Protocol .............. ...............55.... Statistical Power ............ ..... .._ ...............56... Proc edure s............... ....._ ...............57.... Phase A l, Session 1.............. ...............60.... Phase A l............... ...............61... Phase Al-Final Session............... ...............62 Phase B1 (Eight Sessions). ............. ..... ..._ ................62. Phase A2- Return to Baseline (Range of 4-8 Sessions) ............. ...............63 Phase B2- Reimplementation of Intervention (Range of 4-8 Sessions) ..............63 Data Coding ............... ... ... ....__ .......__ ............6 Data Analysis Procedures and Variables of Interest. ....._____ ..... ... ..............65 RM BPC .............. ...............65.... PADE .........__ ......... ._ ...............66... Activity Protocol ................ ...............66... Additional Caregiver Questionnaire .............. ........__ .....__ ...........6 Overview of the Analysis of Single-Subj ect Design Data ................. ................ ...66 Statistical Analysis of Each Aim ................ ...............71........... ... 4 RE SULT S .............. ...............74.... Descriptives .............. ...............74.... Initial Intake M measures .............. ...............74.... Final Outtake Measures ................. ....... .. ...... .... .. ..... ............ .........7 Effect of Acetaminophen on Self-Reported Pain Intensity and Observable Pain Behaviors ............... ........ ..............7 Total Number of Pain Behaviors ....__ ......_____ .......___ ...........7 Participant 1............... ...............78... Participant 2............... ...............80... Participant 3................ ...............82.. Total Duration of Pain Behaviors ....__ ......_____ .......___ ...........8 Participant 1............... ...............84... Participant 2............... ...............85... Participant 3............... .......... ......... .......8 Caregiver and PI Pain Ratings on the PADE .............. ...............89.... Participant 1............... ...............89... Participant 2............... ...............90... Participant 3............... ........ ......... ........9 Effect of Acetaminophen on Excess Disabilities. ................ ......................__.93 Total RMBPC Score ................. ...............94........... .... Participant 1............... ...............94... Participant 2............... ...............96... Participant 3............... ...............97... PADE Part III Subscale Score ....._........___...... .....__ ...........9 Reliability of Caregivers' Ratings of Pain Behaviors .................. ................. ... 100 Pain Behaviors Frequently Displayed by Persons with Dementia .........._...............102 Frequent Pain Behaviors determined by the Coded Activity Protocols ............103 Frequent Pain Behaviors determined by Caregivers' Ratings on Part I of the PA D E ............. .. ... .. .. ... ......... .. ...........10 Frequent Pain Behaviors reported by Caregivers on the Free Response Question ............. .............. 104... ..... 5 DI SCUS SSION .........__........_. .............. 106... Effect of Acetaminophen on Self-Reported Pain Intensity and Observable Pain Behaviors .............. ...............107.... Self-Report of Pain ............ ..... .._ ...............107... Ob servation of Pain B ehaviors ...._.._.._ ........__. ...._.._ ..........10 Proxy Pain Assessments ........._.._.... ........._._. ......._._.......... 11 Effect of Acetaminophen on Excess Disabilities ......___ .........__ ................113 Reliability of Caregivers' Ratings of Pain Behaviors .................. ................. ... 1 17 Pain Behaviors Frequently Displayed by Persons with Dementia .........................119 Validation of the Study Findings ....__ ........._.... ......... ............2 Treatment Integrity ................. ...............120............. Validity of the Findings .....__................. ...............121 .... Nursing Implications .............. ...............123.... Clinical Si gnif icance............... .............12 Lim stations ................. ...............128............. Recruitm ent .............. ...............128.... Sample Size ........._.__......_.._ ...............129.... Tool Choi ce ........._.__......_.._ ............... 13 0... Feasibility .............. ............... 13 1... Future Directions ........._.._.._ ....__. ..............._ 132.. APPENDIX A PERFORMING GRAPHICAL ANALYSES FOR SINGLE-SUBJECT DESIGN DATA ............ ...... ...............137.. B INFORMED CONSENT TO PARTICIAPTE IN RESEARCH-CHRONIC PAIN PARTICIPANT .............. ...............143.... C INFORMED CONSENT TO PARTICIPATE IN RESEARCH-CAREGIVERS ....154 D DAILY CAREGIVER QUESTIONNAIRES ......____ ..... ... ._ .............. ..163 LIST OF REFERENCES ........._... ...... ..... ...............169... BIOGRAPHICAL SKETCH ................. ...............180......... ...... LIST OF TABLES Table pg 1-1 Theoretical and Study Constructs .............. ...............12.... 3-1 Study Constructs and Measures .............. ...............53.... 3-2 Study Design .............. ...............60.... 3-3 Behavioral Definitions .............. ...............67.... 4-1 Descriptive Characteristics for Participants .............. ...............76.... 4-2 Descriptive Characteristics for Caregivers ................. ....................___......76 4-3 Mean Duration (in Seconds) of Individual Pain Behaviors ........._..... ........_......86 4-4 Caregiver and Trained Observer PADE Part I Correlations .............. .................101 A-1 Calculating the M ean .............. ...............137.... A-2 Calculating the Level Change between Phases .............. ....__. ..............138 A-3 Determining the Latency of Change between Phases ................. .....................140 A-4 Calculating 15% Above and 15% Below the Mean ................. ............ .........141 LIST OF FIGURES Figure pg 1-1 Communications model of pain ................. ............. ......... ........ .......9 4-1 Total number of pain behaviors for Pl ................. ...............79........... . 4-2 Mean frequencies for individual behaviors for P 1. ............. .....................8 4-3 Total number of pain behaviors for P 2 ................. ................ ......... ...._81 4-4 Mean frequencies for individual behaviors for P 2. ............. .....................8 4-5 Total number of pain behaviors for P 3............. ......___.....__ ..............83 4-6 Mean frequencies for individual behaviors for P 3. ......................__ ..............84 4-7 Total duration of pain behaviors for P 1. ......................_. ......._..............85 4-8 Total duration of pain behaviors for P 2. .........__ ........... .......___..........87 4-9 Total duration of pain behaviors for P 3. ..._____ ....... .... ..__ ............88 4-10 Caregiver and TO PADE Part I pain ratings for P 1. ............. .....................9 4-11 Caregiver and TO PADE Part I pain ratings for P 2. ............. .....................9 4-12 Caregiver and TO PADE Part I pain ratings for P 3 .............. ....................9 4-13 Frequency of excess disabilities across sessions for P 1. ............. ....................95 4-14 Frequency of excess disabilities across sessions for P 2. .............. ....................97 4-15 Frequency of excess disabilities across sessions for P 3............_ .........._ .....98 4-16 Mean frequencies of excess disabilities related to physical functioning................ 100 4-17 Scatterplot for the CG' s and the TO's total scores on PADE Part I for Pl............101 4-18 Scatterplot for the CG' s and the TO's total scores on PADE Part I for P2............102 4-19 Scatterplot for the CG' s and the TO's total scores on PADE Part I for P3............102 4-20 Total frequency of pain behaviors coded from the activity protocols. ................... 103 4-21 Total frequency of PADE Part I pain behaviors across participants..........._........104 A-1 Presenting the mean for each phase in the graph. ............. ......................3 A-2 Example of a trendline produced in Excel ................ .......... ................ ..139 A-3 Example of calculating the stability of data points within a study phase............... 141 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 AS SESS ING AND TREATING CHRONIC PAIN IN COMMUNITY-DWELLNG OLDER ADULTS WITH DEMENTIA: A SINGLE-SUBJECT APPROACH By Amanda Floetke Elliott August 2006 Chair: Ann Horgas Major Department: Nursing Previous research on pain assessment in persons with dementia has mostly focused on nursing home or hospitalized populations, finding that this subgroup of older adults experiences comparable pain to cognitively intact elders. These studies advocate an analgesic trial to treat pain in this population as untreated chronic pain can lead to excess disabilities, such as depression and functional decline. Currently there are pain treatment guidelines specifically for older adults with dementia. Acetaminophen is the first line drug recommended in pain treatment guidelines for older adults. This feasibility study investigated the effects of acetaminophen in reducing pain behaviors and excess disabilities in community-dwelling persons with dementia. Three participants with severe dementia and their primary caregiver completed 24 daily sessions that allowed the researcher to observe for changes in pain behavior over time following an ABAB study design. During baseline phases (A), pain and excess disabilities were assessed through the completion of daily measurement tools by trained observers and informal caregivers and videotaped behavioral observations of the participant performing an activity protocol (walking, standing, sitting, and lying) designed to elicit pain. Treatment phases (B) consisted of the administration of the treatment drug, Tylenol Arthritis, 1.3 grams every 8 hours while awake in addition to Phase A procedures. Results show that acetaminophen effectively reduced chronic pain behaviors in this sample of older adults with dementia. The frequency of excess disabilities also showed a general decline throughout the study period. Finally, primary caregivers proved to be moderately reliable proxy informants of participants' pain. CHAPTER 1 INTTRODUCTION Older adults in today's society are living longer than they did in previous generations. As a result, there is an increased need for understanding the specific healthcare issues that face this segment of the population. One of the most important, yet often overlooked, healthcare issues that needs to be addressed is that of proper assessment and management of chronic pain. Chronic pain, sometimes referred to as persistent pain, is defined as pain lasting longer than 3 consecutive months in duration. Pain is a complex phenomenon that affects each individual differently. In order to understand the exact nature of an individual's pain experience, it is necessary to elicit his or her own interpretation of this experience. Pain has been described in a variety of ways, such as "an individual's perception of a sensation which is noxious and uncomfortable, and one from which escape or relief is sought" (Weissman & Matson, 1999, p. 31) or as "an unpleasant sensory and emotional experience associated with actual or potential tissue damage, or described of in terms of such damage" (Price, 1988, p. 6). Perhaps the most common definition of pain, suggested by McCaffery and Beebe (1989), has been that pain is "whatever the patient says it is and occurs whenever the patient says it does" (p. 7). Clearly, each definition recognizes the role of the individual in interpreting the experience of pain. Most individuals who experience pain can recognize its onset and can outwardly express this pain, usually through verbal reports. After interpreting their pain, they can seek pain relief through a variety of pharmacologic and/or nonpharmacologic forms. Effective pain management can exert a positive influence in many aspects of an individual's life as pain has been shown to be related to depression and weight loss, and to interfere with activities of daily living, sleep routines, social behavior, functional status, and quality of life (Ferrell, 1995;Villanueva, Smith, Erickson, Lee, & Singer, 2003; Warden, Hurley, Volicer, 2003). However, there are groups of people, such as those with dementia, in whom verbal reports of pain are either unobtainable or unreliable. How does this normal encoding and decoding of the pain experience apply to persons with dementia who, due to the nature of the disease, have barriers in the ability to express themselves? This question serves as the foundation for this study. Background and Significance of the Problem It has only been within the last 15 or so years that researchers have begun to study the assessment of pain in persons with dementia in order to find a way to detect meaningful behaviors indicative of pain (Hurley, Volicer, Hanrahan, Houde, & Volicer, 1992; Villanueva et al., 2003; Warden et al., 2003;). These studies have shown promising results. Recognizing that the presence of dementia complicates pain assessment in these persons due to both the cognitive and verbal deficits that accompany this disease, these researchers have concentrated on identifying observable behaviors such as facial expressions, body movements and postures, as well as vocalizations that are considered to be indicative of pain. The recent progress that has been made in this area has evolved from research studying pain in general. It should be noted that the study of pain has a relatively short history itself, gaining popularity only within the last several decades. First, researchers focused on finding reliable ways to assess pain in the normal adult population. The most commonly used method of pain assessment in this population is through self-report. Simple rating scales allow a person to quantify the intensity of his or her internal painful experience in a measurable way. From these research endeavors, tools such as the Numeric Rating Scale (NRS), the Verbal Descriptor Scale (VDS), and the Visual Analog Scale (VAS) have been developed to measure pain. Each of these tools consists of a scale anchored at 0 (meaning no pain) with increasing numbers that represent increasing amounts of pain, and in the case of the VDS, verbal descriptors that represent incrementally higher levels of pain. Tools like these also allow for measuring the effectiveness of pain treatment by comparing pre-treatment pain intensity ratings to post- treatment pain intensity ratings. As is natural in the course of research, once reliable measures of pain were found in the normal adult population, researchers could begin to expand the study of pain assessment into other populations and to specific causes of pain. Studies have shown that both the VDS and NRS are reliable and easy to use with an elderly population (Gagliese & Melzack, 2003; Herr & Mobily, 1993). Once pain research was expanded to include older adults, researchers found a common theme in most study results in that the 65+ age group had the highest self-reported pain prevalence rates (Gibson & Helme, 1995). Yet, pain in this population is often dismissed as a normal consequence of aging. Recent studies, however, have indicated that this is not the case (Edwards, Fillingim, & Ness, 2003; Gagliese & Melzack; Harkins, 1996; Helme & Katz, 2003). There is evidence that some underlying pathology is involved, with osteoarthritis (OA) of articular j points being one of the most common causes of chronic pain in older adults. In fact, the presence of OA in adults over age 65 is twice that of younger adults (Gibson & Helme). Clearly, research efforts in this area have shown that older adults do experience pain and report higher pain prevalence rates than do younger adults. Building on these findings, researchers have had the basis to propose that pain may affect all older adults experiencing a disease process that typically has pain as a symptom. Even though research studies have found the highest pain prevalence rates in those 65 and older, most investigators believe that older adults underreport the intensity of their pain. This belief has been validated recently by Labus, Keefe, and Jensen (2003) in a review of 30 studies that compared the correlation between self-reported pain intensity and direct observation of pain behaviors, which resulted in only a moderately positive relationship (r= 0.26) on average, across the studies. Several theories have been proposed as to why older adults underreport their painful symptoms. In cognitively intact older adults this could be due to their diminishment of the importance of mild pain, attributing it to the normal aging process (Gagliese & Melzack, 2003; Gibson & Helme, 1995) or because of their fear about what an intensification of painful symptoms might mean (i.e. worsening of a disease process, need for diagnostic testing, or impending death) (Herr, 2002). These theories led the Joint Commission on Accreditation of Health care Organizations (JCAHO) (2001) to recognize pain assessment as the 5th vital sign, requiring healthcare professionals to systematically ask about its presence rather than to rely on the hospitalized patient to volunteer information. There is one fatal flaw in using only self-report to measure pain (whether volunteered by the individual or elicited by another party): this method is not applicable to all populations. For instance, infants can no better report their pain level than can older adults with the compromised cognitive and verbal abilities of dementia. For these reasons, researchers have recently focused efforts on ways to reliably assess pain in persons with dementia. The most valid and reliable way found has been to look for observable behavioral indicators of pain. Keefe (1982) modeled this approach for other investigators when he found that he could reliably assess pain by observing pain behaviors in people with chronic low back pain, looking specifically for guarded movements, bracing, rubbing, grimacing, and sighing. Although the participants in his study were not cognitively impaired, his work in this area allowed researchers to take these behaviors indicative of pain and to observe for them in persons with dementia. In fact, Hadjistavropoulos and Craig (2002) reported findings from studies showing similar behavioral displays of pain observed for both cognitively intact and impaired participants. As mentioned previously, several researchers (Hurley et al., 1992; Villanueva et al., 2003; Warden et al., 2003) have found valid and reliable behaviors that indicate the presence of pain in persons with dementia. While the majority of these research studies have looked at behaviors in persons with dementia residing in nursing homes, it is important to study these behaviors in a community-dwelling sample as well. This is especially true because of the estimated 4.5 million persons with Alzheimer' s disease, more than 70% live at home where nearly 75% of their care is provided by family and friends (Alzheimer' s Association, 2006). Now that pain behaviors have been identified in persons with dementia, research efforts need to shift gears so as to focus on ways to treat this pain. While guidelines exist for pain treatment in older adults in general, there are no specific guidelines for treating pain in persons with dementia, largely because there has been little work done in this area. Rationale and Need for this Study Due to the questionable use of self-report for pain assessment in persons with dementia, researchers have developed and employed behavioral observation strategies in this population with success and have determined that people with dementia do experience pain. The next logical step is to try to do something about the pain. This is an area of research, and clinical practice as well, that warrants further development. In general, pharmaceutical studies have excluded persons with dementia from drug trials, thus preventing this group of vulnerable elders from benefiting from research outcomes (Ancill, 1995). It is necessary to identify effective treatments for pain in persons with dementia because of the possible link between untreated or undertreated chronic pain and excess disabilities, such as depression, declines in functional performance, and sleep and behavioral disturbances (Gibson & Helme, 2000; Villanueva et al., 2003). In relation to persons with dementia, the term 'excess disabilities' has been used to describe "reversible symptoms that are undesirable and temporary extensions of a specific primary disability" (Kolcaba, 2003, p. 3). Because most studies assessing pain in older adults with dementia have been focused on institutionalized persons with moderate to severe dementia, this study will also fill a gap by addressing the pain management needs of persons with dementia who remain residing in the community, typically with informal family caregivers. Several researchers have called for the need of an analgesic trial to treat pain in persons with dementia. Kovach, Weissman, Griffle, Matson, and Muchka (1999) developed the Assessment of Discomfort in Dementia (ADD) protocol which was able to more accurately assess signs of discomfort in persons with dementia, and most importantly, to increase the use of scheduled analgesics to reduce discomfort in these persons. Their findings also support the belief that the pain experienced by people with dementia is amenable to treatment. The authors note that one limitation of these results is that only the use of scheduled analgesics, not prn or "as needed" analgesics, increased. However, prn analgesics, such as acetaminophen, have incredible potential for treating chronic pain, but are often not given because nurses are trained to give prn analgesics in response to verbal reports of pain. Acetaminophen (Tylenol) is one of the most commonly used analgesics for the treatment of chronic pain associated with musculoskeletal conditions. Farrell (2002), reports that pharmaceutically, acetaminophen is the most widely used analgesic and antipyretic agent in the United States and the world. As long as dosing instructions are followed, there are few, if any, risks associated with taking acetaminophen. In fact, two of the maj or benefits of using acetaminophen as an analgesic are its availability over the counter and its relatively low cost. As described below, all of the current treatment guidelines recommend acetaminophen as the first line drug of choice for older adults. The purpose of this study, then, is to evaluate the effectiveness of an acetaminophen trial in reducing the expression of observable pain behaviors and excess disabilities related to chronic pain in community-dwelling persons with dementia, as assessed by trained observers and informal caregivers. Theoretical Framework The specific role of theory in the design and conduct of research studies is that of a foundation from which to organize thoughts and hypotheses. Theories then serve as guides for researchers to use to describe, interpret, and prescribe interventions (Meleis, 1997). Research without theory produces isolated information, but research with theory produces science (Alligood & Tomey, 2002). The choice of a guiding theory is a critical component of the research study design. The Communications Model of Pain (Hadjistavropoulos & Craig, 2002), which addresses the subtleties of pain expression and communication, will serve as the guiding framework for the proposed feasibility study. Within an overall theoretical framework for understanding pain assessment, Hadjistavropoulos and Craig (2002) derived a model of human communication that incorporates both self-report and observational measures of pain. The Communications Model is an appropriate theoretical framework for the proposed research study, as the communication of pain is critical to pain assessment and subsequent treatment. The authors derived the Communications Model from Prkachin and Craig's (1995) model to assist in understanding complex social interactions between persons with pain and their caregivers. This latter model was itself a derivation of Rosenthal's (1982) conceptualization of non-verbal communication. Prkachin and Craig proposed an AB C model focused on facial displays of pain, where (A) is the experience of an internal state that may be (B) encoded into expressive behavior allowing the observer to (C) draw inferences about the nature of the sender' s experience. Hadjistavropoulos and Craig broadened this model to include verbal and non-verbal communication. In their AB C model, (A) represents nociceptive or neuropathic pain processed in the brain and experienced as pain, (B) represents the encoding of pain into either verbal or nonverbal behavior, and (C) represents the potential for this behavior to serve as a message that an observer can decode (see Figure 1-1). The main empirical indicators derived from this model include self-report of pain, the social context of behavior, and observational measures of nonverbal behaviors. Self- report measures are verbal reports that reflect an individuals' subj ective description of the Verbal Programs Automatic Intrapersonal Prcsig message Influences Higher Clarity Mental Obsenrer Bias E te ce E Prc Deodin Higher Mental Automatic Contextual Processing Message Influences Clarity Non-Verbal Programs Figurel-1. Communications model of pain (Hadjistavropoulos & Craig, 2002) experience of pain. The social context of behavior reflects whether this social transaction is instigated by the individual in pain or if it is in response to another' s query. Hadjistavropoulos and Craig (2002) point out that the effectiveness of this social transaction can depend as much upon respondent characteristics as upon the nature of the message itself. For persons with cognitive impairment, self-report measures of pain may not be the most effective way to communicate their pain experience. Rather, observational measures of nonverbal pain (such as facial behavior, limb and torso movements, and changes in habitual patterns of daily living), which focus on signs of pain that are often involuntary and nondeliberate, are more appropriate. One of the key concepts behind this model is that it recognizes that observational measures of automatic expressive behaviors, which focus on involuntary, publicly observable signs of pain, are preferred for measuring pain when the capacity for self-report of subj ective experiences is either not available or when the veracity of self-report is doubted (Hadjistavropoulos & Craig, 2002, p. 553). The authors suggest that this nonverbal behavior usually can be seen as automatic in the sense that the "behavior represents stereotyped, reflexive patterns of response to actual or impending tissue damage that are outside immediate conscious awareness or voluntary control" (p. 553). For the purpose of this feasibility pilot study, designed to measure the effectiveness of an analgesic drug trial in reducing pain and its associated excess disabilities, the presence of chronic osteoarthritic pain in conjunction with normal daily activity will serve as the pain stimulus. Potential influences (including intrapersonal, cultural, and contextual factors) on the expression of pain are no less relevant in persons with dementia than in cognitively intact older adults, and previous research has found that nonverbal pain expressions do include manifestations of affective qualities of the pain experience (Hadjistavropoulos & Craig, 2002) (A, in Figure 1-1). The encoded verbal and nonverbal expressions of the pain experience (B, in Figure 1-1) will then be decoded by the primary caregiver and the trained observer in order to evaluate the effectiveness of the analgesic drug trial (C, in Figure 1-1). Caregivers and the trained observer will also decode the expression of excess disabilities of untreated chronic pain such as depression, declines in functional performance, sleep disturbances, agitation, and behavioral disturbances in order to observe the effect the analgesic trial had on these outcomes. Given the complexities of the pain experience and the issues of validity concerning self-report in persons with dementia, self-report measures alone cannot be expected to capture the full spectrum of the pain experience. This is why Hadjistavropoulos and Craig (2002), incorporate behavioral observations to assess pain in persons with cognitive impairment. Observer reports in conjunction with participant self-report offer the most comprehensive model of pain assessment for persons with dementia. By identifying pain, this model will aid with pain control because the effect of an intervention can be evaluated through subsequent pain assessments following its implementation. Assistance with pain control is needed the most by people who are vulnerable and dependent upon others for generalized care, including persons with dementia. However, caregivers must rely on the availability of reliable and valid information concerning the nature of the individual's distress in order to provide aid, which through this model, can be obtained through behavioral observations. Therefore, the Communication Model will guide the proposed study because it addresses both self-report and observational measures of pain, recognizes the limitations of using self-report measures alone in persons with dementia, and supports a multidimensional approach incorporating observational measures. Specific theoretical constructs, study constructs, and operational definitions are presented in Table 1-1. Statement of Problem and Study Specific Aims As mentioned previously, because pain is so prevalent among older adults, and because older adults with cognitive impairment have compromised ability to reliably self- report their pain, the use of observational strategies for pain assessment in this population is warranted. The use of an acetaminophen trial in this population for pain control is well supported in the literature and thus will serve as the tool for assessing the specificity of observational pain measures. The Communications Model serves as a suitable theoretical framework for designing studies aimed at assessing pain through self-report and Table 1-1. Theoretical and Study Constructs Theoretical Study Constructs Operational Definitions Constructs Verbal Program Pain (Self-reported -Self-reported pain average scores on the NRS pain) PADE: Pain Assessment for the Dementing Elderly ADLs: Activities of Daily Living, including dressing, feeding oneself, and transfers RMPBC: Revised Memory and Problem Behavior Checklist NRS: Numeric Rating Scale observational measures, and for examining the effect of an analgesic trial on reducing pain behaviors and excess disabilities related to untreated chronic pain. Therefore, in order to examine the effectiveness of an acetaminophen drug trial in persons with dementia, this study has 4 specific aims. Aim 1: To investigate the effects of the scheduled administration of acetaminophen (1.3 grams up to three times per day) on self-reported pain intensity and the number of observable pain behaviors exhibited by persons with dementia. Non-Verbal Programs (Observed Pain) Pain -Caregiver reports of pain -Trained Observer reports of pain -Behavioral observation of pain Pain related excess disabilities -Functional performance -Behavioral disruptions -Sum score of PADE Part 1: Physical (observable facial expression, breathing pattern, and posture); and Part 2: Global assessment of overall pain -Sum score of PADE Part 1: Physical (observable facial expression, breathing pattern, and posture); and Part 2: Global assessment of overall pain -Total number of pain behaviors observed during a video-taped activity protocol -Sum score of PADE Part 3: Functional (ADLs) -Total and subgroup scores on RMPBC for memory-related problems, depression, and disruptive behaviors Hypothesis: Regular administration of acetaminophen will decrease the total number of observable pain behaviors (as rated by the caregiver and trained observers) exhibited by the participant, but will have no effect on self-reported pain intensities. Aim 2: To investigate the effects of the scheduled administration of acetaminophen on the frequency and severity of excess disabilities of pain in persons with dementia. Hypothesis: Among persons with dementia, the frequency and severity of excess disabilities of pain, such as depression, agitation, sleep disturbance, behavioral disturbances, and impaired functional performance will decrease from baseline after implementing regular administration of acetaminophen. Aim 3: To determine the reliability of informal caregiver ratings' of the frequency of pain behaviors and excess disabilities exhibited by persons with dementia. Hypothesis: After an initial training session, informal caregivers will display moderate to good reliability with a trained observer in ratings of the frequency of pain behaviors and excess disabilities. Aim 4: To determine the most frequent behavioral indicators of pain displayed by persons with dementia. Hypothesis: Based on findings from similar studies, the most frequently displayed behaviors indicative of pain are expected to be guarding, rubbing, shifting and bracing (Keefe & Block, 1982; Horgas & Elliott, 2005). CHAPTER 2 REVIEW OF THE LITERATURE The following discussion will demonstrate how this proposed study will build on the information gathered by previous researchers in an effort to explore the effects of a pain treatment regimen on observable pain behaviors in persons with dementia. In order to understand the necessity of this treatment regimen, it is important to Birst understand the path laid by previous researchers that supports the need for this study. First, the degree to which pain afflicts both cognitively intact and impaired older adults will be reviewed. Second, the validity of emerging strategies for pain assessment in cognitively impaired older adults will be discussed and analyzed in relation to the more traditional pain assessment strategies for cognitively intact older adults. Included in this discussion will be an overview and definitions of the term "pain behaviors" as well as a section describing the use of informal caregivers as proxy raters. Third, excess disabilities related to untreated and undertreated chronic pain will be explored in order to highlight the need for effective pain management in this population. Fourth, current pain management guidelines for older adults will be discussed, followed by a review of studies highlighting inadequate pain management in persons with cognitive impairment. The proposed study will fill a gap in the existing literature by beginning to investigate pain treatment in cognitively impaired elders. Pain Prevalence Adults ages 65 and over continue to constitute the fastest growing segment of society. Because the number of conditions an individual may develop with pain as a symptom or outcome increases with age, it is appropriate that pain assessment in this population receive attention from both medical and research communities. The prevalence of pain in this population is most often attributed to the presence of chronic musculoskeletal conditions that are common in this age group. A 1998 study conducted by Horgas and Tsai found musculoskeletal conditions, such as arthritis or osteoporosis, to be the most frequently diagnosed painful conditions among nursing home residents. Herr (2002) states that "osteoarthritis alone may be a source of chronic pain in as much as 80% of the population older than age 65" (p. 21). Weissman and Matson (1999) estimate that 80% of community dwelling elderly have at least one chronic disease, that 70% experience some type of pain, and that only 18% take pain medication for chronic pain. Huffman and Kunik (2000) found that 86% of rural community dwelling older adults had pain during the previous year and 59% reported multiple pain complaints. In a study specifically comprised of community- dwelling persons with dementia, only 32% self-reported having current pain (Shega, Hougham, Stocking, Cox-Hayley, & Sachs, 2004), whereas Mitchell, Morris, Park, and Fries (2004) found that 53.4% of their community-dwelling sample with dementia receiving terminal care were experiencing daily pain. Several researchers have found that anywhere from 45% to 83% of elderly residing in nursing homes experience some type of pain (Blomqvist & Hallberg, 1999; Horgas & Dunn, 2001; Weissman & Matson). Ferrell, Ferrell, and Osterweil (1990) found that as few as 29% of a sample of nursing home residents reported having no problems with pain, while 24% reported having constant pamn. Pain Assessment With such a high prevalence of pain in older adults, it is necessary to have successful methods of assessing painful conditions in order to provide individuals with appropriate treatment. This is important because untreated chronic pain may lead to excess disabilities such as impaired physical and social functioning, lowered quality of life, and/or depression (Parmelee, 1993). Eliciting an individual's self-report is the most common method of pain assessment. However, Ferrell, Ferrell, and Rivera (1995) found that approximately 65% of nursing home residents have a specific barrier to the ability to self-report pain due to the presence of cognitive deficits or mental illness. Thus, it is important to consider multiple approaches to assessing pain. Self-Report Several tools exist for measuring self-reported pain, such as the visual analog scale (VAS), the verbal descriptor scale (VDS) and the numeric rating scale (NRS), and research has consistently "supported the use of simple, self-report pain rating scales for chronic pain patients" (Chinball & Tait, 2001, p. 173). Both the VDS and NRS have been shown to be reliable and easy to use with an elderly population (Gagliese & Melzack, 2003; Herr & Mobily, 1993). Typically these scales are presented either vertically or horizontally with values ranging from 0-6, 0-10, or even 0-100, where 0 represents no pain and the highest number represents worst possible pain. When specific questions about pain are presented, the individual is asked to choose a number on the scale that most accurately reflects this pain. However, there are several factors that may complicate pain assessment in older populations and consequently question the validity of self-reports. In cognitively intact older adults, these factors include a tendency for older adults to under-report pain, either because they diminish the importance of mild pain and attribute pain to the normal aging process (Gagliese & Melzack, 2003; Gibson & Helme, 1995) or because they fear what an intensification of painful symptoms might mean (i.e. worsening of a disease process, need for diagnostic testing, or impending death) (Herr, 2002). Older adults may also underreport pain in an effort to maintain a positive self-concept. It is also important to consider that the manner in which a person responds to pain is highly individualized and the impact pain will have depends on both obj ective indicators of its severity and stressfulness, as well as how it is interpreted for personal meaning (Markus & Herzog, 1991). In cognitively impaired individuals, the factors that complicate the use of self- report are vastly different. Snow, Rapp, and Kunik (2005) report that in order to accurately self-report the presence of pain, one must possess the ability to "understand the question in a pain rating, recall pain events in the given time frame, (and) accurately interpret the experience of noxious stimuli as painful events" (p. 22). Communication disorders including reduced receptive and expressive language exacerbate cognitively impaired individuals' inability to report pain (Chinball, Tait, Harman & Leubbert, 2005). Supporting this fact is the finding that chronic pain is less likely to be identified among cognitively impaired nursing home residents than more alert individuals (Sengstaken & King, 1993). It is not clear whether the decreased self-report of pain in persons with dementia is due to actual less pain or a lessened ability to report such pain (Huffman & Kunik, 2000). In either case, measuring pain with self-reports in this population may be unreliable due to the deficits inherent to dementia. Such deficits include compromised cognitive and verbal skills (i.e. memory loss, loss of judgment, confusion, and attention and language deficits) making it difficult for persons with cognitive impairments to recall and/or express pain. These deficits not only affect pain assessment in this population, they also may constrain the ability to assess the effectiveness of pain interventions in that persons with dementia may not be able to reliably report a decrease in pain either, thus potentially leading to further undertreatment of pain in this population. (Chinball et al., 2005). However, it should be mentioned that some researchers have supported the use of self-report in this population. Ferrell et al. (1995) found that at least 83% of nursing home residents were able to successfully use at least one self-report pain scale, even though the study participants' average Mini Mental State Exam score was 12, indicating moderate to severe cognitive impairment. Parmelee (1994) reported that "when questions are phrased simply and straightforwardly, even moderately demented individuals can give valid, reliable information about their pain experience" (p. 289). In a recent study, Pautex and colleagues (2005) found that only 12% of their hospitalized participants with dementia did not understand any one of the four self-report scales they were administered. However, these authors caution that "the ability to complete an assessment does not imply reliability" (p. 527). Because of the conflicting opinions as to the validity of self-report in persons with dementia, it stands to reason that, while it should not be automatically dismissed in this population, neither should self-report be the sole method of pain assessment. Alternative strategies for assessing pain in the cognitively impaired have been the recent focus of research for several investigators. These strategies have revolved around assessing objective behavioral manifestations of pain, and since a maj ority of persons, even those with dementia, were able to use at least one self-reported pain scale, these scales can serve as useful measures against which to compare observational strategies. Observational Strategies If we were to accept McCaffery and Beebe' s (1989) definition that pain is "whatever the patient says it is and occurs whenever the patient says it does," we would have to conclude that if a person does not verbally express that they are in pain then they must not have pain. Marzinski (1991) argues that this definition cannot be used with the nonverbal elderly (including those with dementia) and that behavioral assessment is the only acceptable way of assessing pain in this population. Also, since the use of self-report indices of pain requires verbal skill and comprehension, it is necessary to focus on these alternative measures of pain in persons with compromised verbal abilities. Therefore, to assess pain in persons with dementia, researchers have begun to focus on automatic, less verbal, pain displays. Weissman and Matson (1999) support this focus by finding that in moderately to severely cognitively impaired persons, discomfort is often non-verbally exhibited. The rationale for using observable behaviors as representative of pain is appropriate because observational measures of pain capture behavior that is more automatic and less subj ect to voluntary control, in contrast to self-reports of pain which require higher mental processes and may be more susceptible to purposeful distortion (Hadjistavropoulos & Craig, 2002). This is appropriate for persons with cognitive impairment, where unlike with verbal abilities, there is no barrier to automatic processes. Keefe and Block (1982) were among the first researchers to use direct observation of pain in their work involving patients with chronic low back pain. Their observation methodology consists of instructing the patient to "engage in a standard set of daily activities designed to elicit pain behavior" (Keefe & Smith, 2002, p. 118). The authors observed for the following five overt behaviors that were thought to be consistent with chronic pain: guarded movement, bracing, rubbing the affected area, grimacing, and sighing. Through a series of experiments, these behaviors were shown to be highly reliable and to have good construct, concurrent, and discriminant validity. Thus, this work established several important components of observational methods: 1) pain behaviors can be reliably recorded by trained observers, 2) the frequency of pain behaviors correlates with the patient' s own self-report of pain, and 3) naive observers' ratings of pain also correlate with the frequency of pain behaviors. Through the success of this work, a person' s observable pain behaviors have become essential to understanding the experience and impact of pain (Keefe & Smith, 2002). Because these methods were developed in verbal, cognitively intact populations who can also self-report pain, it is clear that the observation of pain behaviors is crucial to pain assessment in persons with dementia who have compromised cognitive and verbal skills, such as memory loss, loss of judgment, confusion, and attention and language deficits, making it difficult to recall and/or express pain. Advantages of observational methods There are several advantages to using observational methods to assess pain in persons with dementia. These observations are "more objective and accurate measures than can be obtained through anecdotal observation" (Keefe & Smith, 2002, p. 123). Another advantage is that it allows the researcher or clinician to directly observe pain behaviors as well as the effects of chronic pain interventions by comparing the frequency of pain behaviors before and after the implementation of the intervention. If an intervention is an effective pain reliever, the frequency of observable pain behaviors will decline after its implementation. Additionally, another advantage of observational methods is that the researcher can make alternative treatment decisions if intervention is ineffective. A maj or design advantage of obj ective methods is that they allow the researcher to precisely measure the effects of social, psychological, or environmental factors that may influence chronic pain. Keefe and Hill (1985) found pain behaviors to be more likely to occur when patients were moving, such as during walking or transferring from one position to another. Consequently, their research plan included having participants perform an activity-based protocol in order to ensure maximum likelihood of observing pain behaviors. Furthermore, Keefe and Smith (2002) maintain that for persons with musculoskeletal problems (e.g. arthritis), performing daily tasks such as sitting, standing, walking, and reclining for 1-2 minute periods is capable of eliciting pain behavior. Finally, and perhaps most importantly to the proposed research study, observational methods of pain assessment allow a researcher to gather reliable and valid pain (and pain intervention) information in populations whose verbal report of pain is compromised and unreliable, such as is the case for persons with dementia. Limitations of observational methods Despite the apparent utility of observational methods to assess pain in persons with dementia, there are limitations. One limitation is the potential for reactivity (i.e. the tendency for the participant to alter behavior in the presence of an observer) (Elder, 1999; Keefe & Smith, 2002). Due to cognitive decline, persons with dementia are less likely to consciously alter their behavior than are cognitively intact persons (Hadjistavropoulos & Craig, 2002). The potential for observer bias is another limitation of observational methods. Keefe and Smith suggest that this limitation can be minimized by carefully training observers, conducting frequent reliability checks, and holding periodic retraining sessions. Pain Behaviors Measuring pain behaviors in nonverbal persons through direct observation can yield otherwise unobtainable information about an individual's experience of pain. In order to use the term pain behaviors as an outcome measure, it must first be operationally defined in terms of its use in the study. To operationally define pain behaviors and select appropriate measurement tools to fit the study's guiding theory and research questions, other researchers' definitions and measures of pain behaviors must be reviewed. Pain behaviors have been defined as "verbal or nonverbal actions understood by observers to indicate that a person may be experiencing pain and suffering" (Loeser, 2001, p. 19). Specifically, these actions include, but are not limited to, audible complaints, facial expressions, abnormal postures or gait, use of prosthetic devices, avoidance of activities, overt expressions, and verbal or nonverbal complaints of pain, distress, and suffering. This definition captures the most frequently recognized behaviors thought to signify pain and serves as a basis from which other researchers can select the most appropriate behaviors to study pain in their population of interest. All definitions of pain behavior recognize them to be observable, nonverbal behaviors that signify pain to others (Cohen-Mansfield & Creedon, 2002; Keefe & Block, 1982). However, behavioral observation studies differ in their selection of the actual behaviors they choose to study in their population of interest. Fordyce (1976), who popularized the study of chronic pain behaviors, proposed that a thorough behavioral analysis could be just as important as a medical evaluation for evaluating chronic pain and proposed identifying the following as pain behaviors: the report of pain, low levels of activity, taking pain medications, body posturing, and facial expressions (as cited in Keefe, 1982). Subsequent research studies have followed one of two paths; either they are designed to conduct a comprehensive behavioral analysis or they are designed with a specific set of pre-identified behaviors and measures to identify pain. Identifying Pain Behaviors Cohen-Mansfield and Creedon (2002) further developed the definition of pain behavior by interviewing nursing home staff regarding specific behaviors they thought were associated with resident pain. The resulting core group of behaviors was categorized into the following four categories: specific physical repetitive movements (i.e. facial expressions and body postures), vocal repetitive behaviors (i.e. moaning, crying, or screaming), physical signs of pain (i.e. skin discoloration, swollen joints, or a change in vital signs), and changes in behavior from the norm for that person (i.e. changes in mood, movement, or eating patterns). The authors also found that nursing home staff rated vocalizations, rubbing the affected area, and reluctance to move to be among the most prevalent resident pain behaviors. This work led them to develop the following working definition of pain as it applies to evaluating pain behaviors in persons with dementia: "pain is suffering associated with bodily injury or disease, characterized by physical and/or emotional discomfort, which gives rise to a set of distinctive behaviors perceived by caregivers as indicative of that discomfort" (p. 65). Alternatively, Weiner, Peterson, and Keefe (1999) found other salient pain behaviors in nursing home residents (on which residents and caregivers agreed) to be the use of mechanical help, shifting weight when seated, taking or asking for pain medication, moving or walking in a protective fashion, moving extremely slowly, limping, lying down, bracing when seated, clutching the painful area, stiffness, and asking someone to do something to help the pamn. Testing Pre-identified Behaviors Rather than conducting a comprehensive behavioral analysis, other researchers have tested the ability of specific behaviors to signify pain. Keefe and Block (1982) observed for specific behaviors in persons with chronic low back pain by having them perform an activity-based protocol (consisting of activities of daily living) as a pain stimulus. The behaviors they hypothesized would signify pain included: guarded movements, bracing, rubbing, grimacing, and sighing. Not only did the authors find that they could reliably observe these behaviors, but the frequency of these behaviors correlated with patients' self-report, decreased with treatment, were observable by naive observers, and discriminated persons with chronic low back pain from normal and depressed controls. Hadjistavropoulos and colleagues (1998), in an effort to develop an observational strategy to assess pain in persons with dementia, investigated facial reactions to pain using a combination of the extensively tested and well-validated Facial Action Coding System (FACS), nurses' ratings, and student volunteers' ratings. They discovered that facial reactions, especially brow lowering and chin raise, were useful indicators of pain in this population whether rated by FACS trained observers, nurses, or students. Observing pain behaviors can also provide daily documentation of the effects of a pain management program (Keefe & Smith, 2002). Because pain behaviors are readily observable and recordable by trained observers, an observational method of pain assessment is a reasonable way to assess pain in a nonverbal or unreliable population, such as those with cognitive impairments. Researchers have begun to study behavioral assessments by taking these behaviors identified as indicative of pain in cognitively intact older adults and then applying these to cognitively impaired older adults with promising findings (Hurley et al., 1992; Villanueva et al., 2003; Warden et al., 2003). Often in research studies using observational methods to assess behavior, participants are videotaped so that these behaviors may be coded at a later time (Keefe & Block, 1982; Keefe, Crisson, Maltbie, Bradley, & Gil, 1986). In a study conducted by Hurley and colleagues (1992), the authors argue that persons who have lost the cognitive capacity and verbal ability required to communicate must rely on nursing staff and/or other caregivers to recognize, assess, and treat their pain. However, how can these caregivers perform such a task without the knowledge of what they should be looking for to indicate pain in their care recipients? This is why these researchers have focused on behaviors commonly found in people with dementia of the Alzheimer's type who have pain. These behaviors include facial expressions, body movements, certain postures and gestures, and vocalizations. The authors showed that these observed behaviors served as useful external indicators of what the person with dementia experienced internally. Although researchers have begun to make strides in identifying useful ways of assessing pain in persons with dementia, there still exists a critical underassessment of pain in this population leading to large quantities of untreated and undertreated pain. To help abate the chasm, the American Geriatrics Society (2002) put forth a comprehensive framework regarding behavioral indicators of pain. The framework recognizes the following 6 main types of pain behaviors/indicators : facial expressions, verbalizations/vocalizations, body movements, changes in interpersonal reactions, changes in activity patterns or routines, and mental status changes. Underassessment and Treatment of Pain in Persons with Dementia It has been recognized that assessing pain in persons with dementia presents specific challenges, and as a result underassessed and undertreated pain is an unfortunate reality for many people with dementia. An estimated 4.5 million people in the United States have dementia of the Alzheimer' s type; this number is expected to grow to approximately 16 million by the year 2050 (Carter, Rose, Palesch, & Mintzer, 2004). Furthermore, 50% of older adults over age 85 have Alzheimer' s disease with 28% of this population having severe disease (Herr, Bjoro, & Decker, 2006). With prevalence rates of pain in older adults ranging from 45-83% and with no reason to assume that people with dementia experience less pain, it is critical to assess the amount of treatment for pain this population receives. Some factors that may contribute to this inadequate pain assessment and treatment in the healthcare environment are due to "the presence of comorbidities that compete for attention of healthcare staff, a tendency for observers to discount high levels of reported pain, physician fears regarding drug use for pain, and nurse expectations about pain coping...each of these factors is magnified with respect to cognitively impaired older patients" (Chibnall & Tait, 2001, p. 173-4). Several researchers have demonstrated that cognitively impaired individuals receive lower amounts of medication for pain treatment. Horgas and Tsai (1998) found that "nursing home residents with cognitive impairments are less likely to be prescribed and administered pain drugs and receive lower dosages of analgesics when medicated than are their more cognitively intact peers" (p. 240). Also, they found that while 87.6% of participants were prescribed at least one pain drug, only 47% were given an analgesic during the course of their study. Of these drugs, mild analgesics were most commonly prescribed (80%) with acetaminophen being the most frequently prescribed analgesic (59.6%). Chibnall and Tait (2001) found that 61% of cognitively intact elders were prescribed a nonsteroidal anti-inflammatory agent (NSAID) as compared to 48.5% of cognitively impaired elders. Morrison and Siu (2000) found that when "as needed" analgesics are ordered in the elderly, "only 24% to 27% of the prescribed doses are actually administered; this percentage declines with advancing age" (p. 241). Additionally, Pautex and colleagues (2005) found that approximately one-quarter of hospitalized persons with dementia who self-reported pain were not receiving any analgesia. It is imperative that these treatment disparities be overcome. These disparities in pain treatment continue to exist despite the advances made in assessing pain in persons with dementia. Thus, educating those that care for persons with dementia about proper pain assessment and treatment is a crucial component to obtaining appropriate pain management outcomes in this population. Caregiver Assessments According to the Alzheimer's Association and the National Alliance for Caregiving (2004), there are 8.9 million caregivers in the United States caring for someone 50 years or older with dementia. Considering that pain control is "one of the least well managed aspects of professional medical care (and that) untrained family caregivers are routinely given the responsibility of managing pain at home," it is imperative that these caregivers, who have the most day-to-day contact with persons with dementia, possess the skills to assess, treat, and reliably report the pain that the individual experiences (Levine, Reinhard, Feinberg, Albert, & Hart, 2003, p. 20). While many researchers have developed assessment tools and strategies that have proven to be successful and reliable in detecting pain in this population, not all of these tools and strategies are easily implemented or taught to caregivers. In fact in most observational studies, data are coded (either in real-time or from videotaped sessions) by trained research assistants. The researchers who developed the Discomfort Scale for Dementia of the Alzheimer Type (DS-DAT) acknowledge that this scale is most useful for research purposes because it requires more time to train for accurate use than is realistic for caregivers (Warden et al., 2003). In order to translate research into practice as well as to test the construct and social validity of observational methods, these methods must be tested by naive observers (meaning any untrained observer such as a family caregiver, nurse, or nursing assistant) who have the most daily contact with these pain patients. Since caring for persons with dementia is often a time-consuming and stressful job, many caregivers may not have a lot of time to think about assessing the individual's pain. Quick, easily taught, and reliable tools and strategies for caregivers to use to identify pain are needed. In an effort to create a tool that caregivers could reliably use to assess pain in persons with dementia, Villanueva and colleagues (2003) developed the Pain Assessment for the Dementing Elderly (PADE). This tool was tested by a group of caregivers of persons with dementia in long-term care settings. The tool assesses "facial expressions, activities of daily living, and the caregiver' s overall judgment of the resident' s pain" (Villanueva et al., p. 2). The researchers found valid and reliable pain assessments with this group of caregivers after a 1-hour training session. Also with practice, caregivers were able to complete the PADE in 5-10 minutes. Informal Caregivers versus Trained Observers Several research studies have tested their observational methods with naive observers and found promising results. Hadjistavropoulos et al. (1998) compared trained observer' s ratings with the Facial Action Coding System (FACS) to pain intensity ratings by 10 female nurses and 10 female university students after watching videotaped reactions of older adults undergoing venipuncture. All three of these measures were found to be significantly correlated with one another, thus lending support to the validity of informal caregivers' naivee observers') ratings. Likewise, Keefe and Block (1982) asked untrained observers to review videotaped segments and rate subj ect' s pain intensity with three different measures. The total number of pain behaviors (as coded by a trained observer) correlated well with all three ratings from naive observers. Furthermore, Werner, Cohen-Mansfield, Watson, and Pasis (1998) found that family caregivers' and staff members' ratings of senior day care participants' pain severity were highly correlated (r= 0.86). Informal Caregivers versus Participant Self-Report In order to rely on caregiver ratings, it is important to trust that their ratings accurately represent what the care recipient would say themselves. Overall, the reliability of naive observers as proxy raters has not been found to be very strong and even healthcare professionals have been found to be ineffective at both pain assessment and treatment (Herr et al., 2006). Horgas and Dunn (2001) found no significant association between nursing assistants' ratings of resident pain and the resident' s own ratings. Pain was also underdetected by nursing assistants for 37.8% of residents. Most of the research studying caregiver ratings has been done within the context of chronic illness, especially cancer. For cancer symptoms, Nekolaichuk, Maguire, Suarez-Almazor, Rogers, and Bruera (1999) suggest that the best assessment and treatment outcomes for cognitively impaired patients may be obtained by selecting caregivers who best understand and represent the patient's experience. These authors also state that both healthcare providers and other significant caregivers tend to overestimate symptoms of depression and anxiety and they underestimate performance status and quality of life. Nekolaichuk et al., through a study comparing cancer patients, their family caregivers, and nurses' ratings of patient' s symptoms, found that all three respondents were reasonably consistent with one another when ratings were averaged across all symptoms at two measurement times. However, for any one rater on one occasion, reliability estimates were less than 0.70 for most symptoms. Through a review of studies examining ratings of pain by cancer patients and their caregivers, Lin (2001) found that most studies used correlation statistics which may overestimate the actual relationship because this statistic does not take into account error variance. The studies reviewed tended to show that patient and caregiver ratings were significantly correlated and accurate, with some studies showing a tendency for family caregivers to overestimate, rather than underestimate the patient' s pain. In her own study comparing cancer patients' and family caregivers' ratings of pain intensity and interference, Lin found that the Pearson's correlation, intraclass correlation coefficients, and the kappa statistics ofinterrater agreement were statistically significant, lending more support that the caregiver ratings were accurate representations of the patient' s own ratings. However, the author cautions that the kappa values for worst pain (.25), least pain (. 18), current pain (.20), and average pain (.25) do not represent good agreement beyond chance and that family caregivers tended to overestimate patient's pain severity. In a study to identify behaviors signaling pain in nursing home residents, Weiner et al. (1999) also examined the congruence between self-reported resident and caregivers' (nurse and family) ratings of pain behaviors and pain intensity. The authors found that while both the nurse and family caregiver felt similarly confident in their ability to rate the residents' pain, both had poor agreement with the residents' own ratings regarding identification of pain behaviors and rating pain intensity. Interestingly, over 75% of nurses and only 44% of family caregivers felt that the residents showed their pain outwardly and 71% of nurses felt that cognitive impairment made assessments more difficult. Cremeans-Smith and colleagues (2003) found that 55% of spousal caregivers rated their wife's recent osteoarthritis pain severity in agreement with her own rating. Again, for those dyads who disagreed, the caregiver was more likely to overestimate, rather than underestimate, their spouse's pain severity. In the only study to date studying the congruence between caregiver and patient report of pain in community-dwelling persons with dementia, 59% of the 150 patient-caregiver pairs, agreed on the presence or absence of patient pain. Again, caregivers were more likely to estimate their care recipients' pain severity (Shega et. al, 2004). There has been no standard set to gauge how accurate caregiver ratings must be in order to be considered as valid proxies. The maj ority of research utilizing proxy reports from caregivers has shown only moderate agreement with the care recipient, with caregivers tending to overestimate levels of pain and excess disabilities related to chronic pain (such as depression, anxiety, and functional disability) (Cremeans-Smith et al., 2003; DeBettignies, Mahurin, & Pirozzolo, 1993; Lin, 2001; Nekolaichuk et al., 1999; Seltzer & Buswell, 1994; Weiner et al., 1999; Zanetti, Geroldi, Frisoni, Bianchetti, & Trabucchi, 1999). These more negative assessments (i.e. higher ratings of pain intensity) by caregivers have been attributed to the distress and burden they experience when caring for a close relative with dementia, as well as to a greater reported degree of concern over patient disabilities (DeBettignies et al.; Seltzer & Buswell; Zanetti et al.). In situations where relying on caregiver proxies is necessary, they should be seen as reasonably accurate (Lobchuk, Kristjanson, Degnet, Blood, & Sloan, 1997). While the research in this area demonstrates that caregivers' accuracy in assessing patient symptoms is not in total agreement with care recipients, caregiver report is the best option for assessing pain and its excess disabilities in persons with dementia. And, as has been shown, although caregivers' ratings are not perfect, they tend to err on the side of overestimating rather than underestimating the patients' own ratings. However, it should be noted that most of the studies that asked caregivers to give proxy reports for the care recipient did not base this proxy report on direct observations of the care recipient, but rather on global perceptions. Among the many factors that may contribute to caregivers' inability to report accurate proxy ratings, the most important to consider is the tremendous stress and burden that caregivers of persons with dementia are under. Many of these caregivers are themselves elderly, 63 being the average age of caregivers caring for someone age 65+, and one-third of these caregivers are in fair to poor health (Administration on Aging, 2004). Caregivers are at risk for "burnout" as the average person with dementia requires 69-100 hours of care each week (Kahn et al., 1998). Often behavioral disturbances in persons with dementia result in high levels of caregiver burden and are frequently a precipitant to nursing home placement (Kahn et al.; Lantz, 2003). Czaja, Schulz, Lee, and Belle (2003) found that functional limitations and behavioral problems of the care recipient were the primary stressors for the caregiver. This is especially true for caregiving wives, who are more likely to experience caregiving distress and report more frequent problem behaviors in their spouse than caregiving husbands (Ingersoll-Dayton & Raschick, 2004). Some researchers have proposed that these behavioral disturbances and functional limitations may be related to untreated chronic pain (Buffum, Miaskowski, Sands, & Brod, 2001; Douzjian, Wilson, Shultz, Tapnio, & Blanton, 1998; Ferrell, 1995; Gibson & Helme, 2000; Lantz; Kahn et al.; Villanueva et al., 2003). Therefore, it is necessary to educate caregivers in both the importance of, and strategies to, assess and treat pain in their care recipients. Strategies to Strengthen the Validity and Reliability of Caregiver Report The first step in overcoming methodologic problems related to proxy report is by using observational methods (either in combination with the caregiver report or training the caregiver to observe care recipient behavior). There are also strategies to help increase the reliability of caregiver ratings. An initial strategy is to have the caregiver be the person who best understands and represents what the care recipient experiences; this would most likely be the primary caregiver who lives with and is most familiar with the care recipient (Bohac, McNeilly, & Folks, 2003; Nekolaichuk et al., 1999). The most useful way to increase untrained observers' ratings is through education and training. Training does not have to be as intense as is required for formally trained observers. For example, when creating the Pain Assessment for the Dementing Elderly tool, Villanueva and colleagues (2003) obtained valid and reliable caregiver pain assessments after a 1- hour training session. Researchers should also educate caregivers to time their observations with potentially painful events in order to maximize the potential for observing pain behaviors (Weiner et al., 1999). Bohac and colleagues suggest increasing caregiver familiarity with the research instruments being used (with practice sessions) to ensure greater reliability and validity of caregiver reports. Finally, as mentioned previously, caregiver reports have been shown to be influenced by the amount of burden and distress the caregiver feels. Perhaps a strategy aimed at decreasing caregiver burden and distress would in turn, positively influence the reliability and validity of their reports of care recipient behavior. Excess Disabilities related to Dementia and Chronic Pain Investigators whose research focuses on persons with dementia are well aware of the excess disabilities (behavioral disturbances) that accompany disease progression. Research in this area can hardly be conducted without acknowledging the complex behavioral problems that accompany this disease (Teri et al., 1992). Lantz (2003) reports that 65% of community dwelling older adults with dementia have at least one disruptive behavior and 40% have three or more. Rowe, Straneva, Colling, and Grabo (2000) found that 86% of persons with dementia evaluated at a community-based clinic had at least one disruptive behavior, with 86% being verbally agitated, 55% being physically agitated, 33% being verbally aggressive, and 21% being physically aggressive. Overall, the behavior problems commonly recognized to be associated with dementia include: depression, agitation, aggression, anxiety, delusions, impulsivity, repetitive actions, restlessness, disinhibition, resistiveness, physiological risks, impaired cognitive function, difficulties with performance of activities of daily living (ADLs) and instrumental activities of daily living (IADLs), pacing, vocalizations, altered sleep routines and time spent in bed, wandering at night, altered social behavior, and increased health care utilization and costs (Albert, et al. 2001; Ancoli-Israel et al., 2003; Herr et al., 2006; Hinton, Haan, Geller, & Mungas, 2003; Kahn et al., 1998; Lantz, 2003; Opie, Doyle, & O'Connor, 2002; Teri et al). Furthermore, Logsdon, Gibbons, McCurry, and Teri (2002) report that caregivers for persons with severe dementia (MMSE <10) reported higher rates of disruptive behaviors and worse physical functioning in their care recipients. The presence of these behavioral problems and limitations in functional performance is a source of stress and burden to the caregiver. Also, behavioral problems in this population can negatively influence quality of life as well as lead to increased social isolation and risk of falls (Horgas & Margrett, 2001). While previous researchers have targeted reducing these outcomes with interventions such as light therapy (Ancoli- Israel et al., 2003), psychosocial and behavioral interventions (Czaja et al., 2003), or administration of psychotropic medications (Opie et al., 2002), other researchers have found more promising results by assessing and treating underlying chronic pain (Douzjian et al., 1998; Opie et al.). The rationale for this approach stems from the observation that many of the behavioral problems and functional limitations that afflict persons with dementia are also seen in cognitively intact persons who experience chronic pain. For instance, Vlaeyen, Van Eek, Groenman, & Schuerman (1987) identified nine components of chronic pain behavior: anxiety, attention seeking, verbal pain complaints, medication use, general verbal complaints, distorted posture and mobility, fatigue, insomnia, and depressive mood. Ferrell (1995) classifies depression, decreased socialization, sleep disturbances, impaired ambulation, and increased health care use and costs as complications from untreated chronic pain. Thus, there is considerable overlap between excess disabilities related to dementia and those related to chronic pain. Ferrell (1995) emphasizes that in order to maximize mobility and independence, health care providers and caregivers need to assess ADLs because evaluating functional status is an important measure of the success of pain management. Studying pain from osteoarthritis (OA) in rural adults age 45 and older, Jordan, Linder, Renner, and Fryer (1995) found that hip OA, hip pain, knee OA, and knee pain were all associated with self- reported and observed limitations in functional performance. Even the American Geriatrics Society's panel on persistent pain in older adults recommended in 2002 that recent changes in function, verbalizations suggestive of pain, and nonverbal pain behaviors should be assessed in all chronic pain patients. For those with dementia, caregiver reports should be solicited. Several researchers have begun to study the link between pain and specific problem behaviors in persons with dementia. Buffum et al. (2001) found significant positive correlations (r= .50, P= .003) between measures of discomfort (Discomfort Scale) and agitation (Cohen-Mansfield Agitation Inventory) in nursing home residents with dementia, suggesting that agitated behaviors could indicate painful sensations. The authors suggest that although the correlation was strong, the discomfort behaviors may be related to other things, such as hunger or being wet from urinary incontinence, and the best way to evaluate the role of pain in prompting these behaviors would be to administer an analgesic and then observe for changes in behaviors. Three groups of researchers have done this. In an effort to reduce challenging behaviors in nursing home residents with dementia, Opie et al. (2002) initiated 4 distinct groups of interventions: psychosocial strategies, nursing approaches, psychotropic medications, and management of pain. Interventions were tailored to each of the 99 participants, with the maj ority receiving more than one intervention. Changes in pain management were recommended for 18 participants, and while the authors do not report the effects of pain management alone in reducing challenging behaviors, overall their intervention strategies showed a statistically significant improvement in the target behaviors of restlessness, physical aggression, verbal disruption, and inappropriate behavior. The limitation of this study is that the effect of any one intervention on reducing challenging behaviors is unknown; however, their results argue that individually-tailored interventions are effective in reducing challenging behaviors in persons with dementia. Another group of researchers was able to show a strong relationship between treating pain and decreasing problem behaviors. Douzjian et al. (1998) observed that demented residents who were admitted to a skilled nursing facility (SNF) after being discharged from the hospital were often receiving antipsychotic medications without having any documented psychological history. Since the use of antipsychotic medications is strictly regulated in SNFs, they hypothesized the residents' behavior may be linked to pain, so the authors devised what they term an "informal study" by trying to initiate a pain treatment regimen to replace the use of antipsychotic medications. Eight residents were placed on a pain treatment regimen of acetaminophen 650 mg three times a day. Five (63%) of these residents showed a decrease in the number of problem behaviors and the staff was able to discontinue four orders for antipsychotic drugs and two for antidepressants. Most impressively, the authors were able to decrease the facility's use of antipsychotic drugs from 20% to 0% over the study period. This is clinically significant because antipsychotic drugs often work inconsistently and can lead to adverse events such as a higher risk for falls (Opie et al., 2002). In perhaps the most promising study to date, Chinball and colleagues (2005) found that 4 weeks of acetaminophen (3,000 mg/d) was more effective than placebo in increasing participants' engagement with their environment. Specifically, while taking the acetaminophen, participants spent significantly more time in "social interactions, engaged with media, talking to themselves, engaged in work-like activity, and experiencing unattended distress...and spend less time in their rooms, less time removed from the nursing home unit, and less time performing personal care activities" (p. 1921). However, there was no effect of acetaminophen on agitation, emotional well-being, or as needed psychotropic medication use. Similarly, Allen and colleagues (2003) also found that nursing home residents who had received analgesic medication during the course of their 4-week study were more active than those who did not receive any analgesia. Pain Management and Treatment Guidelines Now that researchers have begun to demonstrate successful ways to assess pain in persons with dementia and a correlation between chronic pain and behavioral problems, the next logical step is to further examine pain treatment strategies in this population. Treatment guidelines for the general population with chronic mild to moderate pain state that nonpharmacologic strategies are the appropriate first line defense. These strategies include patient education, self-management programs, and exercise, among others. For obvious reasons, these strategies are not appropriate for persons with impaired memory function. The American Geriatrics Society's 2002 Management of Persistent Pain in Older Adults state that analgesic drugs are safe and effective for use by older people. They recommend following the adage "start low and go slow" in order to assess each individual's response to a given medication. The guidelines report that for most patients "with persistent mild to moderate musculoskeletal pain respond favorably to around-the- clock doses of acetaminophen" (p. S213). This is advantageous because this treatment is capable of providing satisfactory pain relief and has a much lower risk of side effects than NSAID drugs. The maximum dosage for acetaminophen in these guidelines is 4000 mg/24 hours, with 4-6 hour dosing (not to exceed a daily amount of 4 grams), recognizing that the half-life for acetaminophen ranges from 1 to 4 hours. Kovach et al. (1999) recommend that for persons with dementia who cannot report pain symptoms, "the use of a limited trial of analgesics in response to a broad array of physical and behavioral symptoms" may be a good method of determining whether or not the person is experiencing pain and may provide the most appropriate intervention for the patient (p. 417). Ferrell (1995) reports that acetaminophen is the most commonly prescribed analgesic for elderly nursing home patients, and supports its use as the preferred analgesic choice for patients without substantial inflammation because of its lower side- effect profile. In fact, in a study conducted by Chinball and colleagues (2005), where 3,000 mg of acetaminophen was administered in 3 daily 1,000 mg doses for 4 consecutive weeks, there were no adverse events related to the acetaminophen and all post-study liver function tests were normal. Likewise, in a sample of 519, only one person had an allergy to acetaminophen and one other had a drug interaction with acetaminophen (Buffum et al., 2004). The American College of Rheumatology (2000), the European League Against Rheumatism (2003), and the British Medical Journal (2000) all agree that for persons with OA experiencing mild to moderate joint pain, the appropriate first line pharmacologic agent is acetaminophen (Tylenol). This is in agreement with the guidelines derived specifically for older adults by the American Geriatrics Society (2002) described above. It should be noted, however, that none of these guidelines gives specific recommendations for persons with dementia. However, the general scientific opinion is that dementia does not affect the experience of pain, just the memory of it. Therefore, it follows that these guidelines should be applicable and safe in older adults with dementia. In summary, further investigation into pain management in older adults with dementia warrants attention due to the promising results found by previous researchers. Because the prevalence of musculoskeletal pain in older adults in general is large and there is no reason to believe that persons with dementia do not experience equivalent levels of pain as cognitively intact older adults, it is imperative that this pain be assessed and managed effectively. Several methods of assessing pain behaviors in cognitively impaired older adults have been shown to be reliable and valid. However, many need to be evaluated in terms of showing sensitivity to treatment effects. The creators of the Pain Assessment for the Dementing Elderly (PADE) state that a limitation of their validation of this tool was that it lacked an intervention phase; specifically, they state that it would be of interest to see if PADE scores were to decrease in response to analgesic administration (Villanueva et al., 2003). The other area of considerable importance is the relationship between chronic pain and behavioral problems as well as limitations in functional performance in persons with dementia. It is these excess disabilities that place a significant amount of stress and burden on the caregiver. If treating the underlying chronic pain condition can have a positive effect on behavioral problems and functional limitations, this would be of clinical significance to the caregiver as well as lead to a better overall quality of life for the person with dementia. Clinicians have begun to recommend that behavioral problems in persons with dementia need to be considered in the multiple contexts from which they can develop, many of which are other identifiable problems, such as pain, and that treating these behavioral problems should always begin with correcting any underlying medical conditions and alleviating any discomfort (Kahn et al., 1998; Lantz, 2003). Therefore, the proposed pilot study aims to evaluate the effects of an analgesic trial in reducing observable pain behaviors as well as reducing behavioral problems and functional limitations common in persons with dementia. In order to accurately assess and evaluate the effects of any intervention, it is imperative to know the patient' s baseline level of behavior, compare any changes in behavior over time, initiate interventions, and then evaluate the effect of these interventions with the same assessment tools (Buffum et al., 2001). Summary A review of the literature exploring the assessment and treatment of chronic pain in persons with dementia provided support for the need for an analgesic trial to reduce pain behaviors in this population. Furthermore, this review highlighted several issues which were addressed in the present study, including the use of caregiver assessments of care recipient pain and the relationship between chronic untreated pain and excess disabilities. Based on the recommendations of the leading treatment guidelines and the paucity of studies investigating drug treatment for pain in older adults with chronic pain and dementia, acetaminophen was the appropriate treatment choice for this feasibility study. The literature review also supported investigating the effect of around-the-clock dosing of acetaminophen on the frequency and severity of excess disabilities in persons with dementia. Finally, the use of observational measures to assess change in pain behaviors after the implementation of a pain treatment intervention was strongly supported in the literature. CHAPTER 3 IVETHOD S Study Design This feasibility pilot study was conducted using an ABAB single-subj ect design. Through this design, the research participant served as his or her own control through the use of repeated observations which controls for inter-subj ect variability (Crosbie, 1995). This particular design was used in order to investigate the feasibility and effect of implementing a scheduled analgesic trial on participant pain behaviors and excess disabilities related to chronic pain, and also attempted to capture the significance of an individuals' change in behavior. The ABAB design was able to examine the effects of acetaminophen by alternating a baseline condition (phase A- no acetaminophen administered) with the intervention condition (phase B). These phases were then repeated to complete the four phase design. In this type of design, the effects of the intervention are evident if measures of participant pain behaviors and excess disabilities decrease during the first intervention phase, "revert to or approach original baseline levels of performance when treatment is withdrawn, and improve when treatment is reinstated in the second intervention phase" (Kazdin, 1982, p. 110). This longitudinal, feasibility design to study the effects of acetaminophen in persons with dementia is needed as there is a paucity of literature on the effectiveness of acetaminophen to alleviate pain in community-dwelling persons with dementia. Thus, the study was intended to provide important preliminary data about this important, but understudied phenomenon. One benefit of SSD is that it allows for a relatively small N, generally at least three subj ects are required, which is convenient when the target population is difficult to recruit for research studies (as was the case for community-dwelling persons with dementia and their caregivers) (Kazdin, 1982). SSD analysis has traditionally consisted of visual inspection of the graphic display of data, with occasional use of statistical tests to evaluate the reliability of findings, especially in cases with variable or trend-affected baselines (Greenwood & Matyas, 1990; Kazdin). This allows the researcher to view an individual's data points during the baseline observation phase and from the intervention phase on an ongoing basis. Then, judgment for when to switch phases, as well as about an intervention' s effects, is made based on the overall pattern of the data (Hoerster, Hickey, & Bourgeois, 2001; Morgan & Morgan, 2001). A drawback of SSD is that the number of measurement occasions needed is dependent upon how quickly baseline stability is established in order to begin the treatment phase and then how long it takes to observe a treatment effect. Thus, the length of a phase depends on the nature of the phenomenon being investigated. If it is relatively stable in nature and responds quickly to an intervention, the study phases can be relatively short. Likewise, if it is more labile in nature and slower to respond to an intervention, the phases will need to be longer to observe baseline stability and the intervention's effect. Since pain is generally somewhat variable in nature, and the effectiveness of acetaminophen on reducing chronic pain in persons with dementia is unknown, this study departed from the traditional methods for changing phases. Rather than observing for stability in pain behaviors in the initial baseline phase before introducing the intervention, the treatment was give after 8 days of baseline data collection. This allowed for the collection of information on the intra-individual variability of pain behaviors in persons with dementia. Also, due to the unknown effect of acetaminophen or how long it may take to show any effect, the initial intervention phase was created to last the same number of sessions (8) as the baseline phase in order to give the design equivalence. SSD has not been widely employed in nursing research, in part due to some of the criticisms about determining the effect of an intervention through visual inspection. Despite the reliable criteria for visual inspection (such as comparing the magnitude and rate of change to the variability of performance during a phase, the duration of the phase, and the consistency of the effect across phases or baselines), critiques of SSD persist (Kazdin, 1982). One criticism has been that visual inspection of data caters to interventions with strong effects (Greenwood & Matyas, 1990). If an intervention's effects are strong, there will be convincing graphical evidence that the intervention was responsible for the change in behavior, thus virtually eliminating any chance of a Type II error. However, when an intervention's effects are more subtle or more variable, there is a high risk for Type I error as the graphical analyses of this intervention' s effect may find that the intervention failed to effect any change in behavior (Crosbie, 1995). This is, when statistical procedures may be employed in order to test the strength of the intervention's effects when they appear weak or non-existent based on graphical analyses. For this proposed feasibility study, a SSD provided the needed combination of flexibility in sample size and measurement occasion requirements while paying rigorous attention to the nature of individual change. Because the effect of acetaminophen in reducing pain and its related behaviors in community-dwelling persons with dementia is unknown, a SSD, with ongoing analysis of graphical data, allowed the researcher to observe effects immediately. While this design does not allow for comparisons across individuals and does not rely on complex statistics for analysis, it should be noted that the proposed study is a feasibility study, and this design allowed the researcher to carefully follow the trend of the data and adjust the study phases according to participant performance. This design allowed the researcher to examine daily change in behavior and to carefully format "the nature of change in the single case before seeking similarities and differences across cases" (Nesselroade, 1984, p. 275). Sample Five community dwelling older adults (over age 65) with dementia and their primary caregiver were recruited and enrolled to participate in this study in order to obtain a target sample of three pairs to complete the study. Two pairs (40%) were withdrawn by the Principal Investigator. Informed consent was obtained from the legally authorized representative of all participants with dementia and assent was obtained from all participants with dementia. The primary caregiver for each participant also signed an informed consent form to participate in daily sessions with the Principal Investigator over the course of a 4-6 week period, as well as to complete several daily measurement tools and administering multiple doses of acetaminophen on a daily basis according to study protocol. Each caregiver was also the health care surrogate and legally authorized representative of the participant with dementia. Each caregiver and their care recipient also resided together in the same dwelling. Recruitment Procedures Individuals with dementia and their primary caregivers were recruited from the West Central and North Central Florida community. Initial recruitment included working with local chapters of the Alzheimer' s Association through which the PI gained access to present the study to members of Alzheimer' s Caregiver support groups. Flyers were posted and announcements were placed in local church bulletins. The PI also gained permission to attend the University of Florida' s neurology clinics and the Memory Disorders Clinic to identify possible participants. Also, local family practice and geriatric physicians were asked to advertise the study to potentially eligible participants. Inclusion and Exclusion Criteria for Participants with Dementia Inclusion criteria consist of: 1) previously established diagnosis of probable Alzheimer' s Disease (AD) or other related dementia by a physician, 2) score of 23 or less on the Mini Mental State Exam (MMSE), 3) age 65 or over, 4) diagnosis of a painful musculoskeletal condition, such as osteoarthritis (OA), 5) able to swallow an oral pill (as assessed by the principal investigator), and 6) not currently taking prescribed analgesics on a regular basis. As advocated by previous researchers, persons receiving routine aspirin for cardiovascular prophylaxis (81 mg/day) were allowed participate in the study and continue this regimen (Chinball et al., 2005). Participants were excluded if they are unable to walk or have a hypersensitivity to acetaminophen that would prohibit them from taking this drug. Also, persons with major health problems, especially any kidney or liver disease, were excluded. Inclusion and Exclusion Criteria for Caregivers Caregivers recruited for this study were required to be the primary caregiver to the participant with dementia as well as be able to read and write the English language fluently. To be considered the primary caregiver, this person must have provided direct care to the individual with dementia and have had daily contact with him/her so that they were able to observe the participant over the course of the day. Participant Pair 1, Caregiver 1 (CG 1) and Participant 1 (P 1) The first participant pair consisted of a 58 year-old daughter (CG 1) who was the primary caregiver for her 79 year-old mother (P 1). Both were Caucasian and educated at the high school level. The mother was widowed and the daughter was married. Both resided in the same house with the daughter's husband and her adult son. At the time of the study, the daughter had been the primary caregiver for her mother for approximately one year and was receiving respite care one afternoon each week. Participant Pair 2, Caregiver 2 (CG 2) and Participant 2 (P 2) The second participant pair consisted of a 75 year-old wife caring for her 83 year- old husband. Both were Caucasian with a technical or trade school education. They resided together in an independent dwelling with no other residents. At the time of the study, the wife had been functioning as a caregiver for approximately 2 years and her husband attended an adult day care group for approximately 5 hours a day on weekdays. Participant Pair 3, Caregiver 3 (CG 3) and Participant 3 (P 3) The third participant pair consisted of a 67 year-old daughter caring for her 93 year- old mother. Both were Caucasian. The mother had an 8th grade or less education and the daughter had attended some college. The mother was widowed and the daughter was divorced. They resided together in the daughter' s home with no other residents. At the time of the study, the daughter was employed full-time as an administrative program assistant and had an independent aid supervise her mother during the day. She had been serving as her mother' s caregiver for approximately 2 years. Participant Pairs Withdrawn from the Study As mentioned previously, two of the five pairs that were enrolled in this study were withdrawn before completion by the PI. One of these pairs, an African-American daughter caring for her mother, completed the first two study phases. However, during the initial treatment phase (described below) the caregiver was unable to administer the treatment drug according to protocol (i.e. the treatment was administered only one time each day instead of the minimum of two doses to satisfy the study protocol which aimed to provide continual pain relief). The second pair withdrawn consisted of a Caucasian wife caring for her Caucasian husband. This pair completed only the first 6 sessions of the study (the first phase, described below, was 8 total sessions). This pair was withdrawn after it became clear that the participant with dementia was no longer providing assent to participate in the study and was aggravated at being asked to perform the activity protocol by his caregiver. Measures Comprehensive Intake Assessment Once the caregiver expressed willingness to participate in this study, the principal investigator arranged an initial meeting in the residence where the individual with dementia receives care to explain the study, obtain informed consent, and to screen participants for eligibility based on the inclusion and exclusion criteria. Caregivers provided their own consent for study participation and also provided proxy consent for the person with dementia, as each caregiver was also the health care surrogate and legally authorized representative for their care recipient with dementia. (Copies of the informed consent forms are provided in Appendices B and C). Assent from each individual with dementia was also obtained at each session. Each participant with dementia was given the Mini Mental State Examination (MMSE) by a trained research assistant (Folstein, Folstein, & McHugh, 1975). A score of 23 or less was used as a cutoff to verify cognitive impairment in participants with a diagnosis of dementia. Caregivers' cognitive status was judged to be intact by the PI through conversation and training procedures during this initial session; caregivers were not asked to complete formal cognition tests. Demographic information, including age, sex, gender, marital status, education level, and race was collected. Also a medication log sheet was used to compile care recipients' prescribed and over the counter medications. Both caregivers and individuals with dementia were also screened for depression with the 15 item form of the Geriatric Depression Scale (GDS) (Sheikh & Yesavage, 1986), or the Center for Epidemiological Studies Depression Scale (CES-D) if they were less than 65 years of age. The Dementia Rating Scale-2 (DRS-2) (Mattis, Jurica, & Leitten, 2001) was administered to individuals with dementia in order to gain more accurate baseline measures of cognition as well as to thoroughly describe sample characteristics. Caregiver Ratings of Pain Behaviors and Excess Disabilities The Pain Assessment for the Dementing Elderly (PADE) was used to assess pain behaviors and pain-related disability. This tool was originally designed to assess pain in older adults with dementia residing in long-term care facilities (Villanueva et al., 2003). This tool was divided into three parts (i.e. subscales) and contained 24 total questions. Responses were provided using either a Likert scale (scored 1-4, or 0 for a skipped or N/A answer) or multiple choice (scored 1-4) to elicit information about the effects of pain on physical, global assessment, and functional areas. Part I of this measure assessed physical manifestations of pain such as observable facial expressions, breathing patterns, and posture. Responses on Part I were scored so that higher scores were representative of higher distress. Part II consisted of one question which allowed the caregiver to rate, on a Likert Scale (none to severe), their global assessment of the individual's level of pain at the time of observation. Part III assessed functional performance and allowed caregivers to rate the individual's performance of ADLs. Part III was scored so that higher scores represented less independence and higher difficulties with ADLs. (See Appendix D for a copy of the PADE tool). Because the directions for the PADE tool asked caregivers to base their responses to Parts I and II of this measure on a 5- to 10-minute observation of the subj ect, caregivers were asked to complete these sections after observing the participant with dementia perform a 10-minute activity protocol (described below). This measure was chosen because it was designed for caregivers to use to rate global assessments of pain, pain behaviors (e.g. facial expressions, breathing, posture), and ability to perform activities of daily living. Another reason for including this measurement tool is that it has shown significant correlations with measures of agitation in the elders with dementia and has proven to be quick and easy for caregivers to use. PADE intraclass correlation coefficients (ICC) were generally high (interrater ICC= 0.54- 0.95; stability ICC= 0.70-0.98; internal consistency (a) =0.24-0.88). The PADE was also significantly correlated with the Cohen-Mansfield Agitation Inventory (r= .30-.42). The PADE was also able to differentiate individuals who were judged to suffer from clinically problematic pain from those who were not. The Revised Memory and Behavior Problems Checklist (RMBPC) was designed to allow caregivers to measure problems with memory, depression, and disruptive behaviors exhibited by the person with dementia (Teri et al., 1992), and it is the most commonly used tool to measure behavior problems in community-dwelling individuals with dementia (Allen, Kwak, Lokken, & Haley, 2003). This measure consists of 24 items and asks the caregiver to report the frequency of the problem in question and their reaction to this problem (i.e., how much it bothered them) over the preceding week. For the purposes of the proposed study, this was modified slightly to have caregivers report the frequency of and their reaction to problem behaviors for each day of the study. (See Appendix D for a copuy of the RMBPC). The RMBPC has shown good overall reliability for patient behavior (a= 0.84) and caregiver reaction (a= 0.90). The measure also has been shown to be valid as the depression subsection was positively correlated with Hamilton Depression Rating Scale (r=.44) and the memory-related problems subsection was positively correlated with the MMSE (r=.48). Caregivers also completed a third, brief, free-response questionnaire composed of the following two open-ended questions: 1) How typical was this day compared to most days? (i.e. where there any doctor' s appointments, visitors, etc.) and 2) What behaviors did you notice that were most representative of pain in the participant today? (Please include anything that signaled pain to you, even if it was not part of the other two questionnaires). (See Appendix D for a copy of this questionnaire). These measures were selected because they were designed for use by caregivers to rate specific behaviors related to the study's specific aims. Together, these measures also addressed each of the 6 main types of pain behaviors described in American Geriatrics Society's (2002) pain behavior framework. Daily administration of the PADE allowed for frequent assessment of participants' pain behaviors in order to examine daily variations of these behaviors (during the observation phases) and the effectiveness of acetaminophen in reducing these behaviors (during the treatment phases). Also, because measures of ADLs were recorded in this tool, the effectiveness of acetaminophen in increasing participants' functional ability was also assessed. Daily assessments with the RMBPC allowed the researcher to evaluate daily variation in participants' mood and problem behaviors (in observation phases) as well as the efficacy that reducing pain has in secondarily improving mood and decreasing problem behaviors in persons with dementia (in treatment phases). See Table 3-1 for a summary of study constructs and measures. Also, by assessing the typicality of the day, any potential influences on participant behavior or caregiver burden, such as the increased stress of taking the participant to a doctor's appointment, were noted. Also, it was thought that asking caregivers what pain behaviors they are most noticing would allow the researcher to establish which behaviors are most useful in signaling pain to community-dwelling dementia care providers. Trained Observer Ratings of Pain Behaviors The trained observer (i.e., the Principle Investigator or a trained research assistant who was familiar with administering the study tools and who was physically present at the data collection session) also completed the PADE (Parts I and II) daily as well as the frequency portion of the RMPBC for the first and last day of the study. (In this study, the trained observers consisted of the PI and two trained research assistants; one a PhD candidate in nursing and the other a senior high school student in the International Baccalaureate program). Since both the trained observer and the caregiver based their measures on the PADE part I and II on the same 10-minute observation of the participant performing the activity protocol (described below), this provided a way to measure inter- rater reliability between the two groups for assessing pain behavior. Table 3-1. Study Constructs and Measures Constructs Measure Subject Burden Pain -Caregiver and -Sum score of PADE Part 1: -PADE total time to complete trained observer Physical (observable facial for Parts I-III is 5-10 minutes reports of pain expression, breathing pattern, and posture); and Part 2: Global assessment of overall pain -Behavioral -Total number of pain -20 minutes total for activity observation of pain behaviors observed during an protocol, 10 minutes to set up activity protocol and take down equipment and 10 minutes for videotaping -Self-reported pain -average score on the NRS 1 minute Pain outcomes/ correlates -Functional -Sum score of PADE Part 3: -PADE total time to complete performance Functional (ADLs) for Parts I-III is 5-10 minutes -Behavioral -Total and subgroup scores on -RMPBC total time to disruptions RMPBC for memory-related complete is 10-15 minutes problems, depression, and disruptive behaviors PADE: Pain Assessment for the 1)ementing Elderly ADLs: Activities of Daily Living, including dressing, feeding oneself, and transfers RMPBC: Revised Memory and Problem Behavior Checklist NRS: Numeric Rating Scale As mentioned previously, participants also performed a videotaped activity-based protocol derived from Keefe and Block' s (1982) work with chronic low back pain patients. Initial findings using the behavioral observation system showed a highly significant correlation between pain rating with total frequency of observed pain behaviors (r = .71, p < .01). This activity based protocol was previously revised to facilitate assessment of pain behaviors in persons with dementia by Ann L. Horgas, PhD, RN; RO1 #NR05069-01. Using this revised protocol, each participant was asked to perform 4 normal activities of daily living (sitting, standing, lying, and walking in place) as well as transferring to and from each of these activity states. Each activity was performed for a period of one minute, with activities being repeated to reach a behavioral observation time total of 10 minutes. The order of these activities was randomized for each session and the activity protocol was videotaped. The trained observer coded each videotaped activity protocol for the specific pain behaviors described below. Since each participant must have a diagnosis of OA or other painful condition, activities of daily living are deemed to be sufficient stimulants of pain because they represent activities that people with chronic pain often have difficulty with, but that are not too difficult for participants to perform (Keefe et al., 1986). These activities are also ecologically valid as they represent normal and generalizable everyday activities for all older adults who are not so frail as to be confined to a bed or wheelchair. Furthermore, previous studies have shown that increases in observable pain behaviors during activities rather than at rest (Herr et al., 2006). Participants were also asked at each session to self-report their pain level using the numeric rating scale (NRS), both before starting the activity-based protocol and after the protocol was completed. The NRS is the self-report tool of choice for this study because it has been considered conceptually easier to understand by the elderly. This may be because of the added number selections on this scale as compared to other pain rating scales (Herr & Mobily, 1993). The NRS was presented as a horizontal line with 0= no pain as the left anchor and 10= worst pain as the right anchor with equally spaced dashes for each number 1-9. Participants selected the number from 0 to 10 that most accurately identifies the pain in question. This measure has been shown to be both valid and reliable in older adults (Gagliese & Melzack, 2003). See Table 3-1 for a summary of study constructs and measures. Treatment Protocol For the purpose of this study, Tylenol Arthritis Pain Extended Relief was the analgesic of choice for use as the treatment drug. This form of acetaminophen was chosen for its more convenient 3 times per day dosing over the 4 times per day dosing of Tylenol Extra Strength and Regular Strength Tylenol. Tylenol Arthritis has the same safety and side effect profile as the other forms of Tylenol. The difference is that each 650 mg pill of acetaminophen has an immediate release outer layer and a delayed released inner core so each dose (2 pills) provides up to 8 hours of pain relief. Furthermore, "most of the acetaminophen dose is released from the tablet matrix within 5 hours, but peak acetaminophen serum concentrations may be delayed 8 hours or longer following ingestion" (BC Drug and Poison Information Center, 2005, p.1). Acetaminophen is thought to cause analgesia by inhibiting prostaglandin synthesis in the central nervous system (i.e. elevation of the pain threshold) and the therapeutic serum levels for analgesia are 5-20 mcg/mL (Health Digest, 2006; Physician's Desk Reference, 2006). The dose response curve specific for Tylenol Arthritis medication indicate that the peak plasma concentration level (near 9.5 mcg/mL) is reached between 1 and 2 hours after administration, and declines to nontherapeutic levels (less than 5 mcg/mL) 6 hours after administration (McNeil PPC, 2002). The dosing schedule for the treatment drug was designed for each individual participant to follow a continuous dosing pattern during the participant' s waking hours. This means that the caregiver was asked to give two 650 mg pills to the participant every eight hours while the participant was awake. The typical sleep and wake schedule of each participant was taken into account when devising a typical dosing schedule. For example, a possible dosing schedule could then be to take the first dose with the morning meal (9-10am) and a second dose with the evening meal (5-6pm). Alternatively, if the participant is currently taken another medication every eight hours, the caregiver was asked to give the treatment drug along with the other medications) as long as there were no contraindications to doing so. The Tylenol Arthritis medication was provided to the participants free of charge by the PI. A large enough bottle to contain the amount of pills needed for the entire study was supplied. Participants were allowed to retain any remaining pills at the end of the study. The dosing schedule was worked out in advance with each caregiver to fit the daily routine for the individual with dementia. Caregivers were also provided with a medication log highlighting this dosing schedule to serve as a reminder to administer the treatment drug. Based on the participants' daily routines, a 2 time/day dosing schedule (morning after waking and approximately one hour before bed) was established for participants 1 and 3. Since participant 2 was awake more hours of the day, a 3 time/day dosing schedule was established (morning, early afternoon, and late evening). However, it is important to note that although there were dosing differences, each of the participants received the maximum number of doses during waking hours in order to provide continual pain relief without disrupting sleep patterns. Statistical Power Because this study was a feasibility pilot study to test the effects of an acetaminophen trial, the power of this study was not computed. There have been few research studies specifically looking at the statistical power of acetaminophen in reducing pain in persons with dementia. Although Buffum, Sands, Miaskowski, Brod, and Washburn (2004), found that 650 mg of acetaminophen given four times per day was no more effective than prn acetaminophen in reducing discomfort in persons with dementia, this study targeted nursing home residents with severe dementia. It is quite possible that the results may be different in community-dwelling persons with more moderate dementia. Likewise, in an informal study, Douzjian and colleagues (1998) found 650 mg doses of acetaminophen three times a day to be an effective pain reliever. This lends support that using an even stronger dosage (two 650 mg pills up to three times per day) would also be effective in reducing pain. Furthermore, Chinball and colleagues (2005) found moderate-to-strong effect sizes (eta2 = 0. 14-0.29) for acetaminophen in improving outcomes assessed by the Dementia Care Mapping tool in nursing home residents with moderate-to-severe dementia. While acetaminophen is the first line recommended analgesic, it is not the most powerfully acting analgesic available. However, it is reasonable to assume that given its endorsement by leading geriatric treatment guidelines (American College of Rheumatology, 2000; American Geriatrics Society, 2002; British Medical Journal, 2000; European League Against Rheumatism, 2003), that it will produce at least a small effect. Also, the validity and reliabilities of the PADE, RMBPC, and the original activity protocol designed by Keefe and Block (1982) as well as the revised activity protocol (Horgas, 2001), support that these tools will useful in detecting changes in participant behavior. Procedures As shown in Table 3-2, this ABAB single-subj ect study design consisted of an initial baseline daily observational phase (Al) lasting a minimum of eight sessions within a two-week period to observe behavior under conditions before treatment was implemented. As described above, this study departed from the traditional methods for Table 3-2. Study Design Variables ISession 1 Baseline Intervention Baseline Intervention (Baseline) Session Session Session Session 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 121 22 23 24 Inae MeasuresX Otae Measures I X Caregiver IX X Intuton I(Tool I(Medication Administration) Administration) Caregiver Rating -PADE IX IXX XX X XX X XX XX X XX IXX XX X XX X -RMBPC IX IXX XX X XX X XX XX X XX IXX XX X XX X Acetaminophen IIIX XX XX X XX X XX X Activity Protocol IX XX XX X XX IX XX XX X XX X X XX X XX X changing phases. While the target goal was to have daily sessions, due to the unpredictable nature of collecting repeated measurements in persons with dementia, it was thought that it might not be convenient or possible for the caregivers to meet with PI on subsequent days throughout the study period. For this reason, guidelines for a minimum number of sessions within a maximum time-period were established. Typically the initial baseline phase is continued until the participant's behavior appears to be stable "or until it is evident that the response does not improve over time" (Kazdin, 1982, p11.Since pain is labile in nature, it was proposed that at least eight baseline sessions would be needed to observe each individual's typical pain behavior patterns. The baseline phases of this design were necessary to describe the current level of behavior as well as to predict what future behavior would have looked like without an intervention. Once the typical pain behavior pattern was observed, the intervention phase (B1) was initiated. In an effort to create equivalence between study phases, this phase was also eight sessions in length (within a two-week period). Similarly to the baseline phases, the intervention phases describe current behavior and predict future behavior if the intervention were to be continued. However, the intervention phases also test the predictions from the baseline phases. That is, the investigator can test whether behavior in the intervention phase actually departs from the predicted baseline level. The return to baseline phase (A2) began with the withdrawal of the intervention in an attempt to show that the participant's behavior would return to baseline performance without the intervention. This phase was planned to last a minimum of four sessions (if behavior returned to baseline levels rapidly) to a maximum of eight sessions. The unique purpose of this second A phase was to test the prediction from the B1 phase. If behavior returned to A1, levels then it would be clear that the predicted and obtained performance levels differ, suggesting that the intervention withdrawal altered behavior. Following the return to baseline phase, the intervention was reimplemented (phase B2). This phase was also planned to last a minimum of four sessions to a maximum of eight sessions. If in this phase, behavior again improves, the overall pattern of results would strongly suggest that the intervention was responsible for the change in behavior. During the study, the Principal Investigator or a trained research assistant conducted all activity protocols using a digital video camera to capture each session. An effort was be made to try to schedule all videotaped sessions at the same time each day in order to establish a routine that will cause less burden for the caregivers and to try to control for variability in pain experienced at different times of the day. Also, it should be noted that because the half-life of acetaminophen is 1 to 3 hours, the standard washout of 3.5 half-lives would occur within 10 hours (Buffum et al., 2004). Thus, there is no need for a washout period between phases B1 and A2. During any days were the caregivers were unable to meet with the research team, they were told they could complete the daily measurement tools for that day if they chose too. However, CG 1 completed all 24 sessions within 25 days, CG 2 completed the study within 24 days; and CG 3 completed the study within 26 days, so there were not many instances of gaps between visits by the research team. Phase Al, Session 1 Day 1 of the study consisted of participant and caregiver screening for enrollment. A trained research assistant administered the IVMSE separately to the participants with dementia to confirm cognitive impairment. The participants were also given a comprehensive intake assessment in order to thoroughly describe and evaluate sample character sti cs. Consent was obtained from the caregivers, who were each, also the legally authorized representative of their care recipient and assent was obtained from the participants with dementia themselves. Caregivers were asked to provide a current list of all of the medications that the participant received, including any over the counter medications. A medication log for each phase of the study listing all of the participant' s medications was created and given to the caregivers to use as a log to record daily administration of all medications. In an effort to increase the reliability of caregiver ratings, caregivers also received instruction and training in administering the PADE and the RMBPC. Sample videotapes were shown to the caregivers in an effort to increase their knowledge of pain behavior as well as their ability to identify excess disabilities. They were asked to administer these tools themselves in front of the principal investigator to ensure understanding of tool administration. Caregivers received a packet of measurement tools for each phase of data collection. To facilitate data collection, the measurement tools were printed in 14-point font on white paper (to facilitate reading). In addition, the packet was organized and clearly labeled to delineate the correct measure for each session. Also, on this day, the initial videotaped activity protocol was completed. Finally, caregivers and the trained observer completed the PADE and RMBPC regarding participant behaviors, and the caregivers answered the questions regarding the typicality of the day as well as if they noticed any pain behaviors that they thought most indicated pain. Phase Al Throughout this baseline observation phase, the PI or trained RA completed the activity protocol and the PADE, and the caregivers completed both the PADE and the RMBPC during the visit. Since an effort was made to time study sessions at approximately the same time each day, caregivers were instructed to think back over the participant' s behavior since the previous day's session (or for the previous 24 hours if sessions did not occur on subsequent days) and to accurately reflect this behavior on the measurement tools. Additionally, they were asked to evaluate if this was a typical or atypical day for the participant (i.e. were there visitors, were they ill, or did they visit the doctor?) and to describe what behaviors they most noticed as indicating pain. Caregivers also recorded administration of all medications on the log sheet provided. Phase Al-Final Session For the final session in the observation phase (8th SCSsion), the PI instructed caregivers on how to appropriately administer the treatment drug, acetaminophen, and worked with the caregiver to make a convenient dosing schedule/plan. After setting up the schedule (e.g. 900 hours and 1700 hours or morning meal and evening meal) for each individual, the caregiver administered the first dose of acetaminophen at the next scheduled time, after the activity protocol for that day was completed. Caregivers then received a packet for the second study phase including the daily questionnaires and medication log sheets with an additional space for recording when the treatment drug was given. Again, the PI and the caregivers completed the daily behavioral assessment tools and the caregivers provided information regarding the typicality of the day as well as the pain behaviors they noticed. Phase B1 (Eight Sessions) During the second study phase, (the first intervention phase), caregivers continued to administer both the PADE and the RMBPC during the daily visits and recorded if the day was a typical day for the participant as well as the pain behaviors they noticed in the participant. In addition, the caregivers gave the treatment drug, as per the previously established schedule, and recorded the time of administration on the log sheets provided. Participants continued to perform the videotaped activity protocol and the PI or trained RA also completed the PADE. Phase A2- Return to Baseline (Range of 4-8 Sessions) Caregivers continued to log all medication administration during this week but were instructed to discontinue giving the treatment drug. Caregivers were, however, instructed that they could administer rescue doses of pain medication as needed by the participant, but to make sure to log these doses on the medication log sheet. Data collection proceeded as before in Phase Al. Phase B2- Reimplementation of Intervention (Range of 4-8 Sessions) This phase proceeded exactly as the original intervention phase. Caregivers were instructed to resume giving the intervention drug at the previously scheduled dosing times, logging all medications given. Participants continued to perform the activity protocol videotaped by the PI or trained RA, and the both the PI or trained RA and the caregivers completed the measurement tools as previously described. Caregivers also provided information regarding the typicality of the day as well as the most frequent behaviors that they saw in the participants that represented pain. On the last day of this phase, qualitative information regarding the caregivers' perceptions and satisfaction of the pain treatment was gathered. The PI or the trained RA also administered the Geriatric Depression Scale and the Dementia Rating Scale to the participant with dementia and the depression scale to the caregiver as part of an outtake assessment to evaluate changes in depression and cognitive performance over the course of the study. Data Coding Each videotaped activity protocol was coded using Noldus The Observer software the day it was recorded. This enabled the PI (who was the primary coder) to evaluate participant performance on an immediate basis in order to accurately assess behavior and the need for changes in study phases. Since all videotaping was done using digital video cameras, the media files were transferred to the computer using Dazzle Video Pro software and were then opened in The Observer' s viewing window. Each media file was saved on an external hard drive as a backup. A coding program was created in the Observer that allowed independent coders to code the media files for specific pain behaviors. The independent variables entered in this program were participant ID number, study phase, and session number of the phase. Two categories of behaviors were used in this program, state behaviors and event behaviors. State behaviors are behaviors that occur for an unspecified amount of time, this requires that a behavior remain 'active' for the entire duration that this behavior is observed. Event behaviors are behaviors that are observed, but the duration of the behavior is not coded. The Observer requires that all state behaviors in a behavioral class be mutually exclusive; that is, no two behaviors in the same behavioral class can be active at the same time. A second independent trained coder (who was an honor' s student in the undergraduate nursing program) coded 25% of all videotapes to ensure inter-rater reliability and primary coder accuracy. This second coder was blind to the phase of the study in which videotaped sessions had occurred. Coder reliability was measured with the Kappa statistic (K) which not only takes into account the frequency of agreements between coders, but also accounts for the number of agreements one would expect as a function of chance (Blomqvist & Hallberg, 1999). A Kappa of 0.80 was established as the criterion for coder training. Coders trained on practice tapes until their intra-rater reliabilities were greater than a kappa of 0.80 and were considered trained when their inter-rater reliabilities were greater than a kappa of 0.80 on practice tapes when compared with the master coder. For this study, the two raters reached an average inter-rater agreement of kappa =0.84, or 84% on the practice tapes (K range= 0.72-0.96), which meets the established criterion for training (Elder, 1999; Keefe, Bradley, & Crisson, 1990; Keefe & Hill, 1985). Reliability estimates for the actual study tapes were also good at kappa =0.80 (K range=0.62-0.98). This was considered to indicative of good reliability as there were 9 behaviors, 5 activities, and an actual total of 17 codes that coders used for the same observation, and other researchers have proposed that reliability coefficients between .60-.80 indicate good to very good reliability (Blomqvist & Hallberg; Gibson & Helme, 2000). Training strategies to increase inter-rater reliability included informal discussions about coding behaviors, independent study of the training manual (containing definitions and instructions for coding the variables of interest), and practice coding sessions with the 'master coder' and all research assistants The two independent coders coded participants' activity protocols by recording the specified pain behaviors of interest that have been adapted and modified by Horgas (2001) from definitions of behavioral coding categories originally employed by Keefe et al. (1986). See Table 3-3 for the specific behaviors of interest for this study. Data Analysis Procedures and Variables of Interest RMBPC The RMBPC yields both a global summary score and subscores for participants' behavioral disturbance and caregivers' reactivity. Also, frequency and reaction scores for the three specific subscales (Memory-Related Problems, Depression, and Disruptive Behaviors) were computed. PADE The PADE is divided into three parts with Part I representing the physical domain, Part II representing a global assessment domain, and Part III representing the functional domain. Since questions are answered on either a Likert scale (range 1-4, or 0 for skipped answer or N/A) or with multiple choice (1, 2, 3, or 4), sum scores for each of the three parts were computed along with a total measure score. Activity Protocol Coding of the activity protocol yielded information about the duration and frequency of state behaviors (i.e. guarding, bracing, & rigidity) and the frequency of event behaviors (i.e. sighing/nonverbals, rubbing, stopping, grimacing, shifting, and vocalizations). Therefore, the mean time spent in a state behavior and the mean frequencies of all behaviors were calculated by Noldus the Observer software and used in the analyses. Additional Caregiver Questionnaire Caregiver responses as to the pain behaviors that they most frequently noticed to be indicative of pain in the participant were tabulated with frequency counts of specific behaviors listed by the caregiver. Overview of the Analysis of Single-Subject Design Data The overall obj ective of data analysis in any research study is to examine whether a truthful change has been demonstrated and whether or not this change is attributable to the intervention. In single-subject research, this is accomplished by examining the effects of an intervention at different points over time (Kazdin, 1982). In ABAB design, the Table 3-3. Behavioral Definitions ACTIVITY MOVEMENT BRACING COMPLETING PAIN BEHAVIORS (STATE) (STATE) (STATE) ACTIVITY (EVENTS) (STATE Sit = begin coding when Ss' buttocks rest on chair Rigidity = Stiffness of the lower body (from back down to feet) exhibited by Ss maintaining the affected lower body part in a stiff and abnormal position during standing, sitting, or lying down Guarding = abnormally slow, stiff, interrupted, or rigid movement - during4 transfer~rtring ~rt~t~rtrt~ or walking Bracing = holding onto someone or wwth,,rling for support (e.g., walker, bed, cane) participant- initiated behavior. If tester initiated behavior, code as bracing if there is evidence that the participant needed support (e.g., was unstable or falling.). Can occur during sit, stand, walk and transfer. but not during lying down. Rubbing = touching, rubbing, or holding the affected lower body part (e.g., lower back/waist to feet) (palms down) Code 1 completed circular or forward/backward movement as one rub. Shifting = shifting of weight. This includes channinn position or shifting weight (side to side shifts) of the trunk, hips, or lower extremities while doing an activity. [Excludes front to back sway.] Can occur during any activiy; shifting during walking would reflect purposeful shifts in the lower back or trunk. Stopping = interrupting/stoppi ng the activity state during any activity state except transfer ~rt~rt~t~rt~rt~ stopping the activity they are supposed to be doing (Ex: while walking - stopping to stand still or sit down; while standing, stopping to sit down; quitting the activity state) Stand = begin coding when Ss is in an upright position with one or both feet on the floor Table 3-3. Continued ACTIVITY MOVEMENT BRACING COMPLETING PAIN BEHAVIORS (STATE) (STATE) (STATE) ACTIVITY (EVENTS) (STATE Walk = begin coding when Ss is in an upright position, and begins moving feet in marching/walking pattern Lie = begin coding when Ss' head and back touch the bed (horizontal position) Transfer = begins when the 1-minute activity state ends (When the timer rings to end the 1-minute activity, code this as beginning the transfer activity). End transfer when the next 1- minute activity begins Grimacing = Obvious change in facial expression, indicated by frown (tightened lips, corners of mouth pulled back or down) and eye change (narrowed eyes and furrowed brow). Sighing/Non-verbal Vocalization = an obvious exaggerated exhalation of breath, usually accompanied by shoulders rising and falling or change in lip/mouth position OR any spontaneous non-verbal sound such as moaning, groaming, gasping, grunting. Verbal Complaint = spontaneous expression of pain or discomfort in affected body area (e.g., trunk, back, and lower extremities). [Verbal responses to questions or statements made in the context of questioning do not apply.] effects of an intervention are replicated so a judgment as to the intervention's effectiveness is made on the overall pattern of the data. If participant behavior changes each time the intervention is applied and again when it is withdrawn, the effect of the intervention is apparent. These judgments are made based on the visual representation of the data. Visual inspection refers to "reaching a judgment about the reliability or consistency of intervention effects by visually examining the graphed data" (Kazdin, p. 232). As part of the design requirements, participant data are graphed continuously over the course of the study. These graphed data points represent the study behavior(s) of interest. There are several characteristics of the data that are examined in visual analysis of data. These characteristics are either related to the magnitude of change across phases (mean and level) or the rate of change (trend and latency). In addition, overlap of data points between phases and well as stability within a phase are also of interest (Kazdin). A change in mean refers to a shift in the average rate of performance between phases. For this study, higher means of pain behaviors and excess disabilities were expected during the baseline phases and lower means were expected during the intervention phases. A change in level refers to the shift or discontinuity of behavior between the end of one phase and the beginning of the next phase. Since the trend or the slope of the data shows whether behavior systematically increases or decreases over time, a change in trend may be revealed when the intervention is applied or withdrawn. In the present study, decreases in level were expected when shifting from a baseline phase to an intervention phase, and increases in level were expected when shifting from an intervention phase to a baseline phase. Likewise, the trend (slope) was expected to be in the positive (increasing) direction during baseline phases and in the negative (decreasing) direction during intervention phases. The latency of change refers to the period between the onset or termination of one phase and subsequent changes in behavior and is assessed by examining level and trend changes together. Therefore, the latency of change reveals how long after an intervention is applied it takes to change behavior. The closer this change takes place to the change in experimental condition, the more clear the effect of the intervention. In summary, "visual inspection is conducted by judging the extent to which changes in these characteristics are evident across phases and whether the changes are consistent with the requirements of the (study) design" (Kazdin, 1982, p. 237). Kazdin also points out that there are also specific background characteristics to examine in relation to the visual analysis of data. These characteristics speak to the reliability of an effect and are measured through examining the: variability of performance within a particular phase (stability), the duration of the phase, and the consistency of the effect across phases or baselines (overlap). (See the Appendix A for a detailed explanation for performing these graphical analyses). One potential drawback to visual analysis is that only interventions that produce a large effect will be considered to have produced a change. In the case of an intervention that produces more modest effects, statistical evaluation may be warranted. While statistical evaluation is not the norm in single-subj ect research, there are several circumstances in which it can reveal significant intervention effects that may have been missed in visual analysis. Specifically, the statistical evaluation of single-subject data to supplement visual analysis is appropriate when there are unstable baselines, new areas of research are being investigated (intervention effects may be weak), there is increased intrasubj ect variability (where there is little control over the environment and potential influences), or when even small changes in behavior may be important and meaningful (Kazdin, 1982). Appropriate statistical tests include conventional t and F tests (when there is no serial dependency between data points) and time-series analysis. Statistical Analysis of Each Aim Aim 1: To investigate the effects of the scheduled administration of acetaminophen (1.3 grams three times per day) on self-reported pain intensity and the number of observable pain behaviors exhibited by persons with dementia. Hypothesis: Regular administration of acetaminophen will decrease the total number of observable pain behaviors (as rated by the caregiver and trained observers) exhibited by the participant, but will have no effect on self-reported pain intensities. The variables of interest for these analyses included caregivers' and the PI' s or trained RA' s daily responses on the PADE part I and the total frequency and duration of individual pain behaviors as coded with the Observer from the videotaped activity protocol. These totals were graphed against the day of measurement. As described above, the standard in analyzing SSD data consists of visually inspecting dependent measures (i.e. total number of pain behaviors) during independent variable conditions (acetaminophen trial phase) relative to baseline measures (Morgan & Morgan, 2001). The data analysis and graphical interpretation occurred in an ongoing manner such that data were analyzed and graphed as soon as they were collected. At the conclusion of the study, the graph for each participant was interpreted to look for an overall pattern in pain behaviors during baseline phases before treatment and any change in the frequency or duration of behaviors after initiation of pain treatment. Specifically, changes in mean, level, trend, latency, overlap, and stability were examined. If warranted, statistical evaluation using t and F tests to "compare whether differences in means are statistically reliable between, or among, the different phases" (Kazdin, 1982, p. 245) or time-series analysis to "examine whether there is a statistically significant change in level and trend from one phase to the next" (Kazdin, p. 248) were planned. Aim 2: To investigate the effects of the scheduled administration of acetaminophen on the frequency and severity of excess disabilities of pain in persons with dementia. Hypothesis: Among persons with dementia, the frequency and severity of excess disabilities of pain, such as depression, memory problems, behavioral disruptions, and impaired functional performance will decrease from baseline after implementing regular administration of acetaminophen. The variables of interest are the frequency ratings for each of the three subgroups (memory-related problems, depression, and disruption) on the RMPBC and for the measure as a whole which were calculated for each participant daily. Also of interest for this aim are caregiver's responses on the PADE part III. Again, this data was graphed and analyzed as described for Aim 1. Aim 3: To determine the reliability of informal caregiver ratings of the frequency of pain behaviors exhibited by persons with dementia. Hypothesis: After an initial training session, informal caregivers will display moderate to good reliability with a trained observer in ratings of the frequency of pain behaviors. The variables of interest were the caregivers' ratings on the PADE part I and the trained observer's ratings on the PADE part I. To evaluate the reliability of informal caregivers' ratings of pain behaviors, the caregiver' s ratings were compared to those of the trained observer' s for each study session and the inter-rater reliability was calculated using Pearson's correlation coefficient. Aim 4: To investigate the most frequently displayed behavior(s) indicative of pain that is/are displayed most frequently by persons with dementia. Hypothesis: Based on findings from similar studies, the most frequently displayed behaviors indicative of pain are expected to be guarding, rubbing, shifting, and bracing (Horgas & Elliott, 2005; Keefe & Block, 1982). The variables of interest are caregiver ratings on the PADE part I, the total number of individual pain behaviors as coded with the Observer from the videotaped activity protocol, and caregiver responses as to which behaviors they most noticed as being indicative of pain in the participant. The most frequently rated and coded behaviors were then extrapolated and this list of behaviors serves as the behaviors employed most by persons with dementia to relay pain. CHAPTER 4 RESULTS The results of the present study are presented below. First, all descriptive statistics are presented; specifically, scores on the initial cognition and depression measures for each participant are compared to their scores on these measures from the final day of the study. Initial and final caregiver depression scores are also presented. Next, the results for each of the 4 specific aims are presented. Of note, for aims 1 and 2, graphical representations of the variables of interest are presented individually for each participant. Each graph was then analyzed by looking at changes in the mean between phases, the latency of change between phases (which consisted of examining level and trend changes), the overlap of data points between phases, and finally, the stability of data within phases. Descriptives Initial Intake Measures The intake Mini Mental State Exam (MMSE) score for Participant (P) 1 was 10, indicating moderate (near severe) cognitive impairment while the initial Dementia Rating Scale-2 (DRS-2) total score (31) indicated severe cognitive impairment. The intake depression measure indicated no depression (Geriatric Depression Scale score = 0). Initial depression scores for caregiver (CG) 1 indicated mild depression, (Center for Epidemiological Studies Depression Scale score of 19). The intake MMSE score for P 2 was 20, indicating moderate cognitive impairment, while the initial DRS-2 total score (111) indicated severe impairment. The intake depression measure indicated no depression (GDS = 1). Initial depression scores for CG 2 indicated no depression (GDS = 2) as well. The intake MMSE score for P 3 was 5 indicating severe cognitive impairment and the initial DRS-2 total score (65) indicated severe impairment as well. The intake depression measure indicated no depression (GDS = 1) and initial depression scores for CG 3 indicated no depression (GDS = 2). Demographic data along with these scores are presented in Tables 4-1 & 4-2, comparing those pairs who completed the study to the two pairs that were withdrawn from the study. Final Outtake Measures During the last study session, final outtake measures were taken from both the participant and the caregiver. This was done specifically to see if there had been any change in cognition (for the participants) or depression (for both participants and caregivers) during the course of the study. Scores for these outtake measures are presented in Tables 4-1 & 4-2. As can be seen outtake DRS-2 total scores were exactly the same for Participants 1 & 2 as their intake scores. Interestingly, while there was some variation in subscale scores between intake and outtake measures, the total score remained the same. For P 3, the DRS-2 total score improved from 65 to 71, however, this increase in total score did not change her severely impaired classification on this measure. Depression scores also did not show much change from intake assessment to outtake assessment. Participant 1 remained at 0 both occasions, and while Participants 2 and 3 showed declines on the GDS (1 to 0 for P 2, and 3 to 1 for P 3), although neither participant was initially classified as depressed. The same was true for Caregivers 2 and 3 (CG 2 decreased from GDS 2 to 1, CG 3 remained at GDS 2) who were not initially Table 4-1. Descriptive Characteristics for Participants ID Age Sex Racea Educationb Marital Status MMSE Initial DRS-2 Score Final DRS-2 Score 31 111 71 Initial Depression Score (GDS) Final Depression Score (GDS) Participants Included Pl P2 P3 Participants Withdrawn 0 3 aO= Caucasian, 1= Black bl __th grade or less, 2= 9-11th grades, 3= High School, 4= Technical or Trade School, 5= Some College '0= Never been married, 1= Married, 2= Widowed, 3= Separated, 4= Divorced Table 4-2. Descriptive Characteristics for Caregivers ID Age Sex Race" Educationb Marital Status' Relation to Participant Daughter Wife Daughter Initial Depression Score 19 (CES-D) 2 (GDS) 2 (GDS) Final Depression Score 22 (CES-D) 1 (GDS) 2 (GDS) Caregivers Included CG 1 CG 2 CG 3 F 0 3 Caregivers Withdrawn 4 37 F 1 5 0 Daughter 5 77 F 0 3 1 Wife aO= Caucasian, 1= Black bl __th grade or less, 2= 9-11th grades, 3= High School, 4= Technical or Trade School, 5= Some College '0= Never been married, 1= Married, 2= Widowed, 3= Separated, 4= Divorced 12 (CES-D) 5 (GDS) classified as depressed. However, Caregiver 1, the only person in the study to initially have a score in the depressed category (on the CES-D) increased from an initial score of 19 to an outtake score of 22 Effect of Acetaminophen on Self-Reported Pain Intensity and Observable Pain Behaviors The primary specific aim of this study was to assess the effect of scheduled doses of acetaminophen on self-reported pain intensity and the number of observable pain behaviors. Participants were asked to self-report their pain intensity using the numeric rating scale (NRS) both immediately preceding and following the activity protocol. Over the course of the study (24 sessions for 48 total occasions for self-report), P 1 self- reported no pain on all occasions. P 2 used the NRS on four occasions (with this being on both occasions at 2 sessions during the Einal treatment phase) to self-report mild pain intensity (scores of 1s and 2s on the NRS). P 3 used the NRS on six occasions (twice for both occasions at a session and twice preceding the activity protocol only) to self-report pain intensity. On Hyve occasions, during both the baseline and treatment phases, this pain was mild (1s and 2s on the NRS) and on one occasion during the initial baseline phase, this pain was moderate (6 on the NRS). The effect of acetaminophen on observable pain behaviors was evaluated by comparing pain behaviors exhibited by participants' during baseline performance of an activity protocol with their performance of the same protocol while receiving scheduled doses of acetaminophen. Several methods were used to observe for changes in performance. First, the total number and total duration of the previously defined target pain behaviors (see Table 3-3) were coded for using Noldus The Observer software. Changes in performance were also assessed with Part I of the PADE measure which was completed daily by both the CG and the trained observer (TO) (i.e. either the Principle Investigator or trained research assistant who was physically present at the data collection session). In total, there were 4 measures employed to assess the effect of acetaminophen on observed pain behaviors exhibited by persons with dementia. The results for each of these measures are represented graphically for each participant. As part of the analyses of graphed data, changes in mean, the latency of change (i.e. comparing level changes between the end of one phase and the beginning of the next phase and the change in trend or slope between phases), overlap of data between phases, and the stability of data within a phase are presented. Complete descriptions of these components of graphical analysis are presented in Chapter 3. (See Appendix A for examples of how the mean, level, trend, latency, overlap, and stability were calculated). Total Number of Pain Behaviors Participant 1 Figure 4-1 displays P l's frequency of pain behavior for each session during baseline and intervention conditions. Mean: During the initial baseline condition (Al), the mean frequency of pain behavior exhibited during the activity protocol was 32. 1 behaviors per session. This mean frequency decreased to 18.6 after the introduction (Bl) of the intervention (acetaminophen). Upon the return to baseline condition (A2), the mean frequency of behavior increased to 27.5 and then decreased to 17.5 when the intervention was reintroduced (B2). Changes in both level and trend are analyzed together to examine the latency of change between phases. Latency: There was an 11 point decrease in level between phases Al and Bl, however, the trend in phase Al was initially in the unintended direction (slope= -1.44) and flattened (slope= -0.04) in phase Bl. Between phases B1 Total Number of Pain Behaviors for Participant 1 45-Al B1 A2 B2 S20- 1 5 9 13 17 21 Sessions Figure 4-1. Total number of pain behaviors for P 1 across sessions from the coded activity protocols. and A2, there was a small level change in the unintended direction (1 point decrease) but the there was a large change in trend in the intended direction (slope= -0.04 to 5.8). Finally, between phases A2 and B2, there was a 15 point decrease in level and a large change in trend in the intended direction (slope= 5.8 to -2.8). Overlap: There was nooverlap in data points between phases Al and B l, while 2 points overlapped between phases B1 and A2, and 1 point overlapped between phases A2 and B2. Stability: There was 100% stability in the data points in phases Al, A2, and B2, while phase B1 had 62.5% stability in data points. The changes in mean frequencies for each individual pain behavior by phase for Participant 1 are presented in Figure 4-2. Only behaviors which occurred greater than 25 times during the course of the study are included in these analyses. As can be seen, for Mean Frequencyof Individual Pain Behaviors Across Phases for Participant 1 8 m Phase Al m Phase B1 E ~OPhase A2 : O Phase B2 Mean Guarding Mean Rlgidity Mean Bracing Mean Shifting Mean Slghing/Nonverbal Behaviors Figure 4-2. Mean frequencies for individual behaviors across phases for P 1. each behavior except shifting, the mean frequencies are higher in the two baseline phases and are lower during the treatment phases. Participant 2 Figure 4-3 displays P 2's frequency of pain behavior for each session during baseline and intervention conditions. Mean: The mean frequency of pain behavior in theinitial baseline phase (Al) was 33 behaviors per session, which decreased to 22.5 during the first intervention phase (B l). The mean frequency of behavior increased during the return to baseline phase (A2) to 31.3 and decreased to 20. 1 when the intervention was re- introduced (B2). Latency: There was a 9 point decrease in level from the last session in phase Al to the first session in phase B l, and a change in trend from a slope of 1.0 in the intended direction in phase Al to nearly no trend (slope = 0.02) in phase Bl. Although a large change in level between phases B1 and A2 (14 point increase) was observed, the Total Number of Pain Behaviors for Participant 2 45-Al B1 A2 B2 S20- 1 5 9 13 17 21 Sessions Figure 4-3. Total number of pain behaviors for P 2 across sessions from the coded activity protocols. change in trend was in the unintended direction (slope= 0.02 to -1.5). Between phases A2 and B2 there was a 10 point decrease, accompanied by a change in trend in the unintended direction (slope= 0.9). Overlap: There was a 1 point overlap between phases Al and B1 and also between phases B1 and A2. There was no overlap of data points between phases A2 and B2. Stability: There was 100% stability in data for phases A2 and B2, 62.5% for Al, and 75% for phase Bl. The changes in mean frequencies for individual pain behaviors by phase for P 2 are presented in Figure 4-4. Behaviors which occurred greater than 25 times during the course of the study are included in these analyses. As for P 1, for each behavior except shifting, the mean frequencies of each behavior are higher in the two baseline phases and are lower during both the treatment phases. Mean Frequencyof Individual Pain Behaviors Across Phases for Participant 2 6- 5Phase B1 0 Phase A2 Mean Guarding Mean Bracing Mean Rubbing Mean Shifting Mean Gnmacing Mean Sighing/Nonverbal Behaviors Figure 4-4. Mean frequencies for individual behaviors across phases for P 2. Participant 3 Figure 4-5 displays P 3's frequency of pain behavior for each session during baseline and intervention conditions. Mean: The mean frequencies of pain behavior for P 3 were higher across all phases than for either Participant 1 or 2. The change in mean frequencies was as follows: 57.8 behaviors per session for phase Al, 30 for phase Bl, 53.3 for phase A2, and 29.8 for phase B2. Latency: There was a 35 point decrease in level between the end of phase Al and the beginning of phase Bl, however, the change in trend went from a slope= 3.5 in the intended direction to a still accelerating slope of 1.2 in phase Bl. Between phases B1 and A2, there was a 9 point increase in level, however, there was relatively little change in trend (slope= 1.2 to 0.9). There was a 28 point decrease in level between phases A2 and B2, however, the trend continued to accelerate in the unintended direction (slope= 2.1). Overlap: There was also no overlap of data points between any of the phases. Stability: The data were 100% stable in phases A2 and B2, 87.5% stable in phase Al, but had 25% stability in phase B l. Total Number of Pain Behaviors for Participant 3 90-Al B1 A2 B2 S40- 1 5 9 13 17 21 Sessions Figure 4-5. Total number of pain behaviors for P 3 across sessions from the coded activity protocols. Changes in mean frequencies for each individual pain behavior by phase for Participant 3 are presented in Figure 4-6. Behaviors occurring greater than 25 times during the course of the study are included in these analyses. There is a clearly observed pattern of increased mean frequency during baseline phases and decreased mean frequency for intervention phases for guarding, rigidity, stopping, shifting, and sighing/nonverbal behaviors. The pattern is less clear for bracing and verbal complaints, but there is an initial decrease in these behaviors after the initial baseline phase. Total Duration of Pain Behaviors The following section presents study results regarding the effect of acetaminophen on the duration of pain behaviors. As mentioned previously, only certain Mean Frequency or Indiv idual Pain Behav iors Across Phases for Participant 3 HPhase A1 10 gPhase B1 c 8-OPhase A2 2 ~O Phase B Mean Guarding Mean Rlgidity Mean Bracing Mean Stopping Mean Shlitng Mean Mean Verbal Sighing/Nonverbal Complaint Behaviors Figure 4-6. Mean frequencies for individual behaviors across phases for P 3. behaviors were coded both for their frequency and their duration. These behaviors included: guarding, bracing, rigidity, and stopping. Participant 1 Figure 4-7 displays P l's total duration of the pain behaviors (e.g. guarding, rigidity, bracing, and stopping) in seconds for each session across all 4 phases. Mean: The mean durations for phases were as follows: 325.6 seconds per session for Al, 120.7 seconds for Bl, 278 seconds for A2, and 155.4 for B2. Latency: In evaluating the latency of change between phases Al and B1 there was a change in level of a 285.5 second decrease, and although the trend in phase Al was in the unintended direction (slope= -4) there was a pronounced flattening of the trend in phase B 1 (slope= -0.2). Between phases B1 and A2 there was a decrease of 56.3 seconds in level and a marked change in trend from a slope=-0.2 to a slope = 93.0. Finally, between phases A2 and B2 there was a 204.6 second decrease in level as well as a trend lessening to a slope= 12.4. Total Duration of Pain Behaviors for Participant 1 45 -Al B1 A2 B2 400- 350- a, 300- S250- S 200- 100- 1 5 9 13 17 21 Sessions Figure 4-7. Total duration of pain behaviors for P 1 across sessions from the coded activity protocols. Overlap: There was no overlap between phases Al and Bl, and only 1 point of overlap between phases B1 and A2, and A2 and B2. Stability: The data were relatively unstable, finding 62.5% stability in phase Al, 50% stability in phases A2 and B2, and only 25% stability in phase B l. The change in mean duration for each behavior is shown in Table 4-3. The expected pattern of longer durations during the baseline phases and shorter durations during the intervention phases was seen for each behavior; although stopping was an infrequent behavior, occurring less than 5 seconds in each of the latter three phases. Participant 2 Figure 4-8 displays P 2's total duration of the pain behaviors in seconds for each session across all four phases. Mean: The mean duration of pain behaviors was as Table 4-3. Mean Duration (in Seconds) of Individual Pain Behaviors across Phases and Participants Participant B ehavi or Phase Al Phase B 1 Phase A2 Phase B2 1 Guarding 151.2 59.3 121.5 59.7 Rigidity 129.0 48.3 110.8 81.1 Bracing 34.4 11.9 41.9 12.3 Stopping* 10.9 1.2 3.8 2.4 2 Guarding 95.0 55.6 101.2 71.3 Rigidity* 98.0 32.1 47.2 30.6 Bracing 11.5 6.7 10.8 7.1 Stopping* 0.8 3.3 5.3 3.3 3 Guarding 200.4 43.9 137.3 74.8 Rigidity 225.2 112.9 128.2 80.4 Bracing 304.8 257.0 263.7 222.8 Stopping 43.4 12.4 16.6 13.4 * Infrequent behaviors (<25 total occurrences) follows: 205.3 seconds per session during phase Al, 97.7 seconds during phase Bl, 164.6 seconds during phase A2, and 112.2 seconds during phase B2. Latency: Between phases Al and B1 there was a marked decrease in level (159.5 seconds), however, the trend in the initial baseline phase was strong in the unintended direction (slope= -8.3) which flattened in phase B1 (slope= -0.8). Between phases B1 and A2, there was an increase in level of 138.7 seconds and an increase in trend (slope= 18.9) in the intended direction. Between phases A2 and B2, there was a decrease in level (188.2 seconds), however, the change in trend continued to increase to slope= 25.7. Overlap: There was a lot of overlap between data in this participant, 4 points between phases Al and Bl, 3 points between phases B1 and A2, and 2 points between phases A2 and B2. Stability: There was also little stability of the data in all phases (37.5% in Al, 12.5% in B l, 0% in A2, and 50% in B2). Total Duration of Pain Behaviors for Participant 2 450 Al B1 A2 B2 400- 350- vl300- 250- S 200- 0 150- 100- 1 5 9 13 17 21 Sessions Figure 4-8. Total duration of pain behaviors for P 2 across sessions from the coded activity protocols. The change in mean duration for each behavior is shown in Table 4-3. The expected pattern of longer durations during the baseline phases and shorter durations during the intervention phases was seen for each behavior, although rigidity and stopping were infrequent behaviors, with stopping occurring 5 seconds or less in each phase. Participant 3 Figure 4-9 displays P 3's total duration of the pain behaviors in seconds for each session across all four phases. Mean: The mean total duration of pain behaviors was higher for P 3 than for the other two participants. The change in mean for phases was as follows: 773.8 seconds per session for phase Al, 426.2 seconds for phase Bl, 545.9 seconds for phase A2, and 391.5 seconds for phase B2. Latency: Between phases Al and Bl, there was a 209.8 second decrease level and a trend change from a slope of 16.0 Total Duration of Pain Behaviors for Participant 3 1200 Al B1 A2 B2 1000- a 800- .C 600- a 200- 1 5 9 13 17 21 Sessions Figure 4-9. Total duration of pain behaviors for P 3 across sessions from the coded activity protocols. in the intended direction to a slope of -16.0 in the intended direction for the intervention phase. This same effect followed for the latency of change between phases B1 and A2, although the change in level was a decrease of 12.3 seconds. The change in trend was strong with the slope increasing to 111.65 for phase A2. Between phases A2 and B2 there was a 60.6 second decrease for level change, and a change in trend to slope of - 78.1. Overlap: There was one point of overlap between phases Al and B1 and also between A2 and B2. There were 2 points of overlap between phases B1 and A2. Stability: Overall, the stability of phases varied, with 75% stability in phases B1 and A2, 50% in phase Al, but no stability (0%) in phase B2. The change in mean duration for each behavior is shown in Table 4-3. Again, the expected pattern of longer durations during the baseline phases and shorter durations during the intervention phases was seen for all behaviors. |