1 SOCIO-ECOLOGICAL DETERMINANTS OF INJURY AMONG YOUNGER AND OLDER ADULTS INVOLVED IN FATAL MOTOR VEHICLE CRASHES IN THE UNITED STATES By KEZIA DZIFA AWADZI A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLOR IDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2006
2 Copyright 2006 by Kezia Dzifa Awadzi
3 To my parents, Dr. Kwablah Awa dzi and Mrs. Patricia Awadzi, and my grandmother, Agnes Boli
4 ACKNOWLEDGMENTS I would like to express gratit ude to my committee, especially my chair Dr. Duncan, my mainstay throughout my doctoral st udy, co-chair Dr. Classen, who supervised my research and whose Centers for Disease Control and Prevention (CDC) K01 award grant funded the last year and half of my graduate study, and Dr. Allyson Hall, who encouraged me to â€œhurry up and get outâ€ long before she served on my committee. My parents, Dr. Kwablah Awadzi and Patricia Awadzi, have been a major source of inspiration. They have always had high expectations of me and taught me that I can be whatever I want to be as long as I belie ved in myself and worked hard. They provided financial support at various stages of schooling, and ga ve me graduate school survival skillsâ€”a vociferous love of reading, typing lessons at age 12, and computer lessons back in the day when almost all computer functions were executed by command c odes. My sister and friend, Korkoene Awadzi, sustained my endurance level with humor, support, and the unshakable belief that I would make it, my brother, Kwablah Awadzi, Jr., periodi cally sent emails, and my cousin and friend, Christine Naa Norley Lokko, was a source of in spiration with prayer chats and humorous exchanges of our experiences in our doctoral prog rams. I have been blessed to have people who were a source of encouragement and inspirati on to me. To mention a few, I would like to acknowledge my friend Dr. James Okine, who en couraged me enroll in the PhD program and kept tabs on me, and friends Grace Anaglate , Josephine Asmah, Atsuko Kitayama, Adejoke Shitta-Bey, Michiyo Endo, Dr. William Mkan ta, Swathy Sundaram, and Sandra Winter. I recognize Tonja Lindsey of the National Highway Transportation Safety Administration (NHTSA) who answered my ques tions regarding the Fatality Analysis Reporting Systems (FARS) database used in this study, and A bbey Sipp, who provided summaries of state agerenewal license policies.
5 TABLE OF CONTENTS page ACKNOWLEDGMENTS...............................................................................................................4 LIST OF TABLES................................................................................................................. ..........7 LIST OF FIGURES................................................................................................................ .........8 ABSTRACT....................................................................................................................... ..............9 CHAPTER 1 INTRODUCTION..................................................................................................................11 Overview of the Research Topic............................................................................................11 Specific Aims.................................................................................................................. ........11 Driving and Older Adults.......................................................................................................13 Fit of Socio-ecological Determinants of Older Driver Injury to Health Services Research....................................................................................................................... .......15 2 LITERATURE REVIEW.......................................................................................................16 Motor Vehicleâ€“Related Injuries in the US.............................................................................16 Epidemiology of Motor Vehicle Injuries................................................................................18 The Precede-Proceed Model of Health Promotion.................................................................21 Description of the Pr ecede-Proceed Model of Health Promotion...................................21 Domains of the precede-proceed model of health promotion..................................22 Application of the pr ecede-proceed model of health promotion..............................24 The precede-proceed model of health promotion structural model.........................25 Determinants of Injuries among Older Adults........................................................................26 Health......................................................................................................................... .....26 Behavior....................................................................................................................... ...33 Environment....................................................................................................................35 Enabling....................................................................................................................... ....38 Gaps in Literature............................................................................................................41 3 METHODOLOGY.................................................................................................................44 Research Questions and Hypotheses......................................................................................44 Research Question #1......................................................................................................44 Research Question #2......................................................................................................44 Research Question #3......................................................................................................45 Description of the Fatality Analysis Reporting System Database (FARS)............................45 Description of Dependent a nd Independent Variables....................................................48 Data Collapsing and Data Cleaning................................................................................59 Statistical Analyses..........................................................................................................60
6 4 RESULTS........................................................................................................................ .......71 Description of the Sample......................................................................................................71 Univariate Analyses.........................................................................................................71 Bivariate Analyses...........................................................................................................73 Health domain..........................................................................................................73 Behavior domain......................................................................................................73 Environmental domain.............................................................................................74 Predisposing domain................................................................................................77 Reinforcing domain..................................................................................................77 Enabling domain......................................................................................................79 Summary of bivariate results....................................................................................80 Binary logistic regression.........................................................................................80 Summary of Results............................................................................................................. ...84 5 DISCUSSION..................................................................................................................... ..102 Overview....................................................................................................................... ........102 Limitations of the Study................................................................................................102 Procedure of the Study..................................................................................................103 Findings in Light of Research Questions and Hypotheses...................................................104 Injury Prevalence among Independent Vari ables for Younger and Older Drivers.......104 Measures of Association among Independent Variables and Injury.............................105 Age-related License Renewal Policies..........................................................................106 Implications for Policy, Practice and Research.............................................................106 Findings from the final model................................................................................107 Findings from age-related license renewal policies...............................................113 Conclusion..................................................................................................................... .......114 APPENDIX A PRECEDE-PROCEED MODEL OF HEALTH PROMOTION..........................................117 B STRUCTURAL MODEL OF OLDER DRIVER SAFETY................................................118 C INITIAL EXAMINATION OF VA RIABLES & RATIONALE FOR INCLUSION/EXCLUSION BASE D ON CONSENSUS MEETINGS...............................119 LIST OF REFERENCES.............................................................................................................133 BIOGRAPHICAL SKETCH.......................................................................................................142
7 LIST OF TABLES Table page 3-1 Description of independent FARS va riables, variable types and levels............................66 3-2 Age-related renewal policies in 2003................................................................................68 3-3 Reduced renewal cycle requirements in 2003..................................................................69 3-4 In-person renewal requirements in 2003............................................................................70 3-5 Vision, medical, or road test requirements in 2003...........................................................70 4-1 Health, behavioral, environmental, pred isposing and reinforci ng variables of the Precede-Proceed model of health promotion.....................................................................85 4-2 Age-related license renewal polic ies for older adults as of 2003 ( N = 5,747)....................88 4-3 Prevalence of drivers injured in a crash by age group for health, behavior, environment, and predisposing domains............................................................................88 4-4 Prevalence of drivers in jured by type of age-relate d license renewal policy....................92 4-5 Precede-Proceed model of health promotion variables in logistic regression model by domain and statistical significance....................................................................................92 4-6 Binary logistic regression model showing statistically si gnificant age interactions and statistically significant explanatory variab les from the Precede -Proceed model of health promotion with injury (yes/no)...............................................................................93 C-1 Fatality Analysis Report System (FARS) accident level variables.................................119 C-2 FARS vehicle level variables...........................................................................................124 C-3 FARS driver level variables.............................................................................................127 C-4 FARS person level variables............................................................................................130
8 LIST OF FIGURES Figure page 2-1 Motor Vehicle Traffic Fatality Rates by Age Group, 1995. National Highway Traffic Safety Admini stration (NHTSA, 2006).................................................................42 2-2 Motor Vehicle Traffic Fatal ity Rates by Age Group, 1995. NHTSA (2006)..........43 4-1 Drivers injured by age group.............................................................................................95 4-2 Drivers injured in crash by gender.....................................................................................96 4-3 Drivers injured in crash by driver license compliance.......................................................96 4-4 Drivers injured in cras h by restraint system use................................................................97 4-5 Drivers injured in crash by time of day.............................................................................97 4-6 Drivers injured in crashes by vehicle body type................................................................98 4-7 Drivers injured in cras h by most harmful event type.........................................................98 4-8 Previous accident convictions by injury outcome and age group......................................99 4-9 Previous driving while impaired (DWI) convictions by injury outcome and age group.................................................................................................................. ...99 4-10 Previous suspension convicti ons by injury outcome and age group................................100 4-11 Previous speeding convictions by injury outcome and age group...................................100 A-1 Precede-Proceed model of health promotion...................................................................117 B-1 Precede-Proceed model of health promotion structural model from older driver systematic literature review.............................................................................................118
9 Abstract of Dissertation Pres ented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy SOCIO-ECOLOGICAL DETERMINANTS OF INJURY AMONG YOUNGER AND OLDER ADULTS INVOLVED IN FATAL MOTOR VEHI CLE CRASHES IN THE UNITED STATES By Kezia Dzifa Awadzi December 2006 Chair: R. Paul Duncan Cochair: Sherrilene Classen Major: Health Services Research Using a socio-ecological model, the Preced e-Proceed model of health promotion (PPMHP) and the 2003 Fatality Analysis Repor ting System database, this study examined determinants of injuries among older drivers involved in a fatal crash, comparing them with younger drivers (35 to 54 years). The study also explored relationships between age-related license renewal policies and injury outcome for older adults. Significant risk and protective factors for older drivers were number of passeng ers, whether the driver was the registered vehicle owner, the principal point of impact, a nd previous number of mo tor vehicle convictions (e.g., failure to stop at a light). For both age groups, most of the signifi cant risk and protective factors were environmental, with actions taken by the driver just before the crash (most harmful event), having astronomical risks of in jury, with the odds of injury between 31 and 266. Females had a higher risk of injury in a crash compared to males. Behavioral variable s (e.g., restraint system use) and the predisposing variable (vehicle maneuver) were signifi cantly associated with injury for younger and older drivers.
10 Older drivers from states with age-relate d license renewal polices had lower injury prevalence rates compared with drivers with licenses from stat es without any license renewal policies. States requiring in-person renewal had lower injury rates than drivers from states without in-person renewal. However, drivers wi th licenses from states with reduced renewal cycles and vision/medical tests did not significantly differ in in jury rates from drivers with licenses from states without those requirements. Using the PPMHP to examine risk and protective factors for motor vehicle injury among younger and older drivers in the U.S. demonstrated that a socio-ecological approach is needed to understand the factors associated with injuryâ€”an approach not yet utilized in the existing older dr iver literature. Many of the findings showed relevance to drivers from both age groups, with only a se lected few pointing to older adults, meaning that inju ry prevention measures, when developed and implemented, may benefit older and younger drivers alike.
11 CHAPTER 1 INTRODUCTION Overview of the Research Topic This research topic is part of a project funded by the Centers for Disease Control and Prevention (CDC) (Classen, 2004). Th e objective of the project is to develop a public health model to promote safe elderly driving. The pr oject has three specific aims: (a) use a socioecological model as a framework to conduct a system atic literature review (SLR) of older driver studies, (b) evaluate the fit of existing data on older drivers with the model, and (c) develop a model-driven older driver safety program, showin g the interaction of prediction factors and plan for pilot testing (Classen , 2004). To meet the first objective of the project a structural model was created from the results of a SLR (Classen et al., 2006). For the second aim of the project, quantitative and qualitative data fr om existing sources were used to test findings from the SLR. This dissertation is the quantitative phase of the Public Health Model to Promote Safe Elderly Driving project. A national secondar y database, the 2003 Fatality Analysis Reporting Systems (FARS) was used to ascertain relationships among the determinants of motor vehicleâ€“related injuries among older drivers and injury outcomes. Findings from the quantitative and qualitative phases will be integrated to plan an older driver injury prevention program (Classen, 2004). Specific Aims Using the Precede-Proceed model of health promotion (PPMHP) as the organizing framework with the findings from the SLR, th is study examined the relationships among socioecological variables to the level of injury outcomes in motor-vehicle crashes. To identify the age effect in the relationships, a nd better understand the determinan ts of motor-vehicle related injuries among older drivers, I compared two age groups: 35 to 54 years and 65 years and older (using the 35 to 54 age group as a control). The 35 to 54 year age group was selected because it
12 had a lower prevalence of motor vehicle traffi c fatality rates between 1995 and 2005 compared to the 65 years and older group (NHTSA, 2006a). The aims of the study were to: (1) examine the relationship among the socio-ec ological variables to injury within each of the two age groups (old 65 years and older; young, 35 to 54 years); (2) examine if there are age differences betw een the socio-ecological variables and injury outcomes following a motor vehicle crash; (3) examine the relationship be tween state licensure re newal policies and motor vehicle injuries fo r older drivers. Why Examine Older Drivers and Injury-Related Crashes? Motor vehicle injuries are classified as unint entional injuries because the injury occurs over a short time, the harmful outcome was not sought, and the injury resulted from physical energy in the environment (Waller, 1985). Mo st unintentional injuri es (no longer called â€œaccidentsâ€) are avoidable because they are predicta ble events with identifiable origins and risk factors (Christoffel & Gallagher, 2006). Older drivers (65 years and olde r) are more likely than other adult drivers to pose injury and fatality risk to themselves and older passeng ers in their vehicle than the occupants of other vehicles (Braver & Trempel, 2004). Unlike other age groups who display unsafe behaviors such as speeding, alcohol abuse and reckless dr iving (CDC, 2004; NHTSA, 2006b), older driver crashes often stem from physical, sensory, and cogn itive declines as a result of the normal aging process (Carr, 1993). Other contributory crash fa ctors include medical conditions and medication use (Ray, Thapa, & Schor, 1993). On the other hand, older adults are more likely to be wearing seatbelts and to limit their driving to favorable driving conditions than ot her age groups (Bauer, Adler, Kuskowski, & Rottunda, 2003), and are less likely to be driving under the influence of alcohol or substance abuse than other adults (NHTSA, 2006b).
13 The population 65 years and older will represen t about 20% of the entire U.S. population by 2030 (He, Sengupta, Velkoff, & DeBarros, 2 005). In 2004, there were approximately 28 million licensed drivers age 65 and older in th e U.S. (NHTSA, 2006a). The number of older drivers is increasing, with over 40 million pr ojected older adult licensed drivers by 2020 (Centers for Disease Control and Prevention, 2005). In 20 01, except for very young drivers, and based on miles driven, drivers (65 years and older) had the highest rates of fatal crashes (NHTSA, 2001; Li, Braver & Chen, 2003). In fact, in 2005, the 65-plus population comp rised 15% of motor vehicle fatalities (NHTSA, 2006a). Currently, th e 65-plus age group is the third highest for motor vehicle traffic fatalities, with 16 to 20 y ear old drivers being firs t, and 21 to 34 year old drivers being second (NHTSA, 2006a). Past and current trends indicate that the pe rcentage of older drivers on the road are increasing and will continue to do so because of the rise in numbers of older drivers and the aging U.S. population. Apart from high motor vehicl e injuries and fatalities among older adults and the changing demographics in the U.S. population, another reason for understanding the relationship between the determinan ts of motor vehicle injuries a nd older adults is the role of driving in the lives of older American adults. Driving and Older Adults Generally, mobility, that is, the use of vehicl e conveyance or travel from one place to another, is an important facet of American living, and the lack of being able to move from one place to another has practical a nd psychological ramifications to individuals and the society. The mission of the Federal Transit Administration (D OT), based on the Civil Rights Act of 1964, is enhancing the social and economic quality of life for all Americans by providing nondiscriminatory transportation (FTA, 2006).
14 Older adults may stop driving for reasons such as age-related declines and chronic health conditions. Although the availability and accessibility of alternative transportation vary from location to location (e.g., rural areas having fewer alternative tr ansportation opportunities than urban areas) elderly Americans rely primarily on driving automobiles for access to employment, commerce, family, and friends (Houser, 2005; R itter, Straight, & Evans, 2002; Burkhardt, Berger, Creedon, & McGavock, 1998). Older adults may stop driving or self-restrict driving because of age-related declines and acute a nd chronic health conditi ons taking their toll on functioning. In fact, Foley, Harley, Heimovitz, Guralnik & Brock (2002) indicate that the lifespan of Americans exceed their driving life expectancy. Men 70 to 74 years have approximately 7 years of dependency on alterna tive transportation, while women 70 to 74 years have about 10 years of relianc e on alternative transportation. Older Americans view driving as a symbol of independence; in fact, driving cessation among older drivers has been associated with in creased depression (Marottoli et al., 1997) and a decreased societal participati on (Marottoli et al., 2000). Others (Ragland, Satariano, & MacLeod, 2005; Ritter et al., 2002; and Bur khardt et al., 1998) also foun d driving cessation among older adults to be associated with isolation, loss of independence, and depression. In summary, this research is important becau se (a) motor vehicle injury is preventable; (b) older drivers are increasing in numbers and driving longer; (c) older drivers are at the third greatest risk for injuries and fa talities in motor vehicle crashes (NHTSA, 2006a); (d) driving is an integral facet of living for older adults; and (e) the lack of acces sible, affordable and appropriate alternative transportation services for ol der adults have implica tions for public health and possibly health services.
15 Fit of Socio-ecological Determinants of Older Driver Injury to Health Services Research The economic cost of motor vehicle crashes in the United States is $231 billion, of which $36.2 billion (14%) was due to medical costs (Blincoe et al., 2002). For non-fa tal injuries (all age groups), 25% of the total cost resulted from medical care (Blincoe et al., 2002). Although specific information on the economic cost of traffic-related injuries for older drivers is not available, previous studies demons trate the impact motor vehicle injury costs have on society as a whole. Older driver s are likely to get injured in motor vehicle crashes, and their fragility due to age and preexisting medical c onditions make them more likely to require hospitalization, longer and more expensive healthcare, and a le ss complete recovery (Dobbs & Carr, 2005; Richmond, Kauder, Strumpf, & Mered ith, 2002). Freeman, Gange, Munoz, and West (2006) found that older adults who had ceased driv ing, never driven, or who did not have drivers in their home had an increased risk of entering a long-term care facility compared with current drivers. Motor vehicle crashes are the second leadi ng cause of traumatic brain injuries in the U.S. for all age groups (Coronado, Thomas, Sattin, & Johnson, 2005). This study is foundational research for older driver studies in health services settings. By establishing risk and protective f actors associated with motor vehi cle injuries for older drivers on a population-based study, subsequent studies can examine how access and utilization of healthcare services may impact mo tor vehicle injury prevention. This study is germane to health services research in many ways, including heal th education, screening, assessing, and counseling older drivers. Since the healthcare costs of trea ting older adults are predominantly borne by the tax-supported Medicare program, improved unders tanding of this impact has significant implications for healthcare pol icy and healthcare financing.
16 CHAPTER 2 LITERATURE REVIEW In this chapter I discuss the history and ph ilosophy of injury epidemiology, describes the PPMHP, and notes findings from an older driver systematic literature review (SLR). I then discuss studies on motor vehicle-related injuries among older drivers with in the framework of the Precede framework of the PPM HP. I conclude with a discussi on on gaps in the older driver literature. Motor Vehicleâ€“Related Injuries in the US Motor vehicle injury is a pertinent public health concern in the United States and has implication in health services and the U.S. ec onomy as a whole. Older dr ivers have the highest rates of hospital admission resulting from cras hes, followed by senior passengers (Peek-Asa, Dean, & Halbert, 1998) and older drivers are more likely to die or be hospitalized compared with middle-aged drivers (Finison & Dubrow, 2002). Mo tor vehicle crashes are the second leading cause of traumatic brain injuri es in the U.S. for older adults (Coronado, Thomas, Sattin, & Johnson, 2005). Using population-based data from the Center s for Disease Control and Prevention (CDC) traumatic brain injury (TBI) surveillance system from 15 states, Coronado et al. (2005) examined epidemiological and clinical ch aracteristics of 17,657 older adults hospitalized with TBI in 1999. They stratified the sample into three age groups: 65 to 74 years, 75 to 84 years, and 85 years and older. Motor vehicle-related TBI injury rate s were also categorized by driver, passenger, pedestrian, and â€œotherâ€. Common comorbid cond itions for older adults with motor vehiclerelated TBI were hypertension, diabetes, cardiac arrhythmias, chronic pulmonary disease, and fluid and electrolyte disorder s. Motor vehicle injuries were the second cause of TBI hospitalizations for all age groups, regardless of whether the older adult was a driver, passenger,
17 or pedestrian. Specifically for drivers, TBI hospi talizations increased with age among the 65 to 75 (11.9 per 100,000) and the 75 to 84 (15.3 per 100,000) age groups, and declined for the 85 years and older group (12.7 per 100,000). Across a ll age groups, drivers had the highest motor vehicle-related TBI hospitalizations . Older adults who died in hosp ital or were discharged into residential facilities increased with age. Richmond, Kauder, St rumpf, and Meredith (2002) had similar findings in a retrospective study examin ing the characteristics and outcomes of TBI among 38,707 older adults (65 years and olde r) in Pennsylvania between 1988 and 1997. TBI injury from motor vehicle was the second cau se of TBI hospitalization. Risk factors for mortalities, complications, and whether the older adult was discharged to return home or into skilled nursing facility include injury severity, comorbid or medical complications (e.g., cardiovascular and pneumonia), and age. Compared to having a low injury score (0), older adults with injury scores greater than 25 ( OR = 25.51; CI = 14.5.8), injury score 16-25 ( OR = 4.65; CI = 2.9.4), and injury score 10-15 ( OR = 2.76; CI = 1.7.4) had increased risks for mortality. Approximately 33% were discharged into a skilled nursing facility, a rehabilitation center, or another hospital. Freeman, Gange, Munoz and West (2006) f ound significant associations among older adults (65 to 84 years old) who had never driven or had ceased driving and admittance into longterm care facilities. The sample comprised of 1,5 93 older adults living in Maryland selected from a previous study using the Health Care Financ ing Administration Medica re database. Data on their driving status and entry into long-term care facilities were obtained vi a self-report or from a proxy. After adjusting for health conditions and demographic factors using Cox time-dependent regression analyses, former drivers ( HR = 4.85; 95% CI = 3.26.21) and adults who had never driven ( HR = 3.53; 95% CI = 1.89.58) had an increased hazard ratio of admission in a long-
18 term care facility. Also, not havi ng a driver at home was a signi ficant risk factor for admission into a long-term care facility. In summary, compared to younger age groups, older motor vehicle injuries among older adults is associated with increased risks of mortality, exacerbated by comorbid conditions. Secondly, older drivers have highe r hospitalization rates for TBI compared with other persons involved in motor vehicle crashe s, and thirdly, older adults di scharged from hospital may be admitted into another healthcare facility. In th e following section, I discuss the epidemiology of motor vehicle injury in the United States. Epidemiology of Motor Vehicle Injuries In order to understand how to prevent or de crease motor vehicle-related injuries among older adults, one has to comprehend the epidemiol ogy of injury. Injury is defined as damages to the body produced by energy exchange s that are manifested within 48 hours or usually within considerably shorter periods (Haddon, 1980). A few decades ago, injuries were viewed as accidents (random and uncontrollable), with some persons deemed more prone to accidents than others. The term accident is from an era when there were no scientific explanations for, or means of dealing with, plagues and natural disaster s (Haddon, 1999, p.231). In the last few decades, the concepts of injury has evolved to include e nvironmental (e.g., vehicle) and social contexts (Christoffhel & Gallagher, 2006). Injury is cl assified as intenti onal (e.g., homicide) and unintentional (e.g., motor vehicle in juries). Gordon (1949) was the first to use epidemiology in conceptualizing injury prevention. He demonstrated that injury, like other health conditions, had patterns of distribution within the population. These included s easonal variations, geographic, socioeconomic, and rural-urban distributions wh ich behave in similar fashion as infectious diseases (Gordon, 1949; Haddon, 1980, p.412) . Like diseases, injury is caused by a combination of factors including host (e.g., the driver), the agent (e.g., kinetic energy), the environment (e.g.,
19 the highway), and the vector (e.g., the motor ve hicle) with these factors interlinked and influencing one another. Gordon (1949) defined the environment as the physical environment (geographic location), the biol ogic environment, and the socio-economic environment. Haddon (1963) suggested that although the cau ses of accidents are multifaceted, it is fitting to investigate any commonality between th ese various factors and injury in order to prevent injury. Using that appro ach, Haddon classified injuries into two groups: (a) injury caused by interference with normal whole body exchange (e.g., drowning or strangulation), and (b) injuries caused by delivery of energy (mechanical , thermal, or chemical) to the body that exceeds the bodyâ€™s threshold levels, (e.g., motor vehicle injuries). Haddon postulated that if injury cannot be prev ented, then its severity can be minimized in one of the following 10 ways: 1. Avert excess amounts of energies to the body at the early stages of ev ents leading up to the injury. 2. Prevent events from occurring (i.e., preven t or amend the release of the energy). This can be achieved for through improve d vehicle designs (Haddon, 1999). 3. Attempt to remove the person from the surrounding area. 4. If (1) to (3) fail, introduce a barrier that would block or decrea se its effect on the person. 5. Separate in space or time the energy be ing released (e.g., divided highways). 6. Separate by interposition of a material barrier (e.g., airbags in motor vehicles) 7. Modify appropriately the c ontact surface, subsurface, or basic structure (e.g., highway modification).
20 8. Strengthen the structure, living or nonlivi ng that is otherwise damaged by the energy transfer (e.g., minimum standard fo r motor vehicles manufacturers). 9. Move rapidly to detect and evaluate dama ge that is occurring or has occurred. 10. After injury has occurred, improve reperc ussions via emergency care, health care, rehabilitation, and in jury prevention programs (Haddon, 1963, Haddon, 1973). Haddon developed a two-dimensional matrix to he lp organize the factors that contribute to motor vehicle injuries into pre-crash events, crash events, and post-crash events in order to enable minimum loses (Haddon, 1972; Ha ddon 1980). Haddonâ€™s two-dimension matrix examines the sequence of eventsâ€”pre-crash, cras h, and post-crash events â€”in relation to the factors involved (human, vehicles and equi pment, physical environment and roadway, and socioeconomic environment). The advantage of us ing Haddonâ€™s matrix is that different effective interventions can be evaluated on a cell-by-cell level. Haddon advo cated for as much or greater emphasis be put on proven measures in the passi ve measures such as improving the vehicleâ€™s crashworthiness or highway engine ering, rather than the active ar eas of injury prevention that places emphasis on changes in behavior (Haddon, 1972). Gielen and Sleet (2003) define passive strategies as dependence on changing products or environment to make them safer for all regardless of behavior, and passive strategies to injury preventi on as entailing or encouraging people to play an active role in protecting themselves despite h azards in the environment (p.65). Successive researchers such as Gielen (1992) and Runyan (1998) suggest limitations in that although Haddonâ€™s matrix included behavioral intervention strategies, the matrix failed to emphasize them, and greater importance should be placed on active measures (e.g., behavior) as well passive injury intervention strategies. Glan z and Rimer (1995) classified injury-related behavior on three levels: (a) th e intrapersonal level (e.g., the e ffect of individualâ€™s knowledge,
21 attitudes, and beliefs on behavi or); (b) the interpersonal leve l (e.g., the influence of family members, friends, and coworkers on an individualâ€™ s behavior); and (c) the community level (e.g., organizational settings and their influence and othe r society influences such as poverty). Because behavioral injury interventions are usually po licy-related, it would be effective to develop behavioral intervention strategi es on a community level instea d of an indivi dual perspective (Gielen & Sleet, 2003). A single stra tegic approach to reducing mo tor vehicle injuries is not enough, but strategies combining both environmen tal modification and behavioral change are needed (Sleet, 1987). Intervention models such as the Precede-Proceed model of health promotion (PPMHP), approach prevention strategi es from a socio-ecological perspective and accounts for the health (e.g., age), behavioral (e.g., seatbelt use), environment (e.g., highway design), educational (e.g., screen ing and assessment), and policy (e.g., state age renewal policies) perspectives. The Precede-Proceed Model of Health Promotion In this section, I (a) describe the domains of the PPMHP, and (b) discuss the application of the PPMHP from population health st udies and health services research. Description of the Precede-Pro ceed Model of Health Promotion The PPMHP consists of the PRECEDE (pha ses 1) and the PROCEED (phases 5). PRECEDE, the acronym for P redisposing, R einforcing, and E nabling C onstructs in E ducational/E cological D iagnosis and E valuation, is a series of assessments that enables researchers to collect information used to gui de decisions (Green & Kreuter, 2005). PROCEED, the acronym for P olicy, R egulatory, and O rganizational C onstructs in E ducational and E nvironmental D evelopment, is the process of planning and carrying out interventions based on information from the Precede assessments (Appendix A).
22 This dissertation focuses on the PRECEDE framework because the focus of the Public Health Model to Promote Safe Elderly Drivi ng project is to gather information for the PROCEED phase intervention stag e. The PRECEDE framework comprises of four assessments: (a) quality of life, (b) epidemiological assess ment, (c) educational and ecological assessment, and (d) administrative and policy assessment (Green & Kreuter, 2005). Domains of the precede-proceed model of health promotion Social assessment: This is defined as â€œthe assess ment in both objective and subjective terms of high-priority problem s or aspirations for the common good, defined for a population by economic and social indicators a nd by individuals in terms of their quality of lifeâ€ (Green & Kreuter, G-8). In the PPMHP, the social a ssessment phase has one domainâ€”quality of life. During the social assessment phase, needs and as pirations of the target population are assessed from the populationâ€™s perspective on individual and community (e.g., city, state, region, or county) levels (Green & Kreuter, 2005). Epidemiological assessment: The aim of the epidemiological assessment is two-fold: (a) identify health problems or goals that may affect the quality of life of the target population, and to (b) identify etiological factors in the three domains of the epidemiological phase (genetics, behavior, and environment) and th eir direct and combined effect on health. The epidemiological assessment enables researchers to prioritize heal th problems of the community and their direct and combined effect on health (Green & Kreuter, p.81). The International Classification of Functioni ng Disability and Health (ICF) model is incorporated in the health domain of the PPMHP. The ICF model was developed by the World Health Organization in 2001 to include the classi fication of non-fatal disease outcomes. The ICF model consists of (a) body functions, body struct ure, and participation, and (b) environmental and personal factors (Dahl, 2002; Ustun, Chatterj i, Bickenbach, Kostanjsek, & Schneider, 2003).
23 In the ICF model, health comprises of th ree components: (a) body functions (e.g., mental function), (b) body structures (e.g., structures of the nervous sy stems), and (c) activities and participation (mobility). Educational and ecological assessment: During the educatio nal and ecological assessment the researcher examines factors wi thin the predisposing, reinforcing, and enabling domains that are likely to influe nce health behavior, environmenta l factors, and the interaction between genetics, behavior, and environmen t (Green & Kreuter, 2005, p. 14). This phase prioritizes behavioral and environm ental factors related to health and quality of life issues after identifying health determinants and social conditions (Green & Kreuter, p. 147). A predisposing factor is defined as â€œany ch aracteristic of a person or population that motivates behavior prior to the occurrence of the behavior â€ (Green & Kreuter, G-6). Predisposing factors are influenced by sociodemographic factors, such as age, race/ethnicity, and socioeconomic status. They are the beliefs, attit udes, perceived needs, and abilities that govern individual or group behavior a nd are forbearers of behavioral change (Green & Kreuter, 2005), and are therefore important to consider when assessing a target population. An example related to older drivers in the United Stat es is the value placed on independe nce in the American culture. An enabling factor is defined as â€œany charac teristic of the environment that facilitates action and any skill or resource required to attain a specific behaviorâ€ (Green & Kreuter, G-3). Examples of older driver resources are health care providers, and older adult advocate groups. Assessment of enabling factors is critical to th e PRECEDE phase because it enables researchers ascertain available resources to the population unde r investigation. A reinforcing factor is defined as â€œany rewa rd or punishment following or anticipated as consequence of a behavior, serving to strengthen the motivation for the be havior after it occursâ€
24 (Green & Kreuter, G-7). Reinforcement may o ccur through negative or positive feedback from society, professionals, or peers (Green & Kreuter) , or through policy enforcement such as motor vehicle citations. The administrative and policy assessment: This is defined as â€œthe analysis of policies, resources, and circumstances prevailing in an orga nizational situation to f acilitate or hinder the development of the health programâ€ (Green & Kreu ter, G-1). This stage is attained after data have been obtained from the social assessment, health, and the educational and ecological phases of the model. This is the beginning of the PROCEED phase of the model. The PROCEED phase of the model, which is not discussed in detail w ithin the context of my di ssertation, comprises of the following phases: (a) implementation of th e health program, (b) process evaluation, (c) impact evaluation, and (d) outcome evaluation. Application of the precede-procee d model of health promotion The PPMHP has been used over 30 years in more than 950 studies (Green & Kreuter, 2005). It is very suitable for population health studies that examine risks and disease burdens among different social groupings within a popula tion (Labonte, Polanyi, Muhajarine, McIntosh, & Williams, 2005). The PPMHP has been applied in health promotion studies, defined as â€œany planned combination of educati onal, political, regulatory, and or ganizational supports for actions and conditions of living conducive to the health of individuals, groups or communitiesâ€ (Green & Kreuter, 2005, p. 506). The model has also been applied within the heal thcare setting, in areas such as disease prevention, and disease manageme nt programs. It has also been applied as a framework in most unintentional injury research by assessing ri sk factors then to design programs for intervention (Trifile tti, Gielen, Sleet, & Hopkins, 2005) . Areas of relevance for the PPMHP include population-based intervention planning and program evaluation such as compliance with child passenger safety laws in Washington (Chang, Eb el, & Rivara, 2002), and
25 planning an intervention program for child pedest rian injury (Howat, Jones, Hall, Cross, & Stevenson, 1997). The PRECEDE framework has been used to assess behavioral and environmental risk and protective factors to pr event alcohol-related cr ashes (Simons-Morton et al., 1989). Application of the PPMHP in health services research incl ude evaluating programs among health professionals (Chande & Kimes, 1 999), assessing the attitudes and beliefs of health professionals toward domestic viol ence (Sugg, Thompson, Thompson, Maiuro, & Rivara, 1999), assessing physician roles in disease ma nagement (Mann & Putnam, 1989; Mann, Lindsay, Putnam, & Davis, 1997), and improving physicia n services (Solomon, Hashimoto, Daltroy, & Liang, 1998). The PPMHP has been applied in pr eventive programs such as the implementation of breast cancer screen ing (Mahloch, Taylor, Taplin, & Ur ban, 1993; Taylor et al., 1998), cervical cancer screenin g (Taylor et al., 1999), and prostate cancer (Weinrich, Weinrich, Boyd, & Atkinson, 1998). In some studies, researchers ta rgeted both health professionals and patients to evaluate intervention programs in hosp itals (Brink, Simons-Morton, & Zane; 1989). In summary, although the PPHMP has been wi dely applied in population health studies and health services research, up to date, it ha s not been used in motor vehicle injury-related studies for the older adult population. Studies on motor vehicle injury prevention studies have focused on children. In the following section, I de scribe the PPMHP structural model from the SLR, which is the foundationa l work for this dissertation. The precede-proceed model of heal th promotion structural model In this dissertation, I used the PPMHP stru ctural model based on the older driver SLR conducted by the Public Health Model to Pr omote Safe Elderly Driving project in 2005 (Appendix B). The PPMHP structural model is ba sed on the analysis of 201 primary studies on older driver safety. The incl usion criteria were: (a) publishe d and unpublished sources, (b) sources published in English, (c) studies pertaini ng to adults (60 years and older), (d) U.S.
26 sources that were published or completed be tween January 1985 and April 2005, (e) studies relating to safe or unsafe drivi ng (e.g., motor vehicle in juries, motor vehicle fatalities, and motor vehicle citations). Exclusion crite ria were (a) sources with no pr imary findings, (b) duplications of primary studies, and (c) studies pertaining to simulator studies (C lassen et al., 2006). Data from sources were extracted using th e Systematic Process for Investigating and Describing Evidence Based Research (SPIDER) t ool developed by the Public Health Model to Promote Safe Elderly Driving team . Meta-Synthesis and content analysis were used to analyze data from the SLR, and ascertain risk and protectiv e factors for older driver safety. The structural model explained significant re lationships among independent va riables and driving safety outcomes. Findings indicated that majority of studies focused on the health domain (61%) and less on environmental (20%), he alth education (2%), behavior al (10%) and th e predisposing (1%), reinforcing (3%), and enabling (2%) domains (Classen, et al., 2006). Determinants of Injuries among Older Adults Using the PPMHP, the older driver literature ne xt discussed, identifies risk and protective factors for injuries within the context of the health, behavioral, environmental, and enabling domains of the PPMHP. Although I have grouped th e studies under specific domains, they are not mutually exclusive e.g., some environmental studies also contain in formation on behavior and health variables. Literature in this sect ion broadly pertains to the health, behavior, environment, and enabling factors. Health Under the health domain of the PPMHP, risk and protective factors identified from the SLR on older drivers include age, gender, acu te and chronic physical conditions, mental conditions, and number and classes of medicatio ns used by the older dr iver (Classen, et al., 2006).
27 Older adults comprise of 7% of motor vehi cle injuries (NHTSA, 2006a ). Older adults (65 years and older), are the third highest age group fo r fatal injuries (Figure 2-1). Older drivers are likely to get injured in motor vehicle crashes, their fragility due to age and preexisting medical conditions makes them more likely to require ho spitalization, longer and more expensive health care, and a less complete recovery (Dobbs & Carr, 2005; Richmond, Kauder, Strumpf, & Meredith, 2002). Older drivers i nvolved in crashes are more likely to die from injures than younger drivers. Though compared to younger drivers, older adults have less mileage exposure, the risks of crashes, injuries, a nd fatalities increase with age, (Dellinger, Langlois, & Li, 2002). Differences in motor vehicle fatality outcome s for older adults vary within the 65 years and older population (Figure 2-2). Ten-year data (1995 to 2005), indicated differences in motor vehicle fatality rates among five older adult groups: (a) 65 to 69 years old, (b) 70 to 74 years old, (c) 75 to 79 years old, (d) 80 to 84 years old, a nd (e) 85 years and older. The 80 to 84 age group has the highest motor vehicle fatality rates. Over a decade, the 80 to 84 and the 85 years and older groups showed decline in motor vehicle fa talities, however, the age groups 65 to 69 and 70 to 74, maintained stable motor vehicle fatality rates. The 35 to 54 age group has lower motor vehicle fatality rates compared to older drivers (65 years and older). Latest motor vehicle fatality data indicates that older adults (65 years and older) have a motor vehicle fatality rate of 21 per 100,000 residents. Within the older driver popula tion, the 65 to 74 age gr oup has the lowest motor vehicle fatality rate (16.2 per 100,000). The mo tor vehicle fatality rates increase for the 75 to 84 age group (24.9 per 100,000), and the 85 years and over age group (28.8 per 100,000) (National Center for Health Statistics [NCH S], 2006). When exposure (miles driven) is taken into consideration, people age 75 and older have more motor vehicle fatal injuries per 100,000 population and higher fatality rates from motor vehi cle crashes per mile driven than other groups
28 except people younger than 25 (FHWA, 2006 Hi ng, Stamatiadis, & Aultman-Hall, 2003). Similar findings were observed in a study using 1990 FARS data, 1990 General Estimates System (GES) data, and the 1990 Nationwide Personal Transportati on Survey. Very young drivers (16 to 19 years) and older drivers (75 an d above), had higher odds of being fatally injured in a motor vehicle crash after controlling fo r miles traveled (Massie & Campbell, 1993). There are racial differences for motor vehicl e fatality rates (NCHS, 2006). Based on 2003 data, within the male gender, Hispanic or Lati no males have the highest motor vehicle fatality rates per 100,000 resident population (29.2), foll owed by the Black male (28.9 per 100,000), the White non-Hispanic male (28.6 per 100,000), the Amer ican Indian or Alas ka native male (25.4 per 100,000) and the Asian or Pacific Isla nder male (20.1 per 100,000) Among the female gender, the American Indian or Alaska native female has the highest motor vehicle fatality rates across both male and female racial groups ( 32.3 per 100,000), followed by the Asian or Pacific Islander female (16.2 per 100,000), the White non-Hispanic female (15.8 per 100,000), the Hispanic or Latino female ( 13.1 per 100,000), and the Black or Af rican American female (12.4 per 100,000). Gender differences in motor vehicle crash out comes have been found, specifically for older adults. Generally, older female drivers have higher injury (including fatal injuries) rates compared to male drivers. Using the 1975 to 1988 FARS data, Bedard, Guyatt, Stones, & Hirdes (2002) studied driver, vehicle, and crash factors to determine their relationships to fatal injuries, and stratified the data by age to determine any differences among age groups. Results indicated although there were no gender differences for younge r drivers, older male drivers had a higher proportion of fatal injuries compared with older females. However, in a multivariate logistic regression, females had a 54% increased risk of fatal injury compared to males.
29 Using 1982 to 2001 FARS data, Baker, Falb, Voas, & Lacey (2003) studied the driving characteristics of women 70 years and older in the U.S. over the 5-year period and compared with women in age groups 30 to 49 and 50 to 69. The objective of the study was to identify environmental factors (vehicle and road) associat ed with crashes among older females. The areas of interest to the res earchers were times and conditions of motor vehicle crashes, types of crashes, conditions of the vehicles, characteristics of collisions, and propensity of getting injured. A series of loglinear analyses were used to examine the relationships among the environmental variable and crashes, and the es timates used to ascertain the relationships among environmental variables and crash was the likelihood ratio. Females 70 years and older were more likely to have crashes in good weather, in daytime, and on roads with low speed limits. There were al so a significantly larg er proportion of older females who collided with fixed objects in a motor vehicle crash (Baker, Falb, Voas, & Lacey, 2003). For the FARS variable most harmful event , older women had a higher risk of collision with a fixed object ( LR = 1.11) compared with collision with another motor vehicle (not statistically significant), coll ision with a moving object ( LR = 0.94), and non-collision ( LR = 0.92). For the initial point of impact (the angle at which the car was struck first), older women were at a greater risk of being struck in front ( LR = 1.04), compared to the passenger ( LR = 0.93) and driver side ( LR = 0.97), which both had a lesser risk of being struck in the crash. A study of 300 adults 62 to 89 years of age i ndicated that generally for both males and females, driving patterns change with the chan ces of driving every da y decreasing by about 6% after the age of 72, with females having 3 times th e odds of reducing their driving compared with males. Bauer et al (2003) surveyed older driv ers and collected information on driving history, driving patterns, and demographic information. Us ing bivariate analysis and a binary logistic
30 regression method, they predicted changes in driv ing behavior and destin ations based on age and gender (Bauer et al., 2003). Othe r gender differences included driving habits, with women less likely to driver in adverse conditions, and 75% ( p = 0.02) more likely to reduce night driving compared with males, with the odds of driving less everyday 3.14 ( CI = 1.94.13) times more for women than men (p.8). Finison and Dubrow (2002) used the Main e Crash Outcome Data Evaluation System (CODES), a dataset that links poli ce crash data to hospital and deat h certificate data in this study Age was stratified into young drivers (16 to 24 ye ars), middle-aged drivers (25 to 64 years), and older drivers (65 years and over). Bivariate analys es and a multivariate logistic regression were conducted. Outcomes of interest included crashe s, injury, fatalities and hospitalization, and mortality. Independent variables consisted of drivi ng behavior variables such as failure to yield, and environmental variables such as, light condi tions, road conditions, and rural/urban location. On the univariate and bivariate levels, compared with middle-aged drivers, older drivers were more likely to crash at lower speed s, intersections, and in urban area. They were also more likely compared with middle-aged drivers, to driver-relat ed factors such as failure to yield or making a left turn. With regard to the outcome variables, adults were at higher risk of injury resulting in hospitalization and/or death, comp ared with younger and middle-aged drivers, and the risk of fatality and hospitalization increased by age for th e older driver. For older drivers only, female had a 1.62 odds of getting hospitalized or dying from the crash compared with males ( CI = 1.01â€“ 2.59). Age-related declines in sensory conditi ons include loss in vision (e.g., increased sensitivity to glare, reduced contrast sensit ivity, and poor night visi on) and hearing loss.
31 Decreases in cognitive abilities in clude the reduced ability to rapi dly process information and the reduced ability to quickly switch attention between tasksâ€”both of which are necessary skills in practicing safe driving. Physical declines pert inent to driving include the reduced range of motion of the neck and head, slowing down res ponse time to execute vehicle control (Staplin, Lococo, Byington, & Harkey, 2001). These age-re lated conditions are aggravated by medical conditions, medications, and/or medication interactions. Physical and mental conditions such as Al zheimerâ€™s have been shown to influence driving outcomes among older adults. Carr (1993) cl assifies skills required for safe driving into perception (vision, hearing, and ra nge of motion), cognition (cogniti ve-related symptoms may be due to acute or chronic diseas es), and execution (coordination a nd motor). In a study comparing a group with Alzheimerâ€™s disease with a c ontrol group, Rizzo, Anderson, Dawson, Myers, & Ball (2000) found that people with Alzheimerâ€™s disease performed less satisfactorily on attention, visual function, and visual processi ng speed. Older drivers may be prone to vision impairments that may make them more suscep tible to motor vehicle injury. These vision impairments include cataracts, glaucoma, diab etic retinopathy, and age-related macular degeneration. Areas of concern are visual acuity, visual fields, contrast sensitivity, and Useful Field of View (UFOV). Visual acuit y and visual fields are tested in vision tests administered by Divisions of Motor Vehicles in the U.S., although the standards may vary from state to state. However, contrast sensitivity and UFOV ar e not measured by DMVs (Owsley & McGwin, 1999). Among visual and cognitive f actors, visual attention and ment al status were the strongest predictors of motor vehicle crashes (Owsley, Ball, Sloane, & Bruni, 1991; Owsley et al, 1998). The contribution of medications to older driver injuries in motor vehicle crashes can be classified by (a) the number of medications (polypharmacy), (b) cl asses of medications used, (c)
32 side effects of medications, and (d) interaction effects due to a dverse reactions, compliance, and medical errors that may influence the drug metabo lism of older adults differently from other age groups as a result of aging. These include redu ced body mass and basal metabolic rate, reduced proportion of body water, increased proportion of body fat, and decreased cardiac output (Hammerlein, Derendorf, & Lowenthal, 1998). Polypha rmacy is defined as one of the following: (a) the use of many medications at the same time, (b) the prescription of more medications than is clinically indicated, or (c) a medical regi men consisting of five or more medications. Polypharmacy may result in adverse drug reactio ns and interactions (Beer, 1997). Effects of polypharmacy include vision, cognitive, or psycho motor disorders and indicates an increased risk of motor vehicle crashes for older adul ts (Carr, 1993; Lococo & Staplin, 2006). Certain classes of medications have been identified which negatively affect the safety of older drivers because they have the propensity of impairing the function of the central nervous system and, indirectly, the driving of olde r adults (Cowart, & Kandela, 19 85). Classes of medication that have shown some association with unsafe driv ing behavior include be nzodiazepines, cyclic antidepressants, hypoglycemic medications, narco tic analgesics, and antihistamines (Ray, Gurwitz, & Decker, 1992; Ray, Thap a, & Shorr, 1993). Overall, tw enty-eight medications have been identified to have the propensity of bei ng harmful to older adul ts compared with younger counterparts (Beers, 1997). In summary, previous older driver studies in dicate that health f actors (e.g., age, race, gender, and physical health) are associated with injury, fatalit y, and crash outcomes among adult drivers, and that some differences have b een observed among younger, middle-aged, and older drivers.
33 Behavior Behavior factors that are pr edictors of motor-vehicle related injuries include selfregulation, drug use, alcohol and driving, and seat belt use. Generally, older drivers self-regulate, with crashes occurring in broad daylight under no n-adverse conditions, and are more likely to be wearing a seatbelt at the time of the crash, and less likely to be under the influence while driving compared with younger drivers (NHTSA, 2006b). Ball, Owsley, Stalvey, Roenker, Sloane, and Graves (1998) studied 257 li censed older drivers (55 years and older) in Alabama. They examined the association between visual and cognitive impairment and avoidance of taxing driving conditions, and the interre lationships among functional im pairment, driving avoidance, and crash risk. Subjects were given an eye health examination and were given tests to measure visual acuity, contrast sensitivity and visual fi eld sensitivity. The cogniti ve function and the size of the useful field of view were also measured. A structured qu estionnaire (the Driving Habits Questionnaire), was used to collect data on how often the subjects drove (driving exposure) and the evasion of potentially challenging driving si tuations. There were seven items on the Driving Habits Questionnaire pertaining to driving avoida nce. The items covered potentially challenging situations such as driv ing at night, driving in high-volum e traffic, driving on expressways or interstate, driving alone, executi ng left hand-turns across oncoming traffic, and driving in the rain. Ball et al. (1998) performed spearman correlations among the avoidance items on the questionnaire, visual and cognitive functi on, eye health, and dr iving exposure. Older drivers with vision and cognitive problem s were likely to self -regulate the driving, e.g., not driving under adverse c onditions, and drove for fewer times a week. The relationship between the driving avoidance items and mental status, driving alone ha d a strong relationship with mental status. Ball et al ( 1998) also classified the subjects into six groups based on levels of functional impairment. They perf ormed a multivariate analysis of variance to ascertain whether
34 there were significant differences among the groups and their patterns of driving avoidance. The results suggested that there were differences among the groups in their driving avoidance patterns based on their level of visual functi oning. However, there were not significant differences for night avoidance. A separate analysis compari ng subjects with cataract a nd those without any cataract showed that there were significant differences among the two groups, w ith the cataract group having significantly higher avoi dance, but there were no differe nces for avoidance at night driving and making a left tur n. A correlation analysis was performed to investigate the relationship among the driving avoidance items a nd crash history. The numbe r of crashes in the five years prior to the study was significantly a ssociated with driving a voidance items such as avoiding driving in the rain, making left hand turns, and driving in rush hour traffic. In summary, Ball et al. (1998) showed associ ation among visual and function impairment and self-regulation (driving avoidance). The levels of visual impairment were associ ated with the type of driving avoidance. Freeman, Munoz, Turano, and West (2006) had similar results. In a study of 2,520 older adults (65 to 84 years old), they measur ed cognition, depressive symptoms, and comorbid conditions. Visual acuity, contrast sensitivity, and lower periphera l visual fields were also measured at baseline and follow-up two years late r. A questionnaire was administered at baseline and follow up to obtain demographic information, medical history, health behavior, and driving history. The study had three outco mes: (a) driving reduction to fewer than 3,000 miles per year for those who were driving 3,000 or more miles at baseline, (b) night dr iving cessation, and (c) cessation of driving in unfamiliar settings. Statis tical analyses comprised of chi-square tests between each independent variable and outcome va riables and five logistic regressions for the vision variables to control for the age effect. Results from the bivariate analysis showed
35 statistically significant relationships between ag e, gender, race, health condition and decreasing driving mileage. Subjects, who were older, fe male, African-American, cognitively impaired and in fair or worse health cond ition were more likely to decreas e driving mileage at follow up. In summary, older adults app ear to self-regulate th eir driving based on health conditions, how confident they feel, and thei r perceived level of risks. Visi on and cognition are some of the reasons why older drivers may change their driving behaviors. Environment The environment is made up of the economic environment, the physical environment, and the social environment (Green & Kreuter, 2005). Most previous older driver studies pertain to the physical and social environment. The physical environmental factors pertinent to older drivers include vehicle conditions, road conditions, highway design, and traffic pattern s. The outcome of a motor vehicle crash for older drivers may be exacerbated by vehicle desi gns that do not accommodate age-related frailty and fragility (e.g., due to body shrinkage from the normal aging process). For instance, headlights that reduce the incidence of glare at night may improve the safety outcome for older drivers (Schieber, 1994; Wang & Ca rr, 2004). Another vehicle factor that particularly influences types of injuries in motor ve hicle crashes is the position of the seat. Hill & Boyle (2006) demonstrated that for small females, rearward seat positions were associated with chest injuries compared to forward seat positions, while males we re more prone to chest injuries with forward positioned seats. The type of injury incurred app eared to be associated with gender. Regarding highway design, older drivers are more likely to get injured or die in crashes at intersections (NHTSA, 2006b; Awadzi, Classe n, Garvan, and Komaragiri, 2006; CDC, 2004). Other roadrelated factors associated with motor vehicle crashes include road surface conditions, time of day, speed limits, rural vs. urban, and whether the highway was divide d or not. Yan, Radwan,
36 and Abdel-Aty (2005).conducted to investigate environmental (dri ver, vehicle and road) risk factors associated with rear end crashes at inte rsections with signals. The sample was stratified by five age groups ranging from less than 26 years of age, to greater than 75 years of age. Health (age, and gender) and behavioral (alcohol and drug use) were also examined. The data used in the study was the 2001 Florida accid ent database. The quasi-induced exposure method was used to calculated relative accident involvement risk rati os (RAIR). The type of crashes was classified as rear end or non-rear end crashes, and a bina ry logistic regression was used to ascertain measures of association between independent va riables and type of cr ashes. The level of statistical significance was set at p < 0.01. Female drivers were at a less risk of striking another vehicle in the crash ( OR = 0.90; CI = 0.84.96). Drivers who struck another vehicle in a crash were more likely to be less than 26 years of age and male, or older than 75 years, wh ile the drivers whose vehicles were struck, were more likely to be older adults. Alcohol use was significantly associated with crashes. Environmental variables that had significant association with t ype of crash (rear-end crashes) included number of lanes, undivided versus di vided highway, time of crash, rural/urban, and speed limits. The risk of rear-end crashes was hi ghest in 6-lane roads, and on roads that were undivided. Also, the risk of rear-end crashes in the nighttime was less than during the day time ( OR = 0.50; CI = 0.66.80). Drivers on wet ( OR = 3.3; CI = 2.26.84), and slippery roads ( OR = 1.79; CI = 1.64.97), had a greater risk of rear-end cras hes compared with drivers traveling on dry roads. The manner of vehicle and relation to junction are also associated with motor vehicle crashes (Bak er et al., 2003). The social environment comprises of the car e networks passengers, and stakeholders. Services include alternative transportation and community mobility (Gre en & Kreuter, 2005). An
37 example of social environment th at is associated with motor ve hicle crashes among older adults is the presence and number of occupants in the motor vehicle with the older driver. Studies indicate that passengers in the motor vehicle ma y influence unsafe actions displayed by older drivers, thus influencing th e probability of a crash. With 1975 FARS data, Bedard and Meyers (2004) demonstrated an association between the number of occupants in a motor vehicle and unsafe actions performed by the older driver . The objective of the study was to examine whether an older adult driving with passengers would be associ ated with risks of performing unsafe actions. Using FARS data (1975 to 1998), a nd a logistic regression model, Bedard and Meyers used, and compared unsafe actions perfor med by the older driver (stratified by age) when alone, with unsafe actions performed when the older drivers had passengers. Older drivers 65 to 79 years ol d with four or more passenge rs had a 27% decreased risk of performing unsafe driver actions ( OR = 0.73; CI = 0.61.86). However, the presence of passengers in the vehicle could be protective or detrimental, depending on the type of unsafe action performed by the older driver. The presen ce of passengers is a protective factor for speeding, lane-related actions, inexperience, following and driving the wrong way ( OR = 0.37; CI = 0.30.46), neutral for passing and disadvantag eous for obeying signs /warning/right of way, turning, and lane changing ( OR = 1.18; CI = 1.15.22). Hing et al. (2003) used poli ce reports from Kentucky for older drivers (65 years and older) and compared single and multi vehicle crashes and stratified the sample by number of passengers (no passenger, one passenger, and two or more passengers). Hing et al. used the relative accident involvement ratio (RAIR) to calculate the ratio of passengers who were at fault, to those who were not at fault in a crash and controlled for time of day, gender, road condition, road type and number of lanes. Logistic regression was used to ascertain the association among
38 independent variables and at fault crashes. Hing et al established that time of day is associated with whether the presence of occupants is a risk or protective factor. Fo r older drivers during the daytime, having two or more passengers is a risk factor in single and multi-vehicle crashes, but a protective factor in the nighttime. Environmental variables encompass vehicle, highway, and social factors. Highway design, angle at which the vehicle was struck, time of day, road conditions, and the presence and number of passengers in the car, are some variables significantly associated with crashes, injuries and fatalities. Enabling Older driver enabling factors for older drivers are stakeholders that provide resources to facilitate safe driving. In the U.S, these incl ude federal agencies, state agencies, professional organizations, and consumer orga nizations. Older driver stakehol ders provide information on (a) safe mobility programs and guides, (b) driver as sessment and rehabilitation programs, (c) driver self-assessment tools, (d) vehicle adaptations, (e ) social service program s, (f) assisting with transportation options, and (g) increasing community awareness on older driv er safety (Eberhard et al., 2006). Federal agencies: Federal agencies, such as the National Highway Traffic Safety Administration (NHTSA), the Federal Highway Administration (FHWA), and the CDC, conduct internal research and provide da ta on older driver safety issues. For example, the NHTSA finds external researchers to establish national safety standards for motor vehicles in the U.S. and provide annual surveillance data on motor vehicl e crashes and related in juries on demographical levels through its national secondary database s the GES and the FARS. NHTSA has partnered with government and private groups to help older drivers maintain their safe mobility (Eberhard et al., 2006). Areas of emphasis in clude alcohol-impaired driving, older drivers, teenage driving,
39 and child passenger safety (CDC, 2004). The CD C has conducted research on motor vehicle fatalities among adults age 65 and over and found variations by gender, race, and ethnicity. NHTSA and FHWA regularly provide the public with information on behavioral, health, environmental, and vehicle factors that impact safe or unsafe driving pr actices of the elderly population. License renewal policies: State agencies, to different extent s, have developed statutes to monitor and regulate older drivers. Currently, abou t half of the states ha ve some form of agerelated driver licensure renewal laws. There is in dication that the type of renewal policies in a state may be linked to the number of motor vehi cle fatalities among older adults (Grabowski, Campbell, & Morrisey, 2004; Levy, 1995). Driver li censure renewal polici es related to older adults include shorter or â€œaccelera tedâ€ renewal cycles, in-person dr iver license renewals (where the older adultâ€™s fitness may be challenged based on appearance), or special tests, e.g., a vision test (Grabowski & Morrisey, 2001 ). Grabowski et al. (2004) used 10 years of FARS data (1990 to 2000) to investigate the influe nce of elderly licensure laws on motor vehicle fatalities and compared older drivers (65 years and older) to a younger group (35 to 64 years of age). The authors examined variables that included the types of elderly licensure renewal laws such as inperson renewal, vision test, road test, and the renewal period. Besi des that, they included other variables considered to confound, such as state speed limits, t ype of seatbelt laws (primary vs. secondary), blood level alcohol, and administrative license suspension, as well as per capita income and unemployment rates, in the analysis. Findings indicated that st ates with vision test requirements, road tests and, and reduced renewal cycles were not signifi cantly associated with risks of fatality, however, in-person renewal was significantly associated with reduced risk of injury fatality ( RR = 0.83 ; C I = 0.72.96).
40 States vary in their guidelines requirements based on age, length of time until renewal, type of testing required, and whether the licen se is renewed in person or by mail (Wang, Kosinski, Schwartzberg, & Shanklin, 2003; National Academy on Aging, 2001). Six states (Florida, Maine, Oregon, South Carolina, Vermont, and Virginia) and the District of Columbia require a vision test for older adults. The age requ irements for vision tests (that test only visual acuity) vary from state to state. Maryland require s a vision test during li cense renewal for people 40 years and older. New Hampshire and Illinois ar e the only states that require seniors 75 years and older to take a road test. Mo st states have a four or five-y ear renewal policy; however, this policy varies, with some states having up to eigh t years (Wang et al., 2003). Fourteen states have an accelerated renewal policy for older drivers. However, the starting ag e for this accelerated program varies from state to state. The special renewal test requirements for older adults vary from state to state (Levy, 1995; Molnar & Eby, 2005). State licensing and renewal policies may contribute to driving behavior and driving cessation among olde r adults. Koulikov (2005) used the Asset and Health Dynamics of the Oldest Old data to examine th e relationship among state licensing and renewal policies and driving cessation among older driv ers. Results from a factor analysis suggested that states with mental testing, in-person renewal at age 70, and restricted licensing requirements were signifi cantly related to the older adul tsâ€™ decision to stop or reduce driving (Koulikov, 2005). Professional and consumer organizations pr ovide guidelines and resources for older drivers and stakeholders. Examples of professional organizations pertinent to older driver safety are the American Occupational Therapists A ssociation (AOTA) and the American Medical Association which developed policy guidelines for physicians to asse ss health condition of patients that may negatively impair driving (Wang et al., 2003).
41 Consumer organizations have collaborated with state and local governments to make programs available to older adults. For example, the American Association of Retired Persons (AARP), the American Society on Aging, the Amer ican Automobile Association (AAA), and the AOTA developed the CarFit program for older driv ers and driver rehabilitation specialists. The CarFit program enables trained personnel to determ ine the fit of the older adult to the vehicle. AARP has driver safety refresher course for adults 50 years and older to enable them to improve driving skills; learn about normal age-related physical changes; reduc e traffic violations, crashes, and chances for injury; improve safe driving outc omes; and have automob ile insurance discounts (AARP, 2006). The AAA Roadwise Review is an interactive CD-ROM which is a selfassessment tool for older drivers to assess their functional ability to drive safely (AAA, 2005). In summary, older drivers have enabling resour ces on the federal, state, and local levels. On the federal level, the NHTSA provides minimu m safety requirements for motor vehicles, and on the state level, some states have age-renewa l license policies for olde r drivers. Professional and consumer organizations offer resources and se rvices to facilitate sa fe driving among older adults. Gaps in Literature The older driver systematic literature review showed that previous older driver studies focused more on the health domain and less on behavioral, environmental, and health education/policy (Classen et al ., 2006). It is therefore importa nt to use a socio-ecological approach to examine how health, environmen tal, behavior, predisposing, reinforcing, and enabling variables are associated with mo tor vehicle injuries and fatalities. Epidemiological data generally focus on motor vehicle crashes and fatal injuries, and not injuries. While it is important to examine the associations among factors and crashes or fatal
42 injuries, it is also imperative in planning an older driver injury prevention program, to comprehend the relationship among healt h, behavior, environment and injury. The Precede-Proceed model of health promoti on has been used in studies within the healthcare system and injury intervention studies for children; however, it has not been used in addressing risk and protective factors for motor ve hicle-related injuries among older adults. This study will address older driver safety from a socio-ecologica l perspective and explore the relationship between licensure policy for older ad ults and injury prevalence rates. Using younger drivers (35 to 54 years) as a comparison group al lows the examination of risk and protective factors of motor vehicle injuries among older ad ults. Findings from these analyses will be integrated with qualitative data to plan an in tervention program for safe driving among older adults. Figure 2-1. Motor Vehicle Traffic Fatal ity Rates by Age Group, 1995. National Highway Traffic Safety Admini stration (NHTSA, 2006)
43 Figure 2-2. Motor Vehicle Traffic Fata lity Rates by Age Group, 1995. NHTSA (2006)
44 CHAPTER 3 METHODOLOGY In this chapter, I present (a) the research questions and hypotheses of the study, (b) the scope of the FARS secondary database and the pro cedure used to include cas es in the dataset, (c) the sample selection process, (d ) the operationalization of depende nt and explanatory variables, and (e) an outline of the methodology used in an swering the research questions of the study. Research Questions and Hypotheses Using the PPMHP, the PPMHP structural mode l, and a national crash dataset, FARS 2003 (DOT, 2003a), I asked three questions: 1. What is the prevalence of the main determin ants (risk and protective factors) and motor vehicle injury for younger drivers (35 to54 year s) and older drivers (65 years and older)? 2. What are the measures of association among socio-ecological determ inants (behavioral, environmental, predisposing, and reinforcing factors) confounding variables (age), and motor injury (injury: yes/no) for younger and older drivers? 3. What is the prevalence of injuries for older drivers among states with age-related license renewal policies compared to states wit hout any age-related license renewal policies (enabling factors)? Research Question #1 Ho: Younger drivers will not have lower prevalence rates for motor vehicle injuries compared to older dr ivers (65 years and older). Ha: Younger drivers (35 to54 years) will have lower motor vehicle injury prevalence rates compared to older drivers (65 years and older). Research Question #2 Ho: Younger drivers (35 to 54 years) will not have lower measures of associations for environmental variables (e.g., hour of day , and most harmful event ), predisposing variables ( vehicle maneuver ) and injury, compared with older drivers (65 years and older).
45 Ha: Younger drivers (35 to 54 years) will have lower measures of associations for environmental variables (e.g., hour of day , and most harmful event ), and predisposing variables ( vehicle maneuver ) and injury, compared with older drivers (65 years and older). Research Question #3 Ho (1): There will be no difference in moto r vehicle injury prevalence rates for older drivers between states with no age-rela ted licensure renewal procedures (enabling variables) and those with age-rela ted licensure renewal procedures. Ha (1): There will be differences in moto r vehicle injury prevalence rates for older drivers between states with age-rela ted licensure renewal procedures (enabling variables) and states with no age-re lated licensure renewal procedures. Description of the Fatality Analysis Reporting System Database (FARS) This section focuses on the scope and proce dure of the FARS secondary dataset, the rationale for using the FARS dataset, and limitations of FARS. Scope of the FARS database This section covers the description of the FARS dataset and the pr ocedure of the FARS data collection. I discuss the background on th e FARS database, its universe and unit of observation, origin of data in th e dataset, period of the data colle ction, and rationale for using the 2003 FARS dataset in this study. Created by the National Center for Statisti cs and Analysis (NCSA) of the National Highway Traffic Safety Administ ration (NHTSA), FARS contains data on a census of fatal traffic crashes within the 50 states, the District of Columbia, and Puerto Rico. National highway safety communities, state governments, local governments, and researchers use the FARS data for motor vehicle injury and fatality trend analyses.
46 NHTSA has funded the FARS datasets since 19 75. The dates of data collection are from 1975 to 2005. Inclusion criteria for FARS are motor vehicles involved in cr ashes while traveling on traffic ways normally open to the public, and a resulting death within 30 days of the crash. The unit of observation in the FARS dataset is a crash. Data are obtained from coded data elemen ts reported on the following forms containing information pertaining to four main areas: (a) the Accident from, (b) the Vehicle form, (c) the Driver form, and (d) the Person form. In 1987, FARS integrated data from the Multiple Cause of Death file from the National Center for Health St atistics. This dataset provides information on the deceased with matching death certificates on educational level, occupation, specific cause of death, specific injuries, race, and ethnicity. FARS began recording race and ethnicity information of the fatally injured from deat h certificates in 2002 (NHTSA, n.d.). FARS database NHTSA has an agreement with individual U.S. states to gather information from Police Accident Reports, State Vehicle Re gistration Files, State Driver Licensing Files, State Highway Department Data, Vital Statistics, Death Certif icates, Hospital Medical Reports, and Emergency Medical Services Reports in a standard format on fatal crashes occurring in the state. Each fatal crash reported encompasses 125 different coded data elements. On a daily basis, data are entered in local computers, a nd then transferred to th e main NHTSA computer. Data are checked automatically for consistenc y, timeliness, and accuracy (NHTSA, n.d.). Files are available in SAS and sequential ASCII file formats and convertible into other statistical software programs. Rationale of selecting the FARS dataset The FARS database contains health (age, gender, race, physical and mental health conditions), behavioral (r estraint system use, alcohol and dr ug use), environmental (physical and
47 social), predisposing (vehicle maneuver), and re inforcing (number of previous motor vehicle convictions) variables that are useful for analyses using a so cio-ecological framework. I used the 2003 dataset because it was the late st year of available FARS cr ash data, and thus reflected current crash data. Next, I will discuss the process of selecting the dependent, independent, and confounding variables from the 2003 dataset. Process of sample selection Inclusion criteria: Drivers were included when they fell within the two age groups: (a) 35 to 54 years (younger drivers), and (b) driv ers 65 years and older (older drivers). Exclusion criteria: Subjects were excluded when th ere was age-related missing data; were non-drivers of motor vehicles (e.g., passenger s, pedestrians, and drivers of motorcycles); drove heavy trucks (e.g., school buses and motor homes); drove vehicles with unknown body types; or they drove vehicles that do not normally travel on p ublic roads (e.g., snowmobiles and farm equipment). Study design: This study used a cross-sectional design. Sample : The sample size in the study consis ted of 14,038 younger drivers, and 5,744 older drivers ( N =19,782). Process of merging data files: The 2003 FARS sample database s consisted of three main filesâ€”the accident , vehicle , and person files. There were also bl ood alcohol levels files, but these were not included in the analyses. Data f iles were downloaded from the NHTSA website in April 2006 and stored on a secure server at the Co llege of Public Health and Health Professions (University of Florida). The files (initially in SAS format) were merged using three variables: The Vehicle Number , Person Number , and State Number , which, in combination, form unique identifiers. The merged database was convert ed into SPSS 14.0. The data consisted of 178 FARS variables and 102,102 motor vehicle crashes. After the data were saved, a copy of the merged
48 file was created and used in subsequent anal yses. I created a new file excluding cases of nondrivers, cases with unknown ages, cases with crashes involving unknown vehicle types, cases with ages less than 35 years of age and ages between 55 and 64 years of age, and cases involving heavy vehicles. This resulted in a final sample size of 19,782 drivers, comprising of 14,038 younger drivers (35 to 54) years of age, and 5,744 older drivers (65 and over) . Description of Dependent and Independent Variables In this section, I describe th e process of selecting independe nt and dependent variables. For the variables included in biva riate analyses and the binary logistic regression analysis, I describe variables as they were operationaliz ed by NHTSA, and explai n the rationale behind selecting and collapsing leve ls of certain variables. I used the conceptual definitions from the FA RS dataset to operationalize the dependent and independent variables in the study. The con ceptual definitions of all FARS variables are available in the FARS Analytic Reference Guide (1975) and the 2004 FARS Coding and Validation Manual available at www.nhtsa.gov . Independent variables de scribed in this section are limited to the 34 variables used in analys es (1 dependent variab le and 33 independent variables). Dependent variable The dependent variable in this study was injury . In the FARS dataset, the variable injury severity consisted of eight levels of categories: No injury, possible in jury, non-incapacitating evident injury, capacitati ng injury, fatal injury, injury (severity unknown), died prior to crash, and unknown if injured. For the purpose of the analyses, injury was operationalized as whether a driver sustained an injury in th e crash or not (injury: yes/no). Injury Severity was collapsed as follows: (a) Non-incapacitating evident injury, (b) capacitating injury, (c) fatal injury, and (d) injury, severity unknown were pl aced in the injury category injury yes . No injury formed another
49 category. Levels consisting of possible injury, un known if injured, and died prior to crash were treated as missing data, because it was difficult to ascertain whether the group of people in those levels of injury severity were injured or not. Independent variables I used four approaches to sele ct independent variables pertin ent to the research questions. These approaches were: (a) previ ous older driver studies, primar ily, the FARS structural model based upon the SLR (Classen et al, 2006) (A ppendix A), (b) consultation with the Public Health Model to Promote Saf e Elderly Driving team, a NHTSA FARS databa se manager, (c) realism and manageability, and (d) percentage of missing data. These approaches are discussed below. FARS structural model: The structural model (Appe ndix A), developed from the PPMHP was framework used in selecting FARS variables in the study. I selected FARS variables based on whether they fitted within a domain of the PPMHP. I modified the PPMHP structural model by replacing br oad categories from the SLR with specific FARS variables. Consultations: Consultations began prior to analyses . I referred to the FARS manuals for the operationalization of the variables. Appendix C illustrates the initial examination of the 2003 FARS Accident , Vehicle , Person, and Driver sections of the Analyt ical and Validation Manual and the rationale for including and excluding variab les. To establish rigor and reliability in variable selection, there were c onsultations throughout the process. This included consulting with dissertation committee members and the Public Health Model to Promote Safe Elderly Driving Team (including a biostatistician). Cons ultations occurred throughout the six-month process of data examination, variable selection, and collapsing levels variables where necessary. Realism and manageability: Seventy variables were initially selected after first screening of the FARS database. All of the 70 FARS could not be used in the multivariate analysis. It was necessary to use realism to deci de what would be manageable within the scope
50 of the study. One of the methods I used was to examine the operationalized definition of the variables to find out whether some variables had similar conceptual definitions. In such instances, to reduce multicolinearity which could confound the analyses, I selected the variable that best described the factor of interest. An example of such an instance was the environmental variables first harmful event and most harmful event . Both variables captured actions the driver engaged in prior to the crash. However, the first harmful event measured the in itial action, while the most harmful event described the action that may have contributed most to the crash. Univariate analysis showed that frequencies for le vels of each variable were similar, and had the same amount of missing data . Therefore, I selected most harmful event for the multivariate analyses. Percentage of missing data: When conducting the multivariate analyses, I wanted to ensure that the analysis was not compromise d because of many missing data. Early analyses indicated that variables with hi gh percentages of missing data resu lted in a regression model that used as low as 50% of the sample. A way of resolving this was to examine frequencies of independent variables and identify those that had high percentages of missing data. Variables with more than 8% missing data were excluded from subsequent analyses . I chose the cut-off point after examining frequencies and ascer taining the range of missing data among the variables. For example, the variable police reported drug involvement was excluded because it had 69% missing data; that is, e ither it was unknown whether there was an indication of drug use at the time of the crash, or it wa s unknown whether it was reported. Collapsing of data: I used descriptive statistics (e.g., frequencies), to examine data pertaining to all independent vari ables. In instances where a variable had many levels and the frequencies in most levels were fe w, I collapsed the data into fewe r levels to facilitate analysis.
51 For example, most harmful event had 46 levels. This was even tually collapsed into four categories: non-collision, collision with motor vehicle, collision with objects not fixed, and collision with fixed objects. I applied two stra tegies, described below, in the process of collapsing the data. One of the strategies was to use information from the FARS Analytic Reference Guide (1975) and the 2004 FARS Coding and Validation Manual . For variables with several levels, FARS usually coded the variables under broad as well as specific categories. For example, rural vs. urban (FARS code road function class ), had 15 levels. Seven levels of the variable are rural-related (rural principal arterial â€”interstate, rural princi pal arterialâ€”other, rural minor arterial, rural major collector, rural minor collector, rural local road or street, and rural unknown). Seven levels of road function class are urban-related (urban principal arterialâ€” interstate, urban principal arte rialâ€”other, urban minor arterial , urban major collector, urban minor collector, urban local road or st reet, and urban unknown), and one level is unknown . I collapsed this variable, us ing broad categories, into rural vs. urban as the variable had few percentages for every category and I was primarily interested in whether the crash occurred in a rural or an urban area, and not th e type of type of rural or urba n area. I went through this process for all the collapsed variables. The second approach in collapsing the data was to examine variables with many levels and give separate categories to levels that had relatively higher frequencies. Operationalization of independent variables by PPMHP domains In this section, I describe th e 32 independent variables select ed for analyses, within the context of the PPMHP by describing the type of variable (e.g., conti nuous or categorical) and
52 their operational definition in FA RS. I also describe the rationale behind collapsing and recoding the variables used for analyses in this study (Table 3-1). Health domain Age: Age was a continuous variable in the FARS dataset. Age ranged from 0 to 96 years (actual age), and 97 years and older. After the cases that did not meet our inclusion criteria were removed from the sample, the ages of drivers left in the sample were 35 years, and 65+. Thus, age was grouped into two categor ies, the younger and older drivers. Gender: Gender comprised of male and female. Behavioral domain Variables in the behavioral domain were restraint system use , driver drinking , and driver license compliance . Restraint system use: This variable consisted of types of restraints in vehicles and motorcycles for all types of passengers. Restraint system use comprised nine levels: (a) none used/not applicable, (b) shoulder belt (c) lap belt, (d) lap and shoulder belt, (e) child safety seat, (f) bicycle helmet, (g) safety belt used imprope rly, (h) child safety seat used improperly, (i) helmets used improperly, (j) restraint usedâ€”t ype unknown, and (k) unknown. Four levels were eliminated due to the exclusion criteria. As the focus of the domain was on the behavior of the driver rather than the type of restraint system used, the variable was recoded as restraint system used (no/yes). Driver drinking : There were five variables pertaini ng to alcohol in the FARS dataset: (a) police reported alcohol involvement, (b) me thod of alcohol determination by police, (c) alcohol test type/alcohol te st results, (d) driver dri nking and (e) drunk drivers. Driver drinking is a derived variable from the vehicle file. For the purposes of the study, I used driver drinking
53 (no/yes) (option d). Data for driver drinking are obtained from blood al cohol content (BAC) data or police reported alcohol involvement (NHTSA, 2002). Driver license compliance: The FARS dataset contained four variables pertaining to the status of the driver with respect to their driversâ€™ license. The va riables were (a) non-CDL license type/status, (b) commercial motor vehicle license status, (c) license co mpliance with class of vehicle, and (d) compliance with license restrictions. The focus of the study is on whether the driverâ€™s license was valid or not, so the variab le (option c) selected for the analysis was License compliance with class of vehicle . Environmental domain Physical environment: The physical environment comprised of the variables day of week , hour of day , registered vehicle owner , number of lanes , road surface condition, road surface type , road alignment , road profile , rural vs. urban , national highway system , vehicle body type , most harmful event , relation to junction , principal impact , trafficway flow , traffic control device functioning , light condition , weather condition, constr uction/maintenance zone , and airbag deployment . Day of week: This variable was not collapsed, and c ontained cases of crashes for all days of the week (Sunday to Saturday). Hour of day: This was obtained from the variable Accident Time , and comprised two variablesâ€”hour and minute of the crash. The hour (a discrete numerical variable) was used in these analyses. Hour of day ranged from 0 to 24 hours (military time). To ascertain times of the day that may have risk or protecti ve factors for crashes, I collapsed the variable into three levels to enable examination of the influence of various traffic patterns on injury (e.g., peak hours in the morning and evening, and night driving): (a) 9p.m.a.m., (b) 8a.m.p.m., and (c) 2p.m.â€“ 8a.m.
54 Registered vehicle owner: This variable had eight levels: (a) not applicable/vehicle not registered, (b) driver in the cr ash was a registered owner, (c) vehicle was registered as a business/company/government vehicle, (d) vehicle wa s registered as rental vehicle, (e) vehicle was stolen (reported by police), (f) driver less vehicle, and (g) unknown. There were few frequencies (0.8%) for whether the vehicle was a business or rental vehicle. To facilitate further analyses with this variable, it was coded as (a) driver was regist ered owner and (b) driver was not registered vehicle owner. Number of lanes: This was a continuous variable, w ith number of lanes varying from one to seven. Since there were cases with four or more lanes, I collapsed this variable into (a) one lane, (b) two lanes, (c) three la nes and, (d) four to seven lanes. Roadway surface condition: The roadway surface condition was the condition of the road at the time of the crash, which may or may not have contributed to the crash. The levels were as follows: (a) dry, (b) wet, (c) snow or slush, (d) ice, (e) sand, dirt , oil, (f) other and (g) unknown. There were very few cases (4.2%) in the snow or slush, a nd sand, dirt, oil categories. Roadway surface condition was collapsed into two levels: (a) favorable and (b) adverse roadway surface conditions. Roadway surface type : This variable had the following levels (a) concrete, (b) blacktop, (c) brick or block, (d) slag, grav el or stone, (e) dirty, (f) othe r, and (g) unknown. About 85% of the frequencies were in the blacktop category, with very few in the other categories. The variable was collapsed as follows: (a) concrete, (b) blackt op, and (3) other (brick or block, slag, gravel, snow, and dirt).
55 Roadway alignment: This variable comprised three levels: (a) straight, (b) curve, and (c) unknown. The variable was used as it appeared in the database, and the unknown category was treated as missing data in analyses. Roadway profile: This variable had the following levels: (a) level, (b) grade, (c) hillcrest, and (d) sag. About 71% of cases were in the first cate gory of the variable, with very few cases in the other categories, thus data were collapsed into (a ) level and (b) other. Rural vs. urban: Roadway function class had 14 levels , but was broadly classified by NHTSA into rural and urban roadways. The focus of the study was whether the crash occurred in a rural or urban roadway, and thus, the data was collapsed into those categories. National highway system (NHS): This comprises the interstate stem, principal arterial system routes, and some Strategic Highway Network connectors (NHTSA, 2004). The road where the crash occurred was classified as (a) this section IS NOT on th e NHS, (b) this section IS ON the NHS and (c) unknown if th is section is on the NHS. The first two levels were used in the analyses. Vehicle body type: This variable had about 80 levels . It comprised all vehicle types. Heavy vehicles (school buses, mobile homes, heavy trucks) and vehi cles of unknown body types were removed from the databases. The rest of th e data were categorized into three levels: (a) automobile and automobile derivatives, (b) SUVs, and (c) light trucks and pickups. Most harmful event: This is the major event for the vehicle involved in the crash. It consisted of 46 levels, but were grouped under four broad categories used in the analyses in this study: (a) non-collision (e.g., overturn, fire/exp losion), (b) collision with motor vehicle (e.g., motor vehicle in transport on the same roadwa y), (c) collision with object not fixed (e.g., pedestrian or animal), and (d) collision w ith fixed object (e.g., boulde r or utility pole).
56 Relation to junction: Relation to junction is the locati on of the first harmful event that contributed to the crash. The va riable has 18 levels and was gr ouped under three categories, (a) non-junction, (b) non-interchange and (c) interchange area. Principal impact: This is the impact point of the crash. The FARS dataset has both the initial point and the principal point. Both vari ables use clock points to indicate the position of injury 0 for non-collision, (1), and 13 (top) and 14 (undercarriag e). The initial point is the first impact point that produced property damage or personal injury, while the principal point is the impact point that produced the most pr operty damage or persona l injury (NHTSA, 2004). The descriptive analyses (frequenc ies) indicated that both the init ial and the principal point of impact had similar distributions within the various impact levels. To reduce multicolinearity, only one variable, the principal impact was selected in the analyses. Since some of the levels had few frequencies, the variable was collapsed in to the following levels: (a) 1 oâ€™clock, (b) 4 oâ€™clock, (c) 7 oâ€™clock, (d) 10 oâ€™clock, (e) 12 oâ€™clock, and (f) top and undercarriage. Trafficway flow: This variable was grouped as follo ws: (a) not physically divided (twoway traffic-way), (b) not physically divided (with two-way continuous left-turn lane), (c) divided highway, median strip (without tr affic barrier), (d) divided highw ay, median strip (with traffic barrier), (e) one-way traffic-way, and (f) entrance/exit ramp. For the purposes of this research, the variable was collapsed into the following levels: (a) divided highway, (b) not divided highway, (c) one-way traffic-way, (d) not physic ally divided, and (e) entrance/exit ramp. Traffic control device functioning: There are two variables pertaining to traffic control devices in the FARS datasetâ€”traffic control devi ce, and traffic control device functioning. There were 37 traffic control devices grouped under hi ghway traffic signals, re gulatory signs, school zone signs, warning signs, misce llaneous not at railroad crossi ng, at railway grade crossing, and
57 whether or not the device was at a railroad grad e crossing. The focus of the variable in the study was whether the traffic control device was present or not, rather than the type of traffic control device, so for the purpose of the study, the variable traffic control device functioning was used in the analyses. This variable had four levels: (a ) no controls, (b) device no t functioning, (c) device functioning improperly, and (d) device functioning pr operly. The variable wa s collapsed into two categories: (a) no controls, and (b) device functioning properly. Device not functioning and device functioning improperly were removed from subsequent analyses. Light conditions: This was coded as (a) daylight, (b) dark, (c) dark but lighted, (d) dawn, and (e) dusk. The variable was collapsed into three categories: (a) light, (b) dark, and (c) other (dark but lighted, dawn and dusk). Weather conditions: This variable contained the following levels: (a) no adverse atmospheric conditions, (b) rain, (c ) sleet (hail), (d) s now, (e) fog, (f) rain and fog, (g) sleet and fog, and (h) other. This was collapsed into tw o categories: (a) adverse and (b) favorable (no adverse) conditions. Construction/maintenance zone: This pertained to whether the crash occurred in a maintenance or utility zone. The variable was collapsed into two categories: (a) none, and (b) construction, maintenanc e, or utility zone. Airbag deployment: The variable airbag deployment is used to indicate whether the airbag was available for the person and whether it deployed or not, or was turned off. The variable had 11 levels and was recoded as (a ) did not deploy or no ai rbag and (b) deployed. Social environment : This comprised a discrete numeric variable number of occupants . Based on the older driver literature, and freque ncy distribution it was re coded into the following levels: (a) driver only, (b) one passe nger, and (c) two or more passengers.
58 Predisposing domain Vehicle maneuver: This variable captures the driverâ€™s action, or intended action, before commencement of the crash situation (a reflectio n of driver skill), and had 17 levels. These levels were collapsed into the following categor ies: (a) going straight, (b) lane-related, (c) maneuvers (making a right, making a U-turn, pa rking or leaving a parked position, making a controlled maneuver to avoid an object, or back ing up) (d) making a left, and (e) negotiating a curve or changing lanes/merging. Reinforcing domain This domain comprised five discrete numeri cal variables pertaining to the driversâ€™ conviction (ticketed offenses) history within th ree years prior to the crash: (a) number of previous motor vehicle accident convictions, (b) number of pr evious motor vehicle speeding convictions, (c) number of previous motor ve hicle suspension convictions, (d) number of previous DWI (drug and alcohol impairment) co nvictions, and (e) number of other previous motor vehicle convictions (failure to yield, running a stop sign or red light, and lane-related). The reinforcing variables were not collapsed in the analyses. Enabling domain Licensing state: The licensing state was the only vari able applicable to the enabling domain of the PPMHP. This variable was not used in the multivariate analyses. Instead, a subanalysis was performed for older drivers only b ecause this study explor ed the relationship among age-related license renewal policies and injury ou tcomes. States were classified as having agerelated license renewal policie s or not having age-related li censure renewal policies, and bivariate analyses performed to ascertain the relationship with injury, for four age-related renewal policy fact ors I derived from license state . Table 3.2 summarizes states with and without age-related license renewal policies.
59 Data Collapsing and Data Cleaning The next step of the analysis was to ex amine the variables in light of the PPMHP structural model created from the SLR, and th e domains of the PPMHP. A preliminary selection of the variables was done. A codebook was created, using the 1975 FARS Analytical Reference Guide and the 2004 FARS Coding and Validation Manual to ascertain the definitions of each variable. The majority of the variables were categorical or ordinal in nature, with levels of variables varying between 2 and 50. For the purposes of conducting bi variate and regression analyses, it was important to co llapse some of the FARS variab les. The codebook for the initial examination of the FARS variable anal yses is presented in Appendix C. After the initial merging of the FARS variables, we had an initial number of 178 variables. The next step in anal ysis involved attaching labels to the variables in SPSS to ensure accuracy of data analysis. Subsequently, in this stage I performed univa riate statistics of the variables to examine levels of variables for missi ng data. This enabled me to eliminate variables that were applicable to the research questi on but contained too many missing data points. For example, two environmental variab les that contained information on the longitude and latitude of where the crash occurred had 100% missing data. Af ter all the variables ha d been collapsed into smaller levels where applicable, univariate anal ysis was conducted once more. The initial modus operandi was to conduct a bivariate an alysis, select variables that we re statistically significant at the p < 0.05 level, and use those variables for subse quent analyses (particularly, the binary logistic regression). However, due to the large sa mple size, almost all the variables showed up as significant at that level of analysis. The bivariate analysis was used to examine frequencies in the cells, and further collapse data where necessary. For the multivariate analyses, the frequencies of the variables were examined and those with less than 8% missing data were selected for further analyses. This strategy appeared to be a
60 more effective one in the analysis. Thirty-two va riables were selected fo r bivariate and logistic regression analyses. Statistical Analyses Statistical analyses were performed us ing SPSS 14.0 (SPSS, 2005). Statistical analyses consisted of univariate analysis, bivariate analysis, bi nary logistic regression, and a descriptive summary of age-related state license renewal policies for older drivers. Univariate analyses The univariate analysis served three pur poses: (a) to provide information on the distribution of the data, making it easier to see wh ich variables needed to be collapsed, (b) to provide a frequency distribution for the data for the descriptiv e portion of the study, and (c) to provide a means of variable selection for the bi nary logistic regressi on analysis. I performed univariate analyses for the younger group, the older group, and the overall sample. Bivariate analyses I performed bivariate analyses to examin e the relationships between independent variables and injury outcome. The initial objective was to use this level of analysis to select variables for the logistic regressi on. However, this was not used si nce all the variables turned out to be significant at that level of analysis. I performed cross tabulat ions to ascertain the prevalence rates of injury/no injury among the population as well as for the younger and older groups. For numerical variables such as motor vehicle convi ctions, independent t-te sts were performed to ascertain differences in the means between each re inforcing variable and injury outcome for the younger and older groups. Prevalence is defined as â€œthe number of events , e.g., instances of a given disease or other conditions in a given population at a designated timeâ€ (Last, 2001, p.140). The prevalence rate is a measure of the number of people who have a he alth condition in a popul ation at a given time
61 (Mausner & Kramer, 1985, p.44). In the FARS data base, all drivers were involved in a motor vehicle crash and it was not possible to measur e prevalence as defined, as all drivers were exposed. To calculate injury prevalence rates among the independent variables, I used the crosstabulation function in SPSS 14.0 and sele cted the observed frequency option (with percentages for rows, columns and total). I calcul ated the proportion of dr ivers who were injured within each level of a va riable. For example, for gender (younger drivers), 68.8% were male and 31.2% were female. Within the levels of gender the proportion of males injured was 73.0% were male, and 76.8% female, thus, the injury prevalen ce for males in the sample is 73.0%. I also selected the chi-square test opt ion to determine whether there was any statistically significant relationships among independent vari ables and injury (no/yes) at p < 0.05 level. I used a series of bivariate analyses to answer the research question on age-related licensure renewal policies for olde r drivers. The U.S. states we re grouped by factors: (a) state age-related licensure renewal polic y (no/yes) (Table 3-2), (b) re duced renewal cycles for older adults (no/yes) (Table 3-3), (d) in-person re newal (no/yes) (Table 3-4), and (b) vision and medical tests required (no/yes) (Table 3-5). The four categories of age-related license renewal policies are not mutually exclusive categories. I performed a series of chi-square tests to ascertain whether there were si gnificant relationship among the f our age-related renewal policy factors and injury (no/yes). To group the license state variable into four derived variables, I used information on age-related license renewal pol icies in 2003. Information on state age-related license renewal polices was obtained from Physicianâ€™s Guide to Asse ssing and Counseling Older Drivers (Wang, Kosinski, Schwartzberg, & Shankli n, 2003). Thus, Florida, which has a vision test for license renewal for dr ivers over 79 years old, was cla ssified as a state with no agerenewal policy for older driv ers because the effective date for the law was January 2004.
62 Multivariate analyses In selecting the independent variables, I wanted to choose variables th at would result in the best model to answer the research questi ons. Hosmer and Lemeshow (1989) offer three approaches in modeling variables in a regression. One approach is to enter all variables in the logistic regression model regardle ss of whether variables contribute significantly to the model. The downside of this approach is that it may result in overestimated standard errors and estimated coefficients. The second approach is to aim for the most parsimonious model that would still explain the data. Hosmer and Lemeshow (1989) suggest that vari able selection for the logistic regression start with a careful univariate analysis of each independent variable, and a crosstabulation with the dependent variable to ascertain zero data in cells of the table. The third approach for variable selection is to do a selection based on statisti cal criteria, such as a stepwise method. The disadvantage of using statistical met hods to select variable s is that irrelevant variables may be selected by the computer softwa re (p.87). I used the se cond approach in this study. Variables with less than 8% missing data were selected for the logis tic regression analysis. To test the interaction between independent, co nfounding variables and inju ry outcome, a binary regression analysis was performe d. The binary regression model is a method used to ascertain relationships between independent variables and an outcome (d ependent) variable with two levels. The binary logistic regression enables the researcher to find out the estimates for each exploratory variable to the outco me variable. In a binary logis tic regression, the probability of the occurrence of the dependent va riable (varying between 0 and 1) is transformed into odds that express the likelihood of an occurrence relative to the likelihood of a non occurrence (Cox, 1989; Hosmer & Lemeshow, 1989) Unlike a linear regressi on that examines the re lationship of how the dependent variable changes when the independent variable increases by one unit, the logistic
63 regression model examines how th e natural log of the odds that the dependent variable Y = 1 ( ), varies as a function of the linear predictor (J accard, 2001). The relationship between the log odds of the dependent variable (logit) and independent variables is illustrated by the equation: logit ( ) = + 1X1+ 2X2++ kXk, where is the intercept, the regression coefficient, and X the predictor (independe nt) variables (Jaccard, 2001). Apart from the 32 independent variables that were used in the regression model, age interactions were also done for all the variables. Jaccard (2001) defines an interaction effect as when the effect of an independent variable on a dependent variable differs depending on the value of a moderating variable ( p. 12). The interaction model for two predictor variables can be presented as: logit ( ) = + 1X + 2Z+ 3XZ, where X is an independent variable and Z a moderating variable. The statistical outputs obtained from the regres sion model include (a) a classification table, (b) estimated odds ratios, and (c) Hosm er-Lemeshow goodness-of-fit statistic and R2. Classification table: This is used to evaluate the accuracy of the predictive property of the model (Pampel, 2000). Estimated odds ratios: The odds ratios enable one compare two different odds, and estimate the likelihood of getting injured in a cr ash and, thus, enabled us to ascertain which variables were determinants (risk an d protective factors) of injury. Hosmer-Lemeshow goodne ss-of-fit statistic: The Hosmer-Lomershaw estimate enabled me examine the model fit (how well the model fits the data). Pseudo R-square: The Cox and Snell R-square, and the Nagesqure R-square estimates. The Cox and Snell R-square is a measure of the pseudo-variance explained by the binary logistic
64 regression model. The Nagesqure Rsquare is an adjustment of the Cox and Snell estimate, and was reported in the multivariate analyses (Pam pel, 2000). I reported the Nagesqure R-square estimates in this study. Level of significance: For the analyses that examined the relationship among independent variables and injury, and for the binary logistic regression. I chose the leve l of significance to be p < 0.05. Because the study was exploratory in nature , I did not use a Bonferroni correction to account for the large sample size. Most of bivariat e results for injury prev alence had statistical significance at p <0.001, and a Bonferroni corr ection would not have made a difference in the statistically significance in the bivariate results. In the third research question that examined relationships among age-related license renewa l policies and injury outcomes, I applied a Bonferroni correction in the anal yses, because the categories used to group the states were not mutually exclusive. The level of significance was p < 0.0125. In a large sample size, it is important to di fferentiate between stat istical significance and practical significance because the power of the study is depe ndent on the sample sizes. According to Stevens (2002), sample sizes greate r than 200 are likely to have many variables that are significant at p < 0.05. Ways of determining practical or statistical significance include (a) putting the findings in context by examining prev ious research, (b) Using Cohnâ€™s definition of small, medium and large effect sizes, (c) having normative definitions of clinical significance, (d) performing a cost-benefit analysis, and (e) usi ng a null hypothesis that is harder to reject, such as testing whether differences between tw o population is at least 3 standard deviations (Haase, Ellis, & Ladany,1989; Stevens, 2002, p.12). In these analyses, I defined clinical significance mainly by using option (a) that is, pl acing the findings in the context of obtaining historical plausibility through previous research.
65 Limitations of FARS Using FARS encompasses the disadvantages of a cross-sectional study. As data on the explanatory and outcome variables are collected at the same point in time, it is impossible to measure cause and effect as one cannot establish the temporal sequence of events (Mausner & Kramer, 1985). Secondly, because of missing data, some PPMHP variables could not be included in the multivariate analysis. For example, health variables such as drug test results (a variable that contained informa tion on different classes of medica tions) and driver-related factors such as some physical and mental health conditions were not us ed in subsequent analyses. Specifically, FARS has the following limitations: (a) results will not be representative of drivers; (b) all subjects were involved in a fatal crash; thus, comparisons cannot be made with drivers who were crash free; (c) FARS has inad equate information on the social assessment domain of the PPMHP; (d) FARS lacks some so cioeconomic variables (e.g., income). Race, a variable present in the 2003 FA RS data, had information on only the fatally injured, and could not be used in the multivariate analyses in this study. Thus, within the scope of this study, the contribution of some socio-demographic charact eristics to injury outcome could not be ascertained; and (e) some of the FARS variab les had large percenta ges of missing data Strengths of FARS Despite the limitations of the FARS dataset, it meets the criteria as an appropriate database to answer the research questions of this study. The dataset comprise s a variety of socio-ecological variables representing all the major domains of the PPMHP, except the health promotion domain, that enable researchers to examin e crash dynamics and how health, behavior, environment, predisposing, reinforcing, and enab ling variables are associated with injury outcomes in crashes.
66 Table 3-1. Description of independent FARS variables, variable types and levels Precede-Proceed Domain Variable* Number of level/type Description of levels Health Age* (2) nominal 1 = 35 years 2 = 65+ years Gender (2) nominal 1 = male 2 = female Behavior System restrain use* (2) nominal 0 = no 1 = yes Driver drinking (2) nominal 0 = no 1 = yes Driver license compliance* (2) nominal 0 = not valid 1 = valid Environment Day of the week (7) nominal 1 = Sunday 2 = Monday 3 = Wednesday 4 = Thursday 5 = Friday 6 = Saturday 7 = Saturday Hour of the day* (3) nominal 1 = 9p.m.a.m. 2 = 8a.m.p.m. 3 = 2p.m.a.m. Registered vehicle owner* (2) nominal 1 = no 2 = yes Vehicle body type* (3) nominal 1 = Automobiles 2 = SUVs 3 = Vans, light trucks and pickups Number of lanes* (4) nominal 1 = 1 lane 2 = 2 lanes 3 = 3 lanes 4 = 4 to 7 lanes Roadway surface* conditions (2) nominal 1 = favorable 2 = adverse Roadway surface type* (3) nominal 1 = concrete 2 = blacktop 3 = other Roadway alignment (2) nominal 1 = straight 2 = curve Roadway profile* (2) nominal 1 = level 2 = other Road function class (rural (2) nominal 1 = rural
67 Table3 1.Continued. vs. urban)* 2 = urban National Highway System (2) nominal 0 = no 1 = yes Most harmful event (type of collision)* (4) nominal 1 = collision with object not fixed 2 = collision with a moving vehicle 3 = non-collision 4 = collision with a fixed object Relation to junction* (highway design feature) (3) nominal 1 = non-junction 2 = intersection 3 = interchange Principal point of impact (point of impact using the clock method)* (6) nominal 1 = 1 oâ€™clock 2 = 4 oâ€™clock 3 = 7 oâ€™clock 4 = 10 oâ€™clock 5 = 12 oâ€™clock 6 = top and undercarriage Trafficway flow* (5) nominal 1 = divided highway 2 = not divided highway 3 = one way trafficway 4 = not physically divided 5 = entrance/exit ramp Traffic control device functioning* (2) nominal 0 = not present 1 = functioning properly Light conditions* (3) nominal 1 = daylight 2 = dark 3 = other Weather conditions* (2) nominal 1 = favorable 2 = adverse Construction/maintenance zone* (2) nominal 0 = no 1 = yes Airbag deployment* (2) nominal 0 = no 1 = yes Number of occupants (passengers)* (3) nominal 1 = driver only 2 = 1 passenger 3 = 2 or more passengers
68 Table3 1.Continued. Predisposing Vehicle maneuver (driver skill)* (5) nominal 1 = going straight 2 = lane-related changes 3 = other maneuvers 4 = making a left 5 = negotiating a curve Reinforcing Number of previous motor vehicle accident convictions in last 3 years (ticketed) Discrete numerical Number of previous motor vehicle speeding convictions in last 3 years (ticketed) Discrete numerical Number of previous motor vehicle suspension convictions in last 3 years (ticketed) Discrete numerical Number of previous Driving while impaired (DWI) convictions (ticketed) Discrete numerical Number of other previous motor vehicle convictions (failure to yield, running a red light, lane-related errors) in last 3 years (ticketed) Discrete numerical *levels of variable were collapsed Table 3-2. Age-related renewal policies in 2003 States with age-related renewal policies States without age-related renewal policies Alaska Arizona California Colorado Connecticut District of Columbia Hawaii Idaho Illinois Indiana Alabama Arkansas Delaware Florida Georgia Kentucky Massachusetts Michigan Minnesota Mississippi
69 Table3 2.Continued. States with age-related renewal policies States without age-related renewal policies Iowa Kansas Louisiana Maine Maryland Missouri Montana Nevada New Hampshire New Mexico North Carolina Oregon Pennsylvania Rhode Island Utah Nebraska New Jersey New York North Dakota Ohio Oklahoma South Carolina South Dakota Tennessee Texas Vermont Virginia Washington West Virginia Wisconsin Wyoming Table 3-3. Reduced renewal cycle requirements in 2003 States with reduced renewal cycles States without reduced renewal cycles Arizona Colorado Connecticut Hawaii Illinois Indiana Iowa Kansas Maine Missouri Montana New Mexico Pennsylvania Rhode Island Alaska Alabama Arkansas California Delaware District of Columbia Florida Georgia Idaho Kentucky Louisiana Maryland Massachusetts Michigan Minnesota Mississippi Nebraska New Jersey Nevada New Hampshire New York North Carolina North Dakota Ohio Oklahoma Oregon South Carolina South Dakota Tennessee Texas Utah Vermont Virginia Washington West Virginia Wisconsin Wyoming
70 Table 3-4. In-person renewal requirements in 2003 States with in-person renewal requirement States without in-person renewal requirement Alaska Arizona California Colorado Idaho Illinois Indiana Louisiana Alabama Arkansas Delaware Connecticut District of Columbia Florida Georgia Hawaii Iowa Kansas Kentucky Maine Michigan Minnesota Mississippi Missouri Montana Nebraska New Jersey New Mexico Pennsylvania Rhode Island Maryland Massachusetts Oklahoma Oregon New Hampshire New York North Carolina North Dakota Ohio South Carolina South Dakota Tennessee Texas Utah Vermont Virginia Wisconsin Washington West Virginia Wyoming Table 3-5. Vision, medical, or road test requirements in 2003 States with vision, medical or road test requirement States without vision, medical or road test requirement District of Columbia Illinois Maryland Nevada Oregon Pennsylvania Utah New Hampshire Alaska Alabama Arizona Arkansas California Colorado Connecticut Delaware Florida Georgia Hawaii Idaho Indiana Iowa Kansas Kentucky Louisiana Maine Massachusetts Michigan Minnesota Mississippi Missouri Montana Nebraska New Jersey New York North Carolina North Dakota New Mexico Ohio Oklahoma Rhode Island South Carolina South Dakota Tennessee Texas Vermont Virginia Washington West Virginia Oklahoma Wisconsin Wyoming
71 CHAPTER 4 RESULTS In this chapter I describe the younger and ol der driver population w ith respect to the dependent variable (injury: no/yes), independent variables from the health, behavior, environment, predisposing, reinforcing, and enab ling domains of the Prece de-Proceed model of health promotion (PPMHP), and confounding variable (age). The following results are presented: (a) univariate results for younger and older drivers, (b) bivariate results indicating percentages and prevalence rates for injury among younger and older drivers, (c) bivariate results for the exploratory research on the rela tionship between age-related license renewal polices (older drivers only), and (d) results from the binary logistic regression. Description of the Sample In this section I present results of the univari ate analyses for independe nt variables used in the binary regression model. I discuss the univari ate results in relation to the percentage of younger drivers, older drivers, and the overall sample injured in a motor vehicle crash. Univariate Analyses The distribution of crashes for the overal l sample and by age groups (Table 4-1), contained 32 variables organized by the main domains (health, behavior, environment, predisposing, and reinforcing) of the PPMHP. I conducted univariate anal yses with frequencies for categorical variables (27 variables), and means and standard deviations for discrete numerical variables (5 variables). Th e sample population consisted of 71% younger drivers (14,038) and 29% older drivers (5,744). For all drivers, 68.2% (13,492) were male and 31.8% (6,290) were female. Among the younger drivers, there we re 68.8% (9,665) males, and 31.2% (4,373) females, while the gender distribution for olde r drivers was comprised of 66.6% (3,827) male and 33.4% (1,917) female (Table 4-1). Regarding injury severity, 77.4% (15,315) of all drivers
72 experienced some form of injury, 74.1% (10,049) of younger drivers were injured in the crash, and 85.4% (4,906) of older drivers were injured in a crash (Figure 4-1). In the behavior domain, compared to younger drivers, older drivers had lower percentages of non-compliance (hav ing an invalid driversâ€™ licens e for the vehicle they were driving at the time of the crash, indicating alcoho l use, and not wearing a restraint system use). Approximately 19% of younger and older drivers, 18.8% (3,725) of drivers indicated having drunk alcohol at the time of the crash. However, a smaller percentage of older drivers had indication of drinking alcohol while driving (5.1%), compared to younger drivers (24.2%). In general, 66.8% of all drivers wore some form of restraint at the time of the crash. The majority of younger drivers 64.5% (8,295) and 7 2.5% (3,863) of older drivers were wearing a form of restraint during the crash. For environmental variables, observable diffe rences in frequency percentages between younger and older drivers were in the hour of day , light conditions, relation to junction , traffic control device functioni ng, roadway alignment , vehicle body type , registered vehicle owner , principal impact , and most harmful event. For the predisposing domain, vehicle maneuver , observable differences among levels of the variable for the two age groups are that 72% of younger drivers and 64% older drivers had were traveling straight prior to the crash. Also, there was a highe r percentage of older drivers who were executing a left turn (16.8%) comp ared to younger drivers (4.8%). Regarding reinforcing variables, the mean number of convictions for younger and older drivers is presented in Table 4-1. State age-related renewal procedures: Age-related renewal policies in the states were not mutually exclusive (Table 4-2) . About half of the states had so me form of age-related license
73 renewal policy, and 37.9% of older drivers had driv ersâ€™ licenses from states with one or more age-related renewal policies. Most states with renewal policies used the reduced renewal cycle for older drivers alone, or in conjunction with other age-related license renewal policies. Summary of univariate analysis: Using the confounding variable age, with frequencies for nominal data and means and standard deviat ion for numeric data, I showed the frequency distribution among levels of inde pendent variables from the h ealth, behavior, environment, predisposing, and reinforcing domains of the PPMHP. Bivariate Analyses This section consists of bivariate resu lts for PPMHP variables by domain (health, behavior, environment, and predis posing) and injury, and results fo r the chi-square analyses for the four age-related license renewa l policies (enabling variables). Bivariate analyses by age group: Table 4-3 summarizes the results of the bivariate analysis by age to examine relationships among e xploratory variables in the health, behavioral, environmental, and predisposing domains, and the dependent variable (inj ury: no/yes) Results presented for selected variables were used in s ubsequent analyses (logis tic regression), with chisquare analyses for categorical data and inde pendent sample t-test for numerical data. Health domain For younger and older drivers, there were sta tistically significant di fferences in injury prevalence rates among males and females, with a higher proportion of females injured in motor vehicle crashes as shown in Figure 4-2. Behavior domain There were statistically significant differences in the prevalence of drivers injured with respect to the driver drinking variable for younger drivers 2 (1, N = 14,038) = 598.08, p < 0.01,
74 and older drivers 2 (1, N = 5,744) = 11.25, p < 0.01. For both age groups, a higher proportion of drivers who reflected alcohol consumpti on prior to the crash were injured. For driver license compliance (whether the license the driv er was using was valid for the type of vehicle driven at the time of the crash) , in the chi-square test , I observed statistically significant differences in inju ry rates among younger drivers 2 (1, N = 5,562) = 1.04, p < 0.01, but not a statistically significant differ ence in injury rates among older drivers 2 (1, N = 13,921) = 84.44, p < 0.01. For younger drivers, a smaller propor tion of drivers in compliance were injured in the crash compared to those who were not in compliance (Figure 4-3). A statistically significant difference exists among drivers injured who wore a form of restraint (seatbelt, lap belt, or both) during the crash and those who did not. Figure 4-4 presents injury rates for restraint system use, and inju ry for younger and older driv ers. A larger proportion of drivers without a form of restra int were injured, compared to t hose wearing a form of restraint. This finding was statistically signifi cant for both younger and older drivers. Environmental domain Day of week to injury outcome was not statis tically significant for younger drivers 2 (6, N = 14,031) = 10.26, p = 0.11. However, among older drivers, th ere were statistically significant differences in injury outcomes for the day of week ; 2 (6, N = 5,743) = 15.05, p = 0.02. In comparison with other days of the week, fewe r injuries occurred on Sundays. The variables weather conditions , light conditions , and National Highway System indicated statistically significant differences in injury percentages for younger and older drivers. Crashes during favorable weather conditions had lower injury pe rcentages compared to crashes during adverse weather conditions (rain, snow, sleet). For daylight conditions, the gr eatest proportion of younger driv ers injured occurred during dark hours, while for older driv ers, the greatest proportion of injuries occurred in daylight
75 conditions. Drivers not traveling on the National Highway System had lower injury percentages compared to drivers who were on the National Highw ay System at the time of the crash. For both age groups, there were statis tically significant differences among injury rates for hour of day , younger driver, 2 (2, N = 13,936) = 98.21, p < 0.01; older driver, 2 (2, N = 5,722) = 48.73, p < 0.01. A greater proportion of younger and older driv ers were injured in crashes during the 9p.m.â€“ 7a.m.hours (Figure 4-5). For both age groups, drivers traveling on ru ral roads at the tim e of the crash had significantly higher percentages of injuries compared to drivers on urban roads. For number of lanes , there were sta tistically significant differences among younger and older drivers for injury prevalence. Crashes occurring in two-lane roads had the highest percentages of injury for both age groups. Trafficway flow (how the highway was divided) and relation to junction were environmental variables with statistically significa nt differences for injury prevalence in both age groups. However, construction/maintenance zone did not indicate sta tistically significant differences among drivers who crashed in construc tion, maintenance and utility zones, and those who did not. The finding was similar for younger drivers 2 (1, N = 14,038) = 0.31, p < 0.31, and older drivers 2 (1, N = 5,744) = 2.29, p <0.84. Regarding traffic control device functioning, younger and older drivers had statistically significant differences in injury prevalence rate s. Younger drivers had a higher percentage of injury (75.8% injury when traffic control de vices were absent; 68.8% traffic control device present) 2 (1, N = 13,963) = 65.10, p <0.01. However , older drivers had a higher injury percentage when a traffic control was present co mpared to when there was no traffic controls present (84.2% injury when traffic control de vices were absent; 87.2% when a traffic control device was present; 2 (1, N = 5,701) = 12.11, p <0.01. Four of the stat istically significant
76 environmental variables were road-related: (a) roadway surface condition (favorable vs. adverse), (b) roadway surface type (concrete, blacktop, or other), (c) roadway profile (level vs. other), and (d) roadway alignment (straight vs. curve). For both younger and older drivers, the favor able road condition wa s associated with lower percentages of injury. For example, with roadway surface condition , favorable road conditions had a lower proportion of injury cases compared to adverse roadway conditions (e.g., sleet, snow, wet). For vehicle body type , crashes in automobile and automobile derivatives had higher injury percentages for both younge r and older drivers, followed by s ports utility vehicles (SUVs) , and then vans, light trucks and pick-ups; younger driver; 2 (2, N = 14,038) = 134.43, p < 0.01, and older drivers 2 (2, N = 5,744) = 47.75, p < 0.01 (Figure 4-6). The variable a irbag deployment had statistically significant differences in injury rates for younger and older drivers. Within both age gr oups, airbag deployment had statistically significantly higher injury cases than when there was no deployment of the airbag at the time of the crash. Registered vehicle owner had statistically significant differences in injury outcomes for younger and older drivers. Not being the registered vehicle owner had lower injury percentages in comparison to drivers who were th e registered owner of the vehicle driven when the crash occurred. For younger drivers, the number of occupants in the vehicle there was no statistically significant differences in injury percentages for drivers only, drivers with one passenger present, and drivers with two or more passengers present 2 (2, N = 14,038) = 2.93, p < 0.12. However, for older drivers, there was a statistically significant difference in percentages for injury outcomes for number of occupants 2 (2, N = 5,744) = 5.10, p < 0.04, with two or more passengers having lower injury percentage.
77 Principal impact , a variable that uses the clock to indicate the position on the car struck in a crash (e.g., 12 oâ€™clock being head-on collision), indicated stat istically significant differences in injury cases for both older a nd younger drivers, with highest pe rcentages of injury occurring for both age groups in the 7 oâ€™clock area. Most harmful event (type of collision just before the crash occurred) indicated statisti cally significant differences in injury percenta ges among all the four levels. Statistically significan t differences were observed in the collision with object not fixed category, which had very low frequencies of injured drivers for both younger and older drivers (10% younger drivers; 17.0% older driver s), as shown in Figure 4-7. For younger and older drivers, collision with fixed objects (e.g., building or bridge rail ) and non-collision (e.g., overturn or driver injured in vehicle) had the highest injury frequencies. Predisposing domain Vehicle maneuver (capturing the driverâ€™s action or intended action prior to the event leading to the crash) indicated statistically significant differen ces in injury outcomes for younger drivers 2 (4, N = 13,871) = 524.90, p < 0.01; and older drivers 2 (2, N = 5,695) = 48.50, p <0.01. Majority of injuries occurred in both age groups when the dr iver was negotiating a curve or changing lanes (90.8% for younger drivers and 94.1% for older drivers). Older drivers also had high percentages of injuries when making a left turn (94.0%), and going straight (84.5%), while younger drivers had high rates of injury while go ing straight (73.6%), but had lower percentages for lane-related injuries (58.1%), maneuve rs (58.6%), and making a left (58.8%). Reinforcing domain I performed an independent sample t-test to ascertain whether there were differences for older and younger adults regard ing injury outcomes. I also plotted the mean number of convictions by injury (yes) to asce rtain if there were indications of age-interaction effects for the reinforcing variables.
78 Regarding number of previous accident convictions (number of accident convictions within the last three years prio r to the crash) younger drivers ha d a mean of 0.16 convictions for those uninjured in the crash ( SD = 0.43), and 0.16 ( SD = 0.47) convictions for those who were injured. Older drivers had a mean of 0.13 ( SD = 0.38) convictions for th e uninjured drivers and 0.14 ( SD = 0.41) for the injured drivers. There were no statistically signif icant differences in means between the number of previous accident convictions and injury outcome for younger, t (1, 3123) = -0.24, p = 0.98 (2-tailed), and older drivers t (5,433) = -0.21, p = 0.84 (2-tailed). There was no indication of an age interac tion among younger and older drivers and injury outcome for this vari able (Figure 4-8). For the number of previous DWI convictions (driving while impaired) within the last three years prior to the crash, the mean number of previous DWI convictions were 0.02 ( SD = 0.16) for uninjured younger drivers, 0.05 ( SD = 0.24), for injured younger drivers 0.00 ( SD = 0.08) for uninjured older drivers and 0.01 ( SD = 0.00) for injured older driv ers (Figure 4-9). There was a statistically significant differe nce between uninjured younger driv ers and injured younger drivers t (1, 3869) = -6.08, p < 0.01 (2-tailed), but no statistically significant difference between older adults and injury outcome t (5,695) = -0.95, p = 0.34 (2-tailed). There was indication of age interaction among younger and ol der adults (Figure 4-9). Regarding number of previous suspension convictions , there was no statistically significant differences in inju ry outcome for older drivers t (5,695) = 0.14, p = 0.89 (2-tailed), however, there were statistically significant differences in in jury outcomes for younger drivers t (13,869) = -7.209, p < 0.01 (2-tailed). The mean number of convictions for the uninjured younger group was 0.16 ( SD = 0.76) and 0.26 ( SD = 0.92) for the injured younger group, and
79 0.02 ( SD = 0.18) for uninjured older drivers and 0.02 ( SD =0.18) for injured older drivers. There was indication of an age in teraction among younger and ol der drivers (Figure 4-10). Concerning number of previous speeding convictions , the mean number of speeding convictions for uninjured younger driv ers was 0.22 (SD = 0.56), and 0.25 ( SD = 0.68). For uninjured older drivers, the mean number of previous speeding convictions was 0.08 ( SD = 0.31), and 0.06 ( SD = 0.28) for injured older drivers. Olde r drivers had statistically significant differences in injury outcome t (5,695) = 2.0, p = 0.05 (2-tailed), while younger drivers did not show statistically significant differences t (13,869) = -1.40, p = 0.16 There was indication of age interaction between the tw o groups (Figure 4-11). Number of previous other motor vehicle convictions (failure to yield, running a red light, lane-related convictions) and injury outcomes were not statistically significant for younger drivers t (13,869) = -0.36, p = 0.72 (2-tailed) but statistically significant for older drivers t (5,695) = -1.82, p = 0.07 (2-tailed) because the null hypothesis is directional. The mean for uninjured younger drivers was 0.18 ( SD = 0.53), and the mean number of convictions for injured younger drivers was 0.20 ( SD = 0.62). For older drivers, the mean number of previous other motor vehicle convictions was 0.11 ( SD = 0.41) was non-injured drivers, and 0.08 ( SD = 0.33) for injured drivers. There was indication of an age interaction effect (Figure 4-12). The next section of the bivariate analyses shows the injury prevalence rates for younger and older drivers with respect to the enabling variables in the study. Enabling domain Age-related license renewal polices comprised of (a) age-related license renewal policy (no/yes), (b) reduced renewal cycl es (no/yes), (c) in-person rene wal required (no/yes), and (d) vision, medical, or road test requ ired (no/yes). Bivariate result s are presented in Table 4-4.
80 These results suggested lower injury preval ence among older drivers with licenses from states with some form of license renewal policy compared to drivers from states without some form of age-related license rene wal policy (Table 4-4). Bivariate analyses for specific age-related license renewal policies indicated that these di fferences were statistic ally significant for only age-related renewal policies 2 (1, N = 5,725) = 8.67, p <0.01 and states with in-person renewal requirements for older adults 2 (1, N = 5,681) = 21.63, p <0.01. Drivers with licenses issued by states with reduced renewa l cycles for older adults 2 (1, N = 5,725) = 0.30, p = 0.59, and states with vision, medical, or road test requirements 2 (1, N = 5,725) = 0.57, p = 0.46 did not show statistically significant differences in injury outcome. Summary of bivariate results Results of the bivariate analyses indicat e that almost all independent variable representing domains of the PPMHP were statistically significant at p < 0.05. Day of week was not statistically significant for younger drivers, and license compliance was not statistically significant for older drivers. For the enabling vari ables, state age-related renewal polices were statistically significantly associated with injury, but regarding the type of renewal policies, there were statistically significant di fferences among the three age-rela ted renewal policy variables. Binary logistic regression Regression model summary: Overall, the model correctly classified 87.9% of cases in predicting injury outcomes. The Hosmer and Lemoshow test ( p = 0.68), suggested that the model fitted the data well. The Nagelkerke R-Square in dicated that 57.2% of the variance in the model was explained by the data. All odds ratios were at a 95% confidence interval. Most of the statistically signi ficant associations among explor atory variables (main effect and age interaction effects) w ith injury outcome were from the environmental domain (Table
81 4.5). Measures of association for statistically significant explorat ory variables and statistically significant age interaction effect s are presented in Table 4-6. The 32 variables in the logistic regression represented five domains of the PPMHP: (a) health (2 variables), (b) behavior (3 variables), (c) environment ( 21 variables), (d) reinforcing, (5 variables), and (e) predisposing (1 variable). Table 4.5 illustrates the variab les in the logistic regression model, and whether one or more levels of variables were statistically significant at p <0.05. There were 20 statistically significant associat ions associations with age-interactions and 16 main effects. Fourteen statistically significant variables were from the environmental domain, three from the behavioral domain, two from the health domain, one from the predisposing domain, and one variable from the reinforcing domain (Table 4-5). Age interaction effects: Explanatory variables with statistically significant age interaction effects were (a) regist ered vehicle owner, (b) principal point of impact, (c) number of occupants, and (d) number of other pr evious motor vehicle convictions. For registered vehicle owner , a marginally statistically significant variable, older drivers who were not the registered owner of the vehicle they were drivi ng at the time of the crash were 53% ( p = 0.05; CI ; 0.48.00) less likely to be injured in the crash compared to drivers who were registered vehicle owners of the vehi cle they drove when the crash occurred. Regarding principal point of impact , with the 12 oâ€™clock angle (f ront impact crashes) as the referent group, two categories of the variable the 1 oâ€™clock angle ( OR = 1.61; CI : 1.05â€“ 2.47) and 7 oâ€™clock angle ( OR = 4.75; CI = 2.87.86) were statistically significant risk factors for injury.
82 Number of occupants was marginally statistically significant for older drivers. Older drivers with two or more occupants present we re 40% less likely to be injured in a crash, compared with older adults who drove alone ( OR = 0.60; CI : 0.36.01). Main effects: There were 16 variables with statistica lly significant main effects: health (1 variable), behavioral (3 variab les), environmental (11 variable s), and predisposing (1variable) domains. Health: Both age and ge nder had statistically significant associations with injury outcome. Age was an interaction term. Compared to males, females had an increased risk of injury in motor vehicle crashes ( OR = 1.51; CI = 1.29.73). Behavior: Three environmental variablesâ€” driver license compliance , driver drinking , and restraint system useâ€” had statistically significant associations with injury outcomes for younger and older drivers. Drivers who did not have a valid license for the vehicle they were operating at the time of the crash had an increased risk of injury compared to drivers who had a valid license for the vehicle they were driving at the time of the crash ( OR = 1.39; CI = 1.02â€“ 1.90). Not using a system restraint (e.g., seatbelt, la p belt) was associated with about 6 times the odds of injury compared to drivers who wo re a form of restraint during the crash ( OR = 6.20; CI = 5.03.63). For driver drinking , alcohol use was associated with 2 times the risk of injury, in comparison to drivers who had not drunk alcohol ( CI = 1.57.54). Environment: Among environmental variables, except for roadway surface condition , there were no statistically significant a ssociations among road factors such as roadway surface type (e.g., concrete or blacktop), roadway profile (whether the road was level or not), roadway alignment (straight versus curved), construction/maintenance zone , and trafficway flow (how the highway was divided) with injury outcome. Weather conditions and light condition s did not have statistically significant associations with injury
83 Factors with statistically significan t associations with injury were day of week , hour of day , number of lanes , road surface condition , rural vs. urban , vehicle body type , most harmful event , relation to junction , traffic control device functioning , airbag deployment , and National Highway System. For day of week , with Sunday as the day of reference, all days of the week except for Monday were associated with increas ed risk of injuries. Regarding hour of day , the daytime hours were protective factors for injury when compared with the nighttime hours; 8a.m.p.m. ( OR = 0.72; CI = 0.57.90), and 2p.m.p.m. ( OR = 0.63; CI = 0.53.76). For number of lanes , crashes in one-lane roads was protec tive in relation to two-lane roads ( OR = 0.32; CI = 0.12.87). Regarding roadway surface condition , compared to favorable (dry) roadway surfaces, crashes in adverse roadway conditions (e.g., snow, sleet, and rain) had 1.5 times the risk of injury ( CI = 1.16.95). The variable rural vs. urban (road function class) was protective for crashes in urban areas in comparison to cras hes in the rural area ( OR = 0.61; CI = 0.52.71). Concerning vehicle body type , in reference to sports utility vehicles (SUVs), au tomobile and automobile derivatives were risks for injury ( OR = 2.00; CI = 1.64.44). However, drivers of vans, trucks, and light pickups were 24% le ss likely of be ing injured ( OR = 0.77; CI = 0.64.94). Most harmful event was a risk factor for injury, and le vels of the variable had extremely high odd ratios. In comparison with collision w ith an object that was not fixed, drivers who collided with a fixed object had 249 tim es the likelihood of being injured ( CI = 152.61.03), while drivers in collision with moving motor vehicles had 31 times the odds of being injured ( CI = 23.82.31). Drivers in non-collision crashes we re 266 times at risk of injury (CI = 155.37â€“ 454.32).
84 For the relation to junction variable, in comparison to non-junction crashes, intersectionrelated crashes were protective ( OR = 0.59; CI = 0.48.72). Interchange-related crashes were not statistically significant. The variable traffic control device functioning was a protective factor for injury when a traffic control device was absent, with drivers 21% less likely of sustaining an injury ( OR = 0.79; CI = 0.65.95). Similarly, with the variable airbag deployed drivers were 75% less likely to be inju red when the airbag did not deploy ( OR = 0.25; CI = 0.21.29). Predisposing: Vehicle maneuver (actions before initiation of the crash) indicated that compared to going straight, driv ers in lane-related crashes ( OR = 0.64; CI = 0.50.81), maneuvers such as making a right, making a U-turn, parking or l eaving a parked position, making a controlled maneuver to av oid an object, or backing up ( OR = 0.59; CI = 0.38.92), or making a left ( OR = 0.66; CI = 0.51.87) were less lik ely to be injured. Summary of binary logistic regression results: The logistic regression showed associations and age-interaction effects among 32 independent vari ables and injury (no/yes) for younger and older drivers. Statisti cally significant, or marginally significant age interactions were evident for older drivers for registered vehicle owner , principal impact , number of occupants , and other motor vehicle convictions . For all drivers, statistically significant, differences were shown for both age groups. Summary of Results In this chapter I presented results from the uni variate, bivariate, (re gression variables and age-related license renewal policy) , and the binary logistic regre ssion with age-interaction terms. Bivariate findings indicated the majority of th e regression variables had statistically significant relationships for injury outcomes. The bivariat e findings for the exploratory study on age-related license renewal policy indicated th at states with age-related renewal policies had lower injury
85 prevalence rates. While this was true for states with in-person renewal requirements, reduced renewal cycles and vision, medical or road test requirements did not show statistically significant difference in injury rates. Results from the binary logistic regression in dicated that the model had a good fit with the data, and explained more than over half of the variance. Most of the statistically significant risk and protective factors emerged from the e nvironmental domain; how ever, all five PPMHP domains were represented as having statisti cally significant age-rela ted or main effect associations with injury outcomes. Table 4-1. Health, behavioral, environmental, predisposing and reinforc ing variables of the Precede-Proceed model of health promotion Variables Younger Drivers Number (%) N = 14,038 Older Drivers Number (%) N = 5,744 All Drivers Number (%) N = 19,782 Health Age 14,083 (71.0%) 5,744 (29.0%) 19,782 (100%) Gender Male 9,665 (68.8%) 3,827 (66.6%) 13,492 (68.2%) Female 4,373 (31.2%) 1,917 (33.4%) 6,290 (31.8%) Behavior Driver License Compliance Not Valid 1,476 (10.6%) 145 (2.5%) 1,621 (8.3%) Valid for Vehicle Type 12,445 (89.4%) 5,562 (97.5%) 18,007 (91.7%) Driver Drinking No 10,606 (75.6%) 5,451 (94.9%) 16,057 (81.2%) Yes 3,432 (24.4%) 293 (5.1%) 3,725 (18.8%) Restraint System Use None 4,571 (35.5%) 1,464 (27.5%) 6,035 (33.2%) All Types 8,295 (64.5%) 3,863 (72.5%) 12,158(66.8%) Environment Day of Week Sunday 2,002 (14.3%) 696 (12.1%) 2,698 (13.6%) Monday 1,838 (13.1%) 834 (14.5%) 2,672 (13.5%) Tuesday 740 (12.4%) 854 (14.9%) 2,594 (13.2%) Wednesday 1,838 (13.1%) 849 (14.8%) 2,687 (13.6%) Thursday 1,849 (13.2%) 836 (14.6%) 2,685 (13.6%) Friday 2,294 (16.3%) 879 (15.3%) 3,173 (16.0%) Saturday 2,470 (17.6%) 795 (13.8%) 3,265 (16.5%) Hour of Day
86 Table4 1.Continued. Table 4-1. Continued. Table 4-1. Continued. Variables Younger Drivers Number (%) N = 14,038 Older Drivers Number (%) N = 5,744 All Drivers Number (%) N = 19,782 9p.m.a.m. 5,066 (36.4%) 742 (13.0%) 5,808 (29.5%) 8a.m.p.m. 3,112 (22.3%) 2,419 (42.2%) 5,531 (28.2%) 2p.m.p.m. 5,758 (41.3%) 2,561 (44.8%) 8,319 (42.3%) Weather Conditions Non-Adverse 12,048 (86.1%) 5,109 (89.3%) 17,157 (87.0%) Adverse 1,940 (13.9%) 614 (10.7%) 2,554 (13.0%) Light Conditions Daylight 7,712 (55.1%) 4,563 (79.6%) 12,275 (62.3%) Dark 3,863 (27.6%) 622 (10.8%) 4,485 (22.7%) Other 2,416 (17.3%) 533 (9.6%) 2,969 (15.0%) National Highway System Not on the NHS 9,167 (65.5%) 3,926 (68.6%) 13,093 (66.4%) On the NHS 4,830 (34.5%) 1,800 (31.4%) 6,630 (33.6%) Rural vs. urban Rural 8,191 (58.5%) 3,184 (55.6%) 11,375 (57.7%) Urban 5,812 (41.5%) 2,539 (44.4%) 8,351 (42.3%) Number of Lanes One 120 (0.9%) 39 (0.8%) 159 (0.9%) Two 10,413 (75.0%) 4,192 (74.0%) 14,605 (74.7%) Three 1,043 (7.5%) 365 (6.4%) 1,408 (7.2%) Four to Seven 2,302 (16.6%) 1,067(18.8%) 3,369 (17.2%) Trafficway Flow Divided Highway 3,362 (24.1%) 1,359 (23.9%) 4,721 (24.0%) Not Divided Highway 8,892 (63.7%) 3,673 (64.5%) 12,565 (63.9%) One way Trafficway 1,186 (8.5%) 398 (7.0%) 1,584 (8.1%) Not Physically Divided 412 (3.0%) 234 (4.1%) 646 (3.3%) Entrance/Exit Ramp 102 (0.7%) 31 (0.5%) 135 (0.7%) Relation to Junction Non-Junction 9,383 (70.2%) 3,029 (52.7%) 12,867 (65.1%) Intersection-Related 3,766 (26.8%) 2,582 (45.0%) 6,348 (32.1%) Interchange-Related 417 (3.0%) 130 (2.3%) 547 (2.8%) Construction/Maintenance or Utility Zone Not on Construction Zone 13,676 (97.4%) 5,584 (97.2%) 19,260 (97.4%) On Construction Zone 362 (2.6%) 160 (2.8%) 522 (2.6%) Traffic Control Device Functioning No Controls Present 10,626 (76.1%) 3,585 (62.9%) 14,211 (72.3%) Controls Functioning 3,337 (23.9%) 2,116 (37.1%) 5,453 (27.7%) Roadway Surface Condition Dry 11,207 (80.2%) 4,827 (84.4%) 16,034 (81.4%) Adverse 2,764 (19.8%) 891 (15.6%) 3,655 (18.6%)
87 Table 4-1. Continued. Variables Younger Drivers Number (%) N = 14,038 Older Drivers Number (%) N = 5,744 All Drivers Number (%) N = 19,782 Roadway Surface Type Concrete 1,283 (9.6%) 400 (7.4%) 1,683 (8.9%) Blacktop 11,884 (88.7%) 4,985 (91.6%) 16,869 (89.6%) Other 230 (1.7%) 53 (1.0%) 283 (1.5%) Roadway Profile Level 10,026 (73.6%) 4,209 (76.0%) 14,235 (74.3%) Others 3,597 (26.4%) 1,331 (24.0%) 4928 (25.7%) Roadway Alignment Straight 10,869 (77.8%) 4,826 (84.4%) 19,690 (79.7%) Curve 3,101 (22.2%) 894 (15.6%) 3,995 (20.3%) Vehicle Body Type Sports Utility Vehicles 2,460 (17.5%) 373 (6.5%) 2,833 (14.3%) Automobile & Auto Derivatives 6224 (44.3%) 3,895 (67.8%) 10,119 (51.2%) Vans, Light Trucks & Pick-ups 5,354 (38.2%) 1,476 (25.7%) 6,830 (34.5%) Airbag Deployment Did not deploy/None 7,965 (66.1%) 3,079 (61.4%) 11,044 (64.7%) Deployed 4,090 (33.9%) 1,933 (38.6%) 6,023 (35.3%) Registered Vehicle Owner No 4,374 (32.3%) 820 (14.5%) 5,194 (27.0%) Yes 9,185 (67.7%) 4,830 (85.5%) 14,015 (73.0%) Number of Occupants Driver Only 9,212 (65.6%) 3,800 (66.2%) 13,012 (65.8%) One Passenger 2,865 (20.4%) 1,620 (28.2%) 4,485 (22.6%) Two or More Passengers 1,961 (14.0%) 324 (5.6%) 2,285 (11.6%) Principal Impact 1 oâ€™clock 1,620 (12.5%) 906 (16.4%) 2,526 (13.7%) 4 oâ€™clock 868 (6.7%) 290 (5.2%) 1,158 (6.3%) 7 oâ€™clock 1,368 (10.6%) 1,061 (19.2%) 2,429 (13.1%) 10 oâ€™clock 977 (7.5%) 428 (7.8%) 1,405 (7.6%) 12 oâ€™clock 7,732 (59.7%) 2,747 (49.7%) 10,479 (56.7%) Top & Undercarriage 382 (3.0%) 95 (1.7%) 477 (2.6%) Most Harmful Event Collision with Object Not Fixed 1,598 (11.4%) 471 (8.2%) 2,069 (10.5%) Motor Vehicle in Transport 8,129 (58.1%) 3,979 (69.3%) 12,108 (61.3%) Non-Collision 2,234 (15.9%) 473 (8.3%) 2,707 (13.7%) Collision with Fixed Object 2,052 (14.6%) 815 (14.2%) 2,867 (14.5%) Predisposing Vehicle Maneuver
88 Table 4-1. Continued. Variables Younger Drivers Number (%) N = 14,038 Older Drivers Number (%) N = 5,744 All Drivers Number (%) N = 19,782 Going Straight 10,014 (72.2%) 3,666 (64.4%) 13,680 (69.9%) Lane-Related 939 (6.8%) 416 (7.2%) 1,355 (6.9%) Maneuvers 295 (2.1%) 169 (3.0%) 464 (2.4%) Making a Left 663 (4.8%) 956 (16.8%) 1,619 (8.3%) Negotiating a curve or Changing 1,960 (14.1%) 488 (8.6%) 2,448 (12.5%) Reinforcing # Accident Convictions *0.16 (0.46) *0.13 (0.41) *0.15 (0.44) # Suspension Convictions *0.23 (0.89) *0.02 (0.18) *0.17 (0.76) # Driving while impaired (DWI) Convictions *0.04 (0.22) *0.01 (0.08) *0.03 (0.19) # Speeding Convictions # Other Motor Vehicle Convictions *0.25 (0.62) *0.20 (0.58) *0.07 (0.29) *0.09 (0.34) *0.19 (0.55) *0.16 (0.53) * Mean (SD) Table 4-2. Age-related license renewal policies for older adults as of 2003 ( N = 5,747). Age-Related Renewal Policy Number of States* N (%) Age-Related Renewal Policies No 26 3,554 (62.1%) Yes 25 2,171 (37.9%) Reduced Renewal Cycles No 37 4,455 (77.8%) Yes 14 1,270 (22.2%) In-Person Renewal Required No 43 4,629 (81.5%) Yes 8 1,052 (18.5%) Age-Related Testing No 44 5,244 (91.6%) Yes 7 481 (8.4%) *Includes the District of Columbia Table 4-3. Prevalence of driver s injured in a crash by age group for health, behavior, environment, and predisposing domains. Younger Drivers % p -value Older Drivers % p -value Health Gender <0.01* <0.01* Male 73.0 84.1 Female 76.8 88.1 Behavior Driver License Compliance <0.01* 0.19
89 Table 4-3. Continued. Younger Drivers % p -value Older Drivers % p -value Not Valid 84.0 88.3 Valid for Vehicle Type 72.9 85.2 Driver Drinking <0.01* <0.01* No 69.0 85.0 Yes 90.0 92.2 Restraint System Use <0.01* <0.01* None 94.8 97.0 All Types 63.5 81.2 Environment Day of Week 0.11 0.02* Sunday 75.9 82.0 Monday 73.3 87.6 Tuesday 74.9 87.0 Wednesday 71.8 84.8 Thursday 74.6 86.4 Friday 73.9 85.9 Saturday 74.3 83.4 Hour of Day <0.01* <0.01* 9p.m.a.m. 78.5 78.7 8a.m.p.m. 73.8 88.6 2p.m.p.m. 70.1 84.3 Weather Conditions <0.01* 0.04* Non-Adverse 73.3 85.2 Adverse 79.0 87.8 Light Conditions <0.01* <0.01* Daylight 74.1 88.1 Dark 80.4 79.9 Other 64.2 69.4 National Highway System <0.01* <0.01* Not on the NHS 73.3 85.4 On the NHS 75.9 87.4 Road Function Class <0.01* <0.01* Rural 82.9 90.4 Urban 61.9 79.0 Number of Lanes <0.01* <0.01* One 70.0 84.6 Two 77.8 87.0 Three 62.7 80.5 Four to Seven 63.9 81.3 Trafficway Flow <0.01* <0.01* Divided Highway 70.9 84.5 Not Divided Highway 77.8 86.8 One way Trafficway 65.4 82.9 Not Physically Divided 49.0 73.9
90 Table 4-3. Continued. Younger Drivers % p -value Older Drivers % p -value Entrance/Exit Ramp 83.7 90.3 Relation to Junction <0.01* 0.61 Non-Junction 78.6 85.4 Intersection-Related 62.8 85.3 Interchange-Related 73.4 88.5 Construction/Maintenance or Utility Zone 0.62 0.84 Not on Construction Zone 74.1 85.5 On Construction Zone 75.4 81.3 Traffic Control Device Functioning <0.01* <0.01* No Controls Present 75.8 84.2 Controls Functioning 68.8 87.5 Roadway Surface Condition <0.01* <0.01* Dry 72.8 84.8 Adverse 79.5 89.0 Roadway Surface Type <0.01* <0.01* Concrete 72.0 86.0 Blacktop 74.4 85.4 Other 86.5 88.7 Roadway Profile <0.01* <0.01* Level 72.4 84.7 Others 79.8 88.2 Roadway Alignment <0.01* <0.01* Straight 70.5 84.1 Curve 86.8 92.7 Vehicle Body Type <0.01* <0.01* Sports Utility Vehicles 73.9 83.9 Automobile & Auto Derivatives 78.6 87.5 Vans, Light Trucks & Pick-ups 69.1 80.1 Airbag Deployment <0.01* <0.01* Did not deploy/None 68.7 81.7 Deployed 91.0 95.8 Registered Vehicle Owner <0.01* <0.01* No 75.7 82.2 Yes 73.2 85.9 Number of Occupants 0.23 0.08 Driver Only 74.3 85.2 One Passenger 74.7 86.7 Two or More Passengers 72.6 82.1 Principal Impact <0.01* <0.01* 1 oâ€™clock 73.8 84.4
91 Table 4-3. Continued. Younger Drivers % p -value Older Drivers % p -value 4 oâ€™clock 56.6 75.2 7 oâ€™clock 78.0 94.8 10 oâ€™clock 73.6 84.8 12 oâ€™clock 73.7 83.3 Top & Undercarriage 67.8 63.2 Most Harmful Event <0.01* <0.01* Collision with Object Not Fixed 10.0 17.0 Motor Vehicle in Transport 74.6 89.3 Non-Collision 97.2 98.5 Collision with Fixed Object 97.9 99.0 Predisposing Vehicle Maneuver <0.01* <0.01* Going Straight 73.6 84.5 Lane-Related 58.1 81.5 Maneuvers 58.6 76.9 Making a Left 58.8 87.1 Negotiating a curve or Changing 90.8 94.1 Reinforcing** #Previous accidents 0.26 0.57 # Previous suspensions <0.01* 0.92 # Previous Driving While Intoxicated (DWI) <0.01* 0.54 # Previous speeding 0.01 0.12 #Previous other motor vehicle convictions 0.09 0.05* NHS: National Highway System
92 Table 4-4. Prevalence of driver s injured by type of age-re lated license renewal policy Injury % p -value Age-related renewal policy <0.01* No 86.4 Yes 83.6 Reduced renewal cycle 0.59 No 85.5 Yes 84.9 In-person renewal required <0.01* No 86.4 Yes 80.8 Vision, medical or road test required 0.46 No 85.5 Yes 84.2 * p < 0.01 Table 4-5. Precede-Proceed model of health promo tion variables in logistic regression model by domain and statistical significance Domain Variable Significant at p < 0.05 Age interaction effect Health Age X N/A Gender X Behavior Driver license compliance X Driver drinking X Restraint system use X Environment Day of week X Hour of day X Registered vehicle owner X X Number of lanes X Roadway surface condition X Roadway surface type Rural vs. urban X Roadway profile Roadway alignment Vehicle body type X Most harmful event X Relation to junction X Principal impact X X Traffic way flow Traffic control functioning X Light conditions
93 Table 4-5. Continued. Domain Variable Significant at p < 0.05 Age interaction effect Weather conditions Construction/maintenance zone Number of occupants X X Airbag deployment X National Highway System X Predisposing Vehicle maneuver X Reinforcing # previous accident convictions # previous suspension convictions # previous Driving while impaired (DWI) convictions # previous speeding convictions # previous other motor vehicle convictions X X Table 4-6. Binary logistic regre ssion model showing statistically significant age interactions and statistically significant explanatory variab les from the Precede -Proceed model of health promotion with injury (yes/no) Dependent Variable: Injury (yes/no) p OR Lower CI Upper CI Health Domain Gender Male (Referent) Female <0.01* 1.51 1.29 1.73 Behavior Domain Driver license compliance Valid (Referent) Not Valid 0.04* 1.39 1.02 1.90 Driver drinking Not drinking (Referent) (Referent) Drinking <0.01* 2.00 1.57 2.54 Restraint system use Yes (Referent) None <0.01* 6.20 5.03 7.63 Environment Domain Day of week Sunday (Referent) Monday 0.18 1.17 0.93 1.48 Tuesday 0.02 1.33 1.05 1.69 Wednesday <0.01* 1.64 1.27 2.13 Thursday <0.01* 1.46 1.14 1.86 Friday 0.01* 1.36 1.07 1.74
94 Dependent Variable: Injury (yes/no) p OR Lower CI Upper CI Saturday 0.03* 1.29 1.02 1.62 Hour of day 9p.m.a.m. (Referent) 8a.m.p.m. 0.01* 0.72 0.57 0.90 2p.m.p.m. <0.01* 0.63 0.53 0.76 Registered vehicle owner x Age Driver was registered owner x Age (Referent) Driver was not owner x Age 0.05* 0.69 0.48 1.00 Number of lanes Two (Referent) One 0.03* 0.32 0.12 0.87 Three 0.41 0.32 0.68 1.17 Fourâ€“Seven 0.29 0.89 0.75 1.09 Road surface condition Favourable (Referent) Adverse <0.01* 1.50 1.16 1.95 Rural vs. urban Rural (Referent) Urban <0.01* 0.61 0.52 0.71 Body Type SUVs (Referent) Auto & Auto Derivatives <0.01* 2.00 1.64 2.44 Vans, Trucks, & Light Pick-Ups 0.01* 0.77 0.64 0.94 Most harmful event Collision with object not fixed (Referent) Collision with fixed object <0.01* 249.55 152.61 408.03 Motor vehicle in transport <0.01* 30.99 23.82 40.31 Non-Collision <0.01* 265.68 155.37 454.32 Relation to Junction Non-Junction (Referent) Intersection-Related <0.01* 0.59 0.48 0.72 Interchange-Related 0.94 0.98 0.64 1.51 Principal impact x Age 12 oâ€™clock (Referent) 1 â€“ 3 oâ€™clock 0.03* 1.61 1.05 2.47 4 â€“ 6 oâ€™clock 0.50 1.20 0.71 2.05 7 â€“ 9 oâ€™clock <0.01* 4.75 2.87 7.86 10 â€“ 11 oâ€™clock 0.15 1.47 0.87 2.48 Roof or undercarriage 0.95 0.96 0.28 3.33 Traffic control device Functioning (Referent) Not Present 0.01* 0.79 0.65 0.95 Table4 6.Continued.
95 Table 4-6. Continued. Dependent Variable: Injury (yes/no) p OR Lower CI Upper CI Number of occupants x Age Driver only x Age (Referent) One Passenger x Age 0.34 1.18 0.84 1.64 > Two Passengers x Age 0.05* 0.60 0.36 1.01 Airbag deployment Deployed (Referent) Did Not Deployed <0.01* 0.25 0.21 0.29 National highway system On NHS (Referent) Not on NHS <0.01* 0.77 0.65 0.91 Predisposing Domain Vehicle maneuver Going straight (Referent) Lane-related Maneuvers <0.01* 0.64 0.50 0.81 Making a left 0.02* 0.59 0.38 0.92 Negotiating a curve/changing <0.01* 0.66 0.51 0.87 0.35 1.15 0.86 1.54 Reinforcing Domain ** Number previous other MV convictions x Age 0.03* 0.65 0.44 0.97 * p < 0.05 **MV = motor vehicle convictions 77.4% 74.1% 85.4% 68 72 76 80 84 88Percent Injured All DriversYounger DriversOlder Drivers Age Group Figure 4-1. Drivers injured by age group
96 88.1%* 76.8%* 84.1%* 73.0%* 020406080100 Younger Male Older Male Younger Female Older FemaleGenderPercent Figure 4-2. Drivers inju red in crash by gender 85.2% 88.3% 72.9%* 84%* 0 20 40 60 80 100 Not Valid (YD)Valid (YD)Not Valid (OD)Valid (OD) Drivers LicencePercent *p < 0.05; YD = Younger Driver; OD = Older Driver Figure 4-3. Drivers injured in cr ash by driver license compliance
97 84.0%* 72.9%* 88.3%* 85.2%* 020406080100 None (YD) Restraint (YD) None (OD) Restraint (OD) * p < 0.05; YD = Younger Driver; OD = Older Driver Figure 4-4. Drivers injured in crash by restraint system use * p < 0.05; YD = Younger Driver; OD = Older Driver Figure 4-5. Drivers injured in crash by time of day 73.8%* 70.1%* 78.5%* 88.6%* 84.3%* 78.7%* 0 20406080100 8a.mp.m (YD) 2p.mp.m (YD) 9p.ma.m (YD) 8a.mp.m (OD) 2p.mp.m (OD) 9p.ma.m (OD) Hour of DayPercen t
98 73.9%* 78.6%* 69.1%* 83.9%* 87.5%* 80.1%* 0.020.040.060.080.0100.0 SUVs (YD) Autos (YD) Vans, Trucks & Pick Ups (YD) SUVs (OD) Autos (OD) Vans, Trucks & Pick Ups (OD)Vehicle Body TypePercent * p < 0.05; YD = Younger Driver; OD = Older Drive Figure 4-6. Drivers injured in crashes by vehicle body type 99.0%* 98.5%* 89.3%* 17.0%* 97.9%* 97.2%* 74.6%* 10.0%*0.020.040.060.080.0100.0Object not Fixed (YD) Motor Vehicle in Transport (YD) Non-Collision (YD) Fixed Object (YD) Object not Fixed (OD) Motor Vehicle in Transport (OD) Non-Collision (OD) Fixed Object (OD)Collision TypePercent * p < 0.05; YD = Younger Driver; OD = Older Driver Figure 4-7. Drivers injured in cr ash by most harmful event type
99 Older Driver Younger Driver 0.00 0.04 0.08 0.12 0.16 0.20 No injuryInjury Injury OutcomeMean # Accident Convictions Figure 4-8. Previous accident convic tions by injury outcome and age group Figure 4-9. Previous driving while impai red (DWI) convictions by injury outcome and age group Younger Driver Older Driver 0.00 0.02 0.04 0.06 No injuryInjury Injury SeverityMean # DWI Convictions
100 Younger Driver Older Driver 0.00 0.10 0.20 0.30 No injuryInjury Injury OutcomeMean # Suspension Convictions Figure 4-10. Previous suspension convict ions by injury outcome and age group Younger Driver Older Driver 0.00 0.10 0.20 0.30 No injuryInjury Injury Outcome#Mean #Speeding Convictions Figure 4-11. Previous speeding convic tions by injury outcome and age group
101 Younger Driver Older Driver 0.00 0.10 0.20 0.30 No injuryInjury Injury OutcomeMean # Other MV Convictions Figure 4-12. Other motor vehicle convi ctions by injury outcome and age
102 CHAPTER 5 DISCUSSION Overview The main objective of this study was to us e a crash database and a socio-ecological public health model, the Precede-Proceed model of health promotion (PPMHP), to investigate the socio-ecological determinants of motor vehi cle injuries among older drivers (65 years and older) in the United States, using younger (35 to 54 years) drivers as a comparison group. The aims of the study were to (a) describe drivers in the 2003 FARS dataset with respect to their level of injury in motor vehicle crashes, (b) examin e the relationship between the determinants of motor vehicle injuries and injury outcome, and (c) for older driver s, examine the effect of state licensure renewal policies on motor vehicle injuries. Limitations of the Study As previously discussed, the study has the lim itations of a secondary dataset and a crash dataset. Specific to the FARS cr ash dataset, (a) all drivers were involved in a fatal crash (at least one person died), making the inclusion criter ia for the dataset narrow; (b) there was no comparison with a sample of drivers who were not involved in crashes; (c) unreported crashes were not represented in the FARS database; (d) a driver may have been involved in a fatal crash more than once in 2003, and thus more than one crash involving the older driver may have been in the 2003 FARS dataset; (e) da ta collection for environmental, predisposing, and some behavioral variables is based mainly on police ac cident reports, and all states do not routinely report all the FARS elements on the police reports (Massie & Campbell, 1996); and (f) there were not available data on socio-demographic va riables such as income and education that may have given a complete picture of their association with injuries.
103 Specific to the study, for the bivariate analys es, almost all independent variables were statistically significant at p< 0.05 level, possible because of the huge sample size. For the study on state age-related license renewa l policies, some states belong to more than one category (type of policy was not mutually exclusive), and thus, the method was not a rigorous analysis. Secondly, the age-renewal policy reduced renewal cycl e, had a wide state vari ation of the ages at which the policy was implemented, and the renewal cycle (years). Regarding the binary logistic regression, while I was able to determine associations among independent variables and injur y, it was difficult to disentangl e the contribution of complex variables such as relation to junction , vehicle maneuver , and most harmful event , and injury. It was in part because the sample was comprised of drivers who had been involved in a crash. I was also not able to establish whethe r the driver involved in the crash was at fault or not, due to the inadequate data on variables that may have provi ded more information (accident-related factors, violations charged, and ve hicle-related factors). Procedure of the Study I accomplished these objectives in three ways. Using the findings and structure created by a preceding research studyâ€”a systematic literature review on older driver safety, the PPMHP structural model, several c onsultation meetings with the Project to Promote Safe Elder Driving team at the University of Florida, and a National Highway Traffic Safety Administration (NHTSA) consultant, I selected apposite variables for the analyses. I used univariate and bivariate analyses to examine the data and collapse levels of categories where necessary and conducted a bivariate analysis to ascertain the level of injury for exploratory variables for both younge r and older drivers. I performe d a binary logistic regression model to examine measures of associations (odds ratios) between the socio-ecological determinants of motor vehicle injuries and inju ry outcome in a motor vehicle crash. Thirdly, I
104 grouped U.S. states by four factors pertaining to age-related license renewal policies and conducted exploratory bivariate an alyses to investigate the rela tionship among states with and without age-related license renewa l policies and injury outcome. Findings in Light of Research Questions and Hypotheses In this section, I discuss the results from th e perspective of the research questions and hypotheses. Injury Prevalence among Independent Variables for Younger and Older Drivers Research Question #1: What is the preval ence of the main determinants (risk and protective factors) and motor vehicle injury for younger drivers (35 to54 y ears) and older drivers (65 years and older)? Ho: Younger drivers will not have lower prevalence rates for motor vehicle injuries compared to older dr ivers (65 years and older). Ha: Younger drivers (35 to 54 years) will have lower motor vehicle injury prevalence rates compared to older drivers (65 years and older). To achieve this objective, I examined the re lationship among the determinants of motor vehicle injury (exploratory variab les) and injury (dependent vari able) separately for younger and older drivers using chi-square tests for categori cal variables and independent sample t-tests for numerical variables. Older driv ers did have higher injury rate s among all the determinants of motor vehicle injury and injury outcome. However, not all determinants had statistically significant differences in injury prevalence rates. Driver license compliance (whether the driver had a valid license for the vehicle operated at th e time of the crash) had a statistically significant relationship with injury for younger dr ivers, but not for older drivers. Day of the week had statistically significant differences in injury outcomes for older driv ers, but not for younger drivers. Also, all variables in the reinforci ng domain of the PPMHP (numbers of previous
105 accidents, suspensions, DWI (driving while impaired), speeding and other motor vehicle convictions) had statistically si gnificant relationships with in jury outcomes for younger drivers. However, only other motor vehicle convictions (failure to yield, lane-r elated, and running a red light) was statistically significantly different for older drivers. Based on these findings, I reject the alternative hypothesis that young er drivers will have lower moto r vehicle injury prevalence rates. These findings are descrip tive in nature and give an ove rview of younger and older drivers regarding the determin ants of motor vehicl e injury outcomes. Measures of Association among I ndependent Variables and Injury Research Question #2: What are the measur es of association among socio-ecological determinants (behavioral, environmental, pr edisposing, and reinforcing factors) confounding variables (age), and motor injury (injury: yes/no) fo r younger and older drivers? Ho: Younger drivers (35 to 54 years) will not have lower odds ratios for environmental variables (e.g., hour of day , and most harmful event ) and predisposing factors ( vehicle maneuver ) and injury, compared with olde r drivers (65 years and older). Ha: Younger drivers (35 to 54 years) will have lower odds ratios for environmental variables (e.g., hour of day , and most harmful event ) and predisposing factors ( vehicle maneuver ) and injury, compared with olde r drivers (65 years and older) To achieve this objective, I conducted a bina ry logistic regressi on between 32 variables from five domain of the PHMH P (with age-interaction terms) and injury outcome (no/yes). There were four significant age interaction effect s in the model (significant associations between older driver group and injury outcome), three from the behavioral domain ( registered vehicle owner , principal impact , and number of occupants ), and one from the reinforcing domain ( number of previous other motor vehicle convictions ). Although there were three environmental
106 variables for older drivers only, on the whole, stat istically significant risk and protective factors pertained to both younger and olde r adults. I therefore reject the alternative hypothesis. Age-related License Renewal Policies Research Question #3: What is the prevalence of injuries for older drivers among states with age-related license renewal policies compar ed to states without any age-related license renewal policies? Ho: There will be no difference in motor ve hicle injury prevalence rates between states with no age-related licensure rene wal procedures and t hose with age-related licensure renewal procedures. Ha: There will be differences in motor ve hicle prevalence rates between states with age-related licensure and states with no age-related licensure renewal procedures. This exploratory study grouped U.S. states by factors: age-related renewal policy, reduced renewal cycle, in-person renewal, and vision/medical testing. Generally, drivers with licenses from states with age renewal policies had lower injury prevalence rates than drivers with licenses from states with no age-related renewal policies. However, for reduced renewal cycles for older drivers, there was no difference in injury prevalence rates between states with reduced renewal cycles and those without . I therefore reject the null h ypothesis of no difference between states with and without age-re lated policy renewals and accept the alternativ e hypothesis as stated above. Implications for Policy, Practice and Research I will discuss findings from the binary logistic regression model for the 20 significant associations among independent variables and inju ry as well as the bivariate analyses for agerelated license renewal policie s in light of epidemiology, policy, and health services implications.
107 Findings from the final model The risk and protective determinants of injury that were associated for both age groups included health ( gender ), behavior ( driver license compliance , driver drinking , restraint system use ), environment ( time of day , hour of day , number of lanes , road surface condition , vehicle body type , most harmful event , relation to junction , traffic control device , airbag deployment and National Highway System, rural vs. urban ), and predisposing ( vehicle maneuver ). Female drivers were 1.5 times more likely to be injured in a crash compared to male drivers. The gender differences in driver safety outcomes (crashes, injuries and fatalities) are sustained by previous studies (Finison & Dubrow , 2002; Baker, et al., 200 3; Bauer, et al., 2003; and NHTSA 2006a), and lend historical plausibility to the results of this study. This finding has societal implications as olde r women generally live longer than older men. Thus, older women are likely to require transportation in the latter ye ars of their lives. They may, therefore, require interventions such as motor vehicle injury prev ention program. This fi nding has epidemiological implications for research specifically targeted at older female drivers. Behavioral factors (alcohol i nvolvement, lack of seatbelts, and invalid driversâ€™ license) were risk factors for injuries for both age groups. Although 24% of younge r drivers had alcohol involvement compared with 5% of older drivers, ol der drivers were equally at risk for injury in crashes as younger drivers when there was alcoho l involvement. Currently, there are policies at state levels for alcohol use and driving, and seatbelt use. Over th e years, the monitoring of these driving behavioral factors has been associat ed with decreased crashes and injuries among drivers. Continual surveillance of alcohol and seatbelt adhere nce by state governments may be one of the best approaches to decrease injuries among younger and older adults. For environmental factors, registered vehicle owner , not being the registered vehicle owner for the motor vehicle was a protective factor for injury. The reason for this is unclear and
108 there is no available research to buttress this fi nding. However, I postulate that older drivers who drive vehicles not registered in their names are younger and appear healthy (physical, mental, and cognitive) as it is likely that involving a third party would re quire some sort of precursory assessment of the driversâ€™ ability to operate the vehicle. About 10% of older Americans (55 years and older) consent others (family or other) to drive their motor vehicles when the older adult prefers not to drive (Hermanson, 2005). Two reasons offered for this phenomenon was safety and socialization . Older adults (especially those 75 years a nd older) selected someone else to drive because they did not feel safe driving themselv es, or wanted to meet with people (Hermanson, 2005). Older drivers are finding ways of mainta ining independence (having a motor vehicle), and yet practicing safety precautions by having a desi gnated driver when they do not feel safe to drive, and may be a self-regulatory strategy of some older drivers, and has implications for older driver research. The finding also has implica tions for policy, specifically with regard to automobile insurance. If older adults are mainta ining their automobiles so that others can drive them, automobile insurance companies may have to account for car sharing in their premium rates, and yet be flexible e nough to meet the needs of older adults (Hermanson, 2005). Regarding principal impact (the direction at which the vehicle was struck during the crash), the 1 oâ€™clock angle (fr ont passenger side) and 7 oâ€™clock angle (the back side portion of the motor vehicle) were sign ificant injury risk factors for the older driver group. Front-side impact crashes may be high-risk areas for inju ries among older drivers because of kinetic or mechanical forces that directly impact the driver , and the frailty and fragil ity of older drivers. A study on airbags by NHTSA indicated that airbags in passenger cars are most effective in protecting passengers from injuries when the car is struck at the 12 oâ€™clock angel (front impact crashes), with no rollovers. Airbag s are less effective when the vehicle is struck from the 1
109 oâ€™clock, 2 oâ€™clock, 10 oâ€™clock, 11 oâ€™clock, or 12 oâ€™clock with subsequent rollover (Kahane, 1996). However, from this analysis, it is difficu lt to ascertain why th e 7 oâ€™clock angle was a high-risk angle for older drivers. To better unde rstand these mechanisms, another analysis such as a loglinear analysis would have to be performed to unders tand the inter-relationships among certain environmental variables (e.g., speed, relation to junction, most harmful event, vehicle maneuver), and principal impact. However, this finding has implications for policy enforcement interventions in the area of impr oving minimum standards for safety in motor vehicles, such as side airbag protection. The number of passengers in the vehicle at the time of the crash was a marginally significant protective factor for older drivers who had two or mo re passengers in the vehicle. Research by Bedard and Meyers (2004) suggests that the presence of four or more passengers is protective for older drivers 65 ye ars old; this was associated with less risk of crashes for some unsafe actions such as driving the wrong wa y and was not significantly associated with passing, but was associated with hi gher risks of other driver actions such as ignoring signs and warning, with higher risks for drivers 80 years an d older. Baker et al. (2003) found that either two passengers or three or more passengers we re protective for crashes among older female drivers, while driving alone was a risk factor for cr ashes. Hing et al. (2003) related time of day to number of passengers on the safety of older adu lts, with results indicating that while drivers 75 years and older had higher crash ratios than ot her age groups, driving at night was protective for crashes. The results from this study, in conjunction with previous studies, indicate that the presence of passengers may be protective fo r older drivers but the protection may vary depending on other factors such as time of day, ge nder, and type of most harmful event (collision
110 with a fixed object) the driver was engaged in at the time of the crash. The three studies focus on different outcome variables and various combinati ons of exploratory variables for the analyses; thus, to have a clearer picture of the factors involved in the numb er of passengers as a protective factor in injury, more information would be n ecessary. Further analyses such as a loglinear analysis exploring the relationshi p among variables such as health -related factors, time of day, most harmful event, and injury may better explai n the relationship betwee n passengers and injury outcome for older adults. A qualitative study, such as a focus group examining older driversâ€™ perceptions of passengers woul d help buttress findings from the quantitative analyses. The number of previous motor vehicle convic tions was marginally protective for older drivers; that is, for every increase in the numbe r of other motor vehicle convictions, there was a decreased risk in injury. It is difficult to as certain using quantitative data alone why previous number of motor vehicle convictions for failure to yield, running a red light, or lane-related changes was significantly associat ed with reduced risk of inju ries for the older driver group. However, this may have occurred as a result of family and/or caretaker interventions after an older driver got involved in a motor vehicle cr ash. From the findings, policy enforcement of motor vehicle violations may be pertinent to ol der adult injury preventi on. To better examine the influence of reinforcement on injury, further research (e.g., focus group study) that include family/caretaker roles in older ad ultsâ€™ decisions to drive may have to be taken into account. The hour of the day the crash occurred was prot ective for daylight hour injuries. Previous older driver research suggests that daylight hou rs are riskier for crashes involving older drivers compared with other age groups that are more at risk for crashes during the night hours (Finison & Dubrow 2002; Baker et al 2003; NHTSA, 2006 b). The findings app ear contradictory; however, the cited research used crashes as th e outcome variable, while this study focuses on
111 injury outcomes. The findings from this study sugge st that drivers are at higher risks for injury during the daylight hours. Further research may as certain whether there are differences in injury outcomes, controlling for the number of crashes. Two-lane crashes was a protectiv e factor for injury compared to one-lane crashes. Further analyses considering other factors, including trafficway flow (how the highway was divided), might need to be taken into consideration before the implications of these results can be acted upon. Regarding roadway surface conditions , crashes on adverse road co nditions (e.g., wet, oil) had higher risks of injuries comp ared with crashes on dry road conditions. This opposes previous findings (Finison & Dubrow 2002; Baker et al., 2003) implying that crashes are more likely to occur on roads with favorable conditions for older dr ivers. It must be remembered that the data for this study comprised of both younger and older drivers, and has injur y, not crashes, as the outcome variable of interest. Automobile and autom obile derivatives were risk factors for injury compared with SUVs, while light trucks, vans, and pic kups were protective for injury. SUVs are generally larger and therefore the kinetic forces are bigger compared w ith an automobile. A possible intervention is to improve vehicle minimum standards and educat e the public on safe vehicles. NHTSA has devised future laws to improve vehicles by in cluding anti-rollover devices and currently rates vehicles by their crashworthiness, making this rating available to buyers (NHTSA, 2006c). Thus continual policy enforcement of minimum standard s for motor vehicles w ould be beneficial to drivers of all ages in injury reduction. The most harmful event had very high odds ratios for injury for non-collision crashes, collisions with other motor vehicles, and collis ion with fixed objects. E nvironmental engineering of roadways such as replacing fixed objects (e.g., concrete divides) with non-fixed objects, such
112 as cable barriers, may reduce impact severity of crashes and injuries among drivers. For vehicles, improving technology such as anti-rollover de vices may reduce non-collision crashes. Compared to drivers traveling on non-junctions (e.g., rail road cro ssing or bridges), the level intersectionâ€“related crashe s was protective for injury. This may be because intersections are usually more structured (e.g., traffic lights) compar ed with non-junctions, which enable speed reduction and increased awareness of the enviro nment. Research indicates that roads with enhanced intersections (using the Federal Hi ghway Administrationâ€™s recommendations) by and large benefit the safe driving performan ce of both younger and ol der drivers (Classen, Shechtman, Stephens et al., 2006; Shechtm an, Classen, Stephens et al., 2006). In contrast to roads with f unctioning traffic control device s, younger and older drivers on roads without traffic control devices were less likely to be injured in the crash. The reason for this is not clear. I postulate th at roads with traffic control devices may be generally more complex than roads without traffic control devices. From the results of this study not having airbag deployment in the crash was protective for injury. This may be because airbag deploym ent is associated with front and high impact crashes (i.e., serious crashes) a nd thus may be more likely to re sult in injury. One implication of this result is considering follow-up driver educa tion to ensure proper posit ioning of the seat in relation to the airbag. Crashes on the National Highway System was a ri sk factor for injury. It may be because roads not on the NHS are likely to be less tr aveled on and less complex than roads on the National Highway System. Compared to rural areas, crashes in urban areas were less likely to result in injury. This somewhat contradicts previous research that suggests that crashes are more likely to take place in urban areas compared to rural areas (Finison & Dubrow 2002). However,
113 Higher risk of injury in urban areas may be expl ained by the fact that ur ban areas are more likely to have more complex roads and traffic patterns. The high traffic density may be more likely to result in low speed crashes that are less likel y to result in injury. Thus continued speed enforcement may help enforce speed restrictions. Compared to going straight, drivers engaged in lane related maneuvers and making a left turn had a lesser chance of being injured in a traffic cr ash. This may be possibly because performing maneuvers (e.g., making a left turn) re quire the driver to decrease traveling speed, resulting in decreased crash im pact, and risk of injury. Findings from the behavioral domain suggest that regulating driver behavior through enforcement may be a good way of reducing injuri es among all drivers. En vironmental factorsâ€” physical (vehicle factor, highw ay factors) and social (num ber of occupants)â€”contributed significantly as risk and protec tive factors for injuries for bot h age groups. However, vehicle factors such as vehicle owner or principal point of im pact contributed as risk and protective factors for older drivers only. Findings from age-related license renewal policies Findings from the exploratory bi variate analyses of state agerelated license renewal laws showed that reduced renewal cycles and visi on/medical test requi rements did not have statistically significant association with injur y. This may have been because the states had different reduced renewal cycles (years) and, also , the ages for implementing this policy varied from between 60 and 79 years. A more complex anal ysis, such as a hierarchal linear analysis, may be required to investigate the effects of the different y ears of renewal and the age of inception. In-person renewals, however, had decreased prevalence of injury compared to states with no in-person renewals. This is plausible be cause in-person renewals entail some form of
114 physical assessment, which would screen dr ivers who have obvious vision or physical disabilities. Research on the effects of age-renewal po licies on older driver safety outcomes is contradictory. In-person renewal ha s been associated with reduced fatality rates among drivers over 85 years, with other age-renewal policies su ch as vision and road tests insignificantly associated with injury (Grabowski et al., 2004). However, Levy, Vernick, & Howard (1995) found that a state mandated vision te st was associated with reduced fatal crash risk for drivers 70 years and older. It is possible th at states with vision test requi rements may not have differences in injury prevalence rates because state vision re quirements are mainly tests of visual acuity and do not measure other important vision component s such as visual field loss, and contrast sensitivity (Ball, Owsley, Sloane, Roenker, & Bruni, 1993; Owsley, Ball, Sloane, Roenker, & Bruni, 1991). Conclusion This study used the Precede-Proceed model of health promotion (PPMHP) to investigate relationships among socio-ecological variables an d motor vehicle injury outcomes for older drivers (65 years and older) and compared them with younger drivers (35 years). The source of data was the 2003 Fatality Analysis Reporti ng System (FARS), a na tional secondary crash database. Prevalence rates and measures of associations (odds ratios) among independent variables and injury were reported for younger and older drivers. An exploratory sub-analysis (older drivers only), compared injury prevalence rates among stat es with age-related license renewal polices against states without age-related license renewal policies. Regarding age-related renewal policies, states with age-related license renewal polices, and states requiring in-perso n renewal were associated with lower injury rates. The final model
115 yielded that the environmental domain, a domain not well studied in the existing literature (Classen et al, 2006), emerged as predominant to explain important associations with injury. Examining risk and protective factors for motor vehicle injury among younger and older drivers in the U.S. demonstrated that a socioecological approachâ€”an appr oach not yet utilized in the existing older driver literature, is needed to reveal the multiple factors associated with injury. Many of the findings showed relevance to drivers from both age groups, with a selected few pointing to older adults, meaning that in jury prevention measures, when developed and implemented, may potentially benefit ol der and younger drivers alike. The significant findings from the behavioral (alcohol, driver license compliance, and seatbelt use), environmental (imp rovement of vehicle crashworth iness and highway design), and enabling (age related policy renewal) domains hold therefore important implications for injury prevention programs. For example, preliminary in jury prevention strategi es to be examined include continued policy enforcement, envir onmental engineering (e.g. highway design or crashworthiness of vehicles) a nd driver educational programs. Policy enforcement is considered the most e ffective strategy for decr easing motor vehicle injury rates (World Health Organization, 2004). Enforcement of seatbelts and alcohol laws may continue to decrease injury rates among olde r and younger drivers, while age-related license renewal laws may primarily benefit older drivers. In the event of a cr ash, policies designed to improve crashworthiness of motor vehicles may benefit older and younger drivers and passengers alike. This study also emphasizes that differences exist in the dynamics underlying motor vehicle crash outcomes and motor vehicle injury outcomes. For example, a driver may be involved in a motor vehicle crash but may not necessarily be in jured as a result of maintaining low speed or
116 using the in vehicle restraints properlyâ€”both factors associated w ith injury prevention/reduction. Many of the factors contributing to injury may not be explained by the data contained in the aggregated findings of this study. Therefore, if the soci o-ecological determinants associated with injury are to be understood, then further research addressing these specific areas is of utmost importance. Additional research is underway in the Public Health Model to Promote Safe Elderly Driving project to shed more light on the complexities involved in injury outcomes. For example a qualitative study is planned to next examine cr ash related injuries among older adults from their personal perspect ives. Ongoing quantitative research in cludes a log linear analysis to examine the relationships and inte raction effects among variables that had statistically significant associations with injury. Although this study does not have direct implicati ons for health services research at this point in time, it lays the foundation for injury pr evention programs. As such, it creates plausible research and policy making opportunities for working towards accessible and affordable injury prevention programs, especially for older drivers.
117 APPENDIX A PRECEDE-PROCEED MODEL OF HEALTH PROMOTION Figure A-1. Precede-Proceed Model of Health Promotion. [Re-printed with permission from Gr een, L.W., & Kreuter, M.W. (2005). Health Program Planning: An Educational and Ecological Approach . (4th ed.). (Page 10, Figure 1-2) McGrawHill Companies, Inc. NY] Predisposing factors Reinforcing factors Enabling factors Behavior Environment Health Quality of life Health education Policy regulation or g anization Phase 4 Administrative and policy assessment Phase 3 Educational and ecological assessment Phase 2 Epidemiological assessment Phase 1 Social assessment HEALTH PROMOTION Genetics Phase 5 Implementation Phase 6 Process evaluation Phase 7 Impact evaluation Phase 8 Outcome evaluation
118 APPENDIX B STRUCTURAL MODEL OF OLDER DRIVER SAFETY Predisposing factors â€¢Knowledge â€¢Attitudes/ Beliefs/ Values â€¢Perceptions Predisposing factors â€¢Knowledge â€¢Attitudes/ Beliefs/ Values â€¢Perceptions Reinforcing factors â€¢Positive â€¢Negative Reinforcing factors â€¢Positive â€¢Negative Enabling factors â€¢Resource availability/ accessibility â€¢Referrals â€¢Rules/laws â€¢Traffic engineering Enabling factors â€¢Resource availability/ accessibility â€¢Referrals â€¢Rules/laws â€¢Traffic engineering Behavior & lifestyle â€¢Safe (e.g., seat belt use, self-restriction) â€¢Unsafe (e.g., substance use, cell phone use) â€¢Driving history â€¢Driving reduction â€¢Driving cessation â€¢Driving resumption â€¢Lifestyle Behavior & lifestyle â€¢Safe (e.g., seat belt use, self-restriction) â€¢Unsafe (e.g., substance use, cell phone use) â€¢Driving history â€¢Driving reduction â€¢Driving cessation â€¢Driving resumption â€¢Lifestyle Environment â€¢Economic â€¢Physical â€¢Social â€¢Services Environment â€¢Economic â€¢Physical â€¢Social â€¢Services Health Health Safe/Unsafe Driving Safe/Unsafe Driving Health education â€¢Screening/ assessment â€¢Intervention â€¢Counseling Health education â€¢Screening/ assessment â€¢Intervention â€¢Counseling Policy regulation organization HEALTH PROMOTION Body function & structure â€¢Physical condition â€¢Eye â€¢Heart â€¢Systemic â€¢Hearing â€¢Neurological â€¢Medication use â€¢Demographics Activities â€¢Falls â€¢Functional status â€¢ADLs/IADLs â€¢Driving impairment â€¢Disability Participation n = 9 (2%) n = 100 (20%) n = 3 (1%) n = 51 (10%) n = 313 (61%) n = 8 (2%) n = 0 (0%) N = 513 n = 17 (3%) n = 12 (2%) Figure B-1. Precede-Proceed model of health prom otion structural model from older driver systematic literature review. [Reprinted with permission from Classen, S., Ga rvan, C.W., Awadzi, K., Sundaram, S., Winter, S., Lopez, E.D.S. et al (2006). Systematic litera ture review and model for older driver safety. Topics in Geriatric Rehabilitation, 22, 2: 87-98. (Page 95 Figure 2). Lippincott Williams & Wilkins, Inc]
119 APPENDIX C INITIAL EXAMINATION OF VA RIABLES & RATIONALE FOR INCLUSION/EXCLUSION BASE D ON CONSENSUS MEETINGS Table C-1. Fatality Analysis Report Sy stem (FARS) accident level variablesa Element Levels Note PPMHP (yes/no) Codeb Variable # and page # in FARS manualc Case number state number Present in all level files Important for merging N/A N/A A1 Consecutive number Number assigned to forms No N/A A2 p. 24 Vehicle number Number assigned to vehicle No N/A A3 County N/A Physical environment 32 A6 p. 29 City N/A Physical environment 32 A6 p. 29 Accident date Month: 01 Day: 0 Physical environment 32 A8 p. 31 Accident time Physical environment 32 A9 p. 35 National Highway system 0, 1 levels; and 9 Interstate system, principal arterial system routes and strategic network connected Physical environment 32 A10 p. 37 Roadway function class 19 levels Rural: 1 Urban: 11 Physical environment 32 A11 p. 39 Route signing 9 levels Interstate, U.S. highway, state highway, county road. Physical environment 32 A12 p. 45 Traffic identifier *Consider leaving out Physical environment 32 A13 p. 49
120 Table C-1. Continued. Element Levels Note PPMHP (yes/no) Codeb Variable # and page # in FARS manualc Mile point Actual number *Consider leaving out Physical environment 32 A14 p. 53 Global position *Consider leaving out Geographic location of crashes. Expressed in degrees, minutes, and seconds of latitude, and same for longitude. Physical environment 32 A15 p. 55 Special jurisdiction 0 levels *Consider collapsing Where accident occurred in a special jurisdiction Physical environment 32 A16 p. 59 First harmful event About 60 levels, but grouped under: noncollision collision with motor vehicle collision with object not fixed Collision with object fixed. First property damage (including vehicle) or injuryproducing event Physical environment 32 A17 p. 61 Manner of collision 12 levels *Recode as: Front Rear Side Point of impact Physical environment 32 A18 p. 75 Relation to junction 20 levels grouped as (a) interchange, (b) noninterchange and (c) unknown *Recode as: Intersection Entry/exit Hazardous crossing Unknown Location of first harmful event Physical environment 32 A19 p.81
121 Table C-1. Continued. Element Levels Note PPMHP (yes/no) Codeb Variable # and page # in FARS manualc Relation to roadway 01, and 99 *Too detailed. Consider dropping Part of the road, e.g., on roadway, median, shoulder Physical environment 32 A20 p. 89 Trafficway flow 1 levels; and 9 *Recode as: Not divided Divided One-way Exit/entrance Unknown Road design Physical environment 32 A21 p. 95 Number of travel lanes 1 levels; and 9 *Consider dropping Number of travel lanes Physical environment 32 A22 p. 97 Speed limit Actual posted limit; 00 for no statutory limit, and 99 for unknown *Consider dropping Posted speed limit m/hour Physical environment 32 A23 p. 99 Road alignment 1 levels; and 9 Straight or curve Physical environment 32 A24 p.103 Roadway profile 1 levels; and 9 Road profile Physical environment 32 A25 p.105 Road surface type 1 levels; and 9 *Recode as: stable unstable unknown E.g., concrete, brick, slug, dirt Physical environment 32 A26 p.107 Road surface condition 1 levels; and 9 *Recode as: dry adverse other/unknown E.g., dry, wet, snow Physical environment 32 A27 p.109 Construction or maintenance zone 0 levels *Recode as: Yes No Physical environment 32 A28 p.111
122 Table C-1. Continued. Element Levels Note PPMHP (yes/no) Codeb Variable # and page # in FARS manualc Traffic control device Several levels. Can be classified as 1) no controls, 2) highway traffic signals 3) regulatory signs, 4) school zone signs, 5) warning signs, 6) miscellaneous not at railroad crossing; 7) at railroad grade crossing *Recode as: No Yes Physical environment 32 A29 p.113 Traffic control device functioning 0 levels; and 9 Physical environment 32 A30 p.119 Light condition 1 levels; and 9 *Recode as: light dark Physical environment 32 A31 p.121 Atmospheric conditions 1 levels; and 9 *Recode as: adverse non-adverse Physical environment 32 A32 p.123 Hit and run 1 levels Behavior 23 A33 p.125 Notification time EMS Time (military time) Social environment 33 A36 p. 133 Arrival time EMS Time (military time) Social environment 33 A37 p.137 EMS time at hospital Time (military time) Time EMS arrived with victims of accident Social environment 33 A38 p.141
123 Table C-1. Continued. Element Levels Note PPMHP (yes/no) Codeb Variable # and page # in FARS manualc Related factorsaccident level 1 levels; and 99 *Recode as: 1. Highway design 2. Environmental factors Extraneous Personal 3. Care network systems-Tertiary Information on events that may have contributed to crash Physical environment 32 A39 p.145 a School bus related & ra il grade crossing identifier were not included in this list b Codes derived from Older Driver Systematic Literature Review (Classen et al., 2005). c 2004 FARS Coding and Validation Manual *Suggestions
124 Table C-2. FARS vehicle level variablesa Element Levels Note PPMHP (yes/no) Code b Variable # and page #in FARS manualc Case numberstate number Present in all level files Important for merging N/A N/A V1 Consecutive number Number assigned to forms No N/A V2 Vehicle number Number assigned to vehicle No N/A V3 p.153 Number of occupants Actual number Coded for each vehicle involved in the accident Social environment 33 V4 p.155 Registration state 56 levels; 92 State in which vehicle was registered No 33 V5 p.159 Registered vehicle owner 0 levels; and 9 No 33 V6 p.163 Travel speed Actual miles per hour Behavior 21, 22 V15 p. 281 Vehicle maneuver 1 levels; 98 and 99 Lane changes Turning Adjustment to stimuli Parked vehicle Speed variations & stopping Going straight Other/unknown Driverâ€™s action or intended action prior to crash Behavior 32 V16 p.283
125 Table C-2. Continued. Element Levels Note PPMHP (yes/no) Code b Variable # and page #in FARS manualc Crash avoidance maneuver 0 levels No avoidance maneuver Braking Steering Other/steering and braking Not reported Action taken by driver to avoid crash Predisposing 41 V17 p.285 Rollover 0 levels Vehicle overturning during accident Physical environment 32 V18 p.287 Initial point of impact 0 levels Initial point that produced property damage or personal injury (part of vehicle) Physical environment 32 V20 p.291 Vehicle role 0 levels; and 9 Whether vehicle was stuck or did the striking Physical environment 32 V21 p.299 Sequence of events About 50 levels. Categorized as: Nonâ€“collision Collision with motor vehicle collision with objects not fixed Collision with fixed objects Physical environment 32 V32 p.343
126 Table C-2. Continued. Element Levels Note PPMHP (yes/no) Code b Variable # and page #in FARS manualc Most harmful event 1 levels; and 99 Categorized as: Nonâ€“collision Collision with motor vehicle collision with objects not fixed Collision with fixed objects Used when first harmful event is minor for a particular vehicle compared to some subsequent event. Otherwise they are both coded the same per vehicle Physical environment 32 V33 p.355 Related factorsâ€” accident level 1 levels; and 99 (similar to accidentâ€“ level variable) Recoded: 0 = none 1 = 19 (fault on vehicle) 2 = hit and run 3 = vehicle went airborne 4 = other vehicle contributory factors 99 = unknown Information on events that may have contributed to crash Physical environment V34 p.365 a vehicle make, model, body type, model year, vehi cle ID #, bus use, special use emergency use, jackknife, underside/override, exte nt of deformation, manner of l eaving scene, motor carrier ID #, vehicle configuration, vehicle trailing, number of axes, gross vehicle weight rating, cargo body type, and hazardous cargo were not included in this list. b Codes derived from Older Driver Systematic Literature Review (Classen et al., 2005). c 2004 FARS Coding and Validation Manual *Suggestions
127 Table C-3. FARS driver level variablesa Element Levels Note PPMHP (yes/no) Code Variable # and page # in FARS manualc Case number state number Present in all level files Important for merging N/A N/A D1 Consecutive number Number assigned to forms No N/A D2 Vehicle number Number assigned to vehicle No N/A D3 p. 375 Driver presence 1 levels; and 9 *Should not be applicable to our dataset No N/A D4 p. 377 License state 1 levels; 94 Either state where driver got license or residence state of driver Yes 63 D5 p. 381 Driver zip code Actual values N/A D6 p. 385 NON-CDL License type/status 0 levels Categorized as: License type Graduated driver licenses License status *Use license status variables 1 levels; and 9 License status Enabling 21 or 22 63 D7 p. 387 Commercial motor vehicle license status 0 levels; and 9 Status for driverâ€™s commercial driverâ€™s license Enabling 21 or 22 63 D8 p. 397
128 Table C-3. Continued Element Levels Note PPMHP (yes/no) Code Variable # and page # in FARS manualc Compliance with license endorsements 0 levels; and 9 Whether vehicle driven required endorsements and whether driver was in compliance Behavior 21 or 22 63 D9 p. 403 License compliance with class of vehicle 0 levels; and 8 Type of license possessed or not for vehicle driven Behavior 21 or 22 63 D10 p.407 Compliance with license restrictions 0 levels; and 9 Compliance with physical and imposed restrictions Behavior 21 or 22 63 D11 p.411 Driver height Actual feet Health D12 p.415 Driver weight Actual weight in pounds Health D13 p.417 Date of first and last accident, suspension conviction Month and year Actual month and year Behavior or reinforcing 23 or 52 D19 and D20 p. 425
129 Table C-3. Continued Violations charged 0 levels. Categorized as: reckless/hit and run type impairment offenses speedâ€“related offenses rules of the roadâ€“ traffic signs and signals rules of the roadâ€“ turning, yielding, signaling rules of the roadâ€“ wrong side, passing and following rules of the roadâ€“ land usage nonâ€“movingâ€“ license and registration violation equipment license, registration and other violations Behavior or reinforcing (need to find out whether 1) the information is duplicated 2) is this past violations or current? 23 or 52 D21 p. 429 Related factors-driver level 0 levels *need to recode Behavior health 9 22 D22 p. 435 a Previous other harmful MV convic tions not included in the list b Codes derived from Older Driver Systematic Literature Review (Classen et al., 2005). c 2004 FARS Coding and Validation Manual *Suggestions
130 Table C-4. FARS person level variablesa Element Levels Note PPMHP (yes/no) Code Variable # and page # in FARS manualc Case numberstate number Present in all level files Important for merging N/A N/A P1 Consecutive number Number assigned to forms No N/A P2 Vehicle number 00 Number assigned to vehicle No N/A P3 p. 453 Person number 01 Coded consecutively No N/A P4 p. 457 Age 0â€“up to one year 01 (actual age); and 97 65-75 76-85 86+ Health 20a P6 p. 459 Sex 1; and 9 Health 20g P7 p. 461 Person type 1 levels; and 9 Important! Will be used to select inclusion sample (only crashes with person type = 1 will be included) N/A N/A P8 p. 463 Seating position 0; and 99 N/A N/A P9 p. 473 Restraint system use 0 levels; and 99 Type of restraint used None used/NA Shoulder belt Lap belt Shoulder/lap Restraint unknown Unknown Behavior 21/22 P10 p. 479
131 Table C-4. Continued. Element Levels Note PPMHP (yes/no) Code Variable # and page # in FARS manualc Air bag availability 0 levels; and 99 Categorized as: Deployed No deployed Unknown if deployed Not available Record of airbag availability and deployment Behavior Physical environment 21/22 or 32 P11 p. 483 Police-reported alcohol involvement 0 level; 8 and 9 Collapse unknown/not reported Behavior 22 P16 p. 497 Police-reported other drug involvement 0 level; 8 and 9 Collapse unknown/not reported Behavior 22 P19 p. 511 Drug test results 0 Collapse: Type of drugs Behavior Health 9 or 22 P21 p. 519 Injury severity 0 levels; and 9 Collapse: No injury Injury Outcome 81 P22 p. 529 Taken to hospital or treatment facility 0; and 9 Physical environment 33 P23 p. 533 Died at scene/en route 0; 7 levels Outcome 81 P24 p. 535 Death date Month/date/year *Will help determine how long victim lived N/A N/A P25 p. 537 Death time Military time *Will help determine how long victim lived N/A N/A P26 p. 541
132 Table C-4. Continued Element Levels Note PPMHP (yes/no) Code Variable # and page # in FARS manualc Related factorsperson level 0 levels *Recode Add these variables 1 Behavior Health Environmental 22 9 32 P27 p. 543 a ejection, ejection path, extrication, non-motorist location, method of alcohol determination by police, alcohol test type/alcohol test results, method of other dr ug determination, death certificate number, and fatal injury at work variables not included in the list. b Codes derived from Older Driver Systematic Literature Review (Classen et al., 2005). c 2004 FARS Coding and Validation Manual *Suggestions
133 LIST OF REFERENCES American Automobile Association (2005) AAA Ro adwise Reviewâ€” A Tool to Help Seniors Drive Safely Longer. Retrieved May 25, 2006, from http://www.aaasouth.com/acs_news/roadwisereview.asp Awadzi, K., Classen, S., Garvan, C. W., & Komara giri, V. (2006). Determinants of older driver safety from a socio-ecological perspective. Topics in Geriatric Rehabilitation, 22 , 36. Baker, T. K., Falb, T., Voas, R., & Lacey, J. ( 2003). Older women drivers: Fatal crashes in good conditions . Journal of Safety Research, 34 , 399. Ball, Owsley, C., Stalvey, B.T., Roenker, D.L ., Sloane, M.E., & Graves, M. (1998). Driving avoidance and functional impa irment in older drivers. Accident Analysis and Prevention, 30, 3: 313-322. Bauer, M .J., Adler, G., Kuskowski, M. A. & Rottunda, S. (2003). The influence of age and gender on driving pattern s of older adults. Journal of Women & Aging, 15 , 3. Bedard, M., Guyatt, G. H., Stones, M. J., & Hirdes , J. P. (2002). The independent contribution of driver, crash, and vehicle characteristics to driver fatalities. Accident Analysis and Prevention, 34 : 717. Bedard, M., Meyers, J. R. (2004). The influence of passengers on older drivers involved in fatal crashes. Experimental Aging Research, 30 , 25: 205. Bedard, M., Stones, M. J., Guyatt, G. H., & Hirdes , J. P. (2001). Trafficâ€“related fatalities among older drivers and passengers: Past and future trends. The Gerontologist. 41, 751. Beer, M. H. (1997). Explicit crite ria for determining potentially inappropriate medications use by the elderly: An update. Archives of Internal Medicine. 157 :1825. Blincoe, L., Seay, A., Zaloshin ja, E., Miller, T., Romano, E., & Luchter, S., et al. (2002). The economic impact of motor vehicle crashes, 2000 . U.S. Department of Transportation. Washington, DC. 2002. Retrieved, January 20, 2006, from http://www.nhtsa.dot.gov/staticfiles/DOT /NHTSA/Communication%20&%20Consumer %20Information/Articles/Associ ated%20Files/EconomicImpact2000.pdf Braver, E. R., & Trempel, R. E. (2004). Are older drivers actually at higher risk of involvement in collision resulting in deat hs or nonâ€“fatal injuries am ong their passengers and other road users? Injury Prevention, 10, 27. Brink, S. G., Simons-Morton, B. G., & Zane, D. (1998). A hospital-based infant safety seat program for low-income families: Asse ssment of population needs and provider practices.
134 Burkhardt, J. E., Berger, A. M., Creedon, M., & McGavock, A. T. (1998). Mobility and Independence: Changes and cha llenges for older drivers. Bethesda, MD: Ecosemetrics. Carr, D.B. (1993). Assessing older driver s for physical and cognitive impairment. Geriatrics. 48 (5): 46, 51. Centers for Disease Control and Prevention (CDC ), National Center for Injury Prevention and Control (NCIPC). CDCâ€™s Unintentional Injury Activities04. Atlanta (GA): CDC, NCIPC; 2005. Chande, V. T., & Kimes, D. (1999). A health pr omotion intervention for families in a Medicaid managed care plan. Academic Emergency Medicine, 6 (8): 823. Chang, B. A., Ebel, B. E., & Rivara, F. P. (2002) . Child passenger safety: Po tential impact of the Washington State booster seat law on childcare centers. Injury Prevention, 8 : 284. Christoffel, T., Gallagher, S.S. (2006). Injury Prevention and Public Health: Practical Knowledge, Skills and Strategies . (2nd ed.). Sudbury, Mass: Jones and Bartlett Publishers. Classen, S. (2004). Public Health Model to Promote Safe Elderly Driving . CDC Public Health research: Health protection Research Initiative Mentored Research Scientist development Award (Kâ€“O1). Project #: 1 K01 CE0000497. Classen, S., Garvan, C. W., Awadzi, K., Sundaram, S., Winter, S., Lopez, E.D.S et al. (2006). Systematic literature review and struct ural model for older driver safety. Topics in Geriatrics Rehabilitation, 22, 87. Classen, S., Shechtman, O., Stephens, B., Davis, E., Bendixen, R., Belchior, P., et al. (2006). The impact of roadway intersection on driving performance of younger and senior adults: Preliminary results. Topics in Geriatrics Rehabilitation, 21, 2: 18-26. Coronado, V. G., Thomas, K. E., Sattin, R. W., & Johnson, R. E. (2005). The CDC traumatic brain injury surveillance system: Characteris tics of persons aged 65 years and older hospitalized with a TBI. Journal of Head Trauma Rehabilitation, 20, 215. Cowart, V., Kandela, P. (1985). Prescrip tion drugs and driving performance. JAMA. 254: 15. Cox, D.R. (1989). Analysis of Binary Data (2nd ed.) New York: Chapman and Hall. Dahl, T. H. (2002). International Classificati on of Functioning, Disabi lity and Health: An introduction and discussion of its potential imp act on rehabilitation services and research. Journal of Rehabilitation Medicine. 34, 201. Dellinger A. M., Langlois J. A., & Li G. (2002). Fatal crashes among older drivers: decomposition of rates into contributing factors. American Journal Epidemiology . 155, 234.
135 Department of Transportation (2004). 2004 FARS Coding and Validation Manual. National Highway Traffic Safety Administration. Department of Transportation (2002). FARS Analytic Reference Guide: 1975. National Highway Traffic Safety Administration. Department of Transportation (DOT) (2003a). Fatality Analysis Re porting System (FARS) 2003, [data file]. Available from National Highw ay Transportation Safety Administration Website, http://www.nhtsa.gov . Dobbs, B., & Carr, D. (2005). Screening and assessment of medically at-risk drivers. Public Policy & Aging Report, 15, 6. Eberhard, J. W., Stutts, J., Bur khardt, J., Finn, J., Hunt, L., Stap lin, L. et al. (2006). Strategies and tools to enable safe mobility for older adults. Topics in Geriatric Rehabilitation, 22, 3. Federal Highway Admi nistration. (FHWA) Older Driver Safety Facts & Statistics . Retrieved May 19, 2006, from http://safety.fhwa.dot.gov/olde r_driver/old er_facts.htm Federal Transit Administration ( 2006). Civil rights and accessibi lity. Retrieved October 1, 2006, from http://www.fta.dot.gov/civil_rights.html Finison, K. S., & Dubrow, R. B. (2002). A comp arison of Maine crashes involving older drivers. NHTSA: Washington, DC. DOT HS 809407. Foley, D.J., Heimovitz, H. K., Guralnik, J.M., & Brock, D.B. (2002). Drivi ng life expectancy of persons aged 70 years and older in the United States. American Journal of Public Health . 92, 8:1284. Freeman, E. E., Gange, S. J., Munoz, B., & West, S. K. (2006). Driving stat us and risk of entry into long-term care in older adults. American Journal of Public Health, 96, 7: 1254â€“ 1259. Freeman, E. E., Munoz, B., Turano, K.A., & West, S. K. (2006). Measures of visual fuction and their association with driving modification in older adults. Investigative Ophthalmology and Visual Science, 47 : 514. Gielen, A. C. (1992). Health education a nd injury control: Inte grating approaches. Health Education Quarterly. 19, 203. Gielen, A. C., & Sleet, D. (2003). Application of behavior-change theories and methods to injury prevention. Epidemiologic Reviews, 25: 65. Glanz, K., Rimer, B.K. (1995). Theory at a Glance: A guide for Health Promotion Practice . Bethesda, MD: National Cancer In stitute. Publication No. 95-3896.
136 Gordon, J. E. (1949). The epidemiology of accidents. American Journal of Public Health. 39 :504. Grabowski, D. C., Campbell, C. M., & Morrisey, M. A. (2004). Elderly licensure laws and motor vehicle fatalities. JAMA. 291 (23): 2840. Grabowski, D. C., & Morrisey, M. A. (2001). The effect of state regula tions on motor vehicle fatalities for younger and older driv ers: A review and analysis. The Milbank Quarterly, 79, 517. Green, L. W., & Kreuter, M. W. (2005). Health Promotion Planning, An Educational and Ecological Approach (4th ed.). Boston, MA: McGraw Hill Higher Education. Haddon, W., Jr. (1999). The changing appro ach to the epidemiology, prevention, and amelioration of trauma: The transition to approaches etiologically rather than descriptively based. Injury Prevention, 5 :231. Haddon, W., Jr. (1980). Advances in the epidemiology of injuries as a basis for public policy. Public Health Reports. 95 :411. Haddon, W., Jr. (1979). Options for the prev ention of motor vehi cle crash injury. Israel Journal of Medical Science. 16, 45. Haddon, W., Jr. (1973). Energy damage and the ten countermeasure strategies. Journal of Trauma. 13, 4:321. Haddon, W., Jr. (1972). A logical framework for categorizing hi ghway safety phenomena and activity. The Journal of Trauma. 12, 3:193. Haddon, W., Jr. (1963). A note concerni ng accident theory and research with special reference to motor vehicle accidents. Annals of the New York Ac ademy of Sciences. 107 : 635. Hammerlein, A., Derendorf, H., Lowentha l, D. T. (1998). Pharmacokinetic and pharmacodynamic changes in the elderly: Clinical implications. Clinical Pharmacokinetics, 35 : 49.). Haase, R. F., Ellis, M. V., & Ladany, N. (1989). Multiple criteria for evaluating magnitude of experimental effects. Journal of Counseling Psychology, 36 , 511-516 Hermanson, S. (2005). Older car owners : The use of their cars by others. AARP Public Policy Institute, 124: 1-5. Retrieved November 30, 2006, from http://assets.aarp.org/rgcenter/consume/dd124_cars.pdf Hill, J. D., & Boyle, L. N. (2006). The safety implications of vehicle seat adjustments. Journal of Safety Research. 37, 187.
137 Hing, J. Y. C., Stamatiadis, N., Aultman-Hall, L. (2003). Evaluating the impact of passengers on the safety of older drivers. Journal of Safety Research, 34: 343. Hosmer, D.W., & Lemeshow, S. (1989). Applied Logistic Regression. New York: John Wiley & Sons, Inc. Houser, A. (2005). Older drivers and automobile safety AARP Public Policy Institute. Pub ID: FS51R. Retrieved March 6, 2006, from http://www.aarp.org/research/housingmobility/transportation/fs51r_drivers.html Howat, P., Jones, S., Hall, M., Cross, D., & Stevenson, M. (1997). The PRECEDEâ€“PROCEED model: Application to planning a child pe destrian injury prevention program. Injury Prevention, 3 (4): 282. Jaccard, J. (2001). Interaction Effects in Logistic Regression. Thousand Oaks, CA: Sage Publications, Inc. Kahane, C.J. (1996). Fatality reduction by air bags: Analys es of accident data through early 1996. NHTSA Report No DOT HS 808 470. Retrieved November 30, 2006, from http://www.nhtsa.dot.gov/cars/rule s/regrev/evalu ate/808470.html Koulikov, E. (2005). Why older Americans stop driv ing: Effects of state licensing policies on reduction and cessation of driving. Dissertation Abstracts International , 66 (08), 3089, (UMI No. 3186479). Labonte, R., Polanyi, M., Muhajarine, N., & Willi ams, A. (2005). Beyond the divides: Toward critical health research. Critical Public Health , 15 , 1: 5-17. Last, J.M. (Ed.). (2001). A Dictionary of Epidemiology (4th ed.). NEW YORK: Oxford University Press. Levy, D.T., Vernick, J.S., Howard, K.A. (1995). Re lationship between driverâ€™s license renewal policies and fatal crashes involvi ng drivers 70 years and older. JAMA, 274:2840-2846. Li, G., Braver, E. R., & Chen, L. (2003). Frag ility versus excessive crash involvement as determinants of high death rates per vehi cle-mile of travel among older drivers. Accident Analysis and Prevention , 35 , 227. Lococo, K. H., & Staplin, L. (2006). Literature review of polypharmacy and older drivers: Identifying strategies to collect drug usag e and driving functioning among older drivers. NHTSA Report # DOT HS 810 558. Washington DC. Mahloch, J., Taylor, V., Taplin, S., & Urban, N. (1993). A breast cancer screening educational intervention targeting medical office staff. Health Education Research, (8), 4: 567.
138 Mann, K. V., Lindsay, E. A., Putnam, R. W., & Davis, D. A. (1997). Increasing physician involvement in cholesterol-lowering practi ces: The role of know ledge, attitudes and perceptions. Advances in Health Sciences Education, 2, (3): 237. Mann, K.V., & Putnam, R.W. (1989). Physiciansâ€™ perc eptions of their role in cardiovascular risk reduction. Preventive Medicine , 18 , 1: 45-58. Marottoli, R., Medes de Leon, C. F., Glass, T. A., Williams, C. S., Cooney, L. M., Berkman, L. F. (2000). Consequences of driving cessation: Decreased out of-hom e activity levels. Journal of Gerontology, 55B: S334â€“S340. Marottoli, R., Medes de Leon, C. F., Glass, T. A., Williams, C. S., Cooney, L. M., Berkman, L. F. et al. (1997). Driving cessation and increased depres sive symptoms: Prospective evidence from the New Haven EPESE. Journal of American Geriatrics Society, 45 : 202â€“ 206. Massie, D. L., & Campbell, K.L. (1993). Analysis of accident rates by age, gender, and time of day based on the 1980 Nationwide Personal Transportation Survey. University of Michigan Transportation Research Institute, Ann Arbor: MI. Massie, D. L., & Campbell, K.L. (1996). A Review of Out-of-Servi ces Criteria: Relationship between Accidents and Vehicle Defects. Revi ew of National Truck Accident Databases . University of Michigan Transportation Res earch Institute. Ann Arbor: MI. Retrieved October 22, from http://deepblue.lib.umich.edu/ bitstream/2027.42/1171/2/89147.0001.001.pdf Mausner, J., & Kramer, S. (1985). Epidemiologyâ€”An Introductory Text . Philadelphia, PA: W.B. Saunders Company. Molnar, L. J., & Eby, D. W. (2005). A brief look at driver license renewal policies in the United States. Public Policy & Aging Report, 15, 2, 13. National Academy on an Aging Society. Older driv er laws and resources. Retrieved September 30, 2006, from http://www.agingsociety.org/agingsocie ty/links/driverLi cense%20subpage.htm National Center for Health Statistics (2006). Health, United States, 2006: With Chartbook on Trends in the Health of Americans . Hyattsville, MD. National Highway Traffic Safety Administration. (NHTSA) (2006a). Traffic Safety Facts 2005. Older Population . Retrieved October 30, 2006, from http://www-nrd.nhtsa.dot.gov/pdf/nrd-30/ncsa/TSF2004/809910.pdf NHTSA (2006b). Traffic Safety Facts 2004. Older Population . Retrieved January 17, 2006, from http://wwwâ€“nrd.nhtsa.dot.gov/pdf/nrd/ NCSA/TSF2005/OlderPopulationTSF05.pdf NHTSA (2006c). DOT proposes anti-rollover te chnology for new vehicles . (Press release September 14, 2006). Retrieved September 27, 2006, from http://www.nhtsa.dot.gov
139 NHTSA. (2004). Traffic Safety Facts 2003. Older Population. Retrieved August 28, 2004, from http://www-nrd.nhtsa.dot.gov/ pdf/nrd-30/NCSA/TSF2003/809766.pdf NHTSA (2001). Traffic Safety Facts 2000. Older Population . Washington, DC. 2000 National Highway Traffic Safety Administration 2001. National Highway Traffic Safety Administration. (n.d.). Fatality Analysis Reporting System: Fatal Crash Data Overview. Retrieved June 1, 2006, from http://wwwnrd.nhtsa.dot.gov/pdf/nrd-30/ NCSA/FARS/809-726/index.htm Owsley, C., Ball, K., McGwin, G. Jr., Sloane, M. , Roenker, D. L., White, M., et al. (1998). Visual processing impairment and risk of motor vehicle crash among older adults. JAMA. 279 , 1083. Owsley, C., Ball, K., Sloane, M. E., Roenker, D. L., & Bruni, J. R. (1991). Visual/cognitive correlates of vehicle accidents in older drivers. Psychology and Aging, 6, 3: 403-415. Owsely, C., & McGwin, G., Jr. (1999). Vision impairment and driving. Survey of Ophthalmology, 43, 535. Pampel, F.C. (2000). Logistic Regression: A Primer. Thousand Oaks, CA: Sage Publications, Inc. Peek-Asa, C., Dean, B. B., & Halbert, R. J. ( 1998). Traffic-related inju ry hospitalizations among California elderly, 1994. Accident Analysis and Prevention, 30, 389. Ragland, D.R., Satariano, W.A., & Macleod, K. E. (2005). Driving cessation and increased depressive symptoms. Journals of Gerontology Series A: Biological Sciences and Medical Sciences , 60 : 399-403. Ray, W. A., Gurwitz, J., & Decker, M. D. (1992). Psychoactive drugs and the risk of injurious motor vehicle crashes in older drivers. Journal of American Epidemiology. 136:873. Ray, W. A., Thapa, P. B., & Shorr, R. I. (1993). Medications and the older driver. Clinics in Geriatric Medicine, 9 : 413. Richmond, T.S., Kauder, D., Strumpf, N., & Mere dith, T. (2002). Charac teristics and outcomes of serious traumatic injury in older adults. Journal of the American Geriatrics Society . 50, 2: 215-222. Ritter, A. S., Straight, A., & Evans, E. (2002). Understanding Senior Tran sportation: Report and Analysis of a Survey of Consumers Age 50+. Washington, DC: American Association of Retired Persons.
140 Rizzo, M., Anderson, S. W., Dawson, J., Myers, R., & Ball, K. (2000) . Visual attention impairments in Alzheimerâ€™s disease. Neurology. 54: 1954. Runyan, C. W. (1998). Using the Haddox ma trix: Introducing the third dimension. Injury Prevention. 4 : 302. Schieber F. (1994). High-p riority research and development ne eds for maintaining the safety and mobility of older drivers. Experimental Aging Research , 20 , 35. Shechtman, O., Classen, S., Stephens, B., Davis, E., Bendixen, R., Belchior, P., et al. (2006). The impact of roadway intersection on simulate d driving performance of younger and senior adults: Preliminary results. Topics in Geriatrics Rehabilitation, 21, 2: 27-25. Shields, J. (1997). Have we become so accustomed to being passive that weâ€™ve forgotten to be active? Injury Prevention, 3 : 243. Sleet, D. A. (1987). Health education appro aches to motor vehicle injury prevention. Public Health Reports. 102 (6) 606. Simons-Morton, B. G., Brink, S. G., Simons -Morton, D. G., McIntyre, R., Chapman, M., Longoria, J. et al. (1989). An ecological approach to the pr evention of injuries due to drinking and driving. Health Education Quarterly. 16 (3): 397. Solomon, D.H., Hashimoto, H., Daltroy, L., & Li ang, M.H. (1998). Techniques to improve physiciansâ€™ use of diagnostic test s: A new conceptual framework. JAMA, 280 : 20202027. Staplin, L., Lococo, K., Byi ngton, S., & Harkey, D. (2001). Guidelines and recommendations to accommodate older drivers and pedestrians. Publication No. FHWA-RD-01-051. US. Department of Transportation, Federal Highway Administration, Washington, D.C. Retrieved May 16, 2006, from http://www.dotcr.ost.dot.gov/documen ts/ycr/Older_DRivers_Peds.pdf Stevens, J.P. (2002). Applied Multivariate Statistics for the Social Sciences. (4th ed). Lawrence Mahwah, NJ: Lawrence. Erlbaum Publishers Sugg, N. K., Thompson, R. S., Thompson, D. C., Ma iuro, R., & Rivara, F. P. (1999). Domestic violence and primary care: Att itudes, practices, and beliefs. Achieves of Family Medicine, 8 : 301. Taylor, V. M., Schwartz, S. M., Jackosn, J. C., Kuniyuki, A., Fischer, M., Yasui, Y., Tu, S., & Thompson, B. (1999). Cervical cancer sc reening among Cambodian-American women. Cancer Epidemiology Biomarkers & Prevention, 8 , 541.
141 Taylor, V. M., Thompson, B., Montano, D. E ., Mahloch, J., Johnson, K., & Li, S. (1998). Mammography use among women attending an inner-city clinic. Journal of Cancer Education, 13, (2): 96. Trifiletti, L. B., Gielen, A. C., Sleet, D. A., & Hopkins, K. (2005). Behavioral and social theories and models: Are they used in unint entional injury prevention research? Health Education Research, 20, 298. Ustun, T. B., Chatterji, S., Bickenbach, J ., Kostanjsek, N., & Schneider, M. (2003). The International Classification of Functioning, Disability and Health: A new tool for understanding disabi lity and health. Disability and Rehabilitation. 25 (11): 565. Waller, J.A. (1985). Injury Control: A Guide to the Causes and Prevention of Trauma . Lexington, Mass: Lexington Books. Wan, H., Sengupta, M., Velkoff, V. A., & DeBa rros, K. A. U.S. Census Bureau, Current Population Reports, P23, 65+ in the United States: 2005, U.S. Government Printing Office, Washington, DC, 2005. Wang C. C., Kosinski C. J., Schwartzbe rg, J. G., & Shanklin, A. V. (2003). Physicianâ€™s Guide to Assessing and Counseling Older Drivers. Washington, DC: National Highway Traffic Safety Administration. Weinrich, S. P., Weinrich, M. C., Boyd, M. D., & Atkinson, C. (1998). The impact of prostate cancer knowledge on cancer screening. Oncology nursing forum, 25 (3): 527. World Health Organization (2004). World report on traffic injury prevention. Retrieved June 24, 2006, from http://www.who.int/world-health-day /2004/infomaterials/world_report/en/
142 BIOGRAPHICAL SKETCH Kezia Dzifa Awadzi was born in Ghana, West Africa. Her parents, Dr. and Mrs. Awadzi, encouraged their three children to develop a love for reading from an early age. Kezia graduated from the University of Ghan a in 1993 with a bachel orâ€™s degree in home economics and a minor in Food Science. Interested in a writing career, she participated in a national writing competition organized by Step Pub lishers, Ghana in 1994 and won first prizes in the True Life and Drama Categories. Ms. Awa dzi worked on the editorial board of Step Publishers for two years and left for graduate study in Mass Communication at the University of Florida in 1997. She graduated w ith a Masters in mass communication (journalism) in December 1999. Her thesis focused on the U.S. mediaâ€“infl uenced opinions about Africa. Ms. Awadziâ€™s experiences in developing a survey, data collectio n, and data analysis awakened an interest in research. Ms. Awadziâ€™s parents had careers in healthcare and Kezia wanted to do a program in a related field. In Fall 2001, she en rolled in the doctoral program H ealth Services Research in the department of Health Services Research, Manage ment and Policy, University of Florida. While in the doctoral program, Ms. Awadzi worked on proj ects within and outside of health services, including the Biological Control of Brazil ian Peppertree in Florida in the Department of Entomology and Nematology, University of Florida. In January 2005, Kezia had an opportunity to wo rk as a research assistant on the Centers for Disease Control and Prevention projectâ€”the Public Health Model to Promote Safe Elderly Driving. This job made her interested in older driver safety issues , provided exposure to secondary database analyses, and became the focus of her dissertation work. Kezia has published in Topics in Geriatri c Rehabilitation, Consumer Studies in Home Economics, and Environmental Entomology. She graduated in Fall 2006 .