DEVELOPMENT, RELIABILITY AND VALIDITY OF THE HEALTH RISK BEHAVIORS INVENTORY: A SELF-REPORT MEASURE OF 7 CURRENT HEALTH RISK BEHAVIORS

A dissertation submitted to Kent State University in partial fulfillment of the requirements for the degree of Doctor of Philosophy

by Leah A. Irish May, 2011

Dissertation written by Leah A. Irish B.S. University of Wisconsin – Whitewater, 2003 M.A., Kent State University, 2007 Ph.D., Kent State University, 2011

Approved by

_____________________________, Chair, Doctoral Dissertation Committee Douglas Delahanty _____________________________, Members, Doctoral Dissertation Committee Jeffrey Ciesla _____________________________, John Updegraff _____________________________, Carol Sedlak

Accepted by _____________________________, Chair, Department of Psychology Maria Zaragoza _____________________________, Dean, College of Arts and Sciences John R. Stalvey

ii

TABLE OF CONTENTS

List of Figures ......................................................................................................................v List of Tables ..................................................................................................................... vi CHAPTER 1

INTRODUCTION ...................................................................................................1 Behavioral Risk Factors for Physical and Mental Health .......................................2 Physical Activity and Health .............................................................................2 Diet and Health .................................................................................................3 Sleep and Health ...............................................................................................5 Cigarette Smoking and Health ..........................................................................6 Alcohol Use and Health ....................................................................................6 Illicit Drug Use and Health ................................................................................8 Sexual Behaviors and Health ............................................................................9 Summary of Health Risk Behaviors ...............................................................10 Health Behavior Assessments ...............................................................................11 Assessment of Physical Activity .....................................................................11 Assessment of Healthy Diet ............................................................................15 Assessment of Sleep .......................................................................................18 Assessment of Cigarette Smoking ..................................................................19 Assessment of Alcohol and Illicit Drug Use ...................................................22 Assessment of Sexual Behaviors ....................................................................24 Summary of Current Individual Health Behavior Assessments .....................26 Assessments of Multiple Health and Health Risk Behaviors .........................26 The Health Risk Behaviors Inventory (HRBI) ................................................32

2

PILOT HEALTH RISK BEHAVIORS INVENTORY ........................................34 Method ..................................................................................................................34 Participants ......................................................................................................34 Procedures ........................................................................................................35 Measures .........................................................................................................40 Data Analysis ..................................................................................................49

iii

Results ....................................................................................................................53 Exploring the Pilot HRBI.................................................................................53 Validity of the Pilot HRBI ...............................................................................72 Exploratory Factor Analysis ............................................................................87 Part I Interviews ..............................................................................................91 3

REVISED HEALTH RISK BEHAVIORS INVENTORY ................................101 Method .................................................................................................................101 Participants ....................................................................................................101 Procedures ......................................................................................................102 Measures .......................................................................................................103 Data Analysis ................................................................................................104 Results ..................................................................................................................105 Exploring the Revised HRBI .........................................................................105 Reliability of the Revised HRBI ....................................................................109 Validity of the Revised HRBI ........................................................................110 Exploratory Factor Analysis ..........................................................................122 Test-retest Reliability ....................................................................................125

4

DISCUSSION ......................................................................................................127

REFERENCES ................................................................................................................136 APPENDICES .................................................................................................................158 A

HEALTH RISK BEHAVIORS INVENTORY – PILOT VERSION..................159

B

HEALTH RISK BEHAVIORS INVENTORY – REVISED .............................164

C

MEASURES ........................................................................................................173

iv

LIST OF FIGURES

1

Scree plot from exploratory factor analysis of pilot HRBI ....................................88

2

Scree plot from exploratory factor analysis of Revised HRBI ............................123

v

LIST OF TABLES

1

Summary of Health Behavior Measures used to Establish Convergent Validity of HRBI Subscales ...................................................................................52

2

Descriptive Statistics and Internal Consistencies of the Pilot HRBI Subscale and Total Scores .....................................................................................54

3

Descriptive Statistics of the Pilot HRBI Physical Activity Subscale Items ...........55

4

Descriptive Statistics of the Pilot HRBI Diet Subscale Items ...............................56

5

Descriptive Statistics of the Pilot HRBI Sleep Subscale Items .............................57

6

Descriptive Statistics of the Pilot HRBI Smoking Subscale Items ........................58

7

Descriptive Statistics of the Pilot HRBI Alcohol Use Subscale Items ..................59

8

Descriptive Statistics of the Pilot HRBI Drug Use Subscale Items .......................60

9

Descriptive Statistics of the Pilot HRBI Sexual Behavior Subscale Items ............61

10

Descriptive Statistics of Health Behavior Measures Used to Examine Validity of the Pilot HRBI .....................................................................................62

11

Bivariate Correlations Examining Reliability and Validity of the Pilot HRBI Physical Activity Subscale Items ...........................................................................64

12

Bivariate Correlations Examining Reliability and Validity of the Pilot HRBI Diet Subscale Items ............................................................................65

13

Bivariate Correlations Examining Reliability and Validity of the Pilot HRBI Sleep Subscale Items ..........................................................................66

14

Bivariate Correlations Examining Reliability and Validity of the Pilot HRBI Smoking Subscale Items ........................................................................................67

15

Bivariate Correlations Examining Reliability and Validity of the Pilot HRBI Alcohol Use Subscale Items ..................................................................................68 vi

16

Bivariate Correlations Examining Reliability and Validity of the Pilot HRBI Drug Use Subscale Items ............................................................................69

17

Bivariate Correlations Examining the Reliability and Validity of the Pilot HRBI Sexual Behavior Subscale Items .........................................................70

18

Bivariate Correlations between SF-36 Health Outcomes and Pilot HRBI Physical Activity Subscale Items ................................................................73

19

Bivariate Correlations between SF-36 Health Outcomes and Pilot HRBI Physical Activity Subscale Items ................................................................74

20

Bivariate Correlations between SF-36 Health Outcomes and Pilot HRBI Sleep Subscale Items ...................................................................................75

21

Bivariate Correlations between SF-36 Health Outcomes and Pilot HRBI Smoking Subscale Items .............................................................................76

22

Bivariate Correlations between SF-36 Health Outcomes and Pilot HRBI Alcohol Use Subscale Items ........................................................................77

23

Bivariate Correlations between SF-36 Health Outcomes and Pilot HRBI Drug Use Subscale Items ............................................................................77

24

Bivariate Correlations between SF-36 Health Outcomes and Pilot HRBI Sexual Behavior Subscale Items .................................................................79

25

Summary of 7 Factor Solution for Exploratory Factor Analysis of the Pilot HRBI ...................................................................................................89

26

Descriptive Statistics and Internal Consistencies of the Revised HRBI Subscale and Total Scores ........................................................................106

27

Descriptive Statistics of the Revised HRBI Physical Activity Subscale Items .....................................................................................................106

28

Descriptive Statistics of the Revised HRBI Diet Subscale Items ........................107

29

Descriptive Statistics of the Revised HRBI Sleep Subscale Items ......................107 vii

30

Descriptive Statistics of the Revised HRBI Smoking Subscale Items ................107

31

Descriptive Statistics of the Revised HRBI Alcohol Use Subscale Items ..........108

32

Descriptive Statistics of the Revised HRBI Drug Use Subscale Items ...............108

33

Descriptive Statistics of the Revised HRBI Sexual Behavior Subscale Items ....108

34

Descriptive Statistics of Health Behavior Measures Used to Examine the Validity of the Revised HRBI ........................................................................111

35

Bivariate Correlations Examining Reliability and Validity of the Revised HRBI Physical Activity Subscale Items ................................................112

36

Bivariate Correlations Examining Reliability and Validity of the Revised HRBI Diet Subscale Items .....................................................................112

37

Bivariate Correlations Examining Reliability and Validity of the Revised HRBI Sleep Subscale Items ...................................................................113

38

Bivariate Correlations Examining Reliability and Validity of the Revised HRBI Smoking Subscale Items..............................................................114

39

Bivariate Correlations Examining Reliability and Validity of the Revised HRBI Alcohol Use Subscale Items ........................................................114

40

Bivariate Correlations Examining Reliability and Validity of the Revised HRBI Drug Use Subscale Items.............................................................114

41

Bivariate Correlations Examining the Reliability and Validity of the Revised HRBI Sexual Behavior Subscale Items .................................................115

42

Bivariate Correlations between SF-36 Health Outcomes and Revised HRBI Physical Activity Subscale Items ..............................................................118

43

Bivariate Correlations between SF-36 Health Outcomes and Revised HRBI Diet Subscale Items ...................................................................................118

viii

44

Bivariate Correlations between SF-36 Health Outcomes and Revised HRBI Sleep Subscale Items .................................................................................119

45

Bivariate Correlations between SF-36 Health Outcomes and Revised HRBI Smoking Subscale Items..............................................................119

46

Bivariate Correlations between SF-36 Health Outcomes and Revised HRBI Alcohol Use Subscale Items ........................................................120

47

Bivariate Correlations between SF-36 Health Outcomes and Revised HRBI Drug Use Subscale Items ..........................................................................120

48

Bivariate Correlations between SF-36 Health Outcomes and Revised HRBI Sexual Behavior Subscale Items ...............................................................121

49

Discriminant Validity of the Revised HRBI Subscales .......................................121

50

Summary of 5 Factor Solution for Exploratory Factor Analysis of the Revised HRBI ......................................................................................................124

51

Summary of 4 Factor Solution for Exploratory Factor Analysis of the Revised HRBI ......................................................................................................126

ix

CHAPTER 1

INTRODUCTION

The majority of deaths in the United States are caused by chronic illnesses such as heart disease, stroke and cancer (Kung, Hoyert, Xu & Murphy, 2008). It is suggested that much of the disability and mortality associated with these chronic illnesses is caused by modifiable health behaviors (Centers for Disease Control and Prevention (CDC), 2010a). Specifically, recent estimates suggest that approximately half of American deaths are caused by events linked to behavioral risk factors including risky sexual behaviors and illicit drug use with the “big four” behavioral risk factors (physical inactivity, unhealthy diet, cigarette smoking, and excessive alcohol consumption) accounting for approximately 39% of all deaths (McGinnis & Foege, 1993; Mokdad, Marks, Stroup & Gerberding, 2003). Fine and colleagues (2004) examined the prevalence rates of these four health behaviors and found that American adults engage in an average of 1.68 of these unhealthy behaviors. Further, the impact of health behaviors on physical health is cumulative, such that illness risk increases as the number of unhealthy behaviors increases (Rosal et al., 2001). Additional research has highlighted the relationship between health risk behaviors and poor mental health (e.g. Rheingold, Acierno & Resnic, 2004).

1

2 Despite the clinical consequences of health risk behaviors, few attempts have been made toward comprehensively assessing health enhancing and/or health compromising behaviors. The purpose of the present study was to develop a self-report measure of 7 health risk behaviors: 1) physical inactivity, 2) unhealthy diet, 3) inadequate sleep, 4) cigarette smoking, 5) alcohol use, 6) illicit drug use, and 7) risky sex behaviors. Research has suggested that each of these behaviors is associated with negative physical and mental health consequences.

Behavioral Risk Factors for Physical and Mental Health

Physical Activity and Health Despite the fact that approximately 400,000 (16.6%) deaths are caused by poor diet and exercise each year (Mokdad et al., 2004), many Americans continue to lead sedentary lifestyles. Recommendations for physical activity vary somewhat and therefore so do estimates of physical activity. The Healthy People 2010 initiative suggests that adults should participate in moderate activity at least 30 minutes a day for 5 days per week or vigorous activity at least 20 minutes a day for 3 days per week (U.S. Department of Health and Human Services, 2000).

The U.S. Department of Health and

Human Services (2008) released updated guidelines for physical activity suggesting that adults engage in at least 150 minutes of moderate aerobic activity (e.g. brisk walking, bicycling, doubles tennis) per week or 75 minutes of vigorous aerobic activity (e.g. jogging, swimming laps, singles tennis) per week. In addition, adults should participate in muscle strengthening activities (e.g. weight or resistance training, carrying heavy

3 loads, heavy gardening) at least 2 days per week. It was noted that even small amounts of physical activity are better than sedentary lifestyles and was suggested that the health benefits of physical activity increase as the amount of physical activity increases (U.S. Department of Health and Human Services, 2008). A national survey found that, in 2007, 64.5% of American adults met the 2008 guidelines for physical activity, while only 48.8% of the same sample met the recommendations of Healthy People 2010 (Carlson, Fulton, Galuska & Kung, 2008). Regular physical activity has been associated with decreased risk for a number of physical health problems, such as cardiovascular disease (for reviews, see Kohl, 2001; Myers, 2000) and several types of cancer (for reviews, see Batty & Thune, 2000; Thune & Furberg, 2001). Muscle strengthening physical activity is particularly protective against bone density loss (Layne & Nelson, 1999). Further, a sedentary lifestyle is a risk factor for all-cause mortality (Lee & Skerrett, 2001). Physical activity may also have benefits for mental health. Research has found a negative association between physical activity and depression, with some researchers proposing exercise as an alternative treatment for depression (for a review, see Martinsen & Morgan, 1997). Some research has suggested a similar link between physical activity and anxiety (see Raglin, 1997), although this relationship is less clear (e.g. Paluska & Schwenk, 2000).

Diet and Health Dietary Recommendations for Americans 2005 (U.S. Department of Agriculture, 2005) provided suggestions for quantity and types of foods consumed in order to

4 maintain and promote health. Specifically, it was suggested that each day, adults consume 5 servings of fruits and vegetables (2 cups of fruit and 2 1/2 cups of vegetables), 5-8 ounces of grains (particularly whole grains), 3 cups of low-fat milk or dairy products, 6 ounces of protein (preferably lean meats, beans and nuts), and a limited amount of fats and added salts or sugars (U.S. Department of Agriculture, 2005). The relationship between diet and health is quite complex because of the variety of health risks and benefits of a variety of foods. For example, a high-fat diet (e.g. fried foods, fast food, high-fat animal products) may be harmful to cardiovascular health by increasing levels of LDL cholesterol, but also by increasing likelihood of obesity which is a risk factor for a number of health problems (Fung et al., 2001; Must et al., 1999; Stampfer, Hu, Manson, Rimm & Willett, 2000). On the other hand, eating fruits and vegetables has been associated with decreased risk for heart disease, stroke and certain types of cancer (Liu, 2003). Similar health benefits have been reported for diets rich in fiber, including beans, whole grains, and certain fruits and vegetables (Anderson, Smith & Gustafson, 1994). In addition, consumption of processed sugar (e.g. sugar-sweetened beverages, candy, sugarrich desserts) has been associated with increased risk for obesity and Type II Diabetes (Schulze et al., 2004). During times of psychological stress, individuals increase consumption of typically unhealthy “comfort foods”, which result in short-term satisfaction, but have implications for long-term physical health consequences (Dallman, Pecoraro & la Fleur, 2005).

5 Sleep and Health Health behavior researchers are becoming increasingly interested in the relationship between sleep and health. The National Sleep Foundation (2009) recommends that adults sleep for at least 7 hours per night, though they acknowledge that some individual differences exist in optimal sleep quantity. Estimates suggest that only 29.6% of American adults perceive zero nights of insufficient sleep over the past month, with 10.1% reporting having insufficient sleep every night for the past month (McKnightEily et al., 2008). Insufficient sleep has been associated with poorer self-reported physical and mental health, even after controlling for the presence of mood and anxiety disorders (Leger, Scheuermaier, Phillip, Paillard & Guilleminault, 2001). Sleep quality is of equal importance to health, with issues such as insomnia and sleep-disordered breathing often disrupting sleep (Baldwin et al., 2010). Inadequate sleep has been posited to decrease health via dysregulation of the neuroendocrine, nervous, and immune systems, resulting in increased risk for cardiovascular disease, diabetes, infectious diseases, and depression (for reviews see Taylor, 2008; Zaharna & Guilleminault, 2010). Long sleep (sleeping more than 8 hours per night) seems to have similar negative health consequences (Youngstedt & Kripke, 2004). Further, a recent meta-analysis of all-cause mortality concluded that both short and long durations of sleep were associated with an increased risk of mortality (Cappuccio, D’Elia, Strazzullo & Miller, 2010).

6 Cigarette Smoking and Health Tobacco use has been identified as the deadliest behavior in the United States, with over 400,000 (18.1%) deaths in 2000 attributed to tobacco (Mokdad et al., 2004). Tobacco is primarily ingested by smoking cigarettes, and in 2008, 23.9% of Americans aged 12 and older reported smoking at least one cigarette in the past month (SAMHSA, 2009). Research has overwhelmingly supported a consistent relationship between cigarette smoking and several chronic diseases including cancer (particularly lung cancer), heart disease, and chronic pulmonary obstructive disease (for reviews, see Brannon & Feist, 2004; Fagerström, 2002). Research on passive smoking (i.e. secondhand smoke) has demonstrated similar health risks, although to a lesser extent (e.g. Eriksen, LeMeistre & Newell, 1988; He et al., 1999). Cigarette smoking is also positively associated with life stress (see Kassel, Stroud & Paronis, 2003) and psychological disorders (e.g. Feldner, Babson & Zvolensky, 2007; Jane- Llopis & Matytsina, 2006; Quattrocki, Baird & Yurgelun-Todd, 2000).

Alcohol Use and Health Drinking alcohol is a common practice for American adults, with over half of Americans (age 12 and up) reporting having had at least one alcoholic drink (defined as 1 glass of wine, 1 bottle of beer, 1 wine cooler, 1 shot of liquor, 1 mixed drink) in the past month (SAMHSA, 2009). Typically, light alcohol consumption is not associated with increased health problems (see Taylor, 2008). However, almost half of American drinkers (23.3% of all Americans older than 12) report at least 1 occasion of binge

7 drinking (having 5 or more drinks on a single occasion) in the past month, and 6.9% of Americans are classified as heavy drinkers, meaning they engaged in binge drinking on 5 or more occasions in the past month (SAMHSA, 2009). Because of gender differences in alcohol metabolism (Thomasson, 2002), many researchers are beginning to classify binge drinking as 5 or more drinks on a single occasion for males and 4 or more drinks on a single occasion for females (e.g. CDC, 2010b). Chronic heavy drinking is associated with a number of serious health problems including liver disease, cardiovascular disease, and neurological disorders (for reviews, see Brannon & Feist, 2004; Cargiulo, 2007). A recent meta-analysis has also suggested a relationship between heavy drinking and certain types of cancer, including cancer of the liver, oral cavity, stomach, and colon among others (Bagnardi, Blangiardo, LaVecchia & Corrao, 2001). Heavy drinking and substance use disorders are also associated with mood, anxiety, and other psychological disorders (for reviews, see Cargiulo, 2007; JaneLlopis & Matytsina, 2006). According to estimates from the National Survey on Drug Use and Health, 11.1% of adults who engaged in heavy drinking in the past month also suffered from a DSM-IV psychological disorder (Epstein, Barker, Vorburger & Murtha, 2004). Approximately 3.2 million Americans reported comorbid substance abuse and psychological disorders in 2002 (Epsteain et al., 2004). It has been widely observed that individuals often use alcohol as a method of “self-medication” to cope with symptoms of psychological distress (e.g. Leeies, Pagura, Sareen & Bolton, 2001; Robinson, Sareen, Cox & Bolton, 2009). In addition, heavy drinking may disrupt neurological and

8 cognitive functioning (e.g. Cargiulo, 2007), which may, in turn, increase vulnerability to mental illness.

Illicit Drug Use and Health For purposes of the proposed measure, illicit drugs are defined as psychoactive substances (with the exception of tobacco or alcohol) used illegally or inappropriately (i.e. abusing prescription drugs) that can cause physical or psychological dependence (Sarafino, 2008). In 2008, 19.6% of young adults (ages 18-25) and 5.9% of adults aged 26 and over reported using illicit drugs in the past month, with marijuana and prescription drugs (used non-medically) reported most frequently by both groups (SAMHSA, 2009). Because illicit drug use is less common than use of alcohol and tobacco, less is known about the physical health consequences of illicit drug use. However, evidence suggests relationships between certain illicit drug use and several chronic diseases. For example, smoking marijuana has been associated with similar health consequences to smoking tobacco, including respiratory disease and lung cancer (Moore, Augustson, Moser & Budney, 2005; for a review see Hall & Degenhardt, 2009). In addition, cocaine use (which causes constriction of blood vessels and increased heart rate) has been associated with increased risk for cardiovascular disease and stroke (American Medical Association, 2003; Mittleman et al., 1999). Further, studies have demonstrated that illicit drug use is highly comorbid with psychological disorders, although a causal pathway is likely bidirectional (for reviews, see Jane-Llopis & Matytsina, 2006; Moore et al., 2007).

9 Sexual Behaviors and Health Risky sexual behaviors (e.g. unsafe sex, sex with anonymous partners, sex with multiple partners (either direct or indirect) were implicated as the primary cause of 20,000 American deaths (0.8% of all deaths) in 2000 (Mokdad et al., 2004). Sexual risk taking is not as common as some health risk behaviors; approximately 75% of sexually active participants report only one sexual partner at present (e.g. Finer, Darroch & Singh, 1999; Gullette & Lyons, 2005). However, the physical and mental health consequences of risky sex behaviors represent a substantial public health concern. One of the most common physical health consequences of risky sex behaviors is infection with a sexually transmitted disease (STD; Finer et al., 1999). According to the CDC’s 2008 surveillance of chlamydia, gonorrhea, and syphilis, approximately 1.5 million people reported having at least one of these three diseases (CDC, 2009), with some subgroups at particularly high risk (e.g. African-Americans, adolescents and young adults). Further, untreated STDs may cause additional medical issues including chronic pelvic pain, pelvic inflammatory disease and infertility, especially in women (Hook & Handsfield, 2008). Much of the research on risky sex behaviors and physical health has focused on risk for infection with the Human Immunodeficiency Virus (HIV). In 2008, HIV infections were reported by over 50,000 American adolescents and adults, with approximately 80% of infections resulting from sexual behavior (CDC, 2010c). Over time, people with HIV suffer from the health consequences of a weakened immune system and may enter the most severe stage of infection classified as acquired immune

10 deficiency syndrome (AIDS). Although medical treatments have been developed and improved to treat the infection, a cure has not been identified and approximately 75% of the deaths of infected individuals are caused by HIV-related factors (Sackoff, Hanna, Pfeiffer & Torian, 2006). Accordingly, substantial efforts have been made to develop and implement interventions to decrease risky sexual behavior and other risk factors in order to reduce new HIV infections (see Brannon & Feist, 2004; Taylor, 2008). Risky sexual behavior has also been associated with several common psychological disorders including depression, schizophrenia and antisocial disorders, with psychological comorbidity associated with even higher rates of sexual risk behaviors (Ramrakha, Caspi, Dickson, Moffitt & Paul, 2000). However, as with many health risk behaviors, it is difficult to determine the direction of causal relationships. For example, early psychological trauma, such as childhood sexual abuse, may lead victims to engage in risky sex as adolescents and adults. These later sexual experiences may evoke feelings of anxiety or posttraumatic stress, further increasing the likelihood of risky behaviors (e.g. not communicating with partner about safe sex because of anxiety) and also exacerbating psychological symptoms (see Rheingold et al., 2004). Overall, sexual risk behaviors have demonstrated direct and indirect influences on physical and mental health.

Summary of Health Risk Behaviors In sum, a number of behavioral factors have been implicated in the development and maintenance of physical and mental health problems. Consequently, a number of assessments have been developed to measure the current and cumulative prevalence of

11 these health behaviors for use in medical, psychological and intervention research. However, few efforts have been made to develop a standardized, practical and effective tool to measure multiple health risk behaviors, with the vast majority of health behavior assessments measuring only a single health behavior (Babor, Sciamanna & Pronk, 2004).

Health Behavior Assessments A wide variety of screeners, questionnaires, and interviews have been developed to measure a number of health behaviors in adults in the general population. Common themes appear in measures of each individual behavior, highlighting the most relevant aspects of the behavior as it relates to physical and mental health. Despite these commonalities, the length, complexity, content, scoring, and cost of these measures vary greatly depending on the intended purpose and context of use for each instrument. As a result, several excellent assessments are in place for specific areas of research. However, current assessment methods do not adequately address all research aims, creating a need for the development and testing of new instruments. This section will review the content of existing assessment methods for our target health risk behaviors, which helped to identify core concepts that should be included in the creation of the proposed measure.

Assessment of Physical Activity Many self-report instruments have been created to evaluate frequency and/or intensity of physical activity in adults (for reviews, see Babor et al., 2004; Lamb & Brodie, 1990; Laporte, Montoye & Caspersen, 1985; Oliver, Badland, Mavoa, Duncan & Duncan, 2010; Pereira et al., 1997). However, the content of these measures differs

12 widely depending on the target activities. Several focus on a single context or type of activity (e.g. occupation, leisure, sports, aerobic exercise). For example, the Tecumseh Occupational Physical Activity Questionnaire (Montoye, 1971) is a self-report questionnaire that asks about physical activity performed for one’s job(s). Information is collected regarding the participants’ work history (e.g. hours worked per week, months worked per year). In addition, participants report the frequency of travel methods during commute (e.g. walk, bicycle, drive to work), several activities during work (e.g. sitting, standing, heavy lifting) and climbing stairs or ladders during work. Each type of workbased physical activity is weighted according to the intensity (i.e. more intense activities such as walking to work are weighted more heavily than driving to work) and frequency of the activity. These weighted values are summed to provide subscale and overall scores of occupational physical activity (Montoye, 1971). Several other self-report measures focus specifically on leisure activities. The Minnesota Leisure-time Physical Activity Questionnaire (Taylor et al., 1978) is a commonly used interview which lists 63 sports (e.g. basketball, golf), recreational activities (e.g. hiking, sailing), household chores (e.g. gardening, carpentry) and fitness activities (running, weight lifting) that require varying degrees of physical exertion, and asks participants to estimate the frequency of each item over the past year. Each item is assigned an intensity code and is classified as light, moderate, or heavy. Items are also weighted by frequency and summed to create light, moderate and heavy leisure-time activity subscale scores as well as a total. Although instruments like the Tecumseh Occupational Physical Activity Questionnaire and the Minnesota Leisure-Time Physical

13 Activity Questionnaire provide reliable and valid data, they do not provide an overall description of an individual’s physical activity because they are too narrowly focused on specific types of activity. To address this limitation, many of the self-report physical activity measures examine a combination of occupational, leisure-time and household activities in the general population (e.g. International Physical Activity Questionnaire: Booth, 2000; EPIC Physical Activity Questionnaire: Pols et al., 1997). The classification and/or assessment of frequency and intensity are common themes throughout the majority of these instruments. Further, almost all self-report physical activity measures result in a continuous total and/or subscale score, though the metric used is often not comparable across measures. Despite these commonalities, physical activity measures differ widely in other characteristics such as number of items and timeframe being assessed. Overall, standard physical activity questionnaires emphasize frequency and intensity of activity in a variety of contexts. Though the content of physical activity measures is relatively consistent, it must be acknowledged that a number of assumptions are made in the administration of such instruments, arguably more than with other health behaviors. Specifically, researchers assume that their participants interpret constructs such as “vigorous activity” in the way they are intended. Further, researchers assume that participants can fit their own personal activities into the categories laid out by the questionnaire. Often, researchers attempt to assist participants in this task by providing a list of examples (i.e. vigorous activity (e.g. running, soccer, singles tennis), which typically includes only a small number of activities. Altschuler and colleagues (2009) utilized participant interviews to examine

14 participant comprehension of these common techniques used by physical activity assessments. Results revealed a number of discrepancies between the questionnaire assumptions and participant comprehension. First, a number of problems arose with the definition of “intensity” (e.g. vigorous, moderate). Many participants reported uncertainty about these terms, even when increased heart rate and respiration were used to help define them. Further, 17.5% of participants interpreted intensity to mean emotional or psychological intensity, rather than physical exertion. The authors concluded that the use of terms such as “vigorous” or “moderate” to describe intensity may increase measurement error. In addition, they found that participants often identified sweating as an indicator of physical intensity, rather than (or in addition to) heart rate and respiration. Second, these interviews revealed that participants often over-reported certain activities by counting them multiple times, or as a result of misunderstanding. For example, participants pointed out that standing and walking were assessed with separate items, but you cannot walk without standing. Often when activities were measured in multiple contexts (e.g. walking to work, walking briskly for exercise, walking leisurely), participants had difficulty distinguishing between them, and reported the same activity in all categories. Third, many participants had issues with the common practice of listing example activities to help define a term (like vigorous activity). Some participants interpreted the examples to be exclusive lists of activities that met the definition, and others felt the list were too long or could not determine how to fit their own activities into the terms of the questionnaire. Finally, Altschuler and colleagues (2009) found that the large majority of participants did have an accurate understanding of the terms “walking”

15 vs. “hiking” and “jogging” vs. “running”, suggesting that these terms may be used effectively in measures of physical activity. The results of these interviews provide beneficial information for the improvement of current assessment methods and inform the creation of effective new measures.

Assessment of Healthy Diet Self-report dietary assessments are typically designed to determine the extent to which an individual or group complies with nutritional recommendations or guidelines (e.g. Dietary Guidelines for Americans: USDA, 2005). As a result, questions often center around 2 primary themes: adequacy (i.e. consuming the recommended amount of healthy foods/nutrients such as vegetables) and moderation (i.e. limiting consumption of unhealthy foods/nutrients such as saturated fats). Dietary data are often collected using a food diary or food frequency questionnaire, such as the Food Frequency Questionnaire – Short (FFQ-S: Osler & Heitmann, 1996). The FFQ-S lists several healthy and unhealthy foods (e.g. fish, vegetables, ice cream), and asks participants to report the frequency of consumption in the past month on a scale from 1 (never) to 5 (more than once daily). Scores are summed by food group to create subscale frequency scores (e.g. sweets; fruits and vegetables). The FFQ-S provides some general information about eating patterns, but does not allow for very specific investigation of the health quality of foods consumed. In addition, brief versions of food frequency questionnaires often tend to group different foods into single items in an attempt to abbreviate a long list of food options. However, this may be confusing to individuals who eat large quantities of one item, but not another,

16 that are grouped together (Thompson & Subar, 2008). A more detailed food diary which reports specific types of foods and portion sizes may be more informative, particularly if a scoring index is applied. For example, the Healthy Eating Index-2005 (HEI-2005: Guenther, Reedy, Krebs-Smith, Reeve & Basiotis, 2008) was designed to measure the extent to which dietary consumption complies with the Dietary Guidelines for Americans 2005 (USDA, 2005).

Points are awarded for both adequacy (i.e. partially or completely

meeting recommendations for consumption of fruits and/or fruit juice, vegetables, dark green and orange vegetables, grains, whole grains, milk products, meat and beans, and healthy oils like those found in fish and nuts) and moderation (i.e. limiting saturated fat, sodium and calories from saturated fats, alcoholic beverages, and added sugars). Higher HEI-2005 scores suggest a more healthy diet and better compliance to dietary guidelines (Guenther et al., 2008). The HEI-2005 provides an excellent method for quantifying the healthiness of one’s diet. However, this method is more time-consuming than a simple questionnaire because it requires detailed information dependent on participant recall. Further, in order to accurately assess an individual’s typical diet, foods must be recorded on a number of days, increasing participant burden. Several screeners and questionnaires have been created to assess specific types of food consumption. These measures may focus strictly on adequacy of certain healthy foods such as fruits and vegetables (e.g. Five a Day for Better Health Survey: Subar et al., 1995), the moderation of certain unhealthy items such as fat and cholesterol (e.g. Northwest Lipid Research Clinic Fat Intake Scale: Retzlaff, Dowdy, Walden, Bovbjerg & Knopp, 1997), or may assess an abbreviated combination such as the adequacy of fiber

17 and moderation of fat (DINE: Roe, Strong, Whiteside, Neil & Mant, 1994). While these instruments are beneficial in specific lines of research, they do not provide a good overall estimate of dietary health because of their narrow focus. Therefore, a number of more comprehensive, yet brief, dietary assessments have been created to evaluate multiple components of adequacy and moderation (for reviews, see Babor, 2004; Calfas, Zabinski & Rupp, 2000; Thompson & Subar, 2008). One such instrument is the PrimeScreen (Rifas-Shiman et al., 2001) which measures intake of healthy and unhealthy foods and vitamin supplements over the past year. Frequency of consumption of 18 food categories (e.g. deep fried foods, whole grain foods, processed meats) is assessed on a 5-point scale ranging from “less than once per week” to “twice or more per day”. Examples of foods were provided based on the most common food items reported in two large populationbased surveys. In addition, seven questions ask whether participants are taking nutrient/vitamin supplements (e.g. calcium). Items may be interpreted individually or weighted by frequency and scored per nutrient (Rifas-Shiman et al., 2001). The Rapid Eating and Activity Assessment for Patients (REAP: Gans et al., 2006) is similar to the PrimeScreen in that it includes several food frequency items rated on a Likert scale for consumption in an average week. However, the REAP contains additional behavioral components which are not often combined with food frequency questions, such as skipping breakfast and eating in restaurants. All items on the REAP are worded in a negative direction, such that higher scores on the REAP indicate less healthy diets (e.g. inadequate consumption of healthy foods, overeating unhealthy foods). Additional questionnaires have expanded on behavioral or decision-making factors of food choice

18 and preparation. For example, the Eating Patterns Questionnaire (Kristal, Shattuck & Henry, 1990; Shannon, Kristal, Curry & Beresford, 1997) asks participants about healthy diet-related behaviors such as avoiding the use of fat as spread or flavoring, or substituting high-fiber foods for low-fiber foods.

The wide range of available nutrition

assessments reflects the variety of research interests and approaches in the study of diet and health.

Assessment of Sleep Self-report assessments of sleep typically emphasize two primary themes: quantity and quality of sleep. Quantity of sleep is often measured quite simply. For example, the Pittsburgh Sleep Quality Index (PSQI: Buysse, Reynolds, Monk, Berman & Kupfer, 1989) asks participants to report the number of hours of actual sleep per night during the past month. Other measures apply a more complex formula to calculate quantity of sleep. For example, Baker and Driver (2004) asked participants to record their average bedtime and waking time in a sleep diary to estimate the number of hours in bed. Then, sleep onset latency (length of time it takes to fall asleep after going to bed) and length of time of sleep disturbance are subtracted from the hours in bed to produce an estimate of actual time spent sleeping. Several scales have been developed to examine more subjective aspects of sleep quality. For example, the Sleep Quality Scale (SQS: Yi, Shin & Shin, 2006) was recently developed as an all-inclusive assessment of specific aspects of sleep quality such as overall satisfaction with sleep and difficulties with falling asleep, staying asleep, and

19 waking up. The SQS includes 28 items scored using a 4-point Likert scale that assess the frequency of sleep behaviors and problems. Six subscales and an overall total sleep quality score can be calculated for this measure. Similarly, the PSQI (Buysse et al., 1989) includes 19 items which examine seven components of sleep quality as well as quantity. Instruments such as the SQS and PSQI also measure the consequences of poor sleep quality (e.g. daytime drowsiness, functional impairment). Additional measures specifically assess consequences of poor sleep including the Stanford Sleepiness Scale (SSS: Hoddes, Zarcone, Smythe, Phillips & Dement, 1973) and the Epworth Sleepiness Scale (ESS: Johns, 1991). These instruments emphasize the daytime consequences of inadequate sleep and the extent to which participants feel alert and rested. In general, the majority of self-report sleep assessments provide information on adequacy of sleep as well as subjective perceptions of sleep quality.

Assessment of Cigarette Smoking Generally, assessments of cigarette smoking are relatively brief and focus on frequency of lifetime and recent smoking. When determining cumulative risk, measures of lifetime smoking/tobacco use are particularly important, and several methods have been developed to assess lifetime use from relatively simple items (e.g. the 100 cigarette rule-of-thumb: see Brondy, Victor & Diermert, 2009) to more complex assessments (e.g. Lifetime Tobacco Use Questionnaire: Brigham et al., 2008). However, lifetime tobacco use is not a primary focus for many research endeavors which are more interested in current or recent behaviors. Although most self-report measures of cigarette smoking

20 contain some level of lifetime use assessment, many focus primarily on current or more recent smoking patterns. Current smoking status may be assessed with reasonable accuracy with a single question (e.g. “Which of the following best describes your smoking status? I’m a smoker, I smoke daily; I’m a smoker, I smoke occasionally; I’m an ex-smoker, I never smoke now; I’m a non-smoker, I have never smoked; Dickinson, Wiggers, Leeder & Sanson-Fisher, 1989). Over time, questionnaires have evolved to measure nicotine use in more depth (i.e. heaviness of smoking) than that obtained with a single question. An excellent example of this is the Fagerström Test for Nicotine Dependence (FTND: Heatherton, Kozlowski, Frecker & Fagerström, 1991) which is a modified form of the Fagerström Tolerance Questionnaire (FTQ: Fagerström, 1978). The FTND is a 6 item questionnaire designed to evaluate various indicators of current dependence on cigarettes (e.g. number of cigarettes per day, time until first cigarette of the day, smoking while ill). Each question provides 2-4 categorical response choices (e.g. How many cigarettes per day do you smoke? 10 or fewer, 11-20, 21-30 or 31 or more; How soon after you wake up do you smoke your first cigarette? within the first 5 minutes after waking, 6-30 minutes, 3160 minutes or more than 60 minutes after waking; Do you smoke if you are so ill that you are in bed most of the day? yes or no), with each response corresponding to a number of points (0-3) during scoring. Points are summed to create a total measure of severity of dependence on cigarettes. Although the FTND is widely used to identify current dependant smokers, it is not a valid or reliable assessment in samples of light smokers (e.g. Etter, Duc & Perneger, 1999), and is inappropriate for nonsmokers.

21 In 2002, the World Health Organization ASSIST Working Group developed an instrument to assess several dimensions of past and current use of cigarettes, alcohol and illicit drugs. The Alcohol Smoking and Substance Involvement Screening Test (ASSIST: WHO Working Group, 2002) is composed of 8 questions concerning each substance. Specifically, items assess lifetime use, frequency of current use (during the past 3 months), dependence, and problems as a result of use (e.g. social, work, financial). Response options are categorical (e.g. yes/no, yes, but not in the past 3 months, weekly over the past 3 months), and each option corresponds to a point value ranging from 0-4 depending on the number of options for that item. Therefore, the ASSIST is capable of identifying dependent smokers and evaluating the severity of use and its consequences, while also providing accurate assessment of light or nonsmokers (Humeniuk et al., 2008). Assessments of passive smoking, or secondhand smoke, are much less common than assessments of active smoking, but nonetheless important as passive smoking is considered a health risk (see above). Some instruments designed to measure environmental smoke exposure target specific environments such as smoke in the home (e.g. the 7-Day Household Smoking Questionnaire: Glasgow et al., 1998). One good example of a more general passive smoking assessment is the Avoidance of Environmental Tobacco Smoke (ETS) Scale (Martinelli, 1998). This measure is comprised of 10 items which ask about the participants’ behaviors in situations when they are exposed to ETS (e.g.”If I encounter a friend who is smoking, I will sit and talk with him/her while he/she is smoking”; “I do not allow people to smoke in my home”). Each item is answered on a scale from 1 (almost always true) to 4 (almost never true) and

22 the answers are summed to calculate a total score (after reverse coding some items). As would be expected, response patterns differ significantly between smokers (regular and social smokers) and nonsmokers (never smokers and past, but not current smokers). Glasgow and colleagues (1998) suggest that these differences help demonstrate the validity of the measure. In addition, it highlights the limitations of using this, or other smoking measures, in mixed samples of smokers and nonsmokers. Therefore, new general measures of smoking should attempt to be sensitive to all levels of smokers, and include both active and passive smoke exposure.

Assessment of Alcohol and Illicit Drug Use Assessments of the use and abuse of alcohol are often designed and administered independently of assessments of the use and abuse of illicit drugs. However, these two types of assessments tend to be very similar in both structure and content, and therefore will be reviewed together here, collectively referred to as substance use. Typically, selfreport substance use assessments are similar to tobacco use assessments in that they focus on amount of current and/or lifetime use, yet differ in that they tend to place more emphasis on the consequences of use. As a result, many substance use questionnaires focus on classifying or measuring degree of dependence and/or abuse, rather than simply assessing substance use behaviors in general. For example, the Drug Abuse Screening Test (DAST: Skinner, 1982) is a 28-item questionnaire which asks participants about drug-related problems or behaviors that have occurred during the past year (e.g. Have you ever been in trouble at work because of drug abuse?; Do you ever feel bad about

23 your drug abuse?; Do you abuse more than one drug at a time?). Participants respond yes or no to each question, and the yes’s are summed to produce a total score (after reverse coding some items). Instruments like the CAGE (Ewing, 1984) or the Alcohol Dependence Scale (Skinner & Horn, 1984) provide similar basic measures for alcohol dependence. In addition, some measures have been developed to briefly assess problems with both alcohol and illicit drug use (e.g. CAGE-AID: Brown & Rounds, 1995). Although these measures are effective in screening for substance abuse problems, they are not as effective as a simple measure of substance use in the general population. As a result, some measures have been developed with more sensitivity to nonproblem users. For example, the Alcohol Use Disorders Identification Test (AUDIT: Babor, Higgins-Biddie, Saunders & Monteiro, 2001) is a 10-item instrument with questions regarding alcohol use (e.g. How often do you have a drink containing alcohol?), alcohol abuse (e.g. How often do you have six or more drinks on one occasion?), and consequences of alcohol use (e.g. How often have you failed to do what was normally expected of you because of drinking?) over the past year. Participants respond on a Likert scale from 0-4, with labels varying somewhat by question, but overall larger numbers represent more problematic drinking. Items are summed to get a total score, which allows for a more variable distribution of scores for both problem and nonproblem drinkers. As previously described, the ASSIST (WHO Working Group, 2002) measures lifetime use, recent use (past 3 months), and problems related to the use of tobacco, alcohol, and several illicit drugs including cannabis, cocaine, stimulants, inhalants,

24 sedatives/hypnotics, hallucinogens, opioids, and other drugs (used non-medically). The ASSIST is preferable to many other substance use instruments because it provides continuous scores for use and abuse that are appropriate for the general population (not only problem users). Overall, while there are a number of instrument options for screening substance abuse and dependence, few quality instruments exist that are appropriate for more general behavioral research.

Assessment of Sexual Behaviors The assessment of risky sexual behaviors is an important part of HIV-risk and other health research. However, there are relatively few formal measures for this construct. Most often, researchers query participants about individual aspects of sexual behavior and use the responses as stand-alone variables, rather than creating a composite score (for discussion of risky sex measurement methodology, see Fendrich, Smith, Pollack & Mackey-Amiti, 2009; Weinhardt, Forsyth, Carey, Jarowski & Durant, 1998). Further, question formats and response options tend to vary both between and within studies, resulting in a methodologically disparate literature. In general, five main topics appear consistently in various combinations in risky sex research: number of partners, type of partners, frequency of sex, type of sex, and condom use. Number of partners is generally assessed as a single item resulting in a specific number of partners over a given time frame (e.g. Greenwood et al., 2001) or as a classification of having multiple sexual partners over a given time frame (e.g. Boeckeloo et al., 1994). The type of partner may be described in a variety of ways, each of which

25 confers some level of risk. For example, participants who report having sex only with a committed or regular partner would be considered lower risk compared to those having sex with an anonymous partner or a “one night stand” (e.g. Stall et al., 2001; Zenilman et al., 1995). Frequency of sex alone is not generally considered a risk variable. However, questions about frequency of sex are often combined with questions about type of partner (e.g. McLaws, Oldenburg, Ross & Cooper, 1990), which provides a more continuous measure of level or risk (i.e. more frequent sex with a high-risk partner confers more risk than infrequent sex with a high-risk partner). Type of sex is generally classified as oral, vaginal or anal, most often with separate items measuring experience of each type (e.g. Kalichman, Kelly & Stevenson, 1997; Padian, 1990). However, risk is not necessarily inferred from type of sex alone. Often, condom use is incorporated into questions about type of sex (e.g. Kalichman & Rompa, 2001) as well as other sexual risk variables such as frequency of sex (e.g. Cohen & Dent, 1992). Overall, these five characteristics of sexual behavior are the most commonly assessed, and are often presented in various combinations, though rarely used to form composite risk scores. Questions about sexual behavior most often refer to behaviors from a specific timeframe, rather than a lifetime assessment. Kauth and colleagues (1991) have suggested that such measures are reliable for more current timeframes (e.g. past 3 months), but not for longer timeframes (e.g. 1 year), despite the fact that existing measures often assess for longer time frames. Although recent progress has been made (e.g. Fendrich et al., 2009), the lack of standardized, empirically supported questionnaires appropriate for the general population prevents researchers from consistent and comparable results.

26 Summary of Current Individual Health Behavior Assessments In general, researchers have tended toward building substantial research programs around a single, relatively isolated health or health risk behavior. Accordingly, a substantial number of behavioral health assessments focus exclusively on a single category of behavior for use in health research. Within instruments of specific behavior categories, content has remained quite consistent, focusing on core themes of the behavior as it relates to health, such as frequency, intensity, quantity, quality, and other such constructs. However, as the particular focus varies from behavior to behavior and to a lesser degree, from measure to measure within the same behavioral category, the metrics used and resulting scores often differ substantially, making the meaningful combination of these assessments very difficult. The use of these individual assessments may provide very specific and detailed information about behaviors (e.g. How does your first cigarette of the morning make you feel? Do you go sledding?), but this level of precision is neither necessary nor desirable for more general lines of research. Therefore, continued efforts are necessary to develop health behavior assessments that can efficiently evaluate these key features of multiple health behaviors (e.g. frequency, quality) using a comparable metric which allows for examination of cumulative risk across health behaviors.

Assessments of Multiple Health and Health Risk Behaviors Although the large majority of health and health risk behavior assessments focus exclusively on a single type of behavior, there have been some attempts to create more

27 comprehensive measures of multiple behaviors relevant to health. As part of the Prescription for Health program, Glasgow and colleagues (2005) reviewed the existing measures of four health behaviors (physical activity, healthy diet, cigarette smoking and alcohol use) and provided recommendations on a set of questionnaires that could be used to identify at-risk individuals in primary care settings. They based their recommendations on several criteria including practicality (i.e. length), clinical and public health relevance, sensitivity to change, reliability, and validity (Glasgow et al., 2005). The authors recommended that health behavior assessment of adults in primary care should include 22 questions taken from 6 different published and unpublished measures (see Glasgow et al., 2005). These recommendations were later modified and examined in 5358 adult members of the Prescription for Health network (Fernald et al., 2008). Physical activity was assessed using items from the International Physical Activity Questionnaire (IPAQ: Craig et al., 2003) which asks participants to report on the number of days during the past week and the number of hours per day they engaged in vigorous, moderate and light physical activity. Definitions and examples of each level of activity were provided. Eating patterns were assessed using 7 items from Starting the Conversation (STC: Paxton, Ammerman, Gizlice, Johnston & Keyserling, 2007). These questions ask participants to report the frequency of certain types of foods consumed over the past week (e.g. junk food, regular soda, fruits and/or vegetables). Answer options vary from question to question, but all represent frequency values (e.g. 3-4, 5 or more, less than 1). Cigarette smoking was measured using 4 items from 3 different surveys (CDC, 2003;

28 Hughes et al., 2003; Ory, Jordan & Bazzarre, 2002) that asked whether or not cigarettes have been smoked over the past year, past month and past week. Finally, alcohol use was measured with 3 items from the Behavioral Risk Factor Surveillance System (BRFSS) survey (CDC, 2003) which asks participants about frequency and quantity of alcohol consumption over the past month. Information from each subscale was used to determine whether or not the individual should be classified as high-risk for that specific behavior, and the subscales were summed to calculate the total risky health behaviors (possible range=0-4) for each individual (Fernald et al., 2008). This assessment strategy highlights both the strengths and weaknesses of using this “mix and match” approach to measuring multiple health behaviors (e.g. see Babor et al., 2004). Clearly, this mix and match approach of using full or partial versions of well established and validated scales for each individual health behavior is appealing because of the wide variety and quality of measures available. The combination of questionnaires used by Fernald and colleagues (2008) allows for the efficient assessment of multiple major health risk behaviors, which is important in a primary care setting (Glasgow et al., 2005). However, from a psychological research perspective, this assessment strategy has a number of shortcomings. First, there are inconsistencies among the health behavior measures in terms of timeframe and answer options. For example, the alcohol subscale refers to drinking behaviors over the past month, the cigarette smoking subscale refers to multiple timepoints, and the physical activity and eating patterns subscales refer to the past 7 days only. This makes it difficult to identify patterns among the health behaviors or to investigate their temporal relationship with other variables. Further, the answer

29 choices vary greatly from subscale to subscale (e.g. 3-4 desserts per week, 100 cigarettes over lifetime, 2 drinks per day last month, etc.). For the purposes of Fernald and colleagues (2008), the variety of answer choices is not a problem because they are intended to produce a simple risk classification (high-risk vs. low-risk) for each health behavior. This is quick and efficient if the goal is to target high-risk patients for an intervention or treatment, but in many research contexts, a more continuous assessment is preferable. Measuring health behaviors on a continuum may be a more realistic reflection of the true state of health behaviors and allows for more detailed examination. Further, simple risk classification does not allow for addition or comparison of the subscales. Researchers may sum the number of high-risk classifications met, but cannot draw any further conclusions about differences between subscales such as intensity or severity. Overall, while a mix and match approach may be effective for use as a screener in medical settings, it is not ideal for all research purposes. Another approach to the measurement of multiple health behaviors has been to establish and assess a profile of healthy lifestyle factors, including multiple health behaviors. For example, Walker and colleagues (1987) created the Health Promoting Lifestyle Profile (HPLP), which is a modification and abbreviation of the 100 item Lifestyle and Health Habits Assessment (LHHA: Pender, 1982). The HPLP is a 48-item questionnaire designed to assess six health promoting dimensions by measuring frequency of health promoting perceptions (self-actualization, personal responsibility, and interpersonal support) and behaviors (exercise, nutrition, and sleep/stress management) on a scale from 1 (never) to 4 (routinely). High internal consistency and

30 two-week test-retest reliability were found (Walker, Sechrist & Pender, 1987); however, empirical validity for this measure was not established. Rather, Walker and colleagues (1987) consulted a panel of 4 nursing experts to determine convergent validity during the development of the HPLP. The measure is scored by averaging the items for each subscale, and then averaging the subscale means for an overall profile score. This method of scoring limits the range of possible values from 1-4, which provides more information than a simple dichotomous classification and allows for comparison among subscales, but is still a somewhat restricted range of values. Further, while the assessment of multiple health behaviors (exercise, nutrition, and sleep/stress management) is more comprehensive than assessment of a single health behavior, there are still a number of health-promoting behaviors (e.g. avoiding cigarettes, alcohol and illicit drugs) not included in this measure. Also, because this measure is over 20 years old, some individual items would benefit from updating so they are more in line with recent developments in current knowledge and practice. A more comprehensive questionnaire is the Health Behavior Schedule II (HBS-II: Frank, Heiby & Lee, 2007) which is a 209-item (when all items are applicable) selfreport instrument designed to evaluate compliance to 12 health-promoting practices including healthy diet, exercise, flossing, protecting skin from UV rays, wearing a seatbelt, safe sex, wearing a bike helmet, not smoking cigarettes, limiting alcohol, following medical prescriptions, breast self exams, and cervical cancer screening. While the HBS-II covers a broad range of important health behaviors, it was not specifically designed to examine frequency or severity of each behavior. A single item (“How

31 successful have you been at making this a habit?”) is used for each health behavior to determine whether the individual is in compliance with healthy recommendations. The remaining 197 items provide a wealth of information about the reasons for compliance or noncompliance of each behavior (e.g. “How fully do you understand why you should do this?”, “This behavior makes me feel good.”, “How much social support do you get to do this?”). In addition, the HBS-II asks about general and specific factors which may predict compliance to healthy practices (e.g. self-efficacy, access to health care, knowledge of healthy practices). Therefore, this measure provides a thorough assessment tool for researchers and practitioners attempting to modify individuals’ health practices. However, it is not an ideal measure for researchers more interested in a brief and specific assessment of current behaviors, independent of other cognitive and emotional factors. Arguably, the most comprehensive assessment of American health risk behaviors is the Behavioral Risk Factor Surveillance System (BRFSS) with monthly telephone assessments coordinated by the Centers for Disease Control and Prevention (CDC). The specific questions vary from state to state and year to year, but generally include assessment of current physical and mental health and wellness, access to healthcare, lifetime prevalence of specific health problems (e.g. heart disease, diabetes, asthma), and a variety of risk factors including exercise, tobacco and alcohol use, sexual behaviors, injury prevention (e.g. wearing seatbelts), preventative measures (e.g. cancer screening, using sun block), and sleep (CDC, 2010d). Reliability and validity of the BRFSS survey have been documented in several studies (e.g. Mokdad, Stroup & Giles, 2003; Nelson, Holtzman, Bolen, Stanwyk & Mack, 2001; Nelson, Powell-Griner, Town & Kovar,

32 2003). Each health behavior module is brief, typically ranging from 1-5 items per module, and the modules can be customized for specific research questions (CDC, 2010d). However, the BRFSS survey suffers from similar limitations as other health behavior assessments. First, the timeframe of the questions vary both within and between the modules (e.g. lifetime, past month, past 24 hours). Second, the wording and response options differ both within and between modules, making it difficult to compute module scores or compare scores between modules. Finally, the overall design of the survey does not allow for any kind of total health behavior score beyond a sum of dichotomous risk classifications.

The Health Risk Behaviors Inventory (HRBI) The aim of the present study was to fill a void in the existing collection of assessments by developing a brief, self-report questionnaire that evaluates a variety of health risk behaviors in the general adult population. In creating the Health Risk Behavior Inventory (HRBI), we had 4 primary goals. First, the measure should be easy to administer, clear, and relatively brief, taking respondents no more than 15 minutes to complete. Second, the HRBI should provide a general sense of an individual’s actual, recent behaviors. It is understood that some level of detail and precision must be sacrificed in order to obtain such a wide range of information. Therefore, the HRBI was not designed to label individuals as healthy/unhealthy or determine whether respondents meet minimum criteria for health recommendations. Rather, the HRBI allows researchers to identify general patterns and correlates of health risk behaviors and

33 examine change over time. Third, the HRBI should utilize a consistent scoring metric for all items and all subscales. This allows for the appropriate combination of items and scales to create meaningful summary or total scores. Finally, the HRBI should be flexible, so that it can be easily modified for use in various populations and easily changed as new research continues to enhance our understanding of behavioral factors in health. The present study documents the development of the HRBI, and examines its reliability and validity as it compares to existing health behavior assessments.

CHAPTER 2

PILOT HEALTH RISK BEHAVIORS INVENTORY

Method

Participants Participants of all parts of the present study were college students enrolled in psychology courses at Kent State University. The psychology department utilizes the Experiment Management System, which is a web-based interface designed and maintained by Sona Systems that facilitates online scheduling and data collection. This software allowed eligible students to participate in the online components of the present study from any location at any time within the data collection period. The software stores data until it is downloaded and deleted from the system by the researcher. The Sona System was also used to schedule appointments for interviews conducted as part of this study. Students were eligible for participation if they were at least 18 years of age. Students who participated in Part I were automatically excluded from participation in Parts II and III. Three hundred and eighty students completed the online portion of Part I. Forty participants were removed from data analyses (39 people had invalid data; 1 person had over 20% of their data missing), resulting in a final sample size of 340. The majority of

34

35

the sample was female (69.7%), Caucasian (85.9%), and in their freshman year (65.9%). Participants’ ages ranged from 18-44, with a mean age of 19.6 years (SD=3.38). Thirtynine participants (14 males and 25 females) also completed an in-person interview as part of Part I data collection. Of these, 84.6% were Caucasian and 15.4% were AfricanAmerican. The average age of interview participants was 20.7 years (SD=3.75).

Procedures All procedures for the present study were approved by the Kent State University Institutional Review Board.

Development of the pilot Health Risk Behaviors Inventory (HRBI). The HRBI is intended for use in the self-report measurement of 7 health risk behaviors (physical inactivity, unhealthy diet, insufficient sleep, direct and indirect exposure to cigarette smoke, alcohol consumption, illicit drug use, and risky sexual behaviors). The measure is divided into 7 subscales, each representing one of these health risk behaviors. In the pilot HRBI (see Appendix A), each subscale consisted of 9-12 items, written at a seventhgrade reading level. In order to identify appropriate content for the items, an extensive review was conducted of both existing health risk behavior measures and empirical research documenting the health impact of these behaviors. Common themes were identified for measures of each behavior, and emphasized in the writing of the HRBI. For example,

36

although nutrition scales vary widely in specific content and format, items of these scales typically centered around adequate consumption of healthy foods and moderation of less healthy foods and eating behaviors (see above). Therefore, the items in the diet subscale of the HRBI reflect these themes. In addition, when general guidelines were available to identify specific behavioral recommendations to improve health, these guidelines were incorporated into the HRBI items. For example, the official recommendations for adequate sleep are 7-9 hours per night (National Sleep Foundation, 2009). Therefore, the HRBI questions concerning inadequate sleep refer to sleeping less than 7 hours per night. Further, Kent State faculty members with expertise in health psychology were consulted to provide feedback on the content validity of the items and to make any recommendations for improvement of the pilot measure. It is necessary to acknowledge that not all health risk behaviors included in the pilot HRBI were completely or uniformly under the direct control of the participant, and some items may have been influenced by other individuals in the environment (e.g. sleep latency, secondhand smoke exposure, being confronted about alcohol or drug use, etc.). However, these items were included in the HRBI because there are some things that individuals could do to impact the risk, even if it could not be completely controlled. As a result, these items are very consistent with items of other validated health risk behavior measures currently in use. Further, although the HRBI is not necessarily intended for use as a predictor of health, research using a health behavior measure is likely interested in

37

behaviors that are related to health, and therefore removing items directly related to individual health would be counterproductive to the goals of the measure. Items for each subscale were written as statements (e.g. My work involved sitting for long periods). The instructions asked participants to rate the extent to which they felt each statement was true of their actual behavior over the past month on a 5-point Likert scale (1= Never True, 2=Rarely True, 3=Sometimes True, 4=Often True, 5= Always True). This statement-agreement format is often used in psychological questionnaires and allows for a more complex pattern of behaviors than simple yes/no answers. This also reduces concerns about restricted range and floor and ceiling effects because it allows participants to respond on a continuum. Reverse coded items were included to reduce participant boredom and fatigue and improve accuracy of response. It was anticipated that this would also assist in identification of individuals with inaccurate response sets (i.e. participants who answered all questions the same suggesting they were not actually reading the questions). Reverse coded items are indicated with an * in Appendix A. The time frame of one month was chosen for two primary reasons. First, because short-term health risk behaviors may be dramatically altered by life events (e.g. illness, family emergency, vacation), assessing behaviors over the course of one month allows for measuring behaviors following an event, suffers less retrospective reporting bias, and improves the general accuracy of the measure. Second, as the HRBI is intended for

38

identification of current life behaviors rather than life-long health habits, the one-month time frame allows researchers to identify current behavioral status rather than cumulative risk, and allows for the monitoring of behavioral change over time for longitudinal research. The pilot HRBI subscales were scored first by calculating the mean of items in each subscale (possible range for subscale score=1-5). Then, the 7 subscale scores were summed to compute a health risk behaviors total score (possible range=7-35). The measure was intended for optimal flexibility in a variety of research needs. For example, each subscale was independently validated, and therefore may be used as brief individual measures. Further, a subset of subscales may be used (e.g. measuring only physical activity and diet). In such cases, scoring of the measure will be the same, but values will not correspond with those described in this paper. This pilot version of the HRBI was utilized during Part I of data collection which informed further refinement of the measure.

Data collection. Part I of data collection involved two separate but concurrent components. All participants of Part I completed the online portion of this study for course credit. Participants gave consent, provided demographic information, and completed a series of health behavior questionnaires through the Sona System including the pilot version of the Health Risk Behaviors Inventory (HRBI), International Physical Activity Questionnaire (IPAQ: Booth, 2000), Rapid Eating and Activity Assessment for

39

Patients (REAP: Gans et al., 2006), Pittsburgh Sleep Quality Index (PSQI: Buysse et al.,1989), Avoidance of Environmental Tobacco Smoke (ETS) Scale (Martinelli, 1998), Alcohol, Smoking and Substance Involvement Screening Test (ASSIST: WHO ASSIST Working Group, 2002), and 5 items from the BRFSS survey (CDC, 1996, 2003). Characteristics of instruments used in validity analyses are displayed in Table 1. In addition, physical and mental health were assessed using the 36-Item Short Form Health Survey (SF-36: Ware & Sherbourne, 1992). A subset (n=39) of participants from the online component of Part I also completed an in-person interview for additional course credit. Participants were invited to sign up for an interview appointment if they were male, represented a racial minority, or reported engaging in health risk behaviors of low frequency in the sample (i.e. smoking, alcohol use, drug use, risky sexual behavior). During the interview, participants were asked to read the items of the pilot HRBI out loud, and to provide feedback on each HRBI item, including item clarity, length, and applicability as well as insight into their decision making process in choosing an answer for the question. Interviews were taperecorded and transcribed, and this qualitative information was used to further improve the HRBI.

40

Measures

Demographic information. Participants provided basic demographic information including gender, age, race, income, education level, employment, extracurricular involvement and other general information. In addition, participants were asked to respond to the question “Has there been anything going on in your life during the past month that has had a serious impact on your health behaviors or lifestyle? (For example, illness, family emergency, took a vacation, started a new diet, etc.)”

Concurrent validity. Concurrent validity of the HRBI was examined using the 36-Item Short Form Health Survey (SF-36: Ware & Sherbourne, 1992), a self-report measure of physical and mental health. The 36 items may be divided into 8 subscales: 1) limitations in physical functioning, 2) limitations in usual role functioning due to physical health, 3) limitations in usual role functioning due to emotional health, 4) social functioning, 5) vitality, 6) bodily pain, 7) mental health, and 8) general health perceptions. Each item includes 2-6 response options which are assigned values from 0100. Items are averaged to create subscale scores, with higher values indicating better health or function. The SF-36 has been widely used in health research, and its reliability and validity are well-documented (e.g. Brazier et al., 1992; Garratt, Ruta, Abdalla, Buckingham & Russell, 1993; Jenkinson, Wright & Coulter, 1994). Although the SF-36 was initially developed to assess patient outcomes, it has become a common instrument

41

for health assessment in a variety of samples, including college students (e.g. Roberts & Golding, 1999; Stewart-Brown et al., 2000; Waite & Tatchell, 2005). Cronbach’s alpha for the present sample reflected high itnternal consistency (Part I alpha=.93), with subscales ranging from .64 (social functioning) to .90 (limitations in role functioning due to emotional health).

Convergent validity measures. As many authors have noted (e.g. Babor et al., 2004; Pereira et al., 1997), the utility of health behavior assessments varies dramatically depending on the environment, purpose, and participants, which explains the wide variety of available health and health risk behavior assessments. Therefore, there is no clear selfreport “gold standard” to use for validation of new health risk behavior assessments such as the HRBI. The following criteria were used to select appropriate measures for validation in the present study. First, the measure had to assess a single health risk behavior that corresponds with a subscale of the HRBI. It was acceptable for measures to cover multiple health behaviors, but each individual behavior had to receive its own score in order to be used in analyses. Second, for practical purposes, the measure had to have a relatively small number of items. Third, the questionnaire had to assess behaviors, not diagnose behavioral problems (i.e. assess frequency of binge drinking rather than diagnose individuals as alcoholics). Fourth, the questionnaire had to demonstrate good reliability and validity in published research. Fifth, the questionnaire had to focus on behaviors during a recent time frame (i.e. in the past 3 months or more recently).

42

Because of the high number of criteria to consider in the selection of appropriate validation measures, requirement of a timeframe identical to the HRBI (past month) would have severely limited our ability to find suitable measures. Therefore, measures with a timeframe close to the past month (e.g. past 3 months, past 7 days) were considered acceptable if they met all other criteria, but measures with longer timeframes (e.g. past year, lifetime) were not. Based on these criteria, 6 measures were included in this study to assess convergent validity of the HRBI (summarized in Table 1). All materials are displayed in Appendix C. The International Physical Activity Questionnaire (IPAQ: Booth, 2000) was used to measure physical activity over the past week. The IPAQ consists of 27 items which make up 5 subscales. The two items that comprise the sitting subscale were omitted in the present study because they do not reflect a measure of activity. The remaining 4 subscales evaluated include 1) job-related physical activity, 2) transportation physical activity, 3) housework, house maintenance, and caring for family, and 4) recreation, sport, and leisure-time physical activity. Within each subscale, items ask participants to report the frequency of certain types of activities (e.g. vigorous activity, walking) in minutes per day and days per week. Physical activity scores are expressed as METminutes/week, which is calculated by multiplying the metabolic equivalent of the task (MET) for each activity by the number of minutes per day the activity was performed by number of days per week the activity was performed. The IPAQ specifies the values of

43

MET for each type of activity. MET-minutes/week scores are summed across all items to create a total physical activity score. Both long and short forms of the IPAQ have demonstrated good reliability and validity (e.g. Brown, Trost, Bauman, Mummery & Owen, 2004; Craig et al., 2003; Fogelholm et al., 2006; Hagstromer, Oja & Sjostrom, 2006). Internal consistency was relatively low in Part 1 of the present sample (alpha=.63), likely due to a very low internal consistency of the transportation subscale (alpha=.05). However, this apparent inconsistency is likely more of a practical statistical issue than a conceptual one. For example, if an individual’s primary means of transportation is driving a car, they are not likely to engage in much walking or biking for transportation purposes. This would result in low statistical internal consistency because the items would not correlate highly. Therefore these results appear plausible, and do not raise substantial concerns about the use of the IPAQ. The Rapid Eating and Activity Assessment for Patients (REAP: Gans et al., 2006) was used to evaluate healthy and unhealthy eating patterns. The REAP contains 25 items related to eating behaviors and foods consumed in a typical week. Specifically, 4 items ask participants about their adequacy of healthy food consumption (e.g. eating at least 2-3 servings of fruit per day), 10 items refer to moderation or avoidance of unhealthy foods (e.g. sugar-sweetened beverages, fried foods), 9 items refer to choosing healthier versions of foods (e.g. choosing low-fat dairy products like skim milk), and 2 items refer to other eating behaviors (skipping breakfast, eating in restaurants). For each item, participants

44

report whether they engage in each eating behavior usually/often, sometimes, or rarely/never during a typical week, or indicate that the item is not applicable to them. All items are phrased so that a response of usually/often would indicate the least healthy choice and rarely/never would indicate the most healthy choice. Evaluation of the REAP has revealed moderate to high correlations with the Healthy Eating Index-2005 and good 2-week test-retest reliability (r=.86; Gans et al., 2006). Cronbach’s alpha revealed good reliability for the present sample (alpha=.84). Three minor modifications were made to the REAP to make it more appropriate for the present study. First, all items were scored in the reverse of the original method in order to make it more comparable with the HRBI scoring (i.e. will be scored as usually/often=3, sometimes=2, and rarely/never=1 rather than usually/often=1, sometimes=2, rarely/never=3). Item scores were then averaged to create an overall score with a possible range of 1 (most healthy) to 3 (least healthy). Second, the timeframe of “a typical week” was replaced with “the past week”. In addition, 5 items about physical activity, food shopping, cooking, and special diets were not included in the present study as they were covered in other question or were not relevant to the HRBI. The Pittsburgh Sleep Quality Index (PSQI: Buysse eat al., 1989) is a widely used, 19-item questionnaire designed to evaluate self-reported sleep quality over the past month. Four items ask participants to write in information about their typical bedtime, waking time, sleep latency (time until fell asleep), and actual time spent sleeping. The

45

remaining 15 items ask participants to respond to questions about their sleep quality and disturbances on a scale from 0 to 3, with higher scores indicating more sleep-related problems. Seven component scores are created including subjective sleep quality, sleep latency, sleep duration, habitual sleep efficiency, sleep disturbances, use of sleep medications, and daytime dysfunction, which range from 0 (no difficulty) to 3 (severe difficulty). The component scores are summed to create a global PSQI score, with possible values ranging from 0-21. Reliability and validity of the PSQI has been demonstrated in a wide variety of samples (e.g. Buysse et al., 1989; Knutson, Rathouz, Yan, Liu & Lauderdale, 2006; Gentilli, Weiner, Kuchibhatla & Edinger, 1995; Stein, Chartier & Walker, 1993). Reliability of the scaled items of the PSQI was acceptable in the present sample as well (Cronbach’s alpha=.82). As described above, the Avoidance of Environmental Tobacco Smoke (ETS) Scale (Martinelli, 1998) consists of 10 statements that refer to individuals’ behaviors toward ETS. Items are scored from 1 (almost always true) to 4 (almost never true), and summed resulting in a total score with possible values from 10-40, with lower values indicating less exposure to ETS. The Avoidance of ETS Scale does not provide respondents with a timeframe with which to reference their answers (e.g. past year, past month). However, the wording of questions would make it awkward to impose a timeframe for this study. Therefore, participants were instructed to indicate the extent to which the statements are true for them recently. Adding the word “recently” to the

46

instructions was intended to help reduce potential confusion caused by changes in attitude or behaviors over time. Martinelli (1998) reported that the measure had good internal consistency in samples of undergraduates and community women. Further, the Avoidance of ETS Scale successfully differentiated smokers from non-smokers (Martinelli, 1998). Cronbach’s alpha was adequate (alpha=.76) of the present study. The Alcohol, Smoking and Substance Involvement Screening Test (ASSIST: WHO ASSIST Working Group, 2002) is an international assessment tool which measures use and abuse of alcohol, tobacco, and 8 other psychoactive drugs, as described above. A self-report form was used to examine the validity of the smoking, alcohol and illicit drug subscales of the HRBI. The ASSIST includes 8 questions which may be asked with regard to each of the 10 substances. The first question (Q1) asks about lifetime use (no=0, yes=1). If the answer to Q1 is no, the rest of the questions for this substance may be skipped, and participants may move on to the next substance scale. If the answer to Q1 is yes, participants are asked (Q2) how often they have used the substance over the past 3 months from 0 (never) to 4 (daily or almost daily). Additional questions examine current dependence (Q3) and substance-related problems (Q4 and 5) (e.g. legal or financial problems) using the same Likert scale. The remaining 3 items (Q6-8) assess lifetime or current dependence substance-related problems, and risk. Q8 (IV drug use) was asked only once as it is not specific to any particular substance. Several scores may be calculated that were examined during the validation of the HRBI (see Humeniuk et al.,

47

2008 for a description of scoring). The specific substance involvement (SSI) score is calculated by summing the responses for Q2-7 for each individual substance, with higher scores indicating greater involvement with that substance. A total substance involvement (TSI) score may also be calculated by summing the responses to Q1-8 for all substances excluding tobacco and alcohol. The current frequency of use for each substance is simply indicated by the score on Q2 for each substance. In addition, items may be summed which reflect dependence (Q1, 2, 3, 6, 7) or abuse (Q1, 2, 4, 5, 6). The ASSIST has demonstrated good test-retest reliability, though the interval was relatively brief (1-3 days; WHO ASSIST Working Group, 2002). Further, estimates of internal consistency were adequate (WHO ASSIST Working Group) and TSI and SSIs correlated significantly with several well-established measures of substance use (Humeniuk et al., 2008). In the present study, internal consistency was acceptable for the tobacco (alpha=.81) and alcohol (alpha=.70) modules. In Part I, Cronbach’s alpha for the remaining substance modules ranged from.58 (inhalants and cocaine) to .79 (sedatives), with all but the inhalants and cocaine modules above .70.

Examination of the combined substances

(excluding tobacco and alcohol) yielded a Cronbach’s alpha of .82. Five items were taken from the sexual behavior module of the BRFSS survey (CDC, 1996, 2003) to measure risky sexual behaviors including number of partners, condom use, exchange of sex for drugs or money, unprotected anal sex, and treatment for STDs. Each question was answered with either yes or no, with the low-risk response

48

coded as 0 and the high-risk response coded as 1. Items were summed to create a composite score ranging from 0-5. Items were modified somewhat to fit the present study. First, the timeframe assessed by the BRFSS questions ranged from the past week to the past 5 years. In order to create a more consistent timeframe of assessment and to fit with the goals of the present study, all items were reworded to refer to behaviors over the past month. Second, some minor changes were made to the wording of the questions. For example, the BRFSS survey listed several behaviors (unprotected anal sex, prostitution, STDs) and asked participants if they had engaged in any of those behaviors without asking them to indicate which one. For the present study, each question is an individual item. In addition, the question about number of sexual partners was originally worded as asking whether the participant had “only one partner”. However, this does not differentiate between people with no sexual partners (low risk) and people with more than one sexual partner (higher risk). The wording of this question was changed to assess whether participants had “more than one partner”. Finally, several items from the BRFSS sexual behavior module were omitted from present analyses because they were not appropriate for the purposes of the present study. The BRFSS survey was chosen because of its applicability to the general population regardless of age, gender, disease status or sexual orientation. In addition, the BRFSS in general has demonstrated acceptable reliability and validity (see above). Specific information about the sexual

49

behavior module of the BRFSS is limited, although, some studies have supported the consistency of the sexual behavior module (e.g. Hamilton & Morris, 2010).

Data Analysis Analyses for the present study aimed to accomplish 2 primary goals: 1) to examine the quality and utility of the HRBI as a whole, and 2) to examine the properties of individual items which could then inform further improvements to the HRBI. Therefore, parallel analyses were run to explore the normality, reliability, convergent validity and concurrent validity in both the subscale and HRBI sum scores as well as each individual item in the pilot version of the HRBI. Further, factor analyses were utilized to explore and confirm the factor structure of the pilot HRBI.

Exploring the data. Data were first explored to identify participants with invalid data and remove them from further analyses. Individual participants were examined to identify any blatant contradictions in their responses (e.g. reporting unemployment, but also reporting heavy lifting as part of their job; reporting never having tried alcohol, but also reporting that their use of alcohol interfered with their daily lives). Thirty-nine individuals were excluded from Part I analyses as a result of these inconsistencies. Further, data were explored to examine any patterns of missingness or individuals with excessive missing data. Overall, there were relatively few missing data, with only one participant removed from Part I for exceeding 20% missing data points. On average,

50

participants of Part I were missing only 1.3% of data points, and correlations revealed no meaningful patterns in overall missingness (e.g. sex, age, duration of online survey). In addition, the items of the pilot HRBI were explored to help identify items with high frequencies of missingness as that may indicate items which are confusing or difficult to answer. However, the pilot HRBI had extremely low rates of missing data with only two items missing in more than 1% of participants. These items referred to activity at “work”. After all recoding was applied, expectation maximization (EM) imputation was used to impute missing values. Descriptive statistics were conducted to examine the distribution of responses to the individual pilot HRBI items as well as the total and subscale scores for the pilot HRBI and other survey measures.

Reliability of the pilot HRBI. Internal consistency (Cronbach’s alpha) and itemtotal correlations were utilized to establish reliability coefficients for each pilot HRBI subscale. These analyses also helped identify items that did not fit well into the subscales.

Validity of the pilot HRBI. Analyses of validity focused on identifying problem items in the pilot HRBI. Concurrent validity was evaluated by examining the correlations between the pilot HRBI (individual items, subscales, and total score) and the subscale and total scores on the SF-36 (Ware & Sherbourne, 1992) to determine whether participants’ current health behaviors were related to current mental and physical health.

51

Convergent validity was estimated by calculating the correlation between the pilot HRBI subscale scores and scores on a well-established measure of that health risk behavior. For example, scores on the alcohol use HRBI subscale were correlated with scores on the ASSIST alcohol module (WHO ASSIST Working Group, 2002). Correlations were also calculated between individual pilot HRBI items and their corresponding measures. Strong correlations indicated good concurrent and convergent validity. For a summary of measures used in convergent validity analyses, see Table 1.

Exploratory Factor Analysis. An exploratory factor analysis was conducted to examine the factor structure of the items in the HRBI. This analysis helped to determine the subscale structure and to identify any items that did not fit clearly into a pilot HRBI subscale. Sample size estimates for this study were based on the minimum requirements for factor analysis. Several rules-of-thumb provide recommendations for the participants to items ratio, with suggestions ranging from ratios that are quite low (2:1: Kline, 1979) to ratios that are very conservative (20:1: Hair, Anderson, Tatham & Black, 1995). Considering the exploratory nature of Part I, a 5:1 rule-of-thumb was used, such that 340 participants were needed to accurately examine the pilot HRBI’s 68 items (Bryant & Yarnold, 1995).

52

Table 1 Summary of Health Behavior Measures used to Establish Convergent Validity of HRBI Subscales HRBI Subscale

Validity Measure

# of Items

Assessment Timeframe Past week

Range of Possible Scores unlimited

Physical activity

IPAQ

25

Diet

REAP

Sleep

Booth, 2000

25

Past week

1-3

Gans et al., 2006

PSQI

19

Past month

0-21

Buysse et al., 1989

Smoking

ASSIST – tobacco module

Up to 7

Past 3 months

0-23

WHO ASSIST Working Group (2002)

Smoking

Avoidance of ETS Scale

10

In generala

4-40

Martinelli (1998)

Alcohol use

ASSIST – alcohol module

Up to 7

Past 3 months

0-23

WHO ASSIST Working Group (2002)

Drug use

ASSIST – illicit drug modules

Up to 57

Past 3 months

0-184

WHO ASSIST Working Group (2002)

BRFSS survey

5

Past month

0-5

CDC, 1996, 2003

Sexual behavior

Note. aNo timeframe indicated by author.

Author

53

Results

Exploring the Pilot HRBI The individual items and the 7 subscales of the pilot HRBI were examined with simple descriptive statistics to identify any problem items. Specifically, means, standard deviations, ranges, skewness, and kurtosis were calculated for each individual item and each subscale score. Results revealed good distribution and appropriate normality for the HRBI total score and all subscales except the drug use subscale score (see Table 2). The drug use subscale score was positively skewed and had the lowest mean of all subscales (M=1.21) suggesting low endorsement of drug use in this sample. Individual items were flagged as potentially problematic if they had skewness greater than 2, kurtosis greater than 7, or if responses did not represent the full spectrum of possible values (e.g. no participants endorsed the highest score as a response). These results are presented separately for each subscale in Tables 3-9, with potentially problematic items bolded. These items received particular consideration when decisions were made regarding which items to remove during revision of the pilot HRBI in Part II of the present study. In addition, descriptive statistics were examined for the remaining questionnaires intended for use in the validity analyses (see Table 10). Results revealed that IPAQ subscale and total scores were positively skewed, and therefore outliers of each were brought in to the value 3 standard deviations above the mean. However, this failed to eliminate concerns with the skewness and kurtosis of the IPAQ work and household

54

Table 2 Descriptive Statistics and Internal Consistencies of the Pilot HRBI Subscale and Total Scores (n=340) HRBI subscale Physical activity

Range

Mean

SD

Skew

Kurtosis

1.40-4.50

2.98

.61

.20

-.43

Cronbach’s alpha .68

Diet

1.25-4.33

2.80

.56

.00

-.10

.72

Sleep

1.22-4.56

2.77

.52

.20

.51

.66

Smoking

1.00-5.00

1.96

.94

1.48

1.40

.88

Alcohol use

1.11-3.78

1.96

.52

.77

-.21

.60

Drug use

1.00-3.33

1.21

.38

2.37

6.29

.74

Sexual behavior

1.00-3.50

2.35

.60

-.15

-1.35

.73

HRBI total

11.83-23.17

16.03

2.09

.82

.49

--

55

Table 3 Descriptive Statistics of the Pilot HRBI Physical Activity Subscale Items (n=340) Pilot HRBI abbreviated item

Range

Mean

SD

Skew

Kurtosis

1. Sitting at work

1-5

2.72

1.25

.12

-1.10

2. Standing/moving/lifting at work

1-5

2.60

1.22

.33

-.84

3. Sports, high effort

1-5

3.34

1.44

-.27

-1.30

4. 2½ hours aerobic exercise

1-5

3.09

1.23

-.14

-.90

5.Light physical activity

1-5

2.49

1.10

.37

-.64

6. Resistance training twice/week

1-5

3.64

1.43

-.64

-.99

7. Free time physical effort

1-5

2.99

1.11

.05

-.77

8. Free time sat and relaxed

1-5

3.46

.87

-.23

.04

9. 4 hours TV/read/computer per day

1-5

2.76

1.20

.23

-.84

10. Walked/biked to destinations

1-5

2.68

1.23

.43

-.76

Note. Higher scores indicate greater health risk.

56

Table 4 Descriptive Statistics of the Pilot HRBI Diet Subscale Items (n=340) Pilot HRBI abbreviated item

Range

Mean

SD

Skew

Kurtosis

11. 3 or more vegetables

1-5

3.06

1.06

-.22

-.61

12. 2 or more fruits

1-5

2.98

.97

-.12

-.42

13. Whole grains

1-5

2.84

1.10

.20

-.59

14. Extra salt

1-5

2.13

1.26

.52

-1.03

15.Low fat/fat free dairy

1-5

2.92

1.36

.02

-1.22

16. Sugar-sweetened beverages

1-5

2.83

1.24

.13

-1.01

17. Breakfast every day

1-5

2.82

1.32

.09

-1.15

18.Fast food

1-5

2.00

1.11

1.01

.32

19. Fried food

1-5

3.46

1.08

-.31

-.70

20. Packaged/convenience food

1-5

2.89

1.05

.03

-.79

21. Sweets

1-5

2.84

1.05

.16

-.68

22. Too full/felt stuffed

1-5

2.88

.88

-.05

-.55

Note. Higher scores indicate greater health risk.

57

Table 5 Descriptive Statistics of the Pilot HRBI Sleep Subscale Items (n=340) Pilot HRBI abbreviated item

Range

Mean

SD

Skew

Kurtosis

23. Slept at least 7 hours per night

1-5

2.74

.96

.26

-.54

24. Slept less than 7 hours per night

1-5

2.74

.99

.22

-.49

25. Long time to get to sleep

1-5

2.71

1.06

.29

-.46

26. Woke up during night

1-5

2.56

1.03

.28

-.59

27.Slept more than 9 hours per night

1-4

2.12

.85

.28

-.64

28. Difficult to wake up

1-5

3.36

1.07

-.08

-.69

29. Once asleep, stayed asleep

1-5

2.64

1.13

.40

-.64

30. Tired during day

1-5

3.12

.91

-.05

-.30

31. Difficulty concentrating

1-5

2.89

.90

.14

.05

Note. Higher scores indicate greater health risk; Bolded items indicate potentially problematic items.

58

Table 6 Descriptive Statistics of the Pilot HRBI Smoking Subscale Items (n=340) Pilot HRBI abbreviated item

Range

Mean

SD

Skew

Kurtosis

32. Smoked part of all of cigarette

1-5

1.77

1.41

1.55

.74

33. Smoked within hour of waking

1-5

1.31

.94

3.11

8.59

34. Smoked 10 cigarettes per day

1-5

1.25

.81

3.52

11.96

35.Smoked one cigarette per week

1-5

1.62

1.33

1.92

1.99

36.Did not smoke cigarettes

1-5

1.87

1.52

1.34

.00

37. Smoke in car

1-5

1.86

1.43

1.38

.30

38. Avoided people while smoking

1-5

2.86

1.40

.27

-1.24

39. Secondhand smoke

1-5

3.15

1.17

.04

-.95

40. Smoke in home

1-5

1.99

1.55

1.18

-.36

Note. Higher scores indicate greater health risk; Bolded items indicate potentially problematic items.

59

Table 7 Descriptive Statistics of the Pilot HRBI Alcohol Use Subscale Items (n=340) Pilot HRBI abbreviated item

Range

Mean

SD

Skew

Kurtosis

41. 5/4 or more drinks in a day

1-5

2.02

1.22

.83

-.53

42. 5/4 or more drinks several days

1-5

1.46

.85

2.06

3.98

43. Drank some alcohol

1-5

3.98

1.17

-.83

-.33

44.Did not drink alcohol

1-5

2.86

1.58

.14

-1.52

45.Talked to about drinking

1-5

1.32

.83

2.96

8.56

46. Difficult to stop drinking

1-5

1.51

.94

1.94

3.16

47. Interfere with activities

1-5

1.99

1.33

1.13

-.02

48. Drove after drinking

1-5

1.27

.72

2.98

9.05

49.Trouble because of drinking

1-5

1.25

.70

3.36

12.05

Note. Higher scores indicate greater health risk; Bolded items indicate potentially problematic items.

60

Table 8 Descriptive Statistics of the Pilot HRBI Drug Use Subscale Items (n=340) Pilot HRBI abbreviated item

Range

Mean

SD

Skew

Kurtosis

50. Prescription drugs to get high

1-5

1.14

.52

4.72

25.10

51. Drugs not as prescribed

1-5

1.15

.53

4.13

18.42

52. Smoked marijuana daily

1-5

1.35

.88

2.60

6.04

53. Used other illegal drugs

1-5

1.12

.51

5.02

28.28

54. Did not use illegal drugs

1-5

1.86

1.45

1.35

.14

55. Drugs by injection

1-3

1.01

.13

12.48

167.71

56. Asked to stop using

1-3

1.06

.26

4.06

16.83

57. Unable to do what was expected

1-5

1.08

.38

5.90

42.42

58. Trouble because of drug use

1-5

1.08

.43

6.21

42.29

Note. Higher scores indicate greater health risk; Bolded items indicate potentially problematic items.

61

Table 9 Descriptive Statistics of the Pilot HRBI Sexual Behavior Subscale Items (n=340) Pilot HRBI abbreviated item

Range

Mean

SD

Skew

Kurtosis

59. Sex for money/drugs

1-4

1.02

.20

11.94

156.26

60. Did not know partner well

1-4

1.24

.65

2.97

8.27

61. Multiple sexual partners

1-4

1.16

.53

3.65

13.18

62.Committed partner

1-5

1.47

1.04

2.32

4.38

63. Condom used – oral sex

1-5

3.17

1.94

-.19

-1.93

64. Condom used – vaginal sex

1-5

1.96

1.42

1.18

-.13

65. Condom used – anal sex

1-5

1.25

.84

3.65

12.47

66. Sexual history of partner

1-5

1.47

.98

2.17

3.90

67. Sexual partner STDs

1-5

1.24

.76

3.37

11.23

68. Other partners of sexual partner

1-5

1.33

.73

2.00

2.85

Note. Higher scores indicate greater health risk; Bolded items indicate potentially problematic items.

62

Table 10 Descriptive Statistics of Health Behavior Measures Used to Examine Validity of the Pilot HRBI (n=340) Measure Range Mean SD Skew Kurtosis IPAQ 0-23,049 5647.18 5290.88 1.55 2.02 IPAQ-Work 0-18,153 1988.38 3934.81 2.55 6.45 IPAQ-Transportation 0-5,451 1076.47 1180.46 1.92 3.27 IPAQ-Household 0-6.640 796.32 1330.42 2.80 8.35 IPAQ-Leisure 0-10,505 1905.33 2476.72 1.98 3.53 REAP 1.17-2.96 1.93 .32 .07 -.23 PSQI 0-18 5.47 3.12 .97 1.36 PSQI-quality 0-3 1.14 .71 .34 .11 PSQI-latency 0-3 1.26 .97 .36 -.83 PSQI-duration 0-3 .47 .77 1.66 2.09 PSQI-efficiency 0-3 .38 .67 1.94 3.67 PSQI-disturbance 0-3 1.10 .53 .47 1.56 PSQI-medication 0-3 .24 .64 2.94 8.34 PSQI-day dysfunction 0-3 .89 .72 .55 .27 ASSIST-tobacco SSI 0-20 3.22 3.53 1.84 3.68 ASSIST-current tobacco use 0-4 1.24 1.31 1.01 -.05 ASSIST-tobacco dependence 0-13 3.43 3.43 1.44 1.36 ASSIST-tobacco abuse 0-15 2.16 2.33 2.21 7.06 ASSIST-alcohol SSI 0-15 3.52 2.70 1.23 1.79 ASSIST-alcohol dependence 0-12 3.98 2.38 .72 .46 ASSIST-alcohol abuse 0-11 3.09 1.86 .99 1.72 ASSIST-TSI 0-22 3.89 3.76 1.95 4.37 ASSIST-current marijuana use 0-4 .90 1.05 1.56 1.95 ASSIST-current substance use 0-8 .61 .98 3.81 17.53 ASSIST-injection drug use 0-1 .00 .05 18.44 340.00 ASSIST-substance dependence 0-25 4.36 4.37 2.27 5.96 ASSIST-substance abuse 0-19 2.98 3.29 2.29 5.55 BRFSS-sexual behavior 0-3 .53 .65 .98 .46 SF-36 physical functioning 15-100 93.59 12.56 -3.39 13.77 SF-36 role limitations - physical 25-100 90.01 17.01 -1.87 2.63 SF-36 role limitations – emotional 0-100 84.56 20.03 -1.32 1.27 SF-36 energy 0-100 52.93 18.26 -.16 -.05 SF-36 emotional well-being 10-100 68.07 17.81 -.82 .23 SF-36 social functioning 0-100 81.17 21.74 -1.04 .25 SF-36 pain 0-100 82.38 19.27 -1.44 2.12 Note. IPAQ=International Physical Activity Questionnaire; REAP=Rapid Eating and Activity Assessment for Patients; PSQI=Pittsburgh Sleep Quality Index; ETS=environmental tobacco smoke; ASSIST=Alcohol, Smoking, and Substance Involvement Screening Test; SSI=specific substance involvement; TSI=total substance involvement; BRFSS=Behavioral Risk Factor Surveillance Survey; SF-36=36-Item Short Form Health Survey.

63

subscales. Further examination of the data suggested that this was likely a result of the age and class rank of the present sample. Approximately half of the participants reported no employment, which would result in an IPAQ work value of 0. The young age and freshman status of the majority of the sample likely increased the chances that they were living in dorms or with parents which would make low household-related physical activity more likely, although data on living status are not known, and therefore this observation is speculative. As a result, no further transformations were applied to these IPAQ subscales. Similarly, the sleep medication subscale of the PSQI was positively skewed, likely due to the low base rate of sleep medication use in the present sample. Further, low base rates and issues of skewness and kurtosis were identified in several components representing aspects of substance use in the ASSIST. Finally, the physical functioning score on the SF-36 was negatively skewed, reflecting a high level of physical functioning reported by the present sample. Although the lack of normality of these variables may violate general statistical assumptions, they are not extreme violations, and are very plausible considering the demographic and general wellness status of the present sample. Therefore, no additional transformations were applied to these data in an attempt to remain accurately representative to the present sample.

Reliability of the Pilot HRBI Cronbach’s alpha and item-total correlations are displayed for each subscale in Table 2 and Tables 11-17, respectively. Without removing any items, the subscales of

64

Table 11 Bivariate Correlations Examining Reliability and Validity of the Pilot HRBI Physical Activity Subscale Items (n=340) Pilot HRBI abbreviated item

HRBI physical activity subscale .42**

IPAQ total

IPAQ work activity

IPAQ transport ation

IPAQ household activity

IPAQ leisure activity

-.26**

-.20**

-.04

-.06

-.23**

2. Standing/moving/lifting at work

.36**

-.25**

-.29**

-.10

-.11*

-.03

3. Sports, high effort

.69**

-.28**

-.02

-.07

-.08

-.48**

4. 2½ hours aerobic exercise

.70**

-.20**

.00

-.02

-.08

-.38**

5.Light physical activity

.19**

-.04

.01

-.12*

-.03

-.08

6. Resistance training twice/week

.70**

-.25**

-.07

-.04

-.08

-.40**

7. Free time physical effort

.70**

-.29**

-.04

-.11

-.17**

-.42**

8. Free time sat and relaxed

.54**

-.15*

.01

-.09

-.14*

-.22**

9. 4 hours TV/read/computer

.40**

-.13*

-.10

-.02

-.07

-.09

10. Walked/biked to destinations

.33**

.03

.11

-.08

.12*

-.10

Physical activity subscale score

--

-.37**

-.12*

-.13*

-.13*

-.49**

1. Sitting at work

Note. IPAQ=International Physical Activity Questionnaire. *p