THE FLORIDA STATE UNIVERSITY COLLEGE OF SOCIAL SCIENCES

ENGAGING A DEBATE: AN EXPLORATION OF DEPRESSION, ENGAGEMENT, STRESS AND GENDER IN THE NURSING HOME By TINA HEBERT DESHOTELS

A Dissertation submitted to the Department of Sociology in partial fulfillment of the requirements for the degree of Doctor of Philosophy

Degree Awarded: Summer Semester, 2004

The members of the Committee approve the Dissertation of Tina Hebert Deshotels defended on May 21, 2004. Jill Quadagno Professor Directing Dissertation Marie Cowart Outside Committee Member

John Reynolds Committee Member James Orcutt Committee Member

Approved: Ike Eberstein, Chair, Department of Sociology David Rasmussen, Dean, College of Social Sciences

The Office of Graduate Studies has verified and approved the above named committee members.

ii

ACKNOWLEDGEMENTS

This project could not have been completed without the assistance of many individuals at FSU. First and foremost, I would like to acknowledge Jill Quadagno for her undying support, and encouragement throughout the dissertation process. I would like to recognize John Reynolds for his statistical expertise and patience, and Jim Orcutt for his flexibility and dedication to students. I would also like to thank Marie Cowart for her expertise in the field of aging and in the process of life. Finally, I would like to acknowledge all of the faculty in the Department of Sociology at FSU for giving me the training necessary to complete this project. Dr. Brandon Wallace at Middle Tennessee State University gave me access to and interest in statistics in general and this data set in particular. Additionally, a special debt of gratitude is owed to my family for grounding me in a reality of dirty dishes when my mind was floating in a realm of elderly depression.

iii

TABLE OF CONTENTS

List of Tables ............................................................................................ Abstract .................................................................................................

Page vi Page viii

INTRODUCTION.......................................................................................

Page 1

Objectives of the study…………………………………………………… Measuring Depression in the Nursing Home............................................ Implications for Policy ............................................................................

Page 2 Page 5 Page 6

1. DEPRESSION DEFINITION AND DISTRIBUTION………… .................

Page 10

Definition of Depression.......................................................................... Distribution of Depression....................................................................... Summary……………..............................................................................

Page 10 Page 11 Page 15

2. REVIEW OF THE LITERATURE ON DEPRESSION……………………

Page 16

Theories of Engagement .......................................................................... Critiques of Engagement Theories ........................................................... Summary…… ......................................................................................... Research on Aspects of Engagement, Gender and Depression ................. Research on Stress, Gender and Depression............................................. Summary….. ...........................................................................................

Page 17 Page 18 Page 22 Page 23 Page 30 Page 33

3. DATA AND METHODS ............................................................................

Page 36

Hypotheses.... .......................................................................................... The Survey .. ........................................................................................... The Sample .. .......................................................................................... The Measures ..........................................................................................

Page 36 Page 36 Page 39 Page 41

iv

Independent Variables ............................................................................. Control Variables .................................................................................... Analyses ….. ...........................................................................................

Page 44 Page 49 Page 55

4. FINDINGS……. .........................................................................................

Page 56

Preliminary Findings ............................................................................... Negative Binomial Regression................................................................. Summary….. ...........................................................................................

Page 57 Page 59 Page 65

CONCLUSIONS.........................................................................................

Page 66

Distribution of Depression in the Nursing Home Elderly ......................... Depression, Theories of Engagement, Stress, and Gender ........................ Activity Theory ....................................................................................... Disengagement Theory............................................................................ Stressful Aging........................................................................................ Summary…… ......................................................................................... Contributions...........................................................................................

Page 67 Page 70 Page 71 Page 74 Page 76 Page 78 Page 78

APPENDIX

............................................................................................

Page 81

Human Subjects Committee Approval.....................................................

Page 81

BIBLIOGRAPHY.. ..........................................................................................

Page 82

BIOGRAPHICAL SKETCH ...........................................................................

Page 105

v

LIST OF TABLES

Table 3.1 Sample Descriptive Statistics ............................................................

Page 40

Table 3.2 DRS Descriptive Statistics ................................................................

Page 42

Table 3.3 DRS Item Frequencies ......................................................................

Page 43

Table 3.4 SEI Descriptive Statistics..................................................................

Page 44

Table 3.5 Activity Time Descriptive Statistics..................................................

Page 45

Table 3.6 Reduced Engagement Item Descriptive Statistics..............................

Page 46

Table 3.7 Roles Descriptive Statistics...............................................................

Page 46

Table 3.8 Contact Descriptive Statistics ...........................................................

Page 47

Table 3.9 Stress: Conflict Descriptive Statistics................................................

Page 47

Table 3.10 Pain Descriptive Statistics..............................................................

Page 48

Table 3.11 COG Descriptive Statistics ...........................................................

Page 50

Table 3.12 COG Item Frequencies ..................................................................

Page 51

Table 3.13 ADL Descriptive Statistics.............................................................

Page 53

Table 3.14 ADL Item Frequencies...................................................................

Page 53

Table 3.15 Correlations: Dependent and Independent Variables ......................

Page54

Table 3.16 Mean Differences by Gender ........................................................

Page 55

Table 4.1 Comparison of Means DRS by Engagement ....................................

Page 57

vi

Table 4.2 Comparison of Mans :DRS by Stress ...............................................

Page 58

Table 4.3 Comparison of Means: Gender and DRS..........................................

Page 59

Table 4.4 NBR Regression Results..................................................................

Page 61

Table 4.5 NBR Percent Change .......................................................................

Page 62

vii

ABSTRACT

Does engagement impact depression? Five aspects of engagement (an index of social engagement (SEI) “activity time,” “identification with past roles,” “reduced engagement,” and “contact with family and friends”) were used to determine if theories of engagement explain depression in the nursing home. Activity theory suggests high levels of engagement should be related to lower levels of depression. In contrast, disengagement theory suggests high levels of engagement should be related to higher levels of depression. Data from U.S. nursing home residents (n=6,468) were utilized to examine the relationship between engagement and depression to show that neither theory is fully supported across all aspects of engagement. Rather, support was found for exploring a new theory of “stressful aging” in explaining depression in the nursing home. Specifically, results show that stress in the form of negative interactions (“conflict”) and “pain” frequency are better explanations for depression in the nursing home. In addition, the distribution of depression was explored. There has been a notable lack of research on the distribution of depression among the nursing home elderly. Rather, we often assume it is the same in the nursing home as it is in the community. My results show that this is not a safe assumption. Findings suggest that that the nursing home is a unique setting and that it is important to test the relationship between variables of interest rather than assume that it is the same as in the community. Finally, results also show that it is important to consider gender in understanding how engagement impacts depression in the nursing home. Gender differences would have obscured the relationship between engagement and depression if not included. The results confirm that rather than assume that research on the community dwelling aged is transferable to the nursing home setting, research should be conducted on the nursing home elderly as a separate and unique population.

viii

INTRODUCTION

Depression among the elderly is a common and serious problem (NIH 1992). Depressive symptoms are noted in about 15 percent of the elderly living in the community (Koenig and Blazer 1992). Elders in the nursing home are even more likely than those of the community dwelling aged to have indications of depression (NIH 1992; Chandler and Chandler 1988). Prevalence rates of depression in older people residing in nursing homes are three to five times higher than elders living in the community (NIH 1992). “The rate of new cases of depression in nursing homes is striking; 16 percent of residents develop a new episode of major depression over a one-year period, and another 18 percent develop new depressive symptoms” (NIH 1992:2). Researchers estimate that the prevalence of depressive symptoms in the nursing home ranges from 25 to 50 percent, with many residents exhibiting clinically significant levels of symptoms (Minicuce et al. 2002; Streim, Rovner and Katz 1996; Mor et al. 1995; Parmelee Katz and Lawton 1992; 1989). For example, Mor et al. (1995) found that 30 percent of residents exhibited a mood state problem. These feelings are most often manifested in emotions such as sadness, feelings of emptiness, anxiety, or unease. (Morris, Murphy and Nonemaker 1995). Estimates of actual diagnoses of depression are much lower1. It is estimated that between 10 and 1

There are indications that measurements of diagnosis of depression underestimate depression in the nursing home. While diagnosis of depression may at first glance seem an objective scientific measure, it is difficult even among the general population because “where clinicians draw the line between reactions to the slings and arrows of life and clinical depression remains unclear” (Hybels, Blazer, and Pieper 2001:358). However, it is even more difficult in the nursing home. Depression among the elderly in nursing homes occurs in the context of numerous social and physical problems that often obscures or complicates diagnosis. It is often difficult to distinguish from numerous other conditions affecting older people, making diagnostic errors common (Yesavage 1993). This most often results in underestimating the amount of depression. For example, Teresi et al. (2001) estimate that as many as 37 to 45 percent of cases of depression in the nursing home are overlooked. Random samples of 319 nursing home residents using DSM-IIIR criteria found that prevalence estimate for probably and/or definite major depressive disorder was 14.4 percent. The estimate for minor depression was 16.8 percent. The prevalence of significant depressive symptoms (including the category of possible depression) was 44.2 percent (Teresi, et al. . 2001). However, when nursing home personal were asked to observe residents for possible depression, social workers estimated that 19.7 percent of residents had any kind of depression, nurses 39 percent and nurses’ aids 32 percent. The authors conclude that while depressive disorders among nursing home residents is high; depression recognition is relatively low, with 37 to 45 percent of depressed individuals not recognized as depressed. The study suggested that having staff label actions as depression was not as accurate a measure of depression as having them report actions that can then be examined for possible depression (Teresi et al. . 2001). Given the problems with diagnosis of depression, it is important to review research on symptoms. However, this is not always available. As such, when I review literature, I review studies that measure any indications of depression, and make distinctions where appropriate.

1

25 percent of nursing home residents have major depression (Brown et al. 2002; Burrows et al. 2000; Hughes, Lapane and Mor 2000; Streim et al. 1996; Jenike, 1995; Katz et al. 1995; Parmelee et al. 1992; 1989; Abrams, Teresi and Butin 1992; Phillips and Henderson 1991; Rovner et al. 1991). According to some estimates, minor depression approaches 50 percent of nursing home residents (Rosen, Musant and Pollock 2000). Objectives of the Study In this project, I explore the relationship between depression engagement and gender among nursing home residents. My first task is to examine how depression is defined and measured in the elderly. Next, I will explore the distribution of depression related experiences2 among the nursing home elderly. There is a noticeable lack of research on depression among the nursing home elderly. Research on depression among the community dwelling aged finds that age, education, and health problems, specifically Activity of Daily Living limitations (ADL), impact depression related experiences. However, these relationships have been relatively under examined in the nursing home. Although researchers often presume that the relationships are the same as in the community, the nursing home is a different context. Therefore these assumptions need to be explored. Next, I examine explanations for depression related experiences, concentrating on theories of engagement and its critics. There are two general explanations of aging posited by engagement theories. Disengagement theory postulates that as people age, they slowly exhibit an adaptive disengagement from society, a narrowing of activities and relationships, that results in lower levels of depression. These disengagement theorists suggest that engagement levels among the aged will be low and that high levels of engagement will be related to high levels of depression. In contrast, activity theory postulates that activity and connection to others are as important in late life as it is in earlier years. This theory suggests that elders will remain engaged in activities. Elders who are disengaged will have higher levels of depression than elders who remain active. It is important to note that while these are two separate independent theories, both measure a single variable on a continuum. That is, both theories focus on level of activity3. However, each theory envisions different outcomes. In this project, I use engagement as a definer for both theories and test the applicability of each in explaining depression in the nursing home.

2

By depression related experiences, I refer to research that uses depression as defined by the Diagnostic and Statistical Manual (APA 1994). This includes depressive symptoms, diagnosis of depression, psychosocial well being, morale, and happiness if these measures are consistent with the DSM definition. Because of the problems with diagnosis of depression among the elderly in nursing homes it is important to examine research on other indications of depression as well. 3 Studies examining level of activity have used various combinations of activities (physical activity, leisure activity, recreational activity, religious activity, social activity, and social integration (social roles, social interaction, social support, family, and friendship networks) in a variety of combinations to predict depression in the elderly. For example, Moen, Dempster-McLain and Williams (1989) define social integration as “the degree to which individuals have formal attachments to social structure” (p. 635) . They operationalize integration as number of roles or specific attachments, with indicators that measure activity (church attendance and participation in voluntary organizations). This shows the inter-relation of activity and roles.

2

While there is limited support for each position, most research supports activity theory over disengagement theory. Most elders remain engaged in physical, formal, and productive activities (Lefrancois, Leclerc and Poulin 2001; Pennix et al. 1999; Iwarsson et al. 1998; Morgan and Bath 1998; Ross and Drentea 1998; Morgan and Clark 1997; Ljungquist and Sundstrom 1996; Voelkl, Fries, and Galecki 1995; Parker, Thorsland, and Nordstrom 1992; Kaplan et al. 1987; Lin, Woelfel and Light 1985), and in social relations/interactions (Lennartsson and Silverstein 2001; Everard, Lach and Fisher 2000; Bar-tur, Levy-Shiff, and Burns 1998; Rowe and Kahn 1997; Wolinsky, Stump, and Clark 1995; Adelmann 1994; Sabin 1993; Ulbrich and Bradsher 1993). Elders who are engaged in activities are less likely to have depression related experiences compared to those who are less engaged (Pennix et al. 1999; Iwarsson et al. 1998; Morgan and Bath 1998; Ross and Drentea 1998; Morgan and Clark 1997; Ljungquist and Sundstrom 1996; Voelkl, Fries and Galecki 1995). Likewise, the more engaged elders are in social relations, the less depression related experiences they exhibit (Lennartsson and Silverstein 2001; Everard et al. 2000; Bar-tur et al. 1998; Rowe and Kahn 1997; Wolinsky et al. 1995). One problem is that these are studies of the relatively young (under 85) and healthy old. Empirical studies of the very old find that as age and poor health increase, engagement in activities decreases (Adams 2001; Coleman, Ivani-Chalian and Robinson 1999; Atchley 1998; Horgas, Wilms and Baltes 1998; Bennett 1997; Crespo et al. 1996; Johnson and Barer 1992; Davies and Gledhill 1983; Sill 1980). Further, studies that include the very old have found that disengaged elders do not have higher depression related experiences than engaged elders (Ritchey, Ritchey and Dietz 2001; Johnson and Barer 1992; Davies and Gledhill 1983). When examined in the nursing home, the results are mixed. First, not surprisingly, among the nursing home, a population that contains a large percentage of the very old and very sick, disengagement is more common. Indeed, when measuring participation in activities in the nursing home, research finds that disengagement is the rule rather than the exception (Gilbart and Hirdes 2000; Horgas et al. 1998; Resnik, Fries and Verbrugge 1997; Schroll et al. 1997; Mor et al 1995). In addition, while in general most engaged residents have less depression related experiences than those who are not engaged (Gilbart and Hirdes 2000; Mor et al. 1995) this may not be true for less physically healthy residents. Those who are high on physical impairment but have few cognitive deficits experience more mood problems when they are highly engaged than when they are not engaged (Gilbart and Hirdes 2000; Mor et al. 1995). Critics of theories of engagement are many. First, some point out that the quantity of elders’ engagement is not as important as the quality of engagement in predicting depression. This suggests another approach to understanding depression in the nursing home. Specifically, critics argue that stressful interactions (e.g. conflict) and physical stress (e.g. pain) are more important determinants of who is depressed and who is not. Indeed, there is ample evidence to show that conflicted relations are more important then positive interactions in predicting depression among the elderly (Utz et al. 2002; Rock 1984). In addition, pain has consistently been shown to be related to depression in the aged and in the nursing home (Parmalee Katz and Lawton 1993; 1991). However, neither theory of engagement accounts for the quality of engagement. Neither does it account for the impact of physical stress (e.g. pain) on depression. Second, perhaps the most vocal critics of theories of engagement, feminist theorists, argue that theories of engagement are to individualistic. Feminists critique the lack of attention to structural forces by both disengagement and activity theorists. Both disengagement and activity theory, feminists argue, assume that men and women have the same opportunity to engage in activities and roles. Engagement theories also assume that engagement in activity and

3

roles would have the same meaning for men and women. Given that experiences vary by group membership, engagement theories are insufficient when they fail to include gender relations, or the experience of women or other social groups in the context of aging (Blieszner 1994; Reinharz 1986). As such, studies of engagement in activities and social interactions should take into account that engagement will take different forms and have different meaning for different groups4. As such, first, feminist theory argues that because experiences vary be gender, gender should be a primary consideration in attempts to understand depression. Indeed, research shows that elderly women in the community and in the nursing home have more depressive symptoms and are more likely to be diagnosed for depression than elderly men (Brown et al. 2002; Jungmeen and Moen 2002; Topinkova and Neuwirth 1997; Mirowsky 1996; Keith 1993; Dean Kolody, Wood, and Matt, 1992; Krause and Goldenhar 1992; Blazer, Burchett, Service and George 1991). When examining gender differences in engagement and gender differences in the relationship between engagement and depression in the nursing home, research is limited. One study has found that activity levels differ significantly between men and women (Voelkl et al. 1995). However, there has been more research in the community dwelling aged. This research shows that there are gender differences in the levels of engagement and the impact of some aspects of engagement on depression. However, there is reason to suspect that this relationship is different among the very old and especially the very old in nursing homes. First, research that finds gender differences in depression is conducted on the fairly young old (85 and under), or uses aggregate measures, combining 80 and older into one category (e.g. Mirowsky 1996). Research on the community dwelling aged that includes continuous measures of the very old finds that while gender differences remain, differences decrease as age increases, with men coming closer to parity with women (Weissmann, Bruce, Leaf, Florio and Holzer 1991). Therefore, it is possible that gender differences in old age and particularly in the nursing home diminish. First, gender specific duties decline in old age. Second, the nursing home is a setting where ostensibly gender should not matter because all residents are exposed to the same context. Furthermore, studies of gender differences in depression rely on self report. However, research shows that elderly men are less likely to report depressive symptoms when compared to elderly women (Hurley and Sieagal 2001). Therefore, studies that rely on self report to estimate depression related experiences find artificially low levels of depression in men (Hurley and Sieagal 2001). As such, self report is an inaccurate measure of depression and there is reason to suspect the relationship reported. In this study, I first describe rates of depression among my sample of nursing home elderly. Then, I examine the distribution of depression concentrating on variables that have been found to be related to depression among the community dwelling aged. This is important because research on the distribution of depression in the nursing home is scarce. Next, I examine the applicability of theories of engagement and its critics, research on stress, in explaining depression in the nursing home setting. In order to do this, I first examine levels of engagement using five aspects of engagement (an index of “social engagement,” “activity time,” “identification with past roles,” “reduced engagement,” and “contact with family and friends”). I also examine levels of stress (“conflicted relations” and “pain”). Next, I examine the 4

Holstein (1999) has illustrated the gendered implications of theories of engagement by pointing out that measures of engagement are in terms of masculine definitions of productivity. Measuring engagement by focusing on masculine definitions of engagement, feminists argue, leaves out the experiences of women.

4

relationship between these five aspects of engagement and two aspects of stress and depression. I also explore gender differences in the relationship between aspects of engagement, stress and depression. Measuring Depression in the Nursing Home In order to examine these relationships, I use data from 6,468 nursing home residents in 80 different facilities in 12 states across the U.S. (Alabama, Georgia, Indiana, Kansas, Kentucky, Massachusetts, Montana, New Hampshire, South Carolina, Tennessee, Virginia and Washington). Historically, it has been difficult to conduct research in the nursing home. Most U.S. nursing homes maintained only limited information on residents (Morris et al. 1990). In the past, each home identified and adopted its own assessment system. Item definitions varied between facilities, truly comprehensive assessments were rare, and reliable information was often unavailable (Rabins et al. 1987). The Congressional mandate in the Omnibus Reconciliation Act of 1987 (OBRA ’87) has changed that situation. U.S. nursing homes are now required to complete a standardized and comprehensive assessment of all residents’ functional, medical, psycho-social, and cognitive status (Morris et al. 1990; 1991) and use these data to develop an appropriate plan of care. By law, nursing homes receiving Medicare or Medicaid reimbursements are required to assess every resident. Assessments are required at admission, at regularly scheduled intervals thereafter, and on significant change in the resident’s status. The overall system mandated by OBRA ’87 is known as the Resident Assessment Instrument (RAI) and its care assessment component is known as the Minimum Data Set (MDS). The RAI was developed under the direction of the Health Care Financing Administration by a research consortium lead by the Research Triangle Institute in North Carolina. Collaborators included Hebrew Rehabilitation Center for the Aged in Boston, the Center for Gerontology and Health Care Research at Brown University, and the Institute of Gerontology at the University of Michigan (Morris et al. 1990). The RAI contains three major components: (1) a core set of assessment items designed to provide a comprehensive picture of resident’s functional status (known as the MDS for Nursing Home Resident Assessment and Care Screening or simply the MDS); (2) a set of specialized assessment protocols designed to directly link MDS data to care planning (known as Resident Assessment Protocols or RAP); and (3) a user’s manual containing detailed specifications as to how to complete the MDS and Rap assessments, along with item definitions, coding examples, clinical guidelines for using the RAP for care planning and case studies (Morris et al. 1995; 1991). Hallmarks of this assessment tool are its comprehensiveness, use of standardized definitions, and inclusion of items that measure resident deficits, strengths, and preferences. Trained clinical professionals (e.g. nurses, social workers, therapists) assess resident performance over all shifts during the prior seven-day period. Each item has its own explicit definition. Each assessor is to interact directly with the resident, review the record, and gather information on resident performance from direct care and licensed professional staff. Cues are provided for how to ask questions, what to observe, and whom to contact for information (Morris et al. 1991). The MDS is particularly well suited to this project because data are collected on over 300 items from 19 different areas: (1) identification and background information; (2) customary

5

routines; (3) cognitive functioning; (4) communication and hearing; (5) vision; (6) mood and behaviors; (7) psychosocial well-being; (8) physical functioning and structural problems; (9) continence; (10) disease diagnoses; (11) health conditions; (12) oral and nutritional status; (13) dental status; (14) skin conditions; (15) activity pursuits; (16) medications; (17) special treatments and procedures; (18) discharge potential and overall status; and (19) assessment information. In addition, the MDS is important to this project because it contains validated measures of depression. Burrows et al (2000) designed the MDS Depression Rating Scale (MDS-DRS) for nursing homes. The MDS-DRS was tested and validated using the Hamilton Depression Rating Scale and the Cornell Scale for depression, based on DSM-IV criteria and created specifically to be used on geriatric populations. There are two important advantages to the DRS. First, it “draws on continuous observations by staff involved in the daily lives of nursing home residents as incorporated into a routine, standardized assessment protocol” (Burrows et al. 2000:169). Staff are asked if residents have made negative statements (passive suicidal ideation), are persistently angry and irritable with others, express what appears to be unrealistic fears, have repetitive health complaints, have repetitive anxious complaints/concerns (non health related), have sad worried facial expressions, have crying and/or tearfulness. Staff members are not asked to label the symptoms as depression. They are only asked to report what they see. This avoids the limitations of previous studies. That is, it avoids measurement problems associated with diagnosis and labeling of depression among the elderly and among men. Second, the measure is a continuous rather than a categorical measure. “Thus it captures a spectrum of depressive symptomology and allows characterization of individuals with less severe symptoms” (Burrows et al. 2000:170). In this way, depression in the nursing home can be viewed as a matter of degree rather than something one either has or does not have. This measure will permit a more nuanced understanding of depression and how it relates to the variables under question. Implications for Policy Understanding depression in the nursing home is important. Depression in the elderly can have a significant impact on overall heath and desired outcome. “The depressed elderly patient has been shown to have worsened prognosis of concomitant medical conditions, increased use of health care, decreased recovery time and more likelihood of experiencing accelerated physical deterioration” (Cooke and Tucker 2001:498). It is a major cause of morbidity and mortality in elderly populations (Blazer et al. 1991; Klausner and Alexopoulos 1999). These negative consequences extend to the nursing home. Depression has been correlated with an increased risk of mortality in nursing home patients (Rovner and Katz 1993). Indeed, in the nursing home, major depression has been found to increase the likelihood of mortality by 59 percent, independent of physical health measures (Reynolds and Kupfer 1999). In addition, due to the demographic revolutions transforming America and indeed all post industrial societies, it is likely that understanding depression in the nursing home will be increasingly important in the future. First, there is a revolution in longevity, leading to increased life expectancy and the progressive aging of our population (Laslett 1991; Lonergan 1991). Life expectancy among the aged has changed dramatically over the past decades so that now over a

6

quarter of those over 65 can expect to live until they are 90. Those who turned 65 in 1994 are expected to live 17 additional years, an increase of 22 percent since 1960 (Glass et al. 1999). These additional years of life are expected to be years with chronic activity limiting conditions (Crimmins and Saito 1993). These figures point to an increasingly larger demographic presence of very frail older persons. This means that not only will there be more elders, but more elders that require intensive care, and thus an increase in the population that needs nursing home care. In addition, more women are entering the workforce, a trend that is expected to increase. Because women are the primary caregivers of the elderly, this means that there will be fewer people available to care for an increasingly disabled population. The result will likely be a dramatic expansion of the nursing home population. Smaller families, higher rates of divorce, and greater geographic dispersion of family members, means that fewer familial supports will be available in future years, making it likely that those needing nursing home care will increase. When examining the societal costs, the NIMH (1992) notes that depression contributes substantially to the overall disease burden. An estimated $30.4 billion was lost to the direct and indirect costs of depression in 1990 (NIMH. 2001). As it is now, the 1990 Census reported that almost 1.6 million older persons or 5.1 percent of people aged 65 and older (1.4 percent of those age 65-74, 6.1 percent of those age 75-84, and 24.5 percent of persons aged 85 and older) were receiving care in nursing homes (U.S. Congress General Accounting Office 1998). Nursing home care continues to be the single largest component of long-term care expenditures for the elderly population. The total national expenditure for long-term care in 1996 was 125.5 billion, with 69.7 percent of that spent on nursing home care (Levitt et al. 1987). Given the seriousness of the disease, and the increasing population of nursing home residents, depression in the nursing home has been identified as a number one research priority for the twenty-first century (Neugebauer 1999). Understanding depression in the nursing home, by examining what is related to depression and relating empirical findings to theoretical issues, will help ensure that resources are used in the most efficient and productive manner. If we increase our knowledge base, we can increase the capacity for translating knowledge into social action. My study will contribute to understanding depression in the nursing home by addressing the following gaps in the literature. First, the social distribution of depression in the nursing home has been relatively under examined. Often it is assumed that relationships are the same as they are in the community dwelling aged. I intend to challenge this assumption by examining the relationship between age, education, health, gender and depression. Second, much of the current literature on depression among the elderly ignores the oldest old (Blazer 2000). However, there is reason to suspect that theories may not apply to the oldest old (e.g. feminist theory). In addition, many of the theoretical predictions are most likely to occur among the very old (disengagement). In this study I will be able to expand existing theories by examining previously neglected populations. A third problem my study will address is that literature on depression among the elderly often relies on self report, or diagnosis, both of which are problematic measures especially when applied to the very old and men. Because my data relies on staff observations, I will avoid this limitation. Fourth, studies of depression among the elderly often suffer from a lack of uniform data collection. Because a hallmark of the MDS is uniform data collection, I will avoid this limitation. Finally, even when predictors of depression have been studied in the nursing home, they have not been related to theoretical explanations. For example, Mor et al (1995) found some

7

support for disengagement theory among segments of the nursing home population. Among the relatively high cognitive functioning and low physical functioning groups, higher levels of engagement were related to more mood problems. While this could be interpreted as support for disengagement theory, Mor et al. (1995) does not address the possibility that disengagement is possibly adaptive or at least not maladaptive. Rather, they conclude that levels of disengagement are valid measures of the quality of nursing homes. Without fitting empirical evidence within a generalized understanding, the scientific community does not move towards understanding depression. While Bengtson, Burgess and Parrott (1997) point out that much of the current research seems to have disinherited theory, they argue that theory is always present, but implicit, unacknowledged. Unacknowledged theory is unproductive. It prevents a community of researchers from presenting the empirical results “within the context of more general explanations; thus the process of building, revising, and interpretation of how and why phenomena occur is lost” (Bengtson et al. 1997:73). Therefore, “[i]t is important that the theoretical premises under which research proceeds be stated” (Bengtson et al. 1997:73). I will avoid this limitation by consciously relating empirical findings to theoretical explanations. The remainder of my dissertation is organized as follows. In the Chapter 1, I outline how depression is defined and the problems and solutions for detection of depression among the elderly and nursing home elderly where appropriate. How depression is defined is important because different definitions results in different measures, which in turn yields different results. In this chapter, I also examine variables that have been found to be related to depression among the aged. I concentrate on the nursing home elderly, but because this population is often neglected, I also include research on the community dwelling aged. In the following chapter, I review literature that pertains to engagement, stress, gender and depression in the elderly and nursing home elderly where appropriate. First, I examine theories of engagement and the critics. The debate between activity theory and disengagement theory has a long history. I give only a brief summary of the major tenants and concentrate on how these theories can be used to explain depression in the nursing home. Next, I present critics of theories of engagement. Two critics are of particular interest in this project. First, critics argue that theories of engagement are too concerned with the quantity of elders’ engagement and leave out consideration of the quality of engagement. They argue that engagement can be stressful, and it is stress that is important in explaining depression. In addition, elders, and elders in the nursing home in particular, are especially vulnerable to the effect of physical stress in the form of pain. Taken together, these critics suggest that stress, in the form of negative interactions and pain, is most important in determining depression among the elderly. Second, feminists critique theories of engagement. Gender, feminists argue should be of primary concern when examining aging and well-being, in this case depression. Feminists argue that there should be gender differences in depression, in levels of engagement, and in the impact of aspects of engagement on depression. Similarly, there should also be gender differences in stress and in the impact of stress on depression. Based on previous literature, I formulate and present my hypotheses at the end of Chapter 2. In Chapter 3, I describe the methods and data used to address the hypotheses. In Chapter 4, I present the findings. I found that neither theory of engagement is consistently related to depression across all five aspects. Some aspects of engagement support activity theory and some support disengagement theory. However, I do find consistent evidence for critics of engagement theories. First, I found that stress, measuring the quality of residents’ engagement, the number of conflicted relations, far outweighs the number of activities and

8

relations in predicting depression in the nursing home. In addition, I found that pain consistently has the strongest relationship to depression of all the variables in the model. I also found support for feminist theorists’ contention that there are gender differences in the impact of aspects of engagement and stress on depression. Finally, in Chapter 5, I present the conclusions, implications for future research and limitations of the study. I conclude that because neither activity nor disengagement theory is consistently supported in the findings, they are not useful in explaining depression in the nursing home. However, consistently, variables that measure stressful life events (conflicted relations and pain) impact depression in the nursing home more so than other variables in the models. That is, stress is a better explanation for depression in the nursing home. My results suggest evidence for investigating a new theory of “stressful aging” as a better explanation for depression in the nursing home.

9

CHAPTER 1 DEPRESSION: DEFINITIONS AND DISTRITBUTION

A sociological study of the relationship between engagement and depression requires an examination of how depression is defined and a baseline understanding of how depression is distributed across social groups. Defining depression is a result of group consensus and as such is a social event. Understanding how “we” define depression is important because definitions impact outcomes. In addition, the sociological study of depression documents the role of social factors (age, education and gender) in determining the likelihood of being depressed. Knowing how social categories are related to depression will allow a clearer understanding of how engagement is related to depression. In addition, because health is related to both engagement and depression, it is also important to examine the relationship between health variables and depression. In this chapter, I first examine how depression is defined, focusing on definitions of depression among the elderly. Next, I examine how depression is socially distributed across age, education, and gender. Because both engagement and depression are correlated with health status, I also review literature on the relationship between health (ADL limitations) and depression. I concentrate my literature review on the nursing home elderly. However, because this is a neglected population, many of the categories known to be related to depression are under examined or unexamined in the nursing home. Therefore, I also present literature that examines how depression is socially distributed among the community dwelling aged and make postulations about how this would transfer to the nursing home setting. Definition of Depression According to the National Institute of Mental Health, depression is a mood disorder. It is not the same as a passing blue mood. Depression is characterized by persistent sad mood, feelings of hopelessness, helplessness, worthlessness, and thoughts of death or suicide. As such it is “an abnormal state of unpleasant emotion” (APA 1994:32). The definition of depression among the nursing home elderly is based on the American Psychiatric Association’s Diagnostic and Statistical Manuel of Mental Disorders (DSM-IV) (APA 1994). The DSM-IV defines depressive symptoms as: (1) depressed mood, persistent sadness; (2) loss of interest in pleasurable activities; (3) loss of appetite; (4) sleep disturbance; (5) fatigue; (6) feelings of

10

worthlessness and guilt; (7) difficulties in thinking and concentration; (8) psychomotor disturbances; and (9) suicidal notions. Current psychiatric diagnosis of major clinical depression rules out much sadness and malaise attributable to disease, grief, poverty, restricted activity, and physical disability (Kennedy et al. 1989; Boyd et al. 1982). The philosophy views clinical depression as essentially endogenous. That is this philosophy views depression as originating from the individual rather than from individuals’ environment or history. This means that for a diagnosis to pertain, the DSM-IV (1994) requires that symptoms displayed arise from a biological, behavioral, or psychological dysfunction within the individual. Symptoms must not be accompanied by functional impairment, must not be better accounted for by bereavement, and must not be due to medical conditions (Ahmed and Takeshita 1997). As such, the definition of depression is particularly problematic when applied to elders. Elders are more likely than other age groups to experience adverse life events in the form of losses such as disease, grief, poverty, restricted activity, and physical disability (Boyd et al 1982; Kennedy et al 1989). Growing old brings with it numerous biological, developmental and social changes that entail losses, which are risk factors for the development of depression. These include the following: (1) losses in such areas as job, status, health functioning, finances, and loss of relationships with spouse, children, siblings, and friends through death or relocation; (2) decreased adaptive capacity both physical and psychological; (3) neurotransmitter and receptor changes in the brain; (4) increased incidence of physical illness and concomitant medication use; and (5) increased incidence of cognitive impairment, including dementia and delirium (Ahmed and Takeshita 1997). These changes and losses are particularly applicable for elders in the nursing home, making the definition of depression even more difficult in this setting. Responding to the problems of understanding depression in the nursing home, Burrows et al. (2000) used items in the MDS to create a Depression Rating Scale (DRS). Burrows correlated all MDS mood items to two commonly used measures of depression, the 17 item Hamilton Depression Rating Scale (Hamilton 1967) and the 19 item Cornell Scale for Depression in Dementia. Both scales have been validated in disabled and medically ill elderly populations (Rapp; Smith, and Britt 1990), and are widely used in cognitively impaired geriatric populations (Burrows et al 1995; Logsdon and Teri 1995; Alexopoulos et al. 1988a; 1988b). The resulting DRS has seven items (negative statements, persistent anger, unrealistic fears, health complaints, anxious complaints, sad expressions, and crying) with a range of 0 to 14. Although the purpose of Burrows et al. (2000) was to determine a cut off point for when residents were diagnosable as depressed, the scale can also be used as a continuous measure. Distribution of Depression Whatever measures are used, researchers agree depression among the aged in the nursing home is a serious problem. Studies find prevalence of depressive symptoms is between 25 and 50 percent (Wilco et al. 2003; Minnuci et al. 2002; Streim et al. 1996; Katz et al. 1995; Mor et al. 1995; Abrams et al. 1992; Parmelee et al. 1992; 1989; Phillips and Henderson 1991; Rovner et al. 1991; Hyer and Blazer 1982). For example, in a recent study, Wilco et al. (2003), using the MDS-DRS, found that 27 percent of newly admitted nursing home residents had high levels of depressive symptoms. Some estimates of minor depression approach 50 percent of nursing home

11

residents (Rosen et al. 2000). The most common symptoms exhibited in the nursing home are sad, pained facial expression (12.2 percent exhibited five days a week or more) and crying/tearfulness (8 percent exhibited five days a week or more) (Wilco et al. (2003). Between 10 and 25 percent of nursing home residents have an active clinical diagnosis of depression (Brown et al. 2002; Burrows et al. 2000; Abrams et al. 1992; Parmelee et al. 1992, 1989). Depression and age. Because as people age they are more likely to experience a corresponding “normal” level of sadness in relation to losses and changes, symptoms of depression may increase with age, but diagnosis decrease. When examining depressive symptoms in the community dwelling aged, most (but not all) investigators report that depressive symptoms are higher, on average, among older adults than their younger peers (Pennix et al. 1999; Mirowsky 1996; 1992; Krause and Goldenhar 1992; Roberts, Lee and Roberts 1991). For example, Mirowsky (1996) found that depressive symptoms rise among people older than 60. People 80 or older report the most depression followed by those 70-79. “The results leave little doubt that the oldest age groups average high levels of depression, and people 80 years old and older report the most depression” (Mirowsky 1996:180). However, when examining diagnosis of depression, research finds many elderly exhibiting a relatively high number of depressive symptoms relative to the number of diagnoses (George 1992; Koenig and Blazer 1992; Blazer et al. 1991; Weissman et al. 1991). For example, Koenig and Blazer (1992) found that although 15 percent of community dwelling elderly exhibit significant depressive symptoms, only about one percent meet formal criteria for diagnosis of major depression, a prevalence about one-fourth that of adults 18 to 44 years of age (Koenig and Blazer 1992). Research on diagnosis of depression finds that the likelihood of qualifying for a diagnosis of depression declines with age (Brown et al 2002; Blazer et al. 1991). The prevalence of diagnosis of depression was slightly lower in older residents. About 13 percent of elders 65 to 74 had a diagnosis of depression. However, for those aged 75 to 84, about 12 percent had a diagnosis. Only about 9 percent of those aged 85 and older had a diagnosis of depression. The lowest prevalence of diagnosis of depression was in residents aged 85 and older (Brown et al 2002). Similarly research on diagnosis of depression among the nursing home elderly finds that as age increases, the chance of being diagnosed for depression significantly decreases (Jones, Marcantonio and Rabinowitz 2003). Because I am interested in how much sadness elders experience in the nursing home, my study will use symptoms of depression rather than diagnosis. Depression and education. There has been no research on the relationship between education and depression among the nursing home elderly. Research on depression in the community dwelling aged consistently finds that elders with higher educational levels have significantly less depression related experiences than those with less education (Pennix et al. 1999; LaGory and Fitzpatrick 1992; Mirowsky and Ross 1989; Ulbrich, Warheit, and Zimmerman 1989; Norris and Murrell 1987; Holzer et al. 1986; Pearlin et al. 1981; Gove and Geerken 1977). Pennix et al. (1999) compared the relative affect of age and education and found that education accounts for most of the differences found in depression. Indeed, the relationship between education and depression is so well established in the community dwelling aged that Mirowsky (1996) notes, “some depression appears to be caused by advancing age, but is not. It actually reflects greater depression among people with lower education, who tend to be older (p. 197). Researchers examining the relationship between education and depression postulate that education can be viewed as a personal resource. Hess and Warring (1983) note, “[j]ust as the problems of the old are not evenly distributed among all sub populations of the elderly, neither

12

are the resources required to cope successfully” (p. 232). They point out that there is an inverse relationship between problems and resources. “Those with the gravest problems are typically those with the fewest personal and social resources” (Hess and Warring 1983:232). The less access to resources individuals have, the less able they are to cope successfully and the more likely they will be to have depression related experiences. While studies of the impact of education on depression have not been conducted in the nursing home, it seems logical that the benefits of education would carry over into the nursing home. As such, I would expect that as educational levels increase depressive symptoms will decrease. Depression and ADL. Because health is an important correlate of depression and engagement, it will be important to get a baseline understanding of how health measures impact depression. I pay particular attention to Activities of Daily Living (ADL). ADL measures are the most common way to measure health among the elderly (Bailis et al. 2001). ADL is a functional measure based on individuals’ ability to perform activities of daily living that are considered essential to independent living and self-care (e.g. eating, bathing, locomotion, personal hygiene, etc.) (George 1995). In the social sciences, most research uses functional measures, claiming that these measures give an accurate assessment of the outcome of poor health on the elderly population as a whole (George 1995). There have only been a few studies of the relationship between ADL limitations and depression related experiences in the nursing home. Gilbart and Hirdes (2000) found that higher ADL limitations were associated with more mood problems. Other research has found that as ADL limitations increase so do self reports of depressive symptoms (Harralson et al 2002) and diagnosis of depression (Harralson et al 2002; Commerford and Reznikoff 1996). One study found that the most impaired residents were significantly less likely to be diagnosed for depression than those who were least or moderately impaired (Jones et al 2003). By contrast, there is an abundance of research on the relationship between ADL limitations and depression among the community dwelling aged. Empirical research consistently finds that among the elderly in the community, increased depression correlates strongly with increased dependence in preparing meals, shopping, getting out, laundering clothes, and bathing, and with ambulatory problems such as difficulty walking or climbing stairs (Harralson et al. 2002; Minicuci et al 2002; Utz et al. 2002; Camacho, Strawbridge, Cohen et al. 1993; Mirowsky and Ross 1992; Dean Kolody and Wood 1990; Kennedy et al. 1989; Ulbrich et al. 1989; Gurland et al. 1988; Turner and Noh 1988; Arling 1987). Indeed, ADL limitations have been found to be the strongest predictors of distress in old age (Arling 1987). In another study, physical disability accounted for 40 percent of the variance in well-being (Litwin 2000). When controlling for health, Harralson et al. (2002) found that depression related experiences did not differ by age groups. Given the abundance of research that finds that ADL positively impacts depression, I expect this relationship to carry over to the nursing home setting. Depression and gender. While gender differences in depressive symptoms have not been studied in the nursing home, it has often been studied in the community dwelling aged. Elderly women are consistently more likely than men to have depressive symptoms in the community dwelling aged (Bergh, Steen, Waern et al. 2003; Jungmeen and Moen 2002; Hybels et al. 2001; Pennix et al 1999; Mirowsky 1996; Keith 1993; Krause and Liang 1993; Dean et al. 1992; Krause and Goldenhar 1992; Krause, Herzog and Baker 1992; Mirowsky and Ross 1992; Nolen-

13

Hoeksema 1990). Mirowsky (1996) also find that gender differentials in depression are stable across age groups. While there have been no reports of gender differences in depressive symptoms in the nursing home, a few studies have examined gender differences in diagnosis of depression in the nursing home. Brown et al. (2002) found that diagnosis of depression was slightly higher for women (11.4 percent) than for men (9.7 percent). The study did not report if the differences increased or decreased with age. Similarly, Topinkova and Neuwirth (1997) found that women were more prone to suffer from depression than were men in long-term care settings. In another study of 28,367 nursing home residents with Alzheimer’s, depression was equally prevalent in men and women (Ott, Lapane and Gambassi 2000). However, the researchers noted that their study had limitations in that it was not a random sample of all nursing home residents, but a special population of Alzheimer’s patients. Several explanations have been considered for this finding. Research has noted that women have significantly lower educational levels than men (Johnson and Crystal 1999; Ross 1996). Since individuals with lower educational levels have significantly more depressive symptoms than those with higher education, it may be that differences in education explain why women have more depression than do men. However, there have been no studies of gender differences in the impact of education on depression among the elderly. In the community dwelling aged, research has found that women have higher rates of disability (Verbrugge, Lepkowsky and Imenaka 1989; Anderson et al 1999; Ross and Wu 1996). Other research on the community dwelling aged finds that the effect of health events, such as becoming ill, on depressive symptoms is stronger for women than men (Dean et al. 1992). There have been no studies of gender differences in ADL levels or gender differences in the impact of ADL on depression among the nursing home elderly. With regard to gender and depression, while much research finds gender differences in depression and in the impact of age, education, and ADL limitations on depression, there are several criticisms to this postulation. First, these studies seldom focus on the very old (over 85). Studies that control for age find that the gender gap narrows in older age groups, with men coming closer to parity with women in diagnosis of depression (Weissman et al. 1991). Similarly, George (1992) found that gender was a significant risk factor for the onset of depressive disorder for young adults, but not for middle-aged and older adults. A second criticism is that most studies rely on self reports, which may underestimate the prevalence of depression in men, and elderly men in particular. In a sample of community dwelling elderly, Hurley and Siegal (2001) found that men were much more likely to be at risk for under detection of diagnosis because of unreported symptoms than were women. This suggests that men may be as likely or more likely to have depression related experiences as women. However, these symptoms may be less likely to be reported. Given the evidence for substantial gender differences in depression among the nursing home aged and community dwelling aged, I expect that women will have more depressive symptoms in my sample of nursing home elderly than men. In addition, given the fundamental organizing principal of gender, it is important to consider that there may be gender differences in the impact of education and ADL on depression.

14

Summary In this chapter, I examined definitions of depression and how depression is measured in the elderly and nursing home elderly. This examination revealed the importance of using indicators of depression that are distinct from what we might expect from “natural” aging (e.g. sleep disturbances, motor agitation, etc.) in defining depression. It also showed the importance of using symptoms of depression rather than diagnosis in determining how much sadness is experienced by elders. Next, I presented literature that described depression among the nursing home elderly. Based on this literature, I expect there to be between 25 and 50 percent of nursing home residents with at least one depressive symptom. I next presented research that examined the social distribution of depression among the elderly and nursing home elderly where available. I discovered that there is a noticeable lack of research on the social distribution of depression among the nursing home elderly. It is often presumed that the relationships are the same as in the community. I maintain this is not necessarily the case and needs to be explored. Based on this review I formulate the following propositions designed to generate specific hypothesis to determine if the social distribution of depression in the nursing home is the same in the nursing home as it is in the community. I hypothesize that as age increases, depressive symptoms will also increase. I hypothesize that as educational level increases, depressive symptoms will decrease. I expect that those with higher ADL limitations will have more depressive symptoms than those with lower levels of limitations. Women will have more depressive symptoms than men. It will also be important to test for gender differences in the impact of age, education and ADL on depression. Because cognitive limitations5 and race/ethnicity6 are also related to depression and the relationship is uncertain, I will control for cognitive limitations and race/ethnicity.

5

Cognitive impairment presents a particular challenge in the nursing home. First, the relationship between cognitive impairment and organic mental disease remains ambiguous and controversial. “Of all the potential health indicators, cognitive impairment: (a) appears to be least strongly related to social factors; and (b) has received the least attention form social scientists” (George 1995:231). Second, other measures in the study are highly correlated with cognitive limitations. For example, cognitive impairment and physical impairment (ADL) is highly correlated (Won, Lapane and Gambassi 1999). Finally, measures of pain rely on resident report may not be valid for residents who are cognitively impaired. While studies of pain have found that pain reports of subjects whose cognition ranges from intact to moderate impairment are valid (Parmelee, Smith and Katz 1993), a study to test the reliability of the pain item found that it had problems with cognitively impaired residents (Phillips, Chu, Morris and Hawes 1993). Studies found that pain is typically underreported for more cognitively impaired residents, compared with diagnostically similar cognitively intact residents (Parmelee et al. 1993; Sengstaken and King 1993). Given the problems that cognitive impairment presents theoretically, with data accuracy, and with correlations to other variables, many studies control for cognition (Gilbart and Hirdes 2000). 6 The relationship between race/ethnicity and depression in both the community dwelling aged and nursing home aged is unclear. Using the MDS, (Harralson et al. 2002) found that white residents were repeatedly shown to have more depression than black residents. In contrast, Cohen, Hyland and Magai (1998) found no racial differences in depressive symptoms, but whites were significantly more likely to receive a diagnosis of possible depression. Other studies report an interaction between race and education. Among persons of high levels of education, blacks have fewer depressive symptoms than whites; among persons of low levels of education, blacks have more depressive symptoms than whites (Ulbrich et al. . 1989). Given that the relationship is unclear and that there is very little racial variation in my sample (92 percent are White), I will use race as a control variable in testing predictors of depressive symptoms.

15

CHAPTER 2 REVIEW OF THE LITERATURE ON DEPRESSION

This chapter outlines previous theoretical and empirical research on the relationship between engagement and depression. The purpose of this chapter is to use previous studies to formulate specific hypotheses designed to increase out understanding of depression and engagement in the nursing home. This chapter is organized as follows. First, I examine how theories of engagement explain depression among the elderly. Disengagement theory states that normal aging is a gradual process of inevitable disengagement from society by the aging individual and from the individual by society. As such, this theory suggests that reduction in levels of disengagement is normal and if elders do not disengage, they will be more likely to have depressive symptoms than elders who do disengage. In contrast, activity theory states that active and involved aged people are happier and better adjusted than inactive uninvolved elderly people. This theory predicts that elders remain engaged. It also suggests that higher levels of engagement will be related to lower levels of depression. I then present critiques of theories of engagement. In doing so, it is important to note that theories of engagement, while seemingly oppositional theories, have much in common. First, they are both interested in one variable (engagement) in predicting well-being. Critics note that using one variable to explain well-being in aging is insufficient. Rather, it is important to consider the kinds of interactions elders engage in. Indeed, research on stress notes that the quality of interactions is most important in predicting depression. That is, extremely negative interactions (e.g. conflicted relations) are most predictive of depression than are positive interactions. In addition, research on stress notes that physical stress (pain) is also highly predictive of depressive symptoms. Given that pain is a frequent occurrence in nursing homes, these critiques suggest that it should be considered as an important predictor of depression. Second, feminists argue that because gender is a fundamental organizing principal across the life span, men and women in old age have different amounts of social and economic resources (Ginn and Arber 1995; Smith 1987; Hess 1985). Therefore, it is important to consider gender differences both in levels of engagement and in the impact of engagement on depression. In the next section, I present research that examines levels of engagement in the nursing home and how engagement impacts depression related experiences. I also include research that examines the critiques of theories of engagement, research on the relationship between stress (conflicted relations and pain) and depression. Finally I examine gender differences in levels of engagement, levels of stress, and in the impact of engagement and stress on depression. Because

16

many of the relationships are unexamined, or under examined in the nursing home, I include literature from the community dwelling aged. Theories of Engagement Engagement theories represent two views of optimal aging. The theories are two ends of a continuum that uses a single variable (engagement) to predict optimal aging. While the theories were designed to explain optimal aging, optimal aging is a vague category. It has been variously conceptualized as well-being, life-satisfaction, robust aging and a variety of other positive indicators (Moen 1996). In this project, I, like other studies (e.g. Johnson and Barer 1992) am testing the applicability of these theories in explaining the opposite of well-being, the complete lack of well being by using symptoms of depression as the outcome variable. Activity theory. Activity theory (Havighurst, Neugarten and Tobin 1968) argues that activity and involvement is “normal” and that most people remain involved in social life in old age. Activity theorists believe that activity is not only normal but also necessary for a healthy adjustment in old age. “Idleness, not old age, hastens illness and decline” (Havighurst et al. 1968:161). Activities of middle age should be maintained for as long as possible. Substitutes should be found for activities that have to be relinquished (Havinghurst et al. 1968). Activity theorists focus on the importance of engaging in activity and maintaining social interaction for the maintenance of a positive self-concept and self evaluation. According to Rowe and Kahn (1997), engagement, “maintaining close relationships and remaining involved in activities that are meaningful and purposeful,” (p. 46) is essential for successful aging. Influenced by Erikson (1950), activity theory asserts that the capacity to respond to social overtures from others and to initiate meaningful social involvement is an important aspect of human functioning. That is, engagement in life is seen as a necessary part of the development of a mentally healthy person in old age. The loss of social interaction involved in the loss of roles results in a deterioration in self concept and a decline in self-esteem, life satisfaction, and morale (Larson 1978). The more roles people perform, the greater the ability to maintain positive thoughts about some aspects of self and the greater their resources for trading off negative aspects of each role (Thoits 1983). Activity theory (Cavan et al. 1949) claims that as one experiences disability and other age-related declines, social roles may become unattainable. As a way to preserve their self-identity in the face of deficits, aging adults will replace lost activities and roles with new compensatory activities and roles thus becoming re-integrated (Atchley 1998). This suggests that elders remain “engaged” with life. The theory also suggests that disengaged elders will be more likely to have depression related experiences than engaged elders. Disengagement theory. In contrast, Disengagement theory (Cumming and Henry 1961) states that normal aging “is an inevitable, mutual withdrawal…where many of the relationships between an aging person and other members of society are severed, and those remaining are altered in quality” (p. 227). The theory is developmental, with old age seen as a separate phase. Working out of the functionalist paradigm, disengagement theory was an attempt to connect micro individual level theory with macro system level theory. Functionalists posited that the gradual withdrawal of the individual from the system allows for smooth and efficient functioning

17

for the individual and the system (Parsons 1951). That is, it is functional at both the micro and macro level for the individual to withdraw. At the individual level, disengagement theory (Cumming and Henry 1961) posits that the process of aging is an “adaptive narrowing of the older person’s social world and decreasing investment in activities and social relationships” (p. 227). It is functional for the individual who has decreasing ego energy to withdraw in order to adapt to impending death (Cumming and Henry 1961). In this way, growing old involves a gradual and inevitable mutual withdrawal “resulting in decreased interaction between an aging person and others in the social systems he belongs to” (Cummings and Henry 1961:14). Elderly people do this by reducing the number of roles, lessening the variety of roles and relationships and weakening the intensity of those that remain. They shift towards greater autonomy and decreased connection with others. In doing so, disengagement theory suggests elders’ withdrawal involves a fundamental change in both pattern and level of activities. Individuals who disengage entirely drop a large number of their previously customary activities and their overall activity level also drops dramatically (Johnson and Barer 1992). Disengagement can be a beneficial adaptive strategy that enables the individual to conserve energy (Johnson and Barer 1992). In addition, “[b]ecause interactions create and reaffirm norms, a reduction in the number and variety of interactions leads to an increased freedom from control of the norms governing everyday behavior” (Cumming and Henry, 1961:213). As such, disengagement, by enabling the aging person to be relieved of social expectations and constraints, can be satisfying and enhance well-being (Johnson and Barer 1992). At the system level, disengagement theory postulates that it is functional for society that the elderly slowly withdraw. The mutual process of disengagement makes room for younger generations and prevents unnecessary disruptions in the social system caused by deaths among the older population (Cumming and Henry 1961). For example, disengagement allows retirement from roles which young people may fill. According to Cumming and Henry (1961), the process of disengagement is universal and inevitable, although variations in timing and style occur according to the individuals’ psychology, personality, initial type of engagement and life situation. In its original form, Cummings and Henry (1961) explicitly argued against the notion of constraint of the individual by social structure. Disengagement theory postulates that it is natural and inevitable to disengage in old age in response to changing health and social losses. Further, because the theory postulates that disengagement is an adaptive strategy for elderly, it suggests that elderly people who are disengaged will have less depression related experiences than elders who are engaged.

Critiques of Engagement Theories Individual level theory. It is important to note, as mentioned earlier, that while seemingly opposing theoretical positions, activity and disengagement theory have many similarities. Critics argue that while both theories attempt to incorporate micro and macro levels of analysis, they fail “because of the explicit rejection of any notion of social structure” (Marshall 1996:12). Disengagement theory assumes that the disengagement process is an adaptive strategy, initiated by the individual, and that disengagement benefits the individual. Cumming and Henry (1961)

18

note that “[i]t is our assumption that the individual has access to the whole culture, directly or indirectly and furthermore, that he exercises some freedom of choice in selecting his contacts” (p.12). However, the process is far from voluntary in many cases, and is instead initiated by social structure. Age itself can determine people’s social roles “independent of their capacities and preferences” (Moen 1996: 171). Societal attitudes about what is appropriate for the aged affect the extent that elders can remain engaged (Laborsky 1994; Sill 1980; Roman and Taretz 1967). People can also be forced into disengagement by lack of opportunity that results from physical limitations (Atchley 1998). In this way, some disengagement is mandated by structural arrangements and may not necessarily enhance well-being. These problems are perhaps especially the case in the nursing home where opportunities for engagement are limited (Simpson, Woods and Britton 1981). The impact of social structural arrangements that limit what people can do physically and have available for them to do structurally is ignored in disengagement theory. Indeed, in the forward to Lives through the Years (Williams and Wirths 1965), Parsons noted that it was important to take into account structural arrangements. “The location of individuals in different social contexts appeared to lead to differential interaction between persons and context, which in turn lead to differential successful life styles” (p. 3). Similarly, activity theory fails to acknowledge how the focus on maintaining activity may be more important for maintaining social structure than for the individual. Katz (2000) points out that many of the indicators of activity reflect dominant ideals. For example, the narrow focus on activities and relations designed for individual adaptation and satisfaction, in practice, shows a focus on “productive engagement”. This focus reflects an assumption that an individuals’ worth is primarily defined through work and accumulation (Minkler 1984; Estes 1983). Particular kinds of activities that are not “productive” (e.g. napping, watching television, etc.) have been excluded from activity lists even though researchers admit that “some portion of television watching is unquestionably active and stimulating” (Lawton, Moss and Duhamel 1995). Stress. Second, theories of engagement focus on the quantity of residents engagement (e.g. engagement levels), ignoring the negative side of interaction. However, ample evidence suggests that interactions can be a source of stress as well as support (Revenson et al. 1991). Indeed, there is an abundance of research on stress that shows that the “absence of negative social interactions may be more important for mental health than the presence of supportive interactions” (Rook 2001; 1992; 1984; Manne et al. 1997; Krause 1995; Schuster, Kessler and Aseltine 1990). Researchers find that stressful interactions, in the form of conflict in relationships, are more important than levels of engagement in predicting depression (Utz et al. 2002; Derogitis 1986). These findings indicate that it is important to take into consideration whether or not engagement is positive (supportive) or negative (stressful or conflicted). In addition, research on stress can be used to make sense of the relationship between pain and depression. Research on stress explores “pathways though which stress exposure affects psychological well-being” (Lennon 1989). Stress results from the presence of socioenvironmental demands that tax individuals’ ordinary adaptive capacity (Lazarus 1966, Pearlin 1983). Literature on stress notes that “[e]xternal circumstances that challenge or obstruct are labeled stressors, stress refers to internal arousal” (Anehensel 1992:16). As such, “[s]tress is not an inherent attribute of external conditions” (Anehensel 1992:16). Rather, it is experienced by the individual as “stressful”. Stress research emphasizes the social context of stress. What is defined as stressful for some individuals in some contexts will not necessarily be the same for

19

others. “Socio-environmental conditions differ in the capacity to evoke stress, however; some conditions threaten virtually everyone, whereas others are uniformly navigated with ease”(Anehensel 1992:16). In this way, defining what is stressful is related to values (Lazarus and Folkman 1984). Conditions are stressful, when they are threatening. However, the threat that people perceive is closely related to the values they hold, what they define as important, desirable, or to be cherished. Research on stress emphasizes that it is important to be critical of what may be included under the label of “stress” and recognize that what we define as “stressors” is dependent upon what we value. When examining the relationship between stress and depression in the nursing home this is important because what researchers and the general population may define as valuable may not be what residents’ define as valuable. Researchers argue that pain is particularly important as an indicator of stress because individuals experience the event as out of their control (Lennon 1989). Mirowsky and Ross (1990) note that stimuli experienced as out of the individual’s control is particularly stressful. In addition, others note that pain is particularly stressful because it interferes with all aspects of living. It “demands the sufferer’s attention, drains the person of energy, and preoccupies them with the meaning of pain” (Katz et al. 1996:252). Because I am particularly interested in how engagement impacts depression, I focus on the possibility that certain aspects of engagement (interactions) may be stressful (conflicted) and therefore contribute to depression. Therefore, I explore negative interactions in the form of conflicted relations as a form of stress. I examine the idea that levels of engagement are not sufficient in assessing depression, but rather it is important to examine the quality of engagement. I also examine how the stress of pain may contribute to depression in the nursing home. Gender. As mentioned earlier, Feminist theorists critique engagement theories for their lack of attention to gender. Critics argue that “gender is a major dimension of stratification, and that there are significant differences in the way aging affects men and women; aging is a gendered process” (Arber and Ginn 1991:2). Feminists argue that it is important to rethink social theory, “explicitly taking gender relations into account” (Moen 1996:176). Theories of engagement assume that men and women have the same opportunity to engage in activities and roles. The theories also assume that engagement in activity and roles has the same meaning for men and women. Given that experiences vary by group membership, engagement theories, feminists argue, are insufficient when they fail to include gender relations, or the experience of women or other social groups in the context of aging (Blieszner 1994; Reinharz 1986). Studies of engagement in activities and social interactions should take into account that engagement will take different forms and have different meaning for different groups. In this way, including gender does not simply mean adding gender to mainstream theories. Rather, it provides ways to rethink theories of engagement. Specifically, taking gender relations into account addresses the lack of attention to structural forces by both disengagement and activity theorists. Focusing on group experiences, feminist theories provide models to connect the individual to larger social structural variables by examining how individual life experiences are patterned by gender. This links micro and macro levels by addressing both structural and individual levels of theory (Bury 1995; Lopata 1995). Examining gender differences is important not only to give a more accurate description of reality, but also to expand the scope of individual level theories to include structural elements. In this case feminist theory shows that it would be important to pay attention to the possibility that the kinds of engagement measured are “gendered” (particularly salient for men or women).

20

This may affect levels of engagement as well as the impact of engagement on depression related experiences. For example, if we are measuring residents’ level of engagement in crafts versus sporting events, there may be gender differences in levels of engagement and in the impact of engagement on depression. This is particularly true in the nursing home given that opportunities for engagement are limited by what is offered. Conceptual clarity. Finally, critics argue that there is insufficient conceptual clarity to prove or disprove the core premises of either theory of engagement. Indeed, one review of this work concluded that engagement is the least well-conceptualized variable in health studies (Guadagnoli and Mor 1991). As researchers note, it is difficult to determine just what constitutes engagement with life (Lennertsson and Silverstein 2001; Klumb and Baltes 1999). Researchers note that “[d]eveloping a valid and reliable scheme for appraising the salient domains of leisure and productive activities is challenging” (Klumb and Baltes 1999:176). Engagement has been variously conceptualized with such differing and overlapping indicators7 as engagement in social activities, productive activities and fitness activities (Glass, Mendes de Leon, Marottoli et al 1999). These concepts have included various combinations of measurements such as attending classes, attending church or temple, volunteering, (Everard, Lach and Fisher 2000; Glass et al 1999), traveling, entertaining, attending parties, shopping, cooking, paying bills, and doing housework (e.g. Everard et al.2000; Horgas et al. 1998). Leisure activities that are physically demanding have also been used to measure engagement (Seeman, et al. 1995; Simonsick et al. 1993). Some studies use physical activity, sometimes called high-demand activities (swimming, woodworking, walking, and gardening) (Everard et al. 2000; Glass et al. 1999; Pennix et al. 1999). Others study frequency of activity (walking, gardening, and vigorous exercise) (Pennix et al. 1999), solitary activity, sometimes called low demand leisure activity (sewing, reading, watching television, listening to music, hobbies, and art), (Lennertsen and Silverstein 2001; Everard et al. 2000) and social cultural activities (going to movies, theaters, concerts or museums, eating out at restraints, and participating in study groups) (Ljungquist, Berg, and Steen 1996; Wolinsky et al. 1995). Engagement is also measured by social participation variously conceptualized as organizational affiliations, friendship ties, kinship networks, social connectedness, social support, and social integration (Ferraro 1984; Anderson 1983; Bankooff 1983; Thoits 1983; Vachon, Rogers, Lyall et al. 1982; Bahr and Harvey 1980; Durkheim 1951). Problems with defining engagement are particularly true of research on nursing homes. Interactions in the nursing home are complex and adjustment to life in the nursing home is difficult (Gubrium 1991; Shield 1988). While considerably less studied in the nursing home, there are also a variety of measures of engagement applied to this setting. Research has examined social support (Wells and MacDonald 1981), the number and stability of social support

7

For example, some studies define solitary activities in terms of reading or watching television (Longino and Kart 1982; Litwin 2000: Zimmer, Hickey, and Searle 1995), others include participation in hobbies and gardening in their definition (Lennartsson and Silvestein 2001). Yet other researchers have classified activities like gardening or housework as productive activities (Glass et al. . 1999, 1995). In addition, there have been a variety of research projects that focus on the social relational aspect of engagement. In these studies engagement is defined by social relations, including seeing friends or neighbors, talking with others on the phone, helping relationships, and household and kin relationship (Sabin 1993), number of social roles occupied, (George 1995; Richardson and Kilty 1991; Atchley 1989), social interaction and ties (Brook et al. . 1983), and frequency of contacts with relatives, friends, and others (Willey and Siliman 1990).

21

networks (Bitzan and Kruzich, 1990; Wells and MacDonald 1981), and activity levels and friendships (McKee, Harrison, and Lee 1999). Recognizing the need for conceptual clarity, Mor, et al. (1995) have taken up the challenge of formalizing a concept of engagement in the nursing home. Indeed, the purpose of their study was to introduce and describe the validity and reliability of a new measure of social engagement for a nursing home population that was “structurally distinct from measures of depression and anxiety, conflicted relationships, or problematic behaviors” (Mor et al. 1995:1). Theoretically, the measure reflects both social involvement and autonomy in the nursing home. Mor et al (1995) assert that social engagement includes an “ability to take advantage of opportunities for social interaction, limited as they may be in many facilities” (p 1). Further, it includes an ability to initiate actions that “engage residents in the life of the home” (Mor et al. 1995:1). The index measures whether or not the resident is at ease interacting with others, at ease doing planned or structured activities, at ease doing self-initiated activities, establishes own goals, pursues the life of the facility, and accepts invitations into most group activities. While an important step towards conceptual clarity, the measure leaves out many of the aspects of engagement that have been found important in the community. Namely, the measure does not examine family relations, engagement in social roles, change in engagement, and levels of contact with family and friends. The scale also does not leave room for measures of interactions as stressors, or for determining which aspects of social engagement (positive versus negative interactions) better predict depression related experiences. In the community these distinctions have been shown to be important in predicting depression. While a lack of conceptual clarity is a problem in the study of community elders, too much conceptual clarity and therefore limitation, may be a problem in the nursing home. In summary, the meta-construct of engagement with life and its role in promoting successful aging is challenging (Menec 2003; Lennartson and Silverstein 2001). It is difficult to operationalize such a complex set of possible activities that elders engage in and thus to develop valid and reliable indicators for examining the salient domains of activities. However, Lennartson and Silverstein (2001) note that “careful specifications of activity involvement will permit a more informative account of the mechanisms…that link activities to enhanced wellbeing” (Lennartson and Silverstien 2001:341). They suggest that it is important to develop broad conceptualizations of engagement with life “that combine theoretical rigor and desirable measurement properties” (Lennartson and Silverstien 2001:341). Therefore, it is important to measure several aspects of engagement in order to properly understand elder engagement and determine its impact on other outcomes. Summary Activity theory states that active and involved aged people are happier and better adjusted than inactive uninvolved elderly people. This theory suggests that disengagement is associated with an increase in depression related experiences. Disengagement theory states that normal aging is a gradual process of inevitable disengagement from society by the aging individual and from the individual by society. As such, this theory hypothesizes that reduction in levels of disengagement is normal and elders who do not “properly” disengage will have higher levels of depression than elders who disengage. Critics of theories of engagement emphasize that it is

22

important to consider the quality of engagement. This perspective suggests that stress in the form of negative interactions (conflicted relations) and pain will be more important than levels of engagement in determining who is depressed and who is not depressed. Finally, feminist theory argues that men and women will have different levels of engagement and stress. Feminist theory also suggests there will be gender differences in the relationship between engagement, stress and depression. My aim is to determine if these explanations hold true in the nursing home. In the next section, I present studies8 of engagement and depression. I also present evidence for critics who suggest stress and gender will be important considerations. I concentrate my review on five aspects of engagement, “social engagement” (measures of interacting with others, doing planned or structured activities, self initiated activities, establishing own goals, involvement in group activity), “activity time,” “reduced engagement” (reductions in interactions and activities), “strong identification with past roles,” and “absence of contact”. I include research that examines gender differences in levels of these aspects of engagement. I also include research that examines gender differences in the impact of aspects of engagement depression. I then review literature that examines the impact of stress (“conflict” and “pain”) on depression. Finally, based on this review, I formulate hypotheses about the relationship between aspects of engagement, stress, gender, and depression among my sample of nursing home elderly. Research on Aspects of Engagement, Gender and Depression Social Engagement Index (SEI). Using the MDS SEI, (dichotomous measures of whether or not the resident is at ease interacting with others, at ease doing planned or structured activities, at ease doing self-initiated activities, accepts invitations into most group activities, establishes own goals, and pursues involvement in the life of the facility), findings suggest that disengagement is the rule rather than the exception (Gilbart and Hirdes 2000; Horgas et al. 1998; Resnik et al, 1997; Schroll et al. 1997; Mor et al. 1995). For example, Mor et al. (1995) found that average levels of engagement were about two on a six point scale where higher levels indicate greater engagement (Mor et al. 1995). The only item in the engagement scale in which over half (59 percent) were assessed positively was “at ease interacting with others.” Approximately a third of the sample was assessed as “at ease doing planned or structured activities” (38 percent), “at ease doing self-initiated activities” (33 percent), and “accepts invitations into most group activities” (35 percent). Fewer (28 percent) were assessed as establishing their own goals. Fewer still (23 percent) were assessed as pursuing involvement in the life of the facility. Schroll, et al. (1997) had similar results: 58 percent were at ease interacting with others, 33 percent at ease doing structured activities, 38 percent at ease with selfinitiated activities, 30 percent established their own goals, 19 percent pursued facility involvement, and 24 percent accepted invitations to most group activities. Likewise, Resnick et al. (1997) found that 68 percent of the sample exhibited two or fewer measures of social engagement. In a study of nursing homes in the Netherlands, Achterbert et al. (2003) found that 51 percent of newly admitted residents had a low level of social engagement. Resnick et al. 8

I present studies of the nursing home elderly where available. However, because there is a noticeable gap in the literature on the nursing home elderly with regards to my questions, I examine research on the community dwelling aged as well.

23

(1997) found that 68.3 percent of residents scored zero to two on measures of social engagement and 31.6 percent scored three to six. However, much of the disengagement associated with the SEI can be explained by the fact that the nursing home elderly are more likely to have health problems than elders in the community. Research on the relation between levels of engagement and health in the nursing home has examined health measures that include ADL limitation, cognitive limitation, hearing, vision, and communication. These studies find that residents with physical impairment (ADL status) and cognitive impairment have lower levels of engagement as measured by the SEI (Schroll et al. 1997; Mor et al. 1995). When residents are divided into groups based on their level of cognitive and ADL impairment, those relatively high in both cognitive and ADL functioning have the highest levels of SEI (Schroll et al. 1997; Mor et al. 1995). In contrast, those that are low on both cognitive functioning and ADL functioning have the lowest levels of SEI (Schroll et al. 1997; Mor et al. 1995). Interestingly, Mor et al. (1995) found that residents relatively high in cognitive functioning but low on ADL functioning had a highly skewed distribution on social engagement. Seventeen percent scored zero, while 20 percent scored four or higher. There have been no studies of gender differences using the SEI in the nursing home. However, there is some evidence for gender differences when examining indicators used in the SEI in the community dwelling aged. First, levels of engagement in activity differ for elderly men and women (Bennett 2002; 1998; Glass et al. 1999; Lefrancois et al 1998; Crespo et al. 1996). Crespo (1996) finds gender differences in participation in walking, shopping, indoor, outdoor, and leisure activities. “Within these cohorts of older people traditional gender roles continue to exert a strong influence on levels and types of habitual physical activity well into later life” (Crespo et al. 1996). Gender differentials were maintained over an eight year period, with women showing higher levels of activity participation indoors and men showing higher levels of activity participation outdoors (Bennet 1998). In examining older adult friendships studies show that among the oldest groups (75 to 84 years old), men had fewer friends on average than women (Adams and Blieszner 1993). In addition, among their close friend networks, between half and all elders in the study had same sex friends. This is particularly important when thinking about friendships in the nursing home because the nursing home population is predominately female. However, studies of the community dwelling aged using other indicators included in the nursing home SEI find that men are more likely to engage in self initiated acts and establish their own goals (Glass et al. 1999). In the Termen study of gifted adults (mean age 70), men rated achievement goals as more important than did women whereas women rated involvement in relationships as more important than did men (Holahan and Sears 1995). With regard to the impact of SEI on depression, Activity theory postulates that as levels on the SEI increase, levels of depressive symptoms will decrease. Disengagement theory suggests that as levels on the SEI increase, depressive symptoms will also increase. Feminist theory contends that because men and women experience engagement differently, there will be gender differences in the impact of SEI on depressive symptoms. Most research supports activity theory with regards to the SEI. Research finds that those with higher scores on the SEI have lower scores on the MDS-DRS, controlling for low cognitive performance, impairments in vision, and ADL (Achterbert et al. 2003). Mor et al. (1995) and Gilbart and Hirdes (2000) report correlations between social engagement and mood items: (1) expresses sadness/anger/empty feelings over lost roles/status; (2) verbal expressions of distress

24

by the resident; (3) tearfulness, emotional groaning, sighing, breathlessness; (4) sad or anxious mood intrudes on daily life over the last 7 days; (5) motor agitation; (6) failure to eat or take medications; (7) pervasive health concerns; and (8) recurrent thoughts of death. However, Brown et al. (2002) found that there were no significant differences in scores on the SEI among residents diagnosed with depression compared to those not diagnosed. Research on elders in the community finds similar results when examining some indicators used in the SEI. For example, Holahan and Chapman (2002) found that goal setting behaviors (setting their own goals) is positively related to psychological well being in a study of 74 to 79 year olds. Similarly, other research finds that commitment to life goals has positive implications for psychological well-being (Brunstein, Schultheiss and Maier 1999; Brikman and Coates 1987; Emmons 1986) and is negatively related to depression (Holahan 1998; 1988; Holahan and Sears 1995; Brandstadter and Renner 1990). While most research on the relationship between engagement and depression supports activity theory, studies of elderly in the community that include the very old (over 85) have found that levels of engagement do not impact depression (Ritchey et al 2001; Davies and Gledhill 1983; Johnson and Barer 1992). In addition, other research suggests that there may be more support for disengagement theory for some unhealthy elders. Mor et al (1995) finds that for the best functioning group, the SEI is negatively related to mood items. On the other hand, for the most cognitively and physically impaired residents, SEI is positively associated with mood problems. Finally, for the mixed group, those who are high in cognition but low in ADL, there is no relationship between SEI and mood problems (Mor et al. 1995). Since there have been no studies of gender differences in the impact of the SEI on depression in the nursing home, I examine studies in the community that use similar indicators. There have only been a few studies that examine gender differences among the elderly in some of the indicators of social engagement on depression. Men’s participation in outdoor activities decreases their levels of depression more so than it does for women (Bennett 1998). In a study of sixty-eight residents of a retirement center, depression was found to be related to fewer and less satisfying interpersonal relationships for women and lack of involvement in activities for men (Hale 1982). However, other research has examined gender differences in the impact of engagement on well-being, mortality, self esteem and life satisfaction. Research in the community finds that social integration is especially important in women’s lives compared to men’s (Lennartson and Silverstien 2001; Reitzes, Mutran and Verrill 1995; Lee and Shehan 1989; Field and Minkler 1988; Gilligan 1982; Keith et al. 1983). For example, Lee and Shehan (1989) found that participation in voluntary associations was related to women’s but not men’s well being. A study of middle age working men and women (58 to 64 years old) found that activities with relatives and activities with friends from work have a positive effect on selfesteem for women but not for men. In addition, activities performed alone have a positive influence on the self esteem of men but not on the self-esteem of women (Reitzes et al. 1995). Other research has found that there are gender differences in the impact of activity on functional health, with women benefiting from social activities and men from physical activities. (Everard 1999). It is important to note that studies that find gender differences in levels of engagement and in the impact of aspects of engagement on depression are of the relatively young old (below 85). For example, while Crespo et al (1996) asserts that gender differences are stable across age groups, their study only examines the relatively young old (those below 81). Studies that include

25

the very old find support for the contention that gender differences reduce over time (Bennet 1998; Holahan and Sears 1995). For example, Holahan and Sears (1995) find that for respondents over 80, both men and women place less emphasis on achievement and competition whereas relationships continue to be important in the oldest age groups. This indicates a lessening of gender differences among the very old. Based on this literature, I expect to see the following relationships. First, I expect that residents will have low levels on the SEI. Because the indicators of engagement that particularly appeal to men (e.g. goal setting and self initiation of activities) are particularly restricted in nursing homes, I expect men to have significantly lower levels of on the SEI than women. Given the preponderance of evidence, I expect that as SEI levels increase, depression will decrease. However, it will be important to control for health measures to explore this relationship. In addition, there is ample evidence to suggest that SEI will impact women’s depression related experiences more so than men’s in my sample. Activity time. Research examining time engaged in activities also finds that nursing home residents are highly disengaged (Horgas et al 1998; Resnik et al. 1997; Schroll et al. 1997; Mor et al. 1995). Eight percent of nursing home residents participate in activities most of the time, 42 percent some of the time, 44 percent little time, and 7 percent none of the time (Schroll et al. 1997). Almost half spend no more than one-third of their time involved in activities (Resnick et al. 1997). There has been only one study of gender differences in activity time in the nursing home. Voelkl et al. (1995) found that women spend significantly more time involved in activities than do men. However, there has been more research in the community dwelling aged. Research in the community finds that elderly men and women (aged 60 to 80) have significantly different levels of time spent in leisure activities. In a longitudinal study of leisure time physical activity (LTPA) Crespo et al (1996) found that men were consistently more likely to engage in LTPA than women. They examined, jogging, swimming, aerobic dancing, other dancing, calisthenics or floor exercises, gardening or yard work, and weight lifting. Elderly women showed higher levels of activity indoors. Men showed higher levels of activity participation outdoors. With regard to the impact of activity time on depression, activity theory suggests that as activity time increases, levels of depressive symptoms will decrease. Disengagement theory postulates that as activity time increases, depression will also increase. Feminist theory points out that it is important to consider gender differences in the impact of activity time on depressive symptoms. Although these relationships have not been studied in the nursing home most research in the community dwelling aged supports activity theory. Research finds that the more frequently elders engage in social interaction the less likely they are to have depressive symptoms (Lennartsson and Silverstein 2001; Dean et al. 1992). Similarly, in a study of 65 to 74 year old community dwelling aged, Everard (1999) found that total number of activities was positively related to well-being. Pennix et al. (1999) found that the amount of time spent in physical activities (walking, gardening, and vigorous exercise) decreased elders likelihood of having depressive symptoms (CES-D depression scale). Frequency of participation in social groups and in solitary activities (handwork, hobbies, music, and reading) was positively related to wellbeing (Menec 2003). However, other studies that include several measures of engagement along with activity time have found that activity time is not a significant predictor of depression (Everard 2000).

26

Gender differences in the impact of activity time on depressive symptoms have not been studied in the nursing home, but research does show gender differences in the community dwelling aged. In a sample of 1042 over 65 year old women and men Morgan et al. (1991) found clear differences in the organization of activity patterns (Morgan et al. 1991). Among women activity time failed to predict better mental health. Morgan et al. (1991) hypothesized that these gender differences may be attributed to the fact that older women bear more responsibility for housework and are less able to select their activities as freely as men. Given these results, I expect that residents will spend less than half of their time engaged in activities. Given that measures of activity time are limited to indoor activity, I also expect that men will be less likely to engage in activities than women. I also expect that elders with higher levels of activity time will be less likely to have depression than those who have low levels of activity time and that there will be gender differences in the impact of activity time on depression. Reduced engagement. There have been no reports of studies of change in (reduced) engagement in the nursing home. When examining the community dwelling aged, most research finds elders generally maintain the same levels of engagement in social interactions and in customary activities as they did earlier in life (Bukov, Maas, and Lampert 2002; Atchely 1998; 1989). Bukov et al. (2002) found that the best predictor of participation at one point was participation levels at earlier time periods. Research on activities has consistently found evidence for continuity across the life cycle with regards to participation in relationships and activities (Iso-ahola, Jackson and Dunn 1994). For example, Bukov et al. (2002) found in a longitudinal study of social participation that although activity domains change, individuals maintain a constant level of participation. However, these are studies of the relatively healthy and young old. Studies of the old-old (over 85) find support for disengagement theory. Bukov et al. (2002) finds that after age 90, levels of social participation decreases. Indeed, 30 percent of the respondents were socially inactive. Similarly, Johnson and Barer’s (1992) findings indicate that both physical and social losses make it difficult for the elderly to maintain an active social network. “Almost one-half of the sample intentionally remained socially engaged by making new friends and attending senior centers. The remainder, who are generally older and in poorer health, “modify their social world both psychologically and socially” (Johnson and Barer 1992:352). There have been no studies of gender differences in changes in engagement in the nursing home. However, there is evidence that there are gender differences in changes in (reduced) engagement in the community dwelling aged. In a study of 601 individuals aged 65 and older (mean age 74.9), women had rates of decline in social participation and outdoor recreation, while men had rates of decline in exercise, sport, and travel (Lefancois et al. 1998). Other research has found that women’s interests were more congruent than was men’s. Women tended to show a continued emphasis on similar activities in later aging. Although both elderly men and women showed little change between the time periods, women were less likely than men to change activity level (Holahan 2002). Other research has found declines among the community dwelling aged in family contact in men, but not women among the very old (75 to 84 years old). However, these differences became insignificant among the old-old (85 to 93 years old) (Field and Minkler 1988). In a longitudinal study Field and Minkler (1988) found that declines beyond family contact were observed for men but not for women in the very old but not the old-old.

27

With regard to the impact of reduced engagement on depression, disengagement theory suggests that a change in engagement (reduced) will be associated with a decrease in depressive symptoms. In contrast, activity theory postulates that elders who have reduced engagement will have higher levels of depressive symptoms than those who do not. There have been no studies of the impact of change (reduced) engagement among the nursing home elderly. However, in general, research in the community dwelling aged supports activity theory. Morgan and Bath (1998), and Atchley (1998) found that reductions in customary physical activity were associated with lower levels of well-being. However, this research was on the fairly young old (under 85). Johnson and Barer (1992) found that those elders (over 85) who disengaged were not significantly more likely to be distressed than those who remained engaged. Rather than being a maladaptive strategy, as activity theory would predict, Johnson and Barer’s (1992) results indicate that disengagement is psychologically beneficial for the very old. Disengaged elders “redefine their optimal level of social integration and become content with a narrower, more constricted social world” (Johnson and Barer 1992:352). Second, they are likely to “reject those norms which place expectations on them that are incongruent with their capacities” (Johnson and Barer 1992:352). There have been no studies of gender differences in the impact of change in engagement on depression in the nursing home or community dwelling aged. Based on this literature, I expect that there will be very little change (reduced) engagement in the nursing home. I expect those who have reductions in engagement to have more depressive symptoms. While some research has suggested differently (e.g. Johnson and Barer) this research measures reductions over a long period of time (one year). This research also suggests that it will be important to examine gender differences in reduced engagement. In addition, given that there are gender differences in the impact of other aspects of engagement on depression, it will be important to test for gender differences in the impact of reduced engagement on depression in my sample. Absence of contact. There has been no research that reports levels of contact with family and friends among nursing home residents. However, among the community dwelling aged, most elderly have contact with their family and friends (Commerford and Reznikoff 1996; Antonucci 1990; Johnson and Barer 1992). Similarly, there has been no research that reports gender differences in family and friend contact in the nursing home. However, there is considerable research on gender differences in family and friend contact in the community dwelling aged. Feminist theorists’ contention that gender matters in old age is well supported by research on the community dwelling aged. Most studies find that women have higher levels of social contact than men (Fuhrer and Stansfeld 2002; Commerford and Reznikoff 1996; Antonucci 1990). Elderly women tend to have more people in their social networks, to have more family and friends, to have more varied relationships with different types of people, more frequent contact with network members, and receive support from more sources than men (Fuhrer and Stansfeld 2002; Commerford and Reznikoff 1996; Antonucci 1990). With regard to the impact of absence of contact on depressive symptoms, disengagement theory suggests that an absence of contact with family and friends will not be related to depressive symptoms in the nursing home. Activity theory postulates that those who lack contact will have significantly more depressive symptoms than those who do not. Feminist theory points out that there will be significant gender differences in the impact of absence of contact on depressive symptoms.

28

Research on the nursing home elderly supports activity theory. Loneliness, isolation, and lack of social support from family and friends has been found to predict depression (NIMH 2001; Ulbrich and Bradsher 1993) Similarly when measuring well-being, Krause, Lang and Keith (1990) found that elders who had more contact with friends and family had higher levels of well-being than those with low levels of contact. In a study of nursing home residents, Commerford and Reznikoff (1996) found that social support from family was negatively related to depressive symptoms (Beck Depression Inventory 1961). They found that maintaining previous family ties is a significant predictor of psychological adjustment in nursing home populations. Greene and Monahan (1982) found that higher visitation frequency was a significant predictor of lower levels of psychosocial impairment. Fessman and Lester (2000) found social relationships formed with other residents and social relationships with friends and family predicted feelings of depression and loneliness in the nursing home. Residents who reported that they were isolated from other residents, family, and friends were more likely to have feelings of depression than residents who did not feel isolated (Fessman and Lester 2000). There has been no research on gender differences in the impact of family and friend contact on depression in the nursing home. Among the community dwelling aged, some research has found social integration to be especially salient in women’s lives compared to men’s (Bernard 1972). However, other research reached different conclusions. In a study of community dwelling aged (55 to 85 years old) Sonnenberg et al. (2000) found low levels of emotional support was associated with CES-D depression symptoms, but the effect was stronger for men than it was for women. Similarly, Matt and Dean (1993) found that among the old-old, men were more vulnerable to psychological distress than women when they lacked contact. Based on this research, I expect that there will be very few residents who lack contact from family and friends. I also expect that elders who lack contact with family and friends will have more depression than those who do have contact. With regard to gender, I expect that men will be more likely to have an absence of contact than women in my sample. It will be important to test for gender differences in the impact of absence of contact on depression. Attachment to past roles. A description of the frequency of attachment to past roles has not been studied in the nursing home or community dwelling aged. Research has noted that both older men and women typically experience a decline in roles as they age (Elder and Pavaloko 1993; Menaghan 1989; Moen et al. 1992; Smith and Moen 1988). However, Atchley (1999) found little evidence that elders living in the community were preoccupied with the past. There have been no studies that report gender differences in strong attachment to past roles in the nursing home or community dwelling aged. However, studies of gender differences in role attachment among the community dwelling aged finds that “being male or female shapes the nature, organization and patterning of roles, resources, relationships and relevant identities throughout the life course” (Moen 1996; 177). Given that men’s roles are typically associated with higher status, it would seem reasonable to assume that men would be more likely to have strong attachment to past roles. With regard to the impact of attachment to past roles on depression, disengagement theory suggests that those who have a strong attachment to past roles will have significantly more depressive symptoms than do those who do not. Activity theory also agrees that strong attachment to past roles will engender depression because this theory acknowledges that role loss

29

is inevitable9. Feminist theory would argue that there will be gender differences in the impact of strong attachment to past roles on depression. There has been no research on the impact of strong attachment to past roles on depression in the nursing home or community dwelling aged. Research does find that elders experience significant role loss and that role occupancy is important for maintaining quality of life (Atchley 1989; George 1989; 1978; Roscow 1985; 1974; Blau 1973; Lemon, Bengtson and Peterson 1972). Therefore, if nursing home elders have a strong attachment to past roles, roles they can longer access as a resource to ward off psychological distress, they should be more likely to have depressive symptoms than those who do not have a strong attachment to past roles. There have been no studies of gender differences in the impact of strong attachment to past roles. Based on this literature, I expect that most residents will not have a strong attachment to past roles and that men in my sample of nursing home elderly will be more likely to have a strong attachment to past roles than women. I also expect to find support for the assertion that those who have a strong attachment to past roles will have higher levels of depressive symptoms than those who do not. In addition, because men’s past roles have higher status than do women’s (Holahan and Chapman 2002) it makes sense to presume that men’s strong attachment to past roles would be more incongruent with the roles available to them in the nursing home and therefore more predictive of depressive symptoms than if women were strongly attached to past roles women’s. Therefore, I expect the impact of strong attachment to past roles on depression to be greater for men than for women. Research on Stress, Gender, and Depression Critics argue that theories of engagement do not take into account that some aspects of engagement (e.g. interactions) can be stressful. It also does not address how stress, in the form of pain, will impact depression. Conflicted relations. Studies reporting levels of conflicted relations find that conflicted relations are rare (Gilbart and Hirdes 2000; Farber et al. 1991). Gilbart and Hirdes (2000) measured conflicted relations by examining covert/open conflict with or repeated criticism of staff, unhappy with roommate, unhappy with residents other than roommate, and openly expresses conflict or anger with family and friends. The level of conflicted relations has received more attention among the community dwelling aged. It has been variously studied as “social conflict” (MaloneBeach and Zarit 1995; Abbey, Abramis and Caplan 1985), “negative social interactions” (Rauktis, Koeske and Tereshko 1995; Lakey, Tardiff and Drew 1994), “negative social support” (Revenson 1990), and “social strain” (Walen and Lachman 2002; Rook 1992). Similar to the nursing home study, findings show that stress, in the form of conflicted relations, among the community dwelling aged is rare (Akiyama et al. 2003; Sherman 2003; Rook 2001; Revenson et al. 1991). Indeed, research reports that negative interactions, unlike other forms of stressors, decline with age. (Akiyama et al. 2003; Birditt and Fingerman 2003).

9

The difference between disengagement theory and activity theory in relation to how strong identification with past roles will impact depression is that activity theory suggests that it will be important to find new compensatory roles for those lost. Disengagement theory does not think this is the case.

30

There have been no studies of gender differences in the level of conflicted relations in the nursing home. However, there has been considerably more attention paid to gender in the community dwelling aged. Previous research finds that throughout the life course men are more likely to have conflict than are women (Birditt and Fingerman 2003; Wallen and Lachman 2000; Manne et al. 1997). There have been several studies that consider the impact of negative interactions on depression related experiences in the nursing home. Gilbart and Hirdes (2000) find that conflicted relations influences depression. Residents with more conflicted relationships have more depressive symptoms (mood problems) than those with less conflicted relationships (Gilbart and Hirdes 2000). Other studies on the nursing home find that the quality of primary family contacts is negatively associated with depression and positively associated with life satisfaction (Farber et al. 1991). Perceived social support from the family was negatively related to depression. One study of 50 elderly from six different long-term care facilities found that roommate rapport predicted life satisfaction (Kovack and Robinson 1996). Other research has found that the stability of social support networks correlates with a resident’s adjustment to a nursing home (Bitzan and Kruzich, 1990; Wells and MacDonald 1981). There has been no research on the relative impact of positive and negative interactions on depressive symptoms among the nursing home elderly. However, in general, researchers agree that undesirable events are more stressful than other events10 (Thoits 1983). When examined in the community dwelling aged, research on the relative impact of positive and negative aspects of engagement generally supports the idea that negative social interactions have a more potent relationship to depression than do positive or neutral interactions (Rook 2001; 1984). Studies that measure negative social interactions such as criticism, undermining, anger, and conflict find that negative social interactions are more predictive of depressed mood than positive interactions (Rock 2001; 1984; Manne et al. 1997; Krause 1995; Schuster et al. 1990). For example, Rook (2001) found, in a study using diary data collected at two points in time, that “negative exchanges occurred less often but were related more consistently to daily mood than were positive exchanges” (Rook 2001:86). An increase in negative exchanges over a one-year period was associated with an increase in depression (Rook 2001). In the nursing home, Utz et al. (2002) found that stability in social relationships and access to social roles are the most important aspects of engagement and most likely to ward against psychological distress among the elderly. They assert that “intervention efforts should focus on minimizing disruption in daily activities rather than on constructing a new life with new activities and new acquaintances” (Utz et al. 2002:17). Gender differences in the impact of conflicted relations on depression have not been studied in the nursing home. However, in community samples most research reports that conflicted relations have a stronger impact on women’s depressed mood and well being than on men’s (Manne et al. 1997; Walen and Lachman 2000). Specifically, in a study of 2,348 25 to 75 10

Some studies find that positive social interaction is more predictive of depression related experiences than negative social interactions (Sherman 2003; Finch et al. . 1999; Walen and Lachman 2000; Okun and Keith 1998). However, these studies used relatively mild measures of strain compared to research that finds a different relationship. For example, Sherman (2003) measures negative interactions by asking: (1) how often does anyone boss you; (2) how often does anyone not understand you; and (3) how often does anyone get on your nerves. Similarly, other studies measure negative interactions with indicators of feelings of irritation and burden in relationships (Finch et al. 1999, Schuster et al. 1990). Okun and Keith (1998) asked: How often do I feel someone is critical of me?

31

year olds Walen and Lachman (2000) found that conflict with family was correlated with declines in well-being for both men and women, but the relationship was stronger for women. However, it could be that this is the case because women still occupy family roles. Indeed, some research reports no significant gender differences in the impact of conflicted relations on depression (Schuster et al. 1990; Okun and Keith 1998; Sherman 2003). For example, Sherman (2003) found no gender differences in the impact of social strain on depression. They measured strain by asking how often anyone “gets on your nerves”, “bosses you”, or “does not understand you”. However, it is important to note that studies that find no gender differences in the impact of conflict on depression use mild forms of negative interactions. In contrast, studies that find gender differences use more severe forms of stress (e.g. lack of support versus conflict). Based on this review, I expect that there will be very few residents with conflicted relations in the nursing home. Men will be more likely than women to have conflicted relations. I also expect that conflicted relations will influence depressive symptoms in my study such that those with more conflicted relations will have more depressive symptoms compared to those with low levels of conflicted relations. While the relative impact of aspects of engagement has not been studied in the nursing home, Atchley and Barusch (2004) point out that in a nursing home it is difficult to avoid the negative impact of conflicted relations. First, residents must depend on staff to meet their needs. Second, residents are usually not around people who know their past and therefore they cannot draw upon a history of positive interactions. Finally, it is difficult to avoid contact with staff and roommates due to the close proximity. Therefore, I expect conflicted relations will be the most predictive of depressive symptoms of all of the measures of engagement. Pain. Pain is common among institutional long-term care residents. Estimates of the prevalence of pain among nursing home residents range from 24 to 86 percent depending on the intensity and frequency of measurement (Won et al. 1999; Finne-Soveri et al. 2000; Parmelee et al. 1993; Sengston and King 1993; Parmelee, Katz and Lawton 1991; Ferrell, Ferrell, and Rivera 1995). Finne-Soveri (2000) found that daily pain was present in 22 to 24 percent of patients in nursing homes. In another study of a long term care facility, chart reviews and patient interviews found that 24 percent had constant pain, 47 percent had intermittent pain, and only 39 percent had no pain (Ferrell et al. 1995). Previous research has shown that women have pain more frequently than do men in the nursing home (Finne-Soveri et al. 2000; Won et al. 1999). For example, in one study 25 percent of women suffered from daily pain compared to 18 percent of men. However, Geerlings et al. (2002) found that pain has a greater impact on depressive symptoms for men than it does for women. With regard to depression, greater pain frequency has been reported among elderly nursing home residents with depressive symptoms, compared to those without (Won et al. 1999; Casten et al. 1995; Parmelee et al. 1991). Those with pain were almost twice as likely to be depressed compared to residents not in pain (Ferrel et al. 1995; Parmelee et al. 1991). Won et al. (1999) reported that people with daily pain were more likely to have mood disorders than those without pain. Similarly, Gilbart and Hirdes (2000) found that pain predicted more mood state problems in the nursing home. With regard to gender, previous research has shown that women report having pain more frequently than do men in the nursing home (Bergh, et al. 2003; Finne-Soveri et al. 2000; Won 1999). For example, in one study 25 percent of women suffered from daily pain compared to 18 percent of men. However, research has found that pain has a greater impact on depressive

32

symptoms for men than it does for women (Bergh et al. 2003; Geerlings et al. 2002). Researchers hypothesize that pain may cause losses of autonomy, and the psychological responses to such loss may be more pronounced in men (Bergh et al. 2003). Based on this review, I expect that pain will be a common occurrence and that women will have more experiences of pain than men. It will also be important to test for gender differences in the effect of pain on depression. Summary In summary, while disengagement theory enjoys some empirical support, especially when applied to the oldest-old (Johnson and Barer 1992), it has largely been discounted in the gerontological literature (Katz 2000; Achenbaum and Bengston 1994). However, it is possible that disengagement is more frequent and less maladaptive than studies seem to indicate. “Elements of Cumming and Henry’s (1961) disengagement theory are relevant in understanding the social world of the very old. The findings indicate that both physical and social losses make it difficult to maintain an active social network” (Johnson and Barer 1992). When applied to the nursing home, disengagement theory has considerably more support. While in general engaged elders have fewer depression related experiences than disengaged elders, the relationship is impacted by health. Specifically, elders with high cognitive functioning and low physical functioning are found to have more mood problems when they are highly engaged than when they are not. However, the results are not related to disengagement theory. Rather, in practice we can see acceptance of activity theory. For example quality of nursing homes is measured by the amount of social engagement exhibited by residents (e.g. Mor et al. 1995). “We view continued social engagement as a critical component of quality of life, which according to the Institute of Medicine (1986) and the Nursing Home Reform Act, is an appropriate goal of nursing home care (Mor et al. 1995:1). While disengagement and activity theories have many critics, and have generally been discounted since the 1970s (Hendricks 1994), it is important to consider how implicit acceptance of each guides our expectations when thinking about the aged. It is necessary to determine the extent to which these assumptions are supported empirically. Despite its criticisms the enduring legacy of both disengagement and activity theory can be seen in practice. For example, Hendricks (1994), points out that disengagement theory is “in the air” as practitioners operate under an “implicit expectation that growing old is a process of gradually disappearing from view, and assume it is better that way” (p. 753). Researchers also use engagement with life as a measure of “successful aging”. For example, Rowe and Kahn (1987) propose a model of successful aging comprising three components: “engagement with life is one of the three pillars of successful aging….Successful aging is multidimensional, encompassing the avoidance of disease and disability, the maintenance of high physical and cognitive function, and sustained engagement in social and productive activities” (p. 433). Similarly, other researchers use engagement in activity and role occupancy to measure quality of life (Atchley 1989; George 1989, 1978; Roscow 1985, 1974; Blau 1973; Lemon Bengtson and Peterson 1972). Indeed, most research has focused on the negative impacts of social isolation and, conversely the importance of integration into the larger society for the subjective well-being of older individuals (e.g. Chappell 1992; Lee and Shehan

33

1989; Larson 1978). I maintain this assumption needs to be explored empirically in the nursing home setting. In addition, while gender differences in levels of engagement and the impact of engagement on depression is well supported in the community dwelling aged, these relationships have not been tested in the nursing home. Feminist theory suggests that the social organization of gender carries over to the nursing home setting. As such there will be significant differences in levels of engagement and in the impact of aspects of engagement on depression in the nursing home. However, there are a number of problems with this contention. First, while most research in the community shows gender differences, this has not been tested among the very old (over 85), or among the nursing home elderly. Second, when the very old are included, the studies do not control for age, but rather aggregate all women and men together, obscuring possible age differences in the gender depression relationship. Giele (1982) and Rossi (1986) suggest as age increases, gendered normative pressures decrease, allowing men and women to assume the views and roles of the other gender. As such, gender differences should become less pronounced. third, most of the studies rely on diagnosis. Diagnosis of depression among the elderly is fraught with errors, in particular for men. Because men are known to be less likely to report symptoms (Hurley and Siegal 2001), measures of diagnosis may vastly underestimate the depression related experiences of men in the nursing home. There are a number of gaps in the literature that I can address with my sample of nursing home elderly. With regard to activity and disengagement theory, each postulates that levels of engagement are important in ascertaining who is depressed and who is not. Most studies support activity theory. However, among the community dwelling aged, there is also some support for disengagement theory especially among the very old and very sick. Among the nursing home population, this is a relatively under explored area. My study is designed to add to the existing literature by testing the relationship between gender, engagement, and depression holding age and health constant. In addition, I will be able to address gaps in the literature by including the very old, controlling for age, and using data that relies on staff observations rather than self report. In this way, I will be able to avoid the limitations of previous research on the relationship between depression related experiences and engagement among the elderly in the nursing home. In order to do address my questions, I will first describe levels of depression. Next I describe levels of engagement and whether or not there are gender differences. To explore the applicability of theories of engagement in explaining depression in the nursing home, I test the hypothesis that engagement will impact levels of depression. If I find in my population of nursing home elderly, that controlling for age, education, health and cognitive status elders who are engaged are less depressed than elders who are not I have support for activity theory. If on the other hand, I find that, controlling for age, education, health and cognitive status engagement that elders who are engaged are more depressed than elders who are not, I have support for disengagement theory. With regard to critics of theories of engagement, if I find that stress (e.g. conflicted relations and pain), impacts depression I have support for theories of stress. In addition, if I find that the impact of conflicted relations and pain is more predictive of depression than are levels of engagement, I have support for advancing a new theory of ‘stressful aging’ to explain depression related experience in the nursing home. That is, such evidence would suggest that the continued debate over theories of engagement is unproductive and it would be beneficial to consider residents perceptions of interactions as stressful rather than concentrating on the quantity of engagement in explaining depression. If physical stress (pain) is also predictive of depression it

34

will further highlight the importance of considering stress in studies of depression in the nursing home. In addition, if I find that there are important gender differences in engagement and the impact of engagement on depression, then I have support for feminist theories contention that gender matters in old age.

35

CHAPTER 3 DATA AND METHODS

Hypotheses (1) Depression will vary significantly by gender, age, education, and health, controlling for cognitive limitation and ethnicity. (2) As engagement levels increase depressive symptoms will decrease, controlling for cognitive deficits, health, age, race, and education in my sample of nursing home elderly. (3) As stress levels increase depressive symptoms will increase, controlling for the other variables in the model. (4) There will be gender differences in the impact of engagement and stress on depression, controlling for the other variables in the model. This chapter describes the data I bring to bear upon the hypotheses outlined above. In the following sections, I describe the methodology used in this project and give a brief history of the data source, its purpose, and implementation. Next I describe the sample used in this study. I then describe the measures used to test the hypotheses, their validity, and reliability. Finally, I outline the analysis I use to test the hypotheses. The bulk of the analysis will involve examining the relationship between engagement and depression. I also examine the relationship between stress and depression as an alternative explanation for depression in the nursing home. Gender is also a primary topic of investigation. The Survey The hypotheses designed to explore these issues require analysis at the individual level. Therefore, the unit of observation and analysis is the individual. I analyze quantitative assessment data collected in survey form on individual nursing home residents. The sample is comprised of a subset of the MDS. The MDS data are collected on individual nursing home residents in the U.S. and internationally. As mentioned earlier, the data used in this project are obtained from the survey component of the Minimum Data Set (MDS), a component of the Resident Assessment Instrument (RAI). In 1987, in response to recommendations made by the Institute of Medicine

36

(1986), Congress mandated the development of RAI to be used as a standardized assessment and planning tool in all long-term care facilities (Omnibus Reconciliation Act of 1997). The goal was to improve quality of care by creating an accurate and reliable assessment tool that would provide nursing homes staff with information on the care needs of each patient and assist them in the development of individual patient care plans. In addition, the data give a multidimensional view of the patient’s functional capacities that is used to create a profile of a nursing home (Morris et al. 1990). As mentioned earlier, the RAI was developed under the direction of Health Care Financing Administration by a research consortium lead by the Research Triangle Institute in North Carolina. Collaborators included Hebrew Rehabilitation Center for the Aged in Boston, the Center for Gerontology and Health Care Research at Brown University, and the Institute of Gerontology at the University of Michigan (Morris et al. 1990). The RAI contains three major components: (1) a core set of assessment items designed to provide a comprehensive picture of resident’s functional status (known as the MDS for Nursing Home Resident Assessment and Care Screening or simply the MDS); (2) a set of specialized assessment protocols designed to directly link MDS data to care planning (known as Resident Assessment Protocols or RAP); and (3) a user’s manual containing detailed specifications as to how to complete the MDS and RAP assessments, along with item definitions, coding examples, and clinical guidelines for using the RAP for care planning along with case studies (Morris et al. 1995; Morris et al. 1991). By law, nursing homes receiving Medicare or Medicaid reimbursements are required to assess every resident at first admission and annually thereafter. Assessments are also required whenever patients experience a significant change in status (i.e. improvement or decline in multiple areas). Quarterly assessments using an abbreviated form of the RAI are required to monitor the effects of care, identify significant changes in status, and to indicate any necessary changes in care plans. As mentioned earlier, assessments are conducted by trained clinical professionals (nurses, social workers, therapists, etc.) who assess resident performance over all shifts during the prior seven days (in some cases 30 days). The assessor is responsible for interacting directly with the resident, reviewing patient records, medical records, and gathering information on resident performance from direct care, licensed professional staff, and families (Snowden et al 1999; Sgadari et al 1997; Hawes et al 1995; Morris et al 1990). Each item of the MDS has its own explicit definition, coding conventions, and a manual that describes how to ask questions, what to observe, and who to contact for information (Morris et al 1995; Morris et al. 1990). Assessors record the data on paper form and data entry personnel enter the data electronically at the local centers. Nursing homes then submit the data electronically to their respective state centers within seven days of the assessment. As mentioned earlier, the MDS requires the collection of a substantial amount of information. Data are collected on over 300 items from 19 different areas including: (1) identification and background information, (2) customary routines, (3) cognitive functioning, (4) communication and hearing, (5) vision, (6) mood and behaviors, (7) psychosocial well-being, (8) physical functioning and structural problems, (9) continence, (10) disease diagnoses, (11) health conditions, (12) oral and nutritional status, (13) dental status, (14) skin conditions, (15) activity pursuits, (16) medications, (17) special treatments and procedures, (18) discharge potential and overall status, and (19) assessment information. The MDS survey component of the RAI assessment instrument is appropriate for this study for a number of reasons. First, the MDS is collected on individual nursing home residents.

37

Individual level data are needed to test my hypothesis. Second, the MDS collects data on practically the entire population of nursing home residents. It is necessary to have a large number of cases in order to measure multiple variables while controlling for other variables. This data set collects a large amount of data on a population too large to observe directly. This allows me to discern relations between the multiple variables of particular import to my study. In addition, the MDS contains all the variables necessary for my study. Instead of using several instruments for different purposes, one instrument would be preferable. An assessment that gives a holistic view of patient’s health and needs is important (Soderhamn and Berthold 1993). Finally, the variables I will be using from the MDS have demonstrated good to excellent validity and reliability11 Since the implementation of the MDS, several attempts have been made to asses the reliability of the RAI, in particular the MDS. The developers of the RAI tested two versions of the original MDS and RAP in 28 nursing homes in six states using both facility and research nurses to dually assess over 600 residents. The tests were designed to examine the reliability of individual assessment items and overall performance of the instrument. These tests, plus debriefings of nursing home staff that tested the RAI, allowed the developers to assess and substantially improve the inter-rater reliability of the original MDS. The reliability and validity of the MDS continues to be tested with improvements made periodically. No item is added or deleted unless it is field tested to ensure that it performs well and improves or does not change the performance of the entire instrument (Morris et al 1997; 1990). Evidence from inter-rater and test-retest trials has suggested that almost all MDS items can be obtained reliably. The MDS can be used to recognize common geriatric problems (Gambassi et al 1998; Hawes et al. 1995). Hawes et al. (1995) conclude that “the reliability and clinical validity of the MDS instrument and approach to assessment make the MDS useful as both a clinical and a research instrument.” (p. 178). Finally, using a standardized instrument will facilitate conceptual clarity and dialogue across institutions, settings, and nations12. A variety of instruments for assessing depression and engagement have been presented in the scientific literature. Many institutions have also constructed their own assessment instruments without references to the scientific literature in the field. This obstructs the opportunity to compare the results both within and between populations and settings because of unknown reliability of the instruments. Using standardized measures will allow institutions and nations to learn from each other when proposing large and often costly interventions. There are several drawbacks to this method. One drawback is that survey instruments do not measure the meaning individuals give to their behaviors. For example, I may have a measure of how often nursing home residents engage in activities. I will not have a measure of the meaning these activities have for residents. However, that question is another study. I am interested in seeing if engagement impacts depression in the nursing home and how it differs by groups. Another drawback is that data collected on elderly individuals in nursing homes are problematic. That is, data may be “tainted” because residents lack the cognitive ability to 11

Values above .75 represent excellent agreement beyond chance, and values between .40 and .75 represent fair to good agreement beyond chance. 12 The RAI/MDS has been tested for validity and reliability in the U.S. (Hawes et al. 1995) and other countries such as Sweden and Denmark (Sgadari et al. . 1997). A consensus expressed by experts in clinical geriatrics, regarding the most important components of nursing home admission assessment has been found to be congruent with items mandated in the RAI/MDS (Mezey et al. . 1992). The RAI is now used world-wide and a cooperative network between countries, (interRAI), has been established. The RAI/MDS and the manual has also been translated and revised for a Swedish audience (Fries et al. 1997).

38

respond to questions or to be understood by assessors. The MDS items and assessment protocols have been adjusted to account for this by including the use of mutually exclusive response categories, inquiry about resident status over a relevant time period (e.g. seven days), and the use of multiple sources of information (Morris et al. 1991). Finally, a potential drawback to the study is that the MDS is part of an assessment tool. The main purpose of the survey is to assess resident status in order to assure the residents are getting quality care. In this way, I will be using the survey for research, which was not the main intended purpose. However, studies suggest that the “validity of the MDS instrument and approach to assessment make the MDS useful as both a clinical and a research instrument” (Hawes et al. 8:1995). The Sample For this project, I analyze data from a subset of MDS13 that consisted of the actual MDS data transmitted to Center for Medicare and Medicaid Services (formerly the Health Care Finance Administration) from every MDS assessment14 (n=167,617) conducted between July 2001 and March 2003 inclusive of all facilities owned by one large corporation. Personally identifying information (first name, last name, middle initial, social security number, Medicare number, Medicaid number, medical record number, and room number) was omitted. A patient identification number was created to anonymously identify each patient. In order to address my questions, I excluded multiple assessments and selected for the latest full MDS assessments (taken from July 2001to March 2003) on each resident15. This sample consisted of 7,503 individual residents. Because my study is concentrated on the elderly, I excluded residents below 65 (n=483). In addition, I excluded comatose residents (n=23), and short stay residents (residents who are expected to be discharged within 30 to 90 days (n=41). My final sample has survey data on 6,468 individual nursing home residents.

13

I obtained the data on June 13, 2003. I downloaded "tina.zip," a zip compressed file containing an SPSS for Windows data set (pcmds.sav). The file contained the corporations Minimum Data Set data and a text document describing the data set (datadef.txt). I used the information in the text document and the MDS 2.0 data collection form, as a data definition document. From these two, I could find the actual questions asked, responses available, and what the values represented. 14 Full assessments are conducted at admission and then annually. However, a significant subset of items is assessed at least quarterly. It is possible some patients in the data set might not have a full assessment. For example, a patient admitted before July 1, 2001 and discharged in less than a year might have quarterly or other assessments in the data set, but would not have an admission or annual assessment. 15 Assessment type is coded as REASON. Admission assessments are coded one and annual assessments are coded two. Using this field I selected only full assessments. To identify a patients most recent full assessment I sorted the assessments by patient id (PATIENT) and assessment date (OBSDATE) in descending order (latest assessment date first). Then I created a new variable and set it equal to 1 for all cases [COMPUTE COUNT=1.]. Next, I increased the new variable by one if the previous record has the same patient id [IF PATIENT=LAG(PATIENT) COUNT=LAG(COUNT)+1.] This created a variable in which a patients most recent assessment had a value of one and the next most recent a value of two, and so on. I then selected cases where COUNT=1.

39

Table 3.1 Sample Descriptive Statistics Variable Frequency Percent 6468 100 Gender Female 5203 80.4 Male 1265 19.6 6468 100 Age 65-74 724 11.2 75-84 2350 36.4 85-94 2843 43.9 95-highest 551 08.6 6468 100 Ethnicity Indian 7 0.1 Asian 11 0.2 Black 483 7.5 Hispanic 11 0.2 White 5956 92.1 6467 99.99 Education None 67 1 th 8 or less 2363 36.5 9-11 grade 896 13.9 th 12 grade 2046 31.7 Tech/trade 186 2.9 Some 424 6.5 college BA degree 342 5.3 Graduate 143 2.2 Missing 1 .0001

Mean( S/D) n/a 84 (.80)

n/a

3.5 (1.60)

The final data set includes data from individuals who resided in one of 80 NHC nursing homes from July 2001 through March 2003. The nursing homes are located in eleven states. One-hundred fifty-seven individuals live in facilities in Alabama, 203 in Georgia, 344 in Indiana, 221 in Kansas, 277 in Kentucky, 333 in Massachusetts, 976 in Montana, 233 in New Hampshire, 1,269 in South Carolina, 2141 in Tennessee, 91 in Virginia, and 178 in Washington. Eighty percent of the residents are in a for-profit facility, about four percent in a for-profit partnership, one percent in a government facility, eight percent in a non-profit corporation, and five percent in a non-profit “other” type of facility. Most residents reside in a nursing home that is multiowned. Only about 5 percent (350 residents) reside in a singly owned facility while 93.9 percent reside in a multi-owned nursing home. Almost all (99.4 percent) of the facilities are stand alone (not located within a hospital). Most facilities (97 percent) participate in Medicare and Medicaid while two percent participate in Medicare only. The number of beds ranges from 24 to 282 with an average occupancy of 89 percent.

40

As can be seen from Table 3.1, 80.4 percent of the residents are female. The resident age ranges from 65-113 with a mean age of 84. Most residents, 92 percent, are white. The most common educational level is having a high school education16. The Measures Dependent Variable DRS. To measure depressive symptoms, I use the DRS developed from the MDS by Burrows et al. (2000). Burrows et al (2000) developed the DRS from mood items contained in the MDS. MDS items that measure resident mood are: (1) negative statements (e.g. “nothing matters,” I would rather be dead,” “what’s the use”); (2) repetitive questions (e.g. “where do I go? what do I do?”); (3) repetitive verbalizations (e.g. calling out for help); (4) persistent anger with self or others (e.g. easily annoyed, anger at placement in nursing home or anger at care received); (5) self deprecation (e.g. “I am nothing,” “I am of no use to anyone”); (6) expressions of what appear to be unrealistic fears (e.g. fear of being abandoned, left along, or fear of being with others); (7) recurrent statements that something terrible is about to happen (e.g. believes he or she is about to die or have a heart attack); (8) repetitive health complaints (e.g. persistently seeks medical attention, or has obsessive concern with body functions); (9) repetitive anxious complaints or concerns (non health related) (e.g. persistently seeks attention or reassurance regarding schedules, meals, laundry, clothing, relationship issues, etc), (10) unpleasant mood in the morning; (11) insomnia or change in usual sleep pattern; (12) sad, pained, worried facial expressions (e.g. furrowed brows); (13) crying or tearfulness; (14) repetitive physical movements (e.g. pacing, hand wringing, restlessness, fidgeting, picking); (15) withdrawal from activities of interest (e.g. no interest in long standing activities or being with family and friends; (16) reduced social interaction. Mood items are difficult to measure. To obtain reliable and valid results, data collectors are instructed to “initiate a conversation with the resident because some residents talk more about their feelings than others and will either tell someone about their distress, or tell someone only when directly asked how they feel. Other residents may be unable to articulate their feelings (cannot find the words to describe how they feel, or lack insight or cognitive capacity)” (Burrows et al 2000:1119). Collectors observe residents carefully for any indicator, and consult with direct-care staff over all shifts, if possible, and family who have direct knowledge of the resident’s behavior. Collectors also examine the clinical record for relevant information. To develop the DRS, Burrows et al (2000) correlated all MDS mood items to two commonly used measures of depression, the 17 item Hamilton Depression Rating Scale (Hamilton 1967) and the 19 item Cornell Scale for Depression in Dementia. Both scales have been validated in disabled and medically ill elderly populations (Rapp et al. 1990), and are widely used in cognitively impaired geriatric populations (Burrows et al. 1995; Logsdon and Teri 1995; Alexopoulos et al 1988a; 1988b). Four of the mood items (repetitive questions, repetitive calls for help, repetitive physical movements, and reduced social interaction) were excluded because they were not significantly correlated with either one or both of the scales. The remaining 12 mood items were factored and rotated resulting in five factors. First, sad facial expressions, anger and irritability, withdrawal from activities of interest, and unpleasant mood in the morning loaded together and were termed “disturbed mood” (Burrows et 16

The missing value was recoded as the modal value (high school).

41

al 2000). Second, repetitive anxious complaints/concerns (non health related), sleep disturbance, and repetitive health complaints loaded together and were termed “anxiety” (Burrows et al 2000). third, expressions of panic and expressions of unrealistic fears were termed “fear”. Fourth self-deprecation and negative statements loaded together as “loss of meaning” (Burrows et al 2000). Fifth, crying loaded alone and was termed “affect” (Burrows et al. 2000). From here, each of the five factors were analyzed using a serious of step forward regressions which produced “identical subsets of items for the Hamilton Depression Rating Scale and Cornell scale regression models” (Burrows et al. 2000:169). The seven items of the resulting DRS (negative statements, persistent anger, unrealistic fears, health complaints, anxious complaints, sad expressions, and crying) have a range of 0 to 14. When examining scale performance, the DRS achieved correlations of .70 and .71 for the Hamilton scale, and .69 and .70 for the Cornell scale (Burrows et al. 2000). Burrows et al. (2000) also tested the DRS against psychiatric diagnosis using standard DSM-IV criteria. The DRS was accurate in detecting depression 91 percent of the time. “The scale demonstrated excellent sensitivity and acceptable specificity compared with psychiatric diagnosis based on DSM-IV criteria” (Burrows et al. 2000:168). The Cronbach alpha measure of internal consistency yielded an acceptable coefficient in both samples (.75 and .71). Table 3.2 reports DRS statistics for my sample. The range is 0 to 13. The mean is .9185, with a standard deviation 1.58 and a variance of 2.4. The distribution is positively skewed with more than 60 percent of residents exhibiting no symptoms. Indeed only 17 people have a score of 10 or higher. The maximum number of depressive symptoms is 13. Only one person has a score of 13. In addition, the variance is almost three times the mean.

Table 3.2 DRS Descriptive Statistics Variable Frequency Percent 6468 DRS 0 1 2 3 4 5 6 7 8 9 10 11 12 13

3948 987 749 302 230 086 073 043 020 013 009 003 004 001

Range 0-13

61.0 15.3 11.6 04.7 03.6 01.3 01.1 00.7 00.3 00.2 00.1 00.0 00.1 00.0

42

Mean( S/D) .9185 (1.58)

Variance M/M 2.497 0/0

Table 3.3 DRS Item Frequencies Variable Frequency Percent 6468 Negative Statements None 6059 93.7 5 days 340 05.3 6-7 days 69 01.1 6468 Anger None 5936 91.8 5 days 430 06.6 6-7 days 102 01.6 6468 Fears None 6159 95.2 5 days 6-7 days Health Fears None 5 days 6-7 days Anxious None 5 days 6-7 days Sad None 5 days 6-7 days Crying None 5 days 6-7 days

263 46 6468

04.1 00.7

5770 528 170 6468 5751 548 169 6468 4848 1173 0447 6468 5874 0535 0059

89.2 08.2 02.6

Range 0-2

Variance .0898

0-2

.12

0-2

.06

0-2

.17

0-2

.17

0-2

.36

0-2

.11

88.9 08.5 02.6 75.0 18.1 06.9 90.8 08.3 00.9

Table 3.3 examines specific item frequencies. Very few residents (less than 10 percent) exhibit any symptoms six-seven days a week or more. In order of frequency, the most common symptom exhibited is sad facial expression with about 25 percent having symptoms (18 percent five days a week and 7 percent six to seven days a week). Only about 10 percent of residents ever exhibited crying, anxiety (11 percent), and health fears (11 percent). Less than 10 percent of residents were ever observed to have negative statements, be angry or express unrealistic fears. An internal consistency analysis using Cronbach’s procedure on the seven items in the DRS yields a coefficient alpha of .6831. This is below the .70 and .72 found in (Burrows et al

43

2000) samples. Nunnaly (1978) advises that the cut off for reliability should be .7 in order to decrease the chance of making a Type I error. However, this is most necessary in a small sample. The large sample size in my study should help counter the impact of the DRS index’s random error on my estimates and hypothesis testing. Independent Variables. Table 3.4 SEI Descriptive Statistics Variable Frequency Percent M/M 6468 SEI 0 1 2 3 4 5 6 At ease with others-yes Structured activities-yes Self-initiated activity-yes Establishes goalsyes Involved in facility-yes Group Activities

Range 0-6

0517 0972 1246 1271 1072 0719 0664 5588

08.0 15.0 19.3 19.7 16.7 11.1 10.3 86.4

3176

49.2

3214

49.7

1422

22.1

1746

27.0

0-1

4026

62.2

0-1

Mean( S/D) Variance 2.964 (1.748)

3.0477

3/3

0.86 (0.34) 0.50 (0.50) 0.50 (0.50) 0.22 (0.41) 0.27 (0.44) 0.62 (0.48)

0.12

1/1

0.25

0/0

0.25

0/0

0.17

0/0

0.20

0/0

0.24

1/1

Engagement: SEI. The SEI is an additive scale combined from six dichotomous measures in the MDS that focus on sense of initiative and involvement with self, other residents, and direct care staff. Data collectors gather information based upon behavior of the resident (verbal, body language, or actions). Collectors code for the presence or absence of the behavior over the last seven days. These items ask if the resident is: (1) at ease interacting with others; (b) at ease doing planned or structured activities; (c) at ease doing self initiated activities; (d) establishes own goals; (e) pursues involvement in life of facility (e.g. makes or keeps friends, is involved in group activities, responds positively to new activities, assists at religious services);

44

and (f) accepts invitations into most group activities. These items are added together to create an index of social engagement (Schroll et al 1997). The scale ranges from 0-6. The social engagement index has been found to have adequate reliability. The average inter-rater reliabilities of these items were .58 in the U.S. (Schroll et al. 1997). As can be seen in Table 3.4, the mean score for the social engagement index is 2.964 with a standard deviation of 1.748. Most residents (86 percent) are at ease interacting with others, approximately 50 percent are engaged in structured activities and self initiated activities. In contrast, most residents, 22 and 27 percent respectively, are not found to establish their own goals or be involved in the life of the facility. Engagement: activity time. To measure “activity time”, I use the MDS item that measures average time residents’ engage in activities when awake and not receiving treatments or ADL care. The options are: (0) none of the time; (1) little (less than one-third of time); (2) some (from one-third to two-thirds of time); (3) most (more than two-thirds of time). As Table 3.5 shows, the distribution is heavily clustered with most people (85 percent) engaged in activities some of the time (from one-third to two-thirds of the time). Because there are so few observations in the “none” category, I will combine none and little in further analysis.

Table 3.5 Activity Time Descriptive Statistics Variable Frequency Percent Range 6468 100 0-3 Activity Time None Little Some Most

0010 0659 5472 0327

Mean( S/D) 1.95 (0.39)

Variance 0.16

M/M 2/2

00.2 10.2 84.6 05.1

Engagement: reduced. I combine two mood item indicators that measure reduction in activities and reduction in interactions over the last 30 days. The MDS item asks if residents have experienced: (1) withdrawal from activities of interest (e.g. no interest in long standing activities or being with family and friends); and (2) reduced social interaction. Each item is coded where zero indicates no change in engagement, one indicates some reduced engagement, and two indicates that residents have withdrawn from most activities and interactions. I will add these together such that higher scores reflect greater reductions in engagement.

45

Table 3.6 Reduced Engagement Item Descriptive Statistics Variable Frequency Percent Range Mean( S/D) 6468 100 0-4 .1288 Reduced (.5790) Engagement 0 6090 94.2 1 0105 01.6 2 0175 02.7 3 0014 00.2 4 0084 01.3 Withdraw from 6468 100 0-2 .0612 interests (.30) No 6180 95.5 Some 0180 02.8 Most 0108 01.7 Withdraw from 6468 100 0-2 .0676 interactions (.31) No 6140 94.9 Some 0219 03.4 Most 0109 01.7

Variance .3352

M/M 0/0

.0909

0/0

.0967

0/0

Table 3.6 shows the range of scores on the engagement change scale (0 to 4). Most residents (94 percent) show no change in the last 30 days. Because there are so few residents with scores of one to four, I will make this measure dichotomous where zero indicates no change and one indicates reduced engagement. Engagement: identify with past roles. Identification with past roles taps into the resident’s role attachment. The item asks if residents have a strong identification with past roles and life status. The item is coded as one if the feeling is present and zero if it is not. This item has been found to have .51 inter-reliability estimates (Hawes et al. 1995). Table 3.7 shows that 12 percent of the residents in the sample strongly “identify with past roles”.

Table 3.7 Roles Descriptive Statistics Variable Frequency Percent Roles (0) no 56 8 84 7.9 Roles (1) yes 1 784 2.1

Median/Mode 0/0 0/0

46

Independent variable engagement: contact. Another aspect of engagement is absence of “contact”. The MDS item I use to measure this aspect asks whether or not the resident has had an absence of personal contact with family or friends. As Table 3.8 shows, most residents (97 percent) have contact with their family and friends. Only 197 residents do not have contact with family and friends.

Table 3.8 Contact Descriptive Statistics Variable Frequency Percent Median/Mode Contact (0) 6271 97 0/0 Contact (1) 197 3

Stress: conflict.

Table 3.9 Stress: Conflict Descriptive Statistics Variable Frequency Percent Range Mean( S/D) 6468 100 0-5 .1279 Conflict (.4171) 0 6009 89.7 1 0363 08.3 2 0080 01.6 3 0015 00.3 4 0001 00.1 Conflict 0193 03.0 Staff-yes Unhappy 0086 01.4 Roommateyes Unhappy 0115 01.8 Residentsyes 0178 02.8 Conflict Family or Friends-yes

47

Variance .174

M/M 0/0

To measure stress, in the form of conflict, I use an item that asks about problem relationships with staff, roommate and family and friends. I will combine four items: (1) covert or open conflict with or repeated criticism of staff; (2) unhappy with roommate; (3) unhappy with residents other than roommate; (4) openly expresses conflict/anger with family or friends. I use these indicators to create a “conflict” scale. The items are coded as one for presence and zero for absence of problems in the last seven days. Summing the items creates a “conflict” scale with a range of 0 to 4. These items have been found to have .49 inter-rater reliability scores (Hawes et al. 1995). As can be seen from Table 3.9, most residents (about 90 percent) do not have “conflict”. In order of frequency, 193 residents have a conflict with staff, 178 a conflict with family or friends, 115 are unhappy with residents, and 86 are unhappy with their roommate. Stress: pain. I measure stress in the form of physical pain by including an item on frequency of pain. Assessments of pain are included in the MDS. Assessors are instructed to: Review the medical records (including current nursing care-plan) and consult with facility staff members. Ask resident if he/she experienced any of the listed symptoms in the last seven days. A resident may not complain to staff members because he/she may attribute symptoms to old “age”. Such problems can often be remedied. Consult with family member (or other person close to residents) if resident is unable to respond (Morris et al. 1991:42).

When assessing daily pain, all shifts in a one week period are taken into account. Daily pain is recorded when the subject expresses a verbal or nonverbal complaint of pain at least once daily (six-seven days per week) during the observation week. The non-communicative patients are observed (e.g. during all the shifts of the day, during caring procedures and in rest) for any non-verbal indicators of pain, such as moaning, crying, wincing, and other facial expressions. Also, various posturing, guarding, or protecting an area of the body are taken onto account. The item measures frequency with which resident complains or shows evidence of pain in the last seven days. The item is coded: (0) no pain; (1) pain less than daily; and (2) daily pain. Fries et al. (2001) have reported predictive validity of the MDS pain items. In addition, Hawes et al. (1995) report reliability of .46. Table 3.10 shows that 46 percent of the residents have no pain, and only 18 percent have daily pain. Table 3.10 Pain Descriptive Statistics Variable Frequency Percent Range 6468 100 0-2 Pain No Pain 3028 46.8 Less Daily 2229 34.5 Daily Pain 1211 18.7

48

Mean( S/D) .72 (.76)

Variance 0.58

M/M 1/0

Gender. Gender will be coded as a dichotomous variable with females being the reference group. Females are the reference group because they represent the vast majority of nursing home residents (Table 3.1 reports statistics on gender). Control Variables Variable that could conceivably confound the relationship under question will be used as controls in this project. These include demographic characteristics of age, ethnicity, and education. In addition I also include two measures of health, one of mental health (cognitive function) and one of physical health (ADL functioning). However, because I am also interested in how these variables impact DRS, I will examine the relationship between age, education, and ADL limitations on DRS in findings and conclusions. Because cognitive limitations have an uncertain relationship to depression this will be used strictly as a control variable. In addition, because race has an uncertain relationship to depression and there is very little racial variation in my sample, this will also be used strictly as a control variable. More detailed explanations of these variables follow. Control: age. Age is measured by the age (in years) of the resident at the time of assessment. Descriptive statistics on age are reported in table 3.1. I created age groups of 10 years in order to represent descriptive statistics and correlations between age and measures of depression. However, I leave the variable continuous in hypothesis testing. I will top code age so that those with age 101-113 are reported as age 100. Control: race/ethnicity. Race is measured by indicating if the resident is American Indian/Alaskan Native, Asian/Pacific Islander, Black (not of Hispanic origin), Hispanic, or white. Descriptive statistics are reported in Table 3.1. As can be seen, 92 percent of the population is white. Since the vast majority of the population is white and there are so few observations in the other categories, I will create a dichotomous variable where zero is equal to white and one is equal to all others (minorities). As I mentioned earlier, because there is so little racial variation in my sample, race is only used as a control variable in hypothesis testing. Control: education. The education variable in the MDS measures the highest level of education completed. The response categories are: (1) no schooling; (2) eighth grade or less; (3) ninth to eleventh grades; (4) high school; (5) technical or trade school; (6) some college; (7) bachelor’s degree; and (8) graduate degree. Descriptive statistics on education are reported in table 3.1. In analysis, I will combine no schooling, eighth grade or less, and ninth to eleventh grades to create a new category “less than high school”. In hypothesis testing, I create dummy variables with having a “college degree” (BA) or higher being the reference group. That is, I will compare those having “less than a high school” education, a “high school” education, and “less than college” (some college and trade school) to those having a “college degree” (BA) or higher. Control: Cognitive functioning. In this study, I use the direct measure of cognitive function (COG) recommended by Hartmaier and colleagues (1995). I use this measure as a control variable in hypothesis testing. The direct scale consists of seven items related to memory, orientation, and decision making skills, which yield a composite score for each patient with a range of zero (no memory problems, oriented, independent decision making) to nine

49

(severely impaired). The items for memory measure short term and long term memory. The item measures recall of what was learned or known. Each item is coded: (0) memory okay; and (1) memory problem. The recall subscale comprises the following four orientation items that are also part of the direct scale: (1) current season; (2) location of own room; (3) staff names or faces; and (4) awareness of being in an NH. These items have a score range of zero to four (four means the resident is able to recall all four items.) Thus in order to sum all items for the direct scale total score, the recall subscale score must be reverse coded (four is unable to recall any of the four items). Finally, the cognitive skills for daily decision making measures how residents make decisions regarding tasks of daily life. The options are: (0) independent; (1) modified independence (some difficulty in new situation only); (2) moderately impaired (decisions poor, cues or supervision required); and (3) severely impaired (never or rarely made decisions). Spearman-Brown inter-rater reliability coefficients for the MDS cognitive items were found to range from .60 to .88 with an average of .72 (Morris et al. 1990).

Table 3.11 COG Descriptive Statistics Variable Frequency Percent Range 6468 0-9 COG 0 1 2 3 4 5 6 7 8 9

0428 0599 0510 0487 0440 0477 0500 0538 0714 1775

Mean( S/D) 5.5161 (3.1216)

M/M 6/9

06.6 09.3 07.9 07.5 06.8 07.4 07.7 08.3 11.0 27.4

Table 3.11 shows the range of the COG scale in my sample is from zero to nine with most observations clustered at the high end of the scale. Very few residents have no cognitive deficits and most residents (over 50 percent) score six and over. The mean is 5.5161 with a standard deviation of 3.1216. The median is six. The mode is nine. When examining specific items, most residents (77 percent) have a problem with short term memory. When examining long term memory problems, slightly more than half (51 percent) of the population have a problem with long term memory.

50

Table 3.12 COG Item Frequencies Variable Frequency Percent Range 6468 0-1 Short Term Memory Ok-0 1519 23.5 Problem-1 4949 76.5 0-1 Long Term 6468 Memory Ok-0 3148 48.7 Problem-1 3320 51.3 6468 0-1 Recall Season Yes-0 2259 34.9 No-1 4209 65.1 6468 0-1 Recall Room Yes-0 3180 49.3 No-1 3228 50.7 6468 0-1 Recall Staff Yes-0 3733 57.7 No-1 2735 42.3 6468 0-1 Recall Location Yes-0 2843 44.0 No-1 3625 56.0

51

Table 3.12 shoes the specific memory recall orientation items. As can be seen, 65 percent were unable to recall the season, 50 percent did not recall their room location, 42 percent did not recall the staff, and about 56 percent did not know they were in a nursing home. Finally, less than 10 percent of the residents are independent in decision making. At the other extreme, almost 42 percent were severely impaired, with residents never or rarely making decisions. Control: ADL. The most common measures of physical functioning used from the MDS are derived from resident’s ADL classification and is based on six items of activities of daily self-performance. For each of the six items residents are rated on a four point scale regarding dependency: (0) independent (no help or oversight, or help/oversight provided only one or two times during last seven days); (1) supervision (oversight, encouragement or cueing provided three or more times during last seven days, or supervision three or more times plus physical assistance provided only one or two times during last seven days); (2) limited assistance (resident highly involved in activity, received physical help in guided maneuvering of limbs or other nonweight bearing assistance three or more times, or more help provided only one or two times during last seven days); (3) extensive assistance (while resident performed part of activity, over last seven-day period, help was provided three or more times); (4) total dependence17 (full staff performance of activity during entire seven days). Items included are: (1) bed mobility (how resident moves to and from lying position, turns side to side, and positions body while in bed); (2) transfer (how resident moves between surfaces, to and from bed, chair, wheelchair, standing position, excluding to and from bath or toilet); (3) locomotion (how resident moves between locations on the floor); (4) dressing (how resident puts on, fastens, and takes off all items of street clothing, including donning and removing prosthesis); (5) toilet use (how resident uses the toilet room, commode, bedpan, or urinal, transfers on and off toilet, cleanses, changes pad, manages ostomy or catheter, and adjusts clothes); (6) personal hygiene (how resident maintains personal hygiene, including combing hair, brushing teeth, shaving, applying makeup, washing and drying face, hands, and perineum, exclude baths and showers; (7) eating (how resident eats and drinks, regardless of skill, includes intake of nourishment by other means [e.g. tube feeding; total parenteral nutrition]). Items used in this index have been shown to have good to excellent reliability (.69-.92) (Hawes et al 1995). The ADL scale sums the seven items with scores of zero to four. This creates an index that ranges from 0 to 28 with higher values indicating greater functional impairment. I will use the continuous measure to describe the population and as a control in hypothesis testing. However, in order to show correlations between mean number of depressive symptoms and functional impairment, I will create categories of functional impairment. Residents who score greater than the median value will be considered high on functional impairment. Those who score lower than the median value will be considered low on functional impairment (Kiely, Morris and Algase 2000). As can be seen from Table 3.13, the overall scale mean is 16.30, with a standard deviation of 8.75. The median score is 17 and the mode is 28. The reliability of the scale is good. The alpha internal consistency is .9488. This is comparable to Morris et al (1999) who found an alpha of .94. Specific item frequencies are reported in Table 3.14. For dressing, personal hygiene, and toilet use, 22 percent or less are in each of the three least dependent categories; about 31 to 21 percent receive extensive assistance, and 42 to 33 are totally dependent. For bed mobility, 17

The item is coded as eight if the activity did not occur during entire seven days. I will recode eight to be total dependence in order to keep the ordinal measure consistent

52

transfer, and locomotion, between 20 and 30 percent are independent and 20 to 36 percent totally dependent. Eating has the most independent observations with 44 percent being independent. Twenty percent are totally dependent. The least likely observation in all ADL items is for the resident to perform the activity with supervision.

Table 3.13 ADL Descriptive Statistics Variable Frequency Percent Range 6468 0-28 ADL 0-6 7-13 14-18 19-22 23-27 28 Bed Mobility

1169 1252 1167 0888 1136 0856 6468

Mean( S/D) 16.3052 (8.7515)

Variance 76.5881

M/M 17/28

8.1 9.3 8.0 3.7 7.7 3.2 0-4

Transfer

6468

0-4

Locomotion On floor Dressing

6468

0-4

6468

0-4

Toilet

6468

0-4

Personal Hygiene Eating

6468

0-4

6468

0-4

2.31

2 /0

2.15

3 /4

1.69

2 /4

1.34

3 /4

2.09

3 /4

1.47

4/ 4

2.54

1/0

Limited 21.1

Extensive 22.8

Dependence 20.4

20.3 15.3 22.6 15.8 21.8 09.9

23.4 9.9 31.7 20.9 25.8 07.6

30.4 36.5 33.5 42.5 38.7 20.9

1.98 (1.52) 2.39 (1.47) 2.186 (1.64) 2.80 (1.16) 2.70 (1.45) 2.83 (1.22) 1.43 (1.59)

Table 3.14 ADL Item Frequencies ADL Item Independent Supervision Bed 30.3 05.5 Mobility Transfer 19.7 06.1 Locomotion 26.0 12.4 Dressing 06.6 05.6 Toilet 15.4 05.4 Hygiene 07.0 06.7 Eating 44.3 17.3

53

Correlations Between Variables

In this section, I describe correlations among variables in my study. I also present one way ANOVA results that examine gender differences in levels of engagement and stress. This is important in order to get a baseline understanding of how the variables are related.

Table 3.15 Correlations among Dependent and Independent Variables SEI Time Role Reduced Contact Conflict DRS .011 .025* .160** .125** .103** 310** SEI 365** .227** -.069** .029* .114** Time .059** -.110** .003 .059** Roles .039* .042** .194** Reduced .079** .050** Contact .095 Conflict Pain *p