The Gerontologist Vol. 50, No. 1, 36–47 doi:10.1093/geront/gnp067 © The Author 2009. Published by Oxford University Press on behalf of The Gerontological Society of America. All rights reserved. For permissions, please e-mail: journals.permissions@oxfordjournals.org. Advance Access publication on June 23, 2009 Examining Resilience of Quality of Life in the Face of Health-Related and Psychosocial Adversity at Older Ages: What is “Right” About the Way We Age? Zoe Hildon, MA, PhD,1,2 Scott M. Montgomery, PhD,3,4 David Blane, MSc, MD,5 Richard D. Wiggins, PhD,6 and Gopalakrishnan Netuveli, PhD5 terized resilient outcomes relative to qol, included good quality relationships (5.105, confidence interval [CI] 95% 1.323–19.699), integration in the community (10.800, 95% CI 1.227–95.014), developmental coping (3.397, 95% CI 1.079– 10.690), and adaptive coping styles (3.211, 95% CI 1.041–9.910). Implication: Overall results indicate that policies that offer access to protection and help minimize adversity exposure where possible will promote resilience. Purpose: This article examines resilience at older ages, focusing on the relationships between quality of life (qol) and adversity. Our objectives are to identify (a) the basis of adversity, (b) the characteristics of resilient individuals, and (c) the attributes that attenuate the full impact of adversity. Design and Methods: Resilience is defined as flourishing despite adversity. Analysis is carried out in a subsample of the Boyd Orr cohort (aged between 68 and 82 years) using questionnaire data. Adversity was identified as circumstances that produce a significant average decrease in qol (CASP-19 scores). Participants were classified into resilient and vulnerable groups based on high or low qol (CASP-19 scores dichotomized at the median) in the face of significant adversity. Shared characteristics that define these outcomes are reported. Attributes that attenuate the negative impact of adversity were analyzed using stratified logistic regression. Results: Adversity was typified by functional limitation; life getting worse in the domains of health, stress, and general living circumstances; and experiencing a negative life event. The resilient tended to report fewer multiple adversities. Indicators of protective attributes, which also charac- Key Words: Protection, Sociodemographic factors, Social networks, Community-related factors, Coping When Nietzsche (1888/1984) famously noted “that which does not kill us makes us stronger,” he was saying something about the ability to weather adversity—to steel oneself against hardship, to move beyond it, and, in short, to be resilient. Despite the ease with which this expression rolls off the tongue, most of us, from our own experiences, know this expression to be wanting. Sometimes one does not emerge the better from adversity, other times resilience will apply to only one domain of our lives and not carry through to others. Resilience is clearly a splintered concept but nevertheless very relevant to understanding successful aging in Britain today. Within the gerontological literature, successful aging has traditionally been conceived in terms of physiological and cognitive robustness or the staving off of decline in health (Bowling & Dieppe, 2005; e.g., Rowe & Kahn, 2000). Nevertheless, Harris (2008, p. 43) has recently argued that 1 Address correspondence to Zoe Hildon, Department of Clinical, Educational and Health Psychology, Centre for Outcomes Research and Effectiveness, University College London, 1-19 Torrington Place, London WC1E 7HB, UK. E-mail: z.Hildon@UCL.ac.uk 2 Department of Clinical, Educational and Health Psychology, Centre for Outcomes Research and Effectiveness, University College London, UK. 3 Department of Medicine, Clinical Epidemiology Unit, Karolinska University Hospital, Sweden. 4 Clinical Research Centre, Örebro University Hospital, Sweden. 5 Department of Primary Care and Social Medicine, Division of Epidemiology, Public Health and Primary Care, Imperial College London, UK. 6 Social Science Research Unit, Institute of Education, University of London, UK. 36 The Gerontologist be better equipped to manage adversity; the second tries to capture the processes explaining how resilience is achieved (Jacelon, 1997; Richardson, 2002). Analysis identifying resilient characteristics tends to be person focused, grouping people with resilient outcomes and examining the factors that define them (e.g., Garmezy, 1985; Werner & Smith, 1977). As Masten (2001, p. 232) puts it, in a bid “to capture the configural patterns of adaptation that naturally occur, in much the same ways that classification systems for mental disorder organize symptoms into patterns that have been observed to occur together.” This analysis is an effective way of answering the fundamental research question: who are the resilient? In contrast, variable-focused analyses have been used as a starting point for considering resilient processes. These approaches tend to examine the relationship or interplay between adversity and protection, which lead to better-than-expected outcomes (Luthar, Ciccheti, & Becker, 2000; Roosa, 2000), addressing the question, which factors are effective in attenuating the negative impact of adversity? Analysis that consider variables that are mobilized in the face of significant adversity provides one way of identifying more immediate actions, iterative or habitual/routine, performed to achieve resilient ends. Such analysis can suggest practices or assets that are particularly effective when adversity is present. Presumably because they buffer, transform, or negate the full potential impact of psychosocial and health-related crises. Studies examining resilience and protective relationships that focus on older ages (Becker & Newsom, 2005; Butler & Ciarrochi, 2007; Kinsel, 2005; Netuveli, Montgomery, Hildon, Wiggins, & Blane, 2006; Ryff, Singer, Love, & Essex, 1998; Staudinger,Freund,Linden,&Maas,1999;Wagnild& Young, 1990; Windle & Woods, 2004) as well as those on resilience at younger ages have tended to lack focus on the detail of adversity (Hildon et al., 2008). Resilience-centered research, which has considered older populations, has poorly defined the adversity construct or assumed that old age itself is necessarily adverse. Within analyses identifying protective factors promoting resilience at older ages, the precise role of these stated factors are rarely specified or interpreted. Furthermore, although resilient characteristics and the attributes or associated processes that negotiate the impact of adversity have been examined in relation to younger ages, to our knowledge these features of resilience have not been systematically analyzed for older populations. “perhaps we have been striving for the wrong goal. The true quest as we age should not be for successful aging [as traditionally defined], but our goal should be for resilience, an undervalued and not fully examined concept in aging.” A position backed up by the looming reality that, despite ever-hopeful models describing the postponement of morbidity as we grow older (Fries, 1980), a disease-free old age is after all unrealistic for most of us (Kind, Dolan, Gudex, & Williams, 1998; Martin, Meltzer, & Elliot, 1988). Resilience at older ages addresses this certainty and conceptually disputes commonly held assumptions about the elderly individuals. For instance, resilience overrides the idea that once health begins to deteriorate and disability sets in, aging successfully is no longer possible. This is because resilience as a phenomenon is characterized by its improbability and is defined by the presence of hardship, as achieving better-than-expected outcomes or flourishing despite adversity (Garmezy, 1985; Hildon, Smith, Netuveli, & Blane, 2008; Masten, Best, & Garmezy, 1990; Rutter, 1987; Werner, 1990; Werner & Smith, 1977, 1982). Accordingly, the tradition and foci of research on resilience can be seen to have emerged “more by accident than by intent” (Grotberg, 1997, p. 118). Study Aims Historically, research on the topic of resilience originated from studies of risk and development in children and young people (Werner, 1990, 1995). As researchers scrutinized the effects of deprivation and other adversities, seeking to better understand the effects of vulnerability, they were surprised to discover instead that many young people were not beaten down or defeated by socioeconomic or psychosocial adversity. Researchers then began asking not just about what was wrong with young people facing hardship but also what was “right” with them? (Werner, 1990). The presently reported study aims to break down and answer this question in relation to older ages, a stage of life that research on resilience has comparatively neglected (Rowe & Kahn, 2000; Schoon, 2006), focusing on the relationships between quality of life (qol) and health-related or psychosocial adversity. Lessons From the Literature The tradition of resilience-focused research has broadly used two forms of analysis. The first seeks to identify characteristics of resilient people, lists or constellations of factors that identify who might Vol. 50, No. 1, 2010 37 Study Objectives and Research Questions Table 1. Comparison of Boyd Orr Sample Used in Present Analyses (N = 174), With Population of Britain Aged 70–84 Years in the 2001 Census Consequently, we undertake analyses from the established resilience tradition, identifying individual and social indicators that characterize resilient outcomes or point toward resilient processes at older ages. Our objectives are to (a) examine adversity a priori; (b) identify sociodemographic factors, social networks and community-related factors, and styles of coping, which are predominantly associated with resilient qol outcomes when compared with vulnerable ones. In sum, we have three main research questions we ask relative to qol at older ages: What typifies psychosocial and health-related adversity? Who are the resilient? and Which individual or social attributes are effective in attenuating the negative impact of adversity? Population characteristics Female Male Women, married currently Men, married currently Higher educationa Owner occupier Limiting long-standing illness Self-assessed health “good” or “fairly good” Present Boyd Orr sample (%) Census population (%) 54.1 45.9 47.3 80.0 11.6 80.7 40.6 84.1 57.8 42.2 39.1 70.1 11.5 74.8 51.1 70.6 Note: aQualifications beyond Levels 1–3 of National Key Learning Targets (General Certificate of Secondary Education, O Level, A Level, National Vocational Qualifications Levels 1–3), including graduate and postgraduate academic qualifications and some vocational or professional qualifications. Population Sample Study Participants remains broadly representative of the British population in terms of sex, academic qualifications, and housing tenure. The 33 sample members excluded because of missing data tended to be older (56% of the excluded 33 were older than 75 years, compared with 43% of the included), were marginally less likely to be owner occupiers (77% compared with 83%), and were more likely to report their income as enough or more than enough to meet their needs (90% compared with 84%). The cohort we have examined is a stratified subsample of surviving traced members of the Boyd Orr cohort. Originally, between 1937 and 1939, 1,352 families and 7,920 individuals took part in the study. In 1997, a subsample of the child participants, aged between 5 and 14 years at the time of the first survey, was randomly selected (from a possible n = 3,762 cases) using household food expenditure per capita data to stratify the original sample, which had been weighted toward financial deprivation. These individuals were resurveyed between 1997 and 1999 (n = 294) and were found to be broadly representative of their age peers at the 1991 census (Blane, Berney, Davey Smith, Gunnell, & Holland, 1999), with the exception of ethnicity, that is, all study participants were White European British (for further details of the Boyd Orr cohort, see Blane, 2005). Subsequent resurveys have gathered health and qol data. Most recently, between 2003 and 2005, the data for the current analysis were collected by postal questionnaire and home visits from 253 potential respondents (attrition due to death of 14% of the 1997–1998 respondents). The results reported in the current study are based on 174 of the 253 who had complete questionnaire information (nonresponse due to refusal, n = 22; loss through moves and ill health, n = 24; item nonresponse, n = 33). Participants are aged between 68 and 82 years, the mean age was 75 years. Although respondents were somewhat healthier and more likely to be married (Table 1), this sample Measures Resilience of qol Our primary outcome measure was resilience defined as better-than-average qol in the face of significant adversity. Quality of life was measured using a validated continuous scale to capture this concept: the CASP-19 self-completion questionnaire (Hyde, Wiggins, Higgs, & Blane, 2003; Wiggins, Higgs, Hyde, & Blane, 2004). As the name implies, this measure has 19-items that reflect the four dimensions of control, autonomy, self-realization, and pleasure. These four domains are argued to theoretically constitute qol at older ages (Higgs, Hyde, Wiggins, & Blane, 2003). Each item is scored 0–3 using a 4-point Likert scale, producing a scale range of 0 to 57; the higher the score the better the qol. We defined better-than-average qol as a value of CASP-19 above the median. The median cutoff for CASP-19 was chosen because it is seen to realistically capture flourishing or unexpected positive 38 The Gerontologist work (Blane, 2005). This score was dichotomized at the median point of 54 years. outcomes, as behavior that falls within or is higher than the expected average for a normative cohort (Masten & Powell, 2003). Social Networks Psychosocial and Health-Related Adversity Social networks were measured using a number of items. First, we measured the type of contact provided by friends and family, including the quality of relationships, which was measured by aggregating a seven-item 3-point score. The seven items were having people that make me happy, can be relied on, accept me, support and encourage me, make me feel loved, will see me taken care of, and make me feel an important part of their lives. This score was ranked and dichotomized for use in the analysis. Other types of contact were measured dichotomously (yes/no) in relation to family and friends providing practical support or emotional support, or being available to socialize. Density of contact was measured in several ways. First, we asked about the number of close confiding relationships; having more than two close confiding relationships was considered high density. We also asked respondents to count the number of friends and family they were close to; this total was dichotomized into high and low at the median (which was 5 and 6, respectively) to determine if respondents had a wide circle of family and/or friends. Frequency of contact was measured by frequency of telephone contact or meeting up with other people; this was measured separately for friends and family. If contact was reported in the past week, this was considered frequent. The adversities we examine are being limited by illhealth or (in the past 5 years) deteriorating health, having more stress, changing life circumstances, being worse off financially, and experiencing a negative or difficult event such as bereavement. The adversities were measured in several ways. Study participants were asked about change during the previous 5 years in health, financial circumstances, levels of stress, and general living circumstances. Respondents who answered to one or more of these items “a bit worse now” or “a lot worse now” were categorized as having been exposed to adversity, as were participants who reported limitations from their illness and a negative life event. Negative life events were recorded using an open-ended question, coded subsequently for bereavement, the illness of a loved one, and other events experienced as difficult. Hildon and colleagues (2008) have shown that the same event, for example, bereavement (the most commonly reported negative event in old age) or retirement can both be experienced as acute and chronic. For example, when daily routines with partners are fixed, trying to continue these postbereavement or indeed not being able to postretirement can serve to produce constant (chronic) reminders of loss. Therefore, the negative events variable is seen to be able to capture both acute and chronic effects on qol at the time of the survey. In compiling this variable, care was taken to exclude events such as financial problems and personal health problems, which were already measured. Although there is undoubtedly an overlap between these changes and events, all adversities were judged sufficiently distinct to test as a cumulative variable. Community-Related Factors A good sense of community was identified by summing four items developed to capture this domain (Wiggins et al., 2004): “having a lot of friendly neighbors,” “people looking out for each other,” “a good community spirit,” and “having a good mix of people.” Integration in the community was computed by identifying those who reported involvement in paid employment, voluntary work (full time or part time), or community organizations. Sociodemographic Factors Sociodemographic information was compiled from current and previous waves of data collection. This includes a measure for weekly food expenditure per capita as a measure of family income in childhood, dichotomized at the median point of 165 pence. In addition, we consider life course exposure to hazards score, which aggregates the number of years exposure to poor diet, occupational fumes and dust, and physically arduous Vol. 50, No. 1, 2010 Coping Styles Coping was measured using a 20-item coping scale developed based on existing literature (Lazarus, 1980; Staudinger et al., 1999), focus 39 Because adversities constitute changes or events that tend to reduce qol, individuals who maintain the average qol experience for the cohort and better can be seen to be managing exceptionally well or flourishing. The natural comparator to resilient outcomes was identified as vulnerable outcomes, measured as below-average qol (CASP-19 below the median), in the face of at least one validated adversity. Crosstabulation of resilient and vulnerable outcomes with individual and social attributes allowed us to compare the characteristics and levels of adversity exposure of both groups; chi-square test was used to test whether these associations were statistically significant. groups, expert consensus, and factorial analyses (see Appendix). The coping styles were as follows: 1. 2. 3. Avoidance: coping, which sidesteps engaging in solutions or problem solving by, for instance, focusing on negative feelings, for example, feeling like life has no meaning, stuck with a problem, or pretending that problems are not happening. This seems to equate with being overwhelmed by adversity. Adaptation: solution-driven coping, for example, teaching oneself how to live with a problem, which could be equated to integrating adversity. Development: again, solution-driven coping, for example, learning from a problem, which could be equated to moving beyond adversity in a positive way. Examining Attributes That Attenuate the Negative Impact of Adversity To identify which individual or social variables might be effective in attenuating the negative impact of adversity on qol, logistic regression analyses were carried out using CASP-19 as the dichotomous dependent variable. This consisted of stratified analyses comparing the role of sociodemographic, social networks, and community, along with coping styles in subgroups of low and high adversity exposure. High exposure is taken to mean reporting two or more adversities and low exposure to mean reporting only one or no exposures to adversity. All analyses were adjusted for age, sex, and income but variables were not mutually adjusted. There are four possible relationships that can be identified by this analysis. The first is that the tested variable significantly improves qol in the subgroup exposed to high levels of adversity but not when adversity is low. This association can be interpreted as attenuating the effect of adversity on qol or promoting resilience; its effect kicking in only in the presence of adversity. Should a variable significantly improve qol in both subgroups, the factor could be considered to be generally protective, independent of adversity. Alternatively, a variable may significantly improve qol only in the low-adversity group, thus having only a weak protective effect. Last, a variable may have no significant association with qol in both subgroups, which demonstrates its lack of overall importance in terms of protection and resilience. Scores for each of the three dimensions were calculated by summing the endorsements given for each of their items. All three items were ranked and dichotomized into high and low for use in subsequent analysis. Methods of Analysis Analyses were carried out using SPSS 14.0 for Windows. Examining Health-Related and Psychosocial Adversity We first tested whether the five selected adversities were associated with below-average qol. For this purpose, we tested the difference in the mean CASP-19 score between the participants who were not exposed to adversity and those who were exposed to a limiting illness, deteriorating health, more stress, changing life circumstances, being worse off financially, and experiencing a negative life event. This was achieved by considering mean CASP-19 and the five p values using Student’s t test. Only the adversity variables that showed significant reduction in CASP-19 scores, which were hence validated as producing an average adverse effect, were used in subsequent analysis. The cumulative impact of these adversities on CASP-19 scores was examined to assess if the impact of adversities was additive. Results Examining Resilient Characteristics Describing the Sample Resilient outcomes were identified by betterthan-average qol (CASP-19 on or above the median) and exposure to one or more validated adversities. The mean score for the CASP-19 scale was 40.3 (scores ranged from 14 to 56). The distribution of 40 The Gerontologist about 40% and psychosocial about 20%). The combo chart (Figure 1) shows the mean differences in CASP-19 scores between those exposed to a particular adversity and those who were not. Generally, exposure to an adversity was associated with −8 points difference in mean CASP-19 scores. The maximum drop occurred in those who felt life in general was worse (−11.24 points) and the smallest drop was for finance getting worse (−2.13 points), which was also the only difference that did not reach statistical significance. We did not use the financial adversity variable in analysis beyond this point. Analysis of the combined effects of adversity showed that as the number of adversities increased, the impact on qol also increased (Figure 2). This suggests that the five adversities validated by the earlier analysis also had additive effects. For example, the mean CASP-19 score was lower by more than one third (38%) for those exposed to five adversities compared with those exposed to one adversity. CASP-19 scores was negativity skewed; however, repeat analyses using raw CASP-19 scores and a score transformed for normality showed no differences between them (results not shown). Moreover, there were no outliers in our data so that no single observation can be seen to influence results. In terms of the sociodemographic descriptors, not previously reported, 64.4% of this sample had nonmanual occupations (measured using the Registrar General’s Social Class occupational classification schema) and 74% reported access to a car. The mean number of years of exposure to hazard over the life course was 62.6 (scores ranged from 0 to 188), and mean household expenditure per capita in childhood (in pence) was 180.1 (scores ranged from 75 to 452). In relation to social networks, 83.2% of sample participants with complete data reported relying on family for practical support and 85.1% reported relying to them for emotional support; only 28.7% relied on friends for practical support and 19.2% for emotional support. In contrast, 70.7% reported socializing with friends and 51.7% with family. About 79.3% of participants reported meeting or speaking to a family member in the past week and 63.2% reported this type of contact with friends. Mean counts for having a wide circle of family was 4.3 (counts ranged from 0 to 30) and was 5.7 for having a wide circle of friends (counts ranged from 0 to 26). About 28.7% of the sample reported having several (more than two) close confiding relationships, and the mean of scores for the quality of relationships scale was 19.5 (scores ranged from 9 to 21). Relative to community factors, 60.3% of participants with complete data reported integration in the community; the mean score for reporting a good sense of community was 2.3 (scores ranged from 0 to 4). Last, relative to coping styles, the sample mean for scores summing developmental coping was 5.3 (scores ranged from 0 to 9); the mean for adaptive coping styles scoring was 2.7 (scores ranged from 0 to 5); and the mean for scores summing avoidant coping items was 1.5 (scores ranged from 0 to 6). Who Are the Resilient? Of the 174 respondents with complete data, 55 (31%) were classified as resilient, 79 (45%) were classified as vulnerable, and 40 (24%) showed a normative response. The vulnerable group tended to report being exposed to a greater number of adversities—More than one exposure was reported by 87.3% of the vulnerable compared with only 47.3% of the resilient (p < .001). The sociodemographic profile of the resilient and the vulnerable groups was very similar with the exception of age: The resilient were significantly younger compared with the vulnerable (Table 2). There was some suggestion that women might also dominate the resilient group (65.5% women among the resilient as compared with 49.4% in the vulnerable, p = .065). Several aspects of social relations were more prevalent among the resilient than the vulnerable (Table 3), including good quality of relationships (p = .021), practical support from family (p = .019), several close confiding relationships (p = .005), having a wide circle of family (p = .030) and friends (p = .020), contact in the past week with family (p = .049) and friends (p = .019), and social integration in the community (p = .049). Coping styles were found to characterize both resilient and vulnerable outcomes (Table 4). More than 60% of the resilient used highly developmental What Typifies Psychosocial and Health-Related Adversity? The adversities varied in their prevalence with 124 of 174 reporting at least one negative life event (Figure 1). However, other adversities affected smaller proportions of the sample (health related Vol. 50, No. 1, 2010 41 Figure 1. Prevalence of adversities and their impact on CASP-19 scores; the gray square signifies statistically nonsignificant difference in means. results. None of the sociodemographic variables including age suggested an effect whereby the full impact of adversity was attenuated. However, having enough or more income than needed seemed to reduce qol in those who were exposed to higher levels of adversity (0.998, 95% confidence interval [CI]: 0.282–0.769). In contrast, two variables relative to social networks and community as well as (p = .010) and highly adaptive (p = .044) coping styles compared with about 45% of the vulnerable who reported using both these types of strategies. In contrast, highly avoidant coping styles were far more prevalent (67%) in the vulnerable group as compared with the resilient group (46%, p = .018). Which Individual or Social Attributes Are Effective in Attenuating the Negative Impact of Adversity? Table 2. Prevalence of Sociodemographic and Life Course Factors in the Resilient (n = 55) and Vulnerable (n = 79) Groups Analyses examining factors that attenuate the negative impact of adversity are based on n = 95 cases for high exposure to adversity and n = 79 cases for low exposure to adversity. These small numbers might explain the relative lack of positive Sociodemographic and life course factors Aged younger than 75 years Female Married or cohabiting Higher educationa Owner occupier Nonmanual occupationb Car access Enough or more than enough income Low life course hazard exposure High childhood per capita food expenditure Resilient (%) Vulnerable (%) p value 65.5 46.8 .033 65.5 58.2 45.5 83.6 65.5 74.1 87.8 49.4 70.9 48.1 81.0 65.8 72.2 77.6 .065 .128 .763 .697 .965 .806 .154 49.1 52.6 .693 49.1 50.6 .861 Note: aQualifications beyond Levels 1–3 of National Key Learning Targets (General Certificate of Secondary Education, O Level, A Level, National Vocational Qualifications Levels 1–3), including graduate and postgraduate academic qualifications and some vocational or professional qualifications. b Based on the Registrar General’s Social Class occupational classification schema. Figure 2. Quality of life (mean CASP-19 score) in the presence of increasing number of adversities; means and 95% confidence intervals. 42 The Gerontologist Table 4. Prevalence of Coping Styles in the Resilient (n = 55) and Vulnerable (n = 79) Groups Table 3. Prevalence of Social Networks and Community Factors in the Resilient (n = 55) and Vulnerable (n = 79) groups Social networks and community factors Good quality of relationships Practical support from family Practical support from friends Emotional support from family Emotional support from friends Socializes with family Socializes with friends Several close confiding relationships Wide circle of family Wide circle of friends Frequent contact with family Frequent contact with friends Good neighborhood community spirit Integrated into community Resilient (%) Vulnerable (%) p value 66.0 45.5 .021 92.6 77.2 .019 22.2 35.4 88.9 87.3 .788 25.9 15.2 .125 59.3 75.9 81.5 50.6 68.4 58.2 .327 .343 .005 58.5 58.5 75.5 39.2 38.0 58.7 .030 .020 .049 63.0 42.1 .019 48.1 32.9 .077 85.5 70.9 .049 High developmental coping High adaptive coping High avoidant coping Vulnerable (%) p value 67.3 44.2 .010 63.5 46.2 45.5 67.1 .044 .018 is almost double the reported prevalence range for resilience in the literature (Tusaie & Dyer, 2004). Due to sample size limitations, we were forced to use a generous threshold (better than average) to represent flourishing, as a result the prevalence of resilient outcomes in our sample may be somewhat inflated. To some extent, the same bias could be attributed to the findings that describe resilient characteristics of older people or to the prevalence of attributes defining the resilient and the vulnerable groups. However, such limitations are not seen to compromise the analysis of factors that attenuate the impact of adversity (regression analysis). Analysis and use of the adversity construct presented us with further dilemmas; adversities can cluster within individuals, thus creating an overlap between the adversities we used. In children, this had been shown to be the case and without much specificity between adversities in causing negative adult outcomes (Kessler, Davis, & Kendler, 1997). Therefore, our use of adversities additively is justified and was supported by the data that showed a strong gradient relative to CASP-19 scores as these adversities compounded. The disadvantage of this approach was that we were unable to detect whether adversity-specific resilience exists. Just one of the six adversities tested, financial circumstances worsening, did not reach statistical significance, only marginally affecting CASP-19 scores, and was excluded from further analysis. This lack of negative association may be due to a generational effect: the Boyd Orr generation grew up under the shadow of World War II and were likely to have experienced far more poverty in childhood than more recent generations. This could have made them inherently more adaptable to such changes; the steeling effect alluded in the quote by Nietzsche at the beginning of this article. Financial circumstances were examined further as a component of protection in subsequent analysis (having enough or more than enough income). Paradoxically, the better off had less probability of having better-than-average qol. This might be a the two solution-driven coping styles were found to attenuate the expected negative impact of high adversity exposure, having seemingly no effect when adversity was low (Table 5). These factors included good quality of relationships (5.105, 95% CI 1.323–19.699), integration in the community (10.800, 95% CI 1.227–95.014), highly adaptive coping style (3.211, 95% CI 1.041–9.910), and highly developmental coping style (3.397, 95% CI 1.079–10.690). Such attributes can be argued to promote resilience because they are mobilized to buffer, transform, or negate the full potential impact of clearly present adversity. No variables produced generally protective effects (improved outcomes in both strata). Having a wide circle of family and friends improved outcomes only in the low–adversity exposure strata (respectively, 5.988, 95% CI 1.401–25.593; 5.073, 95% CI 1.294–19.888), as did having several close confiding relationships (5.811, 95% CI 1.347–25.059). Discussion Resilience, as we define it here, was found in almost one third of the Boyd Orr participants. This Vol. 50, No. 1, 2010 Resilient (%) .103 Coping styles 43 Table 5. Factors That Confer a Protective Effect in the High–Adversity Exposure Strata (n = 95) But Have No Effect in the Low–Adversity Exposure Strata (n = 79)a Levels of adversity Protective factors Good quality of relationships Integration in the community High developmental coping High adaptive coping High, odds ratio (95% confidence interval) Low, odds ratio (95% confidence interval) 5.105 (1.323–19.699) 10.800 (1.227–95.014) 3.397 (1.079–10.690) 3.211 (1.0 41–9.910) 3.843 (0.917–16.097) 2.976 (0.791–11.197) 3.823 (0.858–17.040) 2.065 (0.543–7.856) a Note: All characteristics were examined, only significant findings are reported here. they also knew to draw on them when needed. Research has shown that existential concerns and agency that produce this dynamic are equally present in older cognitively impaired populations, who are active participants in producing and sustaining qol (Hennessy, 2004). Although psychologically resilient was primarily characterized by the capacity of adaptation and development, sociologically resilient can be defined by a process of exchange. Netuveli, Wiggins, Montgomery, Hildon, and Blane (2008) described resilience as the process that converts social goods into good outcomes. In their longitudinal study of resilience as bouncing back, the authors found that social support before and during adversity promoted resilience. Social capital is widely seen as the combination of social relationships that gives access support (Gray, 2009). Our findings about good quality of relationships and integration within the community are in agreement with Netuveli and colleagues’ and seem to point toward social capital as a key variable in the resilience process. Like in the present study, Netuveli and colleagues also found that the potential for resilience was reduced by concomitant adversities. In lieu of the stratified approach we chose, other authors have used interaction testing to identify relationships between adversity and personal attributes that produce resilience. Interaction testing has been termed notoriously unstable (Luthar et al., 2000) and for this reason can be difficult to interpret. Roosa (2000) argues that in certain highrisk samples, interaction can be disguised as main effects. Specifically, should researchers find a main effect associated with positive adaptation in a highrisk group that is distinct from results on a mainstream (low risk) group, as was the case with our data, this relationship suggests an interaction. The data set was rich in a range of measures that formed the basis of exploratory analysis but limited by small sample size. We recommend more focused quantitative research drawing on a larger sample manifestation of a kind of disappointment paradox (Osika & Montgomery, 2008) whereby expected benefits from an advantageous factor are overshadowed by the disappointment of being exposed to an adversity. This negative result suggests, in terms of resilience at older ages, that insufficient income may not be as threatening and indeed sufficient income may not be as protective as other, perhaps less tangible, circumstances. Accordingly, socioeconomic variables (e.g., home ownership) were not associated with resilient outcomes, in contrast to the range of social network and community and coping characteristics that did characterize resilience. Several of these seemingly protective characteristics also appeared to attenuate the full impact of adversity that was clearly present and threatening (two or more reports of adversity): two of them were psychological (developmental and adaptive coping style) and two were social (good quality of relationships and integration in the community). Because analysis of resilient characteristics showed that the resilient were more likely to have access to these forms of protection in the first place, the resilient can be viewed as more advantaged than the vulnerable from the outset. It is also interesting that avoidant coping was associated with vulnerability. These combined results, taken with the finding that the vulnerable reported more adversities than the resilient participants (who were nevertheless also exposed to significant adversity), suggest that balancing processes are at work. And, that a threshold of adversity, enough to define resilience but not enough to engender vulnerability, is pivotal to these processes. Moreover, balancing processes seemingly hinge on avoiding multiple adversity exposures as well as the capacity to acquire and use psychological and social resources to offset potentially vulnerable outcomes. Indeed, Werner and Smith (1982) found that a distinguishing feature of resilient children was not only that they had access to social support networks but 44 The Gerontologist and experiences, which help to replace losses prompted by bereavement or leaving work. In this vein, we recommend greater funding provision to enable active retirement, particularly for the retiring disabled populations, who are more often unable to access community resources (Llewellyn, 2004). Polices that focus on productive aging and giving older people legitimate roles postretirement are at present difficult to formulate. Many of the tasks taken on by older people, such as grandchild care and domestic duties, are characterized by their informal nature; some might argue that this is perhaps one of the best things about such practices. We would argue that in terms of resilience, the role of policy is to enable social engagement as a positive aspect of aging; productivity in old age seems only to generate beneficial transactions when freely chosen (Hildon et al., 2006). Overall therefore, perhaps the greatest challenge for resilience-centered research is to move beyond findings into the mainstream, where the benefits of social capital and harnessing protective coping are understood by both practitioner and lay people, with the result that ultimately, the focus on the “right” way to acquire and use available resources is integrated and understood as a key aspect of successful aging. size and more diverse elderly populations to substantiate/refute our findings. Because quantitative analysis can suggest process but may not always fully capture the subtleties or dynamics that define them, we also recommend qualitative or mixedmethods analysis. Early in the study of resilience, Werner (1995, p. 84) noted that “if we want to help vulnerable youngsters become more resilient we need to decrease their exposure to potent risk factors and increase their competencies … as well as sources of support they can draw on.” In the same vein, relative to older ages and given the findings on the central role increased adversity plays in resilience and vulnerability, our recommendations for policy are similarly two pronged: we must find effective ways to minimize adversity and its impact where possible, and to promote systems and services that can deliver support when necessary (Hildon et al., 2008). Policies to increase the potential for resilience at older ages should include a focus on improving access to good quality social relationships and integration in the community as well as enabling and engaging positive coping strategies. Such polices should be geared toward helping to maintain enjoyable and supportive social relations and activities (Blane, Wiggins, Montgomery, Hildon, & Netuveli, 2009), and facilitating self-management which builds and draws on positive coping patterns (Skinner & Vaughan, 1983). Interventions that stress self-care through building on problem-solving coping strategies have proved particularly successful in promoting resilient outcomes (Hill-Briggs & Gernmell, 2007; Laforest et al., 2008). As for socially focused policies, these include the recent extension of the London Freedom Pass to all older people in England or facilitating gradual retirements and part-time voluntary positions for older people. It is tempting to focus policy frameworks for older ages around care issues surrounding the Fourth Age, at which time the onset of disability has become severe. However, enabling older people to sustain and develop new places in the community, through the Third Age (known as the period postretirement before the onset of severe disability), is equally important. For example, by making student loans available to retirees (Hildon, Netuveli, & Blane, 2006), or by creating nationalized schemes that channel prospective volunteers’ skills, retirements can be improved. Such policies would enable resilience to be attained and acquired resources would carry through to the onset of severe disability, for example, new and enriching social networks Vol. 50, No. 1, 2010 Funding Economic and Social Research Council Grant L326253061. Acknowledgments We gratefully acknowledge the Boyd Orr respondents for their continued participation and commitment. References Becker, G., & Newsom, E. (2005). Resilience in the face of serious illness among chronically ill African Americans in later life. Journal of Gerontology Series B: Psychological Sciences and Social Sciences, 60, S214–S223. Blane, D. (2005). Cohort profile: The Boyd Orr lifegrid sub-sample— Medical sociology study of life course influences on early old age. International Journal of Epidemiology, 34, 750–754. Blane, D., Berney, L., Davey Smith, G., Gunnell, D., & Holland, P. (1999). Reconstructing the life course: Health during early old age in a follow-up study based on the Boyd Orr cohort. Public Health, 113, 117–124. Blane, D., Wiggins, R. D., Montgomery, S. M., Hildon, Z., & Netuveli, G. (2009). Resilience at older ages: Implications for policy (Working paper). International Centre for Life Course Studies in Society and Health. Bowling, A., & Dieppe, P. (2005). What is successful ageing and who should define it? British Medical Journal, 331, 1548–1551. Butler, J., & Ciarrochi, J. (2007). Psychological acceptance and quality of life in the elderly. Quality of Life Research, 16, 607–615. Fries, J. F. (1980). Aging, natural death, and the compression of morbidity. New England Journal of Medicine, 303, 130–135. Garmezy, N. (1985). Stress-resistant children: The search for protective factors. In J. Stevenson (Ed.), Recent research in developmental psychology (pp. 213–233). Journal of Child Psychology and Psychiatry (Book Suppl. No. 4). Oxford, England: Pergamon Press. Gray, A. (2009). The social capital of older people. Ageing and Society, 29, 5–31. 45 Osika, W., & Montgomery, S. M. (2008). Economic disadvantage modifies the association of height with low mood in the US, 2004: The disappointment paradox. Economics and Human Biology, 6, 95–107. Richardson, G. (2002). The metatheory of resilience and resiliency. Journal of Clinical Psychology, 58, 307–321. Roosa, M. (2000). Some thoughts about resilience versus positive development, main effects versus interaction, and the value of resilience. Child Development, 71, 567–569. Rowe, J. W., & Kahn, R. L. (2000). Successful aging and disease prevention. Advances in Renal Replacement Therapy, 7, 70–77. Rutter, M. (1987). Psychosocial resilience and protective mechanisms. American Journal of Orthopsychiatry, 57, 316–331. Ryff, C. D., Singer, B., Love, G. D., & Essex, M. J. (1998). Resilience in adulthood and later life: Defining features and dynamic processes. In J. Lomranz (Ed.), Handbook of aging and mental health: An integrative approach (pp. 69–96). New York: Plenum. Schoon, I. (2006). Risk and resilience. Cambridge, England: Cambridge University Press. Skinner, B. F., & Vaughan, M. E. (1983). Enjoy old age: A program of self-management. New York: W.W. Norton. Staudinger, U., Freund, A., Linden, M., & Maas, I. (1999). Self, personality, and life regulation: Facets of psychological resilience in old age. In P. B. Bates, & K. U. Mayer (Eds.), The Berlin aging study (pp. 302–328). Cambridge, England: Cambridge University Press. Tusaie, K., & Dyer, J. (2004). Resilience: A historical review of the construct. Holistic Nursing Practice, 18, 3–8. Wagnild, G., & Young, H. M. (1990). Resilience among older women. Image: Journal of Nursing Scholarships, 22, 252–255. Werner, E. (1990). Protective factors and individual resilience. In S. Meisels, & J. Shonkoff (Eds.), Handbook of early intervention (pp. 97–116). Cambridge, England: Cambridge University Press. Werner, E. (1995). Resilience in development. Current Directions in Psychological Science, 4, 81–85. Werner, E., & Smith, R. (1977). Kauai’s children come of age. Honolulu: University of Hawaii Press. Werner, E., & Smith, R. (1982). Vulnerable but invincible: A longitudinal study of resilient children and youth. New York: McGraw Hill. Wiggins, R. D., Higgs, P., Hyde, M., & Blane, D. (2004). Quality of life in the third age: Key predictors of the CASP-19 measure. Ageing and Society, 24, 693–708. Windle, G., & Woods, R. T. (2004). Variations in subjective wellbeing: The mediating role of a psychological resource. Aging and Society, 24, 583–602. Grotberg, E. H. (1997). The international resilience project: Findings from the research and the effectiveness of interventions. In B. Bain, H. Janzen, J. Paterson, L. Stewin, & A. Yu (Eds.), Psychology and education in the 21st century: Proceedings of the 54th Annual Convention of the International Council of Psychologists (pp. 118–128). . Banff, Alberta, Canada, July 24–28, 1996. Edmonton, Alberta, Canada: IC Press. Harris, P. B. (2008). Another wrinkle in the debate about successful aging: The undervalued concept of resilience and the lived experience of dementia. International Journal of Aging & Human Development, 67, 43–61. Hennessy, C. H. (2004). Conclusion. In A. Walker, & C. H. Hennessy (Eds.), Growing older: Quality of life in old age (pp. 225–229). New York: Open University Press. Higgs, P., Hyde, M., Wiggins, R. D., & Blane, D. (2003). Researching quality of life in early old age: The importance of the sociological dimension. Social Policy & Administration, 37, 239–252. Hildon, Z., Netuveli, G., & Blane, D. (2006). Social participation, social support and resilience in older people. In M. Bartley (Ed.), Capability & resilience: Beating the odds (pp. 20–21). London: University College London Department of Epidemiology and Public Health. Hildon, Z., Smith, G., Netuveli, G., & Blane, D. (2008). Understanding adversity and resilience at older ages. Sociology of Health & Illness, 30, 1–15. Hill-Briggs, F., & Gernmell, L. (2007). Problem solving in diabetes selfmanagement and control—A systematic review of the literature. Diabetes Educator, 33, 1032–1050. Hyde, M., Wiggins, R. D., Higgs, P., & Blane, D. (2003). A measure of quality of life in early old age: The theory, development and properties of a needs satisfaction model (CASP-19). Aging & Mental Health, 7, 186–194. Jacelon, C. (1997). The trait and process of resilience. Journal of Advanced Nursing, 25, 123–129. Kessler, R. C., Davis, C. G., & Kendler, K. S. (1997). Childhood adversity and adult psychiatric disorder in the US National Comorbidity Survey. Psychological Medicine, 27, 1101–1119. Kind, P., Dolan, P., Gudex, C., & Williams, A. (1998). Variations in population health status: Results from a United Kingdom national questionnaire survey. British Medical Journal, 316, 736–741. Kinsel, B. (2005). Resilience as adaptation in older women. Journal of Women & Aging, 17, 23–39. Laforest, S., Nour, K., Gignac, M., Gauvin, L., Parisien, M., & Poirier, M. C. (2008). Short-term effects of a self-management intervention on health status of housebound older adults with arthritis. Journal of Applied Gerontology, 27, 539–567. Lazarus, R. (1980). The stress and coping paradigm. In L. Bond & J. Rosen (Eds.), Competence and coping during adulthood (pp. 28–74). New Hampshire, NH: University Press of New England. Llewellyn, G. (2004). Promoting healthy, productive ageing: Plan early, plan well. Journal of Intellectual & Developmental Disability, 29, 366–369. Luthar, S. S., Ciccheti, D., & Becker, B. (2000). The construct of resilience: A critical evaluation and guidelines for future work. Child Development, 71, 543–562. Martin, J., Meltzer, H., & Elliot, D. (1988). OPCS Surveys of disability in Great Britain: The prevalence of disability among adults. London: Her Majesty’s Stationery Office. Masten, A. S. (2001). Ordinary magic: Resilience processes in development. American Psychologist, 56, 227–238. Masten, A. S., Best, K., & Garmezy, N. (1990). Resilience and development: Contributions from the study of children who overcome adversity. Development and Psychopathology, 2, 425–444. Masten, A. S., & Powell, J. L. (2003). A resilience framework for research, policy and practice. In S. S. Luthar (Ed.), Resilience and vulnerability: Adaptation in the context of childhood adversities (pp. 1–25). Cambridge, England: Cambridge University Press. Netuveli, G., Montgomery, S. M., Hildon, Z., Wiggins, R. D., & Blane, D. (2006). Social and psychological determinants of resilience in old age in England. European Journal of Epidemiology, 21, 60. Netuveli, G., Wiggins, R. D., Montgomery, S. M., Hildon, Z., & Blane, D. (2008). Mental health and resilience in older ages: Bouncing back from adversities in British Household Panel Survey. Journal of Epidemiology and Community Health, 62, 987–991. Nietzsche, F. (1984). The twilight of idols. In W.Kaufmann (Ed. & Trans.), The portable Nietzsche (pp. 463–563). New York: Penguin Books. (Original work published 1888). Received December 3, 2008 Accepted March 16, 2009 Decision Editor: William J. McAuley, PhD Appendix: Method of Development and Analysis of a Coping Index Focus Groups The coping items were developed and evaluated using focus groups. Focus group members were aged 68–84 years and lived in Fife (6 participants, 1 man and 5 women), Wirral (8 participants, 4 men and 4 women), and London (6 participants, 2 men and 4 women), which are areas containing sizeable clusters of Boyd Orr cohort members. Initially, ways of coping were explored by guiding discussion using open-ended questions about ways of dealing with specific adversities such as health problems or bereavement. Transcripts were analyzed for the types of coping styles that were described and a draft list of coping items complied. 46 The Gerontologist Drawing on the Literature and Expert Consensus three, and four factors. The best fit was obtained with the three-factor solution (root mean square error approximation = 0.03; see Table A1 for factor loadings). Furthermore, using the same data, we did a confirmatory factor analysis using the three-factor structure. The analysis showed moderate fit (Tucker Lewis Index [TLI] = 0.84), which improved (TLI = 0.91) after modifying the model so that one item, “life has no meaning,” was allowed to correlate with all factors. We used the three factors from the EFA as separate explanatory variables in our analyses. In the second focus group, specific items in this draft were discussed. These items were then reworked based on feedback and existing literature on coping, including Staudinger and colleagues’ (1999) study exploring facets of psychological resilience and Lazarus’s (1980) classic conception of coping as problem-solving or emotion-regulatory functions. Items were finalized based on discussion from a panel of four experts in gerontology and social medicine and feedback from the last focus group that discussed the clarity—rewording, order, and applicability—of specific items. Factor Labeling These factors represent (a) avoidance, or ways of not engaging with adversity, such as feeling stuck with the problem; (b) adaptation, or ways of integrating adversity, for instance, by teaching oneself how to live with a problem; and (c) developmental coping, or moving beyond adversity in a positive way, such as learning from a problem (see Appendix Table 1 for item loadings). Scores for each of the three dimensions were calculated by summing the endorsements given for each of their items. All three items were ranked and dichotomized into high and low for use in subsequent analysis. Factor Analysis Once the item list was compiled and data gathered, we examined the resulting 20-item index. We explored its structure by exploratory factor analysis (EFA) using promax rotation. Because the items were binary, we used MPlus version 4 for the analysis. We also chose promax (oblique) over varimax (orthogonal) rotation because the former allows for factors to be correlated, which is a valid assumption as we have no reason to believe coping skills are independent of each other. The EFA tested two, Appendix Table 1 Factor Loadings for the 20 Items Describing Coping a Component labels and item loadings Coping items Development Learn from the experience Be able to adapt to this change Bounce back Throw myself into activities Past successes mean I will solve this problem too Plan how to deal with the situation Sense of humor will get me through Seek information Be stronger Able to help others if they experience similar events Teach myself to live with this problem I will never really recover Be appreciative of the simple things in life Draw strength from faith Get through this on my own Numb my feelings with drugs and alcohol Pretend it has not happened Improve with the support of others I will be stuck with the problem There is no meaning to life Adaptation Avoidance 0.782 0.652 0.598 0.496 0.682 0.780 0.767 0.572 0.735 0.804 0.850 0.717 0.522 0.494 0.418 0.496 0.524 0.607 0.802 0.848 Note: aThe loadings are of factors generated by promax rotation, highest loadings are indicated and lower factor loadings omitted for convenience. Vol. 50, No. 1, 2010 47