Introduction
Over the last few decades, researchers have increasingly investigated the ways in which individuals respond to illness-related adversities to maintain or regain a sense of wellness in their lives, especially over the latter stages of the life course (Sells et al., Reference Sells, Sledge, Wieland, Walden, Flanagan and Miller2009; Windle, Woods, & Markland, Reference Windle, Woods and Markland2010). A prevalent health condition experienced among people of advanced age is multimorbidity, which is the co-occurrence of two or more chronic illnesses (Salive, Reference Salive2013). It has been estimated that at least two thirds of older adults in many developed countries, including Canada, the United States, and Australia have been diagnosed with two or more chronic conditions (Islam et al., Reference Islam, Valderas, Yen, Dawda, Jowsey and McRae2014; Wister et al., Reference Wister, Coatta, Schuurman, Lear, Rosin and MacKey2016), and these rates rise among the very old. For example, based on United States clinical data, Salive (Reference Salive2013) reports prevalence rates of multimorbidity in the United States of 62 per cent among persons 65–74 years of age, 75.7 per cent among persons 75–84 years of age, and 81.5 per cent among persons ≥ 85 years of age. Multimorbidity is receiving increasing attention as an important health condition, given that it is associated with several deleterious outcomes, including loses of functional ability, social roles, and quality of life (Galenkamp, Braam, Huisman, & Deeg, Reference Galenkamp, Braam, Huisman and Deeg2011), as well increased utilization of health care resources and associated costs (Agborsangaya, Lau, Lahtinen, Cooke, & Johnson, Reference Agborsangaya, Lau, Lahtinen, Cooke and Johnson2013).
Multimorbidity, and the presence of chronic illnesses generally, are typically conceptualized as pathogenic processes with significant adversity often tied to a disablement process defined as difficulty accomplishing desired activities as a result of health or physical challenges (Verbrugge & Jette, Reference Verbrugge and Jette1994). Yet, even in the face of objective illness, most older adults report higher than expected levels of well-being, successful aging, life satisfaction, and perceived health (Wister, Levasseur, Griffiths, & Fyffe, Reference Wister, Levasseur, Griffiths and Fyffe2015), what has been termed the “well-being paradox” (Netuveli & Blane, Reference Netuveli and Blane2008). One interpretation of this apparent contradiction is that some individuals can utilize attitudes, beliefs, resources, and behaviours in order to adapt or cope with the deleterious consequences of multimorbidity, resulting in a sense of healthy aging. Furthermore, in some instances, persons managing co-occurring chronic illnesses may be able to “bounce back” from illness-related adversities, which is called “resilience” (Resnick et al., Reference Resnick, Gwyther and Roberto2019; Windle, Reference Windle2011). One area of interest is the extent to which health behaviours play a role in shaping perceptions of healthy aging among those with multimorbidity, given that a positive view of aging can be instrumental in chronic illness management (Antonovsky, Reference Antonovsky1996; Ong, Bergeman, Bisconti, & Wallace, Reference Ong, Bergeman, Bisconti and Wallace2006; Wister, Kendig, Mitchell, Fyffe, & Loh, Reference Wister, Kendig, Mitchell, Fyffe and Loh2016). Since health behaviours are mutable and have the potential to improve population health and reduce strains on health care systems, it is not surprising that researchers are interested in their role in adaptations to multimorbidity (Short & Mollborn, Reference Short and Mollborn2015).
Thus, when older people with multiple chronic conditions report high levels of healthy aging, this may be indicative of resilience in that they perceive themselves to be coping with adversity (Pruchno & Carr, Reference Pruchno and Carr2017). Perceptions of how well one is aging may be a potentially useful approach to investigating resilience. While as there is a lack of agreement pertaining to resilience measures and operationalization (Cosco et al., Reference Cosco, Kaushal, Hardy, Richards, Kuh and Stafford2017; Cosco, Kok, Wister, & Howse, Reference Cosco, Kok, Wister and Howse2019), global measures, such as appraisals of healthy aging, successful aging, happiness, and well-being, have the potential to capture the inherent complexity of underlying concepts such as resilience (Galenkamp et al., Reference Galenkamp, Braam, Huisman and Deeg2011). Furthermore, while resilience models have been applied to numerous forms of adversity in older persons including chronic illness and multimorbidity (Rybarczyk, Emery, Guequierre, Shamaskin, & Behel, Reference Rybarczyk, Emery, Guequierre, Shamaskin, Behel, Hayslip and Smith2012; Sells et al., Reference Sells, Sledge, Wieland, Walden, Flanagan and Miller2009; Trivedi, Bosworth, & Jackson, Reference Trivedi, Bosworth, Jackson, Resnick, Gwyther and Roberto2011; Wister & Cosco, Reference Wister and Cosco2021; Wister, Kendig, et al., Reference Wister, Kendig, Mitchell, Fyffe and Loh2016), there is a need for further research pertaining to the predictors, processes, and mechanisms associated with resilience, including the role of health behaviours in later life.
Therefore, this article addresses two primary research questions. (1) What are the modifiable behavioural lifestyle factors that are associated with healthy aging, adjusting for key covariates? and (2) Do these patterns differ across multimorbidity types and clusters? This has tremendous import, because it has been estimated that health behaviours account for approximately 40 per cent of health care costs in the United States (McGinnis, Williams-Russo, & Knickman, Reference McGinnis, Williams-Russo and Knickman2002), and furthermore, health promotion interventions need to be tailored and targeted to maximize their potential (Kendig, Browning, Thomas, & Wells, Reference Kendig, Browning, Thomas and Wells2014; Rootman, Pederson, Frohlich, & Dupéré, Reference Rootman, Pederson, Frohlich and Dupéré2012).
Social Determinants and Health Behaviours for Healthy Aging
A social determinant of health model (Raphael, Reference Raphael2016) is incorporated in order to connect healthy lifestyle behaviours and perceptions of healthy aging and to frame the analyses. This approach views health as a social construct that traverses holistic, multi-level, and cumulative process over the life course. Social determinants of health include the nexus of a number of factors often deemed to increase inequality from policy (upstream) down to the individual (downstream) levels (Braverman, Egerter, & Williams, Reference Braverman, Egerter and Williams2011; Short & Mollborn, Reference Short and Mollborn2015). This approach began by emphasizing the social milieu of health: social contexts, such as socio-economic status; living and working conditions; the physical environment, and social support from family, friends, and the community Prus Reference Prus2011; Raphael, Reference Raphael2016; Rootman et al., Reference Rootman, Pederson, Frohlich and Dupéré2012). It has been integrated within federal public policy approaches in Canada, such as the Population Health Model (Hamilton & Bhatti, Reference Hamilton and Bhatti1996). More recently, there has been an application of the original model to health behaviours, what we call the Social Determinants of and Health Behaviour Conceptual Framework (Short & Mollborn, Reference Short and Mollborn2015). It considers both health and healthy lifestyle behaviours as embedded in structure and agency, as well as with genetic and physiological determinants of health (Short & Mollborn, Reference Short and Mollborn2015). As such, they embody not only structural determinants of health (socio-economic opportunities, social support systems, living environments) but also those connected to the role of social-psychological determinants, such as perceptions of healthy aging (Kendig et al., Reference Kendig, Browning, Thomas and Wells2014). The latter builds on Antonovsky’s (Reference Antonovsky1996) salutogenic processes connected to quality of life and well-being, which are distinct from the pathogenic clinical causes and trajectories of psychopathology and disease.
Furthermore, the literature on resilience and aging has investigated a range of areas within which individuals harness resources to foster positive adaptation to illness, including psychological, emotional, spiritual, physical/functional, economic, cultural, and social or ecological domains (Cosco, Wister, Brayne, & Howse, Reference Cosco, Wister, Brayne and Howse2018; Linkov & Kott, Reference Linkov and Kott2019; Ong et al., Reference Ong, Bergeman, Bisconti and Wallace2006; Reference Ong, Zautra and Reid2010; Silverman, Molton, Alschuler, Ehde, & Jensen, Reference Silverman, Molton, Alschuler, Ehde and Jensen2015; Wiles, Wild, Kerse, & Allen, Reference Wiles, Wild, Kerse and Allen2012; Wister, Rosenkrantz, Shashank, Walker, & Schuurman, Reference Wister, Rosenkrantz, Shashank, Walker and Schuurman2020). Applied to multimorbidity, some individuals may (or may not) possess important resilience attributes, such as healthy lifestyle routines, social support resources, economic resources, and social-psychological strengths that may generate positive appraisals of healthy aging (Kendig et al., Reference Kendig, Browning, Thomas and Wells2014; Rybarczyk et al., Reference Rybarczyk, Emery, Guequierre, Shamaskin, Behel, Hayslip and Smith2012; Sells et al., Reference Sells, Sledge, Wieland, Walden, Flanagan and Miller2009; Trivedi et al., Reference Trivedi, Bosworth, Jackson, Resnick, Gwyther and Roberto2011; Windle, Reference Windle2012). We consider positive perceptions of healthy aging among persons with multimorbidity as an indicator of resilience, given that individuals with multimorbidity typically face significant adversities, and appraisals of healthy aging capture elements of successful aging, positive adaptation, comparisons with peers, and temporal dynamics of aging processes, as these appraisals denote a life course element. This approach parallels research that has used self-rated health as a global indicator of complex concepts such as health-related quality of life, and as a predictor of morbidity and mortality (Galenkamp et al., Reference Galenkamp, Braam, Huisman and Deeg2011; Idler & Cartwright, Reference Idler and Cartwright2018).
A core phase of the resilience process is the accessing and activation of internal and external resources from a strength-based lens, including health behavours, social support, and socio-economic status (Richardson, Reference Richardson2002; Sells et al., Reference Sells, Sledge, Wieland, Walden, Flanagan and Miller2009; Resnick et al., Reference Resnick, Gwyther and Roberto2019). Although appraisals of healthy aging do not constitute a formal resilience measure (Cosco et al., Reference Cosco, Kok, Wister and Howse2019), they can be a crude indicator among individuals facing adversity. This article focuses on health behaviours, because they are deemed to be central components of healthy aging among persons faced with the physical, social, and psychological adversities of multimorbidity (Kendig et al., Reference Kendig, Browning, Thomas and Wells2014; Wister, Cosco, Mitchell, & Fyffe, Reference Wister, Cosco, Mitchell and Fyffe2020).
Behavioural/Lifestyle Predictors of Healthy Aging
Health behaviours have been established as primary predictors of well-being in old age, morbidity, and survival. Research has begun to link several lifestyle factors to multimorbidity, including smoking, physical activity, and obesity/nutrition (Agborsangaya, Ngwakongnwi, Lahtinen, Cooke, & Johnson, Reference Agborsangaya, Ngwakongnwi, Lahtinen, Cooke and Johnson2013; Autenrieth et al., Reference Autenrieth, Kirchberger, Heier, Zimmermann, Peters and Döring2013; Canizares, Hogg-Johnson, Gignac, Glazier, & Badley, Reference Canizares, Hogg-Johnson, Gignac, Glazier and Badley2017; Wister, Reference Wister2019). For example, Agborsangaya, Ngwakongnwi, et al. (Reference Agborsangaya, Ngwakongnwi, Lahtinen, Cooke and Johnson2013) found that obesity increased the number and clustering of chronic illnesses, and Canizares et al. (Reference Canizares, Hogg-Johnson, Gignac, Glazier and Badley2017) identified successive cohort increases (being born between 1925 and 1974) in multimorbidity between 1994 and 2010 that were inflated by being obese, a smoker, and sedentary. The positive effects of physical activity as well as the quality and quantity of sleep have also been identified as promoting coping and recovery from multimorbidity (Ezeamama et al., Reference Ezeamama, Elkins, Simpson, Smith, Allegra and Miles2016; Wister, Reference Wister2005). These health behaviours are considered to be health deficits or credits that affect the ability of individuals to manage and respond to both objective (e.g., symptoms) and subjective (e.g., low self-efficacy) illness-related stressors, harness social connections and support, and enhance well-being (Jones et al., Reference Jones, Stathokostas, Young, Wister, Chau, Clark, Duggan, Mitchell and Nordland2018; Kendig et al., Reference Kendig, Browning, Thomas and Wells2014; Pruchno & Wilson-Genderson, Reference Pruchno and Wilson-Genderson2012; Wister, Cosco, et al., Reference Wister, Cosco, Mitchell and Fyffe2020). Thus, we hypothesize that the above health behaviours will be associated (either negatively or positively) with healthy aging in older adults even after accounting for socio-demographic, socio-economic, and illness-related covariates.
Additional covariates are potential predictors of healthy aging based on known epidemiological influences on multimorbidity. These can be separated into modifiable (education, income level, living/housing and urban/rural environment, and social support), and non-modifiable factors (genetics, age, foreign born status) (Canizares et al., Reference Canizares, Hogg-Johnson, Gignac, Glazier and Badley2017; Sells et al., Reference Sells, Sledge, Wieland, Walden, Flanagan and Miller2009; Ungar, Reference Ungar2011; Windle, Reference Windle2012). The type and severity of illnesses contributing to multimorbidity, and perceptions of pain, functional ability, and co-morbidity, represent illness contextual factors that should also be included in an analysis of appraisals of healthy aging and resilience (Ong, Zautra, & Reid, Reference Ong, Zautra and Reid2015; Peters et al., Reference Peters, Cosco, Mackey, Sarohia, Leong and Wister2019; Sells et al., Reference Sells, Sledge, Wieland, Walden, Flanagan and Miller2009). For example, greater perceived pain will act as a barrier to good health, as it can be debilitating to an individual’s ability to function (Trivedi et al., Reference Trivedi, Bosworth, Jackson, Resnick, Gwyther and Roberto2011). In addition, functional status influences the ability to complete activities of daily living.
Methods
Design and Sample
The Canadian Longitudinal Study on Aging (CLSA) data set was utilized in this research. This 20-year panel study of persons 45–85 years of age was initiated in 2010, funded primarily by the Canadian Institutes for Health Research (CIHR), Canada’s federal granting agency for health research. A large set of multidisciplinary data was collected at baseline between 2012 and 2015 (see Raina et al., Reference Raina, Wolfson, Kirkland, Griffith, Oremus and Patterson2009 and Raina et al., Reference Raina, Wolfson, Kirkland, Griffith, Balion and Cossette2019 for full sampling details) for the 51,338 participants in the CLSA.
In order to obtain the sample of participants for this research, two CLSA data sets were merged. This included the comprehensive CLSA data set and the tracking CLSA data set. The comprehensive data set consisted of 30,097 participants who were randomly selected within age/sex strata from within 25 km of dense population data sites, or within 50 km of data collection sites in areas with a lower population density. There were 11 data collection sites located in seven provinces across the country. The tracking data set included a cohort of 20,341 participants who were interviewed with a computer-assisted telephone interviewing device. These tracking participants were randomly selected within age/sex strata from areas outside of data collection site zones throughout the rest of Canada. Sample weights were used to correct for sampling error by age, gender, and geographic location (Kirkland et al., Reference Kirkland, Griffith, Menec, Wister, Hélène and Wolfson2015).
Given our interest in multimorbidity among older adults, our study included only persons 65 years of age or older who reported having two or more chronic conditions. The sample included self-reports of 2 or more of 27 physical chronic conditions in the CLSA. The chronic illnesses used in this study included Alzheimer’s disease, back problems, bowel incontinence, cancer, cataracts, diabetes, epilepsy, glaucoma, heart attack, heart disease, high blood pressure, irritable bowel syndrome, kidney disease, Parkinson’s disease, peripheral vascular disease, lung disease, macular degeneration, multiple sclerosis, osteoarthritis, osteoporosis, migraine headaches, rheumatoid arthritis, stroke, thyroid problems, transient ischemic attack, ulcer, and urinary incontinence. Because there is a lack of uniformity in adding mental health variables, such as anxiety, as a chronic illness rather than as an outcome of physical illness (Schafer et al., Reference Schafer, von Leitner, Schön, Koller, Hansen and Kolonko2010; Sells et al., Reference Sells, Sledge, Wieland, Walden, Flanagan and Miller2009), we conducted a separate parallel analysis of a mental health multimorbidity cluster (described subsequently).
For this study we employed a subset of the participants who were 65 years of age and older and were diagnosed with at least two multimorbid chronic conditions (n = 17,148). We used weighted frequencies for the descriptive analyses, and unweighted frequencies for the multivariate analyses, which when weighted equaled 12,272 (because of oversampling of older age groups) participants.
Measures
The validity and reliability of all relevant measures in the CLSA questionnaires, as well as the references, can be found on the data portal of the CLSA Web site (wwww.clsa-elcv.ca).
Dependent Variable Development
The primary dependent variable is derived from an ordinal scale based on the question: "In terms of your own healthy aging, would you say it is excellent, very good, good, fair, or poor?" The original Likert ordinal measure ranged from “excellent”, “very good”, “good”, “fair”, to “poor”. This healthy aging variable was dichotomized into: “excellent”, “very good”, and “good” (1) and “fair”, and “poor” (0), with 87.3 per cent in the positive appraisal group (see Table 1). We contend that older adults with multimorbidity who report higher levels of healthy aging show higher levels of coping and resilience.
Table 1. Descriptive data for dependent and independent variables for Canadians 65 years of age or older with two or more chronic conditions, CLSA weighted (n = 12,272)
Note: CLSA = Canadian Longitudinal Study on Aging; OARS ADL = Older Americans Resources and Services Activities of Daily Living.
Socio-Demographic Variables
As shown in Table 1, the following socio-demographic independent variables were included: age, gender, educational level, total household income, marital status, and immigration status. Age was an interval variable and ranged from 65 to 85. Sex was a dichotomous variable coded with females as the reference category. The education level variable assessed the highest degree, certificate, or diploma that the participant had obtained. This ordinal variable was measured with the following dummy coded categories: “no post-secondary degree, certificate, or diploma” (reference) “trade certificate or diploma”, “bachelor’s degree”, “university degree or certificate above bachelor’s degree”. Total household income was also dummy coded using the available categories in the data set as possible responses: “less than $20,000” (reference), “$20,000–$49,999”, “$50,000–$99,999”, “$100,000–$149,999”, “$150,000 or above”, and “missing.” Marital status was dichotomized as “single, never married or never lived with a partner”, “widowed”, “divorced”, “separated” (reference), and “married/living with a partner in a common-law relationship”. Immigration status was dichotomized into “born in Canada” and “immigrant”.
Social and Environmental Variables
Number of friends, number of relatives, housing problems, and urban/rural status were included as social-environmental covariates. The variable measuring number of friends assessed how many people a respondent considered to be close friends. Total scores for this continuous variable ranged from 0 to 90. Number of relatives was a count of how many other living family members reported in the participant’s network and had a range of 0 to 100. The housing problems variable was computed based on an additive scale based on a series of questions that asked participants if they had answered ‘no’ to the following: condensation; problems with electrical wiring, plumbing, or heating; infestations; leaking; and noise, as well as answering ‘yes’ to maintenance and repairs. The urban/rural status variable was dichotomized into “rural” and “urban.”
Behavioural Lifestyle Variables
Behavioural lifestyle factors/indicators included body mass index (BMI), physical inactivity, smoking, restless sleep, skipped meals and poor appetite. These were derived from multimorbidity and healthy aging research studies related to lifestyle factors (Agborsangaya, Ngwakongnwi, et al., Reference Agborsangaya, Ngwakongnwi, Lahtinen, Cooke and Johnson2013; Autenrieth et al., Reference Autenrieth, Kirchberger, Heier, Zimmermann, Peters and Döring2013; Canizares et al., Reference Canizares, Hogg-Johnson, Gignac, Glazier and Badley2017; Kendig et al., Reference Kendig, Browning, Thomas and Wells2014; Wister, Cosco, et al., Reference Wister, Cosco, Mitchell and Fyffe2020). BMI was dummy coded into categories using normal (18.5–24.9) as the reference group, underweight (≤ 18.49), overweight (25–29.9), and obese (≥ 30). The physical inactivity measure was based on the Physical Activity Scale for the Elderly (PASE) (Smith, Reference Smith1991). This variable asked how many hours per day a respondent engaged in sitting activities, which has been associated with poor health outcomes (Guralnik et al., Reference Guralnik, Simonsick, Ferrucci, Glynn, Berkman and Blazer1994). This variable used the response set: sitting “less than 1 hour”, “1 hour but less than 2 hours”, “2 hours but less than 4 hours”, and “4 hours or more”. The smoking variable was used based on responses to the question of whether or not a respondent had smoked cigarettes in the past 30 days. Restless sleep was a single item measure taken from the 10-item Center for Epidemiological Studies Depression Scale that assessed how frequently a respondent experienced restless sleep per week. This variable had the possible responses: “all of the time (5–7 days)”, “occasionally (3–4 days)”, “some of the time (1–2 days)”, or “rarely or never (< 1 day)”. Lastly, two eating variables were also included. The first was a variable measuring the degree to which a participant generally described his or her appetite. This variable included the responses: “poor”, “fair”, “good”, and “very good”. The second variable was based on an ordinal response to a question as to the frequency of skipping meals. This was dichotomized into “rarely or never” and “sometimes to always”.
Health Context Variables
Scales measuring pain, the Older Americans Resources and Services (OARS) physical activities of daily living scale, and multimorbidity scales were incorporated in order to capture the illness context. Pain was measured using an ordinal variable based on responses to a question pertaining to the usual intensity of pain or discomfort and included “none”, “mild”, “moderate”, and “severe”. The OARS physical activities of daily living scale measured the degree to which a respondent experienced restriction-free physical mobility. This variable ranged in scores from 2 to 14 with higher scores representing better mobility. Multimorbidity was measured based on a scale of 27 chronic conditions including: Alzheimer’s disease, back problems, bowel incontinence, cancer, cataracts, diabetes, epilepsy, glaucoma, heart attack, heart disease, high blood pressure, irritable bowel syndrome, kidney disease, Parkinson’s disease, peripheral vascular disease, lung disease, macular degeneration, multiple sclerosis, osteoarthritis, osteoporosis, migraine headaches, rheumatoid arthritis, stroke, thyroid problems, transient ischemic attack, ulcer, and urinary incontinence. This base multimorbidity 27 additive scale was included as the final health factor, because individuals can have more than two chronic illnesses. This variable was called “co-morbidity” for the specific cluster conditions because it included concurrent chronic conditions beyond those included in each cluster.
The independent variable with a significant amount of missing data was total household income. The missing data were added as a separate category and were dummy coded. The small amount (< 5%) of missing data for the remaining variables was recoded to the respective means (continuous) or modes (categorical). Missing data were also removed and analyses repeated with no changes to the substantive findings.
Data Analysis
After validating assumptions, binary logistic regression modeling was employed using SPSS 24 in order to determine the associations between the dichotomized healthy aging variable and four sequentially ordered blocks of predictors: (1) socio-demographic, (2) social/environmental, (3) behavioural lifestyle, and (4) illness context. Logistic regression is a technique that has been specifically developed to examine dichotomous associations, especially those that are positively skewed such as the one used in this research. Model odds ratios (ORs), ORs, and significance levels are reported. In addition, logistic regression analyses of healthy aging were conducted on individuals with multimorbidity (two or more conditions), as well as the three multimorbidity clusters. Logistic regression removes the weighting when generating models.
Multimorbidity Cluster Subsample Rationale
Additional analyses were conducted on three multimorbidity clusters. Although the “two or more illnesses” definition has been widely used in multimorbidity studies, some researchers contend that this approach does not account for the fact that illnesses vary in their influence on aging, and furthermore, that some illnesses more than others have been found to occur concurrently (Prados-Torres et al., Reference Prados-Torres, Poblador-Plou, Calderón-Larrañaga, Gimeno-Feliu, González-Rubio and Poncel-Falcó2012). Two of the most prevalent statistical methods include latent class analysis (see Whitson et al., Reference Whitson, Johnson, Sloane, Cigolle, Pieper and Landerman2016) and tetrachoric factor analysis (see Schafer et al., Reference Schafer, von Leitner, Schön, Koller, Hansen and Kolonko2010). Although there is a lack of consensus pertaining to illness clusters to date, we identify, a priori, three separate clusters for subsequent analyses that have a high degree of research support because of their overlap in occurrence and symptoms. These entail: (1) a cardiovascular and metabolic cluster (n = 5,112), including heart disease, diabetes, and high blood pressure (Cornell, Pugh, & Williams, Reference Cornell, Pugh and Williams2007; Freund, Kuna, Ose, Szecsenyi, & Peters-Klimm, Reference Freund, Kuna, Ose, Szecsenyi and Peters-Klimm2012; Holden et al., Reference Holden, Scuffdham, Hilton, Muspratt, Ng and Whiteford2011; Kirchberger et al., Reference Kirchberger, Meisinger, Heier, Zimmermann, Thorand and Autenrieth2012; Ng, Holden, & Sun, Reference Ng, Holden and Sun2012; Schafer et al., Reference Schafer, von Leitner, Schön, Koller, Hansen and Kolonko2010); (2) a musculskeletal cluster (n = 2,245) composed of osteoarthritis, osteoporosis, and chronic lower back problem (Holden et al., Reference Holden, Scuffdham, Hilton, Muspratt, Ng and Whiteford2011; Schafer et al., Reference Schafer, von Leitner, Schön, Koller, Hansen and Kolonko2010); and (3) a mental health cluster (1,156), given that mood disorder, anxiety disorder, and migraine headache are often found to co-occur (Kirchberger et al., Reference Kirchberger, Meisinger, Heier, Zimmermann, Thorand and Autenrieth2012; Schafer et al., Reference Schafer, von Leitner, Schön, Koller, Hansen and Kolonko2010) (Table 2).
Table 2. Hierarchical logistic regression of perceived healthy aging among multimorbid (≥2 conditions, age ≥65), CLSA unweighted (n = 17,148)
Note. *p < 0.05, **p < 0.01, ***p < 0.001.
Ref = reference category; CLSA = Canadian Longitudinal Study on Aging; CI = confidence interval; OARS = Older Americans Resources and Services; BMI = body mass index.
Results
Two or More Multimorbid Conditions Logistic Regression Analysis
Only results in the final model are presented for each subsequent section. Among the sample of persons 65 years of age or older who had two or more chronic conditions, healthy aging was found to be associated with 14 variables (model χ2 =1,175.89, p < 0.001). For the socio-demographic model, a negative association was found for being male rather than female (OR = 0.50, p < 0.001, confidence interval [CI] 0.43–0.58); positive associations for age (OR = 1.03, CI 1.02–1.05, p < 0.001); education level: trade certificate or diploma (OR = 1.24, CI 1.06–1.44, p < 0.05), bachelor’s degree (OR = 1.37, CI 1.11–1.69, p < 0.01), university or certificate above bachelor’s degree (OR = 1.71, CI 1.34–2.17, p < 0.001), compared with no post-secondary degree or certificate; household income: $50,000–$99,999 (OR = 1.47, CI 1.13–1.92, p < 0.01), $100,000–$149,999 (OR = 1.98, CI 1.36–2.89, p < 0.001), $150,000 and higher (OR = 1.88, CI 1.16–2.90, p < 0.01), compared with less than $20,000; and an inverse association with one provincial contrast, Saskatchewan (OR = 0.62, CI 0.43–0.90, p < 0.05), compared with Ontario. Among the social/environmental model, healthy aging was associated with number of friends (OR = 1.02, CI 1.01–1.03, p < 0.001), and number of relatives (OR = 1.01, CI 1.00–1.01, p < 0.05). Turning to the behavioural/lifestyle model, the factors associated with healthy aging include both positive and inverse associations with body mass index: overweight (OR = 1.25, CI 1.05–1.49, p < 0.05), obese (OR = 0.78, CI 0.65–0.93, p < 0.01), compared with normal weight; and positive associations with not smoking (OR = 1.40, CI 1.11–1.77, p < 0.01), compared with smoking; restless sleep: occasionally (OR = 1.52, CI 1.24–1.86, p < 0.001), some of the time (OR = 1.56, CI 1.29–1.88, p < 0.001), rarely or never (OR = 1.67, CI 1.39–1.99, p < 0.001), compared with all of the time; and appetite: good (OR = 2.20, CI 1.53–3.16, p < 0.001), very good (OR = 2.53, CI 1.75–3.65, p < 0.001), compared with poor. The illness context model variables inversely associated with healthy aging was pain: mild (OR = 0.78, CI 0.64–0.95, p < 0.05), moderate (OR = 0.56, CI 0.47–0.65, p < 0.001), severe (OR = 0.42,CI 0.33–0.53, p < 0.001), compared with none; the OARS physical activities of daily living scale was positively associated (OR = 1.30, CI 1.22–1.38, p < 0.001); and the multimorbidity 27 scale exhibited an inverse relationship with healthy aging (OR = 0.80, CI 0.77–0.82, p < 0.001) (Table 3).
Table 3. Hierarchical logistic regression of perceived healthy aging among multimorbid (≥ 2 conditions, ≥ 65 years of age) cardiovascular and metabolic cluster, CLSA unweighted (n = 5,112)
Note. *p < 0.05, **p < 0.01, ***p < 0.001
Ref = reference category; CLSA = Canadian Longitudinal Study on Aging; CI = confidence interval; OARS = Older Americans Resources and Services; BMI = body mass index.
Cardiovascular and Metabolic Cluster Logistic Regression Analysis
Ten variables were found to be associated with healthy aging for the sample of persons 65 years of age and older who had two or more of the following chronic conditions: heart disease, diabetes, and high blood pressure (model χ2 = 369.73, p < 0.001). For the socio-demographic model, these factors were age (OR = 1.02, CI 1.00–1.04; p < 0.05); an inverse association with being male rather than female (OR = 0.59, CI 47–0.74, p <0.001); and positive associations with education level: university or certificate above bachelor’s degree (OR = 1.51, CI 1.04–2.19, p < 0.05), compared with no post-secondary degree or certificate; and, household income: $50,000–$99,999 (OR = 1.78, CI 1.20–2.64, p < 0.01), $100,000–$149,999 (OR = 1.77, CI 1.02–3.07, p < 0.05), compared with less than $20,000. Variables associated with healthy aging among the behavioural/lifestyle model included restless sleep: occasionally (OR = 1.37, CI = 1.00–1.84), p < 0.05, some of the time (OR = 1.53, CI 1.15–2.03, p < 0.01), rarely or never (OR = 1.53, CI 1.18–2.0, p < 0.01), compared with all of the time; and better appetite: good (OR = 2.33, p < 0.001, CI 1.38–3.92, p < 0.01) or very good (OR = 2.64, CI 1.56–4.46, p < 0.001), compared with poor. The illness context model factors associated with healthy aging were less pain: moderate (OR = 0.70, CI 0.54–0.88, p < 0.00), severe (OR = 0.42, CI 0.30–0.59, p < 0.001), compared with none; greater functional status using the OARS physical activities of daily living scale (OR = 1.27, CI 1.16–1.39, p < 0.001); and the cardiovascular and metabolic comorbidity scale (OR = 0.85, CI 0.81–0.89, p < 0.001) (Table 4).
Table 4. Hierarchical logistic regression of perceived healthy aging among multimorbid (≥2 conditions, ≥65 years of age), musculoskeletal cluster, CLSA unweighted (n = 2,245)
Note. *p < 0.05, **p < 0.01, ***p < 0.001
Ref = reference category; CLSA = Canadian Longitudinal Study on Aging; CI = confidence interval; OARS = Older Americans Resources and Services; BMI = body mass index.
Musculoskeletal Cluster Logistic Regression Analysis
For the sample of persons 65 years of age and older who had two or more of the following chronic conditions: osteoarthritis, osteoporosis, and chronic lower back problems, healthy aging was also found to be associated with nine variables (model χ2 = 227.71, p < 0.001). In the socio-demographic model, healthy aging was positively associated with age (OR = 1.04, CI 1.01–1.07, p < 0.05); inversely associated with being male rather than female (OR = 0.67, CI 0.46–0.96, p < 0.05); and being an immigrant (OR = 0.53, CI 0.36–0.78, p < 0.01), compared with being native born. The behavioural/lifestyle model factors associated with healthy aging were high body mass index: being obese (OR = 0.55, CI 0.36–0.84, p < 0.01), compared with being a normal weight; not smoking (OR 1.72, CI 1.01–2.94, p < 0.05); restless sleep: occasionally (OR 1.74, CI 1.09–2.79, p < 0.05), some of the time (OR = 1.72, CI 1.12–2.66, p < 0.05), rarely, or never (OR = 1.74, CI 1.16–2.61, p < 0.01), compared with all of the time; and better appetite: fair (OR = 0.28, CI 0.10–0.77, p 0.05), compared with poor. Variables associated with healthy aging in the illness context model included less pain: moderate (OR = 0.58, CI 0.40–0.86, p < 0.01), severe (OR = 0.51, CI 0.30–0.87, p < 0.05), compared with none; greater functional ability based on the OARS physical activities of daily living scale (OR = 1.32, CI 1.16–1.51, p < 0.001); and inversely associated with the musculoskeletal comorbidity scale (OR = 0.85, CI 0.79–0.90, p < 0.001) (Table 5).
Table 5. Hierarchical logistic regression of perceived healthy aging among multimorbid (≥2 conditions, ≥65), mental health cluster, CLSA unweighted (n = 1,156)
Note. *p < 0.05, **p < 0.01, ***p < 0.001
Ref = reference category; CLSA = Canadian Longitudinal Study on Aging; CI = confidence interval; OARS = Older Americans Resources and Services; BMI = body mass index.
Mental Health Cluster Logistic Regression Analysis
Among the sample of persons 65 years of age or older who had two or more of the following chronic conditions: anxiety disorder, mood disorder, and migraine headache, healthy aging was found to be associated with 11 variables (model χ2 = 162.56, p < 0.001). For the socio-demographic model, healthy aging was positively associated with age (OR = 1.08, CI 1.03–1.13, p < 0.01); inversely with being male rather than female (OR = 0.49, CI 0.30–0.80, p < 0.01); and Province: Newfoundland and Labrador (OR = 0.38, CI 0.16–0.92, p < 0.05), rather than Ontario. Healthy aging was also positively associated with the social/environmental model factors, including: number of friends (OR = 1.07, CI 1.01–1.14, p < 0.05) and urban status (OR = 1.87, CI 1.05–3.34, p < 0.05), compared with rural. The behavioural/lifestyle model factors associated with healthy aging were high body mass index: being obese (OR = 0.54, CI 0.31–0.93, p < 0.05), compared with being normal weight; restless sleep: occasionally (OR = 2.68, CI 1.42–5.08, p < 0.01), rarely or never (OR = 1.80, CI 1.06–3.06, p < 0.05); and better appetite: very good (OR = 3.10, CI 1.27–7.54, p < 0.05), compared with poor. Among the illness context variables, healthy aging was inversely associated with a moderate level of pain (OR = 0.61, CI 0.38–0.99, p < 0.05) compared to none; positively associated with the OARS physical activities of daily living scale (OR = 1.47, CI 1.24–1.74, p < 0.001); and inversely with the mental health co-morbidity scale (OR = 0.89, CI 0.82–0.97, p < 0.05).
Discussion
The concurrence of two or more chronic illnesses is increasingly common among older adults, and presents a set of adversities that limit physiological, psychological and social dimensions of aging (Agborsangaya, Ngwakongnwi, et al., Reference Agborsangaya, Ngwakongnwi, Lahtinen, Cooke and Johnson2013). This article extends our understanding of how and why some individuals are able to perceive themselves as aging in a healthy manner based on an examination of social determinants of health and health behaviours. Perceptions of healthy aging, similar to research using self-rated health, can be a powerful global indicator of health-related quality of life, and may be indicative of resilience among persons facing the everyday adversities connected to multimorbidity (Galenkamp et al., Reference Galenkamp, Braam, Huisman and Deeg2011). It is argued that appraisals of healthy aging, albeit not well understood, reflect comparisons with other older adults, but also capture elements of process linked to aging. Using baseline data from the CLSA, this study examined several major health behaviours that have been associated with morbidity, mortality, and health-related quality of life. Our analyses focus on several types of multimorbidity adversity: (1) older persons reporting multimorbidity – operationalized as two or more conditions; and 2) those reporting two or more illnesses that fall within one of the three major identified clusters of illnesses.
Positive appraisals of healthy aging in the face of multimorbidity adversity are strongly related to several major health behaviours, although both similarities and differences occur across the multimorbidity clusters under examination. For multimorbidity and the three illness clusters, healthy aging is consistently associated with not smoking (except for the mental health cluster), not being obese (except the cardiovascular and metabolic cluster), better sleep, and a better appetite, after adjusting for all variables. Smoking habits, obesity, and food security (availability, access, utilization, stability) represent three health behaviours and/or contexts that typically find their place in health behaviour and aging research (Kendig et al., Reference Kendig, Browning, Thomas and Wells2014; Short & Mollborn, Reference Short and Mollborn2015) and associated research focusing on multimorbidity risk (Agborsangaya, Ngwakongnwi, et al., Reference Agborsangaya, Ngwakongnwi, Lahtinen, Cooke and Johnson2013; Canizares et al., Reference Canizares, Hogg-Johnson, Gignac, Glazier and Badley2017; Skivington, Katikireddi, Leyland, Hunt, & Mercer, Reference Skivington, Katikireddi, Leyland, Hunt and Mercer2015). Smoking can also negatively affect appraisals of healthy aging and resilience among persons with multiple chronic illnesses, perhaps because of its addictive properties. Obesity increases the seriousness of related symptoms, and illness coping responses (Agborsangaya, Ngwakongnwi, et al., Reference Agborsangaya, Ngwakongnwi, Lahtinen, Cooke and Johnson2013; Canizares et al., Reference Canizares, Hogg-Johnson, Gignac, Glazier and Badley2017; Kendig et al., Reference Kendig, Browning, Thomas and Wells2014). We also found that a good appetite augments positive appraisals which is consistent with research that establishes the importance of food security and multimorbidity risk in old age (Sharkey, Reference Sharkey2003). Furthermore, although less studied, sleep quality also enhances a sense of healthy aging among those with multimorbidity and underscores its influence for patterns of recovery and coping that parallels other research (Ezeamama et al., Reference Ezeamama, Elkins, Simpson, Smith, Allegra and Miles2016; Ong et al., Reference Ong, Zautra and Reid2015). These findings concur with many of the associations uncovered in other studies of the role of health behaviours in enhancing appraisals of healthy aging among multimorbid older adults (Stewart & Yuen, Reference Stewart and Yuen2011; Windle, Reference Windle2012; Wister, Cosco, et al., Reference Wister, Cosco, Mitchell and Fyffe2020).
It is well established that lifestyle behaviours tend to be routinized and develop over the life course of individuals. Given their salience as predictors of perceptions of healthy aging, our findings elucidate the ways in which health and healthy behaviours can be understood within a social determinants and health behaviour framework (Short & Mollborn, Reference Short and Mollborn2015) and in terms of understanding resilience as a process and the centrality of resources. Health behaviours are conceptualized as rooted in structural and individual opportunities and constraints forming salient resources that, if available and accessed, can be used to offset the deleterious aspects of multimorbidity and clusters of illnesses in older adults (Wister et al., Reference Wister, Lear, Schuurman, MacKey, Mitchell and Cosco2018). This adversity entails symptoms that restrict functioning; cause pain, stress, anxiety, and discomfort, and restrict the maintenance of healthy living more generally (Ong et al., Reference Ong, Zautra and Reid2015; Sells et al., Reference Sells, Sledge, Wieland, Walden, Flanagan and Miller2009).
Although we examine the role of health behaviours on perceptions of healthy aging among multimorbid individuals, our results are remarkedly consistent with research on lifestyle predictors of multiple chronic illnesses. For example, Skivington et al. (Reference Skivington, Katikireddi, Leyland, Hunt and Mercer2015) found support for very similar behavioural lifestyle factors on multimorbidity (two or more conditions) in a longitudinal study conducted in Scotland. After controlling for socio-demographic covariates, the authors found that multimorbidity was higher among smokers than among non-smokers (OR 1.38, 95% CI 1.20–1.60); for those with BMI 30–35 (OR 1.57, 95% CI 1.22–2.01) and > 35 (OR 2.21, 95% CI 1.40–3.48) compared with those with BMI 20–25; for those with poor diet (OR 1.28, 95% CI 1.05–1.57); and for those in the lowest compared with the highest income tertile (OR 1.29, 95% CI 1.03–1.54). These similarities reinforce the key role that health behaviours play in shaping healthy aging and multimorbidity coping processes.
The absence of support for the effect of physical inactivity on perceptions of healthy aging, while seemingly counterintuitive, is consistent with other multimorbidity research, even longitudinal studies (e.g., Autenrieth et al., Reference Autenrieth, Kirchberger, Heier, Zimmermann, Peters and Döring2013; Canizares et al., Reference Canizares, Hogg-Johnson, Gignac, Glazier and Badley2017). One explanation is that long-term inactivity may be more important for health outcomes, and/or more relevant for particular sub-groups of older adults, such that associations are not supported after adjusting for socio-demographic and social/environmental variables.
Several socio-demographic resources were also associated with healthy aging among those with multimorbidity, indicating positive resilience. Older age, being female, and higher education and income (depending on cluster) lead to positive appraisals of healthy aging. These are in the expected direction except for age. It is possible that individuals compare themselves with others their own age and/or feel that they are aging better if they are older and have lived longer. Females consistently appraise their aging better than males, perhaps because of being greater users of health and community support systems, which mirrors other research related to multimorbidity and resilience (Canizares et al., Reference Canizares, Hogg-Johnson, Gignac, Glazier and Badley2017; Wister, Cosco, et al., Reference Wister, Cosco, Mitchell and Fyffe2020). Furthermore, social support fosters positive appraisals of healthy aging among those with two or more multimorbidity conditions, and those with a mental health cluster (but not for the other two clusters). Harnessing informal social support has been shown to help older people cope with multimorbidity (Sells et al., Reference Sells, Sledge, Wieland, Walden, Flanagan and Miller2009; Wister, Cosco, Mitchell, Menec, & Fyffe, Reference Wister, Cosco, Mitchell, Menec and Fyffe2019; Zautra, Arewasikporn, & Davis, Reference Zautra, Arewasikporn and Davis2010); yet further research is needed to establish types (e.g., caregiving), and sources (spousal, other family, friends) of support on healthy aging, multimorbidity adaptation and resilience. Living in an urban environment fosters higher appraisals of healthy aging for the cardiovascular/metabolic and mental health clusters, which might be linked to access to health care resources. However, having housing problems was not inversely associated with healthy aging. Finally, as hypothesized, perceptions of lower pain levels, higher functional ability, and fewer concurrent chronic illnesses were consistently associated with healthy aging (Peters et al., Reference Peters, Cosco, Mackey, Sarohia, Leong and Wister2019). Clearly, the illness context is central to adaptation and positive appraisals of healthy aging among multimorbid older adults.
Furthermore, the recent COVID-19 pandemic has likely magnified the adversity associated with multimorbidity among older adults by increasing levels of stress, anxiety, depression, social isolation, and the risk of developing the COVID-19 disease. A higher level of adversity can result in health behaviours being more influential resources in fostering healthy aging and resilience among older persons experiencing multimorbidity (Klasa, Galaitsi, Wister, & Linkov, Reference Klasa, Galaitsi, Wister and Linkov2021). Future research needs to also concentrate on the COVID-19 era to fully explore these empirical questions, given the unique illness context that the pandemic entails (Wister & Speechley, Reference Wister and Speechley2020).
The findings in this study are limited in serval ways. First, given that the available baseline CLSA data are cross-sectional, there is a possibility of reciprocal associations between behavioural lifestyle factors and positive perceptions of healthy aging. For example, appetite or sleep quality may not only affect adaptation to multiple chronic illnesses, they may also be influenced by the degree to which an individual is coping with illness. Longitudinal data can help to disentangle causal relationships among these factors to elucidate adaptation and resilience processes over the life course (Cosco et al., Reference Cosco, Kaushal, Hardy, Richards, Kuh and Stafford2017). Second, the healthy aging measure is a single global indicator; therefore, additional measures need to be examined (see Stewart & Yuen, Reference Stewart and Yuen2011; Windle, Reference Windle2011). Further, the degree to which such a global measure captures resilience processes requires further attention, given that variable centered approaches to operationalization have been critiqued (Cosco et al., Reference Cosco, Kok, Wister and Howse2019). Third, research into healthy aging can be enhanced through the study of additional behavioural lifestyles and their interactions with other resources, which will help to map illness trajectories. Finally, this work needs to be applied to additional sub-groups, such as other disease clusters (Sells et al., Reference Sells, Sledge, Wieland, Walden, Flanagan and Miller2009); different racial/ethnic groups; and marginalized communities (Quiñones, Liang, Bennett, Xu, & Ye, Reference Quiñones, Liang, Bennett, Xu and Ye2011; Wiles et al., Reference Wiles, Wild, Kerse and Allen2012), and across a variety of environmental factors (Connelly et al., Reference Connelly, Allen, Hatfield, Palma-Oliveira, Woods and Linkov2017; Ungar, Reference Ungar2011).
Health behaviours are central to perceptions of healthy aging, which in turn, are indicative of resilience in the face of illness. These processes are instrumental in the ability to cope, adapt, and bounce back from multiple chronic illnesses and are fundamental to healthy aging (Kendig et al., Reference Kendig, Browning, Thomas and Wells2014; Windle, Reference Windle2012). Our findings indicate that there are several mutable behavioural lifestyles that are differentially associated with positive illness appraisals. These key lifestyles behaviours appear to affect the degree to which older persons perceive that they are experiencing healthy aging, but are highly dependent on the nature, types and clustering of illnesses. This research serves as a launching point for future studies that can uncover the complexity of these relationships.
Acknowledgements
This research was made possible using the data/biospecimens collected by the Canadian Longitudinal Study on Aging (CLSA). Funding for the CLSA is provided by the Government of Canada through the Canadian Institutes of Health Research (CIHR) under grant reference: LSA 94473 and the Canada Foundation for Innovation. This research has been conducted using the CLSA data set Baseline Tracking v3.4 and Comprehensive v4.0, under Application Number 150914 (https://www.clsa-elcv.ca/). The CLSA is led by Drs. Parminder Raina, Christina Wolfson, and Susan Kirkland. The current study was funded through a CIHR CLSA Catalyst Grant (RN302177 - 373073). The author also thanks Ian Fyffe for assistance with data analyses. The opinions expressed in this manuscript are the author’s own and do not reflect the views of the CLSA.
Ethics Approval
This current project received ethics approval at two levels. Consent to participate was obtained for all participants under the CLSA harmonized multi-university ethics process approved by the Hamilton Integrated Research Ethics Board (HiREB), Hamilton Health Sciences/McMaster University. Written consent was obtained from all CLSA participants prior to enrollment. Individuals who were not deemed to be cognitively functional were excluded from the CLSA study. Simon Fraser University (SFU) was a participating institution in the CLSA data collection, and the SFU Office of Research Services Ethics Committee reviewed all consent material prior to data collection (SFU ORS #2010s0281).
