Defining and Estimating Healthy Aging in Spain: A Cross-sectional Study

Abstract

Purpose of the Study

Using an operational continuum of healthy aging developed by U.S. researchers, we sought to estimate the prevalence of healthy aging among older Spaniards, inform the development of a definition of healthy aging in Spain, and foster cross-national research on healthy aging.

Design and Methods

The ELES pilot study is a nationwide, cross-sectional survey of community-dwelling Spaniards 50 years and older. The prevalence of healthy aging was calculated for the 65 and over population using varying definitions. To evaluate their validity, we examined the association of healthy aging with the 8 foot up & go test, quality of life scores and self-perceived health using multiple linear and logistic regression.

Results

The estimated prevalence of healthy aging varied across the operational continuum, from 4.5% to 49.2%. Prevalence figures were greater for men and those aged 65 to 79 years and were higher than in the United States. Predicted mean physical performance scores were similar for 3 of the 4 definitions, suggesting that stringent definitions of healthy aging offer little advantage over a more moderate one.

Implications

Similar to U.S. researchers, we recommend a definition of healthy aging that incorporates measures of functional health and limiting disease as opposed to definitions requiring the absence of all disease in studies designed to assess the effect of policy initiatives on healthy aging.

In 2015, Spain had a population of more than 46 million people, of whom 18.6% were age 65 or older (Instituto Nacional de Estadística, 2016a). In 2050, that percentage will reach 35.3% (Instituto Nacional de Estadística, 2016b). The aging of the Spanish population presents a number of public policy challenges, including the formidable task of managing the healthcare and social needs of an increasingly larger older population in the context of a publicly-funded health system and difficult economic times (Mato, 2014; Serrano, Latorre, & Gatz, 2014).

One means by which to address the effects of an aging population on national health systems is to promote heathy aging. As the World Health Organization Regional Office for Europe (2012) stated: “Healthy ageing can contribute to the sustainability of health and welfare systems in Europe…by allowing people in higher age groups to remain active, autonomous and fully integrated” (p. 3). Toward that end, Spain recently implemented the Strategy for Health Promotion and Disease Prevention in the National Health System (Mato, 2014; Ministry of Health, Social Services and Equality, 2013), a key aim of which is to foster healthy aging. Elements of the initiative include enhancing health education in primary care facilities and screening for and addressing frailty in older adults (Mato, 2014; Ministry of Health, Social Services and Equality, 2013).

In order to measure the effectiveness of programs and policies designed to promote healthy aging, a valid operational definition of healthy aging is required. As many have noted before us (e.g., Depp & Jeste, 2006; Hung, Kempen, & De Vries, 2010), however, there is lack of agreement about what healthy aging encompasses and even what it should be labeled. Notwithstanding the lack of agreement, one particular conceptualization of healthy aging—termed “successful aging”—has dominated the scientific discourse in this area (Although we prefer the label “healthy aging,” we use the equivalent term “successful aging” when referring to work that specifically used the term.).

“Successful aging”—as articulated by Rowe and Kahn (1987, 1997)—was first introduced in the 1980s and further developed in the 1990s. According to Rowe and Kahn (1997, 1998), “successful aging” includes remaining free of disease, disability, and risk factors for disease; retaining high physical and cognitive functioning; and remaining “actively engaged” (i.e., connected to others and involved in productive pursuits).

Rowe and Kahn’s conceptualization of successful aging has been heavily critiqued (Martinson & Berridge, 2015). Critics of their work have argued that few can achieve the level of health outlined by Rowe and Kahn (Masoro, 2001; Martinson & Berridge, 2015), that it overstates the role of individual behavior in successful aging and ignores the effects of sociocultural context on health and aging (Holstein & Minkler, 2003; Katz & Calasanti, 2015; Martinson & Berridge, 2015; Riley, 1998; Rubinstein & de Medeiros, 2015; Stowe and Cooney, 2015), that it insufficiently incorporates the perspective of older adults (Martinson & Berridge, 2015), and that it is missing elements of the human experience that some deem important (e.g., spirituality; Crowther, Parker, Achenbaum, Larimore, & Koenig, 2002; Martinson & Berridge, 2015). Although we appreciate these critiques, we concur with Stowe and Cooney (2015) that Rowe and Kahn’s conceptualization “warrants modification over disposal” (p. 44).

Importantly, attempts to modify Rowe and Kahn’s conceptualization have already begun. Using data from the U.S. Health and Retirement Study (HRS; Institute for Social Research, 2008), McLaughlin, Jette, and Connell (2012) examined four operational definitions of healthy aging derived from Rowe and Kahn’s conceptualization of “successful aging.” Based on their findings, they recommended altering Rowe and Kahn’s definition in a way that deemphasized “rigid disease criteria” and placed greater importance on functional health and “symptomatic disease,” as they felt this best achieved the goal of “[identifying] a group of individuals for whom the odds of adverse outcomes [were] sizably diminished while…reducing the chances of misclassifying healthy individuals as unhealthy…” (p. 788).

Although their findings are informative, it is unclear if their conclusions hold true in other national contexts such as Spain. Existing research suggests that even when similarly worded items are used, measures may not operate the same way in varying sociocultural contexts (Chan, Kasper, Brandt, & Pezzin, 2012; Jürges, 2007). To the extent that their findings do hold true in other contexts, their work facilitates the establishment of a common definition of healthy aging across the international surveys harmonized with the HRS and serves to encourage cross-national studies of healthy aging. A key strength of such studies is that they can help differentiate age-associated changes in health that stem from inherent aging processes from those that stem from differences in public policies (National Research Council, 2001). For those making difficult decisions about how best to shape public policies in the context of population aging, these types of cross-national comparisons are vitally important.

The purpose of this investigation was to (a) estimate the prevalence of healthy aging in a cohort of older Spaniards using the operational continuum established by McLaughlin et al. (2012), (b) examine demographic correlates of healthy aging, (c) inform the development of a definition of healthy aging in Spain, and (d) foster cross-national research on healthy aging.

Design and Methods

Data for this cross-sectional study come from the pilot survey of the Aging in Spain Longitudinal Study (ELES). In spite of its size (over 1,700 individuals), it is considered a pilot study because it was designed to “pilot” data collection procedures in a sample large enough to calculate precise cross-sectional statistics prior to embarking on a much larger, 20-year longitudinal study.

Details on the ELES have been published elsewhere (Rodriguez-Laso et al., 2014). Briefly, the survey included non-institutionalized people age 50 years and over living in Spain. A commercial household telephone directory served as the sample frame. The sample design involved multiple stages, with stratification by region and size of the city/town of residence. Census tracts (first-stage sampling units), households (second-stage sampling units), and informant sampling units (one individual per dwelling) were selected randomly.

Data collection took place sequentially beginning with a computer-assisted telephone interview (CATI), followed by a nurse visit and a computer-assisted personal interview (CAPI) that included a self-administered questionnaire. A total of 1,747 individuals participated in the CATI and 1,400 in the CAPI. With few exceptions, the variables included in this analysis are from the CAPI. Data were gathered in 2011. The study was approved by the Ethics Subcommittee of the Spanish National Research Council.

Definitions of Healthy Aging

Following McLaughlin et al. (2012), we evaluated four definitions of healthy aging, labeled “Rowe and Kahn” and “Level I”, “Level II”, and “Level III” healthy aging. The criteria for meeting each definition were as follows:

1. Rowe and Kahn: Freedom from disease, disability, and risk factors for disease; high physical function; and cognitive function at or above the median for one’s gender and age group.

2. Level I: Similar to Rowe and Kahn excluding the freedom from risk factors criterion and the threshold for meeting the cognitive component was freedom from severe cognitive impairment.

3. Level II: Similar to Level I except that the disease criterion included only conditions that limited a person’s ability to perform his or her usual activities.

4. Level III: Similar to Level I except that the disease criterion was entirely excluded.

Components of Healthy Aging

Cognitive Function

Participants completed a validated Spanish-version of the Mini-Mental State Examination (MMSE; Folstein, Folstein, McHugh, & Fangiang, 2001; Lobo, Saz, Marcos, & Grupo ZARADEMP, 2002). For Rowe and Kahn’s definition, high cognitive function was defined as a score above the median for the individual’s gender and age group (i.e., ages 65 to 79 or 80+ years). For Level I through Level III healthy aging, freedom from severe cognitive impairment was defined as a score of 24 and above on the 30-item MMSE (Lobo et al., 1999).

Physical Function

High physical function was defined as answering “never” (possible responses included “always,” “almost always,” “hardly ever,” or “never”) to at least 10 questions asking about difficulty carrying out 11 functions adapted from the HRS: walking 100 meters; walking more than 1,000 meters; being seated around 2 hr; standing up from a chair after being seated for a long time; climbing a flight of stairs without resting; climbing several flights of stairs without resting; bending down, kneeling or crouching; extending or lifting arms above the shoulder; pulling or pushing large objects like an armchair; lifting and carrying objects of more than 10 kilograms; and picking up a coin from a table.

Disability

Absence of disability was defined as not having difficulty carrying out any of the following activities adapted from the HRS: getting dressed, walking across a room, bathing or showering, eating, getting out of bed, using the toilet, preparing a hot meal, shopping, making telephone calls, taking medication, and managing money.

Diseases

Being free of disease was defined as not having been diagnosed by a doctor with any of the following conditions: myocardial infarction/heart attack; cardiac failure; angina; osteoarthritis, arthritis or rheumatism; chronic bronchitis, emphysema, or chronic obstructive pulmonary disease; diabetes; depression; anxiety; cerebral embolism/infarction; and malignant tumors/cancer.

Limiting Diseases

If the individual reported any of the above diseases, he or she was asked: “Scoring from 1 to 5, where 1 means not at all and 5 a lot, please tell me if this disease limits your usual activities” (Bayliss, Ellis, & Steiner, 2005). Individuals were classified as free of limiting disease if they were free of all conditions or reported a score of 1 for all diagnosed diseases.

Risk Factors

For consistency with McLaughlin et al. (2012), we considered two risk factors for cardiovascular disease: self-reported medically-diagnosed hypertension and obesity (body mass index ≥ 30 derived from measured weight and height).

Other Variables

Age was measured in years and centered at age 70 (approximate mean age of participants). Education was categorized as less than primary school, primary school, secondary school, and university degree. Classification of marital status included married, widow/widower, single, and divorced/separated. Individuals answering very good or good to the following question were classified as having optimal self-perceived health: “Generally speaking, would you say that your health status is very good, good, fair, bad or very bad?” Quality of life was measured with the Spanish version (Fernandez-Mayoralas et al., 2012) of the Personal Wellbeing Index (PWI), excluding the religiosity item (Rodriguez-Blazquez et al., 2011). Scores range from 0 to 10; individuals scoring 0 or 10 were assigned a missing value (The International Wellbeing Group, 2006). Physical performance was assessed using the 8 foot up & go test (Rikli & Jones, 2001). Interviewers measured the time it took individuals to stand up from a sitting position and sit down again after having walked 8 feet at their maximum speed.

Statistical Analyses

Consistent with McLaughlin et al. (2012), we restricted our population to individuals age 65 and over. For each definition, the prevalence of healthy aging and the percentage meeting each criterion were calculated for the total sample and by gender and age group (i.e., 65 to 79 and ≥80 years). Binary and multiple logistic regressions were used to examine the associations of age, gender, education and marital status with healthy aging according to each definition, where 1 was attributed to those who met the necessary criteria. Covariates were added together and removed one by one if they were determined not to be statistically significant (p < .05) and their removal did not change the coefficient of the principal independent variable by more than 10% nor increase its standard error (Kleinbaum & Klein, 2010). Following Hosmer and Lemeshow (2000), interactions of significant covariates in multivariate models were tested. This was done on an exploratory basis without a priori hypotheses.

To evaluate the validity of the definitions, the association between increasingly demanding definitions of healthy aging and health-related variables were examined using linear or logistic multiple regression. The dependent variables included optimal self-perceived health (logistic regression), quality of life (linear regression), and time to complete the 8 foot up & go test (linear regression). In each model, the principal independent variable was degree of healthy aging: 0 = not experiencing healthy aging, 1 = meets Level III criteria, 2 = meets Level II criteria, 3 = meets Level I criteria, 4 = meets Rowe and Kahn’s criteria. This method of classification allowed us to determine if a gradient effect existed along the continuum of healthy aging. Age, gender, marital status and education were included as covariates. As above, variables that were not statistically significant were removed if their removal did not change the coefficient of the principal independent variable by more than 10% nor increase its standard error. Predicted probabilities (self-perceived health) and mean values (quality of life and 8 foot up & go) were obtained from the final models. Requirements for logistic and linear models were checked and complex sample design features were accounted for through the specification of weights, strata and clusters in the analyses. Because of marked positive and negative asymmetry of the 8 foot up & go times and PWI scores, respectively, the inverse of time and the square of PWI scores were used as the dependent variables. Analyses were carried out with Stata 12 (Stata Statistical Software, 2011).

Results

A total of 922 individuals age 65 and over participated in the ELES pilot study. Of those, 232 (25.2%) were missing data because they did not participate in the CAPI (n = 215) or their data were obtained from a proxy (n = 17). Thus, the final sample included 690 individuals. Data are missing for some variables causing the n to vary slightly across analyses (from 663 to 690).

Mean age of the sample was 74.5 years, 56.8% were women, 52.3% had attended less than primary school, and 60.2% were married (Table 1). Optimal self-perceived health was reported by 53.3% of the sample. Mean score for the PWI was 7.5 and the mean 8 foot up & go time was 9.9 s. Compared to the general population of older adults in Spain (Instituto Nacional de Estadística, 2016c; Instituto Nacional de Estadística, 2016d), the sample was younger, more highly educated, and reported better health.

There was a clear gradient in the prevalence of healthy aging by definition (Figure 1A). The measured prevalence increased as criteria were relaxed, rising from 4.5% for Rowe and Kahn’s definition to 49.2% for Level III healthy aging. As displayed, the Spaniards in this sample fared better than the American sample for all definitions, with observed differences statistically significant for three of the four definitions as evidenced by nonoverlapping confidence intervals. Regardless of the definition used, healthy aging was more prevalent in the younger age group (65 to 79 years) and among men (Figure 1B). Across definitions, estimates varied considerably by age and gender, with no woman aged 80 and over meeting the Rowe and Kahn definition and roughly three-quarters of men age 65 to 79 meeting Level III criteria.

The percentage of adults meeting the components of healthy aging is presented in Table 2. Results from the HRS 2006 survey are presented for reference. The least discriminant condition was freedom from severe cognitive impairment, with more than 90% of adults in all age and gender groups free of severe cognitive impairment. The most demanding component was freedom from disease and cardiovascular risk factors. For all age and gender groups, less than 15% of the Spanish cohort met this criterion. Excluding high cognitive functioning, the percentage of adults meeting each component was greater among men and those aged 65 to 79 years. Except for cognitive functioning, a higher percentage of Spanish than American adults met each component of healthy aging, most notably with respect to physical function, absence of disease, and limiting disease.

Sociodemographic variables associated with healthy aging are presented in Table 3. In bivariate models, age, gender, education, and marital status were significantly associated with Level I, Level II, and Level III healthy aging, with male gender and education beyond primary school increasing the odds of healthy aging and older age and widowhood associated with reduced odds of healthy aging. Patterns were less consistent for Rowe and Kahn’s definition, with education and marital status failing to reach statistical significance. In all multivariate models, age and gender remained significantly associated with healthy aging. Marital status was not statistically significant in multivariate models and education remained significant only in the Level II and Level III models. Moreover, in the case of Level III healthy aging, an interaction with age and education was evident. Specifically, an increasing effect of education was observed with each year under the age of 70 and a decreasing effect was observed for each year over the age of 70 (age was centered at 70).

To evaluate the validity of the various definitions of healthy aging, 8 foot up & go times, quality of life scores, and self-perceived health were regressed on the continuum of healthy aging. Figure 2 contains expected probabilities and means for each outcome for both ends of the educational spectrum because education was significant in the final models. Among those with less than a primary school education, the predicted mean time to complete the 8 foot up & go test was similar for those meeting Rowe and Kahn’s definition, Level I criteria, and Level II criteria (around 7.1 s; Figure 2A). Although there was an increase in predicted times for those meeting Level III criteria (7.54 s) and those not experiencing healthy aging (8.83 s), only the time for the last category was statistically different from the rest. The pattern was similar among those with high education.

A relatively linear decrease in quality of life scores was evident as the definition of healthy aging was relaxed (Figure 2B). Among those with less than a primary school education, scores declined from 8.0 for those meeting Level I criteria and Rowe and Kahn’s definition to 7.2 among those classified as not experiencing healthy aging. Group differences were statistically significant, however, only for those classified as not experiencing healthy aging in comparison to all other groups and those meeting Level III criteria relative to those meeting Level I and Rowe and Kahn’s definitions. The pattern was similar for people with a university degree.

Among those with a university degree, the predicted probability of optimal self-perceived health was highest among those meeting Rowe and Kahn’s definition (92.3%), followed by those meeting Level I criteria (85.1%) and Level II criteria (86.7%; Figure 2C). Although no significant differences in the predicted probability of optimal health were evident for these subgroups, all three had a significantly higher probability of optimal self-perceived health than those meeting Level III criteria (69.4%) and those not experiencing healthy aging (41.9%). Moreover, those not experiencing healthy aging had a significantly lower predicted probability of optimal self-perceived health than those meeting Level III criteria. A similar pattern was evident among those with less than a primary school education.

People with lower education had a lower prevalence of optimal self-perceived health, lower quality of life, and slower physical performance scores at each healthy aging level. Gender was not significant in any model and age for physical performance only. All three models were statistically significant (p < .001).

Discussion

As reported elsewhere (Depp & Jeste, 2006; Fernández-Ballesteros García, et al., 2011; McLaughlin et al., 2012), we found that the prevalence of healthy aging was dependent on the criteria used to define it. In this study, it ranged from 4.5% for the most demanding definition to 49.2% for the least demanding one. Results suggest that freedom from disease and risk factors for disease are key reasons for not meeting the more rigorous definitions.

Consistent with existing research, differences in the prevalence of healthy aging were observed by age and gender, with advanced age (Depp & Jeste, 2006; Hank, 2011; McLaughlin et al., 2010) and female gender (Hank, 2011; McLaughlin et al., 2010) associated with lower odds of healthy aging. Differences by gender were most notable in the youngest age group. Observed gender differences are likely due to the widely observed finding that women have a higher prevalence of disabling conditions such as arthritis and depression and poorer physical function (Crimmins, Kim, & Solé-Auró, 2011; Newman & Brach, 2001). As articulated by Bird and Rieker (1999), the varying health experiences of men and women likely stem from biological factors (e.g., gender differences in body composition; Newman & Brach, 2001) as well as social forces that shape the roles, economic resources and behaviors of men and women (Bird & Rieker, 1999; Spitzer, 2005).

Although educational differences in healthy aging have been reported (e.g., Hank, 2011; McLaughlin et al., 2012), we found an association only for the least demanding definitions. Given that adults with lower levels of education are underrepresented in this sample relative to the broader population of older Spanish adults (Table 1), conclusions about the effects of education are tentative. With that caveat, the educational patterns observed in this study suggest that a higher level of education was less influential in preventing the onset of any disease (i.e., no educational differences were evident for Rowe and Kahn’s definition and Level I healthy aging) than it was in preventing limiting disease and functional health problems (i.e., educational differences were evident for Level II and Level III healthy aging). One possible explanation is that lower levels of education may contribute to suboptimal management and treatment of chronic disease (Brown et al., 2004), increasing the risk of limiting disease and functional health problems. Notably, no protective effect of education was observed among those aged 85 and over for the least rigorous definition (i.e., Level III, which incorporates only functional health). It could be that once one reaches the later stages of older adulthood, the health-related advantages of education are not enough to overcome age-associated biological changes that heighten the risk of functional health problems (e.g., loss of muscle mass; Taffet, 2014). It could also be that those individuals with lower levels of education who reached the age of 85 were a healthier subset of lower educated individuals. Many in this cohort fought in the Spanish civil war (1936–1939) and endured its aftermath. Thus, those reaching the age of 85 with a low level of education may be especially resilient. Alternatively, this could simply be a chance finding, as the interaction of education and age was evident for only one definition and was without an a priori hypothesis.

Using similar definitions, we found a higher prevalence of healthy aging in our sample than in the American cohort examined by McLaughlin et al. (2012). There are several possible explanations. First, there are demographic differences between the two samples. Specifically, the Spanish cohort has a smaller proportion of people age 85 and over and is less racially diverse. Second, there could be cultural differences in reporting difficulty performing physical functions and activities of daily living (Jürges, 2007). Third, diseases could be underdiagnosed in Spain. Because medical consultation in the Spanish Health System is free at the point of delivery, however, this seems unlikely (García-Armesto, Abadía-Taira, Durán, Hernández-Quevedo, & Bernal-Delgado, 2010). Fourth, differences may reflect selective non-response among Spaniards, as prior research (Rodríguez Laso et al., 2013) and Spanish census data suggest that ELES pilot study participants are a healthier subset of older Spaniards. Finally, the observed differences could be real. Spaniards had a somewhat higher life expectancy at age 60 (World Health Organization, 2016) and lower disability-adjusted life year rate (World Health Organization, 2014) in 2012 than Americans. While not investigated in this study, the observed differences could reflect differences in lifestyle factors such as obesity, which is higher in the United States than in Spain (OECD, 2014) as well as life course differences in access to health care. Although the United States has recently taken steps to increase access to healthcare among its citizens (see description of the Affordable Care Act at https://www.hhs.gov/healthcare/about-the-law/read-the-law/), Spain has offered universal health care to its citizens for decades. Comparisons of our prevalence figures with those of other countries are difficult owing to varying definitions of healthy ageing (Hank, 2011). To our knowledge, Spain is the first country outside the United States where McLaughlin and colleagues’ operational definitions of healthy aging have been examined.

A key objective of this analysis was to inform the development of a definition of healthy aging in Spain. To that end, we investigated the relationship between varying operational definitions of healthy aging and physical performance, quality of life, and self-perceived health. Notably, we observed little difference in physical performance across Spaniards meeting Rowe and Kahn’s definition, Level I criteria and Level II criteria. Consistent with the findings of McLaughlin et al. (2012), this suggests that more stringent definitions of healthy aging (Rowe and Kahn’s and Level I healthy aging) offer little advantage over a more moderate definition (Level II healthy aging).

As might be expected, patterns were less definitive for subjective measures of health. Quality of life scores generally increased with the stringency of the definition. Incremental differences were small, however, and subgroup differences were not statistically significant for those classified as experiencing healthy aging according to Rowe and Kahn’s definition and the criteria for Level I and Level II healthy aging. A somewhat clearer demarcation was evident for self-perceived health, with the percentage reporting optimal health notably higher among those meeting Rowe and Kahn’s definition and the Level I and Level II definitions than for the other subgroups. Taken as a whole, these findings coincide with McLaughlin et al.’s (2012) suggestion that a definition of healthy aging that incorporates limiting disease and functional health is likely sufficient to differentiate those with and without significant health problems. More demanding definitions appear to offer little advantage when compared with an objective measure of physical health and self-rated health. Given that those with non-limiting chronic illness seem to have fared as well as those without chronic illness, our findings also suggest that programs and policies that seek to minimize the effects of existing illness are just as vital as those directed at the primary prevention of illness.

There are noteworthy study limitations. First, the only objective indicator of health with which we had to compare definitions was the 8 foot up & go test. A similar test of physical performance, however, has been shown to predict mortality (De Buyser et al., 2013). Second, given the use of secondary data, we were limited to a set of existing measures. Third, the sample represents a healthier segment of the Spanish population and, therefore, may overestimate the prevalence of healthy aging in Spain. Fourth, the cognitive criterion for our Level I, Level II, and Level III definitions was freedom from severe cognitive impairment. Thus, individuals with mild cognitive impairment may have been classified as experiencing healthy aging. Future research efforts might explore the impact of varying MMSE cut-points on the prevalence and validity of the definitions examined in this study. Fifth, this is a cross-sectional investigation. To better understand the causal nature of observed associations, longitudinal research is needed.

European nations are implementing a range of policy and programmatic strategies to promote healthy aging (see July 2014 issue of The European Files at http://www.europeanfiles.eu/), including Spain’s aforementioned Strategy for Health Promotion and Disease Prevention in the National Health System. Measuring and monitoring healthy aging across nations can foster a shared understanding of the types of initiatives that may or may not be effective in promoting healthy aging (National Research Council, 2001; World Health Organization Regional Office for Europe, 2012). As articulated by Goodnow (2008), there is much to be gained by cross-national research efforts, including advancing understanding of “what might be changed, improved, or avoided in one’s own country” (p. 58). We encourage scholars in other nations with data sources harmonized with the HRS to investigate the validity of the proposed “Level II” definition of healthy aging. A shared definition will open new avenues for international comparative research on healthy aging and policy initiatives designed to promote it.

Funding

This work was supported by the Spanish Ministry of Science and Innovation (National R&D&I Plan: ref. CSO2011-30210-C02-01). Funders of the Aging in Spain Longitudinal Study Pilot Survey can be found on the webpage proyectoeles.es.

References