|Home | About | Journals | Submit | Contact Us | Français|
Although the notion of healthy aging has gained wide acceptance in gerontology, measuring the phenomenon is challenging. Guided by a prominent conceptualization of healthy aging, we examined how shifting from a more to less stringent definition of healthy aging influences prevalence estimates, demographic patterns, and validity.
Data are from adults aged 65 years and older who participated in the Health and Retirement Study. We examined four operational definitions of healthy aging. For each, we calculated prevalence estimates and examined the odds of healthy aging by age, education, gender, and race-ethnicity in 2006. We also examined the association between healthy aging and both self-rated health and death.
Across definitions, the prevalence of healthy aging ranged from 3.3% to 35.5%. For all definitions, those classified as experiencing healthy aging had lower odds of fair or poor self-rated health and death over an 8-year period. The odds of being classified as “healthy” were lower among those of advanced age, those with less education, and women than for their corresponding counterparts across all definitions.
Moving across the conceptual continuum—from a more to less rigid definition of healthy aging—markedly increases the measured prevalence of healthy aging. Importantly, results suggest that all examined definitions identified a subgroup of older adults who had substantially lower odds of reporting fair or poor health and dying over an 8-year period, providing evidence of the validity of our definitions. Conceptualizations that emphasize symptomatic disease and functional health may be particularly useful for public health purposes.
As a recent review (1) demonstrates, there is no universally accepted definition of healthy or successful aging. Without question, however, Rowe and Kahn's concept of “successful aging” has played a major role in moving the scientific discourse on healthy aging forward. As of this writing, in fact, their seminal work on successful aging (2) has been cited in nearly 900 scientific articles (3).
According to Rowe and Kahn, an individual experiences successful aging if they are free of disease, risk factors for disease, and disability; have high physical and cognitive functioning; and are socially and productively engaged (ie, actively engaged) (4). This conceptualization has been the subject of criticism, with some questioning how attainable these criteria are for the majority of older adults (5). Indeed, McLaughlin and colleagues (6) found that only 10.9% of older adults in the United States were aging “successfully” in 2004 using a definition informed by Rowe and Kahn's conceptualization.
Although most would agree that the individual elements of Rowe and Kahn's definition are important, an overly restrictive definition of healthy aging is likely to result in the classification of many individuals with inconsequential health problems as unhealthy. This is not ideal if one wishes to differentiate between older adults in reasonably good health from those whose health places them at risk of adverse events. With the broader goal of helping to guide future research aimed at understanding the distribution and determinants of healthy aging, we investigated how sequential modifications to Rowe and Kahn's definition affect (i) the prevalence of healthy aging, (ii) demographic patterns in the phenomenon, and (iii) the validity of the concept.
Data for this study come from the Health and Retirement Study (HRS), a national study of U.S. adults aged 51 years and over (7). The study sample was selected using a complex design that involved clustering, stratification, and oversampling of blacks, Hispanics, and Florida residents (7,8). Participants complete surveys every 2 years, with members of newer cohorts of adults periodically added to the sample to help ensure its representativeness (7).
With the exception of our assessment of mortality (as explained subsequently), we used data from the 2006 wave of data collection. Although data are available for 2008, we utilized 2006 data because it is the latest year for which HRS-imputed cognitive data are available. To examine the relationship between healthy aging and mortality, we used data from participants who completed an HRS survey in 1998. For all analyses, we restricted the analytic sample to self-respondents aged 65 years and older of Hispanic, non-Hispanic black, and non-Hispanic white race-ethnicity (hereafter referred to as Hispanic, black, and white). Individuals of “other” race-ethnicity were excluded because of their small numbers (n = 148 in 1998; n = 168 in 2006). Because 1998 data were used solely to assess subsequent mortality, we also excluded 1998 participants for whom mortality status was unknown (n = 164). The final analytic sample size was 9,068 in 1998 and 9,996 in 2006.
We created four operational definitions of healthy aging (see Table 1). Guided by Rowe and Kahn's conceptualization of successful aging, each definition incorporated one or more of the following dimensions: (i) major illness, (ii) disability, (iii) physical functioning, and (iv) cognitive functioning. It is important to note that we eliminated Rowe and Kahn's active engagement criterion from our definitions. Although social and productive engagement have been linked to positive health outcomes (see, eg, 9,10), we felt that it was important to separate active engagement from our definitions. Failure to do so makes it impossible to study the relationship between factors such as social connections and healthy aging and, likewise, the relationship between healthy aging and productive engagement. Other researchers have taken a similar approach (11).
For the definition of healthy aging that we refer to as Rowe and Kahn’s, respondents met the illness criterion if they (i) had never been told by a doctor that they have arthritis, cancer, chronic lung disease, diabetes, heart disease, psychiatric illness, or had a stroke and (ii) were free of two physiological risk factors for disease—self-reported hypertension and obesity (BMI ≥ 30 based on self-reported height and weight). For the definition that we refer to as Level I Healthy Aging, the illness criterion was the same except that individuals need not be free of physiological risk factors for disease.
Recognizing that the mere presence of disease does not mean that an individual experiences any sequelae of consequence, the major illness criterion for our next definition (ie, Level II Healthy Aging) was further relaxed to include only symptomatic disease. In the HRS, respondents who report being told by a doctor that they have a particular condition are then asked a series of follow-up questions about that condition. We used participants’ responses to those questions to ascertain symptomatic disease status. Although the follow-up items varied by condition, classification of symptomatic disease was generally based on the use of medication (eg, “Are you currently taking any medication or other treatments for your arthritis or rheumatism?”), rehabilitation therapies (eg, “Are you receiving physical or occupational therapy because of your stroke or its complications?”), invasive treatments such as surgery (eg, “In the last two years, have you had surgery on your heart?”), and/or self-reported limitation in a respondent’s ability to engage in his or her usual activities as a result of the specified illness (eg, “Does your lung condition limit your usual activities, such as household chores or work?”). For our final definition, Level III Healthy Aging, the presence of disease was ignored altogether. In this last case, we presumed that our functional criteria were sufficient to differentiate between older adults with and without disease of significance.
For all definitions, respondents were deemed free of disability if they reported no difficulty performing each of six activities of daily living (ie, walking across a room, dressing, bathing or showering, eating, getting in or out of bed, and using the toilet) and five instrumental activities of daily living (ie, making phone calls, managing money, preparing a hot meal, shopping for groceries, and taking medications) (12).
Those who reported no greater than one difficulty (because of health problem) with any of 11 tasks—ranging from picking up a dime from a table to climbing several flights of stairs without resting (12)—were classified as having high physical functioning. This criterion was consistent across all definitions of healthy aging.
Cognitive functioning was assessed with a series of tasks patterned after the Telephone Interview for Cognitive Status (13) (see 14 for details). For Rowe and Kahn's definition, individuals were classified as having high cognitive functioning if they obtained a cognitive score at the median or higher (ie, a score of 22 out of 35). This is consistent with an earlier study of Rowe and Kahn's concept of successful aging (6). For the remaining three definitions, participants met the cognitive functioning criterion if they were free of cognitive impairment. Because “high cognitive functioning” is not necessary for independent living, we believe that freedom from cognitive impairment is of greater societal significance. Corresponding with Langa and colleagues (15), individuals who achieved a score greater than 10 were classified as being free of cognitive impairment.
Self-rated health, ranging from poor to excellent, was used to assess the concurrent validity of each definition of healthy aging. Predictive validity was assessed based on the occurrence of death between 1998 and 2006 among those who did and did not meet the criteria for each definition of healthy aging in 1998.
For the vast majority of variables utilized in this investigation, less than 5% of data were missing in either 1998 or 2006. To minimize losses due to missing data, we performed multiple imputation using IVEware (16). Values for missing cognitive data were imputed prior to release of the data (see 17).
To examine the impact of relaxing the definition of healthy aging on prevalence estimates and demographic differences in healthy aging, we calculated (i) the percentage of respondents meeting each definition and (ii) odds ratios for healthy aging by age, education, gender, and race-ethnicity in 2006. To assess concurrent validity, we compared the odds of reporting fair or poor health for those who did and did not meet each definition of healthy aging in 2006. To assess predictive validity, we examined the odds of mortality between 1998 and 2006 for those who did and did not meet each definition in 1998. To further ascertain how well each definition was able to differentiate between older adults with and without consequential health issues, we calculated sensitivity and specificity values for each definition using subsequent death as the gold standard for health status. All data were weighted to reflect the U.S. population of adults aged 65 years and older. Standard errors were adjusted for features of the complex sample design.
As shown in Table 2, the young–old, women, and married individuals comprised more than half the sample in both 1998 and 2006. More than two thirds of the sample had at least a high school diploma in both years, although the percentage was greater in 2006 (76.9%) than in 1998 (69.9%). Approximately 87% of the weighted sample in both years was white, with Hispanic and black adults comprising about 5% and 8% of the sample, respectively.
As expected, the percentage of older adults experiencing healthy aging varied by definition. The highest percentage was for Level III Healthy Aging (35.5%), which ignores the presence of chronic illness (see Table 3). The lowest was for Rowe and Kahn's definition (3.3%).
An examination of the percentage of adults meeting each component of healthy aging indicates that illness and physical functioning were limiting factors. When the illness criterion included any history of the seven aforementioned illnesses as well as risk factors for disease, just 6.2% of older adults met the illness criterion. When the criterion was limited to symptomatic disease, however, the percentage increased to 32.5%. Whereas less than half of respondents met the physical functioning criterion, nearly three quarters were free of major disability. Almost 97% of respondents were free of cognitive impairment.
Odds ratios for healthy aging for major demographic subgroups are displayed in Table 4. As compared to the young–old, the odds of healthy aging were 45%–61% lower for those aged 75–84 years and 66%–86% lower among those aged 85 years and older. Across definitions, the odds of healthy aging were 22%–47% lower for women than men. Compared to those with a college or higher education, the odds of healthy aging were 29%–47% lower for those with some college education, 33%–54% lower for those with a high school diploma, and 59%–82% lower for those with less than a high school diploma.
With the exception of our Level III definition (odds ratio = 0.79, 95% CI = 0.64–0.99), no significant differences in the odds of experiencing healthy aging were evident for Hispanic and white adults. Black adults generally had lower odds of experiencing healthy aging than white adults, but the magnitude and significance of the difference varied by definition. For Rowe and Kahn's definition, the odds of black adults experiencing healthy aging were 81% lower than for white adults. Respective figures for Levels I, II, and III Healthy Aging were 10%, 24%, and 31%.
In 2006, 29.8% of older adults reported fair or poor health. Across definitions, the odds of reporting this level of health were 86%–93% lower for those who met the criteria for healthy aging relative to those who did not (see Table 5). Approximately one third (34.9% weighted) of those who completed a survey in 1998 died between 1998 and 2006. The odds of dying over the 8-year period were 73% lower among those meeting Rowe and Kahn's definition as compared with those who did not. The difference was smaller for the remaining definitions, with the odds of dying between 60% and 64% lower for those classified as experiencing healthy aging relative to those not classified as such.
Based on the sensitivity values displayed in Table 6, Rowe and Kahn's criteria correctly classified 98% of those who died between 1998 and 2006 as unhealthy in 1998. That percentage decreased as the criteria for healthy aging were relaxed (Level I = 92%, Level II = 86%, Level III = 73%). The near-perfect performance of Rowe and Kahn's criteria in this regard came at a cost, however. Using Rowe and Kahn's criteria, only 8% of those who survived the 8-year period were classified as healthy in 1998 (specificity = .08). That percentage increased steadily as the criteria were relaxed, reaching 49% for our Level III definition.
Although the scientific literature is replete with references to healthy aging, actually measuring the phenomenon is challenging. Unless one wishes to limit the study of healthy aging to those with near-perfect health, one must find a balance between “what is possible” and the reality that as people grow older decrements in health are likely to occur. Although, for instance, some older adults remain free of chronic illness until very late in life, over 80% of older adults have at least one chronic condition (18). Thus, although it is “possible” to remain disease free into older adulthood, it is unlikely for most—especially as one reaches advanced old age. More importantly, the mere presence of any particular decrement in health may be relatively inconsequential for both the individual and larger society.
In this investigation, we examined four definitions of healthy aging—all informed by Rowe and Kahn's concept of successful aging (4). Specifically, we assessed how relaxing the operational criteria affects prevalence estimates, demographic patterns in healthy aging, and the validity of the concept. Not surprisingly, we found that the measured prevalence of healthy aging increased steadily as we relaxed the criteria, rising from 3.3% to 35.5%. To the extent that definitions of healthy aging are used to identify segments of the population that might benefit from intervention efforts, these differences are not trivial. Using the definition that we refer to as Rowe and Kahn’s, roughly 1 million older Americans were classified as experiencing healthy aging in 2006. For Level III Healthy Aging, that number increased to over 11 million.
Across definitions, the odds of healthy aging were lower among those of advanced age, those with lower levels of education, and women than for their corresponding counterparts. These findings are consistent with a broad literature documenting higher levels of disease and functional limitations among those of advanced age and those with lower levels of education (see, eg, 18–20) and a higher prevalence of functional health problems among older women (21), suggesting that our definitions are capturing known health disadvantages in the older population.
We found a complex pattern of results by race and ethnicity. Although Hispanic adults had significantly lower odds of meeting the criteria for our purely functional definition of healthy aging, no significant difference was evident for definitions that incorporated disease prevalence. This finding is consistent with work documenting as good or better chronic disease profiles but worse functional health among Hispanic older adults, with differences in the latter often attenuated when factors such as socioeconomic status are controlled (see, eg, 19). In contrast with Hispanic–white differences, black older adults generally had significantly lower odds of meeting each definition than older white adults. The magnitude of the difference varied by definition, however, with the difference most striking for the definition that incorporated obesity and hypertension. Given the higher prevalence of hypertension among older black adults (22) and obesity in older black women (23), this is not unexpected. Variation in observed racial and ethnic differences highlight, however, how findings about racial and ethnic disparities in healthy aging may be particularly sensitive to the operational definition of healthy aging employed.
Although differing definitions and varying degrees of statistical adjustment make it difficult to compare findings across other recent studies (6,24–27), healthy aging is generally reported to be lower among those of advanced age and lower socioeconomic status, which is similar to what we observed here. Gender differences tend to be less consistent, indicating that—like race and ethnicity—gender differences in healthy aging are sensitive to the measurement and analytic strategies employed. Even in this study—where we found that women had significantly lower odds of being classified as experiencing healthy aging across the conceptual continuum—the magnitude of the difference varied across definitions. The largest difference was for our purely functional definition (ie, Level III Healthy Aging), suggesting that gender differences in functional aspects of health are greater than differences in chronic disease prevalence as measured in this study.
We found that by using each definition, we were able to identify a group of older adults who were markedly less likely to report fair or poor health and to die over an 8-year period than their respective counterparts, providing evidence of the validity of all four definitions. For both self-rated health and death, the odds ratios were lowest for Rowe and Kahn's definition. This suggests that Rowe and Kahn's criteria—as operationalized herein—identified a subset of older adults with particularly low odds of adverse outcomes. It should be acknowledged, however, that the precision of the estimates for Rowe and Kahn's definition is lower than for the less stringent definitions due to the small number of older adults who meet the criteria.
Our analysis of the sensitivity and specificity of each definition in predicting mortality adds an additional dimension to our understanding of the validity of the definitions examined. Results indicate that the more stringent definitions were most sensitive (ie, those who died were readily classified as unhealthy in 1998) but least specific (ie, most of those who survived were not classified as healthy in 1998). In conjunction with the results of our examination of the odds of fair or poor self-rated health and death, these findings shed insight into how best to operationalize healthy aging. In particular, they support the use of definitions that emphasize freedom from symptomatic disease and functional limitations over those that incorporate more rigid disease criteria. The former identify a group of individuals for whom the odds of adverse outcomes are sizably diminished while, at the same time, reducing the chances of misclassifying healthy individuals as unhealthy—characteristics that are essential if one seeks to distinguish the truly healthy from the truly unhealthy. Although they would benefit from additional refinement, we believe that our Level II and Level III definitions provide a useful starting place for establishing a realistic benchmark for healthy aging.
Several limitations in our investigation should be noted. First, our study involved secondary analysis of an existing data set. As such, we worked with available measures in establishing our operational definitions of healthy aging—not all of which were ideal. A broader range of chronic conditions, for example, may have enhanced the ability of our definitions to differentiate between older adults with and without consequential health problems. Likewise, our assessment of physical functioning was based on self-report measures with a limited range of response options. A more precise means of assessing physical functioning may have improved the performance of our definitions.
Second, our assessment of symptomatic disease status was based largely, though not completely, on the use of medical treatments. Some individuals undoubtedly needed but did not receive treatment for their conditions. These individuals would have been classified as not having symptomatic disease, which may have reduced the ability of our Level II definition to differentiate between older adults with and without health problems of consequence.
Third, we examined only all-cause mortality. It is possible that the association between healthy aging and mortality varies by cause of death.
Fourth, because our primary reasons for examining demographic differences in healthy aging were to (i) determine if our definitions followed known demographic patterns in health and (ii) assess how those patterns changed as we varied the definition, we examined only unadjusted findings. We acknowledge, however, that demographic patterns in healthy aging may differ in adjusted analyses.
Lastly, our analysis included only community-dwelling adults who were able to respond to survey items independently. Nursing home residents and those requiring proxy respondents are, therefore, not represented in this analysis. Given the higher prevalence of cognitive impairment and other health problems in these segments of our older population, our estimates of the percentage of older adults meeting each healthy aging criterion and definition are undoubtedly higher than we would have observed had these subgroups been included.
In this investigation, we sought to contribute to the scientific discourse on the measurement of healthy aging. For general public health purposes, our results suggest that a definition that focuses on symptomatic disease and functional health outcomes may be more useful than one that is based on rigid disease criteria. Our Level II and Level III definitions provide a useful starting place for future work directed at capturing this phenomenon.
This work was supported by postdoctoral fellowships to S.J.M. from the National Institute on Disability and Rehabilitation Research of the U.S. Department of Education (grant number H133P050001) and the Eunice Kennedy Shriver National Institute of Child Health & Human Development (grant number T32HD007422). The Health and Retirement Study is sponsored by the National Institute of Aging (grant number NIA U01AG009740).
The content of this article is solely the responsibility of the authors and does not necessarily represent the official views of the Eunice Kennedy Shriver National Institute of Child Health & Human Development or the National Institutes of Health. The Health and Retirement Study is conducted by the University of Michigan. The following data files were used for this analysis: RAND HRS Data (version I); HRS Tracker 2006 (version 3.0); and HRS Cognition Imputations 1992–2006 (version 1.0).