[a] The CDC does feature “age disparities” more prominently in its health disparity research portfolio; however, the issue is seemingly conceived primarily as a dichotomous one, with most attention given to youth obesity prevention and issues with aged individuals (“the elderly”).
Overall, looking at the articles chosen for this paper, few authors explained in-depth how decisions were made regarding the analytical representation of age. Authors’ attempts at conceptualizing age can be divided into three different categories: 1) using a “proxy construct” based on government programs and associated data sets; 2) nominal (ad hoc) categorization; and 3) employing “development milestones” as a means to understand age. Legal definitions regarding eligibility criteria for government insurance served as a “proxy construct” for some authors to classify individuals. Khan et al. based their definition of “elderly” individuals on the starting age for enrollment in the Medicare insurance program (65 years old).3 In addition to providing a linked data set for analysis, government insurance also provided a conceptual cut-off for other authors to define what constituted an “adult”. Sommers et al. defined “non-elderly adults” spanning from 18 years of age up to 63 years of age, using the eligibility of Medicare insurance as the analytical bookend for the end of “adulthood.”4
Second, some authors simply used nominal categories as a means to classify individuals. Howell et al. used the nominal categories of “children” to describe individuals between the ages of 0-12 years, while “adolescents” were classified as being 12-18 years of age.5 In her review, White includes diverse and overlapping age ranges corresponding to the labels “adolescent:” 10-18 years; 19-23 years and “young adult:” 18-23 years; 19-29 years .6
Finally, some authors used a more textured approach to link the notion of age with other variables in order to employ “development milestones” as a means to understand the progression of time. Lurie coupled analysis of different age groups with differences in educational levels (0-6yrs as pre-elementary school-aged (pre-ESA); 6-12yrs as ESA; and 13-18yrs as post-ESA) in understanding expansion in State Children’s Health Insurance Program (SCHIP) coverage.7 Hong and Kim took one step further by conceptualizing age coupled with other important developmental touchstones, such as marital status and number of children, to stratify individuals of the same age into different qualitative categories for out-of-pocket coverage of health expenses.8
Measurement of Age
Most data were collected from population-level surveys yielding individual self-reports or parent reports when the focus of the interview concerned a minor under the age of 17 years. The majority of surveys identified in this review used open-ended questions to ascertain individuals’ ages. Alternately, “What is your date of birth?” or other questions requesting birth year, were used to calculate age from a set time point (i.e. Dec. 31).
Authors rarely reported age disparities related to health insurance without combining single-year ages into a group or range. While most studies described respondents’ age ranges with no rationale for classifications, the few justifications provided varied across authors. In an article exploring the effect of SCHIP expansions by children’s age, one author utilized age groups previously set by a federal mandate of income eligibility levels for public health insurance.7 In another analysis, health insurance status was examined from ages 13 through 32 years, using 2-year age intervals (i.e. 13–14, 15–16); the authors justified this classification as a means of improving the statistical precision of their estimates.9 They further defended the inclusion of the 20 year age span in their analysis as a way to encompass a wide spectrum of insurance patterns across the developmental periods of adolescence, young adulthood and beyond young adulthood.9
Within these general labels, the years corresponding to childhood, adolescence, young adulthood, and older adulthood varied considerably. Age-defining limits and subcategories for childhood identified in the literature included: 0-4 years (pre-school) vs. 5-17 years (school age)6 and pre-elementary school age (younger than 6 years) vs. elementary school age (6-12 years) vs. post-elementary school age (13-18 years).7 Most often, adolescence was defined as 10-18 years, which contains boundaries for early adolescence (10-13 years), mid-adolescence (14-16 years), and late adolescence (17-18 years). Alternate definitions for young adulthood included 18-21 years and 18-24 years. In a review of some analyses, a distinction was made between 18-21 years and 22-24 years, reflecting cut-off points for private insurance coverage based on parental insurance policies or eligibility for SCHIP.6 Additional definitions for adulthood included 18-63 or 64 years, covering the years before Medicare eligibility and also referred to as non-elderly adults4 and 19-64 years, sometimes described as working age adults.6
With respect to authors’ analytic approaches, the age domain was operationalized in a number of ways, including age as a stratifying variable with the insurance outcome displayed as a comparison across age-defining groups.5 Age was also presented as a cohort with rates of insurance coverage evaluated in depth for a specific group born within a certain time period.10 In some instances, age was presented as interval groupings within defined periods of the life course (e.g. 2-year intervals during adolescence).6,9 Age was presented as a continuous variable, bound by upper and lower limits (e.g. 18-64 year olds defined as pre-Medicare eligibility; children 0-18 years).3,4,7 In examples where age was offered as a continuous variable, it was often used as a covariate (“control variable” in the economic literature3) when evaluating other predictors of being insured vs. uninsured or assessing relationships between health insurance coverage and other health outcomes.3,4 In one study, age was uniquely transformed by a life-cycle framework to reflect contextual importance when considering age as a developmental reference (incorporating reference age in years, marital status, and the presence of children under age 18 years in the household).8 The resulting life-cycle stage was used as an independent variable in establishing factors salient in the determination of proportions of household budget given to health care expenditure.8
Within the age disparity domain, insurance coverage was presented in absolute and relative terms. Examples of absolute reference included: (1) the presentation of absolute differences in insurance coverage by age groupings at one point in time; and (2) the report of change in absolute differences over time, sometimes defined as gaps in coverage for particular age groups and within subgroups (e.g. among adolescents, differences in rates based on race, ethnicity or income).4 Relative differences among age groups were also portrayed based on other predictors, such as region of residence, gender, and race and ethnic classifications.4,9
Descriptive analysis appeared to be the dominant analytic technique used when evaluating insurance coverage with relationship to age. Trend data documented changes over time for concepts such as overall rates of coverage and proportions of individuals reporting episodes of non-coverage within age groups.4,10 When looking across specific age groupings, summary analyses reported on overall patterns such as a U-shaped pattern identified for coverage or an inverted U-shape characterizing lack of insurance coverage.9
Comparisons for insurance coverage by the age domain generally followed four patterns: (1) young adults vs. children; (2) young adults vs. older age adults (30 years and older); (3) children vs. young adults vs. older age adults (30 years and older); and (4) no comparison group, but reporting on percentages of coverage for a specific age group (e.g. the span of young adulthood, 19-29 years) with the goal of looking at the insurance outcome as a function of critical transition points or looking at overall trends for insurance retention.4
Findings for the insurance outcome within the age domain were most often presented as comparison percentages or percent differences, such as comparing age groups with the highest proportion versus the lowest proportion of full-year or partial-year coverage. Additional analytic schema used to present findings included odds ratios identifying predictors of coverage vs. non-coverage within specific age groupings;9 Kaplan-Meier survival curves for continuous Medicaid enrollment;4 Cox proportional hazard ratios for Medicaid disenrollment;4 crude and adjusted relative risk for use of preventive services based on insurance status where age is used as a discrete variable among a grouping of 51-61 year olds;10 linear regression coefficients revealing life-cycle stages and other characteristics determining household budget share allocated to health care expenditure;8 and weighted averages of health insurance coverage by age groupings, providing difference-in-difference estimates pre and post SCHIP expansion.7
Data sources identified in the literature included the National Health Interview Survey;9 the Medical Expenditure Panel Survey;4 the Health and Retirement Survey;10 the Survey of Income and Program Participation;7 the Medicare Current Beneficiary Survey;3 Medicaid;5 the Civilian Health and Medical Program of the Uniformed Services;5 and the Consumer Expenditure Survey.8 The analyses performed were secondary analyses using data publicly available or available through primary sources that required special agreement or registration (e.g. Health and Retirement Survey). Some data sets which were administered at more than one time point allowed for longitudinal analyses.3,4,8,10 In addition to being able to evaluate data longitudinally, the data sets provided large sample sizes which facilitate stratification in the event of narrowly defined age groupings9 and weighting procedures4,7,9,10 which permit some comment on a nationally representative focus.
Discussion and Conclusion
This paper examined how age is conceptualized, measured and analyzed in the literature in relationship to health insurance coverage. There are likely a number of reasons underpinning the difference in how age is conceptualized as a domain vis-à-vis other domains. First, age is inherently quantifiable and easily standardized across populations without the need of conversion; that is, although individuals may differ across a spectrum of possible characteristics, a man who is 35 years old can be standardized across populations with another man who is 35 years old with little conceptual confusion, although the comparison may not result in a comprehensive understanding of underlying differences. Second, conceptualization of age in health science studies is generally predicated upon accumulated insights from medical epidemiology; that is, as an individual gets older, there is an increase in physical susceptibility[b] to both morbidity and mortality from certain diseases. This accepted consensus regarding the function of an increase in age and imputed direction of causality is not easily found in other domains.
Literature reviewed in this analysis revealed little explanation by authors for their choice in how age is measured or represented analytically. Perhaps the paucity of explanation regarding age is explained by researchers’ passive use of data garnered from existing surveys (secondary data) that already established categories for age, many simply imported into their studies. For example, Kahn and her co-authors did not provide an explanation for the lower bound of their definition of “elderly” as including individuals 65 years of age;3 one could certainly question what the inherent qualitative difference is between an individual who is 64.5 years of age versus 65 year of age besides the onset of Medicare coverage and a corresponding data set. Conversely, Hong and Kim offered a nuanced approach to defining age as an analytic construct with their presentation of a life-cycle framework.8 Indeed, their conceptualization of age is an intriguing one that opens up other analytical possibilities: although two individuals may be the same age, they may face different environmental pressures and have different living conditions that may account for differences in health besides merely a quantitative representation.
The data sets represented in the analyses present some challenges. Limitations of the data sets include bias and potential inaccuracies introduced with self-report data; although with specific reference to the age domain, there was no comment from any of the references about concerns regarding the validity of information. Studies employing a cross-sectional design are only able to make provisional comment on insurance status within age domains.9 Given the requirements defining the boundaries of confidentiality and pressures arising from political discourse, there is an inability to break-down much of the data to examine state or local level questions. Selection effects may limit data (i.e. who participates in surveys; who remains for follow-up in longitudinal design); this is explicitly noted in one reference.3 One study makes comment of limitations in generalizability of findings, given their focus on a specific age group, individuals aged 51-61 years.10
Using age as an analytic construct has many formidable strengths including quantification and standardization across diverse populations, which may not be feasible in the case of other domains, and a paucity of missing data. However as noted above, the analytical weaknesses of using age flow directly from the putative strengths; for while it may be useful to compare health disparities among individuals 35-40 years old, an individual’s age group provides limited information regarding potential underlying differences between individuals that may lead to disparities. Indeed, because all individuals of the same age group are not homogenous and likely face different risks from genetic and environmental factors, some scholars attempt to provide greater insight on disparities by combining the stratification of age with other variables, including race, gender, and income group.4,9 However, even this stratification approach lacks the ability to isolate key developmental factors which underlie the inherent value of age.
The life-cycle analytical framework provides a unique way to conceptualize a deeper understanding of age within a contextual process that confers meaning beyond the numerical value. Coupling age with these important characteristics not only facilitates better insight into how individuals in the same age grouping may differ, but also may engender a more comprehensive understanding of how other contextual factors interact with each other to produce health disparities. The life-cycle construct exhibits similarities with sociological constructs for latent life pathways. Both frameworks delineate the importance of timing and sequencing of transitions (e.g. leaving home, finishing school) and the acquisition of adult roles (e.g. marriage and parenthood) as influential measures of health. The frameworks are situated within a larger construct recognizing the importance of life course study that allows for recognition of overarching social and historical impact on opportunities and exposures producing health. Given the greater depth offered in understanding underlying mechanisms for health outcomes, we suggest future research be directed to the incorporation of life-cycle models in the study of health disparities. Inclusion of a life-cycle framework will likely present significant challenge, as it will necessitate reconciliation with current large data sets whose variables of record are determined by bureaucratic interests tied to specific funding structures. As the US population embarks upon a historical shift in demographic structures, the inclusion of life-cycle constructs in analytic frameworks may lead to substantial changes in the labeling and understanding of different life stages as they relate to health outcomes. The salience of incorporating context is also evident given that the conceptualization of age will likely need to be revised as health care reform establishes a new system of coverage that ultimately replaces the current piecemeal model that has produced to the “U” and “inverted U” shapes for insurance outcomes described in this paper. In sum, the use of contextual constructs that incorporate age as well as overarching social structural changes occurring within the US likely will be the most instrumental in further enriching current applications of the age domain in future research analyses.
Pober D, Freedson P, Kline G, McInnis K, Rippe J. Relationship of age to selected fitness and health related measures in healthy adults ages 40 to 79 Years. Clinical Exercise Physiology. 2002;4(2):108.
Guendelman S, Wyn R, Yi-Wen T. Children of working poor families in California: The effects of insurance status on access and utilization of primary health care. Journal of Health & Social Policy. 2002;14(4):1.
Khan N, Kaestner R, Lin S-J. Effect of prescription drug coverage on health of the elderly. Health Services Research. 2008;43(5):1576-1596.
Sommers BD. Loss of health insurance among non-elderly adults in Medicaid. Journal of General Internal Medicine. 2009;24(1): 1-7.
Howell EM, Buck JA, Teich JL. Mental health benefits under SCHIP. Health Affairs. 2000;19(6):291-297.
White PH. Access to health care: Health insurance considerations for young adults with special health care needs/disabilities. Pediatrics. 2002;110(6): 1328-1335.
Lurie IZ. Differential effect of the State Children’s Health Insurance Program expansions by children’s age. Health Services Research. 2009;44(5):1504-1520.
Hong G-S, Kim SY. Out-of-pocket health care expenditure patterns and financial burden across the life cycle stages. The Journal of Consumer Affairs. 2000;34(2):291-313.
Adams SH, Newacheck PW, Park MJ, Brindis CD, Irwin CE Jr. Health insurance across vulnerable ages: Patterns and disparities from adolescence to the early 30s. Pediatrics. 2007;119:e1033-e1039.
Sudano JJ, Baker DW. Intermittent lack of health insurance coverage and use of preventive services. American Journal of Public Health. 2003;93(1):130-137.
[a] An examined look at the NIH health disparities portal (http://ncmhd.nih.gov/) and the CDC health disparities portal (http://www.cdc.gov/omhd/Topic/healthdisparities.html).
[b] This proposition is true for physical disease but not necessarily for mental disease. For example, individuals have a higher susceptibility to mental disease in their adolescence years than in their 20s and 30s with the second modal peak thought to occur near middle-age.