|Home | About | Journals | Submit | Contact Us | Français|
We examined associations between material resources and late-life declines in health.
We used logistic regression to estimate the odds of declines in self-rated health and incident walking limitations associated with material disadvantages in a prospective panel representative of US adults aged 51 years and older (N=15 441).
Disadvantages in health care (odds ratio [OR]=1.39; 95% confidence interval [CI]=1.23, 1.58), food (OR=1.69; 95% CI=1.29, 2.22), and housing (OR=1.20; 95% CI=1.07, 1.35) were independently associated with declines in self-rated health, whereas only health care (OR=1.43; 95% CI=1.29, 1.58) and food (OR=1.64; 95% CI=1.31, 2.05) disadvantage predicted incident walking limitations. Participants experiencing multiple material disadvantages were particularly susceptible to worsening health and functional decline. These effects were sustained after we controlled for numerous covariates, including baseline health status and comorbidities. The relations between health declines and non-Hispanic Black race/ethnicity, poverty, marital status, and education were attenuated or eliminated after we controlled for material disadvantage.
Material disadvantages, which are highly policy relevant, appear related to health in ways not captured by education and poverty. Policies to improve health should address a range of basic human needs, rather than health care alone.
The past century has witnessed tremendous advances in medical care and technology, along with gains in life expectancy. Yet, these gains in life expectancy have been unequally distributed and have come to a halt for some disadvantaged groups of Americans.1 Throughout the life course, poor persons fare worse than higher-income individuals on key health indicators. The poor have lower self-rated health, a higher prevalence of chronic conditions, and higher mortality.2,3 Health disparities by race/ethnicity appear similarly entrenched.3,4
The association between socioeconomic status (SES) and health continues into old age and is evident across the income gradient.5,6 Higher SES, measured in terms of education, income, or occupational prestige, is associated with decreased mortality among persons aged 65 years and older,7 whereas lower levels of education, income, and occupation contribute to higher levels of morbidity and mortality in older individuals.5,7,8 The life-course model posits that accumulated disadvantage can contribute to health status in old age.5,9 The socioeconomic gradient in health persists in old age despite participation in Medicare, which provides nearly universal health insurance coverage.5,10 Further improvements in population health will require attention to factors in addition to health care that drive health disparities.11,12
Researchers have called for better measurement of characteristics associated with SES other than income,13,14 including direct measurement of material resources.15 Material resources, the goods and services that income leverages, have been proposed as critical factors in determining population health, and unequal distribution of these resources may contribute to health disparities.16,17 Unmet needs related to health care, food, and housing are interrelated indicators of material hardship,15 but only a few cross-sectional studies have simultaneously considered how multiple forms of material hardship may relate to health.15,18–20 Instead, previous research has considered the health effects of individual forms of material disadvantage. Inadequate health insurance is related both to lower use of appropriate health services and to poorer health outcomes.21–23 Food insecurity is related to higher rates of functional impairment among persons aged 60 years and older24 and to poorer health.25,26 Home ownership and other shared household amenities and assets are related to better self-rated health through multiple pathways, including housing conditions and neighborhood environments.27–30 In the present study, we examined simultaneously the population distribution of these 3 basic human needs—health care, food, and housing—and the later-life health consequences of material disadvantage in these domains. We anticipated that each of these material resources would contribute independently to declines in self-rated health and walking ability, even after we controlled for the effects of standard socioeconomic indicators such as education and poverty.
We examined data from the 2004 and 2006 Health and Retirement Study (HRS), which is a nationally representative panel study of Americans aged 51 years and older.31,32 These 2 HRS waves included a new cohort of participants—the early baby boomers—born from 1948 to 1953. Baby boomers account for a disproportionate share of the US population, and there are important differences in wealth and health between baby boomers and earlier cohorts.33–35 Spouses or partners of HRS participants were excluded from this analysis if they were less than 51 years of age.
We chose outcome measures to reflect midlife and older individuals’ underlying health status. Self-rated health is an important predictor of mortality across age, gender, and racial/ethnic groups.36–38 Participants reported whether their health was excellent, very good, good, fair, or poor in both 2004 and 2006. Consistent with prior research, a major decline in self-rated health was defined as either a decline from excellent, very good, or good health in 2004 to fair or poor health in 2006 or a decline from fair health in 2004 to poor health in 2006.39
Walking ability is a powerful predictor of incident disability, institutionalization, and mortality in older persons.40–42 Participants reported whether they had any difficulty walking across a room, difficulty walking 1 block, or difficulty walking several blocks in 2004 and 2006. We defined incident walking limitation as a report of new onset of difficulty in any of these 3 areas. For each walking measure, less than 2% of respondents reported that they did not do the activity; these participants were coded as having difficulty.
Our objective was to examine associations between these health measures and indicators of material disadvantage in the domains of health care, food, and housing. Health care disadvantage was assessed in 2 ways. First, we identified a group that included both individuals without any form of health insurance (uninsured) and those with a high ratio of out-of-pocket health spending to income (underinsurance). For participants with household incomes of less than 200% of the federal poverty line, underinsurance was defined as out-of-pocket expenditures exceeding 5% of household income; for higher income participants, the underinsurance threshold was 10%.23 Out-of-pocket expenditures included deductibles, copays, and any health care cost not covered by insurance. For the purposes of this study, the 4% of participants without any current health insurance were combined with the underinsured group. Second, participants were considered to have foregone medications if they reported taking less medication than was prescribed because of cost at any time during the past 2 years.22
Food disadvantage was assessed with 2 items.24,25,43 Participants were considered to have food insufficiency44 if they answered no to the question, “In the last 2 years, have you always had enough money to buy the food you need?” Participants also reported whether anyone in the household received government food stamps at any time during the past 2 years.
Housing disadvantage was assessed with 4 items. First, participants were classified as owners, renters, or other (e.g., living with family), with renting considered an indicator of material disadvantage on the basis of published reports of poorer health among renters when compared with homeowners.27,29,45 Participants also reported on housing quality; those who reported fair or poor quality housing were compared with those who reported good quality housing. Housing costing 30% or more of monthly household income was considered unaffordable.46 Participants who reported fair or poor neighborhood safety were compared with those reporting higher levels of safety.
Demographic covariates were self-reported, including age, gender, marital status (married versus unmarried), years of education, race/ ethnicity (non-Hispanic White, non-Hispanic Black, or Hispanic; other race/ethnicity [n=423] was excluded), and income. We calculated poverty status by using self-reported income and household composition.47 When poor health or other characteristics precluded survey completion, a proxy respondent (usually a spouse or other family member) completed the survey on behalf of the participant; we controlled for 2004 proxy status. In 2004, participants reported whether a doctor had ever told them they had a heart problem (including heart attack, coronary artery disease, angina, congestive heart failure, or other heart condition), cancer (excluding minor skin cancer), stroke, chronic pulmonary disease, or diabetes. Participants were classified as current or former cigarette smokers or as nonsmokers. With the use of self-reported height and weight to calculate body mass index, participants were classified on the basis of National Heart, Lung, and Blood Institute guidelines as underweight, normal weight, overweight, or obese.48
Analyses were conducted with Stata version 10 (Stata, College Station, TX) to account for the complex sample design and provide estimates representative of the noninstitutionalized US population. The 2004 HRS response rate was 87.8%. Of the 17890 age-eligible respondents in 2004, 16025 also participated in 2006 (89.6%). Between 2004 and 2006, 1000 participants (5.6%) died and 865 were lost to follow-up (4.8%). An additional 190 participants were excluded because of missing data on material resources and 394 were excluded because of missing data on covariates. Baseline information on the 15441 remaining participants was used to examine cross-sectional associations between demographic characteristics and material disadvantage (Table 1). The χ2 test was used for descriptive comparisons across groups.
We used logistic regression to predict the odds of decline in self-rated health and incident walking limitations after adjustment for covariates. By definition, a decline could not be observed in participants already in poor health in 2004. Excluding 1153 participants with poor health at baseline, as well as 20 with missing self-rated health data in 2006, yielded a final sample of 14268 for analyses of decline in self-rated health. For the analysis of incident walking limitations, we similarly excluded 821 participants who already reported difficulty walking across a room in 2004 (the worst category), as well as 11 with missing walking data in 2006, for a final analytic sample of 14609.
As illustrated in Table 1, we observed differences in the baseline distribution of material disadvantage across age, gender, racial/ethnic, and education groups in all 3 resource domains examined (health care, food, and housing). Differences between older (≥65 years) and younger (51 to 64 years) respondents were significant for all individual components in the health care and food disadvantage domains but were more modest in the housing domain. Participants under 65 years of age, most of whom were not yet eligible for Medicare, reported substantially more problems with foregone prescriptions because of cost (11.5% vs 6.6%). By contrast, uninsurance or underinsurance was more common in the older age group, which reflected a higher prevalence of underinsurance because of both higher health costs and lower incomes. Younger participants fared significantly worse than did adults 65 years and older on indicators of food disadvantage; they reported a higher occurrence of both insufficient funds to pay for necessary food and more frequent receipt of food stamps. On every individual indicator studied, women were significantly more likely than were men to be disadvantaged.
The most pronounced differences in material disadvantage occurred across racial/ethnic and educational strata. We observed consistent, substantial, and statistically significant race/ethnicity-based material resource differentials in all 3 domains, as well as for every individual indicator examined. Non-Hispanic Black and Hispanic respondents were far more likely than were non-Hispanic White respondents to report foregoing needed prescriptions because of cost and had a significantly higher prevalence of uninsurance or underinsurance. Similarly pronounced racial/ethnic differentials were observed for health care and food disadvantage. Food insufficiency and food stamp receipt were far more common for non-Hispanic Black and Hispanic respondents than they were for non-Hispanic White respondents. More than 20% of non-Hispanic Black and Hispanic respondents lived in low-quality housing compared with 7.8% of non-Hispanic White respondents. Housing was unaffordable for 23.4% of non-Hispanic Black, 26.7% of Hispanic, and 12.3% of non-Hispanic White respondents. The majority of non-Hispanic Black (60.7%) and Hispanic (53.5%) respondents reported living in unsafe neighborhoods, compared with 22.2% of non-Hispanic White respondents. Across all 3 racial/ethnic groups studied, the majority of participants reported material disadvantage in at least 1 domain. Such disadvantage was particularly common among non-Hispanic Black and Hispanic participants, who often reported multiple unmet needs.
Disparities by educational attainment also were pronounced in our study population. In the domains of health care, food, and housing, and on every individual indicator of disadvantage, respondents with less than a high school education were far more likely than were those with high school diplomas to report material deficits. Examples included the markedly higher occurrence of foregone prescriptions (13.6% vs 8.5%) and uninsurance or underinsurance (46.2% vs 27.5%), as well as large differences in the prevalence of renting rather than owning a home (21.8% vs 10.6%), living in low-quality housing (23.0% vs 7.6%), and living in an unsafe neighborhood (47.8% vs 23.4%). Food insufficiency (5.5% vs 2.0%) and food stamp receipt (13.1% vs 2.5%) were also substantially more common in respondents without a high school diploma.
As illustrated in Table 2, disadvantaged individuals were markedly more likely than were their advantaged counterparts to experience declines in self-rated health and incident walking limitations. More than 1 in 10 participants reported a decline in self-rated health (12.3%) or incident walking limitations (15.8%) between 2004 and 2006. For all individual and summary measures of health care, food, and housing disadvantage, we observed higher rates of worsening health among participants without adequate material resources.
The results of logistic regression models designed to determine the independent health effects of each demographic indicator and domain of disadvantage are reported in Table 3. The analysis of declines in self-rated health (Table 3) excluded participants who already reported poor health in 2004. In model 1, which predicted a decline in self-rated health with the use of only demographic characteristics, declines were more common among older, Black or Hispanic (compared with White), and unmarried respondents, as well as those with less than a high school education and those living in poverty.
Next, in model 2, we assessed the association between dichotomous indicators of health care, food, and housing disadvantage and decline in self-rated health. Disadvantage in each domain was associated with significantly elevated odds of decline in self-rated health, with the strongest associations observed for food disadvantage (odds ratio [OR]=2.10; 95% confidence interval [CI]=1.65, 2.68). These results were largely unchanged after we controlled for demographic characteristics (model 3). The relations between being non-Hispanic Black and being unmarried and health declines were rendered nonsignificant after we controlled for material disadvantage, and the estimated effects of poverty, Hispanic ethnicity, and education were attenuated. Even after we further controlled for baseline self-rated health, co-morbid conditions, weight status, and smoking status (model 4), health care, food, and housing disadvantage were independently associated with the odds of a decline in self-rated health. The effect was strongest for food disadvantage (OR=1.69; 95% CI=1.29, 2.22), followed by health care (OR=1.39; 95% CI=1.23, 1.58) and housing (OR=1.20; 95% CI=1.07, 1.35) disadvantage. Effect estimates for each type of disadvantage were comparable with those observed for a range of comorbid conditions, including diabetes and stroke.
Comparable results for incident walking limitations are also shown in Table 3, again showing the strongest effect for food disadvantage (OR=1.64; 95% CI=1.31, 2.05), followed by health care disadvantage (OR=1.43; 95% CI=1.29, 1.58). Housing disadvantage was not a significant predictor of walking limitations after we controlled for baseline walking limitations, comorbid conditions, weight status, and cigarette smoking status (model 4).
We used similar models that controlled for all covariates to examine the independent contributions to both outcomes of the component indicators that constitute health care, food, and housing disadvantage (results not shown). Both indicators of health care disadvantage (foregone prescriptions because of cost and uninsurance or underinsurance) were significantly associated with self-rated health declines and incident walking limitations. We observed significantly elevated odds of decline in self-rated health among food stamp recipients (OR=1.48; 95% CI=1.10, 2.00), as well as elevations approaching statistical significance among participants reporting insufficient money for food (OR=1.47; 95% CI=0.92, 2.36; P=.104). For incident walking limitations, we observed elevated point estimates for both insufficient money for food (OR=1.47; 95% CI=0.95, 2.26; P=.082) and food stamp receipt (OR=1.26; 95% CI=0.96, 1.65; P=.095), although neither was statistically significant. In the housing domain, only living in an unsafe neighborhood was significantly associated with decline in self-rated health (OR=1.17; 95% CI=1.02, 1.34), whereas both unsafe neighborhood conditions (OR=1.17; 95% CI=1.05, 1.30) and poor housing quality (OR=1.20; 95% CI=1.03, 1.41) were associated with higher odds of incident walking limitations.
The predicted probability of declines in self-rated health and incident walking limitations are provided in Figure 1. The probabilities shown are based on the number of domains disadvantaged, with control for all covariates in model 4. We observed a monotonic pattern of more frequent declines in self-rated health and walking limitations among respondents with a higher number of domains disadvantaged, such that the highest risk was observed in respondents disadvantaged in all 3 domains.
Our findings provided evidence that health is shaped by unmet needs for adequate food, housing, and health care. We observed that most Americans over 50 years of age experienced at least 1 material disadvantage in these domains. The most common problems were low neighborhood safety (27.7%) and uninsurance or underinsurance (30.9%). Consistent with previous reports on racial/ethnic and socioeconomic disparities, non-Hispanic Black and Hispanic respondents and those without a high school education were markedly more likely to report disadvantage for every indicator studied.49 For example, nearly 9% of non-Hispanic Blacks (vs 2% of non-Hispanic Whites) experienced disadvantage in all 3 domains studied, as did 7% of participants with less than a high school education (and 2% of those with high school education). These differences are both a telling reminder of the distribution of disadvantage in the United States and a likely explanation for profound and enduring health disparities. Strong relations between health declines and non-Hispanic Black race, low education, poverty, and unmarried status were attenuated or eliminated after we controlled for indicators of material disadvantage. Although not the primary focus of this article, these findings suggest that differential access to necessary material resources may be 1 pathway through which non-Hispanic Blacks and other disadvantaged populations experience poorer health outcomes.
The effects we observed were substantial and sustained even after we controlled for a range of covariates, including baseline health status and comorbidities. In general, the estimated health effects of material disadvantage were similar to associations observed between comorbid illness and declines in health. For example, our results suggest that food disadvantage is as strong a predictor of later health declines as is heart disease, cancer, stroke, pulmonary disease, or diabetes.
These results reaffirm a large body of research demonstrating the toll of inadequate health care access and underinsurance on the health of Americans.21–23,39,50 Our findings also emphasize the interconnectedness of material resources and the need for multifaceted policies to improve population health.12,51 Policy interventions to address 1 material domain may have spillover effects on other domains, as illustrated by the introduction of Medicare Part D prescription drug coverage in 2006, after the start of this study. Medicare Part D may have reduced financial constraints for older adults in a way that both improves access to needed medications and frees money to purchase other necessities, like food.52 Similar arguments could be made about how policy initiatives to address neighborhood safety and housing quality are likely to impact population health.
Both researchers and policymakers have recognized the need to coordinate efforts to ensure access to adequate material resources across domains.3,53–55 For example, the Department of Housing and Urban Development has combined the delivery of housing services with onsite health care for low-income older persons and persons with disabilities.55 Material disadvantage may be reduced substantially with appropriate planning for retirement, but this will also require improving the current levels of financial literacy among older persons in the United States.33 Efforts to improve the health of the nation may be more effective if they simultaneously address a range of basic needs instead of individual social or economic domains alone.
Material disadvantage, and policies to remediate that disadvantage, may influence health through several pathways, including direct physical effects, behavioral influences, and stress or mental health effects. For example, food disadvantage may operate through each of these pathways. Food disadvantage was strongly associated with health declines, which supports the contention that current food stamp benefit levels of approximately $1 per meal may be inadequate to maintain health.26 Nutritional compromise may lead to frailty and associated functional impairments.24,44,56 Insufficient money for food may lead individuals to choose inexpensive, unhealthy foods, leading to the paradox of overweight among adults with food insufficiency.57 Being unable to provide needed food for oneself or one’s family represents a stressor, and both acute and chronic stress have been linked to a variety of adverse physiologic responses.58,59 Finally, food disadvantage, the least common form of disadvantage in this study, may simply serve as a sensitive indicator of extreme disadvantage. Similarly, housing disadvantage may affect health through multiple pathways,60,61 with poor housing quality presenting environmental hazards (e.g., poor indoor air quality), and unsafe neighborhood conditions leading to reduced physical activity and increased stress.62
Several limitations should be taken into account in interpreting the results of this study. First, we relied on self-reported material disadvantage and health outcomes. Differential reporting of material disadvantage may bias estimated differences across demographic groups. Additionally, the validity of self-rated health as a predictor of mortality may vary across socioeconomic groups, although evidence is mixed.63–65 Second, we used longitudinal data to predict change in health accounting for baseline characteristics, but this analysis cannot prove conclusively that material disadvantage causes poor health. Other factors, including health earlier in life,66,67 self-efficacy,68 and institutional barriers (e.g., racism)69 may affect both health and late-life access to material resources. Furthermore, some indicators are related to both need for and use of services. For instance, only participants who needed prescriptions were at risk of forgoing prescriptions because of cost. Similarly, underinsurance was based on the ratio of health expenses to income; this measure may reflect a high need for health services (poor health) and inadequate health benefit arrangements. Third, the large number of comparisons conducted may increase the probability of type 1 error. Finally, our sample excluded 2449 individuals with missing data, including 1000 who died and 865 who were lost to follow-up between 2004 and 2006. Participants with missing data had a higher number of domains disadvantaged and had worse self-rated health and a higher prevalence of walking limitations at baseline. They were also older and more likely to be non-Hispanic Black or Hispanic, to have less than a high school education, to be in poverty, and to be unmarried; in addition, they also had a greater prevalence of chronic conditions. Thus, we are likely have underestimated the prevalence of material disadvantage in 2004.
Despite these limitations, this study demonstrated the importance of considering health care, food, and housing as determinants of population health and health disparities. Each of these factors contributed to declines in self-rated health and incident walking limitations—2 important indicators of future morbidity and mortality risk—in this nationally representative sample of adults over 50 years of age. Older adults with multiple forms of material disadvantage were at particularly increased risk of health decline and functional impairment. Strategies to improve population health and to reduce health disparities must address a range of basic human needs, including affordable, quality health care, food, and housing.
This study was supported by the Robert Wood Johnson Foundation Health & Society Scholars program. B.J. Soldo and J. McCabe were supported by National Institute on Aging (grant R01 AG023370 and P30 AG012836). J. A. Pagán acknowledges the support of the Department of Defense (grant W81XWH-06-1-0334), the Agency for Healthcare Research and Quality (grant R24HS017003), and the Centers for Disease Control and Prevention (grant 1H75DP001812-01).
The authors thank Sarah Sanchez for her research assistance.
ContributorsD.E. Alley and C. Cannuscio were responsible for study design, interpretation of results, and drafting of the article. B.J. Soldo, J.A. Pagan, and D.A. Asch assisted in the interpretation of results and critical revisions of the article. J. McCabe and S.H. Field assisted with data coding and analysis. M. deBlois assisted with drafting of the article.
Note. The contents are solely the responsibility of the authors and do not reflect the views of the study sponsors. The authors had no conflicts of interest to report.
Human Participant Protection
Institutional review board approval was granted by the University of Pennsylvania.
Dawn E. Alley, Department of Epidemiology and Preventive Medicine, Division of Gerontology, University of Maryland School of Medicine, Baltimore.
Beth J. Soldo, Population Aging Research Center, University of Pennsylvania, Philadelphia.
José A. Pagán, Department of Health Management and Policy, School of Public Health, University of North Texas Health Science Center, Fort Worth, and the University of Texas-Pan American, Edinburg.
John McCabe, Population Aging Research Center, University of Pennsylvania, Philadelphia.
Madeleine deBlois, Department of Society, Human Development, and Health at the Harvard School of Public Health, Boston, MA.
Samuel H. Field, Frank Porter Graham Child Development Institute, University of North Carolina, Chapel Hill.
David A. Asch, Center for Health Equity Research and Promotion, Philadelphia Veterans Affairs Medical Center and the Department of Health Care Management and Economics, University of Pennsylvania, Philadelphia.
Carolyn Cannuscio, Department of Family Medicine and Community Health and the Leonard Davis Institute for Health Economics and the Center for Public Health Initiatives, University of Pennsylvania, Philadelphia.