|
TRENDSIncreasing Obesity Rates And Disability Trends
Are older Americans becoming more or less disabled? Unhealthy body weight has increased dramatically, but other data show that disability rates have declined. We use data from the Health and Retirement Study to estimate the association between obesity and disability, and we combine these data with trend estimates of obesity rates from the Behavioral Risk Factor Surveillance Survey. If current trends in obesity continue, disability rates will increase by 1 percent per year more in the 50–69 age group than if there were no further weight gain.
Numerous studies have documented that Americans have become increasingly obese, defined as a body mass index (BMI, weight in kg divided by height in m2) of 30 or more. More than one in five U.S. adults are classified as obese based on self-reported weight, and almost one in three based on objectively measured weight, more than are current smokers (22 percent).1 The rise in obesity is particularly troubling given that obesity is associated with increased chronic physical illnesses and mortality; its negative consequences may even exceed those of smoking or problem drinking.2 Nevertheless, it is unclear whether the "obesity epidemic" has as much of an impact on population health as improvements in medical care, other changing health behavior (such as reduced smoking), education, and lifetime disease exposure. Mortality and disability rates among the elderly have been declining at a rate of 1 percent or more per year, and some researchers argue that this trend will continue.3 It appears that the effects of obesity on disability so far have been dominated by other factors. But could the ongoing "obesity epidemic" noticeably alter this conclusion? Or are the disability implications of weight gain negligible when compared with the combined effects of other social trends? This paper analyzes trends between 1985 and 2002 and extrapolates them to 2020. We only ask what would happen if trends in body mass were to continue, holding everything else constant, but do not try to predict changes in medical technology or other trends. Our predictions combine estimates about the effects of obesity on health from the Health and Retirement Study (HRS) with time series estimates of obesity based on the 1985–2002 Behavioral Risk Factor Surveillance Survey (BRFSS).4
We use data from the HRS to estimate the relationship between weight and disability/ health status. The HRS is a national panel survey of 9,825 primary respondents born during 1931–1941, who were first surveyed in 1992; follow-up surveys were conducted in 1994, 1996, 1998, and 2000. Obesity trends are estimated from the 1985–2002 BRFSS, a national cross-sectional telephone survey. The BRFSS, like the HRS, uses self-reported height and weight. Among the respondents ages 50–69 in the 2000 wave of the HRS and the 2000 BRFSS, 41 percent were overweight but not obese (BMI: 25–29.9); 26 percent in the HRS and 24 percent in the BRFSS were obese or severely obese. Dependent variables. Our main dependent variables are two measures of disability or, more precisely, functional limitations. The first measure is based on limitations with activities of daily living (ADLs) and assesses difficulty with performing any of five different tasks: bathing, eating, dressing, walking across a room, and getting in or out of bed. The second measure indicates reports of "impairment or health problem that limits the kind/amount of paid work." We use "disability" as shorthand for our measures of functional limitations. There are many other definitions of disability, which may have little overlap with our definitions, such as eligibility for disability benefits, residency in nursing homes, or dependency on caregivers. We assess health with two measures: the percentage of respondents who consider themselves in fair or poor (versus excellent, very good, good) health; and the number of chronic health problems (doctor-diagnosed eight conditions: high blood pressure, diabetes, cancer, lung disease, heart problems, stroke, psychological problems, and arthritis). Our final set of dependent variables assesses health care use. We analyze total costs (in constant 1992 dollars), number of doctor visits, and the probability of an inpatient stay during the prior two years. The specific services most closely reflecting dependence/disability are too rare in our sample to study: home health care (2 percent of the sample) and nursing home care (<0.5 percent). Explanatory variables. The primary explanatory variable is weight class: normal (BMI 18.5 or more but less than 25), overweight (BMI 25 or more but less than 30), moderately obese (BMI 30 or more but less than 35), and severely obese (BMI 35 or more). Other explanatory variables are age (dummies for age groups at 50–54, 60–64, 65–70; age group 55–59 is the reference), race (black, Hispanic, other race; white is the reference), insurance status (public, private; uninsured is the reference), marital status (married, divorced, widowed; never married is the reference), education (college, some college, high school; less than high school is the reference), census region (Northeast, Midwest, West; South is the reference), an indicator for each survey wave, and other behavioral variables, such as current daily smoking and heavy drinking (three or more drinks per day). Study methods. The first step was to estimate the relationship between weight and health-related outcomes in the HRS data. We pooled the data across all waves, generating a sample of people ages 50–69, subset to people whose weight did not fluctuate by more than 10 percent between adjacent waves and a BMI greater than 18.5.5 The estimation is cross-sectional; that is, we estimated the prevalence of functional limitations by obesity status, not the incidence of new functional limitations. We adjusted standard errors using the Huber/ White nonparameteric correction for multiple observations on the same people. There is no evidence that this cross-sectional analysis overestimates the effects of obesity. Using BMI from prior waves as a regressor on health in later waves gives very similar effects of obesity; the same is true in an incidence analysis of new health problems between 1992 and 1998. In both cases, the point estimates of obesity’s adverse effects are slightly higher than in the model we present here. However, these two approaches also reduce the age range and sample size and increase standard errors; we therefore prefer the cross-sectional estimates. We estimated probit models for dichotomous outcomes and linear models for other dependent variables. Based on specification tests for interactions between weight class and age, sex, race/ethnicity, we separately estimated models for men and women but included only main age and race/ethnicity effects (no interaction with weight class). The second step in our analysis was to use the 1985–2002 BRFSS data, adjusting for differences in state participation and demographics, to estimate a time trend for weight class and obtain adjusted rates of the prevalence of different weight classes. We use an ordered logit model with a linear time trend that can differ by age, sex, and race/ethnicity. In the final step, we combined the estimates from the HRS and BRFSS to produce predictions of the health-related outcomes among people ages 50–69, both historically (1985–2000) and extrapolated through 2020. In aggregating the rates, we held demographics constant to isolate the effect of weight status.
The associations between weight and health, disability, and health care are shown in Exhibit 1  . Severely obese older people are more than twice as likely as people of normal weight to be in fair/poor health and have about twice as many chronic medical conditions, which replicates earlier studies of the effects of obesity.6 Severe obesity is associated with 68 percent (men) or 60 percent (women) higher health care costs than normal weight; moderate obesity increases health care costs by 18 percent (men) to 31 percent (women). Averaging across these four groups (men/women and moderate/severe obesity), we found a 44 percent increase in health care costs compared with people of normal weight, a result similar to other national studies.7 The effects of obesity on work limitations are similar in size, but the effect on ADL limitations is much stronger, especially in the heaviest weight group. For men, moderate obesity is associated with a 50 percent increase in the probability of ADL limitations, and severe obesity, with a 300 percent increase. This is not an artifact of our model specification: The difference in the onset of new ADL limitations from 1992 to 1998 between normal-weight and severely obese men, classified by 1992 weight, is 13.4 percentage points (an incidence analysis), compared with the 12.6-percentage-point difference in Exhibit 1. Even larger effects exist for women: The probability of ADL limitations doubles with moderate obesity and quadruples with severe obesity. This highlights the need to distinguish moderate and more severe levels of obesity, a distinction rarely made in studies, especially since the growth rate of severe obesity between 1985 and 2000 has been twice that of moderate obesity.8
Exhibits 2 and 3 show the adjusted obesity rates historically (1985–2000) and extrapolated to 2020. The 1985–2000 numbers differ from descriptive BRFSS statistics primarily because we held population characteristics constant and weighted results to be representative for all states, not only states participating in the BRFSS.
Our main results link weight predictions with the effects on health, shown in Exhibit 4 . First consider the predicted changes between 1985 and 2000. Weight changes during these years have raised the proportion of men ages 50–69 in fair or poor health by 6.2 percent (1.3-percentage-point increase from a baseline prevalence of 21.0 percent) and by 8.5 percent for women, compared with the counterfactual scenario of no weight gain during that period. For ADL limitations, the measure most comparable to previously reported declines in disability, the increase caused by obesity was more striking during these fifteen years: 9.7 percent for men and 13.0 percent for women. The excess health care costs of obesity (over and above those for a person of normal weight) increased by 40.2 percent for men and 47.7 percent for women.
If the obesity trend were to continue through 2020, without other changes in behavior or medical technology, the proportion reporting fair or poor health would increase by 11.7 percent for men and 14.1 percent for women compared with 2000. Similar, although slightly smaller, changes would occur in chronic conditions and service use. In terms of costs, the "excess" costs of obesity would increase from 14.3 percent in 2000 to 21.4 percent in 2020 for men and from 12.7 percent to 19.9 percent for women. The largest effects of increased weight, however, would be on disability, with the prevalence of any ADL limitation increasing 17.7 percent for men and 21.8 percent for women.
Obesity is associated with large effects on limitations in ADLs and work, two variables that have not been studied in this context but that can provide a linkage to the literature on disability among the elderly. Despite all the attention obesity has received in recent years, a large body of research on disability has shown that older Americans have become healthier, and some believe that this trend continues. How can we reconcile these two areas? The findings are not necessarily contradictory. Declining disability studies analyze the effects of all societal changes, whereas we isolate a single change—namely, changing obesity rates—holding everything else constant. It is possible that regardless of how strong the adverse effects of obesity are, the other effects dominate. This is a plausible explanation when changes in obesity rates are minor, as until recently. Our study population is also slightly younger (50–69 years) than what is usually considered the elderly (65 and older), and obesity rates have not yet increased much in absolute terms among the elderly: Among Americans ages 50–69, the prevalence of severe obesity increased six percentage points between 1985 and 2000; among Americans over age seventy, it increased only three percentage points. As a consequence, the increases in functional limitations attributable to obesity among the near-elderly that we document for 1985–2000 would have been smaller for the elderly during the same period and could easily have been overwhelmed by other factors. In that case, our results anticipate the obesity consequence among the elderly by a decade. But as obesity becomes more and more prevalent among the elderly, it will be more and more difficult for other social trends to counter its adverse health effects. Unless the factors underlying past trends toward better health become even stronger, Americans who will be ages 50–69 in 2020 may not have better health and functioning than this age group has now. Study limitations. Our simple simulation study has many limitations. First, obesity has become a policy issue, and this attention may change trends, although there is no evidence of this yet. Nevertheless, there is hope that the high rates predicted for 2020 might not materialize. Also, the health and disability effects of obesity presented here are based on cross-sectional association in the 1990s, and it is unlikely that this relationship will remain constant over time. However, the biases can go either way. Improved medical care could reduce the effect of obesity on disability and health. On the other hand, the current relationship between obesity and health estimated from the HRS does not reflect the increased exposure duration to obesity over time. In other words, we do not model that a sixty-year-old in 2020 is likely to have been obese for more years than a sixty-year-old in 2000, and ignoring the increased exposure time underestimates health effects. Increased mortality associated with obesity may also reduce observed disability rates among the elderly, although the primary effects of obesity are on morbidity. Concluding comments. Our results may go against the conventional wisdom about disability among the elderly that suggest a continuing decline in disability, but this may be because of the lower age in our study population and because we used more recent data, giving us an advance view of future changes among the elderly. Two new studies seem to support our conclusion that rising obesity might reduce or even reverse the decline.9 As with smoking and problem drinking, a time lag intervenes between obesity and the development of chronic health problems. The largest declines in smoking happened between 1960 and 1980, whereas the largest increase in obesity occurred in the past decade and primarily among younger cohorts. The full long-run consequences of increased obesity rates at the aggregate level are probably not yet visible, but there are enough warning signs to suggest that past trends showing better health among the elderly may be ending—unless other societal changes or medical advances can compensate for the effects of unhealthy weight gain.
The authors thank two anonymous referees for helpful comments on an earlier draft. Roland Sturm (sturm{at}rand.org) is a senior economist at RAND in Santa Monica, California, where Jeanne Ringel is an economist and Tatiana Andreyeva, a RAND Graduate School fellow.
This article has been cited by other articles:
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Top
Data And Methods
Study Results
