Chronic Medical Conditions and the Sex-based Disparity in Disability: The Cardiovascular Health Study

Abstract

Background.

Older women experience disability more commonly than their male peers. This disparity may be due, in part, to sex-based differences in the prevalence or the disabling effects of common medical conditions. The objectives of this analysis were to (a) quantify the extent to which excess disability in women is explained by higher prevalence of selected medical conditions and (b) evaluate whether the same conditions have differing effects on disability in men and women.

Methods.

We analyzed cross-sectional data from 5,888 community-dwelling older men and women. Disability was defined as difficulty with greater than or equal to one activity of daily living. Thirteen medical conditions were assessed by self-report, testing, or record review.

Results.

Controlling for age, race, education, and marital status, women were more likely to experience disability (odds ratio = 1.70, 95% confidence interval = 1.36–2.11). Higher prevalence of arthritis and obesity in women explained 30.2% and 12.9%, respectively, of the sex-based difference in disability rates, whereas male prevalent diseases like vascular conditions and emphysema narrowed the disability gap. Women with arthritis, hearing problems, coronary artery disease, congestive heart failure, stroke, and claudication were more likely to exhibit disability compared with men with the same conditions (p < .001).

Conclusions.

Efforts to lessen sex-based inequality in disability should focus on reducing the prevalence of arthritis and obesity. Future generations may see greater functional disparity if rates of vascular disease and emphysema rise among women. Several conditions were more often associated with disability in women, suggesting additional sex-based differences in the disablement process.

ALTHOUGH women have better survival rates than men throughout the life span, women experience higher rates of morbidity and functional limitations (1–8). Although the female disadvantage in morbidity and self-rated health decreases with advancing age, the sex gap widens for the prevalence of disability (9). Disability, defined here as difficulty in accomplishing an activity of daily living (ADL), may be more common among older women for several reasons.

First, self-reported disability could be biased, with women more likely than men to admit difficulty with task performance. However, previous work suggests that there is high correlation between self-reported ability and observed performance on similar tasks with no significant sex differences in reporting accuracy (2,10). Second, sex-based disparity in disability likely reflects some innate differences between men and women based on genetics, hormones, and structural anatomy (3,8). These innate differences tend to give men an advantage in physical performance (11,12). Starting from a lower performance baseline, a woman may need to experience less functional decline than a man to fall below her threshold for disability, which is set by both performance ability and environment (13). Third, men and women experience different patterns of psychosocial risk factors and health conditions, which influence life span and the disablement process (8,14–16). Although it is recognized that women are disproportionately affected by disabling but nonlife-threatening medical conditions (such as arthritis) (3,17), the extent to which common conditions account for excess disability among older women has not been fully described.

The impact of disease on disability at the population level is a function of disease prevalence as well as disease severity and disabling effect. Previous analyses that explored the extent to which the sex gap in disability is explained by sex-based differences in disease prevalence yielded somewhat conflicting results, possibly because they were conducted in relatively small or restricted populations (4,6). The previous studies did not seek to estimate the individual contributions of particular chronic conditions to sex-based differences in disability rates nor did they investigate sex-based differences in the disabling effects of particular diseases other than arthritis (4).

The current analysis expands on previous work by considering sex-based differences in the prevalence and the disabling effects of 13 common medical conditions in a large, well-defined, and diverse cohort of older adults. The objectives of this analysis are to (a) quantify the degree to which differing prevalence of selected medical conditions accounts for sex-based disparity in disability and (b) evaluate whether the same conditions are associated with different rates of disability in men and women. Because our data set lacks measures of disease severity, we could not explore this potential aspect of sex-based differences in disability. Nevertheless, the analyses suggest priority disease targets for reducing disability among older women. Unlike previous studies, we also identify male prevalent conditions and quantify the degree to which these conditions lessen the observed gender gap in disability. The results extend our understanding of the relative importance of particular medical conditions as variables affecting future population trends in disability.

METHODS

Study Population

The Cardiovascular Health Study (CHS) was a population-based cohort study of persons aged 65 years or older recruited from four communities in the United States: Sacramento County, California; Allegheny County, Pennsylvania; Forsyth County, North Carolina; and Washington County, Maryland. Details of sampling and study design have been described elsewhere (18,19). Potential participants were excluded if they were institutionalized, nonambulatory at home, unable to be interviewed, receiving hospice care, receiving radiation or chemotherapy for cancer, or not expected to remain in the area for 3 years. The original cohort included 5,201 men and women enrolled between 1989 and 1990. To improve minority representation, an additional 687 African Americans were recruited between 1992 and 1993. The cross-sectional analyses presented here make use of baseline data from the original and the African American cohorts and are drawn from the updated CHS database, which has incorporated minor corrections through 2001. The study was approved by the institutional review board at each site, and this analysis was approved by the Duke Institutional Review Board.

Disability

Functional status was assessed by self-report of the following ADLs: bathing, dressing, walking around the home, getting out of a bed or chair, using the toilet, and eating (20). Consistent with previous literature, disability was defined as difficulty with or inability to perform one or more of these tasks (21–23).

Medical Conditions

Guided by clinical judgment and known associations in the medical literature (24–27), we identified 13 medical conditions in the data set, which were potential contributors to disability. The presence or absence of each condition was determined as follows. Vision and hearing problems were assessed by self-report. Arthritis, bronchitis, and emphysema were assessed by self-report of physician diagnosis. Fracture was based on self-report of a fracture in the previous year. Reported presence of claudication, stroke, congestive heart failure, and coronary heart disease (angina or myocardial infarction) was confirmed by review of medications and medical record, and the variables have been modified over the years to reflect the newest and most accurate information about the presence or absence of these conditions at baseline (21). Diabetes was defined as taking insulin or oral hypoglycemics or a measured fasting glucose >126 mg/dL. Body mass index was calculated from measured height and weight, and obesity was defined as a body mass index of 35 or greater. Cognitive function was assessed in the original cohort with a 35-point Mini-Mental State Examination (cognitive impairment defined as a score of 30 or less (28)) and in the African American cohort with a 100-point Modified Mini-Mental State Examination (cognitive impairment defined as a score of 80 or less (29)).

Other Variables

Age was calculated from date of birth and expressed in years. Current marital status, years of education, and race (black vs. non-black) were determined by self-report. For each participant, we calculated a comorbidity score as the total number of the 13 medical conditions that were present in that individual at baseline. Depression was defined as a score of 10 or greater on the 10-item version of the Center for Epidemiological Studies-Depression scale (CESD-10) (30). Because depression is a common disabling condition that may affect men and women differently, we report its sex-specific prevalence in this population. However, we did not include depression among the 13 medical conditions for two reasons. First, we wished to focus on physical rather than psychiatric contributors to disability. Second, the causative pathway between disability and depression is highly bidirectional, meaning that depression may be a cause or a result of disability, and data suggest that depression and disability are mutually reinforcing over time (31). Although bidirectional causation could exist between disability and some physical conditions (ie, obesity), epidemiology suggests that medical disease, including obesity, tends to precede disability (32,33).

Statistical Analysis

Descriptive statistics were used to characterize the cohort with respect to the 13 medical conditions and other demographic and health-related variables (SAS software version 9.2, Cary, North Carolina). Chi-squared and t tests assessed the significance of unadjusted differences in these variables in men and women.

We then performed several analyses based on a medical model of disability, whereby medical conditions generally precede disability in a causal pathway (34). Guided by this model, we refer to “effects” of disease on disability, recognizing that the true direction of cross-sectional associations cannot be determined. Our cross-sectional analyses had two objectives: (a) to estimate the extent to which differing prevalence of selected medical conditions in men and women explains sex-based disparity in disability and (b) to examine whether the disabling effects of particular medical conditions vary between men and women.

To accomplish the first objective, we sought to quantify the indirect effects of female sex on disability as mediated via each medical condition. We constructed multiple logistic regression equations in which female sex was an independent variable and the dependent variable was either a particular medical condition or disability. Indirect effects via each medical condition were estimated as the product of two regression coefficients as follows: (the coefficient for female sex when the dependent variable is a particular medical condition) × (the coefficient for that medical condition when the dependent variable is disability) in models that include sex, age, race, education, marital status, and the medical conditions (35). We performed these analyses with MPLUS software, which is especially suited to estimations that involve multiple mediators and a dichotomous dependent variable (35,36). We report indirect effects as percentages of the total effect of female sex on disability, which was estimated in a model that included age and race as well as sex. Reported in this manner, the indirect effects represent the percentage of the female excess in disability, which can be explained by uneven prevalence of the medical condition in men and women.

To accomplish the second objective, we examined whether the effects of the 13 medical conditions on disability were different in men and women. For this analysis, we modeled the rate (rather than the odds) of disability because Rothman and others have shown that, when the dependent variable is dichotomous, additive models are superior for assessing interactions between risk factors (37–39). We used Rothman's Relative Excess Risk due to Interaction statistic to test the significance of interactions between sex and each of the 13 medical conditions (40). When a significant interaction exists between sex and a medical condition, it can be said that the condition's effect on disability differs in men compared with women. We employed SAS methods to estimate the Rothman's Relative Excess Risk due to Interaction statistic based on a Poisson-distributed model (41).

RESULTS

Of the 5,888 CHS participants, 3,393 (58%) were women (Table 1). A higher proportion of women were African American, and the average age of female participants was slightly lower. Compared with men, women were less likely to be married and had fewer years of education. Women were more likely to be obese and to report arthritis, fractures, vision problems, and bronchitis. Conversely, men reported higher prevalence of claudication, stroke, coronary heart disease, congestive heart failure, diabetes, hearing problems, and emphysema. Women had higher comorbidity scores, reflecting a greater number of coexisting medical conditions, and women were more likely to endorse symptoms of depression.

Table 1.

Characteristics of the Cohort

Characteristic	All participants, N = 5,888	Women, N = 3,393	Men, N = 2,495	*p Value
Age in years	72.8 ± 5.6	72.5 ± 5.5	73.3 ± 5.8	<.001
Years of education	13.7 ± 4.8	13.4 ± 4.5	14.1 ± 5.1	<.001
Race, % black	16	17	14	.005
Married, %	66	54	83	<.001
Medical conditions, %
Arthritis	52	57	44	<.001
Obesity (BMI ≥ 35)	5	8	2	<.001
Fracture	5	6	4	<.001
Vision problems	7	9	5	<.001
Hearing problems	8	7	9	.001
Diabetes	16	15	19	<.001
Claudication	3	1	4	<.001
Stroke	4	3	6	<.001
Congestive heart failure	5	4	5	.04
Coronary heart disease	20	16	25	<.001
Bronchitis	22	27	16	<.001
Emphysema	4	3	6	<.001
Cognitive impairment	30	30	29	.54
Comorbidity variables
Comorbidity score (number of 13 medical conditions)	2.3 ± 1.6	2.4 ± 1.6	2.2 ± 1.5	.001
Depression, % (CESD-10 ≥ 10)	14	17	10	<.001
Disability
Number of limitations in ADLs	0.13 ± 0.51	0.16 ± 0.57	0.08 ± 0.4	<.001
ADL disability (limitation in any task), %	8	10	6	<.001

Notes: ADL = activity of daily living; BMI = body mass index; CESD-10 = 10-item Center for Epidemiological Studies-Depression scale.

^*p Values are based on tests of significant difference between men and women.

Women had significantly more activity limitations than men. In unadjusted logistic regression models, women's odds of disability were 83% higher than men's odds (odds ratio [OR] = 1.83, 95% confidence interval [CI] = 1.49–2.23, p value < .01). The association between female sex and disability persisted in models that adjusted for age, race, education, and marital status (OR = 1.70, 95% CI = 1.36–2.11, p value < .01) and in models adjusted for age, race, education, marital status, and the 13 medical conditions (OR = 1.54, 95% CI = 1.21–1.96, p value < .01).

Objective 1: Indirect Effects

Disabling conditions that are more common in women explain some of the female excess in disability. The magnitude of the indirect effect mediated via a given condition is determined by (a) the relationship between female sex and the condition (ie, how much more common the condition is among women) and (b) the relationship between the condition and disability (ie, how much more common disability is among people with the condition).

Table 2 summarizes the relationship between each medical condition and disability (Column 1) as well as the magnitude of the indirect effect of female sex on disability, which is mediated via that condition (Column 2). With the exception of cognitive impairment, each medical condition was significantly associated with disability (Table 2, column 1). The following conditions were associated with at least a twofold increase in the odds of ADL disability (in descending order of strength of association with disability): obesity, arthritis, stroke, emphysema, hearing problems, and congestive heart failure. The second column of Table 2 reports the percentage of the sex-based disparity in disability, which is mediated via each medical condition. For example, the higher prevalence of arthritis among women explains 30.2% of the sex-based difference in ADL disability.

Table 2.

Association of Medical Conditions With Disability and Percent of Sex-based Disparity in Disability Mediated via Each Condition

	Association with Disability	Percent of Sex-based Disparity in Disability Mediated via Condition, %‡
Medical Conditions	Odds Ratio (95% Confidence Interval)†
Arthritis	2.67 (2.12–3.37)**	30.2**
Obesity	3.30 (2.42–4.50)**	12.9**
Fracture	1.61 (1.09–2.33)*	2.0*
Vision problems	1.95 (1.46–2.60)**	5.5**
Hearing problems	2.09 (1.55–2.81)**	−2.7**
Diabetes	1.56 (1.23–1.97)**	−7.5**
Claudication	1.82 (1.20–2.96)*	−5.9**
Stroke	2.41 (1.67–3.47)**	−6.2**
Congestive heart failure	2.09 (1.48–2.95)**	−9.8**
Coronary heart disease	1.29 (1.02–1.64)*	−9.8*
Bronchitis	1.45 (1.16–1.82)**	12.5**
Emphysema	2.18 (1.50–3.19)**	−6.7**
Cognitive impairment	1.19 (0.95–1.50)	0

Notes: †Models include age, sex, race, education, marital status, and 13 medical conditions.

^‡We report the indirect effect via each medical condition, expressed as a percentage of the total effect of female sex on disability. Negative percentages indicate that the disabling condition was less common among women and therefore diminish rather than contribute to excess disability among women.

*p < .05; **p < .01.

On the other hand, disabling conditions that are more common in men decrease the sex-based disparity in disability. The indirect effects via these conditions are represented by negative values in Table 2. The magnitude of such effects was generally modest (<10% of the total effect of female sex on disability). However, if disease prevalences in the population changed such that these conditions were equalized among men and women (and assuming that the disabling effects of each condition were the same in men and women and all other factors remained constant), we had expect that the female:male odds ratio for disability would rise from 1.83 to 2.25.

Objective 2: Interactions

Next, we considered whether the effect of the medical conditions on disability varied by sex (Table 3). When significant interactions existed, they were all in the same direction, with the disabling effect of medical conditions always greater among women. Women with arthritis, hearing problems, coronary artery disease, congestive heart failure, stroke, and claudication experienced higher rates of ADL disability compared with men with the same conditions (p < .001).

Table 3.

Comparing the Effects of Medical Conditions on the Rate of Disability in Men and Women

	Regression Coefficient
Condition	Men	Women	p Value†
Arthritis	.04**	.10**	<.001
Obesity	.17**	.17**	NS
Fracture	.06**	.06**	NS
Vision problems	.11**	.11**	NS
Hearing problems	.03	.14**	<.001
Diabetes	.05**	.05**	NS
Claudication	.05*	.21**	<.001
Stroke	.07**	.18**	<.001
Congestive heart failure	.07**	.21**	<.001
Coronary heart disease	.02	.08**	<.001
Bronchitis	.05**	.05**	NS
Emphysema	.09**	.09**	NS
Cognitive impairment	.02*	.02*	NS

Notes: All models included age, race, education, and marital status. Coefficients indicate regression effects on the rate of disability. NS = nonsignificant.

^†These p values indicate whether a sex by condition interaction is present. p values >.05 are NS, indicating that the medical condition is associated with a similar disability rate in men and women.

**p < .01; * p < .05. These p values indicate whether a sex-specific coefficient is significantly different from zero.

The significant interactions remained significant (p < .05) after a Hochberg adjustment for multiple testing (42). The data in Table 3 are derived from models that adjusted for age, race, education, and marital status. We also considered interactive effects in models that further adjusted for the 13 medical conditions, but those models exhibited nontrivial multicollinearity (tolerances <0.5) and thus are not presented. The associations between variables in those models were typically smaller in magnitude but were all in the same direction.

Comorbidity and Disability in Men and Women

Because the analyses described earlier suggested that the disabling effects of particular diseases tend to be greater among women, we considered comorbidity as a health-related factor that differs in men and women and may influence the disabling effects of a particular disease. As reported in Table 1, women had higher medical comorbidity scores (a simple count of the 13 conditions), meaning that on average, they had a higher number of concurrent diseases. Controlling for sex, age, race, education, and marital status, comorbidity score was associated with disability (OR = 1.80, 95% CI = 1.68–1.93). Women's higher comorbidity scores accounted for 18.0% of the sex-based disparity in disability rates.

We also considered sex-based differences in depression, which represents a psychiatric comorbidity that could influence the disabling effects of index medical conditions. In this population, the prevalence of depression was higher among women than among men (17% vs 10%, respectively). Controlling for age, race, sex, education, marital status, and the 13 medical conditions, depression was strongly associated with disability (OR = 2.63, 95% CI = 2.09–3.29). The association between depression and disability differed by almost twofold in men and women, with a regression coefficient of .08 in men and .15 in women (p < .001 for test of significant interaction).

DISCUSSION

To our knowledge, this is the first study to quantitatively evaluate the role of specific medical conditions in sex-based functional disparity in a large community-based cohort of older adults. Not only did women experience higher prevalence of some disabling conditions, but in the presence of certain conditions, women were more likely than men to exhibit disability. Unlike earlier studies that examined changes in the association between sex and disability after controlling for multiple health-related variables at once (4,6), we estimated the degree to which individual conditions contributed to sex-based inequity in disability. Our findings suggest particular medical conditions that should receive high priority in efforts to reduce disability among older women.

The two health conditions that contributed most significantly to women's greater burden of disability were obesity and arthritis. In the case of arthritis, women were more likely than men to have arthritis, and consistent with a previous study (17), women with arthritis experienced a higher rate of disability than men with arthritis. In contrast, the strong association between obesity and disability did not differ significantly in men and women, but obesity was about four times as common in women. Efforts to reduce weight gain among young and middle-aged women may substantially reduce their risk of disability in later years, especially because obesity is a risk factor for arthritis and other conditions. Unfortunately, however, the prevalence of obesity in the United States is rising with women disproportionately affected (43). If current trends continue, our findings indicate that the obesity epidemic could sustain a wide sex-based gap in disability among older Americans.

This study suggests other condition-specific points of emphasis for reducing the burden of disability in women. Higher rates of vision problems and bronchitis among women accounted for 5.5% and 12.5%, respectively, of the sex-based difference in disability. Because these conditions are not life threatening, clinicians may sometimes assign their management a lower priority. However, it is possible that aggressive symptom management or appropriate referrals to physical therapy or low vision rehabilitation may be useful in reducing sex-based functional disparity.

The sex-based gap in disability was reduced by the fact that men were more likely than women to experience certain disabling conditions, including congestive heart failure, coronary heart disease, claudication, diabetes, stroke, and emphysema. If rates of these conditions become more comparable between men and women, we would expect to see even greater discrepancy in disability, particularly because women with congestive heart failure, coronary heart disease, claudication, and stroke experienced higher disability rates than men with these conditions. Recent gains in life expectancy have been larger for men than for women such that the sex gap in longevity is narrowing (44). Potential explanations for this phenomenon are that rates of cardiovascular disease have improved less in women (though they are decreasing in both sexes) and that other smoking-related diseases, such as emphysema, are increasing among women (45,46). Our findings highlight the fact that such trends in women's health could have troubling implications for quality as well as quantity of life.

It is notable that even after adjustment for many covariables, female sex remained significantly associated with disability. Certainly, additional mediators may exist, and covariables may be measured imperfectly. Alternatively, the disablement process may differ in men and women such that even if risk factors were evenly distributed between the sexes, disability rates would remain higher among women (9). In keeping with this notion, we observed that the cross-sectional association between several medical conditions and disability differed in men and in women, and in each case, the condition was associated with greater disability among women. Although the cause of the associations cannot be determined from this analysis, the finding is consistent with previous work, which has suggested that women are more likely than men to develop disability when faced with similar medical conditions (47,48).

This observation has several possible explanations. First, women's tendency to have lower baseline physical performance than men may place them closer to a threshold for disability before experiencing any disease or age-related loss of function. Second, women may experience greater severity of disease, perhaps owing to sex-related disparities in health care, such as limited access or delayed referral (49). Finally, other factors more prevalent in women could interact with a given medical condition to elevate women's risk of functional consequences and disability. Such factors might include medical or psychiatric comorbidity (17), environmental or nutritional risks, caregiver demands, or lacking support systems (3).

Although a thorough investigation of such factors was beyond the scope of this analysis, we note that women were indeed more likely to have higher medical comorbidity scores and depression, both of which were associated with disability. Furthermore, the association between disability and depression was almost twice as strong in women compared with men. This finding suggests that interventions to address the psychological needs of persons with disability may be particularly beneficial to women, regardless of whether women with disability are more likely to become depressed or depression is more likely to lead to disability in women or both.

Several limitations should be considered in the interpretation of results. First, the majority of data about health conditions and disability were derived from self-report, and the use of a suboptimal assessment tool for cognitive impairment may explain the unexpected finding that cognitive impairment was not significantly associated with disability. Second, the data lack measures of disease severity. Third, from this cross-sectional analysis, it is not possible to determine the direction of causation in observed associations. Finally, the statistical approach treats diseases as independent stochastic events, an assumption not always met in older adults.

A main strength of this analysis is that the CHS constitutes a large diverse cohort with meticulous collection of data. Our findings provide new insight into the role of particular medical conditions in the disablement process and highlight points of emphasis for efforts to reduce sex-based functional disparity.

FUNDING

This work was supported by the National Institutes of Health (AG-023629, CHS was supported by contract numbers N01-HC-85079 through N01-HC-85086, N01-HC-35129, N01 HC-15103, N01 HC-55222, N01-HC-75150, and N01-HC-45133; grant number U01 HL080295 from the National Heart, Lung, and Blood Institute, with additional contribution from the National Institute of Neurological Disorders and Stroke; and R01 AG-15928, R01 AG-20098, and AG-027058 from the National Institute on Aging; and R01 HL-075366 from the National Heart, Lung, and Blood Institute, University of Pittsburgh Pepper Center P30-AG-024827), the John A. Hartford Foundation, Duke Pepper Center P30-AG-028716, and K23-AG-032867. A full list of principal CHS investigators and institutions can be found at http://www.chs-nhlbi.org/pi.htm.