Total Health Care Expenditures. The NMES includes detailed, corroborated data on 1987 health care expenditures. Total annual medical care expenditures for individuals were examined.
Statistical Analyses
Because of the complex survey design of the NMES, analyses were conducted with the statistical package SUDAAN.40 SUDAAN uses the method of Taylor series linearization to produce appropriate standard errors in surveys involving cluster sampling. Weights provided on the public-use tapes were used to adjust for survey oversampling and nonresponse. The results reported provide national estimates of frequency distributions and means.The relationship between family care and total health care expenditures was examined by a 2-step approach. First, for each subject the presence of any expenditures during the survey period was determined, thus allowing comparison of the proportion of spenders with nonspenders. Second, among spenders, we assessed total health care expenditures by family care and individual care category. Univariate analyses provided national estimates of the proportions of spenders and nonspenders and, among spenders, total annual medical expenditures. Analyses that adjusted for other characteristics included multiple logistic regression to assess the relationship between family care status and spending or nonspending, and multiple linear regression to assess the relationship between family care status and total annual medical expenditures. Log transformation of the outcome (total medical expenditures) was performed to normalize the skewed distribution of expenditures. The method of Duan and colleagues1 was used to retransform the logarithm-based parameter estimates into dollars. Because each covariate chosen for these analyses could potentially contribute to confounding, fully saturated multivariate models are presented.
Results
Baseline characteristics of the sample stratified by family care category are shown in Table 1. This table also provides national population estimates for each covariate.
The mean age was slightly older for family care adults (35 years) than individual care adults (34 years). Similarly, the mean age was slightly, but not significantly, older for family care children (10 years) than individual care children (9 years). Education was lower for family care adults (20% with less than high school education) than individual care adults (11%), although income levels were similar across groups. Family care adults were more likely to be uninsured (17%) than individual care adults (11%). Similarly, more family care children were uninsured (17%) than individual care children (14%), although this finding was not statistically significant. Family care adults were more likely to be women (56%) than individual care adults (51%), and family care adults were more likely to be single parents (23%) than individual care adults (14%). Race/ethnicity was similar across groups. Rural residence occurred more frequently in family care (39%) than individual care (24%) families. Fewer family care families lived in the northeastern United States (12%) than individual care families (23%). Self-reported health status was similar across groups for adults, although fair or poor health status occurred less often in family care (5%) than individual care (8%) children. The mean number of unhealthy behaviors was slightly greater for family care (2.1) than individual care (1.9) adults, and current smoking occurred more frequently in family care (32%) than individual care (25%) adults. Mean medical skepticism scores were similar across groups.
Nonspenders
After adjustment for covariates, the association between family care (as opposed to individual care) and likelihood of having any expenditures did not differ significantly for children (adjusted odds ratio [AOR] for having no expenditures = 1.2; 95% confidence interval [CI], .8 - 2.2) or adults (AOR for having no expenditures = 1.3; 95% CI, .9 - 1.9). Family-level analyses showed that less than 1% of families had no health care expenditures, and the proportion of families without expenditures did not significantly differ between family care and individual care families.
Health Care Expenditures
For children, after the exclusion of nonspenders, there were similar unadjusted annual median total health care expenditures for subjects with family care ($192) and individual care ($195). However, for adults, unadjusted annual median expenditures were lower for people with family care ($343) than for those with individual care ($383).
Table 2 and Table 3 present the relationship between family care and total health care expenditures after simultaneous adjustment for each covariate. For children, the association between family care, in contrast to individual care, was not statistically significant (b = -.10; 95% CI, -.45 to .07). After multivariate adjustment, measures significantly associated with lower expenditures for children included low income and the lack of health insurance, while good and fair or poor health status were associated with greater expenditures Table 2. Retransforming the results for children provided a point estimate that family care was associated with 9% ($19) lower expenditures. Retransforming other significant covariates revealed reductions of 43% ($84) for lacking insurance, 40% ($77) for income less than 200% of the poverty level, and 23% ($44) for income between 200% and 400% of the poverty level. There were increases in expenditures of 3% ($7) per additional year of age and an increase of 52% ($103) for good and 236% ($458) for fair or poor health compared with excellent health status.