Estimating health-related quality of life in populations through cross-sectional surveys

Background: Health-related quality of life (HRQL) is optimally reported from the individual vantage point; consequently, prevalence estimates from the National Health Interview Survey (NHIS) could be misleading, because approximately one third of the information is proxy-provided. Objective: The objective of this study was to discern the degree to which disparities in the crude prevalence of reduced HRQL is explainable by case-mix differences between proxy- and self-reportees. Subjects: We studied the cross-sectional data on 96,091 persons from the 1994-1995 NHIS Supplement on Disability. Research Design: We conducted a study of reduced HRQL expressed as a respondent type indicator (proxy vs. self) adjusting for socioeconomic differences and limitations in function. Measures: We studied reduced HRQL indicated by 1) perception of poor health, 2) perception of disability, 3) 30 or more bed days in the last year, or 4) 20 or more doctor visits in the last year. Results: The crude estimated prevalence of reduced HRQL for proxy reports was approximately half that of self-reports. Although the unadjusted odds ratios (ORs) for proxy- compared with self-reportees ranged from 0.51 (95% confidence interval [CI], 0.47-0.55) to 0.59 (95% CI, 0.55-0.64), case-mix adjustment resulted in a significant directional change for poor health perception, and all adjusted ORs were substantially closer to 1.0 than their unadjusted counterparts. Conclusion: Adjustment for case-mix explained away most of the proxy-/self-report disparity, suggesting that a major component of differences in prevalence between proxy- and self-respondents is case-mix-related. Consequently, compared with excluding proxy reports, inclusion of proxy reports results in relatively less bias when the NHIS-D is applied to obtain prevalence estimates for the indicators studied.