Mortality prediction by quality-adjusted life year compatible health measures: findings in a nationally representative US sample

Abstract

Background: Responses to single-item categorical self-rated health (SRH) measures predict mortality, but performance of quality-adjusted life year (QALY) compatible health measures in this regard has not been much investigated.

Objectives: To examine mortality prediction and discrimination by 4 QALY compatible health measures, a reference single-item categorical SRH measure, and 1-year declines in those measures.

Research design: Cox survival and area under the curve (AUC) (discrimination) analyses of the 2000 to 2002 Medical Expenditures Panel Survey respondent data linked to the National Death Index through 2006, with and without adjustment for sociodemographic characteristics (age, sex, race/ethnicity, education, and income).

Subjects: A total of 22,259 respondents aged 18 to 90.

Measures: EQ-5D summary index (EQ-5D); predicted EQ-5D (pEQ-5D) derived from the SF-12; SF-6D; EQ visual analog scale (EQ VAS); and a single-item categorical SRH measure.

Results: Adjusted mortality hazard ratios for 0.1 point decrements in QALY compatible health measure scores were: EQ-5D, 1.24 [95% confidence interval (CI): 1.20, 1.29); pEQ-5D, 1.40 (95% CI: 1.34, 1.47); EQ VAS, 1.30 (95% CI: 1.26, 1.35); SF-6D, 1.37 (95% CI: 1.30, 1.43)]. In adjusted AUC analyses, baseline scores on all study health measures discriminated mortality risk, but the pEQ-5D, EQ VAS, and single-item categorical SRH measures were statistically superior in this regard. One-year declines also predicted mortality for all study health measures, but only pEQ-5D decline discriminated mortality risk in adjusted AUC analyses.

Conclusions: For all of the study health measures, baseline scores and 1-year declines in scores predicted mortality. The pEQ-5D also consistently discriminated mortality risk in adjusted AUC analyses.