Abstract
OBJECTIVES: To further validate an index predicting mortality in community-dwelling older adults. DESIGN: A comparison of the performance of the index in predicting mortality among new respondents to the National Health Interview Survey (NHIS, 2001-2004) with that of respondents from the original development and validation cohorts (1997-2000) and a test of its performance over extended follow-up (up to 9 years) using the original cohorts. Follow-up mortality data were available through 2006. SETTING: NHIS. PARTICIPANTS: Twenty-two thousand fifty-seven new respondents to the NHIS (2001-2004) and 24,139 respondents from the original development and validation cohorts (1997-2000). MEASUREMENTS: A risk score was calculated for each respondent based on the presence or absence of 11 factors (function, illnesses, behaviors, demographics) that make up the index. Using the Kaplan-Meier method, 5-year mortality estimates were computed for the new and original cohort respondents and 9-year mortality estimates for the original cohorts. RESULTS: New respondents were similar to original cohort respondents but were slightly more likely to be aged 85 and older, report diabetes mellitus, and have a body mass index of 25.0 kg/m² or greater. The model performed as well in the new cohort as it had in the original cohort. New respondents with risk scores of 0 to 1 had a 2% risk of 5-year mortality, whereas respondents who scored 18 or higher had a 69% risk of 5-year mortality (range 3-71% risk of 5-year mortality in the development cohort). The index also demonstrated excellent calibration and discrimination in predicting 9-year mortality (range 7% risk for scores of 0-1 to 92% risk for scores of ≥ 18, original validation cohort extended). CONCLUSION: These results further justify use of this index to estimate life expectancy in clinical decision-making.