Multimorbidity measures differentially predicted mortality among older Chinese adults

Abstract

Objectives: This study aimed to examine and compare the associations between different multimorbidity measures and mortality among older Chinese adults.

Study design and setting: Using the Chinese Longitudinal Healthy Longevity Survey 2002-2018, data on fourteen chronic conditions from 13,144 participants aged ≥65 years were collected. Multimorbidity measures included condition counts, multimorbidity patterns (examined by exploratory factor analysis), and multimorbidity trajectories (examined by a group-based trajectory model). Mortality risk associated with different multimorbidity measures was each analyzed using Cox regression. C-statistic, the Integrated Discrimination Improvement (IDI), and the Net Reclassification Index (NRI) were used to compare the performance of different multimorbidity measures.

Results: Participants with multimorbidity, regardless of measurements, had a higher risk of death compared with people without multimorbidity. Compared with the mortality prediction model using age and sex, C-statistics showed added discrimination (over 0.77, all P < .05) for models with multimorbidity measures. Multimorbidity trajectory showed integrated discrimination and net reclassification improvement for mortality prediction compared to condition count (IDI = 0.042, NRI = 0.033) and multimorbidity pattern (IDI = 0.041, NRI = 0.069).

Conclusion: Adding multimorbidity measures significantly improved the performance of a mortality prediction model using age and sex as predictors. Trajectory-based measures of multimorbidity performed better than count- and pattern-based measures for mortality prediction.

Keywords: Condition count; Mortality; Multimorbidity measures; Multimorbidity pattern; Multimorbidity trajectory; Older adults.