A comparison of self-rated health, frailty index, and multimorbidity as predictors of four- year all-cause mortality: a multi-cohort study in 30+ adult populations

Background Identifying individuals at increased risk of adverse health outcomes is a key objective in public health and clinical care. Self-rated health (SRH) predicts several adverse health outcomes, but its ability to predict mortality compared with clinical health indicators is less well studied. The aims were to (i) assess the overlap between SRH, frailty, and multimorbidity; (ii) compare how they predict mortality; and (iii) explore how their discrepancies relate to mortality risk. Methods Data from five cohorts (N > 60,000; ages 30–99) from three Nordic countries included baseline measures of SRH, frailty (Rockwood frailty index), and multimorbidity (number of diseases). Age- and sex-adjusted Cox models were used to estimate four-year mortality risk. Predictive ability was assessed via Harrell’s c-index, goodness-of-fit, sensitivity, and specificity. Results Across cohorts: (i) SRH overlapped substantially—but not completely—with frailty and multimorbidity; (ii) all three indicators predicted four-year all-cause mortality well (c-index 0.7–0.9), with average c-index difference of ~ 1 percentage point. Poor SRH and frailty had high specificity but lower sensitivity, while multimorbidity captured more deaths with less precision; (iii) poor SRH predicted higher mortality risk both in individuals with and without frailty or multimorbidity, whereas good SRH attenuated the risk even among those with these conditions. Stratified analyses showed consistent patterns by sex, age, and educational attainment. Conclusions SRH has a substantial, but not complete overlap with frailty and multimorbidity. All three predicted four-year mortality with similarly high accuracy, with no clear superiority. SRH was a strong predictor both with and without frailty or multimorbidity.