The M3 multimorbidity index outperformed both Charlson and Elixhauser indices when predicting adverse outcomes in people with diabetes

Abstract

ObjectivesMultimorbidity is common among those living with diabetes. The purpose of this manuscript was to assess the performance of the recently developed M3 multimorbidity index within a large, well-defined cohort of patients with diabetes. Study Design and SettingWe compared regression model performance between three morbidity indices (M3, Charlson, and Elixhauser) for important adverse outcomes (hospitalization, amputation, and mortality) among 217,207 patients with diabetes and compared the real-world applied impact of these indices for modeling these diabetes outcomes against two key exposure variables (ethnicity and deprivation). ResultsThe M3 Index outperformed both Charlson and Elixhauser indices in terms of the prediction of hospitalization events, amputation events, and death. When applied to models that were investigating the relationship between (a) ethnicity and (b) deprivation on risk of these outcomes, the M3 Index also explained more confounding due to multimorbidity than either Charlson or Elixhauser (e.g., shift in odds ratio toward the null when predicting hospitalization: M3 Index 60%, Charlson 40%, and Elixhauser 24%). ConclusionThese results give additional support for the utility of the M3 index in clinical populations, adding to previous observations regarding its performance in a general population setting.