New ICD-10 version of the Multipurpose Australian Comorbidity Scoring System outperformed Charlson and Elixhauser comorbidities in an older population

Objectives To translate, validate, and compare performance of an International Classification of Diseases, 10th revision (ICD-10) version of the Multipurpose Australian Comorbidity Scoring System (MACSS) against commonly used comorbidity measures in the prediction of short- and long-term mortality, 28-day all-cause readmission, and length of stay (LOS). Study Design and Setting Hospitalization and death data were linked for 25,374 New South Wales residents aged 65 years and older, admitted with a hip fracture between 2008 and 2012. Comorbidities were identified according to the MACSS, Charlson, and Elixhauser definitions using ICD-10 coding algorithms. Regression models were fitted and area under the curve (AUC) and Akaike Information Criterion assessed. Results The ICD-10 MACSS had excellent discriminating ability in predicting inhospital mortality (AUC = 0.81) and 30-day mortality (AUC = 0.80), acceptable prediction of 1-year mortality (AUC = 0.76) but poor discrimination for 28-day readmission and LOS. The MACSS algorithm provided better model fit than either Charlson or Elixhauser algorithm for all outcomes. Conclusion This work presents a rigorous translation of the ICD-9 MACSS for use with ICD-10 coded data. The updated ICD-10 MACSS outperformed both Charlson and Elixhauser measures in an older population and is recommended for use with large administrative data sets in predicting mortality outcomes.