Marc Simard1, Caroline Sirois1,2,3, Bernard Candas2,4. 1. Quebec National Institute of Public Health. 2. Department of Social and Preventive Medicine, Faculty of Medicine, Laval University. 3. Centre of Excellence on Aging of Quebec, Integrated University Health and Social Services Centres of the Capitale-Nationale. 4. National Institute of Excellence in Health and Social Services, Québec, QC, Canada.
Abstract
OBJECTIVES: To validate and compare performance of an International Classification of Diseases, tenth revision (ICD-10) version of a combined comorbidity index merging conditions of Charlson and Elixhauser measures against individual measures in the prediction of 30-day mortality. To select a weight derivation method providing optimal performance across ICD-9 and ICD-10 coding systems. RESEARCH DESIGN: Using 2 adult population-based cohorts of patients with hospital admissions in ICD-9 (2005, n=337,367) and ICD-10 (2011, n=348,820), we validated a combined comorbidity index by predicting 30-day mortality with logistic regression. To appreciate performance of the Combined index and both individual measures, factors impacting indices performance such as population characteristics and weight derivation methods were accounted for. We applied 3 scoring methods (Van Walraven, Schneeweiss, and Charlson) and determined which provides best predictive values. RESULTS: Combined index [c-statistics: 0.853 (95% confidence interval: CI, 0.848-0.856)] performed better than original Charlson [0.841 (95% CI, 0.835-0.844)] or Elixhauser [0.841 (95% CI, 0.837-0.844)] measures on ICD-10 cohort. All weight derivation methods provided close high discrimination results for the Combined index (Van Walraven: 0.852, Schneeweiss: 0.851, Charlson: 0.849). Results were consistent across both coding systems. CONCLUSIONS: The Combined index remains valid with both ICD-9 and ICD-10 coding systems and the 3 weight derivation methods evaluated provided consistent high performance across those coding systems.
OBJECTIVES: To validate and compare performance of an International Classification of Diseases, tenth revision (ICD-10) version of a combined comorbidity index merging conditions of Charlson and Elixhauser measures against individual measures in the prediction of 30-day mortality. To select a weight derivation method providing optimal performance across ICD-9 and ICD-10 coding systems. RESEARCH DESIGN: Using 2 adult population-based cohorts of patients with hospital admissions in ICD-9 (2005, n=337,367) and ICD-10 (2011, n=348,820), we validated a combined comorbidity index by predicting 30-day mortality with logistic regression. To appreciate performance of the Combined index and both individual measures, factors impacting indices performance such as population characteristics and weight derivation methods were accounted for. We applied 3 scoring methods (Van Walraven, Schneeweiss, and Charlson) and determined which provides best predictive values. RESULTS: Combined index [c-statistics: 0.853 (95% confidence interval: CI, 0.848-0.856)] performed better than original Charlson [0.841 (95% CI, 0.835-0.844)] or Elixhauser [0.841 (95% CI, 0.837-0.844)] measures on ICD-10 cohort. All weight derivation methods provided close high discrimination results for the Combined index (Van Walraven: 0.852, Schneeweiss: 0.851, Charlson: 0.849). Results were consistent across both coding systems. CONCLUSIONS: The Combined index remains valid with both ICD-9 and ICD-10 coding systems and the 3 weight derivation methods evaluated provided consistent high performance across those coding systems.
Authors: Uma V Mahajan; Hammad A Khan; Xiaofei Zhou; Shaarada Srivatsa; Christina H Wright; Adam H Bates; Martha Sajatovic; Nicholas C Bambakidis Journal: Neurocrit Care Date: 2022-09-16 Impact factor: 3.532
Authors: Wayne A Ray; Cecilia P Chung; C Michael Stein; Walter Smalley; Eli Zimmerman; William D Dupont; Adriana M Hung; James R Daugherty; Alyson Dickson; Katherine T Murray Journal: JAMA Date: 2021-12-21 Impact factor: 157.335
Authors: Jessica J Kirby; Sajid Shaikh; David P Bryant; Amy F Ho; James P d'Etienne; Chet D Schrader; Hao Wang Journal: J Clin Med Res Date: 2021-04-27