The Australian Injury Comorbidity Index to Predict Mortality.

STUDY OBJECTIVE: Existing comorbidity indices such as the Charlson comorbidity index are dated yet still widely used. This study derives and validates up-to-date comorbidity indices for hospital-admitted injury patients, specific to mortality outcomes. METHODS: Injury-related hospital admissions data for 2 cohorts of patients in the Australian state of Victoria were linked to mortality data: July 2012 to June 2014 (161,334 patients) and July 2006 to June 2015 (614,762 patients). Logistic regression models were fitted, and results were used to derive binary and weighted comorbidity indices to predict mortality outcomes. The indices were validated with data from New South Wales (Australia). RESULTS: There were 11 comorbidity groups identified as associated with inhospital death (cohort 1), 13 with 30-day mortality, and 19 with 1-year mortality (cohort 2). The newly derived weights for comorbidities were very different from the Charlson comorbidity index weights for some conditions. The area under the curve statistics for inhospital death, 30-day mortality, and 1-year mortality were similar for the newly derived binary comorbidity indices (0.920, 0.923, and 0.910, respectively), the Charlson comorbidity index (0.915, 0.919, and 0.906, respectively), and the Elixhauser comorbidity measure (0.924, 0.923, and 0.908, respectively). The false-negative rates for the new binary indices (15.8%, 15.8%, and 16.3%, respectively) were statistically equal to those of the Charlson comorbidity index (17.4%, 16.3%, and 16.5%, respectively) and the Elixhauser comorbidity measure (15.2%, 14.8%, and 16.3%, respectively). CONCLUSION: The newly derived Australian Injury Comorbidity Indices, which are a binary representation of individual conditions associated with the outcome of interest, are useful in quantifying the effect of comorbidity among injury patients. They include a shorter list of conditions than existing indices such as the Charlson comorbidity index and Elixhauser comorbidity measure, are up to date, and consider the individual association of each condition over a summed score such as the Charlson comorbidity index. Indices that quantify the effect of comorbidities should consider the population, disease prevalence, and outcome of interest and require periodic updating.

Effect of comorbidity on injury outcomes: a review of existing indices.

PURPOSE: Accounting for comorbidity in predicting outcomes for patients is vital in clinical care, epidemiological research, and health service planning. The aim of this study was to review published literature to compare the performance of existing comorbidity indices and their use in injury populations. METHODS: A thematic literature search for comorbidity indices and/or injury outcomes was conducted. Methods, results, and recommendations from selected articles were abstracted, documented, and compared; comparisons of results were made in terms of the indices' ability to predict outcomes, using the C-statistic, R2, and odds ratios. RESULTS: Fifty-two articles relating to the derivation and/or validation of comorbidity measures were found. The most commonly used measures were the Charlson Comorbidity Index (CCI) and the Elixhauser Comorbidity Measure (ECM). The ECM was found to outperform the CCI in terms of predictive ability, although the CCI was more widely used. Derivation of study-specific weights to the CCI added more predictive power to the index. CONCLUSIONS: Existing literature that compared the predictive abilities of the ECM and CCI favors the ECM. This literature review did not identify a measure specifically designed for general injury populations. Development of an injury-specific comorbidity measure will be timely and assist future research in injury epidemiology.