Assessing the Predictive Accuracy of Popular Comorbidity Indices in Total Ankle Arthroplasty Outcomes

Submission Type: Ankle Arthritis Research Type: Level 3 - Retrospective cohort study, Case-control study, Meta-analysis of Level 3 studies Introduction/Purpose: As the incidence of TAA increases, identifying ways to stratify a patient’s risk for adverse outcomes becomes imperative. Two potential tools, the Charlson Comorbidity Index (CCI) and Elixhauser Comorbidity Index (ECI), have been widely supported and utilized across various fields of medicine and orthopaedic surgery. Despite encouraging results in TAA literature, the validity of the CCI and ECI as predictive tools for postoperative complications and functional outcomes following TAA remains ultimately unexplored. This study aims to describe and compare the predictive capacity of the CCI and ECM on the outcomes of TAA to further understand the reliability of such comorbidity indices in orthopaedic research. Methods: The Nationwide Readmissions Database (NRD) was queried from 2015 to 2020 to identify 29,705 patients undergoing primary TAA. Patients’ comorbidity was measured via CCI and ECI score, with patients who experienced adverse postoperative outcomes (any complication, readmission, mortality, extended length of stay (LOS), and adverse discharge) identified. Each index’s predictive ability was measured using the c statistic, a measure of area under the receiver operating characteristic curve (AUC). The value for AUC ranges from 0.50 to 1.0, indicating the discriminative ability of each index in assigning probability of the examined outcome, with 0.50 indicating no ability to discriminate and 1.0 indicating perfect ability to discriminate. AUC was categorized into poor (AUC <.70), acceptable (0.70< AUC < 0.80), excellent (0.80< AUC < 0.90), and outstanding (0.90< AUC < 1.0) regarding the predictive capability of each model. The indices were also compared to a base model, which considered age, sex, and primary payer. Results: The overall cohort was majority male (54.2%) with a mean age of 64.15 (range 17-90) years, and the majority of patients had Medicare as their primary expected payer (58.9%). Elixhauser comorbidity index provided a superior predictive ability for any complication (ECI AUC=0.61, CCI AUC=0.59; p<.001), extended LOS (ECI AUC=0.69, CCI AUC=0.65; p<.001), and adverse discharge (ECI AUC=0.70, CCI AUC=0.68; p<.001) as compared to the Charlson comorbidity index. Examination of each model’s predictive ability found mortality was the only variable that both ECI (AUC=0.88, 95%Confidence Interval [CI]= 0.77–1.00) and CCI (AUC=0.88; 95% CI=0.79–0.96) were able to predict with excellent capability. Additionally, the ECI predicted adverse discharge with acceptable capability (AUC=0.70; 95%CI=0.69-0.71). Conclusion: In this study, the ECI outperformed the CCI in predicting any complication, extended length of stay ≥4 days, and adverse discharge to a facility in patients undergoing total ankle arthroplasty. However, the only variable that was excellently predicted was 180-day mortality, in which there was no difference in the predictive capability of each index. These findings indicate that while ECI and CCI are capable of predicting postoperative mortality following TAA, there is a need for alternative models that offer better predictive capability when examining other postoperative adverse outcomes.