Comparison of Elixhauser and Charlson Comorbidity Indices in Predicting Short-Term Outcomes in Minimally Invasive Thoracic Operations
*Alexander T. Kim1, *Li Ding2, *Brooks V. Udelsman1, *Scott M. Atay1, *Sean C. Wightman1, *Takashi Harano1, *Graeme M. Rosenberg1, *Justin D. Blasberg3, *Daniel J. Boffa3, Anthony W. Kim1
1. Division of Thoracic Surgery, Keck School of Medicine of the University of Southern California, Los Angeles, CA, United States. 2. Department of Population and Public Health Sciences, Keck School of Medicine of the University of Southern California, Los Angeles, CA, United States. 3. Section of Thoracic Surgery, Yale School of Medicine, New Haven, CT, United States.
Background: The Elixhauser (ECI) and Charlson-Deyo (CCI) comorbidity measures are two commonly used risk adjustment models, differing in number and categorization of comorbidities. This study aims to comparatively evaluate their ability in predicting short-term outcomes (in-hospital mortality, postoperative complications, nonroutine discharge, and 30-, and 90- day readmissions) in minimally invasive pulmonary lobectomy (MIL) and minimally invasive Ivor Lewis esophagectomy (MIE), two commonly performed index thoracic operations.
Study Design: The Healthcare Cost and Utilization Project National Readmission Database was queried from 2016 to 2018. MIL and MIE were identified using relevant International Classification of Diseases, 10th Edition (ICD-10) codes. Socioeconomic factors, comorbidities, and hospital characteristics were also analyzed. Multivariable logistic regression analyses were performed to assess the contributions of the ECI and CCI to outcome variables. Receiver Operating Characteristic (ROC) curves were plotted and regression models compared using the area under the ROC (AUC). Discriminative ability of the models was assessed using the concordance statistic derived from AUC. An AUC>0.7 was defined as the threshold for discriminative ability.
Results: In the MIL cohort, the CCI better predicted in-hospital mortality (AUC, 0.7866; 95% CI, 0.7549-0.8182) compared to the ECI (AUC 0.7561; 95% CI, 0.7214-0.7908, p=0.003) Figure 1a. Nonroutine discharges were predicted better with the CCI (AUC 0.6427; 95% CI, 0.6362-0.6492) compared to ECI (AUC 0.6399; 95% CI, 0.6333-0.6464, p=0.01) but only marginally and with an insufficient discriminative ability. Both displayed poor discriminative ability and no significant differences regarding postoperative complications, 30-day, and 90-day readmissions for MIL. In the MIE cohort, there was no significant difference in predicting inpatient mortality between the two models (ECI 0.7991; 95% CI 0.7199-0.8783 vs. CCI 0.8179; 95% CI 0.7454-0.8905, p=0.26) Figure 1b. ECI (AUC 0.6791; 95% CI, 0.633-0.7252) better predicted 90-day readmissions compared to CCI (AUC 0.6379; 95% CI, 0.5896-0.6861, p=0.02), but with insufficient discriminative ability. Both models exhibited poor discriminative ability and no significant differences regarding nonroutine discharges, 30-day, and 90-day readmissions for MIE.
Conclusions: Based upon two commonly performed index thoracic surgery operations, CCI better predicted in-hospital mortality than ECI for only minimally invasive pulmonary lobectomy. Despite some differences between CCI and ECI, both indices lacked the discriminative ability in predicting other postoperative outcomes. These variations in predictive ability indicate that postoperative outcome measures may differ by comorbidity index employed and by surgery type, suggesting the need for careful selection of the appropriate risk adjustment models especially once patients are deemed fit for a thoracic operation.
Figure 1 demonstrating receiver operating characterstic curves with corresponding c-statistics for the Elixhauser (ECI) and Charlson (CCI) comorbidity risk adjustment models in both minimally invasive pulmonary lobectomy and minimally invasive Ivor-Lewis esophagectomy. 1a demonstrates CCI is a better predictive model for inpatient mortality in minimally invasive pulmonary lobetomy. 1b demonstrates no significant difference between ECI or CCI in predicting mortality in minimally invasive esophagectomy.
Back to 2024 Posters