Developing Mixed-effects Models to Compare the Predictive Capability of Various Comorbidity Indices on Short-term Patient Outcomes Following Lumbar Fusion Procedures

Introduction: As the incidence of operative spinal pathology grows, so do the rates of lumbar spinal fusion procedures. Comorbidity indices can be used preoperatively to predict complications in patients requiring surgery. However, there is a paucity of research defining optimal comorbidity indices in patients undergoing spinal fusion. The aim of this study was to utilize modeling strategies to evaluate the predictive validity of various comorbidity indices and combinations thereof.

Methods: We used the Nationwide Readmissions Database (NRD) from 2016-2019 to identify patients who received spinal fusion surgery. Patient frailty was queried using the Johns Hopkins Adjusted Clinical Groups indicator. Elixhauser Comorbidity Index (ECI) scores were collected to enable measurement of comorbidity status. Receiver operating characteristic (ROC) curves were created for relevant complications, including mortality, non-routine discharge, cost, length of stay (LOS), and readmission.

Results: A total of 750,183 patients were included. Non-routine discharges occurred in 161,077 (21.5%) patients. The mean adjusted all-payer cost for the procedure was $37,616.97±$27,408.86 (Top Quartile: $45,409.20) and the mean length of stay (LOS) was 4.1 days±4.4 days (Top Quartile: 8.1 days). Models using Frailty+ECI as the primary predictor consistently outperformed other models with statistically significant p-values, determined by comparing AUCs, for most complications. However, for prediction of mortality, the model using Frailty+ECI was no better than the model using ECI alone (p=0.23), and for prediction of cost, the ECI model outperformed the the models using frailty alone (p < 0.0001) and the model using frailty+ECI (p < 0.0001).

Conclusion : This investigation is the first to utilize big data modeling strategies to delineate the predictive utility of comorbidity indices for the prognostication of patients undergoing lumbar fusion surgery. With the knowledge gained from our models, spine surgeons, payers, and hospitals may be able to identify vulnerable patients more effectively within their practice who may require additional resource utilization in the preoperative setting.