Predicting Costs Exceeding Bundled Payment Targets for Total Joint Arthroplasty.

Abstract

BACKGROUND\nThe Center for Medicare and Medicaid Services has instituted bundled reimbursement models for total joint arthroplasty (TJA), which includes target prices for each procedure. Some patients exceed these targets; however, currently there are no tools to accurately predict this preoperatively. We hypothesized that a validated comorbidity index combined with patient demographics would adequately predict excess cost-of-care prior to hospitalization.\n\n\nMETHODS\nTwo thousand eighty-four primary unilateral TJAs performed at a single tertiary center were retrospectively examined. Data were extracted from medical records and a predictive model was built from 30 comorbidities and 7 patient demographic factors (age, gender, race, body mass index, American Society of Anesthesiologists score, smoking status, and marital status). Following parameter selection, a final multivariable model was created, with a corresponding nomogram for interactive visualization of probability for excess cost.\n\n\nRESULTS\nSix hundred twelve patients (29%) had cost-of-care exceeding the target price. The final model demonstrated adequate predictive discrimination for cost-of-care exceeding the target price (area under the receiver operator characteristic curve: 0.747). Factors associated with excess cost included age, gender, marital status, American Society of Anesthesiologists score, body mass index, and race, as well as 7 Elixhauser comorbidities (alcohol use, rheumatoid arthritis, diabetes, electrolyte disorders, neurodegenerative disorders, psychoses, and pulmonary circulatory disorders).\n\n\nCONCLUSION\nA novel patient model composed of a subset of validated comorbidities and demographic variables provides adequate discrimination in predicting excess cost within bundled payment models for TJA. This not only helps identify patients who would benefit from preoperative optimization, but also provides evidence for modification of future bundled reimbursement models to adjust for nonmodifiable risk factors.