Abstract 16468: Upcoding of Complication May Account for Declining Failure to Rescue Rates Following Major Cardiac Surgery

Abstract

Introduction: To guide public reporting and payment reform initiatives, payers are increasingly tying reimbursement to hospital performance on administratively-derived quality measures. This has created tension with the existing fee-for-service model where “upcoding” of post-operative complications may increase hospital revenue. Hypothesis: We hypothesized that current complication measures may inaccurately reflect hospital quality, due to increased upcoding. Methods: We used Medicare data (2000-8) for patients undergoing cardiac surgery. We compared rates of post-operative pulmonary complications and pulmonary-specific failure to rescue (FTR, ie mortality after a pulmonary complication) under one of two definitions: (1) allowing diagnostic codes solely for pulmonary failure to define a complication (unconditioned definition, UCD), or (2) conditioning the identification of a complication on having a treatment code for mechanical ventilation or intubation (procedure-conditioned definition, PCD). We considered PCD to be a better reflection of true hospital quality. Results: We studied 1,248,765 operations in 1027 hospitals. Over this time period, pulmonary complications decreased 0.3% under the PCD criteria but increased 14.3% under an UCD, p<0.001. Since the PCD rate remained essentially unchanged (6.2% vs 5.9%, p<0.001), this divergence was driven almost entirely by the prodigious increase in UCD complications (12.5% vs 20.3%, p<0.001), Figure. There was no change in PCD-FTR rate (15.4% vs 15.5%, p=0.876) but a sizable decline in UCD-FTR rate (18.9% vs 11.7%, p<0.001). Conclusions: Our findings suggest that current hospital surgical performance measures are sensitive to upcoding, and may provide qualitatively different reflections of quality depending on how they are derived. Measure developers should consider the implications of hospital upcoding, especially given its increasing incidence.