Variation and temporal trends in Medicare Severity Diagnosis Related Group billing for carotid endarterectomy across Vascular Quality Initiative centers

BACKGROUND: Hospitals use Medicare Severity Diagnosis Related Group (MS-DRG) coding to bill for each admission, with higher payment provided for those with complications or comorbidities (CCs) or major CCs (MCCs). We analyzed variation in MS-DRG coding for carotid endarterectomy (CEA) across Vascular Quality Initiative (VQI) centers to identify opportunities for improvement. METHODS: We matched CEA cases in the VQI registry with Medicare claims data from 2013 through 2023. A multivariable stepwise logistic regression model was created to identify variables associated with CC or MCC billing (with odds ratios [ORs]). We examined variations in CC or MCC billing percentages among centers and applied a multivariable model to compare observed-to-expected (O:E) CC or MCC billing rates to adjust for patient differences. Billing trends over time were analyzed using linear regression. RESULTS: Among 75,064 primary CEA admissions for Medicare beneficiaries, 39.5% were billed with CCs or MCCs (31.1% CCs, 8.4% MCCs) and 60.5% without. Eleven VQI clinical variables that indicated CCs were significantly associated with CC or MCC billing: dialysis dependence (OR, 22.30), postoperative neurological events (OR, 5.25), length of stay >2 days (OR, 4.98), pre-existing heart failure (OR, 4.08), any postoperative complication (OR, 3.70), reoperation (OR, 3.65), preoperative neurological event (OR, 1.70), higher American Society of Anesthesiologists class (OR, 1.50), urgent/emergent procedure (OR, 1.28), postoperative intravenous blood pressure management (OR, 1.27), and institutional transfer (OR, 1.16). Across 371 VQI centers, CC or MCC billing rates varied 5.6-fold, ranging from 15% to 84% of admissions. After risk adjustment, observed-to-expected ratios still demonstrated 4.2-fold variation, from 0.45 to 1.90 across centers. The observed rate of CC or MCC billing increased over time, from 35.5% in 2015 to 46.1% in 2023 (P < .001). In contrast, the expected adjusted rates predicted by the model remained stable (P = .269). A discrete inflection point was identified between 2016 and 2017, after which higher billing rates persisted (P < .001). CONCLUSIONS: Substantial variation exists for MS-DRG billing for CEA across VQI centers. Billing rates for CCs or MCCs increased from 2013 through 2023. These findings suggest potential underbilling at some centers and overbilling at others. Leveraging Medicare-linked VQI registry data and the proposed model could help to improve coding and billing accuracy, optimize reimbursement, and reduce Medicare audit risk.