SS05 - Optimizing Hospital Billing by Using Data from the Vascular Quality Initiative (VQI)

Objectives: Over one billion dollars in peripheral vascular care is billed to Medicare annually, and in recent years, vascular surgical procedures have seen an increase in charge-to-reimbursement ratio. It is important, therefore, that care is accurately represented to ensure a sustainable future for the provision of vascular care. Proposed solutions such as direct physician involvement in case coding increase accuracy, but pose heavy administrative burden on surgeons. We hypothesized that data from VQI could be used to identify cases appropriate for a Diagnosis-Related Group (DRG) with complication/comorbidity (CC) or major complication/comorbidity (MCC), to avoid underbilling such cases.

Methods: The Medicare-linked VQI PVI registry was queried to identify patients who underwent PVI procedures at 183 centers between 2010-19. Patients were included if the index DRG represented a major cardiovascular procedure. Cases were grouped into “without CC/MCC” and “with CC/MCC.” Baseline characteristics for each group were compared. A bootstrap method of stepwise regression with 100 samples was created to predict whether a case was likely to have CC/MCC. Based on the remaining variables, we generated an expected number of cases with CC/MCC at each participating institution, and compared this to the observed number of cases with CC/MCC. Model performance was evaluated with a correlation coefficient.

Results: 41,177 patients met the inclusion criteria. Univariate comparison showed that patients with CC/MCC were younger (72 years vs 73 years, p< 0.001) and more likely to have urgent (30.6% vs 18.49%, p< 0.001) or emergent (5.1% vs 3.9%, p< 0.001) procedures. [Table 1]. Stepwise regression identified that gender, history of heart failure, dialysis, stress testing, percutaneous coronary intervention, prior aneurysm repair, prior amputation, concomitant endarterectomy, target lesion re-treatment, hematoma, and length of stay were the key predictors of CC/MCC. The c-statistic for the model was 0.816, representing strong association. The expected number of cases with CC/MCC exceeded the observed number at 67/183 centers, suggesting underbilling at 36.6% of centers. [Figure 1]. The mean Medicare payment for cases without CC/MCC was $10,721 and for cases with CC/MCC, $17,804.

Conclusions: Our VQI-based model demonstrates a strong c-statistic for predicting if procedures should be billed with CC/MCC, and demonstrates that 37% of centers may be underbilling PVI procedures. For centers participating in the VQI, this may be a time- and cost-efficient method of improving coding accuracy.