Optimizing hospital billing by using data from the Vascular Quality Initiative

Hospitals are reimbursed for inpatient admissions by Medicare based on the principal Medicare severity diagnostic related group (MS-DRG) assigned to that admission. MS-DRGs with complication or comorbidity (CCs) or major CCs (MCCs) increase reimbursement to reflect the added cost of caring for complex patients. Undercoding DRG complexity has been shown to result in unrecovered reimbursement and financial losses for hospitals. We hypothesized that granular, objective data collected by trained abstractors in the Society for Vascular Surgery Vascular Quality Initiative (VQI) could be used to improve MS-DRG coding accuracy by detecting CCs and MCCs that might be missed in the standard coding process.

The Medicare-linked VQI Peripheral Vascular Intervention (PVI) registry was queried to identify patients who underwent admission for PVI procedures at 230 centers between 2010 and 2019. Patients were included if the index DRG represented a PVI procedure. Cases were grouped into those without CC/MCC and those with CC/MCC. PVI registry characteristics for each group were compared. Logistic stepwise regression was used to identify variables predicting CCs/MCCs billing. The model was then used to calculate the variation in the expected number of admissions qualifying as CCs/MCCs at each center, compared with the observed number actually billed across centers.

We analyzed 40,822 admissions associated with PVI treatment, of which 76% of which were billed with CCs/MCCs. This rate varied substantially across the 230 hospitals, from 48% to 100%. Stepwise regression identified that preexisting congestive heart failure, diabetes mellitus, dialysis, prior amputation and dependent functional status, plus post-treatment cardiac, renal, pulmonary or access site complications, amputation, and longer length of stay were independently associated with hospitals billing an MS-DRG with CCs/MCCs. The model was highly accurate with a c-statistic of 0.82. The expected number of cases billed with CCs/MCCs exceeded the observed number at 89 of 230 centers, suggesting underbilling at 39% of centers.

Data available in the VQI PVI registry accurately identified admissions billed by VQI hospitals with MS-DRGs for CCs or MCCs. This multivariable model could be used to prepare reports for participating hospitals to identify cases likely to justify MS-DRG coding with CCs or MCCs, for special attention by coders. This would not be intended to supplant existing coding systems, but rather to provide additional validation and benchmarking that could help to avoid undercoding and loss of hospital revenue. This could provide added value to VQI hospitals at no additional cost, to help offset the cost of VQI participation.