Statistical Method for Monitoring Non-Acceptable Diagnosis Related Group (DRG)

摘要

The Medicare program, private insurers, and managed care organizations reimbursehospitals for in-patient admissions using the Diagnosis Related Group (DRG). The DRG is determined from a complicated algorithm based on patient medical records. Previous studies have generated concerns of 'DRG upcoding', where incorrect DRG codes may be selected to obtain a higher reimbursement. Insurers rely on expensive manual audits of claims to verify the appropriateness of the underlying DRG coding. As part of a larger statistical system, we developed a hierarchical Bayesian logistic regression for detecting claims with incorrect DRG coding using insurer claims data, together with results from a manual audit. Estimates were developed from an insurer's 1993-5 audited claim data and applied to 5,278 additional audited claims from the same time frame (1,671 claims were coded incorrectly). We propose that a claim should be investigated if the predicted recovery is more than the cost of auditing that claim. Of the 5,278 claims, the proposed system achieves 98% of the recovery of a complete audit at 88% of the investigation cost. Results from the model can be used to select claims for future audits.