Rule-Based Detective Model for Medical CPT Coding Errors in an Outpatient Community Health Medical Center

Medical coding has been in place for years in the healthcare industry and constantly evolving. Errors in medical coding can happen for multiple reasons such as improper understanding of medical coding guidelines, the complexity of the coding systems, and continuous changes in medical coding. This paper develops a rule-based model that detects certain CPT medical coding errors in an outpatient rural setting. The Python-based model can detect three types of predefined errors: under-coding, over-coding, and mismatched code. Each error type contained several scenarios of mistakes. The model was evaluated by medical coding experts using three datasets for validation (1,922 visits), testing (2,132 visits), and implementation (2,031 visits). The model identified 80 percent of the errors flagged by the experts during their review. Mismatched codes are the most common errors and 90% of these errors the model identified were corroborated by the expert panel. The model performance is consistent as the detection accuracy is highly similar for each error scenario using the three evaluation datasets. Based on the results, the model can effectively reduce the labor time and cost needed for manual audits by medical coders, generate more revenue, reduce costs of miscoded visits and claims, and improve coding practices and accuracy. The expected annual monetary benefit of the model at FLCH ranges from $70,000 to $100,000. One of the strengths of this model is its generalizability where it can be implemented at any outpatient medical center, as medical coding is standardized and universal.