Rule-Based Prediction of Medical Claims' Payments: A Method and Initial Application to Medicaid Data

Imperfections in healthcare revenue cycle management systems cause discrepancies between submitted claims and received payments. This paper presents a method for deriving attributional rules that can be used to support the preparation and screening of claims prior to their submission to payers. The method starts with unsupervised analysis of past payments to determine normal levels of payments for services. Then, supervised machine learning is used to derive sets of attributional rules for predicting potential discrepancies in claims. New claims can be then classified using the created models. The method was tested on a subset of Obstetrics claims for payment submitted by one hospital to Medicaid. One year of data was used to create models, which were tested using the following year's data. Results indicate that rule-based models are able to detect abnormal claims prior to their submission.