Driving Impact in Claims Denial Management Using Artificial Intelligence

Abstract

A healthcare provider’s ability to quickly and efficiently process claims and quantify denial rates is critical to ensure smooth revenue cycle management and medical reimbursement. But the hospitals and medical practitioners are receiving more claim denials from payers, with the average rate of denial steadily increasing year over year. The recent COVID-19 pandemic has further accelerated the denial rate. An accurate denial detection algorithm can help to reduce the burden on healthcare providers. In this study, we propose a boosting-based machine learning framework to predict the likelihood of claims being denied along with the reason code at a line level. Prediction at a line level provides a finer-grained explanation to the administrative staff by pointing out the specific line for corrections. The list of important features provides an interpretable solution to the healthcare providers which enables them to create the right edits and correct the claim before going out to the payer. This in turn helps the healthcare provider dramatically improve both net patient revenue and cash flow. They can also put a check on their costs, as fewer denials mean less rework, resources, and time devoted to appealing and recovering denied claims. The denial model showed good performance with Area Under the Curve (AUC) of 0.80 and 0.82 for professional and institutional claims respectively. According to our estimates, the model has the potential to save 15%–50% of the denial cost for a healthcare provider. This in turn would have a tremendous impact on the healthcare costs as well as help make the healthcare process smoother.