Sustainabilty of Machine Learning in Health Claims Automation in the Kenyan Insurance Industry

Abstract

Background Hospitals in Kenya have embraced the use of hospital management systems to tracking patient’s data. The automation of this data has improved the efficiency of the insurance system and ultimately the settlement of healthcare claims. Claims processing is the most important function for any insurance company. The speed and convenience with which the claims are settled has a bearing on the general reputation of the insurance company. Problem The insurance health industry in the country faces difficulties which include increasing fraud, rising costs, and a high number of claims ratios. There are approximately 30 million claims in the country every year that must be reviewed, approved, and paid every year. The major challenge affecting this is the process is not fully automated. Purpose This research analyzed the various machine learning models to determine the accuracy of health claims automation. Method This was achieved by using, the supervised machine learning model was used to determine the accuracy. The supervised machine learning models used were K- Nearest Neighbor, Naïve Bayes and LDA- Linear Discriminant Analysis. Findings Experiments were conducted and K -Nearest Neighbor was the best model to determine the accuracy of claims with a 99.9% accuracy and it produced the best results with smaller and larger datasets. Conclusion ML is a possible solution for the challenges that are facing insurers due to lack of automation. Applying ML not only in a health insurance industry but also other insurance industries. e.g., general insurance can be used to improve the business, increase income of the insurer, and reduce costs. The timely payment of a claim to a hospital will also improve the relationships between the insurer and the hospital.