Machine learning-based risk-adjusted capitation: an efficient payment model for Kenya’s primary healthcare at the NHIF

Abstract

In Kenya, the National Health Insurance Fund (NHIF) aims to provide quality and affordable care to its members. To achieve this, efficient provider payment mechanisms are essential. Capitation—a prepayment model where providers receive fixed monthly payments per enrolled member regardless of actual services delivered (for example, KSh 1,000 per member per month for primary care services)—incentivizes cost containment and efficiency. However, when all providers receive identical payments regardless of their patient population’s health risks, those serving sicker populations face financial disadvantages. This research explores the application of K-means clustering to risk-adjusted capitation for more equitable provider payments. Unlike traditional risk adjustment methods that rely on predetermined categories, K-means clustering algorithmically identifies natural groupings of patients with similar characteristics. For instance, the algorithm can differentiate between low-cost healthy young adults and high-cost elderly patients with chronic conditions, allowing payment rates to be calibrated accordingly. Our analysis identified two optimal patient clusters with average claim amounts of 4,848 and 4,930, both significantly lower than the dataset’s overall average claim of 5,917, indicating meaningful population segmentation. These clusters, verified through Principal Component Analysis (PCA), demonstrated homogeneity in cost patterns and demographic factors. By integrating this approach, NHIF could replace its current uniform capitation rates with differentiated payments—higher rates for providers serving predominantly high-risk populations and standard rates for those with healthier enrolees. The findings demonstrate how advanced analytics can transform healthcare financing by enabling more precise resource allocation, reducing provider incentives to avoid high-risk patients, and promoting proactive preventive care aimed at keeping patients healthier. This data driven approach offers NHIF a practical pathway toward more sustainable, equitable healthcare financing that aligns provider incentives with improved population health outcomes. Keyword: NHIF, K-means, PCA, Davies-Bouldin Index (DBI), Calinski-Harabasz Index, Risk- Adjusted Capitation