Integration of machine learning HIV Risk Scoring Model into KenyaEMR
On April 19, 2022, Data for Impact (D4I) held the next webinar in a series focusing on integration in global health monitoring, evaluation, and learning. This webinar, led by Palladium’s Antony Ojwang and Jonathan Friedman, explored the integration of machine learning models with electronic medical records, achieved through the Kenya Health Management Information Systems (KeHMIS) project.
Under KeHMIS, Palladium developed a model to predict the risk a patient will test positive for HIV using information collected and available during routine client screening. The model was developed in R and piloted via a web app in Homa Bay county. For a more sustainable solution, Palladium sought to integrate the model into KenyaEMR, the EMR developed by KeHMIS and based on OpenMRS, so that the model could generate predictions without internet connectivity. To do this, Palladium converted the model from R to Predictive Model Markup Language (PMML) and used a Java library to integrate the model directly in KenyaEMR.
During the webinar, the presenters walked through the process of building, testing, and integrating the machine learning model and shared lessons learned on model integration and systems implementation.
Antony Ojwang is Palladium’s EMR Architect and Technical Lead under the KeHMIS project where he provides overall technical leadership for the architecture and implementation of KenyaEMR and mHealth products. He holds an MSc in Applied Computing (Health Informatics) and a BSc in Information Sciences (Information Technology).
Jonathan Friedman is a senior technical adviser for data science at Palladium, overseeing machine learning initiatives across global health projects covering predictive analytics, optimization, and anomaly detection. He previously led data science projects for the US Internal Revenue Service and Department of Defense and has a BA in International Affairs and an MA in International Economics.