Machine Learning Models for Patient Screening Using Routinely Collected Data in Primary Care

Screening is crucial to preventing the health consequences associated with undiagnosed diseases. Electronic health records (EHRs) from primary care can be leveraged with machine learning (ML) techniques to create new tools for patient screening in general practice. The aim of this narrative review is to discuss the recent literature on the development and validation of predictive ML models designed for the early detection of health conditions using readily available patient data. The PubMed, Web of Science, Scopus, and IEEE Xplore databases were searched for studies published within the last five years. Twenty-one studies were found, covering a variety of health conditions. ML-based tools can function as independent screening tests or can enhance existing screening methods. Moreover, ML models can be employed to screen for conditions for which screening approaches have not yet been developed. However, primary care EHRs alone are not always a sufficient source of data for effective screening. Poor data quality can result in erroneous or biased predictions. Despite these limitations, the application of ML for screening has shownpromising results, and further research in this area is warranted.