Artificial Intelligence and Electronic Health Records: A Systematic Review in Dentistry, Family Medicine, Infectious Disease

Abstract

Background: Electronic health records (EHRs) generate large volumes of structured and unstructured clinical data that are increasingly mined by artificial intelligence (AI) and machine learning (ML) tools to support diagnosis, risk prediction, and care management. Dentistry, family medicine, and infectious disease care each depend heavily on EHR documentation, yet evidence on EHR-based AI applications across these three fields has not been jointly synthesized.
Objective: To systematically review primary studies that developed, validated, or applied AI/ML models using EHR or electronic dental record (EDR) data in dentistry, family medicine, or infectious disease settings.
Methods: PubMed/MEDLINE, Scopus, Embase, Web of Science, and the Cochrane Library were searched for English-language randomized controlled trials, cohort, case-control, cross-sectional, and case-series studies published between 2009 and 2026. Screening and data extraction followed the PRISMA 2020 statement. Risk of bias was assessed across participant-selection, predictor, outcome, and analysis domains. Findings were synthesized narratively because of methodological and outcome heterogeneity.
Results: Of 1,848 records identified, 12 studies met the inclusion criteria (4 in dentistry, 4 in family medicine, 4 in infectious diseases), comprising sample sizes from 218 to over 130,000 patients. Gradient-boosted ensemble models, particularly XGBoost, were the most frequently top-performing algorithms across all three domains. Reported discrimination was generally moderate to good where quantified, and natural language processing of unstructured EHR text improved prediction in infectious disease applications. Most studies were single-center, retrospective, and lacked external validation, and risk of bias was predominantly rated as “some concerns.”
Conclusion: EHR-embedded AI models show consistent, domain-spanning potential to augment early risk identification in dentistry, family medicine, and infectious diseases, but methodological heterogeneity and limited external validation currently constrain clinical translation. Prospective, multicenter, externally validated studies with standardized outcome reporting are needed.