Leveraging Machine Learning to Enhance Preventive Care Continuity: A Predictive Analytics Case Study from Qassim in Alignment with Saudi Vision 2030

Background: Preventive health screening is a cornerstone of population health, but many patients fail to return for follow-up care, undermining early disease detection. This issue is highly pertinent in Saudi Arabia’s Qassim region, aligning with the national Vision 2030 healthcare transformation. Machine learning (ML) offers a promising predictive approach to identify patients at risk of "non-return" to enable targeted interventions. Objective: To develop and evaluate an ML-based model for predicting patient non-return after preventive screening in the Qassim region, identify associated risk factors, and align the findings with the Saudi Model of Care reforms. Methods: A retrospective observational analysis of electronic health records from Qassim’s primary care screening program (2019–2024) was conducted. The primary outcome was "non-return" within 6 months of an indicated follow-up. Multiple ML algorithms were evaluated using 10-fold cross-validation. Results: Among 18,752 screened patients, 5,230 (27.9%) did not return for follow-up. Ensemble tree-based methods performed best. The random forest classifier achieved the highest predictive performance (AUROC 0.812, accuracy 78.5%). Key predictors of non-return included extended lead time until the scheduled follow-up, prior appointment no-shows, and a lack of critical clinical findings during the initial screening. Conclusion: The developed ML model successfully predicts patient loss to follow-up with high accuracy. Integrating such predictive analytics into routine primary care enables early, personalized interventions, directly supporting Saudi Arabia’s healthcare efficiency and preventive care goals.