Abstracting and Indexing

• PubMed NLM
• Google Scholar
• Semantic Scholar
• Scilit
• CrossRef
• WorldCat
• ResearchGate
• Academic Keys
• DRJI
• Microsoft Academic
• Academia.edu
• OpenAIRE
• Scribd
• Baidu Scholar

Introduction: Chronic non-communicable illnesses, including type 2 diabetes, hypertension, chronic kidney disease, cardiovascular disorders, and chronic obstructive pulmonary disease, are increasingly common in India’s semi-urban areas. Because primary-care clinics are rarely used for proactive screening, timely diagnosis and prevention in these low-resource settings remain difficult. This study assessed whether incorporating machine-learning–based prediction tools into routine primary-care visits can practically and effectively uncover previously undiagnosed chronic diseases.

Methods: In a prospective observational study at Katihar Medical College (Bihar), we enrolled 250 adults from April 2023 to March 2024. Demographic details, clinical findings, and basic laboratory results were collected and used to train logistic-regression and random-forest algorithms. Model performance was judged by area under the receiver-operating-characteristic curve, sensitivity, specificity, and calibration. The resulting risk scores were presented to physicians through a point-of-care digital dashboard.

Results: Predictive models identified high-risk profiles in 56.8% of participants. Confirmatory diagnoses revealed 128 new cases of chronic disease, with random forest models achieving AUROC values ranging from 0.78 to 0.89 across conditions. Physicians reported improved diagnostic confidence and patient engagement. The tool demonstrated high clinical utility and integration feasibility.

Conclusion: Predictive analytics can effectively augment routine primary care in early disease identification.