JMIR Publications

ABSTRACT

Background:

Nephrolithiasis affects up to 15% of the population and often remains undetected in asymptomatic individuals. Currently, no standardized or practical screening method exists. Imaging-based diagnostic tools such as ultrasound and CT are costly, operator-dependent, or involve radiation, making them unsuitable for large-scale screening.

Objective:

This study aimed to develop a low-cost, rapid screening model for clinically significant nephrolithiasis using machine learning (ML) and simple clinical parameters.

Methods:

We conducted a multi-hospital study using data from 3 hospitals in Kaohsiung, Taiwan (2012–2021). Adults without renal colic were included. ML models were trained and tested using ten routine variables: gender, age, BMI, gout, diabetes, eGFR, urine pH, red blood cells, specific gravity, and bacteriuria.

Results:

Among 6528 participants, the best-performing model achieved an AUC of 0.968 (95% CI, 0.956–0.980), sensitivity of 0.873, and specificity of 0.947. The model was developed using only Asian population data, which may limit generalizability.

Conclusions:

This ML-based model enables efficient, non-invasive, and large-scale kidney stone screening using routine health data. It can be integrated into health check-ups or telemedicine platforms to facilitate early detection and proactive management.