Using ChatGPT-4 to Analyze 24-Hour Urine Results and Generate Custom Dietary Recommendations for Nephrolithiasis

Area of Science:

• Nephrology
• Artificial Intelligence
• Dietary Management

Background:

• Nephrolithiasis (kidney stones) incidence is rising, necessitating improved recurrence prevention strategies.
• Dietary modifications are key for prevention, but targeted counseling based on 24-hour urine analysis is underutilized.
• Artificial intelligence (AI) may offer novel solutions for analyzing complex patient data and personalizing treatment.

Purpose of the Study:

• To evaluate ChatGPT-4's capability in analyzing 24-hour urine collection results.
• To assess the AI model's effectiveness in generating custom dietary advice for nephrolithiasis prevention.
• To explore AI-assisted analysis and counseling for personalized kidney stone management.

Main Methods:

• Eleven unique prompts with synthesized 24-hour urine results were input into ChatGPT-4.
• The AI was instructed to provide five dietary recommendations per prompt.
• Responses were evaluated for accuracy, completeness, and appropriateness by urologists, a nephrologist, and a dietitian.

Main Results:

• ChatGPT-4 achieved high average scores for accuracy (5.2/6), completeness (2.4/3), and appropriateness (2.6/3).
• The model correctly identified normal urine values but struggled with consistently detecting abnormalities, especially calcium and citrate.
• ChatGPT-4 failed to address 3 out of 10 introduced abnormalities and required further refinement for dependable performance.

Conclusions:

• ChatGPT-4 demonstrates potential as a tool for dietary management in nephrolithiasis.
• AI-assisted analysis can provide personalized recommendations but needs improvement in identifying and addressing specific urine abnormalities.
• With refined prompting, physician oversight, and continued training, AI can form a basis for personalized medicine and reduce administrative burdens in managing kidney stones.