Evaluation of Diagnostic Recommendations Embedded in Medication Alerts: Prospective Single-Arm Interventional Study

Area of Science:

• Health Informatics
• Machine Learning in Healthcare
• Clinical Decision Support

Background:

• Potentially inappropriate prescribing in outpatient settings leads to adverse outcomes and inefficiencies.
• Clinical decision support systems (CDSS) show promise but are limited by incomplete medical records.

Purpose of the Study:

• To evaluate a machine learning-based CDSS for improving diagnostic recommendations.
• To ensure prescribed medications have documented diagnoses and meet appropriateness criteria.

Main Methods:

• Prospective, single-arm interventional study over one year in hospital outpatient departments.
• Machine learning algorithms trained on national health insurance data provided diagnostic recommendations.
• Outcome measures included alert and acceptance rates, with descriptive and trend analyses.

Main Results:

• The system (MedGuard) processed 438,558 prescriptions from 125,000 patients.
• An overall alert rate of 2.28% and a diagnostic recommendation acceptance rate of 56.55% were observed.
• Accepted recommendations led to prescription adjustments or added diagnoses; ophthalmology had the highest acceptance (96.59%).

Conclusions:

• Embedding diagnostic recommendations in ML-based CDSS alerts can enhance diagnostic completeness and outpatient safety.
• Refining alerts for specialty-specific workflows and validating in diverse settings are crucial for future efforts.