Artificial Intelligence Integration Into Hospital Medication Management: Combining Medication Demand Predictive Analysis, Medication Pattern Basket Analysis And Patient Feedback Sentiment Analysis

Abstract

Background: Hospital systems face persistent challenges in efficient medication administration, particularly in balancing inventory levels and addressing patient satisfaction. While AI has shown promise in healthcare, there remains a lack of integrated multi-model frameworks tailored to regional healthcare environments.

Objective: This study aims to develop, implement, and evaluate an integrated artificial intelligence framework that combines medication demand prediction, pattern basket analysis, and patient sentiment analysis to optimize medication management in outpatient pharmacies in hospitals.

Methods: A regression-based demand prediction model was created using a 36-month dataset (January 2022–December 2024) that included 300 different types of medications. In order to find co-prescription patterns, basket analysis was performed on 200 prescriptions (70 acute care, 130 chronic disease cases), and sentiment analysis was used to categorize patient feedback into positive, neutral, and negative groups.

Results: The demand prediction model for monthly medication orders had an R2 value of 0.918 and forecasting accuracy of 83.3%. Basket analysis was used to find clinically significant co-prescription relationships, which pharmacists validated. Sentiment analysis produced an 89.4% validation accuracy. Restrictions include data from three hospitals only, and privacy issues have limited generalizability. The true synergy occurs within the integration of the three models, where these outputs are triangulated into three actionable pillars.

Conclusions: The integrated AI framework demonstrates significant potential to enhance clinical safety and inventory efficiency in hospital pharmacies. Future research should focus on cross-hospital validation and larger datasets to increase scalability.