Avaliação da eficácia de um sistema de IA na redução de erros das prescrições medicamentosas
Saved in:
| Hovedforfatter: | |
|---|---|
| Publication Date: | 2025 |
| Format: | Master thesis |
| Sprog: | por |
| Source: | Repositório Institucional da UPF |
| Download full: | https://repositorio.upf.br/handle/123456789/9624 |
Summary: | With the growing number of new medications and an aging population, the use of drugs to manage chronic diseases has increased significantly. This makes it essential to closely monitor medical prescriptions to ensure dosages remain within recommended guidelines and to identify potential drug interactions before they become an issue. In this study, we used the NoHarm.ai tool, designed to detect medications prescribed in doses that fall outside standard ranges. Our objective was to evaluate whether the platform accurately flagged prescriptions with dosages that were either above or below established recommendations. We also aimed to assess whether patients could be at risk due to these dosing discrepancies. Within this context, artificial intelligence functions as an initial screening stage in a hybrid review system for prescription analysis. We conducted a prospective observational cohort study to examine prescribing practices in a 22-bed ICU at a tertiary hospital. Previously, we had performed a systematic review, which revealed a growing number of studies focused on leveraging AI to help hospital pharmacists detect drug interactions during prescription reviews. The findings of our observational study validated the effectiveness of the AI tool in recognizing unusual dosage patterns and highlighted its detection quality. Of the notifications tracked during this research, 85% of the interventions were accepted, and in 82% of these cases, prescription dosages were adjusted as a result. These results suggest that the AI tool contributed to safer prescribing practices and improved the overall quality of patient care by increasing trust and reliability in the process. Overall, further studies and similar tools. |