Practical Exploration on Assisting Prescription Rationality Evaluation Based on DeepSeeK Large Language Model
10.3870/j.issn.1004-0781.2025.12.026
- VernacularTitle:DeepSeeK大语言模型辅助处方合理性评估的实用性
- Author:
Hongyan MA
1
;
Xiao ZHOU
1
;
Biyao ZHANG
1
;
Chongde SHEN
1
Author Information
1. 南京医科大学附属无锡人民医院药学部,无锡 214023
- Publication Type:Journal Article
- Keywords:
DeepSeek;
Large language model;
Prescription evaluation;
Prescription optimization;
Clinical decision support
- From:
Herald of Medicine
2025;44(12):2057-2061
- CountryChina
- Language:Chinese
-
Abstract:
Objective To explore the application of the DeepSeek large languagemodel in assessing prescription ration-ality and evaluating its implementation effectiveness.Methods A random sample of 5 099 outpatient and emergency prescrip-tion records from our hospital in a specific month of 2024 was collected.The DeepSeeK large language model employs vector data-base technology for prescription review.An evaluation metrics analysis was conducted between the DeepSeeK's reviews and phar-macist evaluations.Performance metrics including accuracy,precision,recall,specificity,and F1 were systematically calculated.Error pattern analysis encompassed false positive rate,false negative rate,and categorical comparison of discrepant prescriptions.Statistical deviation assessments were performed using x2 and Kappa coefficient to quantify inter-rater consistency.Results This study contains a total of 5 099 data,pharmacist group's review results are reasonable for 4 879 cases and unreasonable for 220 cases.In contrast,The Deepseek model's review results in the AI group are reasonable for 4904 cases and unreasonable for 195 cases.The evaluation accuracy rate of the DeepSeeK model is approximately 93.82%,the precision rate is approximately 25.64%,the recall rate is approximately 22.73%,the specificity is approximately 97.03%,and the F1 value of the model is ap-proximately 24.13%.Conclusion With rapid advancements in artificial intelligence,large language models like DeepSeek,leveraging their superior language comprehension and knowledge reasoning capabilities,can effectively assist pharmacists in pre-scription review,offering new technical support for clinical medication safety.
