Accepted for/Published in: JMIR Medical Informatics
Date Submitted: Sep 18, 2019
Date Accepted: Mar 24, 2020
Date Submitted to PubMed: Mar 26, 2020
Warning: This is an author submission that is not peer-reviewed or edited. Preprints - unless they show as "accepted" - should not be relied on to guide clinical practice or health-related behavior and should not be reported in news media as established information.
Development and Validation of Multi-view Inappropriate Medication Use Prescription Detection Model: Machine Learning Study
ABSTRACT
Background:
Inappropriate use of medicines remains a universal challenge for medical practice. Although increasing attention has been paid to inappropriate prescription medication use, the current policies do not meet the demand for early detection and timely intervention.
Objective:
Develop an inappropriate medication use prescription detection model using multi-view topic modeling method combined with the topic matching method, and assess the validity of the model.
Methods:
A multi-view extension of the latent Dirichlet allocation topic modeling algorithm was chosen to generate the diagnosis-medication topic models indicating the underlying health status of patients using diagnosis and medication as variables, with data from the Chinese Monitoring Network for Rational Use of Drugs (CMNRUD) database. Each topic consisted of a set of diagnoses, medications that were highly related to each topic, and the percentages for each of these. Topic mapping was used to calculate the similarities of distribution of the diagnoses and medications on the topics built in the previous step and find the inappropriate medication use prescriptions by setting a threshold. The prescription manual review result by experts in the Beijing Regional Prescription Review (BRPR) database was used as the gold standard to assess the validity of the model. We also conducted a sensitivity analysis by randomly sampling the validation prescriptions and compared the model performance.
Results:
A total of 44 million prescriptions was used to generate topics using the diagnoses and medications from the CMNRUD database. A 15,000 random samples of the BRPR database were used for validation, and the model had a sensitivity of 81.8% and a specificity of 47.4%, and the positive predictive, negative predictive values were 14.5%, 96.0%, respectively. The model showed preferable stability under different sampling proportion.
Conclusions:
Multi-view topic modeling method with the topic matching method can detect inappropriate medication use prescriptions combining. Considering the mediocre specificity, with moderate sensitivity, this model can be used as a primary screening tool and will likely complement and improve the manual review.
Citation
The author of this paper has made a PDF available, but requires the user to login, or create an account.
Copyright
© The authors. All rights reserved. This is a privileged document currently under peer-review/community review (or an accepted/rejected manuscript). Authors have provided JMIR Publications with an exclusive license to publish this preprint on it's website for review and ahead-of-print citation purposes only. While the final peer-reviewed paper may be licensed under a cc-by license on publication, at this stage authors and publisher expressively prohibit redistribution of this draft paper other than for review purposes.