A Multiview Model for Detecting the Inappropriate Use of Prescription Medication: Machine Learning Approach

Abstract

Background: The inappropriate use of prescription medication has recently garnered worldwide attention, but most national policies do not effectively provide for early detection or timely intervention.

Objective: This study aimed to develop and assess the validity of a model that can detect the inappropriate use of prescription medication. This effort combines a multiview and topic matching method. The study also assessed the validity of this approach.

Methods: A multiview extension of the latent Dirichlet allocation algorithm for topic modeling was chosen to generate diagnosis-medication topics, with data obtained from the Chinese Monitoring Network for Rational Use of Drugs (CMNRUD) database. Topic mapping allowed for calculating the degree to which diagnoses and medications were similarly distributed and, by setting a threshold, for identifying prescription misuse. The Beijing Regional Prescription Review Database (BRPRD) database was used as the gold standard to assess the model's validity. We also conducted a sensitivity analysis using random samples of validated prescriptions and evaluated the model's performance.

Results: A total of 44 million prescriptions were used to generate topics using the diagnoses and medications from the CMNRUD database. A random sample (15,000 prescriptions) from the BRPRD was used for validation, and it was found that the model had a sensitivity of 81.8%, specificity of 47.4%, positive-predictive value of 14.5%, and negative-predictive value of 96.0%. The model showed superior stability under different sampling proportions.

Conclusions: A method that combines multiview topic modeling and topic matching can detect the inappropriate use of prescription medication. This model, which has mediocre specificity and moderate sensitivity, can be used as a primary screening tool and will likely complement and improve the process of manually reviewing prescriptions.

Keywords: inappropriate use of prescription medication; latent Dirichlet allocation; multiview learning; prescription review; topic model.

Figure 1

Graphical representation of the latent Dirichlet allocation model. K: number of topics; M: number of documents; N: number of words in each document; x: observed words in the document m; z: topic of nth word in a document m; θ: topic distribution for document m (document-topic distribution); φ: topic-word distribution; α: hyperparameter of θ; β: hyperparameter of φ.

Figure 2

Graphical representation of the multiview latent Dirichlet allocation model. K: number of topics; M: number of prescriptions; N^A: number of diagnoses per prescription; N^B: number of medications per prescription; x^A: diagnosis (type A feature); x^B: medication (type B feature); z^A: topic of x^A; z^B: topic of x^B; φ^A: topic-diagnosis distribution; φ^B: topic-medication distribution; β^A: hyperparameter of φ^A; β^B: hyperparameter of φ^B; θ: prescription-topic distribution; α: hyperparameter of θ.

Figure 3

Summary of the performance of multiview latent Dirichlet allocation model with TM detection methods under different thresholds. Horizontal axis: thresholds of TM methods (from 1 to 5). Vertical axis: percentage of SEN, SPE, PPV, and NPV. K: number of topics; NPV: negative-predictive value; PPV: positive-predictive value; SEN: sensitivity; SPE: specificity; TM: topic mapping.