JMIR Publications

ABSTRACT

Background:

Early detection of adverse events (AEs) and their management are crucial to improve anticancer treatment outcomes, and listening to patients’ subjective opinions (“patients’ voices”) can make a major contribution to improving safety management. Recent progress in deep leaning (DL) technologies has enabled various new approaches for the evaluation of safety-related events based on patient-generated text data, but few studies have focused on improvement of real-time safety monitoring for individual patients. In addition, no study has yet been performed to validate DL models for screening patients’ narratives for clinically important AE signals that would require medical intervention. In our previous work, novel DL models have been developed to detect AE signals for hand-foot syndrome (HFS) or AEs limiting patients’ daily lives (AE-L) from cancer patient-authored narratives, aiming ultimately to utilize them as safety monitoring support tools for individual patients.

Objective:

This study was designed to evaluate whether our DL models can screen clinically important AE signals that would require intervention by healthcare professionals. The applicability of our DL models to patients’ complaint data at pharmacies was also assessed.

Methods:

Pharmaceutical care records at community pharmacies were utilized for this evaluation of our DL models. The records followed the SOAP format, consisting of Subjective (S), Objective (O), Assessment (A), and Plan (P) columns. Because of the unique combination of patients’ complaints in the S column and the professional records of pharmacists, this was considered a suitable data source for the present purpose. Our DL models were applied to the S records of cancer patients, and the extracted AE signals were assessed in relation to medical actions and prescribed anticancer drugs.

Results:

From 30,784 S records of 2,479 patients with at least one prescription history of anticancer drugs, our DL models extracted true AE signals with more than 80% accuracy for both HFS and AE-L. The DL models were also able to screen AE signals requiring medical intervention by healthcare providers. The extracted AE signals could reflect side effects of anticancer drugs used by the patients, based on analysis of the types of prescribed anticancer drugs and the time-course relationship between AEs and drug administration. “Pain or numbness”, “Fever” and “Nausea” were common symptoms that each accounted for more than 20% of the extracted AE signals.

Conclusions:

Our DL models were able to screen clinically important AE signals that would require intervention to treat the symptoms. It was also confirmed that these DL models could be applied to patients’ subjective information recorded in pharmaceutical care records accumulated during pharmacists’ daily work.