JMIR Publications

ABSTRACT

Background:

Many patients with chronic diseases, especially those with multiple diseases, may need to take a variety of drugs every day to control their condition1, which may lead to a decline in patient medication compliance. The decline of compliance is often manifested in the behavior of taking medicine not according to the prescribed time, not taking medicine or interrupting taking medicine. This has a negative impact on the therapeutic effect, resulting in varying degrees of maximal consequences and an increased risk of complications.

Objective:

In order to solve the problems of poor multiple medication adherence and difficult medication scheduling in the elderly, an intelligent medication scheduling system for patient medication reminder was established, and its performance was evaluated through testing and compared with the artificial intelligence GPT-4.

Methods:

Medication time constraint with one drug (MTCOD) data were designed by integrating the physicochemical properties of the drug, dining constraints, and chronopharmacology, and drug interactions (DI) were considered to design the medication time constraint with multi-drug (MTCMD) data. Combining the above elements and taking into account the patient's situation and their needs, an optimal medication scheduling program was developed for them. Based on the above rules, a universal medication schedule (UMS) system is established, including a complete drug information database and a medication reminder system. The UMS system and GPT-4 were used to test the prescriptions of 20 elderly patients with chronic diseases, and 12 experts of clinicians, pharmacists or nurses were invited to evaluate the medication time of the system in terms of accuracy, safety, adherence, and usefulness using a Likert 5-level scale.

Results:

The drug database has been constructed and loaded with 1,926 items of basic drug information and 350,351 items of DI data, and 2,229 items of MTCOD and 87,248 items of MTCMD have been originally created. The medication reminder system has been constructed and clinically tested to generate a universal medication schedule, which can be used by medical staff to develop personalized medication schedules for patients. The expert evaluation results showed that the UMS group had high accuracy and safety scores, which were significantly better than the performance of the GPT-4 (P<0.001) and close to the results of professionals.

Conclusions:

Universal medication scheduling system provides data support for intelligent medication reminders, automatic identification of drug interactions, and monitoring system, which is more accurate and safer compared with GPT-4, and can be used as a powerful supplement to the clinical work of pharmacists.