Artificial intelligence-based approaches for monitoring medication adherence among cardiovascular disease patients: a scoping review

Abstract

Cardiovascular diseases (CVDs) are still a leading cause of death worldwide, and the impact of the disease in lower-income countries is dire. Patients must take medications on a regular basis for effective management of CVD; however, the level of efficacious self-management is far below what is required. Artificial Intelligence (AI) shows promise in evaluating different data sets, aiding in clinical decision processes, and facilitating active engagement in the self-management of medication. The purpose of this scoping review was to analyse the available literature on the approaches that have been employed in using AI to improve medication adherence in patients with CVD. This research was done according to the PRISMA-ScR guidelines. A comprehensive literature search was conducted on June 19, 2025, across PubMed, Embase, CINAHL, Scopus, and Web of Science using pre-established eligibility standards. Screening was done using Rayyan across two stages. Data was extracted on the characteristics of the study, AI-based approaches and their methods, measures of adherence, and the outcomes reported. Sixteen studies were included based on the inclusion criteria. The included studies varied in setting, sample size, and disease conditions. AI-based approaches comprised natural language processing, reinforcement learning, and machine learning algorithms such as neural networks, random forests, support vector machines, and decision tree models. Interventions were delivered through smartphone applications, clinical workflow systems, Internet of Things (IoT) enabled devices, and chatbots. Medication adherence was assessed using pharmacy refill data, electronic health records (EHR), validated surveys, device-based monitoring, administrative claims, and biochemical testing. Several AI-based approaches showed improvements in adherence; however, across studies, reports were inconsistent, and methodological variation limited comparability. Evidence from low- and middle-income countries remained limited. AI-based approaches have been applied to monitor and support medication adherence among patients with CVD; however, the outcome measures are inconsistent due to limited population diversity and variable study quality, restricting the comparability across existing evidence. Standardized adherence metrics, rigorous methodologies, and broader evaluations in diverse settings are essential to strengthen future research and support real-world decisions.

Ähnliche Arbeiten