Use of AI to Predict and Support Medication Adherence in Patients With Breast Cancer: Systematic Review.
Oral medications are commonly used in the treatment of breast cancer (BC), despite high rates of nonadherence. As adherence is fundamental for optimal treatment, finding ways to effectively improve it is important. Artificial intelligence (AI) is being widely applied to health care.
This review aims to offer an overview of the contribution of AI to medication nonadherence among patients with BC, suggesting future research directions, and highlighting existing gaps.
Four databases (PubMed, Embase, Scopus, and Web of Science) were searched. Studies were eligible if they used AI to predict, monitor, or support medication adherence in patients with BC, were published in English, and had full text available. Reviews, conference abstracts, editorials, and studies with mixed samples were excluded. This review was conducted following PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) guidelines and was registered in PROSPERO (International Prospective Register of Systematic Reviews) database (CRD42024587020). The risk of bias was evaluated using the PROBAST (Prediction Model Risk of Bias Assessment Tool) and the Downs and Black's methodological quality scale.
Ten studies were included in this review. Most of them (k=9) focused on developing machine learning models to predict medication nonadherence. These studies used a range of techniques, including logistic regression, artificial neural networks, and random forests. Model performance varied widely, with area under the receiver operating characteristic curve values ranging from 0.61 to 1.00. Predictors of nonadherence were clustered into 4 groups: clinical, disease, and treatment-related factors (eg, side effects and comorbidities); behavioral factors (eg, prior nonadherence); psychosocial factors (eg, quality of life and self-efficacy); and sociodemographic factors (eg, age and income). The only intervention study identified evaluated an AI-based chatbot and reported promising results, showing a 20% increase in adherence among participants who engaged with its reminder feature. All included studies were at high risk of bias, mainly due to the absence of model calibration or insufficient reporting, and their findings should therefore be interpreted with caution. A further limitation was the lack of attention to implementation: predictive accuracy alone is insufficient, and future work must also address actionability, safety, and cost-effectiveness to enable real clinical use. Progress in this area will require coordinated efforts among researchers, developers, clinicians, and policymakers to support the responsible development and implementation of these tools into routine care.
To our knowledge, this is the first systematic review of AI applications specifically targeting medication adherence in BC. It focuses on both predictive and interventional studies, mapping current AI applications within this specific clinical context. The findings highlight gaps in the implementation phase and emphasize the need for future research integrating a coordinated, multidisciplinary approach involving researchers, AI specialists, policymakers, and health care teams.
This review aims to offer an overview of the contribution of AI to medication nonadherence among patients with BC, suggesting future research directions, and highlighting existing gaps.
Four databases (PubMed, Embase, Scopus, and Web of Science) were searched. Studies were eligible if they used AI to predict, monitor, or support medication adherence in patients with BC, were published in English, and had full text available. Reviews, conference abstracts, editorials, and studies with mixed samples were excluded. This review was conducted following PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) guidelines and was registered in PROSPERO (International Prospective Register of Systematic Reviews) database (CRD42024587020). The risk of bias was evaluated using the PROBAST (Prediction Model Risk of Bias Assessment Tool) and the Downs and Black's methodological quality scale.
Ten studies were included in this review. Most of them (k=9) focused on developing machine learning models to predict medication nonadherence. These studies used a range of techniques, including logistic regression, artificial neural networks, and random forests. Model performance varied widely, with area under the receiver operating characteristic curve values ranging from 0.61 to 1.00. Predictors of nonadherence were clustered into 4 groups: clinical, disease, and treatment-related factors (eg, side effects and comorbidities); behavioral factors (eg, prior nonadherence); psychosocial factors (eg, quality of life and self-efficacy); and sociodemographic factors (eg, age and income). The only intervention study identified evaluated an AI-based chatbot and reported promising results, showing a 20% increase in adherence among participants who engaged with its reminder feature. All included studies were at high risk of bias, mainly due to the absence of model calibration or insufficient reporting, and their findings should therefore be interpreted with caution. A further limitation was the lack of attention to implementation: predictive accuracy alone is insufficient, and future work must also address actionability, safety, and cost-effectiveness to enable real clinical use. Progress in this area will require coordinated efforts among researchers, developers, clinicians, and policymakers to support the responsible development and implementation of these tools into routine care.
To our knowledge, this is the first systematic review of AI applications specifically targeting medication adherence in BC. It focuses on both predictive and interventional studies, mapping current AI applications within this specific clinical context. The findings highlight gaps in the implementation phase and emphasize the need for future research integrating a coordinated, multidisciplinary approach involving researchers, AI specialists, policymakers, and health care teams.
Authors
Pezzolato Pezzolato, Voskanyan Voskanyan, Cutica Cutica, Marzorati Marzorati, Pravettoni Pravettoni
View on Pubmed