High nonadherence in the US leads to avoidable mortality.
Eric Pulice, MPA, DBA candidate, is an assistant professor of health care management, College of Business and Aviation, Fairmont State University in Fairmont, West Virginia.
Alberto Coustasse, MD, DrPH, MBA, MPH, is a professor of health care management and administration, Lewis College of Business, Marshall University in Huntington, West Virginia.
A study conducted in 2019 showed that approximately 50% of the 187 million patients in the US health care system did not follow their medication plan as prescribed.1 This means they failed to adhere to the drug regimen or to take the medication for the entire prescribed duration. Adherence rates for most medications used to treat chronic conditions, such as diabetes and hypertension, usually fall in the range of 50% to 60%, even with patients who have good insurance and drug benefits.2 This rate of adherence leads to an estimated 125,000 avoidable deaths each year and $100 billion annually in preventable health care costs.1 The World Health Organization identified economic, social, health care, patient, provider, and therapy-related factors for nonadherence.3 However, medication nonadherence is not considered a significant issue by most practicing physicians.4,5
Traditional methods, such as pill counts, have been unreliable in accurately measuring patient medication adherence.6 Current technologies addressing medication adherence include electronic pill bottles, ingestible sensors, and video-based systems. These technologies offer diverse ways to monitor medication adherence with expectations of accuracy and increased patient adoption because of a lack of obtrusiveness for patients; however, successful clinical implementation remains infrequently documented. Further, telephonic-based e-health interventions to promote medication adherence have been implemented with mixed results.4,7 Nevertheless, the global digital medication adherence market is projected to reach $6.8 billion by 2026, indicating substantial investment and potential for growth.4
The inverse correlation between medication adherence and quality of life underscores the importance of adherence in improving overall patient well-being.8 Nonadherent patients have been found to have a higher risk of major cardiovascular events, emphasizing the impact of adherence on disease-specific outcomes.9,10 Nonadherence has also been linked to potentially inappropriate increases in treatment intensity, indicating the broader impact on health care utilization and patient safety.11 The absence of successful interventions to enhance medication adherence, coupled with the crucial role of adherence in achieving better patient outcomes, underscores the necessity to investigate alternative approaches.
Artificial intelligence (AI) and machine learning are valuable tools for predicting medication adherence for various conditions, such as diabetes and hypertension, for which AI has been shown to achieve 70% to 80% accuracy in identifying nonadherent individuals.12 To date, the FDA has approved over 60 AI-equipped medical devices, indicating a substantial trend in the integration of AI in health care.13 Machine learning, a subset of AI, can analyze large, intricate data sets using sophisticated algorithms and techniques, extracting valuable insights and patterns.14,15 Machine learning has been effectively utilized to predict medication adherence in various medical conditions, such as opioid use disorder, type 2 diabetes, an hypertension.12,16,17 Moreover, machine learning techniques have been employed to forecast the likelihood of nonadherence in individuals with type 2 diabetes, demonstrating the potential to identify patients susceptible to suboptimal medication adherence.18
AI has also been employed to develop tools for providing comprehensive assessments of patients’ adherence behaviors and psychological factors influencing adherence levels. For instance, AI-based tools have been designed to evaluate self-reported adherence to medication, offering insights into patients’ adherence behaviors and barriers.19,20 These interventions were customized to meet the individual needs of patients and could include motivational messages, behavioral interventions, and psychological support for individuals with chronic conditions, such as affective disorders and schizophrenia.21
Despite its potential, integrating AI into medication adherence faces challenges. For example, health care professionals often face limitations in general knowledge and awareness of available technological-based solutions for medication adherence.22,23 Moreover, variance exists in knowledge and perspectives regarding the utilization of AI among physicians and medical students. These differences can result in differing levels of acceptance and understanding of the value of AI technologies when utilized in health care.24
However, there is a growing understanding that medical education must play a formal role in the adoption and use of AI technologies. Specifically, medical education should, going forward, involve teaching future physicians about the applications and limitations of using AI tools in medical practice.25 Additionally, as a part of that education process, ongoing discussions around the ethical, legal, and societal concerns regarding AI implementation and use will continue to be important.
