Use of Artificial Intelligence in the Screening and Treatment of Chronic Diseases

Introduction

Chronic health diseases are physical or mental conditions that last more than one year and cause functional restrictions that require monitoring and treatment (Hajat and Stein, 2018). Patients with multiple chronic conditions experience frequent hospitalizations, die prematurely, require consultation with multiple specialists, and accrue high healthcare utilization rates (Navickas et al., 2016). They are also on multiple medications with complex drug interactions and have difficulty adhering to treatment that adversely impacts the quality of life and outcomes (Navickas et al., 2016). A significant proportion of the aging population (patients >65 years of age) is at higher risks of these chronic health diseases, with incidence impacting almost one-third of the world’s population and rates of patients experiencing greater than four chronic diseases at a time expected to double between 2015 and 2035 (Marengoni et al., 2011; Kingston et al., 2018).

As a result of its increasing prevalence, a 2015 Global Burden of Disease report correlates 70% of deaths to chronic disease complications (Global Burden of Disease Study 2015, 2016).

In addition to high morbidity and mortality, managing chronic disease complications results in high economic costs and significant productivity losses to communities. In 2016, chronic diseases cost the United States 1.1 trillion dollars and placed substantial burdens on health systems nationwide (The Costs of Chronic Disease in the U.S., 2019)- The World Health Organization (WHO) anticipates a shortage of approximately 12.9 million healthcare providers globally by 2035 (Kingston et al., 2018).

The complexities of chronic disease management, coupled with primary care physician time limitations and worldwide healthcare provider shortages, contribute to suboptimal treatment for patients (approximately 46.3% of patients with chronic diseases do not receive care per the recommended guidelines) (McGlynn et al., 2003). Further, current chronic disease management strategies are not equipped to address unique patient complexities.

The healthcare industry’s current infrastructure suffers from workforce shortages, underfunded systems, increases in demands, and an explosion of chronic disease prevalence and increased cost. Novel healthcare delivery models are urgently needed to address the complexities and multimorbidities that result from increasing chronic disease prevalence (Navickas et al., 2016). Al is a combination of promising technologies that may play a crucial role in future healthcare model success. This chapter will focus on Al’s potential role in successful healthcare delivery, its implications for three chronic health diseases (diabetes, cardiovascular disease or CVD, and dementia), the challenges and considerations for Al technologies in the healthcare environment, and its practical applications.

Role of Artificial Intelligence in Chronic Disease

Machine learning (ML) is a subset of Al that utilizes computer algorithms to analyze data and correlate relationships between risk factors (variates) and health conditions (result variables or covariates). ML models provide scalable solutions that may address the global burden of increased chronic disease prevalence. They may also be used to evaluate unstructured datasets and synthesize disease predictions to provide personalized and patient-specific healthcare recommendations. Contrary to traditional statistical modeling, ML processes potentially improve predictions and classifications because they are not dependent on preconceived variable assumptions. Healthcare systems already have key elements to enhance chronic healthcare delivery: electronic medical records (EMRs) with rich data, evidence-based guidelines, and infrastructures with computing power.

ML promises significant improvements to care compared to current risk score utilization. Physicians routinely use risk scores to make healthcare management decisions. However, these risk scores use limited numbers of clinical variables with modest predictive accuracy in individual subjects. ML leverages big data, recognizes complex patterns, and refines predictive accuracies via repetition and adjustment. Utilizing analytical data methods with patient data and evidence-based guidelines may allow healthcare providers to instantaneously tailor patient-specific recommendations to promote appropriate preventative and therapeutic personalized medicine strategies.

Personalized medicine’s primary objective is to utilize patient insights and features to derive patient-specific risk factors and subsequently customize treatment options based on the individualized risk profile (Redekop and Mladsi, 2013). ML models may play a vital role in personalized medicine by helping to merge diverse healthcare datasets, electronic health records (EHRs), imaging data, genetic data, biomarker analyses, and data from mobile sensors (Rumsfeld et al., 2016). Deploying ML systems may foster personalized medicine, predict the risk of disease/complications, and therefore recommend actionable interventions and prevent acute events. Technological advances offer healthcare providers with countless tools to fully understand their individual patients’ needs and provide the best possible care. The explosive use of digital health and mobile apps creates patient habit and activity datasets that can be merged with EMR data. Using these data sources, ML models may analyze real-time patient needs and guide healthcare providers to enhance the management of chronic diseases such as diabetes, CVD, and dementia.