Forecasting Patient No-show and long-term Health Trajectory using Bayesian Analysis
Bengt Arnetz, Department of Family Medicine, College of Human Medicine, Michigan State University, USA
Background: The healthcare cost explosion and quality deficiencies represent major challenges to the United States’ financial, social, and population health. By some estimates, 30% of healthcare resources are wasted. Another challenge is to develop a better understanding of the long-term trajectory of patients with chronic diseases, e.g., hypertension, diabetes, and metabolic syndrome. Clinical questions that beg better answers include: Which patients are most likely to have the worst health trajectory, and therefore require intensive medical interventions? Which patients will manage fine with only minimal interventions? What are the optimal mixture of various treatment strategies, e.g., pharmaceutical, environmental, and/or behavioral for specific patient groups? Another challenge to healthcare systems is patients that for various reasons do not show-up for their medical appointments. This results in underutilization of sparse resources and, possibly, worse patient health outcomes. Approach: Bayesian analysis that incorporates close cooperation between statisticians and clinicians represents a promising new approach to address such challenges. It does not suffice to merely apply machine learning and artificial intelligence. By involving clinicians, specific models can be developed and tested in cooperation with the statisticians. Through an interactive, self-corrective, and reiterative model developing process, the end-result will assist clinicians and healthcare systems to apply an evidence-based strategy to improve patient outcomes, processes, and resource utilization. A team of Bayesian statisticians, health systems researchers, and clinicians are in the process of applying this interactive and reiterative process to determine whether chronic disease patients, whom use around 80% of the healthcare resources, can be managed more efficiently. Furthermore, can clinics, especially those serving underserved and vulnerable patients, better predict no-shows and implement policies that facilitate patients’ ability to attend medical appointments? Conclusion: Healthcare costs represent almost 20% of the United States gross national product, the highest in the world. However, based on outcomes, the United States are not even ranked among the top 10. By forging a closer collaboration with key stakeholders, it is likely that we will achieve a marked improvement in patient outcomes, at the same time as the pressure on an already strained healthcare system is decreased.