The Chronic Rhinosinusitis (CRS) Problem and an AI Driven Personalized Medicine Solution - SUMMARY: Chronic rhinosinusitis (CRS) is a highly prevalent disease that inflicts a severe quality-of-life (QOL) impairment, with local symptoms of nasal blockage, congestion, loss of smell, nasal drainage, headache, and recurrent infections. CRS is also appreciated for its prominent role as a compounding factor in other respiratory diseases such as allergic rhinitis, asthma/COPD, and cystic fibrosis. General health concerns including fatigue, sleep dysfunction, anxiety/depression, excessive medication usage, frequent physician visits, and productivity loss often accompany nasal symptoms and contribute to overall health detriment. As a result of local symptom burden and overall health disturbance, CRS has been ranked among the most severe health ailments, alongside congestive heart failure and end-stage renal disease. Among existing therapeutic options for CRS, outcomes vary widely. The stepwise approach to CRS management results in slow and measured escalation of care only after repeated attempts at various medical therapies. As nuanced understanding of disease trajectories are uncovered through deep phenotyping of individuals with CRS, a precision approach to this chronic disease is on the horizon. A patient’s decision to commit to aggressive CRS treatment, such as surgery, includes many personal factors but inherently relies on the likelihood for symptom improvement. Not surprisingly, the degree of symptom improvement gained is the primary contributor to patient satisfaction with the intervention. Though numerous risk factors have been associated with symptom severity and outcomes in CRS, there is a notable void in translation of these observations to real-world, clinical decision-making. This is due to the many possible predictor variables, unclear disease natural history, and weakness of traditional statistical approaches in dealing with noisy data. Indeed, weighing these many considerations is a complex clinical task. Our long-term goal is to create a personalized predictive model that can help individual patients understand their likely outcomes with available therapeutic options. Translating research observations into a comprehensible platform for shared decision-making, precision medicine, and rational allocation of system resources (i.e., cost-benefit analyses) would be truly innovative in CRS. Our central hypothesis is that machine learning and artificial intelligence approaches can be applied to myriad demographic and clinical data to infer clinical outcomes for individuals deciding when to pursue surgical treatment for CRS. The proposed research is significant because of its potential to: (1) illuminate how disease manifestations occur in some but not all CRS patients, (2) predict patient trajectories and inform clinical interventions, (3) support real-time clinical decision making, (4) develop new patient-centered treatment algorithms, and (5) serve as an initial study of predictive analytics informing the decision for elective surgical therapies in chronic disease.
