Budget Constrained Machine Learning for Early Prediction of Adverse Outcomes for COVID-19 Patients

Abstract

Background: Machine learning (ML) based risk stratification models of Electronic Health records (EHR) data may help to optimize treatment of COVID-19 patients, but are often limited by their lack of clinical interpretability and cost of laboratory tests. We develop a ML based tool for predicting adverse outcomes based on EHR data to optimize clinical utility under a given cost structure. This cohort study was performed using deidentified EHR data from COVID-19 patients from ProMedica Healthcare in northwest Ohio and southeastern Michigan. Methods: We tested performance of various ML approaches for predicting either increasing ventilatory support or mortality and the set of model features under a budget constraint was optimized via exhaustive search across all combinations of features. Results: The optimal sets of features for predicting ventilation under any budget constraint included demographics and comorbidities (DCM), basic metabolic panel (BMP), D-dimer, lactate dehydrogenase (LDH), erythrocyte sedimentation rate (ESR), CRP, brain natriuretic peptide (BNP), and procalcitonin and for mortality included DCM, BMP, complete blood count, D-dimer, LDH, CRP, BNP, procalcitonin and ferritin. Conclusions: This study presents a quick, accurate and cost-effective method to evaluate risk of deterioration for patients with SARS-CoV-2 infection at the time of clinical evaluation.

Deep learning-assisted sensitive detection of fentanyl using a bubbling-microchip

“…AI technologies that emphasize development of learning models while taking cost constraints into account are loosely called "budget-sensitive" learning models. 8,9 While all fields of medicine can benefit from such approaches, it can be argued that point-of-care diagnostics (POC), which highly values costeffectiveness, tends to benefit the most. Furthermore, a budget sensitive model synergizes well with the more recent wave of POC diagnostics that have incorporated more "smart" elements in the assay workflow.…”

Dynamic modeling of hospitalized COVID-19 patients reveals disease state dependent risk factors

“…Understanding of risk factors influencing disease severity is critical for efficient clinical management of COVID-19 patients. Studies have shown that risk factors, such as obesity, sex, and age are highly correlated with adverse outcomes in COVID-19 patients [1][2][3][4][5][6] . Furthermore, recent studies suggest such risk factors also may affect certain aspects of COVID-19 progression, specifically disease onset 7 and time-to-death 8 .…”