Sumanth Swaminathan and James Morrill presented Vironix Health software at the 2021 SIAM Annual Meeting - July 19th to 23rd. The topic of the presentation was: Machine Learning for Triage Recommendations. The organizer of the presentation from SIAM was Manuchehr Aminian from California State Polytechnic University. The entire abstract is featured on SIAM latest news blog - Click Here
There is major interest in the benefits of high-frequency monitoring and analysis of time series relating to health and biology. Much of this comes from the realm of wearable devices, where common questions arise, such as inferring the health state of individuals. Similar questions may arise in traditional monitoring approaches, but applying modern time series analysis and machine learning techniques.
Viral and chronic respiratory illnesses are characterized by rapid health deterioration episodes that manifest as either emergency care events (in the case of chronic lung disease) or rapid infection spread (in the case of Covid-19 and Influenza). In this study, we present an approach to identifying clinically significant flare-ups of respiratory illness using globally available clinical characteristic data and physician provided triage labels. The methodology leverages Bayesian Inference and Monte Carlo methods to generate hypothetical patient scenarios for training triage algorithms. Machine-learning prediction models are evaluated in out-of-sample validation tests of accuracy, sensitivity, specificity, ppv, and npv when detecting health deterioration and recommending the most appropriate responsive care. Algorithm recommendation and exacerbation identification time series data from real patient trials is then used to risk stratify chronic lung patients for future health downturns using signature methods coupled with classical machine-learning.