Author: Animesh (Aashoo) Tandon
Infants with single ventricle heart disease have high morbidity and mortality in the interstage period, between stages I and II of palliation. The risk factors for interstage mortality are not well-understood, making predicting adverse events and designing effective interventions difficult. Currently, there are no methods of using real-time continuous physiologic data to predict adverse events, such as mortality or unplanned reinterventions, in the interstage period. There is, therefore, a critical need to collect and evaluate real-time continuous physiologic data in the interstage period. The long-term goal for this research program is to integrate the use of predictive analytic algorithms for real-time physiologic data into clinical decision-making in congenital heart disease across all time scales, in order to improve outcomes for interstage patients. The central hypothesis for this proposal is that wearables can yield useful data in infants with congenital heart disease, based on previous work in the neonatal intensive care setting. The rationale for this proposal is that if the use of real-time physiologic data shows predictive power, it could significantly change management for interstage patients in a myriad of ways. The central hypothesis will be tested through these two specific aims: 1) Show proof-of-concept regarding the use of a wearable pulse oximeter and heart rate monitor for interstage single ventricle patients, while at home, and 2) Identify vital sign patterns that signify an increased adverse event risk in interstage patients. For Aim 1, 15 patients will be enrolled locally, and a wearable continuous physiologic monitor will be applied. Parents will give feedback about whether the device was tolerated by the infant patient, as well as their own ease-of-use. The physiologic data obtained from the device will be evaluated for signal accuracy and usability for future algorithm development. For Aim 2, 50 patients will be supplied with the device for home use. Outcome data (EMS called; admission to the hospital; mortality; other interstage red flag events) and potential predictors such as oxygen saturation, heart rate, heart rate variability, and pulse oximetry signal strength patterns preceding events will be extracted from the data stream and from parent history. Machine learning techniques will be applied: first, training datasets for vital signs preceding adverse events and event-free times will be analyzed for patterns; then validation datasets will be used to test the derived algorithm. The primary analysis will use binary outcomes (with and without adverse events) with generalized linear or non-linear mixed models. The proposal is innovative because continuous physiologic data, and machine learning algorithms applied to those data, have not been collected for interstage patients at home. The proposal is significant because a model that can predict adverse events in interstage patients may fundamentally change morbidity and morbidity, as well as alter the clinical surveillance, for interstage patients. Ultimately, the knowledge gained from this proposal may be generalized as a paradigm for home monitoring for other high-risk populations.
Co Author/Co-Investigator Names/Professional Title: Jeffrey B. Anderson, MD, MPH, MBA; Associate Professor, Cincinnati Children's Hospital Medical Center Thomas M. Zellers, MD, MscMM; Professor, University of Texas Southwestern Medical Center