Early Career
Status: Funded - Open
Summary
BACKGROUND: The wavelengths of light used by pulse oximeters overlap with the absorption spectrum of melanin resulting in extensive retrospective literature evaluating the overestimation of oxygenation by pulse oximetry in patients from races that may have darker skin tone. These inaccuracies are associated with higher rates of missed hypoxemia, insufficient supplemental oxygen, delayed intervention, and increased mortality. GAP: Previous studies relied on race and ethnicity as poor surrogates for skin tone. This has limited understanding of the problem and our ability to improve device performance. HYPOTHESIS: Aim 1: To determine the dynamic overestimation of saturation by pulse oximetry based on skin tone. We hypothesize that overestimation by pulse oximetry will vary over time based on dynamic patient-level factors allowing for modelling and improved accuracy of pulse oximetry. Aim 2: To determine the feasibility of using standardized photography to measure skin tone in place of high-cost spectrophotometers. We hypothesize that artificial intelligence will allow for accurate measurement of skin tone. METHODS: Patients admitted to the pediatric cardiac ICU will have spectrophotometric measurement of their skin as well as simultaneous oximetry and blood gas measurements at least daily over 72 hours. The primary outcome will be missed hypoxemia (SaO2<88% while SpO2≥92%). Additionally, standardized photography will be obtained to attempt to replicate spectrophotometry measurements using deep learning. RESULTS: Pending. IMPACT: Improved pulse oximeter performance through the inclusion of skin tone would lead to better diagnosis of hypoxemia. This would allow for more equitable oxygen supplementation across diverse patient care settings leading to better pediatric patient outcomes. Measurement of skin tone without a spectrophotometer would greatly increase the clinical availability of these measurements. Website Link: https://pediatrics.vumc.org/person/joseph-starnes-md-mph