Development and Validation of Correction Equations to Enhance Quality of Infrared Technology

About This Grant

The goal of this proposal is to investigate the contributors to infrared medical device errors and to develop correction algorithms to improve the accuracy of pulse oximeters and temporal thermometers in patients with acute respiratory failure. Over one million adults are admitted to intensive care units (ICUs) in the US every year for acute respiratory failure requiring mechanical ventilation. In these patients, it is essential to have accurate oxygen saturation and temperature measurements. Pulse oximeters and temporal thermometers, which measure oxygen saturation and body temperature, are the two most ubiquitous infrared devices in the ICU. However, these devices are significantly less accurate in Black patients, leading to delays in recognition and treatment of life-threatening hypoxemia and infection. There remains a gap in knowledge about the factors that contribute to racial disparities in infrared device accuracy. While it is hypothesized that infrared devices perform differently across racial groups due to differences in skin melanin, this hypothesis has not been rigorously tested in prior studies. Further, there are clinical factors such as anemia and chronic lung disease that could also affect the accuracy of these devices. There are known racial disparities in comorbidities such as anemia and chronic lung disease, which could then mediate the inaccuracy observed in infrared devices. In this proposal, we will quantify the role of melanin in association with infrared device errors in a multi-center, racially diverse cohort of critically ill patients with acute respiratory failure (Aim 1); we will use structural equation modeling to develop innovative correction algorithms that incorporate melanin, bilirubin, dyshemoglobins, age, sex, perfusion, and comorbidities to correct device errors (Aim 2); we will validate the correction algorithms and evaluate the clinical impact of the algorithms in accurately identifying clinically significant measurements (Aim 3). Through this proposal, we will develop an innovative publicly available de-identified dataset of critically ill patients enrolled across 9 ICUs in 3 hospitals with quantified melanin levels integrated with electronic health record (EHR) data that will enable researchers to investigate inaccuracies in other medical devices. Successful completion of this grant will significantly enhance our scientific knowledge of how individual patient factors influence the accuracy of infrared medical devices in real-world clinical settings. Although device development is often seen as the primary solution to addressing infrared device errors, it takes decades, is expensive, and may not comprehensively correct for all the complexities of physiology in critical illness. The ultimate public health impact of this work is developing EHR-based correction algorithms (Device+) as a complementary solution to device development. The Device+ solution will significantly impact clinical practice by improving the accuracy of vital sign measurements, which will lead to improved population-level as well as individual-level outcomes in critically ill patients.