Precision Assessment Algorithm for Reducing Disaster-related Respiratory Health Disparities

点击翻译按钮获取中文摘要

ABSTRACT Weather and climate disasters are responsible for over 13,000 deaths and $1.7 trillion additional costs over the last 40 years in the USA. Older adults are particularly susceptible to respiratory symptoms, disease exacerbation, unscheduled health care utilization, and decreased quality of life after disaster exposure to particulates, mold, and flooding. Our preliminary data reveal those with Black racial identities possess fewer resources to prepare for disaster. Profound racial disparities observed in the COVID-19 pandemic illustrate the devastating sequelae of long-standing macro-level disparities of segregated housing and sociopolitical networks. The long-term goal of this work is to eliminate racial disparities in large scale disaster health outcomes. The short-term goal of this research is to identify pathways of equal opportunity and disaster affirmative action interventions. The objective here is to create a software prototype of a machine learning algorithm with a novel, valid and reliable assessment tool of disaster vulnerability for older adults with chronic obstructive respiratory disease, prioritizing equality of opportunity to reduce racial bias and disparity. Our specific aims are to 1) Empirically validate a novel assessment tool of disaster vulnerability using self-reported items and scoring system, 2) Refine the validated instrument with a machine learning based algorithm for precision prediction of household emergency preparedness for disaster, 3) Assess racial disparities, data and algorithm bias for Black participants in household hazard vulnerabilities and our instrument development process, and 4) Test interoperability with existing customer software platforms as a plug-in software add-on. We will accomplish these aims using a mixed-methods approach, recruiting 20 expert panel members and up to 600 potential end-user participants, working to over-sample those who reside in predominantly Black communities and Black racial identities. The knowledge gained from this study will provide foundational work to develop precision interventions to reduce post-disaster respiratory symptoms, disease exacerbation, unscheduled health care utilization, and decrements in respiratory quality of life. The results of this study will inform the next generation of electronic health record and patient reported outcomes applications, ensuring the validity, prognostic accuracy, and machine learning models are most relevant to those at highest risk for racial disparities: those with Black racial identities. Health care providers can use our software tool to target and optimize disaster telehealth service lines and increase intervention precision to reduce disaster morbidity and mortality disparities.