RSNA Health Equity Committee/R&E Foundation
Emerging Issues Grant Recipient Fireside Chat
Project Abstracts
Emerging Issues Grant - Interventions/Solutions to Health Care Disparities
Farouk Dako, MD, MPH
University of PennsylvaniaCommunity Support Program for Lung Cancer Screening
Lung cancer screening (LCS) using low-dose CT (LDCT) for high-risk individuals is a critical public health tool for reducing lung cancer mortality. Despite the recent expansion of the inclusion criteria to be more equitable, uptake and adherence remains low and is lowest among low-income and racial/ethnic minority populations and in individuals with a negative baseline screening study (Lung-RADS 1&2). Community-based programs promoting cancer screening uptake have demonstrated increased screening engagement compared to clinic-based programs in reaching disadvantaged individuals but suffer limited data demonstrating impact on a population level. Community-based participatory research (CBPR) utilizes a partnership approach, bringing together researchers and community stakeholders as equal partners throughout the research process to contribute expertise and share in decision making. CBPR has demonstrated the ability to reduce cancer health disparities and reduce mistrust between academia and surrounding communities through reciprocal learning.
The overall objective of this project is to demonstrate the impact of a community support program (CSP) on adherence to LCS follow-up in an urban environment. It utilizes novel population level data that includes geospatial analysis of Philadelphia neighborhood-level data of patients who have received a LDCT for LCS. This allows for targeted community level interventions and the ability to evaluate impact on a population level. We will target individuals in our healthcare system residing in Philadelphia with a negative baseline screening CT and an upcoming or missed follow-up screening CT. Participants will be divide, based on residential location, into receiving clinic + community support vs routine clinic only (control). Primary analysis will measure the 12-month follow-up LDCT adherence rate difference between clinic + CSP vs clinic only group and its association with CSP enrollment. Planning and evaluation of our program will be performed using RE-AIM framework with emphasis on measuring reach, effectiveness, and sustainability
Judy Gichoya, MD
Emory UniversityDebiasing Image-Based AI Models for Population Health
The long-term goal of this project is to accelerate the translation of image based predictive models used for disease screening into clinical care through rigorous testing and external validation to mitigate bias while preserving model performance. Leveraging our extensive experience in developing diverse datasets and validating model bias, we propose to 1) Perform external validation of three image based screening models that have been shown to mitigate disparities - the Mirai breast cancer screening model, the Algorithmic prediction model for knee osteoarthritis and an algorithm that predicts atherosclerotic disease risk from chest Xrays (CXR); and 2) Develop and apply debiasing techniques to the classification models in Aim 1 and validate performance on the diverse Emory dataset. In aim 1, we will perform external validation of the three predictive models on a gold standard test set from Emory mammogram (EMBED), musculoskeletal (MURA)and Chest Xray dataset reporting on individual and subgroup model performance for different races and genders. We expect the model performance to drop when tested on the new Emory dataset which has more Black patients. To fix this performance drop, we will train an adversarial debiasing network which has two arms - a predictor for the disease condition/outcome and an adversary that predicts a protected variable (race or gender). This network penalizes the model for learning the protected features. Model performance will be evaluated across stratified cohorts after debiasing and model bias evaluated using the True Positive Disparity (TPR) with a disparity measure of between 0.8 and 1.25.
If successful, this project will contribute additional evidence and guidelines for model testing and validation while promoting fairness and mitigating biases to accelerate translation of these image-based models into the clinic. We will provide bias mitigation tools for research by sharing our models and comprehensive test subsets of our diverse data.
Soterios Gyftopoulos, MD
NYU LangoneUnderstanding and Improving Osteoporosis Screening in U.S. Asian Populations: A Mixed Method Analysis Based Education Initiative
Osteoporosis is a common condition that predisposes to increased fracture risk, morbidity, and mortality. The majority of osteoporosis studies have focused on non-Hispanic white women. However, epidemiological findings have shown that Asian women and men have the highest age-adjusted prevalence of osteoporosis in the U.S. In addition, the term “Asian” represents a large, heterogeneous population representing more than 20 ethnicities, and the U.S .census projects that, as the fastest growing race group in the US, this racial group will encompass 46 million citizens by 2060. Limited studies have also shown differences in bone density screening when the dominant Asian origin groups are evaluated separately. A true understanding of osteoporosis in the U.S. Asian population requires more information in the context of the individual ethnicities. Given the potential consequences of osteoporosis-related fractures, osteoporosis screening is an important yet underutilized tool that can identify Asian patients most at risk for fracture and can allow for the timely administration of preventative medications and lifestyle changes. There has been limited work done on osteoporosis screening in U.S. Asian populations and no evidence-based programs to increase screening in targeted Asian populations. Identifying patients at greatest risk of missing osteoporosis screening is an important first step in increasing utilization rates as it would allow organizations to focus resources on reaching and recruiting Asian populations that need improved bone healthcare the most. Coupling these data with an understanding of the facilitators and barriers to screening for these groups will provide a comprehensive strategy to improve the preventative care provided to those in these at-risk patient populations. We will use a culturally adaption framework and health literacy approach to translate findings to inform the development of resources and materials to improve the preventive care provided to these populations.
Christina Lebedis, MD
Boston University Medical CenterAddressing Disparities in Hepatocellular Carcinoma (HCC) Surveillance Using Implementation Strategies
The overall goal is to implement an innovative evidence-informed intervention to improve hepatocellular carcinoma (HCC) surveillance in at-risk patients and reduce the disproportionate toll of HCC in minorities and socially disadvantaged populations. HCC surveillance by imaging improves HCC outcomes and is recommended by all major societies. HCC disproportionately affects marginalized populations and poor rates of surveillance widens disparities in outcomes. The commonest reason for poor surveillance is underuse of HCC surveillance by providers, making it the highest yield target for improvement. We have developed the Patient Navigation for HCC (PN-HCC) surveillance intervention, an evidence-informed multi-strategy intervention, that combines patient navigation and dashboarding, to improve HCC surveillance rates. Our program is in line with RSNA R&E Foundation’s program announcement which seeks ‘new exploratory intervention to mitigate the negative impact of healthcare disparity.’ A key exploratory strategy in the intervention is the change in site of initiation of surveillance from provider outpatient clinics to department of radiology, where HCC surveillance will be operationalized by a patient navigator and electronic health record dashboard. To implement this study, we will be leveraging excellent resources at Boston Medical Center, New England area’s largest safety net hospital, where ~57% population are marginalized and there is a high burden of liver disease. This is a multidisciplinary collaboration with radiology (LeBedis) and gastroenterology (Mohanty) with input from the Center for Implementation and Improvement Sciences at Boston Medical Center. Our project aims are: Specific Aim 1: Implement the PN-HCC surveillance intervention in GI/radiology and evaluate implementation outcomes using the RE-AIM framework (reach, adoption, effectiveness, implementation, maintenance) with a focus on health inequities. Specific aim 2: Assess the barriers and facilitators of the PN-HCC Surveillance intervention in primary care by semi-structured interviews with key stakeholders from BMC and affiliated community health centers, guided by the Consolidated Framework for Implementation Research.
Bhavika K. Patel, MD
Mayo Clinic ArizonaSustainability of a Pilot Ride Share Intervention: Creating a Plan To Enable Large-Scale Inclusive Patient Access to Preventative Screening
Screening and preventative disparities are especially concerning in the setting of breast cancer, the most common cancer for women where survivability is high with early detection and screening. Hispanic women are more likely to be diagnosed with breast cancer at a later stage. Structural barriers, such as lack of transportation, are commonly reported barriers to breast cancer screening. In collaboration with community health partners in Maricopa County, we have created a pilot ride-share program to enhance access to mammograms by inviting uninsured, Hispanic women to participate in the screening as part of a national trial. Failure to sustain the ride-share program will likely reduce observed benefits, making widespread adoption and scale-up to other populations and clinical trials unlikely. Sustainability refers to the structures and processes that facilitate the ability to maintain programs and their benefits over time. The objective of this study is to evaluate and plan for the sustainability of a no-cost ride-share program to reduce structural barriers to access and use of breast cancer screening and improve clinical trial retention. Aim 1. Explore structures and processes that hinder or facilitate sustainability of a no-cost ride share program for uninsured, Hispanic patients in Maricopa County and identify potential strategies to support sustainability. Aim 2. Examine the role of the no-cost ride-share program on key sustainability outcomes - reach and maintenance including estimated cost and trial retention. Exploratory Aim 2. Explore differences in sustainability outcomes by patient characteristics This study will inform a plan for future sustainability and potential scale-up of transportation-related strategies to alleviate healthcare disparities in our surrounding communities. The long-term goal is to enable widespread implementation of a no-cost ride share program to reduce structural barriers to access to screening modalities, in addition to clinical trial recruitment and retention, among historically underserved, minority populations.
Gelareh Sadigh, MD
University of California, IrvineEmpowering Latinx patients Lung Cancer Screening Uptake (Empower Latinx)
Lung cancer is the leading cause of cancer-related mortality, with Hispanic patients having lower survival rates than non-Hispanic Whites. Lung cancer screening (LCS) with low-dose computed tomography (LCDT) decreases the mortality rate of lung cancer by 20%, when compared with chest X-ray alone, and has been covered by both private and public health insurance since 2015 for eligible populations. However, adherence to LCS remains low. Hispanic patients are reported to have a lower adherence rate to LCS compared to non-Hispanic Whites, highlighting the need to increase uptake of LCS among Hispanic communities. Our proposed intervention, Empower Latinx, informed by the Health Belief Model, is a multi-level intervention including (1) Primary care provider (PCP) notifications of patients’ LCS eligibility and self-reported barriers to receiving LCS; (2) culturally tailored patients’ LCS education; (3) patients’ referral to financial navigation resources; and (4) patients’ reminder to discuss LCS during PCP visit. We propose to conduct a pilot feasibility patient-level randomized trial of 60 Hispanic LCS-eligible English or Spanish patients with a scheduled PCP visit at four primary care clinics including two Federally Qualified Health Centers (FQHC) to compare Empower Latinx vs. usual care enhanced by brief patient LCS education. We will assess barriers to LCS among Hispanic patients who are LCS-eligible (Aim 1) and test the feasibility of our intervention among enrolled patients (Aim 2). The primary endpoint is the order of LDCT within 4 months assessed with self-reported patient surveys and EMR data extraction. In Aim 3, we will describe experiences with the intervention by conducting semi-structured qualitative interviews with 10 patients and 5 providers. Our study will assess barriers to LCS in the Hispanic population and provide actionable results to refine our intervention for future testing in a multi-institutional larger randomized controlled trial to improve health outcomes among the Hispanic population.