With support from the Gordon and Betty Moore Foundation, the Kaiser Permanente Division of Research (KP-DOR) Augmented Intelligence in Medicine and Healthcare Initiative (AIM-HI) has selected five proposals that evaluate the implementation of existing Artificial Intelligence/Machine Learning algorithms that enhance diagnostic decision-making to achieve the following objective:
To advance research methods, identify best practices for scalability, and build capacity for effectively implementing and rigorously evaluating the use of AI/ML algorithms for diagnostic decision-making in real-world settings.
Precision Resuscitation with Crystalloids in Sepsis (PRECISE trial)
Primary PI: Sivausbramanium Bhavani, MD, MS, Emory University
Co-investigators: Andre Holder, MD, MS, Rishikesan Kamaleswaran, PhD, Colleen Kraft, MD, MS, Kirk Easley, PhD, David Murphy, MD, PhD, Nicole Franks, MD, David Wright, MD, Greg Martin, MD, MS, Craig Coopersmith, MD, Emory University
The Precision Resuscitation with Crystalloids in Sepsis (PRECISE) trial is a pragmatic, single-blinded, patient-level randomized controlled trial that will be conducted across the Emory Healthcare system, including tertiary academic, hybrid academic-community, and community hospitals. In the PRECISE trial, the team will apply their machine learning algorithm to identify a subset (Group D) of patients and randomize them to usual care or intervention, an electronic health record (EHR)-alert that nudges clinicians to use balanced crystalloids instead of normal saline, to evaluate a potential reduction in a 30-day mortality rate.
Advancing Novel Approaches and Best Practices for Effective AI-Enabled Diagnosis using Randomized Trials, Algorithmovigilance, and Proactive Risk Assessment
Primary PI: Peter J. Embí, MD, MS, Vanderbilt University Medical Center
Co-investigators: Laurie L Novak, PhD, MHSA, Megan Salwei, PhD, MS, Colin Walsh, MD, MA, Shannon Walker, MD, Allison P. Wheeler, MD, MSCI, Amanda Mixon, MD, MSPH, Benjamin C. French, PhD, MS, Adam Wright, PHD, Bradley Malin, PhD, Sharon Davis, PhD, MS, Michael E Matheny, MD, MPH, Vanderbilt University Medical Center
This proposal plans to evaluate hospital-acquired venous thromboembolism (HA-VTE)-specific AI-driven clinical decision support (AI-CDS) systems across rural and urban sites, and adult and pediatric patients. The project team also plans to study and refine novel approaches for real-world algorithmovigilance, going beyond to focus on real-time and sustained monitoring of AI tools that are deployed and actively used in the health care system.
Diabetic Retinopathy Screening Point-of-Care Artificial Intelligence (DRES-POCAI)
Primary PI: Fatima Munoz, MD, MPH, Centro de Salud de la Comunidad de San Ysidro, Inc. dba: San Ysidro Health
Co-PI: Sonia Tucker, MD, MBA; Co-investigators: Edgar Diaz-Pardo, MD, Oliver Solis, OD, San Ysidro Health; Nicole Stadnick, PhD, Marva Seifert, PhD, University of California, San Diego (UCSD); Chaithanya Ramachandra, PhD, Malavika Bhaskaranand, PhD, Eyenuk, Inc.
San Ysidro Health (SYH), a Federally Qualified Health Center (FQHC) serving over 121,000 San Diego County resident from medically underserved minority communities that are low-income, underinsured, with high rates of diabetes mellitus (DM) and low rates of diabetic retinopathy (DR) screening (54%), has identified an innovative method to address unmet medical needs in diabetic eye care by enhancing and modifying existing clinical practices through integrating point-of-care (POC) artificial intelligence (AI) technology for DR screening. The DRES – POCAI intervention will be integrated into SYH’s electronic health records (EHR) and increase access to DR screening within the primary care setting. This new model will improve quality performance levels, facilitate clinical decision-making process, accelerate timely identification of DR, improve linkage to DM guideline-concordant care, educate/engage patients, and therefore leading to an improved overall management of patient eye health.
High-Throughput Precision Identification of Cardiac Amyloidosis in a Diverse Population
Primary PI: David Ouyang, MD, the Smidt Heart Institute at Cedars-Sinai
Co-investigators: Paul Cheng, MD, PhD, Palo Alto Veterans Affairs Medical Center; Faraz Ahmad, MD, MS, Northwestern Medicine; Jacob Abraham, MD, Providence Heart Institute
The proposal evaluates the implementation of a machine-learning based opportunistic screening algorithm to help detect an underdiagnosed cardiac amyloidosis (CA), a rare fatal disease especially underdiagnosed among Black patients and the elderly. This project will apply a previously validated and published algorithm to existing echocardiogram images in an automated screening workflow to identify those at highest suspicion for cardiac amyloidosis for downstream evaluation by specialists. In addition to deploying the AI tool, the project will evaluate if the screening program is effective at identifying patients with cardiac amyloidosis, improves health outcomes, and is embraced by clinicians in the health system.
Generalizing An AI/ML Model for Pediatric Asthma Care in Safety Net Health Settings
Primary PI: Cesar Termulo, Jr., MD, Parkland Health
Co-investigators: Yolande Pengetnze, MD, MS, George (Holt) Oliver, MD, PhD, Parkland Center for Clinical Innovation (PCCI); Amrita Waingankar, MD, MBA, Parkland Community Health Plan (PCHP); Naomi Gebrelul, MD, Foremost Family Health Centers; Sharon Davis, DO, Los Barrios Unidos Community Clinic
Parkland has developed an AI/ML risk prediction model leveraging EHR data to identify rising asthma risk in pediatric patients. The model distributes risk reports to frontline providers and is integrated into Parkland’s EHR to trigger point-of-care alerts during outpatient visits for Very-High- or High-Risk patients. This proposal will test the generalizability of Parkland’s EHR model in two large Federally Qualified Health Centers (FQHCs): Los Barrios Unidos (LBU) and Foremost Family Health Centers (Foremost), which serve some of the most underserved Dallas communities.
Contact Us: Questions about the initiative can be directed to AIM-HI@kp.org.
