University Of Alabama At Birmingham receives $200K Trans-Nih Research Support grant from the National Institutes of Health

Implementing human-centered ai in acute care to facilitate prep delivery - project summary people who use drugs (pwud) are at high risk for acquiring hiv, yet they are rarely prescribed hiv pre- exposure prophylaxis (prep), particularly in acute care settings such as the emergency department and hospital, where many have their only meaningful contact with the healthcare system. Existing hiv risk assessment tools perform poorly in these environments partly because many characteristics key to establishing hiv risk are documented primarily in the electronic health record (ehr) as unstructured clinical text and are not captured by structured data elements. Innovative approaches are needed to improve identification of prep-eligible pwud and to support non-specialist physicians in delivering prep care in these settings, particularly in under-resourced settings like the u.s. Deep south. The overall objective of this mentored K23 award is to develop dr. Bradford into an independent physician-scientist with expertise at the intersections of infectious diseases, addiction medicine, and the development of explainable, human-centered clinical artificial intelligence (ai). The proposed research will integrate qualitative methods, human-centered design, and ai-based approaches to design and validate a clinical decision support (cds) system to facilitate prep delivery by generalists in emergency and inpatient settings. While dr. Bradford has deep expertise in hiv prevention, addiction medicine, and quantitative methods, he requires advanced skills in human-centered design and ai pipeline development to build an independent research program in this important topic area. Dr. Bradford’s training objectives parallel his research aims and include (1) gaining advanced training in human-centered design and qualitative methods, (2) developing expertise in explainable ai and language model-based ehr pipelines, and (3) acquiring skills in rigorous model validation, including assessment of bias. These objectives will be achieved through mentored, hands-on research with a multidisciplinary team spanning hiv prevention, clinical ai, and biostatistics. Dr. Bradford’s three research aims focus on developing and validating a human-centered, ai-enabled cds approach to improve prep delivery for pwud in acute care. Aim 1 will use qualitative interviews and human-centered design to inform the design of a cds prototype which will undergo iterative prototyping and usability testing. In parallel, aim 2 will develop and evaluate methods to identify substance use- and hiv risk- related behaviors from unstructured ehr text, enabling more accurate identification of patients who may benefit from prep. Aim 3 will validate these methods in our local deep south health system, assessing predictive performance and potential bias using an existing longitudinal ehr-based cohort of hiv incidence. Each of these aims will provide the opportunity to learn key methods for designing, developing, and assessing the performance of ai-based pipelines for hiv prevention. This K23 award will not only advance prep for pwud but also position dr. Bradford as an independent physician-scientist leader at this critical inflection point in healthcare ai.