Monday, February 10, 2025
2/10/2025
Acute kidney injury (AKI) is a frequent complication among critically ill children and is independently associated with increased morbidity and mortality. While AKI identification remains largely reactive, early prediction of AKI provides opportunities for mitigation strategies to improve patient outcomes. Existing pediatric AKI risk prediction models are primarily driven by single-center data with limited translation into clinical decision support (CDS) at the point of care. While CDS systems are designed to improve care quality and patient safety, they have failed to make meaningful improvements while contributing substantially to provider alert fatigue. This is especially true in pediatric critical care, where alerts and alarms are ubiquitous and often ignored. Disregarding important alerts may lead to patient harm. The addition of machine learning methods to augment decision support has been the focus of substantial hype, however clinicians remain skeptical over concerns such as utility, credibility, and validity. Within this clinical context, this application addresses two specific knowledge gaps. First, without validated and generalizable prediction tools, the best method for identifying patients at risk for developing AKI in the pediatric ICU (PICU) is unknown. Second, current CDS systems are burdensome and there exist no clear recommendations on useful, usable implementations for AKI decision support. The long- term goal of my research program is to improve the care and outcomes of critically ill children with acute kidney injury while reducing provider alert frustration through personalized, shareable CDS. As a step toward this goal, this career development award will demonstrate success of an integrated CDS pipeline for external validation, cutting-edge technology to facilitate distributed implementation, and usability assessment. The research aims of this proposal are to: 1) establish baseline external validity of an existing AKI prediction model and extend with new predictors to improve accuracy within two novel multi-center datasets, each with >80,000 PICU encounters; 2) overcome interoperability barriers using Fast Healthcare Interoperability Resource (FHIR)-based EHR technology and prospectively evaluate this implementation’s performance; and 3) implement principles of user-centered design to improve clinician acceptance. With the support of my mentoring committee, I will gain advanced knowledge and develop expertise in user-centered design, qualitative methodology, clinical trial design, and implementation science in decision support. This work is innovative as it leverages two brand-new, highly granular datasets to achieve external validation, implements an AKI prediction tool using new FHIR-based technology, and utilizes mixed-methods cognitive informatics tools to perform usability testing. However, the truly innovative aspect of this work is the integration of all three aims into a single pipeline. This work provides the foundation for a prospective clinical trial evaluating clinical effectiveness, as well as incorporation of biomarkers and physiologic data streams, and the development of evaluation measures for decision support systems.
HHS’ Tracking Accountability in Government Grants System (TAGGS) website provides access to detailed descriptions of grants, loans, aggregated direct payments and other types of financial assistance awarded by the HHS Operating Divisions (OPDIVs) and the Office of the Secretary Staff Divisions (STAFFDIVs).