Sepsis is a frequently encountered clinical entity in pediatric intensive care units (PICUs) leading to significant morbidity and mortality. Blood cultures (BCs), as the only way to identify blood stream infections, represent a cornerstone of sepsis diagnosis and treatment. The overuse of BCs, however, has been shown to result in additional laboratory tests, unnecessary antibiotic use, prolonged hospitalization, and increased healthcare costs. Existing clinical decision support (CDS) tools for improving BC utilization focus on the use of analytic approaches to determine the pretest probability of bloodstream infections. These CDS tools cannot address the naturalistic and intuitive nature of clinical decision-making and, therefore, may not get adopted by providers and improve decision quality. To improve BC utilization and reduce its negative impacts on healthcare quality, safety, and efficiency, innovative CDS tools drawing upon the strengths of both analytical and naturalistic decision-making are needed. In this study, we propose to develop an electronic health record-embedded CDS tool for improving BC utilization, which will use the strength of analytical decision-making in synthesizing large amounts of data to facilitate naturalistic decision-making. We will apply a sociotechnical systems approach, i.e., the Systems Engineering Initiative for Patient Safety 2.0 model, and a user-centered design method, i.e., contextual design, to guide the development of the CDS tool, which is an iterative process including three steps: First, we will examine individual and team cognitive work associated with obtaining a BC (Aim 1). We will conduct shadowing with contextual inquiry and critical decision method interviews with providers to identify their information needs and strategies used when making decisions regarding obtaining a BC. Second, based on findings from Aim 1, we will develop the CDS tool in collaboration with providers (Aim 2). Focus group design sessions will be conducted with providers to (1) translate findings from Aim 1 into design requirements, (2) design and evaluate mock-ups, and (3) develop an implementation plan. Outcomes of the focus group design sessions will be used to develop and iterative evaluate prototypes of the CDS tool. Finally, we will implement the CDS tool in the PICU at the Johns Hopkins Hospital and assess its use and impacts on BC utilization and patient outcomes. While qualitative observation and interview data and computer-generated quantitative data will be collected to assess the use of the CDS tool, an interrupted time series design will be used to assess the impacts of the CDS tool on BC utilization and patient outcomes. This study will demonstrate that the CDS tool developed to use analytical decision-making to facilitate naturalistic decision-making can reduce BC utilization in the PICU without influencing patient outcomes. The CDS tool and the approach used to develop the CDS tool will be generalizable to a broader set of healthcare issues beyond BCs.
In this self-contained health information technology study, we propose to develop an electronic health record- embedded clinical decision support tool based on the understanding of individual and team cognitive work to facilitate data-driven naturalistic decision-making and reduce unnecessary BC utilization in critically ill children.
