Introduction: This proposal aims to understand and improve the care of Veterans treated for pneumonia through informatics. Background: Pneumonia is the most common infectious cause of death in the United States, with an estimated 50,000 deaths and $8-10 billion in costs annually.12 Twenty-five thousand Veterans are seen per year in the Emergency Department, and 20,000 are hospitalized.13 Risk adjusted 30-day mortality rates range among VA facilities from 11% to 18%. The resources utilized for pneumonia also vary widely, with the cost of hospital care estimated at up to 25 times that of outpatient care.20 To improve the standard of pneumonia care, intense research and quality improvement efforts have been focused upon establishing evidence-based practice guidelines for pneumonia that represent a consensus of research and opinions generated from our best academic hospitals and pneumonia experts. However, we continue to see variation in both management and outcomes across the VA as in other systems, suggesting that guidelines may be difficult to apply to real practice. The recent advancement of our electronic health record allows us to measure actual practice at both the population and the individual level. By studying variation in triage and antibiotic decisions for patients with pneumonia, we can learn from our own population to generate evidence that includes the previously underrepresented patients, settings and scenarios, bringing it closer to real practice. Career Aims: My career goals are to: 1) Develop skills in advanced population analytics, 2) Gain a foundation in behavioral sciences, and 3) Develop a deeper understanding in informatics. Research Aims: My research aims build toward my career aims: (1) Examine sources of variation in triage and antibiotic selection decisions for pneumonia across VA emergency departments. (2) Characterize the cognitive processes influencing decision-making in pneumonia among providers at emergency departments demonstrating variation in decision-making. (3) Design and test an informatics tool that supports decision- making, integrates with workflow, and allows us to learn from clinical experience. Methods: Aim 1 A. will develop and test models of practice variation for triage and antibiotic selection and b. will identify emergency departments with high and low decision-making thresholds across the VA system. Aim 2 will use mixed methods to characterize the cognitive process of clinical reasoning and contextual influences impact decision-making through interviews with providers. Aim 3 will utilize a real-time informatics tool to test current best practice recommendations against physician decisions by providing physicians the opportunity to disagree with any recommendations, and collect information on reasons for disagreement. Impact: This work will directly inform the development of informatics tools for pneumonia for the VA. Ultimately, while my clinical interest is in pneumonia, the skills I develop will be applied to many clinical problems in medicine where decision-making occurs in the setting of uncertainty.

Pneumonia is a common but challenging disease, requiring providers to make difficult decisions under uncertainty in diagnosis, etiology, and consequences. Implementation of practice guidelines based upon traditional evidence-based medicine has made overall improvements in care, but this approach fails to account for the complexity in decision-making at the individual level for many clinical scenarios. The electronic health record within the VA provides a unique opportunity to build a research foundation that learns through the experience of the population, individuals, and also at the point of care using informatics tools. My research integrates big data, interviews with providers, and an informatics tool to generate evidence and cognitive support that are relevant to more clinical environments and patients. This will ultimately enhance decision-making and lead to better, more individualized care for patients.