C. J. Hu1,2, R. P. Garg2, L. Janczewski2, J. D. Slocum2, A. Berry1, M. Rafferty1,6, L. Shamsuddin3,4, J. Huml1, N. Nadig1,5, A. Stey1,2 1Feinberg School Of Medicine – Northwestern University, Chicago, IL, USA 2Northwestern Medicine, Department Of Surgery, Chicago, IL, USA 3Northwestern Medicine, Northwest Region, Department Of Pulmonary And Critical Care, Chicago, IL, USA 4Rosalind Franklin University Chicago Medical School, North Chicago, IL, USA 5Northwestern Medicine, Department Of Pulmonary And Critical Care, Chicago, IL, USA 6Shirley Ryan AbilityLab, Chicago, IL, USA
Introduction:
Shock is associated with a 50% mortality rate depending on the shock etiology. Patients in shock are seen by many different providers because of the different shock etiologies. This study sought to characterize the diagnostic concordance of shock etiology among physicians caring for shock, and to identify relationships between diagnostic concordance and mortality. We hypothesized that patients with in-hospital mortality would have lower diagnostic concordance between providers.
Methods:
This was an observational retrospective cohort study conducted across a health system of 11 acute care hospitals. Inclusion criteria were adult patients (≥ 18 years) hospitalized with shock between 2018-2022 and evaluated by at least two different providers. Shock was defined based on the inclusion of “shock” in a history and physical, progress, or consult note on hospital day 0-4, as captured in Electronic Health Record data. Diagnostic concordance between providers of shock etiology was estimated using cosine similarities. 15-dimension vectors representing the 15 potential shock etiologies documented by each provider were created for each patient encounter. Cosine similarity scores were calculated between vectors for each hospital day for each patient encounter. The average cosine similarity score was compared between hospital days, patient characteristics, discharge disposition, and mortality.
Results:
Overall, there were 26,277 eligible patient encounters. The largest proportions of patient encounters were 61-70 years old (25.6%), male (55.0%), and non-Hispanic White (72.8%). Most patient encounters (85.4%) had only one shock etiology, but 13.1% of encounters documented two shock etiologies and 1.5% documented three. Shock etiology documentation among patient encounters increased from day 0 (3,822) to day 1 of hospital stay (4,546), then decreased over days 2-4 (3,628 on day 4). Average cosine similarity scores increased from 0.858 on day 0 to 0.862 on days 1-3, rising to 0.873 on day 4. Female patients had lower cosine similarity scores between providers across all hospital days (e.g., day 4 score; males: 0.875, females: 0.870). Patients who died in hospital or eloped exhibited lower cosine similarity scores across all hospital days compared to discharged patients (Figure).
Conclusion:
This study demonstrated variability in shock etiology diagnostic concordance between providers across time. Patients who died in the hospital had lower diagnostic concordance between providers across all hospital days.
