Hostname: page-component-586b7cd67f-vdxz6 Total loading time: 0 Render date: 2024-11-23T10:31:50.371Z Has data issue: false hasContentIssue false

P082: Predictive ability of the quick Sepsis-related Organ Failure Assessment score among patients with infection transported by paramedics: a Bayesian analysis

Published online by Cambridge University Press:  02 May 2019

S. Alex Love
Affiliation:
University of Calgary, Calgary, AB
D. Lane*
Affiliation:
University of Calgary, Calgary, AB

Abstract

Core share and HTML view are not available for this content. However, as you have access to this content, a full PDF is available via the ‘Save PDF’ action button.

Introduction: The quick Sepsis-related Organ Failure Assessment (qSOFA) score was developed to provide clinicians with a quick assessment for patients with latent organ failure possibly consistent with sepsis at high-risk for mortality. With the clinical heterogeneity of patients presenting with sepsis, a Bayesian validation approach may provide a better understanding of its clinical utility. This study used a Bayesian analysis to assess the prediction of hospital mortality by the qSOFA score among patients with infection transported by paramedics. Methods: A one-year cohort of adult patients transported by paramedics in a large, provincial EMS system was linked to Emergency Department (ED) and hospital administrative databases, then restricted to those patients with an ED diagnosed infection. A Bayesian binomial regression model was constructed using Hamiltonian Markov-Chain Monte-Carlo sampling, normal priors for each parameter, the calculated score, age and sex as the predictors, and hospital mortality as the outcome. Discrimination was assessed using posterior predictions to calculate a “Bayesian” C statistic, and calibration was assessed with calibration plots of the observed and predicted probability distributions. The independent predictive ability of each measure was tested by including each component measure (respiratory rate, Glasgow Coma Scale, and systolic blood pressure) as continuous predictors in a second model. Results: A total of 9,920 patients with ED diagnosed infection were included. 264 (2.7%) patients were admitted directly to the ICU, and 955 (9.6%) patients died in-hospital. As independent predictors, the probability of mortality increased as each measure became more extreme, with the Glasgow Coma Scale predicting the greatest change in mortality risk from a high to low score; however, no dramatic change in the probability supporting a single decision threshold was seen for any measure. For the calculated score, the C statistic for predicting mortality was 0.728. The calibration curve had no overlap of predictions, with a probability of 0.5 (50% credible interval 0.47-0.53) for patients with a qSOFA score of 3. Conclusion: Although no single decision threshold was identified for each component measure, a calculated qSOFA score provides good prediction of mortality for patients with ED diagnosed infection. When validating clinical prediction scores, a Bayesian approach may be used to assess probabilities of interest for clinicians to support better clinical decision making. Character count 2494

Type
Poster Presentations
Copyright
Copyright © Canadian Association of Emergency Physicians 2019