Background Currently there exists no widely accepted, objective tool for determining the probability of shock in the emergency department (ED). Associated mortality outcomes remain unacceptably high. Using variables that are associated with shock and available in the ED we aimed to derive an objective tool for estimating the probability of shock

Methods In a prospective, longitudinal study conducted in the ED, Cardiff, adult patients aged ≥18 years, presenting with NEWS ≥3 were recruited. We derived a model consisting of 8 variables, and validated it using bootstrap (1000 iterations; random number seed: 978). The primary outcome was combined 30-day mortality or ICU admission.

Results 361 patients were recruited (mean age 69.4 ±15.9 years; male 62.7%; 80 positive outcome) to the study. The multivariate model variables were: Glasgow Coma Scale 3–8 (Odds ratio (OR) 9.1 (95% CI 1.2–69.4) p=0.0324), 9–12 (OR 1.3 (95% CI 0.4–4.2) p=0.6561); capillary return >4 (OR 8.7, (95% CI 2.9–26.1) p=0.0001), 3–4 (OR 3.7 (95% CI 1.3–10.4); p=0.0113); temperature ≤35°C (OR 5.4, (95% CI 1.8–16.3); p=0.0045); bilirubin >34 (OR 4.9, (95% CI 1.6–14.5); p=0.0045); oxygen supplement (OR 1.9, (95% CI 1.0–3.7); p=0.06); respiratory rate >24 (OR 2.6, (95% CI 1.3–5.1); p=0.0069); and albumin <35 (OR 1.9, (95% CI 0.9–3.8); p=0.09).

The model was a 100-point scale where 0 was low probability and 100 was maximal probability of shock. The AUC was 0.801 (95% CI 0.756–0.841), accuracy 82%, p<0.0001, Hosmer & Lemeshow test 9.19 (p=0.326). At a cut off >35, and Youden Index 0.4557 the sensitivity and specificity were 56.2% (95% CI 44.7–67.3) and 89.32% (95% CI 85.1–92.7) respectively. At a fixed specificity of 80%, the estimated sensitivity was 60.5% (95% CI 49.4–71.3). At a fixed sensitivity of 80%, the estimated specificity was 62.3% (95% CI 45.9–71.2).

Conclusions An accurate 7-point ED tool for shock has been derived and internally validated.