Objective: To identify independent predictors of severe pneumonia in a local population, and create a simple severity score that would be useful in the ED.
Methods: Data on the clinical features of patients presenting to hospital with community-acquired pneumonia were collected. Multivariate logistic regression was used to identify independent predictors of death, requirement for ventilatory or inotropic support, and these combined. These predictors were used to modify an existing severity score, and its performance was tested in a second cohort of patients.
Results: A total of 392 patients in the derivation, and 330 in the validation cohorts. Independent predictors of 'death and/or requirement for ventilatory or inotropic support' were: systolic blood pressure (BP) <90 mmHg (OR 3.49 [95% CI 1.12-10.38]); acute confusion (OR 5.48 [95% CI 2.74-10.99]); oxygen saturations < or =90% (OR 3.49 [95% CI 1.77-6.89]); and respiratory rate > or =30/min (OR 2.65 [95% CI 1.35-5.21]). Age >65 years was not an independent predictor in this patient group (OR 0.52 [95% CI 0.23-1.16]). This information was used to propose that severe pneumonia could be predicted by two or more of: acute confusion; oxygen saturations < or =90%; respiratory rate > or =30/min; and either systolic BP <90 mmHg; or diastolic BP < or =60 mmHg. In a separate cohort, the performance of this score was similar to other tools.
Conclusion: This provides a practical tool that can be used to 'flag' impending patient demise. Its advantages are that it is simple, uses predictive variables, does not require invasive testing, and removes bias regarding patient age. Like other tools, its accuracy is not perfect, and it should only be used to augment clinical judgement.