# Clinical Gestalt to Predict Bacterial Infection and Mortality in Emergency Department Patients: A Prospective Observational Study

## TL;DR

This study shows that emergency doctors' initial clinical impressions can help identify patients with bacterial infections and predict mortality risk.

## Contribution

Demonstrates the utility of clinical gestalt in predicting bacterial infection and mortality in ED patients with suspected infection.

## Key findings

- Physicians' clinical gestalt effectively differentiated bacterial from non-bacterial infections.
- Higher gestalt scores correlated with increased antibiotic prescriptions.
- Gestalt showed moderate accuracy in predicting 30-day mortality.

## Abstract

Time to treatment is a significant predictor of mortality in emergency department (ED) patients with bacterial sepsis. Strategies for the early detection of bacterial infection and sepsis are lacking. Clinical gestalt is a tool for assessing and synthesizing the entire clinical picture, focusing on the first clinical impression at presentation.

This study aimed to assess ED physicians’ clinical gestalt for the prediction of bacterial infection and mortality in ED patients presenting with signs and symptoms of infection.

Prospective, observational study with a 30-day follow-up.

Patients aged 18 or older presenting to the ED with signs and symptoms compatible with an infection and abnormal vital signs were included.

ED physicians recorded their clinical gestalt using a visual analog scale (VAS) to assess the likelihood of bacterial infection and responded to a dichotomous question regarding the probability of a patient’s death. The main outcome was the confirmed diagnosis of an acute bacterial infection. Final diagnoses, based on laboratory and follow-up information, were adjudicated by an expert panel.

In total, 444 patients were included. Median age was 68 years [IQR 51, 80] and median National Early Warning Score (NEWS) was 5 [IQR 3, 7]. Median VAS for physicians’ clinical gestalt regarding bacterial infection likelihood was 8.2 [IQR 6.7, 9.0] of 10 in patients with bacterial infection, 2.3 [IQR 1.2, 4.3] in patients with viral infection, 4.6 [IQR 4.0, 7.3] in patients with an infection due to another pathogen, and 2.3 [IQR 1.1 6.2] in patients with no acute infection (p-value = <0.001). Clinical gestalt’s sensitivity regarding 30-day mortality was 57.1% [95%CI 37.2 to 75.5] with a specificity of 83.4% [95%CI 79.7 to 86.9].

In this study of ED patients presenting with signs and symptoms of infection, clinical gestalt was shown to be useful differentiating between bacterial and infections of other causes. Antibiotic prescription rate increased with the likelihood of bacterial infection according to physician gestalt. Lastly, simple heuristic prognostication of mortality (likely vs. unlikely) carried some, but limited, prognostic value.

The online version contains supplementary material available at 10.1007/s11606-025-09440-7.