# Recognition of Serious Infections in the Elderly Visiting the Emergency Department: The Development of a Diagnostic Prediction Model (ROSIE)

## TL;DR

This study developed a diagnostic model called ROSIE to help identify serious infections in elderly emergency department patients using easily measured clinical features and biomarkers.

## Contribution

The novel contribution is a validated clinical prediction model for serious infections in older adults using systolic blood pressure, oxygen saturation, and C-reactive protein.

## Key findings

- The model achieved an AUROC of 0.82, indicating good discriminatory power.
- Systolic blood pressure, oxygen saturation, and C-reactive protein were the final predictors retained in the model.
- Adding procalcitonin did not improve the model's discrimination.

## Abstract

Background/Objectives: Serious infections in older adults are associated with substantial mortality and morbidity. Diagnosis is challenging because of the non-specific presentation and overlap with pre-existing comorbidities. The objective of this study was to develop a clinical prediction model using clinical features and biomarkers to support emergency care physicians in diagnosing serious infections in acutely ill older adults. Methods: We conducted a prospective cross-sectional diagnostic study, consecutively including acutely ill patients (≥65 year) presenting to the emergency department. Clinical information and blood samples were collected at inclusion by a trained study nurse. A prediction model for any serious infection was developed based on ten candidate predictors that were further reduced to four ad interim using a penalized Firth multivariable logistic regression model. We assessed discrimination and calibration of the model after internal validation using bootstrapping. Results: We included 425 participants at three emergency departments, of whom 215 were diagnosed with a serious infection (51%). In the final model, we retained systolic blood pressure, oxygen saturation, and C-reactive protein as predictors. This model had good discriminatory value with an Area Under the Receiver Operating Characteristic (AUROC) curve of 0.82 (95% CI: 0.78 to 0.86) and a calibration slope of 0.96 (95% CI: 0.76 to 1.16) after internal validation. Addition of procalcitonin did not improve the discrimination of the model. Conclusions: The ROSIE model uses three predictors that can be easily and quickly measured in the emergency department. It provides good discriminatory power after internal validation. Next steps should include external validation and an impact assessment.