# Predicting ICU transfer for high-risk patients upon medical admission via the medical intensive care prediction score (MICAPS)

## TL;DR

This study developed a new scoring system called MICAPS to predict which patients admitted to the hospital will need intensive care, helping doctors make early decisions and manage resources better.

## Contribution

The novel contribution is the creation and validation of the MICAPS score, a predictive tool for early identification of patients likely to require ICU transfer.

## Key findings

- MICAPS achieved an area under the ROC curve of 0.809, indicating strong predictive accuracy.
- Key predictors of ICU transfer included abnormal respiratory rate, low oxygen saturation, and Glasgow Coma Scale score below 9.
- The model had a sensitivity of 67.4% and specificity of 81.3% at a score threshold of ≥40.

## Abstract

Early identification of patients at risk for admission to the medical intensive care unit (MICU) at the time of medical admission is crucial for optimizing resource utilization and improving patient outcomes. No standardized, unified scoring system exists to predict MICU requirements for early medical admissions (EMA). This study aimed to develop and validate a predictive scoring system, the Medical Intensive Care Admission Prediction Score (MICAPS), to identify patients at high risk of transfer to the MICU based on demographic data, triage hemodynamics, and limited presentation-day laboratory data.

This retrospective cross-sectional study included 11,847 adult patients admitted to medical floors via the emergency department (ED) at Hamad General Hospital, Qatar, between January 2019 and December 2019. Cerner® was used to extract relevant data. Multivariate logistic regression identified significant predictors of MICU admission, and regression coefficients were used to develop the MICAPS model. ROC curve analysis and bootstrapping methods were employed to validate the model’s performance and accuracy.

Of 11,847 patients admitted to medical services, 909 (7.7%) were transferred to MICU. Significant predictors included male gender (OR: 1.41, 95% CI: 1.17-1.70), age ≤ 60 years (OR: 2.15, 95% CI: 1.72–2.68), abnormal respiratory rate (OR: 2.35, 95% CI: 1.48–3.72), oxygen saturation < 88% (OR: 1.95, 95% CI: 1.30–2.92), Glasgow Coma Scale < 9 (OR: 6.54, 95% CI: 4.91–8.71), RRT activation (OR: 3.82, 95% CI: 3.19–4.56), and abnormal laboratory values such as WBC ≥ 10 (OR: 1.29, 95% CI: 1.08–1.54) and lactate > 1.7 mmol/L (OR: 1.96, 95% CI: 1.64–2.34). MICAPS demonstrated good predictive power, with an area under the ROC curve of 0.809 (95% CI: 0.79–0.82), a sensitivity of 67.4%, a specificity of 81.3%, and a positive likelihood ratio of 3.60 at a score of ≥ 40.

MICAPS is a simple-to-apply scoring system that enables the identification of patients early in their hospitalization who may require ICU care later during the hospital stay. It can support early clinical decision-making and optimize resource allocation in emergency departments, medical floors, and critical care settings. Further multicenter prospective validation is needed to assess its generalizability in the real world.

Not applicable.

The online version contains supplementary material available at 10.1186/s12873-025-01448-w.