Overview
A recent study published in the Journal of Critical Care highlights the potential of artificial intelligence (AI) in improving resource efficiency in intensive care units (ICUs) for patients suffering from severe community-acquired pneumonia (CAP).
Study Details
- Date: November 18, 2025
- Conducted by: D’Or Institute for Research and Education (IDOR)
- Participants: 16,985 adult CAP admissions across 220 ICUs in 57 Brazilian hospitals during 2023
Challenges in ICU Management
Severe CAP poses significant challenges for ICUs, requiring extensive resources such as:
- Prolonged hospital stays
- Respiratory support
Traditional evaluation methods often overlook patient severity, complicating fair comparisons between hospitals and hindering effective management strategies.
Introducing SLOSR
To tackle these issues, researchers developed the Standardized Length of Stay Ratio (SLOSR) using machine learning techniques. This tool aims to:
- Predict the appropriate length of ICU stay based on patient risk
- Facilitate accurate comparisons across hospitals
- Identify both overuse and underuse of resources
Methodology
The study employed a retrospective, multicenter approach, analyzing various factors including:
- Age
- Comorbidities
- Need for mechanical ventilation
- Disease severity
A machine learning model was utilized to predict the expected length of stay, allowing researchers to calculate the SLOSR as the ratio of observed to predicted times. Rigorous statistical validation was performed to ensure the model’s accuracy.
Key Findings
- Median Length of Stay: 4 days
- Patients Requiring Ventilatory Support: Approximately 28%
- Model Performance: Strong explanatory power with low prediction errors
Implications for Healthcare
The findings suggest that SLOSR could serve as a valuable tool for hospitals and healthcare managers, enabling:
- Performance evaluation of ICUs adjusted for patient severity
- Identification of efficient resource use and areas of waste
However, researchers emphasize the need for further studies to assess the method’s applicability in diverse healthcare contexts.
For more information, refer to the study by Quintairos et al. in the Journal of Critical Care (2025). DOI: 10.1016/j.jcrc.2025.155208.