An improved medical admissions risk system using multivariable fractional polynomial logistic regression modelling

Area of Science:

• Medical Informatics
• Clinical Prediction Models
• Acute Medicine

Background:

• In-hospital mortality prediction for acute medical patients is crucial for resource allocation.
• Existing tools may not fully utilize routinely collected data.
• Need for validated risk assessment tools in emergency departments.

Purpose of the Study:

• To develop and validate an in-hospital mortality risk prediction tool.
• Utilize routinely collected physiological and laboratory data.
• Targeting unselected acutely ill medical patients.

Main Methods:

• Analysis of 10,712 patients from St James's Hospital (SJH) and 3,597 from Nenagh Hospital.
• Development of a 5-day in-hospital mortality risk score (MARS) using logistic regression.
• Inclusion of nine variables: age, vital signs, urea, potassium, hematocrit, and white cell count.

Main Results:

• The MARS score demonstrated high predictive accuracy.
• Area Under the Receiver Operating Characteristic Curve (AUROC) was 0.93 for SJH and 0.92 for Nenagh Hospital.
• The model showed good calibration with Hosmer and Lemeshow goodness-of-fit tests (P > 0.05).

Conclusions:

• The MARS system effectively estimates in-hospital mortality using routine emergency department data.
• This tool can assist clinicians in prompt and accurate resource allocation.
• Further studies are needed to validate the impact of MARS on clinical outcomes.