Predicting 30-day mortality using point-of-care testing; an external validation and derivation study

Area of Science:

• Emergency Medicine
• Clinical Prediction Models
• Health Informatics

Background:

• Early risk stratification in the emergency department (ED) is crucial for prioritizing patient care.
• Existing prediction models often rely on laboratory parameters, leading to delays due to blood testing.
• Point-of-care testing (POCT) offers a solution for rapid results, preventing diagnostic delays.

Purpose of the Study:

• To externally validate a laboratory-based mortality prediction model in an ED setting.
• To assess the performance of a POCT-based model against a traditional laboratory-based model.
• To determine if POCT can provide comparable accuracy for early risk stratification.

Main Methods:

• Retrospective review of 34,437 adult ED patients (2012-2016).
• External validation of a laboratory-based model (age, glucose, urea, sodium, hemoglobin, platelets, WBC).
• Comparison with a POCT-equivalent model (glucose, urea, sodium, hemoglobin) using Area Under the Curve (AUC).

Main Results:

• The laboratory-based model showed an AUC of 0.812 after refitting.
• The POCT-equivalent model achieved a comparable AUC of 0.790.
• A slight decrease in performance was observed with the POCT model.

Conclusions:

• POCT models demonstrate similar performance to laboratory-based models for predicting 30-day mortality in the ED.
• POCT facilitates rapid risk stratification, potentially improving patient management.
• Further validation is needed, but POCT shows significant potential for ED decision-making.