A prediction model for thrombocytopenia after neurosurgery: a retrospective study

Area of Science:

• Neurosurgery
• Hematology
• Medical Informatics

Background:

• Thrombocytopenia is a common complication after major surgery, particularly neurosurgery.
• It is associated with adverse clinical outcomes, necessitating predictive tools.

Purpose of the Study:

• To develop and validate a logistic regression model for predicting thrombocytopenia in patients post-neurosurgery.
• To identify key predictive variables for early detection of thrombocytopenia.

Main Methods:

• Retrospective analysis of 1,109 neurosurgery patients.
• Logistic regression model development using least absolute shrinkage and selection operator (LASSO) regression for variable selection.
• Model performance evaluation using training and test datasets, including subgroup analysis.

Main Results:

• The logistic regression model achieved high discriminative ability with an AUC of 0.916 (training) and 0.883 (test).
• Key predictors included hypertension, SOFA score, bilirubin, albumin, INR, thrombocytocrit, systolic blood pressure, and vasoactive drug use.
• The model demonstrated good calibration but low recall, indicating potential for early intervention support.

Conclusions:

• The developed logistic regression model can effectively predict thrombocytopenia following neurosurgery.
• This model can serve as a valuable clinical decision-support tool for timely interventions.
• Further refinement may improve recall for enhanced clinical utility.