Prediction of respiratory insufficiency in Guillain-Barré syndrome

Abstract

Objective: Respiratory insufficiency is a frequent and serious complication of the Guillain-Barré syndrome (GBS). We aimed to develop a simple but accurate model to predict the chance of respiratory insufficiency in the acute stage of the disease based on clinical characteristics available at hospital admission.

Methods: Mechanical ventilation (MV) in the first week of admission was used as an indicator of acute stage respiratory insufficiency. Prospectively collected data from a derivation cohort of 397 GBS patients were used to identify predictors of MV. A multivariate logistic regression model was validated in a separate cohort of 191 GBS patients. Model performance criteria comprised discrimination (area under receiver operating curve [AUC]) and calibration (graphically). A scoring system for clinical practice was constructed from the regression coefficients of the model in the combined cohorts.

Results: In the derivation cohort, 22% needed MV in the first week of admission. Days between onset of weakness and admission, Medical Research Council sum score, and presence of facial and/or bulbar weakness were the main predictors of MV. The prognostic model had a good discriminative ability (AUC, 0.84). In the validation cohort, 14% needed MV in the first week of admission, and both calibration and discriminative ability of the model were good (AUC, 0.82). The scoring system ranged from 0 to 7, with corresponding chances of respiratory insufficiency from 1 to 91%.

Interpretation: This model accurately predicts development of respiratory insufficiency within 1 week in patients with GBS, using clinical characteristics available at admission. After further validation, the model may assist in clinical decision making, for example, on patient transfer to an intensive care unit.