Predicting intensive care uniteacquired weakness in the first week of an intensive care unit stay: A multicentre external validation study

Abstract

Background: To diagnose intensive care uniteacquired weakness (ICU-AW) in a timely manner, we previously constructed a prediction model based on multicentre data from 642 patients (development cohort), focussing on the presence of ICU-AW on days 3*5 of ICU admission.Objective: The aim of this study was to investigate the external validity of the original prediction model in a new multicentre cohort (10 of the 80 original ICUs) and investigate frailty as a new predictor.Methods: Newly admitted patients with an ICU stay for >48 h were included. Predictors were prospectively recorded, and an outcome of ICU-AW was defined by a Medical Research Council (MRC) mean score <48. We assessed calibration and discrimination in the original prediction model in the validation cohort (411 patients). We then updated the model in the validation cohort by adding frailty, measured using FRAILEspña and the Clinical Frailty Scale-España.Results: Of the 351 patients with an MRC score in the validation cohort, 195 (55.5%) developed ICU-AW. Model calibration and discrimination in the original model were good with these patients (calibrationin- the-large was 0.17 [95% confidence interval {CI}: {-0.07; 0.40}], slope was 0.93 [95% CI: {0.66; 1.21}], and area under the receiver operating characteristic curve was 0.723 [95% CI: {0.67; 0.78}]. However, when the model validation included unconscious patients (unfeasible for MRC score assessment) classified as having ICU-AW, there was a tendency to underestimate ICU-AW. Model updating did not improve performance (net reclassification improvement was -0.2% with Clinical Frailty Scale-España and -0.3% with FRAIL-España.Conclusions: The initial prediction model for ICU-AW shows good performance in this new independent multicentre validation cohort, which confirms that the predictor variables of ICU-AW for the first 5 days of ICU stay are older age, being female, not being conscious for MRC assessment, and receiving renal replacement therapy. Protectors are active mobility and hyperactive delirium. The frailty variable does not enhance the predictive model.