A Modified Physical Disability Screening Model After Treatment in the Intensive Care Unit: A Nationwide Derivation-Validation Study Publisher



Moayed MS¹ ; Vahedianazimi A¹ ; Goharimoghadam K² ; Asgharijafarabadi M³ ; Reiner Z⁴ ; Sahebkar A^{5, 6, 7, 8} Show Affiliations
Source: Journal of Clinical Medicine Published:2022

Abstract

Background: Many of the survivors of critical illnesses in the intensive care unit (ICU) suffer from physical disability for months after the treatment in the ICU. Identifying patients who are susceptible to disability is essential. The purpose of the study was to modify a model for early in-ICU prediction of the patient’s risk for physical disability two months after the treatment in the ICU. Methods: A prospective multicenter derivation–validation study was conducted from 1 July 2015, to 31 August 2016. We modified a model consisting of three risk factors in the derivation group and tested the modified model in the validation group. They were asked for their physical abilities before being admitted, two months after discharge from the ICU by a binary ADL staircases questionnaire. The univariate and multivariate logistic regression was used to modify physical disability components in the derivation data set. Receiver operating characteristic curves were used to determine the sensitivity and specificity of the threshold values in the validation group. Results: Five-hundred nineteen survivors were enrolled in the derivation group, and 271 in the validation. In multivariable analysis, the odds ratio (OR) of physical disability significantly increased with educational level ≤ elementary school (OR: 36.96, 95%CI: 18.14–75.29), inability to sit without support (OR: 15.16, 95%CI: 7.98–28.80), and having a fracture (OR: 12.74, 95%CI: 4.47–36.30). The multivariable validation model indicated that education level, inability to sit without support, and having a fracture simultaneously had sensitivity 71.3%, specificity 88.2%, LR+ 6.0, LR− 0.33, PPV 90.9, and NPV 64.9 to predict physical disability. Applying the coefficients derived from the multivariable logistic regression fitted on the derivation dataset in the validation dataset and computing diagnostic index sensitivity 100%, specificity 60.5%, LR+ 2.5, LR− 0.003, PPV 80.8, and NPV 100. The modified model had an excellent prediction ability for physical disability (AUC ± SE = 0.881 ± 0.016). Conclusions: Low education level, inability to sit without support, and having a fracture in a modified model were associated with the development of physical disability after discharge from ICU. Therefore, these clinical variables should be considered when organizing follow-up care for ICU survivors. © 2022 by the authors. Licensee MDPI, Basel, Switzerland.