A predictive model for the classification of emergency intensive care unit patients with systemic inflammatory response syndrome based on a similarity network fusion algorithm.

Background: Intensive care unit-acquired weakness (ICU-AW) is a prevalent and severe neuromuscular complication in critically ill patients. It is a consequence of critical illness and is characterized by systemic inflammatory response syndrome (SIRS)-induced metabolic stress and multiple organ dysfunctions. Moreover, ICU-AW is one of the most important factors affecting the prognosis of patients with SIRS, Electrophysiological examination is an effective method for early identification and monitoring of the course of the disease and is essential for accurate diagnosis of critical illness neuromyopathy (CINM). The data-intensive ICU environment is ideal for implementing the similarity network fusion (SNF) method. The objective of this study was to establish and validate a ICU-AW predictive model in SIRS patients, providing a practical tool for early clinical prediction.

Methods: Clinical characteristics, demographic data, longitudinal neurophysiological data, and disease severity indicators of the enrolled patients were recorded. The patient data included nerve conduction, F-wave, and direct muscle stimulation (DMS) data from 94 follow-up visits as well as various scores, including Medical Research Council (MRC), sequential organ failure assessment (SOFA), acute physiology and chronic health evaluation II (APACHE II) scores and C-reactive protein (CRP). This algorithm was used to analyze electrophysiological data of emergency intensive care unit (EICU) patients with SIRS and fully exploit their similarities in age, sex, body mass index, and electrophysiological data by fusing the similarity networks of these patients with different sets of attributes. Existing patients was performed a clustering analysis and predicted the classification of new patients using spectral clustering and label propagation algorithms on the fusion network, respectively.

Results: Classification prediction model categorical of ICUAW in Patients with SIRS was highly consistent with the clinical diagnosis and had high accuracy and discriminative ability. The model captures the importance of advanced age and lung infections as risk factors for ICU-AW and also demonstrates the significant prognostic value of DMS in EICU patients with SIRS and its ability to predict the development of clinical muscle weakness.

Conclusions: Electrophysiological abnormalities are a critical feature of both ICU-AW and non-ICU-AW. Modeling the prediction of SIRS patients progressing to ICUAW which is conducive to early intervention, mechanism studies, and patient rehabilitation.