[Establishment and validation of a risk prediction model for intensive care unit-acquired weakness]

To explore the risk factors of intensive care unit-acquired weakness (ICU-AW), and to establishment and verify its risk prediction model.

A modeling group of 231 patients who met the inclusion criteria and were admitted to the intensive care unit (ICU) of the First Hospital of Jiaxing from July 2019 to June 2020 was collected by convenience sampling method. According to whether they developed ICU-AW, they were divided into ICU-AW group (55 cases) and non ICU-AW group (176 cases). The clinical data were collected concerning patients' individual information, disease-related factors, treatment-related factors and laboratory indicators, and the differences of the above indexes between two groups were compared. Logistic regression was used to analyze the ICU-AW risk factors and a risk prediction model was constructed. Calculate the area under ROC curve (AUC) to test the prediction effect of the model. At the same time, 60 patients who admitted to ICU from July to October 2020 and met the standards were collected to verify the model.

Compared with non ICU-AW group, there were more males in ICU-AW group [61.8% (34/55) vs. 44.3% (78/176), P < 0.05], with higher levels of systemic inflammatory response syndrome (SIRS), sepsis, immobilization and the use of neuromuscular blockers [SIRS: 30.9% (17/55) vs. 3.4% (6/176), sepsis: 12.7% (7/55) vs. 2.3% (4/176), immobilization: 72.7% (40/55) vs. 39.2% (69/176), the use of neuromuscular blockers: 50.9% (28/55) vs. 14.2% (25/176), all P < 0.05], and acute physiology and chronic health evaluation II (APACHE II) score, blood lactic acid level and duration of mechanical ventilation, length of hospital stay were all increased [APACHE II score: 18 (15, 24) vs. 12 (8, 17), blood lactic acid (mmol/L): 2 (1, 2) vs. 1 (1, 2), duration of mechanical ventilation (days): 7 (4, 12) vs. 2 (2, 5), length of hospital stay (days): 10 (6, 16) vs. 5 (3, 9), all P < 0.05]. SIRS, APACHE II score, duration of mechanical ventilation and blood lactic acid were included to construct a risk prediction model [odds ratio (OR) values were 4.835, 1.083, 1.210, 1.790, P values were 0.018, 0.013, 0.015, 0.013]. The model equation was P = exp [-5.207+(1.576×SIRS)+(0.079×APACHE II)+(0.191×duration of mechanical ventilation)+(0.582×blood lactic acid)]. Internal verification: Calibration diagram showed the calibration curve above the ideal curve, AUC = 0.888, 95% confidence interval (95%CI) was 0.839-0.938; when the cut-off value was 0.166, the sensitivity was 89.1%, the specificity was 75.6%, and the maximum index was 0.649. External verification: Calibration diagram showed that the calibration curve was above the ideal curve, and the plotted AUC = 0.853, 95%CI was 0.753-0.953. When the cut-off value of the corresponding predictive risk value was 0.367, the sensitivity was 68.8%, the specificity was 86.4%, and the maximum approximate index was 0.552.

The risk prediction model of ICU-AW constructed in this study has good consistency and prediction efficiency, which can provide reference for medical personnel to identify high-risk groups of ICU-AW patients in the early stage and provide targeted interventions in advance.