Development of an Early Warning Model for Intensive Care Unit-Acquired Weakness in Mechanically Ventilated Children

A prospective cohort study of mechanically ventilated children was established to systematically analyze epidemiological characteristics. The modified Pediatric Medical Research Council (MRC) muscle strength scale (pMRC) combined with simplified bedside neuroelectrophysiological testing (measurement of common peroneal nerve compound muscle action potential amplitude) was used to determine the occurrence rate, subtype distribution (CIP/CIM/Mixed), and natural disease course of intensive care unit-acquired weakness (ICU-AW) among mechanically ventilated children in China. An age-stratified model was applied to analyze differences in the occurrence rate of ICU-AW among children of different age groups. A Cox regression model was employed to quantify the dose-response relationship between dynamic parameters-such as duration of mechanical ventilation, cumulative doses of sedative and analgesic drugs, and glycemic variability-and the development of ICU-AW, and to construct a risk prediction nomogram.

Clinical parameters-including demographic characteristics, disease types, critical illness scores, treatment indicators such as mechanical ventilation parameters, laboratory indicators (e.g., inflammatory and biomarkers, metabolic genes), imaging data (muscle and diaphragmatic ultrasound, electrophysiology), molecular biomarkers, and muscle biopsy data-were integrated. Data mining and machine learning techniques were applied to develop an early warning model for ICU-AW based on Cox regression. A logistic regression preliminary screening model was constructed by integrating demographic characteristics and biomarkers. Quantitative parameters from muscle ultrasound (e.g., diaphragmatic excursion, muscle thickness) were incorporated, and dynamic risk assessment was optimized using the Random Forest algorithm. The sensitivity and specificity of the model were evaluated.

1. Study Design: This study employed a multicenter prospective cohort design.

1. Grouping: Pediatric patients undergoing mechanical ventilation were grouped based on the occurrence of ICU-AW at the study endpoint. The diagnostic criteria required meeting any one of the following: i) New-onset muscle dysfunction or weaning difficulty (unrelated to the primary disease, or in non-neuromuscular diseases with normal cardiopulmonary function); ii) Daily MRC score assessment for enrolled PICU patients [for patients under analgesic sedation, drug dosages were adjusted to maintain a modified Comfort-B sedation score between 11 and 23]. Patients with a total MRC score \< 48 on two consecutive assessments 24 hours apart, and excluding those with other neuromuscular disorders, were assigned to the ICU-AW group; those with a total MRC score ≥ 48 were assigned to the non-ICU-AW group; patients who could not be evaluated were excluded. Subtype criteria: CIP: In addition to the above criteria, also meeting iii) Slowed sensory and/or motor nerve conduction velocity; CIM: In addition to criteria i and ii, also meeting iv) Reduced CMAP amplitude with prolonged duration, and normal SNAP amplitude; Mixed: Meeting both criteria iii and iv.
2. Sample Size Estimation (Data Management + Statistical Methods): The occurrence rate of pediatric ICU-AW was estimated at 2%. With a permissible error of 0.03 and a confidence level of 0.95, the sample size was calculated as 1500 cases using PASS software. Assuming a non-response rate of 10% among study subjects, the required sample size was N = 1500 / 0.9 = 1666 cases. Statistical analysis was performed using Stata 15.0 software. Continuous variables following a normal distribution in measurement data are presented as mean (SD); comparisons of means between groups were conducted using one-way and multi-way ANOVA. Those with a skewed distribution are presented as median (range or interquartile range, IQR). Count data (binary or multi-category) are expressed as incidence rates or composition ratios, with comparisons of incidence rates or composition ratios performed using the chi-square test. Logistic regression analysis was used to examine the relationship between biomarkers and the occurrence and development of ICU-AW, as well as the weight and contribution of each variable in the early warning model. ROC curve analysis was employed to assess model performance. A P-value \< 0.01 was considered statistically significant for all analyses.

2. Case Collection and Data Analysis: An electronic database was established to collect clinical data from pediatric patients undergoing mechanical ventilation, including clinical baseline characteristics, laboratory test results, and imaging data. Each research center designated dedicated research personnel to begin enrolling study cases on the same start date (cases already hospitalized on the start date who met the inclusion criteria were enrolled). These personnel were responsible for data cleaning, organization, and standardization to ensure data quality. For all enrolled cases, demographic characteristics, clinical features, and laboratory test information were recorded in CRF forms on the day of enrollment (D0), day 3 (D3), day 10 (D10), the day of PICU discharge (Ddis), or the day of death (DD). Patient examinations for data collection were performed only when deemed appropriate by the physician. If an examination was not performed, the variable value was assumed to be normal or consistent with the previous measurement. Specific recorded information included:

1. Demographic characteristics: Age, sex, length/height, weight, body mass index.
2. Baseline clinical characteristics: Past medical history and underlying diseases, primary diagnosis upon PICU admission, critical illness score (PELOD-2), nutritional status (skinfold thickness/mid-upper arm circumference), CK, myoglobin, neurofilament light chain, a panel of twelve inflammatory factors, urinary titin, blood glucose, MRC muscle strength assessment, gastrocnemius muscle electrophysiology, rectus femoris muscle ultrasound, diaphragm thickness/excursion, treatments received (corticosteroids, e.g., dexamethasone, prednisone, methylprednisolone, hydrocortisone succinate; neuromuscular blocking agents, e.g., vecuronium), mechanical ventilation (duration in days, mode, fraction of inspired oxygen, mean airway pressure, tidal volume ml/kg, respiratory rate, P0.1), days on cardiopulmonary bypass (CPB), days on extracorporeal membrane oxygenation (ECMO), days of rehabilitation, whole-exome sequencing (WES) of family lineage, etc. Patient examinations for data collection were performed only when deemed appropriate by the physician. If an examination was not performed, the variable value was assumed to be normal or consistent with the previous measurement.
3. Determination of inflammatory markers: All inflammatory marker determinations were performed using EDTA-anticoagulated plasma samples. These included: Pro-inflammatory cytokines: IL-1β, IL-6, IL-8, TNF-α, IFN-γ, GM-CSF; Anti-inflammatory cytokines: IL-10, IL-13; Chemoattractant (CX3CL1), soluble intercellular adhesion molecule-1 (sICAM-1), soluble E-selectin (sE-selectin), and soluble P-selectin (sP-selectin). Each inflammatory marker was quantitatively determined using standardized immunoassay methods, such as ELISA (Enzyme-Linked Immunosorbent Assay).
4. Follow-up indicators: Nutritional status (skinfold thickness/mid-upper arm circumference), CK, myoglobin, blood glucose (D0, D3, D7, D10), MRC score, muscle ultrasound, diaphragm ultrasound, electromyography, muscle biopsy (if necessary) (D0, D3, D7, D10).
5. Outcome measures: The occurrence rate of ICU-AW in pediatric patients undergoing mechanical ventilation on day 3 or day 10. Defined as MRC score \< 48, or slowed nerve conduction velocity on electromyography; CIP: Normal or mildly reduced nerve conduction velocity, reduced CMAP amplitude, reduced mixed SNAP amplitude; CIM: Normal or mildly reduced nerve conduction velocity, reduced CMAP amplitude, reduced muscle excitability to direct stimulation, increased CMAP duration, normal SNAP; or definitive diagnosis by muscle biopsy.