The Pediatric Data Science and Analytics Subgroup of the Pediatric Acute Lung Injury and Sepsis Investigators Network: Use of Supervised Machine Learning Applications in Pediatric Critical Care Medicine Research.

OBJECTIVE: Perform a scoping review of supervised machine learning in pediatric critical care to identify published applications, methodologies, and implementation frequency to inform best practices for the development, validation, and reporting of predictive models in pediatric critical care. DESIGN: Scoping review and expert opinion. SETTING: We queried CINAHL Plus with Full Text (EBSCO), Cochrane Library (Wiley), Embase (Elsevier), Ovid Medline, and PubMed for articles published between 2000 and 2022 related to machine learning concepts and pediatric critical illness. Articles were excluded if the majority of patients were adults or neonates, if unsupervised machine learning was the primary methodology, or if information related to the development, validation, and/or implementation of the model was not reported. Article selection and data extraction were performed using dual review in the Covidence tool, with discrepancies resolved by consensus. SUBJECTS: Articles reporting on the development, validation, or implementation of supervised machine learning models in the field of pediatric critical care medicine. INTERVENTIONS: None. MEASUREMENTS AND MAIN RESULTS: Of 5075 identified studies, 141 articles were included. Studies were primarily (57%) performed at a single site. The majority took place in the United States (70%). Most were retrospective observational cohort studies. More than three-quarters of the articles were published between 2018 and 2022. The most common algorithms included logistic regression and random forest. Predicted events were most commonly death, transfer to ICU, and sepsis. Only 14% of articles reported external validation, and only a single model was implemented at publication. Reporting of validation methods, performance assessments, and implementation varied widely. Follow-up with authors suggests that implementation remains uncommon after model publication. CONCLUSIONS: Publication of supervised machine learning models to address clinical challenges in pediatric critical care medicine has increased dramatically in the last 5 years. While these approaches have the potential to benefit children with critical illness, the literature demonstrates incomplete reporting, absence of external validation, and infrequent clinical implementation.

Nonresuscitation Fluid Accumulation and Outcomes After Pediatric Cardiac Surgery: Single-Center Retrospective Cohort Study.

OBJECTIVES: Postoperative patients after congenital cardiac surgery are at high risk of fluid overload (FO), which is known to be associated with poor outcomes. "Fluid creep," or nonresuscitation IV fluid in excess of maintenance requirement, is recognized as a modifiable factor associated with FO in the general PICU population, but has not been studied in congenital cardiac surgery patients. Our objective was to characterize fluid administration after congenital cardiac surgery, quantify fluid creep, and the association between fluid creep, FO, and outcome. DESIGN: Retrospective, observational cohort study. SETTING: Single-center urban mixed-medical and cardiac PICU. PATIENTS: Patients admitted to the PICU after cardiac surgery between January 2010 and December 2020. INTERVENTIONS: None. MEASUREMENTS AND MAIN RESULTS: There were 1,459 postoperative encounters with 1,224 unique patients. Total fluid intake was greater than maintenance requirements on 3,103 of 4,661 patient days (67%), with fluid creep present on 2,624 patient days (56%). Total nonresuscitation intake was higher in patients with FO (defined as cumulative fluid balance 10% above body weight) versus those without. Fluid creep was higher among patients with FO than those without for each of the first 5 days postoperatively. Each 10 mL/kg of fluid creep in the first 24 hours postoperatively was associated with 26% greater odds of developing FO (odds ratio [OR] 1.26; 95% CI, 1.17-1.35) and 17% greater odds of mortality (OR 1.17; 95% CI, 1.05-1.30) after adjusting for risk of mortality based on surgical procedure, age, and day 1 resuscitation volume. Increasing fluid creep in the first 24 hours postoperatively was associated with increased postoperative duration of mechanical ventilation and PICU length of stay. CONCLUSIONS: Fluid creep is present on most postoperative days for pediatric congenital cardiac surgery patients, and fluid creep is associated with higher-risk procedures. Fluid creep early in the postoperative PICU stay is associated with greater odds of FO, mortality, length of mechanical ventilation, and PICU length of stay. Fluid creep may be under-recognized in this population and thus present a modifiable target for intervention.

Recognizing Pediatric ARDS: Provider Use of the PALICC Recommendations in a Tertiary Pediatric ICU.

BACKGROUND: For almost 50 years, pediatricians used adult guidelines to diagnose ARDS. In 2015, specific criteria for pediatric ARDS were defined. However, it remains unclear how frequently providers recognize pediatric ARDS and whether recognition affects adherence to consensus recommendations. METHODS: This was a mixed-method, retrospective study of mechanically ventilated pediatric subjects after the release of the pediatric ARDS recommendation statement. Pediatric ARDS cases were identified according to the new criteria. Provider recognition was defined by documentation in the medical record. Pediatric ARDS subjects with and without provider recognition were compared quantitatively according to clinical characteristics, adherence to lung-protective ventilation (LPV), adjunctive therapies, and outcomes. A qualitative document analysis (QDA) was performed to evaluate knowledge and beliefs surrounding the Pediatric Acute Lung Injury Consensus Conference recommendations. RESULTS: Of 1,983 subject encounters, pediatric ARDS was identified in 321 (16%). Provider recognition was present in 97 (30%) cases and occurred more often in subjects who were older, had worse oxygenation deficits, or were bone marrow transplant recipients. Recognition rates increased each studied year. LPV practices did not differ based on provider recognition; however, subjects who received it were more likely to experience permissive hypoxemia and adherence to extrapulmonary recommendations. Ultimately, there was no differences in outcomes between the provider recognition and non-provider recognition groups. Three themes emerged from the QDA: (1) pediatric ARDS presents within a complex, multidimensional context, with potentially competing organ system failures; (2) similar to historical conceptualizations, pediatric ARDS was often considered a visual diagnosis, with measures of oxygenation unreferenced; and (3) emphasis was placed on non-evidence-based interventions, such as pulmonary clearance techniques, rather than on consensus recommendations. CONCLUSIONS: Among mechanically ventilated children, pediatric ARDS was common but recognized in a minority of cases. Potential opportunities, such as an opt-out approach to LPV, may exist for improved dissemination and implementation of recommended best practices.