An explainable machine learning model for prediction of high-risk nonalcoholic steatohepatitis.

Early identification of high-risk metabolic dysfunction-associated steatohepatitis (MASH) can offer patients access to novel therapeutic options and potentially decrease the risk of progression to cirrhosis. This study aimed to develop an explainable machine learning model for high-risk MASH prediction and compare its performance with well-established biomarkers. Data were derived from the National Health and Nutrition Examination Surveys (NHANES) 2017-March 2020, which included a total of 5281 adults with valid elastography measurements. We used a FAST score ≥ 0.35, calculated using liver stiffness measurement and controlled attenuation parameter values and aspartate aminotransferase levels, to identify individuals with high-risk MASH. We developed an ensemble-based machine learning XGBoost model to detect high-risk MASH and explored the model's interpretability using an explainable artificial intelligence SHAP method. The prevalence of high-risk MASH was 6.9%. Our XGBoost model achieved a high level of sensitivity (0.82), specificity (0.91), accuracy (0.90), and AUC (0.95) for identifying high-risk MASH. Our model demonstrated a superior ability to predict high-risk MASH vs. FIB-4, APRI, BARD, and MASLD fibrosis scores (AUC of 0.95 vs. 0.50, 0.50, 0.49 and 0.50, respectively). To explain the high performance of our model, we found that the top 5 predictors of high-risk MASH were ALT, GGT, platelet count, waist circumference, and age. We used an explainable ML approach to develop a clinically applicable model that outperforms commonly used clinical risk indices and could increase the identification of high-risk MASH patients in resource-limited settings.

EVIDENCE FOR BENEFICIAL USE OF THE COMPENSATORY RESERVE MEASUREMENT IN GUIDING INTRAOPERATIVE RESUSCITATION: A PROSPECTIVE COHORT STUDY OF ORTHOTOPIC LIVER TRANSPLANT RECIPIENTS.

ABSTRACT: Introduction: The compensatory reserve measurement (CRM) is a continuous noninvasive monitoring technology that provides an assessment of the integrated capacity of all physiological mechanisms associated with responses to a hypovolemic stressor such as hemorrhagic shock. No prior studies have analyzed its use for intraoperative resuscitation guidance. Methods: A prospective observational study was conducted of 23 patients undergoing orthotopic liver transplant. Chart review was performed to identify timing of various intraoperative events. Data were compared based on predefined thresholds for existence of hemorrhagic shock: CRM lower than 40%, systolic blood pressure (SBP) lower than 90 mm Hg (SBP90), and heart rate (HR) higher than 100 beats per minute (HR100). Regression analysis was performed for predicting resuscitation events, and nonlinear eXtreme Gradient Boosting (XGBoost) models were used to compare CRM with standard vital sign measures. Results: Events where CRM dropped lower than 40% were 2.25 times more likely to lead to an intervention, whereas HR100 and SBP90 were not associated with intraoperative interventions. XGBoost prediction models showed superior discriminatory capacity of CRM alone compared with the model with SBP and HR and no difference when all three were combined (CRM-HR-SBP). All XGBoost models outperformed equivalent linear regression models. Conclusion: These results demonstrate that CRM can provide an adjunctive clinical tool that can augment early and accurate of hemodynamic compromise and promote goal-directed resuscitation in the perioperative setting.

Long-Term Neurocognitive Outcomes in Pediatric Nonfatal Drowning: Results of a Family Caregiver Survey.

BACKGROUND: Drowning is a leading cause of brain injury in children. Long-term outcome data for drowning survivors are sparse. This study reports neurocognitive outcomes for 154 children hospitalized following drowning. METHODS: A survey for parent caregivers was distributed online. Likert scale items assessed 10 outcome variables in four domains: motor (three), perception (three), language (three), and social/emotional (one). Cluster analysis, outcome relative risk, and descriptive statistics were applied. RESULTS: Of 208 surveys received, 154 met inclusion criteria. Coma was the most common admission status (n = 137). Cluster analysis identified three outcome groups: Mild (n = 39), Moderate (n = 75), and Severe (n = 40). Motor impairment with cognitive and perceptual sparing (deefferentation) was present in Moderate (P < 1 × 10-26) and Severe (P < 1 × 10-12) but absent in Mild. Locked-in state was endorsed in both Moderate (83%) and Severe (70%). The strongest predictor of good outcome (Mild) was hospitalization with no medical intervention (relative risk [RR] = 6.7). Responsivity on admission (RR = 4.2) or discharge (RR = 12.22) also predicted good outcome. In-hospital prognostication and counseling predicted outcome weakly (RR = 1.3) or not at all. CONCLUSIONS: Long-term outcomes in pediatric drowning ranged widely. Overall, motor impairments exceeded perceptual or cognitive (P < 1 × 10-18), with "locked-in state" endorsed in most (93 of 154). The strongest predictors of good outcome were the lack of necessity for interventions and responsivity on admission or discharge. The eponym "Conrad syndrome" is proposed for locked-in state following nonfatal drowning in children.