Estimating risk of prolonged mechanical ventilation after liver transplantation in children: PROVE-ALT score.

BACKGROUND: Children at high risk for prolonged mechanical ventilation (PMV) after liver transplantation (LT) need to be identified early to optimize pulmonary support, allocate resources, and improve surgical outcomes. We aimed to develop and validate a metric that can estimate risk for Prolonged Ventilation After LT (PROVE-ALT). METHODS: We identified preoperative risk factors for PMV by univariable analysis in a retrospective cohort of pediatric LT recipients between 2011 and 2017 (n = 205; derivation cohort). We created the PROVE-ALT score by mapping multivariable logistic regression coefficients as integers, with cutoff values using the Youden Index. We validated the score by C-statistic in a retrospectively collected separate cohort of pediatric LT recipients between 2018 and 2021 (n = 133, validation cohort). RESULTS: Among total 338 patients, 21% (n = 72) were infants; 49% (n = 167) had cirrhosis; 8% (n = 27) required continuous renal replacement therapy (CRRT); and 32% (n = 111) required management in hospital (MIH) before LT. Incidence of PMV post-LT was 20% (n = 69) and 3% (n = 12) required tracheostomy. Independent risk factors (OR [95% CI]) for PMV were cirrhosis (3.8 [1-14], p = .04); age <1-year (8.2 [2-30], p = .001); need for preoperative CRRT (6.3 [1.2-32], p = .02); and MIH before LT (12.4 [2.1-71], p = .004). PROVE-ALT score ≥8 [Range = 0-21] accurately predicted PMV in the validation cohort with 73% sensitivity and 80% specificity (AUC: 0.81; 95% CI: 0.71-0.91). CONCLUSION: PROVE-ALT can predict PMV after pediatric LT with a high degree of sensitivity and specificity. Once externally validated in other centers, PROVE-ALT will empower clinicians to plan patient-specific ventilation strategies, provide parental anticipatory guidance, and optimize hospital resources.

Child Opportunity Index and Pediatric Intensive Care Outcomes: A Multicenter Retrospective Study in the United States.

OBJECTIVES: To evaluate for associations between a child's neighborhood, as categorized by Child Opportunity Index (COI 2.0), and 1) PICU mortality, 2) severity of illness at PICU admission, and 3) PICU length of stay (LOS). DESIGN: Retrospective cohort study. SETTING: Fifteen PICUs in the United States. PATIENTS: Children younger than 18 years admitted from 2019 to 2020, excluding those after cardiac procedures. Nationally-normed COI category (very low, low, moderate, high, very high) was determined for each admission by census tract, and clinical features were obtained from the Virtual Pediatric Systems LLC (Los Angeles, CA) data from each site. INTERVENTIONS: None. MEASUREMENTS AND MAIN RESULTS: Among 33,901 index PICU admissions during the time period, median patient age was 4.9 years and PICU mortality was 2.1%. There was a higher percentage of admissions from the very low COI category (27.3%) than other COI categories (17.2-19.5%, p < 0.0001). Patient admissions from the high and very high COI categories had a lower median Pediatric Index of Mortality 3 risk of mortality (0.70) than those from the very low, low, and moderate COI groups (0.71) ( p < 0.001). PICU mortality was lowest in the very high (1.7%) and high (1.9%) COI groups and highest in the moderate group (2.5%), followed by very low (2.3%) and low (2.2%) ( p = 0.001 across categories). Median PICU LOS was between 1.37 and 1.50 days in all COI categories. Multivariable regression revealed adjusted odds of PICU mortality of 1.30 (95% CI, 0.94-1.79; p = 0.11) for children from a very low versus very high COI neighborhood, with an odds ratio [OR] of 0.996 (95% CI, 0.993-1.00; p = 0.05) for mortality for COI as an ordinal value from 0 to 100. Children without insurance coverage had an OR for mortality of 3.58 (95% CI, 2.46-5.20; p < 0.0001) as compared with those with commercial insurance. CONCLUSIONS: Children admitted to a cohort of U.S. PICUs were often from very low COI neighborhoods. Children from very high COI neighborhoods had the lowest risk of mortality and observed mortality; however, odds of mortality were not statistically different by COI category in a multivariable model. Children without insurance coverage had significantly higher odds of PICU mortality regardless of neighborhood.

Profile of Fluid Exposure and Recognition of Fluid Overload in Critically Ill Children.

OBJECTIVES: Fluid overload is common in the PICU and has been associated with increased morbidity and mortality. It remains unclear whether fluid overload is a surrogate marker for severity of illness and need for increased support, an iatrogenic modifiable risk factor, or a sign of oliguria. The proportions of various fluid intake contributing to fluid overload and its recognition have not been adequately examined. We aimed to: 1) describe the types and amounts of fluid exposure in the PICU and 2) identify the clinicians' recognition of fluid overload. SETTING: Noncardiac PICU in a quaternary care hospital. PATIENTS: Pediatric patients admitted for more than 24 hours. DESIGN: Prospective observational study over 28 days. INTERVENTIONS: Data were collected on the amount and type of fluid exposure-resuscitative boluses, blood products, enteral intake, parenteral nutrition (total parenteral nutrition), or modifiable fluids (IV fluids and medications) indexed to the patients' admission body surface area on days 1 and 3. Charts of patients admitted for 3 days who developed 15% fluid overload were reviewed to assess clinicians' recognition of fluid overload. MEASUREMENTS AND MAIN RESULTS: One hundred two patients were included. Day 1 median fluid exposure was 2,318 mL/m (1,831-3,037 mL/m; 1,646 mL/m [1,296-2,086 mL/m] modifiable fluids). Forty-seven patients (46%) received fluid boluses, and 16 (16%) received blood products. Day 3 median fluid exposure was 2,233 mL/m (1,904-2,556 mL/m; 750 mL/m [375-1,816 mL/m] modifiable fluids). Of the 54 patients, one patient (1.9%) received a fluid bolus and two (3.7%) received blood products. In our cohort, 47 of 54 (87%) had fluid exposure greater than 1,600 mL/m on day 3. Fluid overload was not recognized by the clinicians in 30% of the patients who developed more than 15% fluid overload. CONCLUSIONS: Although resuscitation fluids contributed more to fluid exposure on day 1 compared with day 3, fluid exposure frequently exceeded maintenance requirements on day 3. Fluid overload was not always recognized by PICU practitioners. Further studies to correlate modifiable fluid exposure to fluid overload and explore modifiable practice improvement opportunities are needed.

Comparing Vasoactive-Inotropic Score Reporting Strategies in the PICU Relative to Mortality Risk.

OBJECTIVES: High Vasoactive-Inotropic Scores have demonstrated association with poor outcomes in pediatric cardiac ICUs and are being calculated more frequently in studies of critically ill noncardiac patients. Available studies differ in their approach to assigning Vasoactive-Inotropic Scores, making direct comparisons difficult. The goal of this investigation is to compare multiple approaches to Vasoactive-Inotropic Score assignment to determine their strength of association with mortality in a general pediatric intensive care population. In doing so, we aim to help validate the use of the Vasoactive-Inotropic Score in noncardiac patients and to help inform future studies of the relative strength of available approaches in assigning this score. DESIGN: Retrospective chart review. SETTING: PICU at an academic freestanding children's hospital. PATIENTS: Two-thousand seven-hundred fifty-two consecutive patients admitted over a 17-month time period were screened for receiving inotrope or vasopressor therapies regardless of disease process. Four-hundred seventy-four patients met inclusion criteria. INTERVENTIONS: None. MEASUREMENTS AND MAIN RESULTS: For each patient treated with continuous infusions of vasoactive medications, a Vasoactive-Inotropic Score was calculated (and then recalculated) every time they had a documented dose change. Multiple strategies were evaluated to generate receiver operating characteristic curves in relation to mortality. Area under the curve was greatest when evaluating the maximum Vasoactive-Inotropic Score (Max Any) during the initial treatment course (0.788) with an increasing relative risk as the score increased. The Vasoactive-Inotropic Score at 48 hours after treatment initiation had next highest area under the curve (0.736). Primary diagnosis categories were also analyzed, and area under the curve was greatest for the cardiovascular group (0.879). CONCLUSIONS: Increasing Vasoactive-Inotropic Scores for patients in the PICU are associated with mortality risk. The scoring strategy used can influence the strength of the association, as can the primary diagnosis category.

Hepatobiliary Dysfunction and Disseminated Intravascular Coagulation Increase Risk of Mortality in Pediatric Hemophagocytic Lymphohistiocytosis.

OBJECTIVES: Hemophagocytic lymphohistiocytosis poses significant challenges due to limited tools to guide clinical decisions in a population at high risk of death. We sought to assess whether disseminated intravascular coagulation and hepatobiliary dysfunction, significant comorbidities seen in critical care settings, would identify hemophagocytic lymphohistiocytosis patients with increased risk of mortality. DESIGN: Retrospective chart review. SETTING: Single-center PICU. PATIENTS: All patients admitted to a tertiary care children's hospital diagnosed with hemophagocytic lymphohistiocytosis from 2005 to 2012. INTERVENTIONS: None. MEASUREMENTS AND MAIN RESULTS: Forty-three patients were diagnosed with hemophagocytic lymphohistiocytosis with median age of 61 months. The 5-year overall survival was 51% (22/43). Univariate analyses revealed ferritin levels greater than 10,000 (ng/mL), international normalized ratio greater than 1.5, or platelet counts less than 100,000/µL at initiation of dexamethasone were individually associated with mortality. Development of disseminated intravascular coagulation, hepatobiliary dysfunction, or both increased the likelihood of death in hemophagocytic lymphohistiocytosis patients (relative risk; 95% CI) (6; 1.4-34; p < 0.05), (4.1; 1.8-10; p < 0.05), and (7.5; 1.8-42; p < 0.05). Of 12 autopsies performed, 75% had at least one active infection, 66% had chronic lymphopenia, 50% had lymphocyte depletion in the spleen, thymus, or bone marrow, 42% had evidence of microvascular thrombosis, and 92% had evidence of hepatocellular injury. CONCLUSIONS: Hemophagocytic lymphohistiocytosis continues to have high mortality with hemophagocytic lymphohistiocytosis-1994/2004 (dexamethasone/etoposide), the current standard of care for all children with hemophagocytic lymphohistiocytosis. Hemophagocytic lymphohistiocytosis patients who developed disseminated intravascular coagulation, hepatobiliary dysfunction, or both had higher risk of death with mortalities of 60%, 77%, and 77%, respectively. Phenotypic classifications are urgently needed to guide individualized treatment strategies to improve outcomes for children with hemophagocytic lymphohistiocytosis.

The Use of Nesiritide in Children With Congenital Heart Disease.

OBJECTIVE: We evaluated the use of nesiritide in children with critical congenital heart disease, pulmonary congestion, and inadequate urine output despite conventional diuretic therapy. DESIGN: We conducted a retrospective analysis of 26 consecutive patients, comprising 37 infusions occurring during separate hospitalizations. Hemodynamic variables, urine output, and serum creatinine levels were monitored prior to and throughout the duration of therapy with nesiritide. In addition, the stage of acute kidney injury was determined prior to and throughout the duration of the therapy using a standardized definition of acute kidney injury-The Kidney Disease: Improving Global Outcomes criteria. SETTING: Cardiac ICU. PATIENTS: Pediatric patients with critical congenital heart disease, pulmonary congestion, and inadequate urinary output despite diuretic therapy. INTERVENTION: Nesiritide infusion. MEASUREMENTS AND MAIN RESULTS: The use of nesiritide was associated with a significant decrease in the central venous pressure and heart rate with a trend toward a significant increase in urine output. During the course of therapy with nesiritide, the serum creatinine and stage of acute kidney injury decreased significantly. The decrease in stage of acute kidney injury became significant by day 4 (p = 0.006) and became more significant with time (last day of therapy compared with baseline; p < 0.001). During 12 of the 37 infusions, the stage of acute kidney injury decreased by two or more (p < 0.001). CONCLUSIONS: Nesiritide had a favorable impact on hemodynamics and urine output in children with critical congenital heart disease and pulmonary congestion, and there was no worsening of renal function.

Fluid Overload and Kidney Injury Score: A Multidimensional Real-Time Assessment of Renal Disease Burden in the Critically Ill Patient.

OBJECTIVE: Interruptive acute kidney injury alerts are reported to decrease acute kidney injury-related mortality in adults. Critically ill children have multiple acute kidney injury risk factors; although recognition has improved due to standardized definitions, subtle changes in serum creatinine make acute kidney injury recognition challenging. Age and body habitus variability prevent a uniform maximum threshold of creatinine. Exposure of nephrotoxic medications is common but not accounted for in kidney injury scores. Current severity of illness measures do not include fluid overload, a well-described mortality risk factor. We hypothesized that a multidimensional measure of renal status would better characterize renal severity of illness while maintaining or improving on correlation measures with adverse outcomes, when compared with traditional acute kidney injury staging. DESIGN: A novel, real-time, multidimensional, renal status measure, combining acute kidney injury, fluid overload greater than or equal to 15%, and nephrotoxin exposure, was developed (Fluid Overload Kidney Injury Score) and prospectively applied to all patient encounters. Peak Fluid Overload Kidney Injury Score values prior to discharge or death were used to measure correlation with outcomes. SETTING: Quarternary PICU of a freestanding children's hospital. PATIENTS: All patients admitted over 18 months. INTERVENTION: None. RESULTS: Peak Fluid Overload Kidney Injury Score ranged between 0 and 14 in 2,830 PICU patients (median age, 5.5 yr; interquartile range, 1.3-12.9; 55% male), 66% of patients had Fluid Overload Kidney Injury Score greater than or equal to 1. Fluid Overload Kidney Injury Score was independently associated with PICU mortality and PICU and hospital length of stay when controlled for age, Pediatric Risk of Mortality-3, ventilator, pressor, and renal replacement therapy use (p = 0.047). Mortality increased from 1.5% in Fluid Overload Kidney Injury Score 0 to 40% in Fluid Overload Kidney Injury Score 8+. When urine output points were excluded, Fluid Overload Kidney Injury Score was more strongly correlated with mortality than fluid overload or acute kidney injury definitions alone. CONCLUSION: A multidimensional score of renal disease burden was significantly associated with adverse PICU outcomes. Further studies will evaluate Fluid Overload Kidney Injury Score as a warning and decision support tool to impact patient-centered outcomes.

The use of nesiritide in patients with critical cardiac disease.

OBJECTIVE: We evaluated the use of nesiritide in children with critical CHD, pulmonary congestion, and inadequate urine output despite undergoing conventional diuretic therapy. DESIGN: We conducted a retrospective analysis of 11 patients with critical CHD, comprising 18 infusions, each of which occurred during separate hospitalisations. Haemodynamic parameters were assessed, and the stage of acute kidney injury was determined before and throughout the duration of therapy using a standardised definition of acute kidney injury - The Kidney Disease: Improving Global Outcomes criteria. Patients Children with critical CHD, pulmonary congestion, and inadequate urinary output despite undergoing diuretic therapy were included. Measurements and main results The use of nesiritide was associated with a significant decrease in the maximum and minimum heart rate values and with a trend towards a significant decrease in maximum systolic blood pressure and maximum and minimum central venous pressures. Urine output increased but was not significant. Serum creatinine levels decreased significantly during the course of therapy (-0.26 mg/dl [-0.50, 0.0], p=0.02), and the number of patients who experienced a decrease in the stage of acute kidney injury of 2 or more - where a change in the stage of acute kidney disease of 2 or more was possible, that is, baseline stage >1 - was highly significant (five of 12 patients, 42%, p<0.001). CONCLUSIONS: Nesiritide had a favourable impact on haemodynamics, and its use was not associated with deterioration of renal function in patients with critical CHD.

Resuscitation Bundle in Pediatric Shock Decreases Acute Kidney Injury and Improves Outcomes.

OBJECTIVE: To investigate the impact of an early emergency department (ED) protocol-driven resuscitation (septic shock protocol [SSP]) on the incidence of acute kidney injury (AKI). STUDY DESIGN: This was a retrospective pediatric cohort with clinical sepsis admitted to the pediatric intensive care unit (PICU) from the ED before (2009, PRE) and after (2010, POST) implementation of the SSP. AKI was defined by pRIFLE (pediatric version of the Risk of renal dysfunction; Injury to kidney; Failure of kidney function; Loss of kidney function, End-stage renal disease creatinine criteria). RESULTS: A total of 202 patients (PRE, n = 98; POST, n = 104) were included (53% male, mean age 7.7 ± 5.6 years, mean Pediatric Logistic Organ Dysfunction [PELOD] 8.9 ± 12.7, mean Pediatric Risk of Mortality score 5.3 ± 13.9). There were no differences in demographics or illness severity between the PRE and POST groups. POST was associated with decreased AKI (54% vs 29%, P < .001), renal-replacement therapy (4 vs 0, P = .04), PICU, and hospital lengths of stay (LOS) (1.9 ± 2.3 vs 4.5 ± 7.6, P < .01; 6.3 ± 5.1 vs 15.3 ± 16.9, P < .001, respectively), and mortality (10% vs 3%, P = .037). The SSP was independently associated with decreased AKI when we controlled for age, sex, and PELOD (OR 0.27, CI 0.13-0.56). In multivariate analyses, the SSP was independently associated with shorter PICU and hospital LOS when we controlled for AKI and PELOD (P = .02, P < .001, respectively). CONCLUSION: A protocol-driven implementation of a resuscitation bundle in the pediatric ED decreased AKI and need for renal-replacement therapy, as well as PICU and hospital LOS and mortality.

Using Time Series Analysis to Predict Cardiac Arrest in a PICU.

OBJECTIVES: To build and test cardiac arrest prediction models in a PICU, using time series analysis as input, and to measure changes in prediction accuracy attributable to different classes of time series data. DESIGN: Retrospective cohort study. SETTING: Thirty-one bed academic PICU that provides care for medical and general surgical (not congenital heart surgery) patients. SUBJECTS: Patients experiencing a cardiac arrest in the PICU and requiring external cardiac massage for at least 2 minutes. INTERVENTIONS: None. MEASUREMENTS AND MAIN RESULTS: One hundred three cases of cardiac arrest and 109 control cases were used to prepare a baseline dataset that consisted of 1,025 variables in four data classes: multivariate, raw time series, clinical calculations, and time series trend analysis. We trained 20 arrest prediction models using a matrix of five feature sets (combinations of data classes) with four modeling algorithms: linear regression, decision tree, neural network, and support vector machine. The reference model (multivariate data with regression algorithm) had an accuracy of 78% and 87% area under the receiver operating characteristic curve. The best model (multivariate + trend analysis data with support vector machine algorithm) had an accuracy of 94% and 98% area under the receiver operating characteristic curve. CONCLUSIONS: Cardiac arrest predictions based on a traditional model built with multivariate data and a regression algorithm misclassified cases 3.7 times more frequently than predictions that included time series trend analysis and built with a support vector machine algorithm. Although the final model lacks the specificity necessary for clinical application, we have demonstrated how information from time series data can be used to increase the accuracy of clinical prediction models.

Time series analysis as input for clinical predictive modeling: modeling cardiac arrest in a pediatric ICU.

BACKGROUND: Thousands of children experience cardiac arrest events every year in pediatric intensive care units. Most of these children die. Cardiac arrest prediction tools are used as part of medical emergency team evaluations to identify patients in standard hospital beds that are at high risk for cardiac arrest. There are no models to predict cardiac arrest in pediatric intensive care units though, where the risk of an arrest is 10 times higher than for standard hospital beds. Current tools are based on a multivariable approach that does not characterize deterioration, which often precedes cardiac arrests. Characterizing deterioration requires a time series approach. The purpose of this study is to propose a method that will allow for time series data to be used in clinical prediction models. Successful implementation of these methods has the potential to bring arrest prediction to the pediatric intensive care environment, possibly allowing for interventions that can save lives and prevent disabilities. METHODS: We reviewed prediction models from nonclinical domains that employ time series data, and identified the steps that are necessary for building predictive models using time series clinical data. We illustrate the method by applying it to the specific case of building a predictive model for cardiac arrest in a pediatric intensive care unit. RESULTS: Time course analysis studies from genomic analysis provided a modeling template that was compatible with the steps required to develop a model from clinical time series data. The steps include: 1) selecting candidate variables; 2) specifying measurement parameters; 3) defining data format; 4) defining time window duration and resolution; 5) calculating latent variables for candidate variables not directly measured; 6) calculating time series features as latent variables; 7) creating data subsets to measure model performance effects attributable to various classes of candidate variables; 8) reducing the number of candidate features; 9) training models for various data subsets; and 10) measuring model performance characteristics in unseen data to estimate their external validity. CONCLUSIONS: We have proposed a ten step process that results in data sets that contain time series features and are suitable for predictive modeling by a number of methods. We illustrated the process through an example of cardiac arrest prediction in a pediatric intensive care setting.

Fluid Overload and Kidney Injury Score as a Predictor for Ventilator-Associated Events.

Objective: The Pediatric and Neonatal Working group developed new ventilator associated events (VAE) definitions for children and neonates. VAE includes ventilator-associated condition (VAC), infection-related ventilator-associated complication (IVAC), and ventilator-associated pneumonia (VAP). Acute kidney injury (AKI) and fluid overload (FO) have been associated with worse clinical outcomes of ventilated children. Fluid Overload and Kidney Injury Score (FOKIS) is an automatically calculated score that combines AKI and FO in one numeric quantifiable metric. This study analyzed the association between FOKIS and VAE. Design: Retrospective matched case control study. Setting: A freestanding children's hospital. Patients: A total of 168 who were ventilated > 2 days. Interventions: None. Measurements and Main Results: We identified 42 VAC cases (18 IVAC and 24 non-infection-related VAC cases). Controls were matched to cases for age, immunocompromised status and ventilator days prior to VAC. VAC cases had longer ICU days, median (IQR), 28.5 (15, 47) vs. controls 11 (6, 16), p < 0.001; longer ventilation days, 19.5 (13, 32) vs. 9 (4,13), p < 0.001; and higher hospital mortality, 45.2 vs. 18%, p < 0.001. VACs had a higher incidence of AKI, 85.7 vs. 47.3%, p < 0.001; higher peak daily FO% within 3 days preceding VAC, mean (SD), 8.1(7.8) vs. 4.1 (3.4), p < 0.005; and higher peak FOKIS, 6.4(3.8) vs. 3.7(2.8), (p < 0.001). Multivariate regression model adjusted for severity of illness identified peak FOKIS (odds ratio [OR] 1.29, 95%CI: 1.14-1.48, p < 0.001) and peak inspiratory pressure (OR 1.08, 95%CI: 1.02-1.15, p = 0.007) as risk factors for VAC. Conclusions: The FOKIS and its clinical variables were associated risk factors for ventilator-associated events. Further studies will determine the utility of FOKIS as a predictor for VAEs.

Epilepsy is associated with ventricular alterations following convulsive status epilepticus in children.

OBJECTIVE: Convulsive status epilepticus can exert profound cardiovascular effects in adults including ventricular depolarization-repolarization abnormalities. Whether status epilepticus adversely affects ventricular electrical properties in children is less understood. Therefore, we sought to characterize ventricular alterations and the associated clinical factors in children following convulsive status epilepticus. METHODS: We conducted a 2-year retrospective, case-control study. Children between 1 month and 21 years of age were included if they were admitted to the pediatric intensive care unit with primary diagnosis of convulsive status epilepticus and had 12-lead electrocardiogram (ECG) within 24 hours of admission. Children with heart disease, ion channelopathy, or on vasoactive medications were excluded. Age-matched control subjects had no history of seizures or epilepsy. The primary outcome was ventricular abnormalities represented by ST segment changes, abnormal T wave, QRS axis deviation, and corrected QT (QTc) interval prolongation. The secondary outcomes included QT/RR relationship, beat-to-beat QTc interval variability, ECG interval measurement between groups, and clinical factors associated with ECG abnormalities. RESULTS: Of 317 eligible children, 59 met the inclusion criteria. History of epilepsy was present in 31 children (epileptic) and absent in 28 children (non-epileptic). Compared with the control subjects (n = 31), the status epilepticus groups were more likely to have an abnormal ECG with overall odds ratio of 3.8 and 7.0 for the non-epileptic and the epileptic groups respectively. Simple linear regression analysis demonstrated that children with epilepsy exhibited impaired dependence and adaptation of the QT interval on heart rate. Beat-to-beat QTc interval variability, a marker of ventricular repolarization instability, was increased in children with epilepsy. SIGNIFICANCE: Convulsive status epilepticus can adversely affect ventricular electrical properties and stability in children, especially those with epilepsy. These findings suggest that children with epilepsy may be particularly vulnerable to seizure-induced arrhythmias. Therefore postictal cardiac surveillance may be warranted in this population.

Organ dysfunction is associated with hyperglycemia in critically ill children.

OBJECTIVE: To determine the association between organ dysfunction and hyperglycemia in critically ill children receiving intravenous insulin. DESIGN: Retrospective chart review (cohort study). SETTING: Pediatric intensive care unit in a university hospital. PATIENTS: n = 110 patients; inclusion criteria: ICU hospitalization from May 2005 to May 2006; insulin drip to manage hyperglycemia. EXCLUSION CRITERIA: insulin drip <48 h; diabetic patients. MEASUREMENTS: Duration of hyperglycemia: sum of hours of hyperglycemia (> or =126 mg/dl). Hypoglycemia (blood glucose <40 mg/dl). Organ dysfunction was determined per International Pediatric Sepsis Consensus Conference criteria. Multiple logistic regression models determined the association between > or =3 compared to <3 organ dysfunctions and hyperglycemia, hypoglycemia, and mortality, after adjustment for confounding variables (age, gender, PRISM score, vasopressors, steroids). MAIN RESULTS: Organ dysfunction > or =3 compared to <3 after adjustment for confounders was associated with intermittent hyperglycemia of > or =24 h (OR 6.1, CI 1.8-21.2; p = 0.004). Hyperglycemia trended towards significance with mortality [3.2 (CI 0.9-11.6, p = 0.079)]. Hypoglycemia, after adjusting for the above, was not associated with mortality. CONCLUSIONS: Organ dysfunction (> or =3 versus <3) was significantly associated with hyperglycemia for > or =24 h and hypoglycemia. Hyperglycemia trended toward significance with mortality in critically ill children. There was no association between hypoglycemia and mortality.

An alternative treatment strategy for pump thrombus in the DeBakey VAD Child: use of clopidogrel as a thrombolytic agent.

Our institution has used the DeBakey VAD Child as a bridge to heart transplantation in select pediatric patients. Pump thrombus is a potentially serious complication with few available treatment options. Only surgical device exchange or the use of tissue plasminogen activator are reported in the literature. We report the use of clopidogrel as an alternative thrombolytic agent in the presence of pump thrombus in an adolescent patient with a DeBakey VAD Child heart pump.

Generating a mortality model from a pediatric ICU (PICU) database utilizing knowledge discovery.

Current models for predicting outcomes are limited by biases inherent in a priori hypothesis generation. Knowledge discovery algorithms generate models directly from databases, minimizing such limitations. Our objective was to generate a mortality model from a PICU database utilizing knowledge discovery techniques. The database contained 5067 records with 192 clinically relevant variables. It was randomly split into training (75%) and validation (25%) groups. We used decision tree induction to generate a mortality model from the training data, and validated its performance on the validation data. The original PRISM algorithm was used for comparison. The decision tree model contained 25 variables and predicted 53/88 deaths; 29 correctly (Sens:33%, Spec:98%, PPV:54%). PRISM predicted 27/88 deaths correctly (Sens:30%, Spec:98%, PPV:51%). Performance difference between models was not significant. We conclude that knowledge discovery algorithms can generate a mortality model from a PICU database, helping establish validity of such tools in the clinical medical domain.