Nutrition Screening in the Pediatric Intensive Care Unit: Evaluation of an Electronic Medical Record-Based Tool.

Hospitalized, critically ill children are at increased risk of developing malnutrition. While several pediatric nutrition screening tools exist, none have been validated in the pediatric intensive care units (PICU). The Children's Wisconsin Nutrition Screening Tool (CWNST) is a unique nutrition screening tool that includes the Pediatric Nutrition Screening Tool (PNST) and predictive elements from the electronic medical record and was found to be more sensitive than the PNST in acute care units. The aim of this study was to assess the performance of the tool in detecting possible malnutrition in critically ill children. The data analysis, including the results of the current nutrition screening, diagnosis, and nutrition status was performed on all patients admitted to PICUs at Children's Wisconsin in 2019. All 250 patients with ≥1 nutrition assessment by a dietitian were included. The screening elements that were predictive of malnutrition included parenteral nutrition, positive PNST, and BMI-for-age/weight-for-length z-score. The current screen had a sensitivity of 0.985, specificity of 0.06, positive predictive value (PPV) of 0.249, and negative predictive value of 0.929 compared to the PNST alone which had a sensitivity of 0.1, specificity of 0.981, PPV of 0.658, and NPV of 0.749. However, of the 250 included patients, 97.2% (243) had a positive nutrition screen. The CWNST can be easily applied through EMRs and predicts the nutrition risk in PICU patients but needs further improvement to improve specificity.

Evaluation of an electronic medical record-based Paediatric Nutrition Screening Tool.

BACKGROUND: Nutrition screening is recommended to identify children at risk for malnutrition. A unique screening tool was developed based on American Society for Parenteral and Enteral Nutrition (ASPEN) recommendations and embedded in the electronic medical record to assess for nutrition risk. METHODS: The components of the tool included the Paediatric Nutrition Screening Tool (PNST) and other elements recommended by ASPEN. To evaluate the screening tool, retrospective data were analysed on all patients admitted to acute care units of Children's Wisconsin in 2019. Data collected included nutrition screen results, diagnosis and nutrition status. All patients who received at least one full nutrition assessment by a registered dietitian (RD) were included in analysis. RESULTS: One thousand five hundred seventy-five patients were included in analysis. The following screen elements were significantly associated with a diagnosis of malnutrition: any positive screen (p < 0.001), >2 food allergies (p = 0.009), intubation (p < 0.001), parenteral nutrition (p = 0.005), RD-identified risk (p < 0.001), positive risk per the PNST (p < 0.001), BMI-for-age or weight-for-length z-score (p < 0.001), intake <50% for 3 days (p = 0.012) and NPO > 3 days (p = 0.009). The current screen had a sensitivity of 93.9%, specificity of 20.3%, positive predictive value (PPV) of 30.9% and negative predictive value (NPV) of 89.8%. This is compared with the PNST which had a sensitivity of 32%, specificity of 94.2%, PPV of 71% and NPV of 75.8% in this study population. CONCLUSION: This unique screening tool is useful for predicting nutrition risk and has a greater sensitivity than the PNST alone.

Novel Pediatric Height Outlier Detection Methodology for Electronic Health Records via Machine Learning With Monotonic Bayesian Additive Regression Trees.

OBJECTIVE: To create a new methodology that has a single simple rule to identify height outliers in the electronic health records (EHR) of children. METHODS: We constructed 2 independent cohorts of children 2 to 8 years old to train and validate a model predicting heights from age, gender, race and weight with monotonic Bayesian additive regression trees. The training cohort consisted of 1376 children where outliers were unknown. The testing cohort consisted of 318 patients that were manually reviewed retrospectively to identify height outliers. RESULTS: The amount of variation explained in height values by our model, R2 , was 82.2% and 75.3% in the training and testing cohorts, respectively. The discriminatory ability to assess height outliers in the testing cohort as assessed by the area under the receiver operating characteristic curve was excellent, 0.841. Based on a relatively aggressive cutoff of 0.075, the outlier sensitivity is 0.713, the specificity 0.793; the positive predictive value 0.615 and the negative predictive value is 0.856. CONCLUSIONS: We have developed a new reliable, largely automated, outlier detection method which is applicable to the identification of height outliers in the pediatric EHR. This methodology can be applied to assess the veracity of height measurements ensuring reliable indices of body proportionality such as body mass index.