309 - Predictors of Gastrostomy Tube (G-tube) Risk at Discharge in Infants with Chronic Tube Feeding and Development of a Gastrostomy Risk Calculator

Publication Number: 3298.309

Enas Alshaikh, Nationwide Children's Hospital, columbus, OH, United States; Zakia Sultana, Nationwide Children's Hospital, Columbus, OH, United States; Sudarshan R. Jadcherla, Nationwide Children's Hospital, Columbus, OH, United States

Background: Nasogastric tube feeding remains a common method of providing enteral nutrition to neonates admitted to the NICU, particularly among premature infants. Enteral and oral feeding milestones are dependent on both, infant- and process-driven factors. Achieving safe and effective oral feeding are important consideration for discharge from the NICU. However, for infants with persistent feeding difficulties, gastrostomy tube (G-tube) placement offers a viable alternative, but the timing and decisions remain imprecise. Early identification of infants at high risk for G-tube placement can provide anticipatory guidance to the parents and providers, optimize feeding outcomes, and facilitate timely interventions.

Objective: To:1) examine clinical characteristics and feeding milestones in relation to discharge outcomes, 2) identify risk factors for G-tube prediction in NICU infants, and 3) develop an easily usable G-tube prediction risk calculator.

Design/Methods: Data from 1552 nasogastric (NG) tube-fed infants (48% female, gestational age < 37 weeks) referred to our infant feeding disorders program were analyzed. The dataset included demographic, growth, feeding milestone, and comorbidity variables. A logistic regression model was developed to predict G-tube placement at discharge. Using Least Absolute Shrinkage and Selection Operator (LASSO) regression with L1 regularization on 114 set of variables, 10 clinically relevant predictors were identified. The dataset was randomly partitioned into training (70%) and testing (30%) subsets, and model performance was assessed using sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV).

Results: Out of 1552 chronically tube-fed complex infants with missed full oral milestone and had complete data, 38% were discharged on G-tube, and 1.4% underwent fundoplication. Clinical characteristics of infants overall, PO or G-tube at discharge are shown in Table-1. The predictors of G-tube at discharge and the G-tube calculator are shown in Table-2 and ROC curve model in Figure-1. Using random partitioning model to train (70%) and to test (30%) and classification threshold of 0.6, sensitivity was 74%, specificity was 66%, PPV was 22% and NPV was 95%.

Conclusion(s): By understanding the physiologic and clinical predictors of G-tube dependence, clinicians can provide anticipatory guidance to families and tailor care strategies to support and coordinate development of feeding skills and milestones. Use of G-tube calculator can be easier in the evaluation of therapies and counselling.

Table 1. Clinical characteristics among oral fed and Gastrostomy tube (G-tube) fed infants at discharge
Data is presented as mean ± SD, %, and median (min-max).

Table 2. Association between clinical characteristics and Gastrostomy tube (G-tube) prediction at discharge
Odds ratios (OR) and confidence intervals (CI) were estimated using unadjusted and multivariable logistic regression models to predict exclusive G-tube feeding at discharge. Unadjusted models assessed each predictor individually, while adjusted models included all covariates simultaneously. Variables X1-X10​ represent subject characteristics included in the regression formula for brevity. For example, an infant with the following characteristics-2 parental complications, birth weight z-score of -1, female sex, 5-minute APGAR of 4, presence of congenital anomalies and neurological disorders, achievement of full enteral feeding by day 14 of life, a 10-day interval from first enteral feeding to first oral feeding, and first oral feeding achieved at 34 weeks postmenstrual age-had an estimated 81% probability of being discharged with a G-tube.

Figure 1. Receiver operating characteristic (ROC) curve of modeling probability of Gastrostomy tube (G-tube) at discharge
The ROC curve demonstrates the trade-off between sensitivity and 1 − specificity. The area under the curve (AUC) was 0.75, indicating moderate to good discriminatory performance. The diagonal gray line indicates the reference line of random classification (AUC = 0.5).