108 - Temporal Changes of Heart Rate Variability as a Predictor Tool for Necrotizing Enterocolitis Using Bayes Rule Analysis.

Publication Number: 2103.108

Peter J. Porcelli, Wake Forest School of Medicine of Wake Forest Baptist Medical Center, Winston-Salem, NC, United States; Will King, Medical Predictive Science Corporation, Charlottesville, VA, United States; Haley McCalpin, Boston Children's Hospital, Boston, MA, United States; Stephen M. Downs, Wake Forest School of Medicine of Wake Forest Baptist Medical Center, Winston Salem, NC, United States; Parvesh Garg, Wake Forest School of Medicine of Wake Forest Baptist Medical Center, Winston Salem, NC, United States

Background: Necrotizing enterocolitis (NEC) is the most common serious gastrointestinal disease of the newborn with an incidence up to 13% in VLBW infants and a 20-50% mortality rate. Also, medical costs for VLBW infants with NEC averages $180k, while surgical cases can reach $400k.
Although many treatment options exist, predictive tools for NEC are rare. Earlier suspicion and treatment of sepsis in preterm infants by as little as 3 hours reduced morbidity and medical costs, and earlier NEC diagnosis could have a similar beneficial effect. We have previously shown increased heart rate variability (HeRO) scores are effective early predictors of late onset sepsis and meningitis.

Objective: Apply Bayes decision support information, and compare to frequentist analysis, to describe heart rate variability as a temporal, stepwise predictor of necrotizing enterocolitis.

Design/Methods: We collected demographic, therapeutic, survival, and HeRO data from infants diagnosed with NEC at Atrium/Brenner Children's Hospital from 8/2020-5/2024. During each of the 16 days leading up to and including NEC diagnosis, we calculated the mean HeRO Score for each patient with NEC and compared those with scores from the general NICU population using both traditional metrics (e.g. sensitivity and odds ratio) and decision support metrics (positive and negative likelihood ratios). Because the HeRO score represents multiples of infectious risk, we identified a HeRO score > 1 as a positive infectious risk for the likelihood and odds ratios.

Results: Demographic data for 51 patients showed a median birth weight of 1020 gm and 28wk-5d gestational age (Table 1). The survival rate was 78.4%, and 41.2% were surgical cases. The likelihood ratio indicating a HeRO score >1 rose over the 3 days before NEC diagnosis along with an elevated odds ratio of 10.4 the day before NEC diagnosis (Table 2). Mean HeRO scores increased in surgical patients through day 6 before NEC diagnosis, but were consistently elevated in most medical NEC cases with a score near 2 (Figure).

Conclusion(s): In this representative sample, HeRO Scores, and the resulting positive likelihood ratio, and odds ratio all rose during the days before NEC onset, with different trends between surgical and medical cases. These trends support loss of heart rate variability as an inflammatory marker which may prove useful for earlier diagnosis of NEC. Likelihood ratios, as per Bayes rule, may provide more flexibility compared to frequentist techniques. As HeRO offers continuous noninvasive monitoring of infectious risk, linking electronic records to HeRO monitoring could expand the knowledge dataset.

Table 1. Patient Demographic and Disease Data.


Table 2. HeRO Analysis. Likelihood of HeRO Score >1 Before NEC Diagnosis.


Figure. HeRO scores in Medical vs. Surgical NEC patients before NEC Diagnosis.
Figure. HeRO scores in Medical vs. Surgical NEC patients before NEC Diagnosis.
This figure shows daily Surgical (Left, n = 21) patient HeRO scores increasing up to 6 days before the infant was diagnosed with NEC with the blue line indicating mean HeRO scores. HeRO scores were relatively level in Medical patients (Right, n = 30) during the days before onset of NEC.
A HeRO score of 1 is considered neutral reflecting no greater risk of infection than the general population. Rising HeRO scores reflect multiples of infectious risk, such that a HeRO score of 2 indicates an infectious risk 2x that of the general population. The changing dot colors indicate the relative risk related to the HeRO score with green low risk changing to orange/ red as infection risk rises.