Introduction
Patients with CHD are at increased risk of intestinal injury. Factors that contribute to mucosal barrier fragility include haemodynamic changes, abnormal systemic vasculature, ductal-dependent systemic circulation, hypoxia, anaemia, reperfusion injury, inflammation, and more. Reference Becker, Hornik and Cotten1–Reference Bell, Suna and Marathe4 The inability to meet the oxygen demands of the intestine in response to feeding leads to mesenteric ischaemia, which causes intestinal barrier dysfunction, changes to the gut microbiome, dysmotility, post-operative feeding intolerance, and post-operative necrotising enterocolitis. Reference Jeffries, Wells, Starnes, Wetzel and Moromisato5 The reported incidence of necrotising enterocolitis in the CHD population ranges from 2 to 13%, with the highest rates found in patients with left-sided obstructive lesions. Reference Kelleher, McMahon and James3,Reference Bell, Suna and Marathe4,Reference McElhinney, Hedrick and Bush6
The 1978 Bell’s criteria Reference Bell, Ternberg and Feigin7 and the 1986 modified Bell’s criteria, Reference Walsh and Kliegman8 developed from research in the premature neonatal population, remain the gold standard for diagnosis of necrotising enterocolitis. Bell’s criteria use a standardised method to grade necrotising enterocolitis based on clinical and radiographic features. Stage 1, or “suspected NEC,” is defined by mild systemic signs and early radiographic features that are non-specific. There are some similarities in the pathophysiology of necrotising enterocolitis in premature infants and those with CHD. In both groups, ischaemic insults causing intestinal injury can result from hypotension, anaemia, blood transfusion, and the innate inflammatory response of the gut to pathogens. In infants with CHD, particularly those with a large patent ductus arteriosus or systemic to pulmonary shunt, episodic ischaemia from diastolic steal or flow reversal leads to impaired diastolic mesenteric blood flow, resulting in the colon being disproportionately affected in CHD patients due to its watershed regions of blood supply. Reference Bubberman, van Zoonen and Bruggink2 Furthermore, the use of medications in the CHD population, such as prostaglandins, proton pump inhibitors, and opioids that reduce gastric acid production and prolong enteral transit time, can facilitate bacterial translocation and increase the risk of necrotising enterocolitis. Lastly, feeding practices vary greatly, and delayed feeding may also contribute to cellular atrophy and increased gastrointestinal permeability. Reference Bell, Suna and Marathe4,Reference Floh, Slicker and Schwartz9–Reference Scahill, Graham, Atz, Bradley, Kavarana and Zyblewski11
Diagnosis of necrotising enterocolitis in CHD patients is difficult as emesis, abdominal distention, and hematochezia (findings that align with Stage I necrotising enterocolitis by Bell’s criteria) may also occur in situations of feeding intolerance without necrotising enterocolitis. Reference Bell, Suna and Marathe4 Necrotising enterocolitis, in the setting of CHD, significantly increases mortality, with many studies reporting rates exceeding 25%. Reference Jeffries, Wells, Starnes, Wetzel and Moromisato5,Reference Martini, Beghetti and Annunziata12 As a result, the management of suspected necrotising enterocolitis in this population has been conservative, typically involving bowel rest and antibiotics for 7–14 days. However, no literature currently guides decision-making regarding the diagnosis and treatment of necrotising enterocolitis in the CHD population.
Given this uncertainty, we performed a quality improvement project to create a novel diagnostic and treatment algorithm for necrotising enterocolitis in infants with CHD at a single centre. Our goals were to improve early diagnosis of necrotising enterocolitis, categorise disease severity, minimise antibiotic duration, and return patients to enteral feeding as soon as safely possible.
Methods
Setting and context
The project is a single-centre study conducted at a 17-bed cardiac ICU at Norton Children’s Hospital in Louisville, Kentucky.
Preintervention
In 2021, we established a multidisciplinary quality improvement team consisting of cardiac intensivists, a neonatologist, cardiologists, paediatric general surgeons, a cardiothoracic surgeon, nurse practitioners, speech therapists, registered dieticians, and a pharmacist. A subgroup from this team was formed to focus on the diagnosis and treatment of necrotising enterocolitis. It was determined that significant practice variability existed in nil per os (NPO) times, antibiotic duration, and clinical monitoring. The quality improvement team identified factors that influenced the decision-making for suspected necrotising enterocolitis diagnosis and the re-initiation of enteral nutrition. This, in combination with known literature on necrotising enterocolitis diagnosis, was used to create a scoring tool. Reference Walsh and Kliegman8,Reference Casals and Spaeder13–Reference Schat, Schurink and van der Laan19 The elements of the scoring tool were selected based on published literature and included diagnostic imaging, known risk factors for necrotising enterocolitis development, clinical presentation of necrotising enterocolitis, and markers that indicate patient illness severity to stratify risk. Points were assigned to each element based on severity, with higher values assigned to factors that reflect an increased risk for true necrotising enterocolitis, thus contributing to a larger total score. For instance, normal values or findings were assigned a score of 0, while the most severe findings, such as free air in the abdomen, were assigned 4 points (Supplemental Table S1). In the absence of extensive data for retrospective design using prediction models, this tool was developed based on clinical knowledge and current literature.
Total scores were grouped into five categories: <4, 4-8, 9-12, 13-17, and >18, with higher scores indicating a greater likelihood of necrotising enterocolitis. Clinical management steps were recommended for each severity group. The treatment algorithm indicated time NPO, recommendations on how to resume feeds, duration of antibiotic therapy, and imaging schedule (Supplemental Table S1).
A key driver diagram delineated primary drivers for diagnosis and treatment of necrotising enterocolitis as (1) standardising NPO and antibiotic days, (2) reducing the potential overdiagnosis of necrotising enterocolitis, and (3) reducing time back to full feeds after necrotising enterocolitis treatment (Fig 1).
Figure 1. Key driver diagram.
Cohort identification
We identified preintervention patients through a retrospective chart review of all patients with a diagnosis of CHD and necrotising enterocolitis (Interational Classification of Diseases (ICD-10) -10 codes as follows: P77.1, P77.2, P77.3, P77.9, K92.1, K63.89. Current Procedural Terminology (CPT) codes as follows: 44,120, 44,125, 44,141) as well as myocarditis or cardiomyopathies cared for in the cardiac ICU from January 1 2019 to January 1 2022. Postintervention patients were identified prospectively from October 1, 2022 to November 1, 2023.
Inclusion criteria
Both pre- and post-operative patients under six months old with CHD, as well as patients with ductal stents, myocarditis, or cardiomyopathy admitted to the cardiac ICU, were included. Patients admitted to the ICU after a catheter-based intervention (e.g., ductal stent placement) were also included. Once patients were transferred out of the ICU, they were no longer included. Patients with more than one episode of suspected necrotising enterocolitis concern were defined as a separate subsequent event.
Exclusion criteria
Patients who could not achieve full feeds after concerning for necrotising enterocolitis secondary to persistent feeding intolerance lasting more than 3 months were excluded.
Planning and implementing the study interventions
Interventions
We implemented the necrotising enterocolitis scoring tool and treatment algorithm in all patients in the cardiac ICU less than six months of age with clinical suspicion of necrotising enterocolitis. The clinical trigger for scoring was defined as any concern for necrotising enterocolitis that would normally prompt a provider to hold feeds and initiate a diagnostic evaluation. Education on the inclusion criteria, elements of the scoring tool, and the application to clinical practice was provided through virtual and in-person sessions with various provider groups (physicians, advanced practice providers, surgeons, and dietitians). These sessions included interactive elements such as group discussions, practice scoring, and question-and-answer opportunities to enhance understanding. The tool was printed and kept on the unit as well as accessible via a quick response code. Reinforcement of appropriate tool use was overseen by the site primary investigator (JFD). Each time the tool was implemented, the primary investigator was notified and independently scored the patient for accuracy. Several multidisciplinary meetings were held to promote discussion and gather feedback for the duration of the study period.
Evolution of the interventions
To monitor compliance, the primary investigator performed weekly audits. The tool was modified based on feedback and discussion. Modifications made to the tool based on the feedback of the providers included establishing and defining the clinical trigger and defining the time interval between scores.
Validity data were gathered from the scoring tool after six months of use. Six months following implementation, the scoring tool was validated. Six providers (three advanced practice providers and three intensivists) were asked to utilise the tool to score four artificial patient scenarios created by the primary investigator. Participants were also asked for feedback on the use of the scoring tool. The tool demonstrated 88% reliability in producing consistent scores between individuals. Feedback from this validation phase was incorporated to improve replicability. Changes included formatting adjustments to enhance readability and user-friendliness and modification of the “prodrome” category by separating “new vasoactive use” into its risk category with different assigned points for stable vasopressor (unchanged dose of vasopressor prior to and during concern for necrotising enterocolitis) use versus escalation of vasopressors during period of concern for necrotising enterocolitis. These improvements led to 100% accuracy in score replication during a repeat validation cycle. A timeline and description of the Plan-Do-Study-Act cycles and the evolution of the protocol are presented in Supplemental Figure S1.
Measures
Collected data included general demographics, primary cardiac lesion, operative data (type of surgery and cardiopulmonary bypass time), vasoactive-inotropic score Reference Gaies, Jeffries and Niebler20 , anticoagulant use; ventilator requirement (yes/no as well as mean airway pressure); Bell’s Stage; nutritional data at the time of necrotising enterocolitis suspicion, during necrotising enterocolitis treatment, and after necrotising enterocolitis treatment, physical and laboratory components from the scoring tool as well as surgical complexity (Society of Thoracic Surgeons-European Association for Cardio-Thoracic Surgery (STAT) category). Reference Jacobs, Jacobs and Thibault21
Outcome measures
Outcome measures include days NPO, antibiotic treatment duration, days on total parenteral nutrition, and days to full enteral feeding after necrotising enterocolitis diagnosis.
Process measures
Compliance was monitored weekly, and the necrotising enterocolitis score and treatment plan were incorporated into the daily progress notes. Compliance was measured through the use of the scoring tool and associated treatment algorithm without deviation. Deviation was allowed and recorded at the provider’s discretion.
Balancing measures
Balancing measures include the time to reach full feeds after cardiac surgery (which is defined as at least 150 ml/kg/day of 24 kcal/ounce feeds for this institution), progression to a diagnosis of surgical necrotising enterocolitis, an increase in vasoactive-inotropic score by more than 5, length of stay, mortality, and the method of feeding at discharge.
Statistical analysis
Demographic and clinical characteristics are expressed as mean ± SD or median/interquartile range as appropriate. For the comparison of continuous variables between two groups, we used a Student’s t-test for normally distributed data and a Wilcoxon rank sum test for skewed data. Categorical variables were compared using contingency tables or Fisher’s exact test. A p value of<0.05 was considered significant. All data were analysed using R version 4.3.1. 22 Quality improvement Macros were used to produce statistical process control charts. 23
Ethical considerations
The study was approved, and a waiver of informed consent was granted by the University of Louisville and Norton Children’s Hospital Institutional Review Board. We followed the Standards for Quality Improvement Reporting Excellence Guidelines (SQUIRE 2.0) in this report. Reference Kelz, Schwartz and Haut24
Results
A total of 29 patients were initially included in the study (14 preintervention and 15 postintervention). There were 47 episodes of necrotising enterocolitis (19 preintervention and 28 postintervention). After excluding two patients for persistent feeding intolerance>3 months, a total of 27 patients were included (14 preintervention and 13 postintervention). Thirty-nine episodes of necrotising enterocolitis were analysed (19 preintervention and 20 postintervention) (Supplemental Fig S2). The surgical volume per STAT category at this single centre in the pre and postintervention time intervals was not different. All episodes of necrotising enterocolitis in the cohort occurred post-procedurally. One of the two patients excluded had seven episodes of bloody stools and concern for necrotising enterocolitis during the study period and never advanced past half-volume feeds. The other patient was listed for transplant and did not advance past 1/3 volume feeds until after the transplant.
Preintervention versus postintervention groups
Table 1 summarises the demographic and clinical characteristics. Baseline characteristics, including age at admission, weight, cardiac lesion, surgery, STAT category, bypass duration, and use of ECMO, were similar between the two groups. Patients in the postintervention group were born at an earlier gestational age (37 weeks vs. 38 weeks and 5 days, p = 0.027). The incidence of necrotising enterocolitis was highest in patients with hypoplastic left heart syndrome (50% preintervention and 31% post, p = 0.33). Nearly half of the patients who met the criteria for necrotising enterocolitis underwent an operation classified as STAT 5 (42% preintervention and 55% postintervention, p = 0.33).
Table 1. Clinical characteristics and descriptive features of patients at time of necrotising enterocolitis score
a. 150ml/kg of 24kcal.
b. Heparin dosing 15U/kg/hr.
c. Lovenox dosing 0.75 mg/kg Q12.
d. PTT goal 60-90s.
e. Xa goal 0.5-1 units/ml.
f. DTT goal 60-90s.
The risk factors and descriptive features of patients at the time of necrotising enterocolitis score, as well as clinical and laboratory changes, are also listed in Table 1. Markers of severity of illness, such as ventilation and requirement of vasopressors, were similar pre and postintervention. Nutritional data, including feed volume, advancement rate, and formula choice, did not differ between pre and postintervention groups. There was no difference in the amount of patients on fortified feeds in the pre and postintervention groups. However, more patients in the preintervention group had feeds fortified within 24 hours before concerning for necrotising enterocolitis. Elevated lactic acidosis was seen more often in the preintervention group, with 44% of episodes of necrotising enterocolitis with a lactic acid>2.5mmol/L versus only 5% in the postintervention group (p = 0.021). A greater decrease in platelet count was also seen in the preintervention group, with 37% of episodes of necrotising enterocolitis associated with a decrease in platelets versus 0% in the postintervention group (p = 0.011). Other clinical and laboratory changes, including pH, anticoagulation method, change in sodium or creatinine, change in urine output, abdominal exam, grossly bloody stools, or pneumatosis seen on abdominal radiographs, did not differ between the two groups.
Measures
Primary outcome measures
In the postintervention group, patients were NPO for fewer days (2 vs. 7 days, p = 0.004), on antibiotics for fewer days (3 versus 7 days, p = 0.02), on total parenteral nutrition for fewer days (0 vs. 20 days, p = 0.01) and returned to full feeds after diagnosis of necrotising enterocolitis in fewer days (8 versus 18 days, p = 0.013) (Table 2). When analysed by Bell’s criteria, Bell’s Stage I patients had significantly fewer NPO days (2 versus 7, p-0.004), antibiotic days (2 vs. 7, p = 0.045), and a shorter time to full feeds (5 vs. 14.5, p = 0.023) postintervention (Supplemental Table S2). The Bell’s Stage II and III patients had no statistical difference in primary outcomes; however, the median NPO days and antibiotic days were fewer in the postintervention group.
Table 2. Outcomes
Continuous variables: median and interquartile range.
Categorical variables: number and percentage.
Process measures
Compliance with the use of the necrotising enterocolitis scoring tool and treatment algorithm was 100%. The median initial score at the time of necrotising enterocolitis concern was 5 and ranged from 3 to 10. Of the 20 episodes of necrotising enterocolitis in the postintervention group, 18 of these scored<4 at the recommended repeat scoring time interval, and feeds were restarted according to the treatment algorithm. On two occasions, the repeat score prompted continued NPO status.
Balancing measures
Table 2 also shows there was no change in balancing measures between groups. Time to full feeds after cardiac surgery, progression to surgical necrotising enterocolitis diagnosis, length of stay, change in vasoactive-inotropic score, mortality, and method of feeding at discharge were not different between the two groups.
Discussion
Given the unique pathophysiological and clinical differences in the aetiology and diagnosis of necrotising enterocolitis in infants with CHD, as well as the need for consistent definitions and comparisons across the literature, a standardised approach is ideal. Our study suggests that the implementation of a novel scoring tool and treatment algorithm when there is a clinical concern for necrotising enterocolitis in CHD patients less than six months of age reduces the duration of NPO time and antibiotics without an increase in either disease severity or progression to surgical necrotising enterocolitis. This is particularly true for patients with Bell’s Stage I suspected necrotising enterocolitis when management is standardised. Reference Bubberman, van Zoonen and Bruggink2
It is well-reported that withholding enteral feeds and malnutrition in the CHD population is associated with increased mortality and morbidity. Reference Kelleher, McMahon and James3,Reference Wong, Cheifetz, Ong, Nakao and Lee25,Reference McKean, Kasparian, Batra, Sholler, Winlaw and Dalby-Payne26 Feeding restrictions may compromise metabolic reserves important for neonates undergoing cardiopulmonary bypass. While the use of total parenteral nutrition provides nutrition, it does not maintain and restore the intestinal microbiome. Additionally, enteral feeding is essential for the health and growth of the intestine and its mucosal barrier. In infants with CHD, who are at risk of developing severe post-operative bacterial infections, avoiding prolonged total parenteral nutrition use and initiating early enteral feeding are protective. Reference Yoshimura, Miyazu and Yoshizawa27
The majority of patients in this study were Bell’s Stage I, without true pneumatosis on abdominal radiographs. This scoring tool is unique in that it allows risk stratification of patients based on objective data and provides standardised treatment. Historically, at our institution, concern for necrotising enterocolitis, despite no evidence of pneumatosis, would often result in 5–7 days of antibiotics and NPO time. Following the implementation of the scoring tool and treatment algorithm, antibiotic duration and NPO were decreased. This reduction was not statistically significant for Bell’s Stage II and III patients, likely due to our small sample size.
The incidence of necrotising enterocolitis in premature infants with CHD is higher than in those with prematurity alone. Compared to the preintervention group, the postintervention group in our study had a younger gestational age. Nevertheless, no increased development of surgical necrotising enterocolitis was observed. This finding could be secondary to small sample size capture and variability in CHD lesions. However, there was no difference seen in cardiac diagnosis and STAT category between the pre and postintervention groups. Multiple patients in the preintervention group had symptoms associated with surgical necrotising enterocolitis, including lower platelets and elevated lactic acid. Although not statistically significant with the small sample size, the preintervention group also contained more patients that progressed to surgical necrotising enterocolitis. Within the preintervention group, all mortality was attributable to six episodes of necrotising enterocolitis that occurred in three patients (one patient had two episodes of necrotising enterocolitis and one patient had three episodes of necrotising enterocolitis). Death was due to intestinal ischaemia in only one of the three patients. No patient in the postintervention group died secondary to necrotising enterocolitis. One patient died in the postintervention group (with two episodes of necrotising enterocolitis) secondary to respiratory failure not associated with necrotising enterocolitis. While no statistically significant difference in necrotising enterocolitis development rate, severity of illness markers, and death was noted, the postintervention group had a trend towards fewer necrotising enterocolitis events and fewer necrotising enterocolitis-attributed deaths, suggesting that such differences might be detectable given a higher sample size.
There was no difference in the number of patients receiving fortified nutrition between the pre and postintervention groups. However, in the preintervention group, a significant increase was noted in the number of patients who had feeds fortified 24 hours before the diagnosis of necrotising enterocolitis. This could have potentially influenced the time taken to achieve full feeds, as fortification is a known risk factor for necrotising enterocolitis development.6 12 In addition, more patients were NPO at the time of necrotising enterocolitis diagnosis in the preintervention group, suggesting that the preintervention group was generally sicker. It should be noted, however, that the treatment algorithm standardised the resumption and advancement of feeds and it is possible this contributed to the improved outcomes seen in the postintervention group. Although not statistically significant, more patients received breast milk in the postintervention group at the time of diagnosis. Breast milk has been shown to be protective against the development of necrotising enterocolitis. Reference Kelleher, McMahon and James3 The treatment algorithm does not indicate changing feeds when a patient is refed; however, this represents a potential area for further investigation.
The clinical trigger for concern about necrotising enterocolitis is often the presence of grossly bloody stools. However, in CHD patients on anticoagulation, this clinical sign can be multifactorial and may not necessarily indicate intestinal ischaemia. A patient with a ductal-dependent lesion and a grossly bloody stool would automatically receive a score of 5, placing them in the treatment algorithm of 48 hours of NPO and antibiotics. While this could be seen as overtreatment, this sub-population is at the highest risk for clinically significant necrotising enterocolitis. Reference Becker, Hornik and Cotten1 The tool provides an objective measurement that can be reassuring in such cases. It recommends reassessing the patient after 48 hours, and if there is no development of true pneumatosis or worsening clinical condition, the algorithm suggests it is safe to resume feeding, even if bloody stools persist. For patients whose symptoms may resemble necrotising enterocolitis but could be attributed to other diagnoses (i.e., anticoagulation, milk protein intolerance, feeding intolerance, ileus) and are at low risk for surgical necrotising enterocolitis, the tool prevented overtreatment while also reducing NPO time and antibiotic days.
Although the necrotising enterocolitis scoring tool incorporates near-infrared spectroscopy, we could not compare pre and postintervention patients for significant changes in near-infrared spectroscopy due to a lack of available data in the preintervention group. Reference Casals and Spaeder13 Somatic or renal near-infrared spectroscopy has been shown to detect gut perfusion changes before the development of necrotising enterocolitis and may provide an opportunity for timely interventions that improve necrotising enterocolitis outcomes. Reference Howarth, Banerjee, Leung and Aladangady15,Reference Mishra, Mathur, Mohamed and Maheshwari18,Reference Schat, Schurink and van der Laan19 As the necrotising enterocolitis literature continues to evolve, we recommend incorporating additional variables into the scoring tool for ongoing validation and refinement.
Given the small sample size, we were not able to identify which components of the scoring tool and treatment algorithm had the greatest influence on the improved outcomes. We hope to expand the use of the scoring tool and treatment algorithm to other institutions. Larger numbers of patients will allow us to identify whether the improved outcomes are primarily due to earlier identification, risk stratification, or the treatment algorithm itself. Future expansion of this project could also include the implementation of the scoring tool and treatment algorithm in high-risk patients for necrotising enterocolitis admitted from home to the general ward, to cardiac step-down units, and to neonatal intensive care units that care for infants with CHD.
Limitations
Our study has several limitations. First, the analysis is limited by the small sample size. Not all patients in our cohort had the same underlying anatomy or surgical repair, but due to the rarity of this complication, further sub-division was not feasible. Although the annual surgical volume did not differ between the two time periods, the total number of patients was not equal, with 586 in the pre-intervention period and 201 in the post-intervention period. The similar patient numbers in the preintervention and postintervention groups, despite the different time frames analysed (3 years versus 1 year), likely indicate that some patients with suspected necrotising enterocolitis were not identified in the preintervention period. This may be secondary to ICD-based patient identification in the preintervention period versus real-time data collection in the postintervention period.
Second, the generalizability of this single-centre study may be limited. Centres with different infrastructure, care teams, and practice variations may not achieve the same outcomes. For example, our institution does not routinely use abdominal ultrasound to aid in the diagnosis of necrotising enterocolitis, though growing literature suggests that ultrasound can provide real-time assessment of vascular perfusion, bowel wall thickness, improved sensitivity for detecting pneumatosis, altered peristalsis, and peritoneal fluid characteristics.17 While ultrasound may offer increased sensitivity, this advantage depends on the expertise of the sonographer. As a result, it was not a key element in our scoring tool, nor was it used to identify pneumatosis at this institution.
Another variation seen between centres is the location of patient discharge. At our institution, all pulmonary artery bands, shunts, and ductal stents are admitted and discharged from the ICU and may affect variable outcomes such as ICU length of stay. We aim to continue validating the necrotising enterocolitis scoring tool across institutions as it is implemented and as the algorithm is refined.
Conclusions
The implementation of a scoring tool and treatment algorithm for CHD patients ≤6 months of age with suspected necrotising enterocolitis was associated with fewer days on antibiotics, fewer NPO days, improved time to return to full feeds after concern for necrotising enterocolitis, and no increase in progression to surgical necrotising enterocolitis. This collaborative effort from a multidisciplinary team improved the standardisation of care and patient outcomes. These initial promising results from this intervention need to be validated and refined across multiple centres.
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/S1047951125000368.
Acknowledgements
None.
Financial support
This research received no specific grant from any funding agency, commercial, or not-for-profit sectors.
Competing interests
None. The authors have no financial interests to declare in relation to the content of this article.
Open access
