Standardizing the Evaluation and Management of Necrotizing Enterocolitis in a Level IV NICU

The MGH NICU is a 38-bed, Level IV referral center caring for nearly 800 critically ill infants annually. Baseline NEC rates per year among VLBW infants, as defined per the Vermont Oxford Network between 2010 and 2020, ranged from 1.4% to 12.2%, with a median of 7.0%.¹⁹  NEC requiring surgical intervention in VLBW infants over the same period was 0% to 6.4%, with a median of 2.8%.¹⁹  The center provides consulting services from pediatric surgery, pediatric infectious disease, pediatric radiology, nutrition, pediatric pharmacy, lactation, and occupational and speech therapy. The unit is staffed by neonatologists (in-house 24/7), neonatal nurse practitioners, NICU hospitalists, NICU fellows, and pediatric residents utilizing Epic Systems (Verona, WI) as an electronic medical record (EMR) and ordering system. The institution sponsors formal training in process improvement through the Clinical Process Improvement Leadership Program²⁰  and multiple team members participated during this project. This work was undertaken as a QI initiative at MGH, and therefore was not formally supervised by the institutional review board per their policies (2021P001903).

A multidisciplinary workgroup with representatives from neonatology (faculty, fellows, residents, nurse practitioners, and physician leadership) and nursing (registered nurses and nursing leadership), as well as pediatric surgeons, pediatric radiologists, pediatric infectious disease specialists, NICU dieticians, and pediatric pharmacists developed a driver diagram aimed at determining primary and secondary drivers of variability in NEC management (Fig 1). As observed in Figure 1, many variables were identified within our unit, particularly differences in physician training practices, inconsistency in defining disease severity, and medical team turnover. We learned that team members preferred uniformity and a better way of anticipating and responding appropriately to patient care plans. Likewise, the multidisciplinary construction of the driver diagram highlighted areas of poor communication. A crucial point was the need to use a standardized classification of NEC severity, such as the Bell’s staging system.^21,22 

The driver diagram (Fig 1) also describes how our potential change ideas would address root causes of variation. The development of a NEC management guideline by our multidisciplinary group would foster collaboration among services (neonatology, pediatric surgery, pediatric infectious disease, pediatric radiology, nursing, and nutrition). Additionally, a guideline removes decision-making on the basis of experience and training, which can vary by practitioner and moment to moment.⁹  The workgroup hypothesized that a guideline derived from available evidence and local expert consensus would shift the decision-making process away from personal expertise and toward a standardized plan. Lastly, the guideline would both inform decision-making and educate.

Using this proposed, unit-developed guideline, we aimed to align our current management of NEC, thereby decreasing variation of practice within our unit regarding antibiotic choice, antibiotic days, and nil per os (NPO) days without increasing morbidity or mortality. Specifically, the guideline narrowed antibiotics to 1 agent (piperacillin/tazobactam) and recommended a course of treatment (antibiotics and NPO time) on the basis of the severity of NEC disease (Bell’s staging criteria²² ). All aims were planned on a 2-year timeline.

A NEC evaluation and management guideline was developed by our working group starting in February 2018 (Fig 2). We incorporated our best multidisciplinary interpretation of current literature using modified Bell’s staging^{1,2,6,7,23–26}  to advise initial laboratories,^6,27,28  imaging,^27,29–40  NPO course,⁴¹  antibiotic choice and course,⁸  gastric decompression,⁴¹  breast milk for oral care,^42–46  feeding readvancement,^47–50  and pediatric surgery and infectious disease consultation.^8,26  Piperacillin/tazobactam was chosen as the preferred single-agent antibiotic. Piperacillin/tazobactam has the fewest compatibility concerns with other commonly used intravenous medications and parenteral nutrition, compared with alternative antibiotic regimens.⁵¹  In addition, piperacillin/tazobactam is generally less nephrotoxic than most aminoglycosides,⁵²  with less concern for increasing ampicillin and gentamicin resistance to common bacterial pathogens,⁵³  and has good coverage for pathogenic intestinal bacteria, including anaerobic species.^53,54 

The working group met monthly with pediatric surgery, pediatric radiology, pediatric pharmacy, NICU nursing, NICU nutrition, and pediatric infectious disease between February 2018 and September 2018. After final approval by all stakeholders, the guideline was presented and approved by our Newborn Collaborative Practice Committee, a multidisciplinary group that reviews and approves guidelines for inpatient newborn and infant care. Guidelines approved via this route are reviewed and updated at least every 3 years. The newly developed guidelines were posted to our clinical practice guideline online warehouse and printed and posted in the clinician workstation in December 2018. Furthermore, monthly, one-hour lectures on NEC and the application and use of the guidelines were provided to new residents, fellows, and other interested NICU staff starting in March 2018. Neonatal attending physicians were educated on new guidelines through regularly scheduled divisional meetings. Updates were presented to staff during these meetings annually. An educational bundle was presented to the nursing staff that included NEC's clinical presentation and pathophysiology, the nursing assessment and care of the infant with NEC, and the guidelines for the workup and management of NEC. The education sessions were conducted over a two-week timeframe, twice daily at “board rounds,” which occur at the start of each shift, with all nurses of the shift present. The nursing staff was required to review an educational module after implementation of the new guidelines in December 2018.

After PDSA Cycle 1, feedback from staff suggested that other tools within our workflow could help support the guideline's efforts to educate and standardize care. This included using our EMR system to inform the ordering of tests and treatments in alignment with the established guideline.⁵⁵  The first tool developed was a “NEC order panel” that prompts providers to order the initial laboratory testing, NPO status, Replogle tube placement, imaging, and piperacillin/tazobactam as the antibiotic when evaluating a patient with clinical concern for NEC. The second tool was a documentation template familiar to providers who write notes in our unit. The template prompted documentation of Bell’s staging and reminded the user of the staging definitions. The staging choice within the template then guided the provider to the recommended treatment course (including antibiotic choice, antibiotic duration, and duration of NPO), nutrition recommendations, indication for consultants, and laboratory/imaging frequency. This PDSA cycle was initiated in January 2019.

Primary outcome measures included: (1) ratio of observed-to-expected antibiotic days per NEC case based on the guideline (Fig 2), (2) ratio of observed-to-expected NPO days per NEC case based on the guideline (Fig 2), and (3) percent piperacillin/tazobactam utilization in NEC cases. Using a ratio of observed-to-expected treatment days based on the guideline, rather than the absolute number of days, accounts for differences in NEC severity/Bell’s staging. An “ideal” observed-to-expected ratio is 1.0, regardless of stage. In a unit with no interpractice variability, where the guideline was being closely followed and patients receive the recommended antibiotic and NPO courses, the resulting ratio would be 1.0. Ratios >1.0 indicate a course longer than recommended, and <1.0 indicate a shorter course.

Process measures included educational sessions for nurses and trainees, as well as EMR order panel usage. Balancing measures were percentage of cases requiring surgery, resulting in death, and treated for fungemia.

Baseline data were collected retrospectively for 2 years (January 2016 to January 2018) before the first PDSA cycle. Earlier data collection was constrained by absence of a reliable EMR. Qualifying patients were identified utilizing the Epic system report functionality to obtain antibiotic orders that were administered to patients in the NICU. Antibiotic search terms included clindamycin, meropenem, metronidazole, and piperacillin/tazobactam. From this preliminary list, inclusion was determined through manual data collection by a second team member via chart review and cross-referenced with data collected by our Vermont Oxford Network data collector. Because PDSA cycles began in February 2018, data collected before this date were marked as “preintervention” and subsequent as “postintervention.” NEC was defined as previously described per the VON definition.¹⁹  Using this strategy, patients with isolated spontaneous intestinal perforation (SIP) noted during surgical intervention were not included. Patients with severe congenital gastrointestinal anomalies were excluded from all analyses. Deceased patients were excluded from the primary analysis because a truncated treatment course would not reflect management variability or guideline compliance. Patients of all NEC stages were included.

Data were collected and stored using Research Electronic Data Capture (Nashville, TN). The data were clustered into 6-month intervals (January–June and July–December). QI Macros SPC Software (Denver, CO) within Microsoft Excel (Redmond, WA) was used for the primary analysis. The ratios of observed-to-expected days of antibiotics and NPO were plotted as u-charts (Figs 3 and 4), given they represent discrete “count” data with variable areas of opportunity. Percentage piperacillin/tazobactam utilization and the balancing measures (percentage of cases requiring surgery, percentage of cases resulting in death, and percentage of cases treated for fungemia) were plotted as p-charts given they represent discrete “classification” data (Figs 5 and 6 and Supplemental Fig 7). Process change and stability were analyzed using QI Macros SPC Software, applying Montgomery standard rules of special cause variation.⁵¹ 

We performed a secondary analysis stratifying cases by Bell’s Stage of NEC, comparing outcome measures pre- versus postintervention. We did this analysis in two ways. First, we excluded deceased patients (n = 99), as in the primary analysis above, to prevent inclusion of falsely short durations of antibiotic and NPO days caused by deaths before treatment completion. To be conservative, we performed the analysis again, including deceased patients (n = 114). We used Mann-Whitney tests for continuous outcomes of average antibiotic and NPO days, and Fisher’s exact tests for the binary outcome of piperacillin/tazobactam usage. These analyses, shown in Table 1 and Supplemental Table 3, were performed in Graph Pad Prism software version 9.1.2 (San Diego, CA).

A total of 124 cases of NEC were observed between January 2016 and June 2021. After excluding cases in patients with severe congenital gastrointestinal anomalies, 114 cases of NEC remained. Fifteen patients (13.2%) died before discharge, leaving 99 surviving cases (86.8%) of NEC for the primary analysis. A chart review of Stage IIIB cases confirmed that there were no cases of SIP within our Stage IIIB NEC patients. There were 2 cases of SIP in our unit during the study period. There were no significant differences in the staging of cases per epoch pre- versus postintervention by χ² (P value .35) (Table 2).

A control chart (u-chart) plotting “ratio of observed-to-expected antibiotic days per NEC case” over time reveals special cause variation with the maintenance of this effect over several 6-month intervals (Fig 3). Process change analysis demonstrates a baseline mean ratio of 1.94, with a postintervention mean of 1.18.

A control chart (u-chart) plotting “ratio of observed-to-expected NPO days per NEC case” over time reveals special cause variation with the maintenance of this effect over several 6-month intervals (Fig 4). Process change analysis indicates a baseline mean ratio of 1.69 with a postintervention mean of 1.14.

A control chart (p-chart) plotting “percentage of piperacillin/tazobactam utilization per NEC case” over time reveals special cause variation with the maintenance of this effect over several 6-month intervals (Fig 5). Process change analysis indicates a baseline mean of 30% with a postintervention mean of 91%.

In our secondary analysis of cases stratified by Bell’s stage, excluding deceased patients, the average antibiotic days were significantly reduced for Stage IIA (P value .031) postintervention (Table 1). The average NPO days were significantly reduced for Stage IA (P value .031) and stage IIA (P value .026) postintervention (Table 1). Piperacillin/tazobactam utilization per NEC case was significantly increased postintervention for Stages IA and IIA (P value .008 and .018, respectively) (Table 1). When deceased patients were included, the overall trends remained unchanged, and piperacillin/tazobactam utilization per NEC case was significantly increased postintervention for Stage IIB (P = .036) (Supplemental Table 3).

PDSA cycle 1: Planned monthly educational lectures with physician trainees were recorded as being completed 84.8% (39 of 46) of the time. This averages to 5.1 lectures per every 6-month interval. The nursing education bundle and module were considered complete after at least 80% of the 75 staff nurses were “signed off” as receiving formal education.

PDSA cycle 2: The NEC order panel within our EMR was used in 32% (16 of 50) of NEC cases after its creation in January 2019. The order panel was used in 34% (16 of 47) of NEC cases that used piperacillin/tazobactam. Note template utilization could not be tracked in an automated fashion and, therefore, was determined not to be a feasible process measure with the project’s time frame.

Our sample size is small and limits our assessment of important outcome measures. However, in their respective control charts (p-charts), no special cause variation was noted in any of the following balancing measures: percentage of cases resulting in death, percentage of cases requiring surgery, or percentage of cases treated for fungemia (Fig 6 and Supplemental Fig 7).

NEC is a devastating disease with high morbidity and mortality in the preterm and VLBW neonatal populations. Despite the severity of this condition, there is little evidence supporting a single therapeutic strategy. However, the literature does strongly support the importance of standardization in medical practice.^14,16  Once variability in practice is narrowed, the focus can turn toward improving outcomes. Although many studies, including QI projects, have importantly focused on decreasing NEC rates,^15,56  few have aimed to standardize NEC treatment. In this QI project, we aimed to reduce the variation in the management of NEC in a Level IV NICU.

We demonstrated that developing a guideline with supportive constructs in the EMR and accompanying education effectively decreased variation in the management of NEC in our unit over 3.5 years. We found special cause variation with decreased ratios of observed-to-expected antibiotic days (1.94 to 1.18) and NPO days (1.69 to 1.14). Limiting antibiotics and NPO days are important for neonatal health. Antibiotics in the neonatal period have been shown to cause changes in the microbiome, leading to long-term health implications.^57,58  Similarly, periods of NPO days can lead to villous atrophy, increase in feeding intolerance, poor growth, and prolonged duration of central lines.^59,60  Both outcome ratios show significant gains in a relatively short period of time. With continued practice, our hope is that these ratios may continue to trend even closer to the ideal of 1. However, the group simultaneously acknowledges that NEC diagnosis is often nuanced and difficult to fit into strict Bell’s staging criteria definitions and may lead clinicians toward a more conservative approach.

Before our guideline, 30% of NEC patients were treated with piperacillin/tazobactam. After this initiative, piperacillin/tazobactam was the antibiotic of choice in 91% of patients with NEC. Although there is no clear evidence for selecting one antibiotic regimen over another in the management of NEC,^61,62  we showed that adherence to the guideline was robust after its initiation. Our findings could have implications for other neonatal diseases in which management is unnecessarily varied, and a similar approach could be adopted to decrease variation. Finally, there was no increase in the balancing measures of surgical NEC, fungal infections, or deaths with implementing the guideline. In the future, the promotion of responsible antibiotic stewardship may be aided by the use of narrower antibiotic regimens in less-severe forms of NEC.⁶³ 

Our project interventions are examples of “just-in-time” learning, a technique when education is embedded directly into a provider's workflow. This method can increase the utilization and effectiveness of a new guideline.⁵⁵  The built-in documentation template and order panel guide the user through recommendations rather than depending on a provider's memory or manual referencing of the guideline. Thereby, the preferred management plan becomes automatic.

One limitation to our study is the small number of NEC cases annually within our center. In addition, because many of our patients did not require surgery, it is possible that some patients with a presumed diagnosis of NEC managed medically had a SIP and were misdiagnosed because of difficulty differentiating these entities clinically.⁶⁴  Because there are no widely accepted treatment recommendations for NEC, the development of our guideline was on the basis of review of existing literature and local consensus. We were limited to 2 years of baseline data because of a change in the EMR in 2016. Finally, because of the timeline of our project, we were able to collect data on short-term balancing measures but not on long-term outcomes such as neurodevelopment.

Development of a guideline and supporting EMR constructs decreased variation in the management of NEC without increasing morbidity or mortality. Guideline adherence led to a decrease in excess use of antibiotics and NPO days, which may reasonably lead to improved long-term quality outcomes. Even in other pediatric and neonatal diseases with similarly limited evidence-based treatment standards, providers should consider implementing locally developed guidelines to benefit from decreasing practice variation and to optimize quality and safety outcomes.

First and foremost, we thank our patients and their families. We thank the nurses, residents, fellows, attending physicians, and support staff involved in the care of these patients in the MGH NICU. We thank Lee Hang, PhD, for biostatistical support through the Harvard Catalyst.

EMR

electronic medical record

MGH

Massachusetts General Hospital

NEC

necrotizing enterocolitis

NPO

nil per os

PDSA

Plan, Do, Study, Act

QI

quality improvement

SIP

spontaneous intestinal perforation

VLBW

very low birth weight