Decreasing Time to Full Enteral Feeds in Hospitalized Children With Medical Complexity Experiencing Feeding Intolerance

This was a QI study based in the model for improvement¹⁷  conducted on the HM service at a large quaternary care pediatric health system with >700 hospital beds. The HM service cares for >8000 patients per year, including >500 children who have neurologic impairment and/or require medical technology assistance. At the main campus, the HM service has 5 clinical teams, including a dedicated, interdisciplinary complex care team. The complex care team caps at 10 patients and is staffed by an attending physician and/or HM fellow, 4 residents, 2 nurse practitioners, a social worker, a dietitian, a care manager, and a pharmacist. CMC are also frequently cared for on the other general HM teams.

Before our work, when patients were admitted to the HM service with feeding intolerance, they would typically be started on intravenous fluids with enteral feeds withheld. Once the decision was made to transition to enteral feeds, a low-calorie enteral electrolyte solution (eg, Pedialyte) was often initiated at a low volume and rate. After enteral feed reintroduction, management varied greatly, including decisions to start, stop, or alter feeding plans on the basis of perceived tolerance or intolerance, even in patients with similar presentations and underlying conditions (Fig 1).

We assembled a multidisciplinary team in June 2019 to improve the care of CMC dependent on enteral nutrition who experience feeding intolerance during hospitalization. Team members included 2 nurse practitioners, a bedside nurse, 2 resident physicians, a HM fellow, 3 HM attending physicians, and a registered dietitian with experience caring for CMC. We also engaged key stakeholders as consultants, including 5 parents of CMC, 2 gastroenterology physicians who care for children with intestinal failure, formula room technicians, additional nurses, and an ICU dietitian.

Based on the baseline data, which indicated a median time to full enteral feeds of 3.5 days, our team chose a goal of 2.5 days. This would result in a clinically meaningful increase by an entire day of full enteral nutrition. This also aligned with parent team members’ goals to maximize time at home, because patients could potentially be ready for discharge sooner.

To inform intervention development, our team also reviewed the electronic health record (EHR) for 60 children with feeding intolerance in the 6 months before our study to determine the most commonly documented reasons for feeding intolerance and constructed a Pareto chart (Fig 2).

Based on chart review, process observation, input from key stakeholders, and through a modified failure modes and effects analysis,¹⁸  our improvement team identified key drivers of decreasing time to goal enteral feeds (Fig 3), including standardized approach to managing feeding intolerance, parental buy-in and shared understanding of parental goals, timely delivery of formula to the room, and provider knowledge. Key drivers guided intervention development; interventions were then tested and refined through sequential Plan-Do-Study-Act cycles.¹⁷ 

By engaging expert stakeholders and considering feeding protocols for critical care populations, we created a feeding intolerance management algorithm.^11–16  The algorithm included guidance for evaluating the most common treatable causes of feeding intolerance (Fig 2), when to consult gastroenterology experts, caution for when not to feed patients (eg, enterocutaneous fistula), and how and when to titrate feeds. In addition to providing a framework for feed titration, the algorithm also deemphasized low-calorie electrolyte solutions on the basis of existing evidence^19–23  and concern for delays in reaching goal feeds. The algorithm recommended partnering with families throughout hospitalization to obtain their buy-in to advance feeds using the algorithm, obtain their input on signs of feeding intolerance that would indicate failure, and determine appropriate discharge regimen based on the family’s comfort with titrating feeds at home.

The algorithm was then presented to broader stakeholder groups, including pediatric residents, hospitalists, and the interdisciplinary inpatient complex care team for feedback and then tested initially on individual patients. Team members reached out to hospital medicine teams on which patients with feeding intolerance were admitted and asked them to consider trialing the feeding algorithm. On the basis of qualitative feedback from patient parents and providers after each test, iterative changes were made. Feedback from Plan-Do-Study-Act cycles resulted in several adjustments to the algorithm, including prioritizing collaboration with parents as an early step and notation of additional exclusions (ie, bilious output). We conducted 5 iterative rounds of improvement before distributing the final algorithm (Fig 4). The algorithm was also posted on the resident resource Web site within our institution’s intranet for easy access.

To address the “provider knowledge” driver, presentations to stakeholder groups detailed rationale for the project and interventions, including resident educational conferences, HM provider meetings, and inpatient nutrition support meetings. Education on the use of the feeding algorithm was provided. Additionally, the Pareto chart (Fig 2) illustrated that causes of feeding intolerance are usually reversible; once treatment is initiated, feeds can likely be restarted. For example, we found that a large percentage of patients experienced feeding intolerance because of acute gastroenteritis. Subsequently our education to the residents emphasized that rapid reinstitution of an appropriate diet has been associated with decreased stool volume and duration of diarrhea. Ongoing education was sustained with monthly E-mail provider reminders and inclusion of algorithm introduction at monthly resident HM rotation orientation sessions.

To address the driver of timely formula delivery to the bedside, we explored the process of how formula was ordered and delivered to patients. Our team discovered a misperception among staff, resulting from when the formula room was under construction, that formula could only be ordered at certain times, and noted that outdated information remained posted on the nursing units. Our team subsequently facilitated dissemination of updated information through nursing leadership, and bedside nurses began calling for formula on-demand again.

Our team noted through chart review that patients admitted overnight with feeding intolerance were often placed on intravenous fluids without attempting feeds. Residents noted that they were unable to order certain specialized formula recipes after the formula room closed in the evening. Formula room staff confirmed that a courier was available to retrieve formulas after hours and that providers could call ahead to request formula to be prepared for patients who had not yet arrived to the inpatient unit. This information, including the number for the courier, was shared with residents and nurses and added to the feeding algorithm.

At the launch of improvement efforts, QI team members identified patients admitted for feeding intolerance and contacted attending physicians reminding them to consider using the feeding algorithm if appropriate, which provided real-time reminders of this new tool. Through chart review, team members also identified patients who were eligible to be managed by feeding algorithm but were managed in an alternate manner and then reached out to the medical team to elicit the reasons why the feeding algorithm had not been used. This served as a form of audit and feedback²⁴  and allowed our team to collect information on barriers to using the feeding algorithm (eg, feeding algorithm not readily accessible for reference on rounds) and lead to interventions (eg, posting of the feeding algorithm on the resident resource Web site within our institution’s intranet).

We retrospectively collected baseline time to goal enteral feeds and LOS for hospitalized children with GT or GJ tubes who experienced feeding intolerance and were admitted to a HM service from February 1, 2019 to September 1, 2019 from the EHR. Data were then followed biweekly throughout the intervention period to understand the impact of our interventions and facilitate rapid-cycle learning.

Given that there is not a single, standardized billing code for feeding intolerance, we identified patients by manually reviewing charts of children with GT or GJ tubes admitted to HM. We included patients who, on the basis of provider documentation experienced symptoms consistent with gastrointestinal dysfunction (eg, vomiting, retching, pain, agitation, and/or abdominal distension) that occurred with administration of feeds and subsided with cessation of feeds.⁶  We excluded patients who received chronic total parenteral nutrition or who received a surgical consult during admission, because these patients’ feeds were often managed by the surgical team because of an underlying condition for which the management included nil per os status (eg, obstruction, pneumatosis intestinalis). We excluded scheduled admissions.

Given that feeding order timestamps were not always representative of when patients achieved goal feeds, time to goal enteral feeds was determined through chart review of progress notes and measured in days. Goal enteral feeds were defined as the patient’s home, baseline feeds which were typically documented in the history and physical or in the patient’s problem list under “nutrition diet or plan,” or the enteral feeding regimen acceptable for discharge as determined by the team and family and documented in the chart. The day goal enteral feeds were documented in the daily progress note as achieved was used as the timestamp for our primary measure. Some parents preferred to continue feed titration after discharge. If a child was achieving full feeds before, we counted the day of discharge as the day goal enteral feeds were achieved. Our primary measure of median time to goal enteral feeds was measured in increments of 10 patients, given the unpredictable patient volumes of children with feeding intolerance on the HM service week to week. The secondary measure of LOS, calculated as time in days from admission to hospital discharge, was also followed over time. We tracked pediatric ICU (PICU) transfers and 7- and 30-day readmissions events as balancing measures to ensure increasing time to feeds was not placing patients at risk for adverse events.

We examined cohort demographic and clinical characteristics using descriptive statistics. A run chart tracked the effect of interventions on median time to goal enteral feeds and median LOS. We used established run chart rules to identify nonrandom signals and determine when to change the centerline.²⁵  Age of the baseline group and intervention group were compared by using a 2 sample t test. Sex, race, route of enteral feeds (GT versus GJ tube), readmission rates (7- and 30-day), and rates of PICU transfer events were compared by using a χ² test. Underlying reasons for feeding intolerance were compared between the baseline and intervention groups using Fisher’s exact test.

Of 225 patient encounters during the study period (February 2019 through January 2021), there were a total of 146 unique patients. We identified 60 patient encounters during the baseline period (February 1, 2019 to September 1, 2019) and 165 patient encounters during the intervention and sustain period (September 1, 2019 to June 15, 2021). Patients in the baseline period did not differ significantly from patients in the intervention and sustain period (Table 1). ∼85% of children were discharged on full volume and frequency of feeds in both the pre- and postintervention time periods. The most common underlying causes of feeding intolerance were viral gastroenteritis, respiratory infections, ileus, constipation, and urinary tract infection (Table 1).

Baseline median time to goal enteral feeds for children with GT or GJ tubes who experienced feeding intolerance during hospitalization was 3.5 days (Fig 5). We documented a reduction of median time to goal feeds to 2.5 days within 6 months of improvement interventions, which was temporally associated with widespread educational efforts to orient clinicians to the feeding intolerance management algorithm. This reduction was sustained even during the COVID-19 pandemic. Improvement was sustained with subsequent interventions, which included continued monthly resident orientation session education, monthly E-mail provider reminders, and feedback.

For the duration of the intervention period, median LOS remained unchanged at 4.5 days (Fig 6). Although there was no significant difference in 7- or 30-day readmission rates or PICU transfers during hospitalization, 7-day readmissions were increased in our intervention group when compared to our baseline group. (Table 1).

Using improvement science methods, we decreased median time to goal enteral feeds from 3.5 days to 2.5 days in children with medical complexity who are GT or GJ tube dependent admitted to the HM service over 1 year. Our key interventions involved implementing a standardized feeding intolerance management algorithm, provider education, correction of formula room misinformation, optimized physician ordering practices, reminders, and near–real-time feedback. A nonrandom signal of change was temporally associated with introduction of a standardized feeding algorithm, suggesting it was a particularly impactful intervention.

Previous studies have shown that standardized management of other common pediatric diagnoses (eg, pneumonia, urinary tract infections) in otherwise healthy children improves quality of care.^26–28  Additionally, there is a growing body of evidence within the neonatal and PICU populations, who like CMC are at risk for malnutrition and feeding intolerance, that supports using standardized feeding protocols can improve outcomes, particularly shortening the time to reach full enteral feeds.^11–14  Using a standardized feeding guideline to guide enteral feeding advancement in preterm infants, Chu et al reduced average time to full feeds by 5 days and central line days by 35%, without increasing complications.¹⁶  In the PICU, Ziemba et al used an early enteral nutrition feeding protocol to safely increase the percentage of critically ill children who reached goal enteral feeds in the first 48 hours of admission by 20%.¹³  Our study has expanded on this work by focusing on CMC, a population of children frequently excluded from research and quality improvement efforts,²⁹  highlighting the feasibility of implementing a standardized management protocol to treat a common issue in a complex, heterogenous group of patients.

The feeding algorithm also promoted family and provider collaboration on feeding plan development. Promoting parental engagement in decision-making for hospitalized children is a key component of family-centered care endorsed by the American Academy of Pediatrics Committee on Hospital Care.³⁰  Previous studies have shown that effective provider-family communication leads to safer care.^31,32  Khan et al conducted a prospective, multicenter study utilizing a structured, standardized mode of communication with families on rounds and achieved a significant decrease in harmful errors and nonpreventable adverse events, with increased family experience scores.³¹  Families of CMC bring experience with their child’s previous illnesses and knowledge of their child’s diseases, medications, and medical technologies. Our experience using the feeding algorithm reinforces that caregivers of CMC bring valuable viewpoints and expertise on the care of their children and should be actively engaged during hospitalization. Most relevant to the success of implementing our feeding algorithm, partnering with families early in the process was critical for gaining buy-in. Our hypothesis is that many families were open to trying something different in part because they were provided a clear explanation of the plan with opportunity to have input.

During our project, we identified delays in formula order and delivery were not because of inefficient processes, but simply lack of knowledge of existing processes. Through discovery of misinformation about formula room policies (eg, on-demand delivery and overnight courier availability), our study also emphasized that sharing and sustaining important clinical process knowledge in a large, complex health care system is challenging.

Although patients were able to achieve goal feeds more quickly, there was no change in LOS. We hypothesize this is because of the complicated and multifaceted nature of medical readiness for discharge in CMC³³  and that CMC may experience multiple reasons for hospitalization simultaneously (ie, bacterial pneumonia leading to an increased in oxygen requirement in conjunction with feeding intolerance).³⁴  An overall improvement of time to full enteral feeds is valuable, even without LOS change, because it allows for full nutrition to an acutely ill cohort of children who are at high risk of undernutrition, growth failure, micronutrient deficiencies, and osteopenia.^35–37 

It should be noted that our post intervention group had a higher rate of PICU transfer and 7 day readmission than our preintervention group. Patients in our intervention group who were transferred to the PICU and readmitted within 7 days were chart reviewed, and there was no clear documentation tying PICU transfer or 7 day readmission to the way feeds were initiated. Further, our 7 day readmission rates were low in the baseline cohort when compared to our typical 7 day readmission rates for complex patients, and the 7 day readmission rates for the intervention cohort are more reflective of our baseline readmission rates for our inpatient complex care service (10.8% during the study period). Although these higher rates of transfer and 7-day readmission did not reach statistical significance, future directions should include continued monitoring of 7 day readmission rates to determine if significance emerges, adequately powered studies looking at differences in readmission and PICU transfer for patients who receive standardized feeding protocols versus those that do not.

This study occurred at a single center at an institution with a dedicated, highly-resourced HM complex care team, and generalizability of results may be limited. However, this algorithm was also used on our general HM teams; thus, the feeding intolerance management algorithm can likely be applied to other institutions who care for CMC. Because of data retrieval constraints relating to no reliable, standardized code for feeding intolerance, manual chart review for our primary measure was required, which has inherent limitations in reliance on accurate documentation of achievement of goal feeds in progress notes. Although this method did not capture exactly when the patient achieved goal feeds, we thought this method of measurement was superior to other avenues of measurement such a formula order timestamps, given that it overestimated time to goal feeds rather than underestimated it. We were also unable to measure usage of the feeding algorithm as a process measure; however, we noted improvement following wide distribution of the algorithm. Lastly, given that many of our interventions were people-dependent, our project will benefit from continued implementation of high reliability interventions (ie, integrating principles of the feeding algorithm into EHR orders).

Through education, correction of misinformation, partnership on care plans with families, and a standardized feeding intolerance algorithm, we decreased time to full enteral feeds for CMC admitted with feeding intolerance. Ongoing efforts focus on sustainability, including integration of elements of the feeding intolerance management algorithm into EHR order sets and spread to other service lines within our institution that frequently care for CMC. Our project also identified the need for further clinical research into management variation of feeding intolerance as it relates to hospital and nutritional outcomes.

The authors thank the parents and providers who participated in the creation of the feeding intolerance algorithm including Jamie Swartzel, Conor Merritt, MD, Lacey Yeager, MSN, APRN, Stephanie Oliveria, MD, CNSC, and Marialena Mouzaki, MD, MSc, and Setenay Kara PhD, who provided QI support during the intervention period of this project.