Quality Initiative to Reduce Extrauterine Growth Restriction in Very Low Birth Weight Infants Free

The incidence of extrauterine growth restriction (EUGR), defined as discharge weight <10th percentile for postmenstrual age (PMA), remains high in very low birth weight (VLBW) infants.^1–4  Multiple potentially better clinical practices (PBPs), including early and aggressive parenteral nutrition, the use of human milk, and early initiation and steady increase of enteral feeds, have been demonstrated to improve infant growth.^5–7  Although these proven PBPs exist, there is still wide variability among neonatal providers and centers in their implementation and attainment, including inconsistent initiation of parenteral and enteral nutrition and interruptions in enteral feeds. Quality improvement techniques can be used to decrease variability in practice and improve outcomes.⁸ 

As recommended by national and state perinatal guidelines, the majority of VLBW infants born in Tennessee are cared for in level III and IV NICUs.^9,10  Upon review of their respective data, many of these centers found that their incidence of EUGR was above the national average. This realization spurred the selection of this project and the development of the statewide initiative (Supplemental Information).

Using the Institute for Healthcare Improvement collaborative learning model,¹¹  the Tennessee Initiative for Perinatal Quality Care (TIPQC) aimed to decrease the incidence of EUGR among VLBW infants cared for in Tennessee NICUs by 10% from May 2016 to December 2018.

Nine NICUs from across the state (academic and nonacademic [Supplemental Information]) participated in the quality improvement initiative from May 2016 through December 2018. Data were prospectively captured on all VLBW infants (birth weight 500–1499 g and appropriate for gestational age) born between May 1, 2016, and December 31, 2018 (32 months), with a gestational age of 23 to 32 weeks. Infants with known genetic or chromosomal anomalies and those who underwent bowel surgery were excluded. Outborn infants admitted after the first 12 hours of life (HOL) were also excluded. To assess extrauterine growth, all included infants had a minimum 4-week hospital stay.

This project was introduced and kicked off at the annual statewide TIPQC collaborative meeting in spring 2016. At this meeting, a tool kit of evidence-based PBPs was discussed by experts in neonatal nutrition. After reviewing the literature, a multidisciplinary development team identified the PBPs. The PBPs included in the tool kit were (1) starting parenteral nutrition (glucose and protein) and intravenous lipid emulsion within 24 HOL, preferably soon after admission to the NICU (ie, within 12 HOL)¹² ; (2) starting enteral nutrition (mother’s or donor’s milk when mother’s milk is not available) within 48 to 72 HOL¹³ ; (3) standardizing human milk fortification when enteral intake is 50 to 100 mL/kg per day¹⁴ ; (4) assessing weight and nutritional intake daily¹³ ; (5) discontinuing the measurement of gastric residuals as a measure of feeding tolerance unless there are other clinical indications¹⁵ ; and (6) promoting intermittent bolus gastric feeding and implementing measures to minimize nutrient loss in tubing and feeding systems.¹⁶ 

Each participating NICU was responsible for developing protocols and implementing standardized practices chosen from the tool kit. Each team monitored their improvement by collecting the agreed-upon data and performing plan-do-study-act cycles.¹⁷  In these plan-do-study-act cycles, the teams included tests of change that were not part of the tool kit, such as working with their respective pharmacies to have stock total parenteral nutrition solutions available at all times and to have more frequent delivery of intravenous lipid emulsion, allowing for earlier administration. The teams met every month by webinar and in person at semiannual regional learning sessions. During the monthly and semiannual meetings, discussion included quality improvement methods, review of the state aggregate data, successes and challenges to implementing the PBPs, and the latest evidence related to the topic and PBPs. The quality improvement methods discussed included use of checklists, flowcharts, Pareto charts, driver diagrams, Ishikawa diagrams, and process maps. The involvement of parents in the team projects was also strongly encouraged throughout the process.

TIPQC was started in 2008 by a group of neonatologists and obstetricians from across the state led by Dr Judy Aschner with support from TennCare and the Tennessee Department of Health. TIPQC seeks to improve health outcomes for all mothers and infants in Tennessee by engaging key stakeholders in a perinatal quality collaborative that identifies opportunities to optimize maternal and infant outcomes and implement data-driven provider- and community-based performance improvement initiatives. When work began on this project, TIPQC had already facilitated 16 other maternal and infant statewide quality improvement projects.

The following key data were collected on each infant: gestational age, sex, and birth weight. The infant’s body weight measured closest to 36 weeks PMA (between 34 and 38 completed weeks) was also captured.

The primary outcome of the project was the incidence of EUGR at 36 weeks PMA defined as a weight of <10th percentile for PMA.¹⁸  The secondary outcomes were (1) growth velocity (exponential 2-point method)¹⁹  and (2) change in weight z score.^20–22  Both secondary outcomes were measured between birth and 36 weeks PMA. Sex-specific body weight percentiles and z scores were calculated as per Fenton and Kim.¹⁸ 

The 4 captured process measures were the HOL when parenteral protein and intravenous lipid emulsion were initiated and the number of days to first enteral feed and attainment of full enteral caloric intake (110–130 kcal/kg per day). The diagnosis of necrotizing enterocolitis (NEC), based on Bell criteria,²³  was captured as a balancing measure.

Although formally defined structure measures were not captured by each participating hospital, participating centers completed a postproject survey to assess their nutritional practices and their perception of whether participation in the project assisted in achieving the PBPs and benefited their nutritional management of VLBW infants. The potentially achieved benefits included in the survey were (1) no longer checking residuals in infants receiving tube feedings, (2) the development of a standard definition of feeding intolerance, (3) a policy/guideline in place that discusses human milk fortification, (4) a policy/guideline in place that discusses the proper handling of human milk, and (5) a policy/guideline in place that discusses what to do if the baby is not gaining an average of 10 to 20 g/kg per day.

Statistical process control charts were created for the process, primary, and secondary outcome measures. A p-chart was used to analyze the primary outcome, whereas X-bar charts were used to analyze the secondary outcome and process measures. The center line was calculated (and initially fixed) as the mean of the first 4 months of data (the baseline period). Upper and lower control limits were set at 3 SDs, with the lower control limit constrained to be >0 for the X-bar charts of the process measures. Shewhart control chart rules were used to indicate special cause variation (after the baseline period), ≥1 points outside the upper or lower control limit in the direction of improvement, or runs of ≥8 consecutive points on 1 side of the center line. Shifts in the center line were implemented in the data point immediately following the indication of special cause variation. Statistical process control charts were created using the qcc (version 2.7) package in R (version 4.0.2).²⁴ 

Because of the low monthly frequency (<5) of VLBW infants who met the project criteria at most of the 9 participating hospitals (because of the relative size of each NICU and varying rates of infants born appropriate for gestational age), the hospitals were aggregated into 2 groups: 4 regional perinatal center (RPC) hospitals versus 5 non-RPC hospitals. We theorized that the hospitals would perform and improve similarly within these 2 groupings. This aggregation grouped hospitals with similar facilities and monthly frequency of VLBW infants while also ensuring a large enough monthly subgroup size. Our state’s perinatal regionalization system began in the early 1970s and provides the necessary statewide infrastructure for high-risk perinatal care. The RPCs are all long-standing NICUs at academic centers with similar substantial resources for supporting and monitoring nutrition in VLBW infants, including full-time neonatal nutritionists.

A total 1367 VLBW infants met inclusion criteria, which represents ∼40% of all VLBW infants born in Tennessee during this period.²⁵  Between the RPC hospitals and the non-RPC hospitals, >60% of the infants were 28 to 32 weeks gestation at birth and had a median birth weight of ∼1100 g (Table 1).

Although no special cause variation was indicated in the RPC hospitals for the primary or secondary outcomes, these hospitals showed movement in the direction of improvement for all 3 outcome measures in the last 6 to 7 months of the project (Fig 1). In the baseline period of the RPC hospitals, the incidence rate of EUGR, the mean growth velocity, and the mean weight z score were 31.0%, 13.7 g/kg per day, and −0.87, respectively.

Like the RPC hospitals, no special cause variation was indicated in the non-RPC hospitals for the reduction of EUGR (Fig 1A). The incidence of EUGR in the baseline period for the non-RPC hospitals was 44.2%. Special cause variation was technically indicated in the non-RPC hospitals for growth velocity (Fig 1B), but the average quickly decreased back to the baseline center line, so a shift was not implemented. The mean growth velocity in the baseline period was 12.9 g/kg per day in the non-RPC hospitals. Special cause variation was also indicated in the non-RPC hospitals for change in weight z score (Fig 1C) but the shift was small (mean of −1.11 in the baseline period to a mean of −1.07).

Special cause variation was indicated in both RPC and non-RPC hospitals for both HOL when parenteral protein was initiated (Fig 2A) and HOL when intravenous lipid emulsion was initiated (Fig 2B). The mean HOL when parenteral protein was initiated shifted in the RPC and non-RPC hospitals from a baseline value of 5.1 and 3.7 hours, respectively, to 3.7 and 2.3 hours, respectively. There was a larger decrease in the mean HOL when intravenous lipid emulsion was initiated, specifically shifting in the RPC and non-RPC hospitals from a baseline value of 22.0 and 19.4 hours, respectively, to 11.7 and 9.1 hours, respectively.

No special cause variation was indicated in either group for the number of days to first enteral feed, but the non-RPC hospitals showed a movement in the direction of improvement in the last 7 months of the project (Fig 2C). The mean number of days in the baseline period was 1.7 in both RPC and non-RPC hospitals.

Although the RPC hospitals did not show special cause variation in improving their number of days to attainment of full enteral caloric intake (110–130 kcal/kg per day), the non-RPC hospitals did (Fig 2D). The mean number of days in the RPC hospitals was 16.3 in the baseline period, and the mean number of days in the non-RPC hospitals improved from 15.8 during the baseline to 12.0.

The incidence of NEC was overall low among the VLBW infants in the RPC and non-RPC hospitals (<5%) and did not change significantly during the project years (Table 2).

One of the many benefits to the monthly huddles and semiannual meetings was the opportunity for the teams to discuss the logistics of their successes, such as how they worked with their respective pharmacies to have parenteral amino acid solutions stocked in the NICUs and had intravenous lipid emulsions delivered twice a day rather than the previous practice of once a day, therefore allowing for earlier initiation. Additional discussions among teams included best practices for measuring linear growth, the importance of daily weights and nutritional calculations, best practices for delivering enteral nutrition through tube feedings to minimize nutrient loss in the feeding tubes, and how and when to provide human milk fortification.

The postproject survey revealed that the majority of the teams believed that being a part of this project helped to make nutrition more of a priority in their NICUs. At least half of the centers also identified additional benefits as a result of their participation in this project (Table 3), including initiating enteral feedings within 1 to 2 days of birth and no longer checking gastric residuals in infants receiving tube feedings.

Through their involvement in this collaborative project, our participating hospitals were able to show significant improvements in implementing some of the potentially better nutritional practices and increase awareness about the importance of nutrition and growth in VLBW infants. Although special cause variation was not indicated in any of the outcome measures (primary or secondary), it was indicated for the reduction in specific process measures: the HOL when parenteral protein was initiated, the HOL when intravenous lipid emulsion was initiated, and the number of days to attainment of full enteral caloric intake (among the non-RPC hospitals). Importantly, these positive results were noted with no major change in the incidence of NEC.

Strengths of our project were the participation by 4 of the 5 RPC hospitals in our state and the inclusion of both academic and nonacademic units. The large number of infants enrolled represents ∼40% of all VLBW infants born in Tennessee during the period of this quality improvement initiative. Additionally, through this project, we were able to bring the best practices for nutrition to the state’s level III and IV NICUs. These best practices include early initiation of parenteral and enteral nutrition, elimination of routinely checking gastric residuals to assess feeding tolerance, early fortification of human milk, and a more aggressive approach to increasing enteral feeding volumes.

We also found external evidence of the improvement we saw in the project. Eight of the 9 hospitals participating in this project are part of the Vermont Oxford Network (VON). According to the VON reporting system for VLBW infants,²⁶  the median percentage of EUGR in all VLBW infants who survived to discharge across these TIPQC VON hospitals dropped from 58.3% to 50.9% between 2016 and 2018, a 13% reduction in EUGR. In comparison, the median percentage of EUGR in VLBW infants who survived to discharge across all VON hospitals in the United States dropped by only 3% (from 48.8% to 47.2%) in the same time frame. The VON data also revealed the TIPQC rate of EUGR approaching the national rate in 2018. Unfortunately, the data also indicate that improvement as not sustained in 2019 after the completion of our initiative.

There were several other limitations to this initiative. Although the initiative lasted 32 months, we may have seen continued improvement (or at least sustainment) with more time and support for the implementation of the identified best practices. The need to integrate many different hospital departments may also be seen as a limitation. For example, some facilities were challenged by their pharmacy’s requirements, thus hindering their ability to start intravenous lipid emulsion as soon as the providers would have preferred. An additional limitation is that our project did not include race, ethnicity, or other demographic data. At the time this project was developed, the inclusion of such data was not part of our purview. Process measures were only formally defined and captured for 2 of the PBPs. In addition, data around the receipt of breast milk (mother’s or donor’s) and formula feeding were not collected, which affected the outcome, process and balancing measures of this project. We also lacked adequate monthly sample size at all 9 participating hospitals to analyze the outcome and process measures at an individual hospital level and compare the improvement across the hospitals to see whether the initiative had more influence at some hospitals than others. Failure to have each participating hospital formally capture when it implemented each PBP in the tool kit also made this analysis difficult. Likewise, the participating hospitals did not implement each PBP at the same time. In turn, although we saw improvement in some measures, we may have seen more convincing evidence of special cause variation if our initiative and participating hospitals had been more systematic and aligned in the sequence and initiation of the PBPs.

Although we monitored specific process measures within the first few days of life, we did not capture data on caloric and protein intakes throughout the hospital stay. We also did not capture whether interruptions or decreases in enteral feedings occurred (eg, feeding intolerance, NEC concerns, blood transfusions, and indomethacin administration). We tried to limit the data burden for the participating hospitals, but the lack of more frequent capture of intake, interruptions, and growth metrics supports why we saw an improvement in our process measures but not in our outcome measures.

The success of our quality improvement project reveals the benefits of collaborative learning and quality initiatives, as other projects have done before.^5,7 

The authors thank Tamara J. Wallace, DNP, APRN, NNP-BC, and Peter Grubb, MD, for their contribution to the development of this improvement project in 2016.