Introduction

Diagnosis and management of the feeding process in preterm infants can be variable from admission to discharge among multi-disciplinary providers in the neonatal intensive care unit (NICU). Such multi-disciplinary practice variation can impede discharge planning and prolong length of hospital stay (LOHS) thereby escalating the economic burden of preterm birth [1]. For example, from the data during 2008-16, the average expenditure for an infant born at 24-week gestational age (GA) was $603,778 (standard deviation $509,165) and these costs continued to increase after discharge [2,3,4,5]. Presence of comorbidities such as bronchopulmonary dysplasia (BPD), necrotizing enterocolitis (NEC) and low birth weight double the mean NICU costs for premature infants [2]. The American Academy of Pediatrics recognizes independent safe oral feeding as a fundamental milestone to achieve before a preterm infant can be discharged home [6]. This goal is affected by co-morbidities such as apnea or respiratory problems, as well as the patient-parent-provider process during the NICU stay [6,7,8,9,10,11,12,13]. Dysphagia is more common in preterm infants which can often delay discharge [14]. Thus, safe and adequate oral intake is dependent on multi-system interactions and neuromuscular regulatory processes involving the airways, breathing, swallowing, cardiorespiratory function, and intestinal motility [15,16,17,18,19,20,21]. On the other hand, the development of a successful oral feeding plan for each infant requires understanding the infant’s pathophysiology as well as the NICU processes and policies, including how they may change over time.

Maturation of feeding skills accelerates with adequate nutrition and consistent practice particularly during the phase of rapid growth, and this development can be delayed by the slower implementation of enteral/oral feeding opportunities [22,23,24]. Lack of safe full oral skills leads to discussion of alternative feeding pathways for discharge via either a nasogastric tube or gastrostomy which are often resisted by parents and increase the socio-economic burden to families and the health care system [11, 25, 26]. The development of standardized feeding approaches has been found to decrease the incidence of home tube feeds, accomplish earlier acquisition of feeding milestones, and decrease LOHS in premature infants [10, 15, 25, 27,28,29,30].

In 2010, a simplified, individualized, milestone-targeted, pragmatic, longitudinal and educational (SIMPLE) feeding program was developed to enhance the attainment of feeding milestones in the tertiary care NICU setting at Nationwide Children’s Hospital in Columbus Ohio, which is an all-referral level IV NICU [1]. The SIMPLE feeding program was designed through an evidence-based process using available research and knowledge with the goal of optimizing feeding milestones and decreasing both the need for tube feedings at discharge and LOHS. The comparison group were patients admitted during the 15 months prior to the inception of the program, who satisfied the same inclusion and exclusion criteria [1, 29].

While the program became part of our unit culture, persistent variability was to be expected due to the large number of providers, inherent practice variability, and time-related changes including personnel turnover, new research evidence, improved survival rates, and changes in policies at the national and institutional level. To document the continued effectiveness of our feeding program over time, including time-related variations in care such as the COVID-19 pandemic, we analyzed our 10-year experience with a view to develop mathematical models to predict outcomes and practices with the wealth of data which had been collected. Given this rationale, our aims in this study were to: (1) Compare the differences in clustered gestational age groups in the attainment of feeding milestones, discharge feeding outcomes, and resource utilization, (2) Develop models to predict outcomes based on feeding milestones and co-morbidities, and (3) Assess maintenance and variation within the SIMPLE feeding program over a 10-year period based on acquisition of feeding milestones and outcomes. Specifically, we tested the pre-determined hypotheses that the improvement in feeding milestones initially seen with the standardized feeding initiatives in premature infants would be maintained and that the feeding milestones would be distinct based on GA category.

Materials/subjects and methods

Patients

Inclusion and exclusion criteria remained consistent over time since the inception of the SIMPLE feeding program [15]. Briefly, all preterm infants who were born at ≤32 weeks GA, were ≤34 weeks PMA on admission, and survived to discharge without exclusion criteria were screened. In this way, the patient cohort was less heterogeneous. Exclusions included intraventricular hemorrhage (IVH) grade III or IV, neonatal abstinence syndrome, genetic/chromosomal defects, congenital birth defects, surgery other than patent ductus arteriosus ligation, transfer prior to achieving the full oral feeding milestone and death prior to discharge. All excluded infants had complex diagnoses and therapeutic approaches requiring referrals for sub-specialty interventions. The excluded conditions did not fit into a standardized feeding approach and required a more personalized approach based on co-morbidities and results of other tests and interventions. Patients were stratified based on GA into periviable (<24 weeks GA) [31, 32], extremely preterm (24 to <28 weeks GA) [32] and very preterm (≥28 to 32 weeks GA) [32].

NICU setting

The study setting was an inpatient level IV NICU which provides all-referral, tertiary care for premature infants where age at referral as well as patient acuity are variable. Multiple providers (neonatologists, nurse practitioners, lactation consultants, occupational therapists, speech therapists, registered dieticians, pharmacists, and registered nurses) were involved with infant feeding. The Nationwide Children’s Hospital Institutional Review Board (IRB) approved this project, and informed consent was not deemed necessary. Permission was granted by the IRB for use of the patient data associated with this project.

Interventions

The SIMPLE program used pragmatic feeding milestone criteria as described previously [1]. These milestones were targeted specifically to the population treated in our particular level IV referral only NICU where complex cases are often seen, many with extended respiratory support and limited total fluid volumes as was seen in our baseline patients. Milestones in another program should be individualized to the population and baselines in that center [29]. Briefly, feeding milestone targets were (a) start of trophic feeding by no later than day of life 3, (b) progression to full enteral feeds defined as 120 mL/kg/day at no later than day of life 14, (c) first oral feeding prior to 34 weeks postmenstrual age (PMA) and (d) full oral feeds defined as at least 120 mL/kg/day by 38 weeks PMA. Infants included in the SIMPLE program received focused multidisciplinary feeding rounds twice a week which included neonatologists, neonatal nurse practitioners, nurses, nutritionists, lactation consultants, occupational therapists, and parents. Personalized education was provided to all participants during feeding rounds regarding feeding milestones, risk factors for feeding difficulties, importance of nutritive and non-nutritive stimulation, gut motility, gastroesophageal reflux, aerodigestive symptoms, feeding tolerance, growth, nutrition, and concerns for dysphagia. Targeted milestones were displayed at the bedside as a reminder to the care team. In addition, didactic sessions were provided addressing progress of the program, accountability metrics and education on varying topics related to feeding, including interesting case studies. Data was collected each week including demographics, diagnoses, feeding milestones, and discharge criteria. Bedside feeding rounds were maintained through June of 2020 when bedside feeding rounds were stopped because of restrictions at the bedside due to COVID-19. At this time virtual feeding rounds were implemented. The feeding team continued to work closely with the point-of-care providers with evidence-based practices including recognizing feeding readiness [33], initiating and advancing cue-based feeds [34], early oral stimulation [18], kangaroo care and breast feeding [35].

Data metrics

The goal of the SIMPLE feeding program was achievement of full oral feeds and decreased LOHS. Analysis of the data included demographic characteristics, feeding milestone data, presence of the co-morbidities BPD and/or IVH (grade I or II), LOHS, discharge feeding outcomes, and feeding outcomes at 6 months and 1 year. Data was collected weekly in a secure database and verified by data managers.

When the SIMPLE program was developed, initial data compared the 92 baseline patients (admitted January 1, 2009–March 15, 2010) with 92 SIMPLE feeding program patients (admitted June 15, 2010-April 30, 2012) [1]. Despite similar clinical characteristics, the feeding milestones and LOHS were significantly improved in the SIMPLE program patients [1]. The current study examines the outcomes and maintenance of the feeding milestone trends and annual variations across the entire 10 years of implementation (June, 2010-June, 2020), not including the 92 baseline patients described in the initial SIMPLE publication [1].

Statistical analysis

SIMPLE program patients (n = 434) were compared among the three gestational age categories using a chi-square or Fisher’s exact test to analyze categorical variables. Kolmogorov-Smirnov was used to test the normality assumption. And then, depending on the result of the normality test, Kruskal-Wallis test or Analysis of variance (ANOVA) was used to analyze continuous variables. Continuous data are presented as mean (SD) and categorical data are presented as n (%) or as stated.

A Pearson correlation examined the relationship between PMA at discharge and timing of acquisition of each feeding milestone including day of life at first enteral feed, day of life at full enteral feeds, PMA at first oral feed, and PMA at full oral feeds. Subsequently, each milestone outcome was analyzed over the 10-year course of the SIMPLE program using generalized linear models. All models were adjusted for GA categories and BPD. Multiple comparison using Bonferroni adjustment was conducted between the starting year of 2010 and the following years.

To explore and uncover structure in the relationship between the LOHS and the feeding milestones as predictors, a scatter plot and non-parametric model (generalized additive model) were fitted. Not all changes in LOHS regarding feeding milestones were simply linear. Therefore, negative binomial regression was fitted to estimate the change in slope for different gestational age periods. The break points or cut points (knots) are decided by data-driven graphical representations. To choose the best breakpoint, multiple models with different knots that were identified from previous analysis were fitted. Akaike’s Information Criterion (AIC) was used to compare between all these regression models. The model with the smallest AIC was used and the knot at 26 weeks GA was considered the best breakpoint. Negative binomial was used to handle the overdispersion problem in the highly skewed LOHS. SAS (version 9.4,) was used to perform the analysis, and a significance level was considered at alpha = 0.05.

Results

Demographic and Outcome characteristics

All 605 NICU infants with GA ≤ 32 weeks admitted between June of 2010 and June of 2020 were screened for eligibility for the SIMPLE feeding program. Exclusions (N = 171, 28.3%) included: 22 admitted after 34 weeks, 9 died prior to discharge, 11 discovered with genetic anomalies, 24 with IVH grade III or IV, 62 had surgery and/or NEC, and 43 were transferred to referring hospitals prior to the acquisition of feeding milestones. The remaining 434 (71.7%) infants met inclusion criteria with 24 subjects <24 weeks GA (periviable) [31, 32], 197 subjects 24 to <28 weeks GA (extremely preterm) [32], and 213 subjects ≥28 weeks GA (very preterm) [32]. The mean number of patients admitted to the program annually over the 10 years was 39 ± 14 (mean ± SD). Demographic, co-morbidity, feeding and discharge outcome data are displayed in Table 1. Co-morbidities, respiratory support at discharge, tube feeding at discharge, gastrostomy tube placement, and tube feeding at 6 months of age were all significantly higher with lower GA. Feeding milestones were attained significantly earlier at increased GA and subjects were discharged at an earlier PMA with decreased LOHS (Fig. 1A–F). The correlation with timing of reaching specific feeding milestones and PMA at discharge (Fig. 2A–D) were significant for earlier initiation of enteral and oral feeding associated with earlier discharge. Regardless of GA at birth, oral feeding capabilities were maintained and improved beyond discharge when measured at 6 months and 1-year corrected age.

Table 1 Demographics, Comorbidities and Feeding and Discharge Outcomes by Gestational Age.
Full size table
Fig. 1: Feeding Milestones and Hospital Stay by Gestational Age.
figure 1

Boxplots illustrate the distribution among the three gestational groups for feeding (A–D) and economic milestones (E, F). Note that the diamond (◊) represents the mean, and the solid line inside the boxplot represents the median.

Full size image
Fig. 2: Correlation of Feeding Milestones with PMA at Discharge.
figure 2

Scatter plots show relationships between PMA at discharge and feeding milestones of DOL at first (A) and full (B) enteral feeds and PMA at first (C) and full (D) oral feeds). The line represents the regression best fit. r: Pearson correlation. A significant correlation was noted between acquisition of all feeding milestones and earlier PMA at discharge, specifically with PMA at first oral feed and PMA at full oral feed.

Full size image

Prediction of Estimated LOHS and Annual Data Trends

An estimation equation for LOHS was developed using GA, presence of BPD, PMA at first oral feed, PMA at full oral feeds and DOL at full enteral feeds (Table 2A). This equation showed a close correlation with observed and predicted LOHS (Table 2B) using our cohort of 434 subjects. To assess predictive performance for the LOHS model, five-fold cross-validation was carried out with a calculation of the Pearson’s correlation coefficient (r) between the observed and predicted values. In five-fold cross validation, the dataset was split into five equal sized groups with four groups forming the calibration dataset and the fifth group the validation set. This process was repeated five times, so that each observation was in the validation set exactly once.

Table 2 A. Equation to Estimate LOHS. B. Verification of Equation with Study Sample.
Full size table

With the SIMPLE program, a lower LOHS and PMA at discharge were maintained over the years (Fig. 3A–F) with less variation over time noted by a decrease in the number of outliers over the 10 years. The feeding milestone goals were maintained, with a reduction in variation as evident by lower standard deviation over the years. However, among the outliers, five infants had LOHS greater than 300 days, of whom, four had severe BPD, were on positive pressure ventilation until at least 60 weeks PMA and were discharged with gastrostomy tubes. The final patient had significant hypoventilation in sleep which did not resolve until after 65 weeks PMA, although the patient reached full oral feeds by 46 weeks and went home orally-fed. There were also 53 infants who were discharged >30 days after reaching full oral feeds due to persistent bradycardia and desaturations, decline in their oral feeding, or attempts to wean off respiratory support. Of the total number of subjects <24 weeks GA, 50% of them were in this group who had a delayed discharge after reaching full oral feeds.

Fig. 3: Feeding Milestones, LOHS, and PMA @ Discharge over 10 Years.
figure 3

Boxplots illustrate the yearly distribution for feeding milestones (A–D) and economic outcomes (E, F); note that improvements were maintained and variability was decreased throughout the study period. The star represents the mean and the solid line the median. The discharge outliers are shown as blue dots and diamonds; (•) represent those with full oral feeding; and the diamond (◊) represents tube feeding at discharge.

Full size image

Discussion

Salient findings and implications

Prolonged LOHS adds significant economic burden for parents and health care systems overall with feeding method being an important determinant for the discharge of high-risk infants from the NICU. In the SIMPLE program participants, we compared the differences in clustered GA groups in the attainment of clinically important feeding outcomes and found that lower GA at birth was associated with later PMA at attainment of feeding milestones. The more premature an infant is at birth, the less swallowing experience they have in utero. Swallowing of amniotic fluid begins as early as eleven weeks gestation with sucking developing at about 18 weeks gestation [21, 36]. By 28 weeks gestation, the fetus swallows around 250 mL/kg/day of amniotic fluid [36]. Development of swallowing function requires neuromuscular stimulation, sustained practice, and time for the maturation of the sensory-motor apparatus involved with the development and adaptation of safe aerodigestive reflexes [22, 23]. After birth, premature infants are not able to nutritively swallow for a prolonged period of time which increases with lower GA. This skill must then be relearned when oral feeds are introduced. Premature infants have modifiable neurologic and aerodigestive physiology that makes timely interventions imperative [37]. Thus, time (growth and maturation) and practice (repetition and adaptation of aerodigestive skill with the feeding process) are the most effective treatment strategies for improving feeding difficulties in the preterm infant [37, 38].

Predictive models for LOHS based on GA, feeding milestones and co-morbidities were developed and the final model was tested to determine actual versus predicted outcomes, and found very close correlation between actual and predicted LOHS. The attainment of the three most important feeding milestones, i.e., full enteral feeds, 1st oral feed, and full oral feeds varied among the three GA groups. These milestones, along with GA at birth and the presence or absence of BPD, were excellent predictors of LOHS. Half of the infants in the SIMPLE program had varying levels of respiratory support at discharge, and the need for respiratory support at discharge was inversely proportional to gestational age at birth. BPD rates were essentially 100% in the <24 weeks GA group, and decreased with increased gestational age at birth, with the prevalence of BPD reaching 36% in those born at greater than 28 weeks GA. As NICUs are taking care of more extreme premature infants, it is important to identify appropriate milestones and set realistic expectations for these infants born at the lowest GA. Predictive prognostic models can assist caregivers and parents with joint decisions about long term treatment courses and preparation for discharge [39]. The results of this study can be used in developing maturation-specific safe discharge feeding plans, and in providing anticipatory guidance to parents and providers as well as to payors.

Practice variation was examined over the 10-year period by measuring feeding milestones, feeding outcomes, and LOHS by year. The SIMPLE feeding program has shown that earlier acquisition of feeding milestones led to earlier discharge, which in turn led to less health care costs, better parent satisfaction, and improved quality of life for the infants and families [1]. This analysis of a decade of data demonstrated that the program maintained these outcomes while minimizing practice variation. However, to ensure that these outcomes were sustained required a concerted team effort from the feeding care providers in continuing the SIMPLE feeding program. In order for any feeding program to be successful requires that multiple disciplines (physicians, parents, nurses, patient care assistants, dieticians, therapists, etc.) have common goals for a one-team approach to feeding [29]. Ensuring adherence to the SIMPLE feeding strategy required continuous education to maintain staff competencies across the entire 10-year period. This was accomplished by understanding the physiology of aerodigestive behaviors, using objective procedures to understand the pathophysiology of feeding difficulties, providing personalized and precise evidence based solutions, utilizing a pragmatic approach, and including the parents as care providers [40]. To our knowledge, this is the first study that has examined the sustained, long-term impact of a novel, milestone-targeted, standardized feeding program in the minimization of practice variation leading to lower LOHS.

Limitations

The study included limitations in both the population included and generalizability. The data was collected from an all-referral Children’s Hospital with tight inclusion and exclusion criteria aimed at ensuring uniformity of care within the NICU for this select population. The exclusion criteria were aimed at those infants who had diagnosis-related feeding issues that required more personalized feeding plans, which would not be compatible with the established goals for feeding milestones for preterm infants without these diagnoses. Over the ten years of the program, the proportion of infants admitted to our facility who met the inclusion criteria declined due to higher acuity with more infants meeting exclusion criteria. Furthermore, the number of reverse transports of infants increased over the time of the study due largely to bed-space issues. However, our findings are likely generalizable to other mostly referral-based Level IV NICUs in free-standing Children’s Hospitals.

Another limitation within this study was the lack of comparable data for outcomes from our own program, since all qualifying babies were included, or from other programs due to our unique program. Even within the strict inclusion/exclusion criteria, the patients from an all-referral children’s hospital are more complex with more co-morbidities than the general population from most birthing hospitals. While the gastrostomy rates seem relatively high (overall 7.8% in the first year of life and 12.4% with BPD), they are lower than that reported from the Children’s Hospitals Neonatal Database (CHND) which included 24 academic level IIIC NICU’s (including our NICU) across 20 states who had ≥25 beds and >400 admissions annually with >50% of admissions born outside of the hospital. Grover et al. using CHND data found an 18% and 23% rate of gastrostomy in infants with BPD born at 28–32 weeks GA and <28 weeks GA, respectively, during a 16-month period between 2010 and 2011 [41]. Murthy, et al. in an overview of NICU patients of all gestational ages using CHND data found that some amount of tube feeding was prescribed at discharge in 11.4% of patients, whereas we found that the overall rate of any tube feedings at discharge was 6.9% in the SIMPLE program [42].

Conclusions

Undoubtedly, the earlier data showed that the SIMPLE feeding program improved outcomes with standardization [1], while the current data demonstrates the sustained ability to maintain and improve patient outcomes over a decade, that even included the pandemic. The SIMPLE program maintained these improvements and the process has now become routine on our unit. As the treatment of infants at younger GA becomes increasingly prevalent, research is needed to identify evidence-based feeding therapies that will improve aerodigestive maturation and clinical outcomes while minimizing morbidities in the periviable population who are at highest risk for adverse feeding outcomes. Since GA, BPD, and feeding milestones have a major impact on LOHS, early interventions during the NICU stay to enhance acquisition of airway-digestive skills while implementing safe feeding programs remain critical. In order to address the limitations, further research and quality improvement investments are needed utilizing (a) prospective studies, (b) comparative studies among institutions to examine feeding milestones, and (c) studies targeting the more complex and rare disease populations with multiple comorbidities. These types of studies have the potential to dramatically improve patient outcomes and reduce long-term healthcare costs while improving the quality of life for patients, parents, and providers.