Abstract
BACKGROUND:
Late preterm infants are at high risk for medical complications and represent a growing NICU population. While 34-weeks’ gestation infants are generally admitted to the NICU and 36-weeks’gestation infants stay in mother-baby, there is wide practice variation for 35-weeks’gestation infants. The objective of this study was to compare short-term outcomes of 35-weeks’ gestation infants born at two hospitals within the same health system (DUHS), where one (DRH) admits all 35-weeks’ gestation infants to their level II NICU and the other (DUH) admits all 35-weeks’ gestation infants to mother-baby, unless clinical concern.
METHODS:
We conducted a retrospective cohort analysis of 35-weeks’ gestation infants born at DUHS from 2014–2019. Infant specific data were collected for birth, demographics, medications, medical therapies, LOS, ED visits and readmissions. 35-weeks’ gestation infants at each hospital (DRH vs DUH) that met inclusion criteria were compared, regardless of unit(s) of care.
RESULTS:
726 infants of 35-weeks’ gestation were identified, 591 met our inclusion criteria (DUH –462, DRH –129). Infants discharged from DRH were more likely to receive medical therapies (caffeine, antibiotics, blood culture, phototherapy, NGT), had a 4 day longer LOS, but were more likely to feed exclusively MBM at discharge. There were no differences in ED visits; however, more infants from DUH were readmitted within 30 days of discharge.
CONCLUSIONS:
Our findings suggest admitting 35-weeks’ gestation infants directly to the NICU increases medical interventions and LOS, but might reduce hospital readmissions.
Abbreviations
Late preterm
Neonatal intensive care unit
Gestational age
Duke University Health System
Mother-baby unit
Duke University Hospital
Duke Regional Hospital
Length of stay
Emergency Department
Birth weight
Antenatal steroids
Nasogastric tube
Total serum bilirubin
Transcutaneous bilirubin
Total bilirubin
Maternal breast milk
Donor breast milk
Intravenous
Cesarean section
Direct antiglobulin test
Introduction
Although significantly more mature and larger than their extremely preterm counterparts, late preterm (34 0/7 –36 6/7 weeks, LP) infants are still physiologically immature and have limited compensatory responses to the extra-uterine environment, and therefore, are at disproportionately higher risk than term infants for medical complications. These complications include respiratory distress, apnea of prematurity, hyperbilirubinemia, hypothermia, hypoglycemia, feeding difficulties, as well as hospital readmission, neurodevelopmental impairment, and death [1–9]. In the United States, the increase in LP births is a major contributor to the rise in the preterm birth rate and represents an increasing population in the neonatal intensive care unit (NICU) [10–13]. In fact, two recent studies by the Vermont Oxford Network and in California found infants ≥34 weeks’ gestation accounted for ≥74% of all NICU admissions, with ≤15% classified as “high acuity” admissions [14, 15].
Despite this large number, the optimal initial location for the care of LP infants is unclear. As mentioned above, many LP infants require specialized care and are at higher risk for morbidity, mortality and hospital readmission. However, admission to the NICU often requires separation of the mother-infant dyad, may interfere with breast feeding, increases the risk of infection, increases costs, increases anxiety to the family and may not be necessary for most LP infants, especially at later gestational ages (GA) (35–36 weeks) [16–18]. Most centers generally admit infants < 35-weeks’ gestation to the NICU, while ≥36 weeks’ gestation infants typically ‘room-in’ with their parents after birth in mother-baby/post-partum units, unless there is clinical concern. However, for 35-weeks’ gestation infants, there is no consensus and practice varies widely [15, 20].
The Duke University Health System (DUHS) is in a unique position to compare the short-term outcomes of 35-weeks’ gestation infants cared for in a mother-baby unit (MB) versus the NICU. Within the DUHS, there are two birthing centers six miles apart: Duke University Hospital (DUH), a large tertiary referral center, and Duke Regional Hospital (DRH), a community teaching hospital. The newborn units are staffed by pediatricians, neonatologists and nurse practitioners from the same group and most work at both locations. At DRH, 35-weeks’ gestation infants are automatically admitted to their level II NICU and cared for by neonatologists, regardless of clinical stability. This policy was instituted in late 2009 at DRH because of the high rate of readmissions in this population. However, at DUH 35-weeks’ gestation infants are admitted to the MB unit and cared for by general pediatricians, unless there is clinical concern. Comparing outcomes of these similar groups of infants can provide valuable information to help determine the optimal initial location for care of the increasing population of 35-weeks’ gestation infants. The primary objective of our study was to compare short-term outcomes of 35-weeks’ gestation infants between infant populations initially cared for in the NICU (DRH) versus MB (DUH), including length of stay (LOS) and post-discharge morbidity (emergency department (ED) visits and hospital readmission). The secondary objective was to compare in-hospital outcomes of these infants including: respiratory support, medication use, phototherapy and feeding problems.
Methods
Study population
We conducted a retrospective cohort analysis of 35-weeks’ gestation infants born at DUH and DRH (Durham, NC) from 2014 to 2019. We used Duke Enterprise Data Unified Content Explorer, a web-based query previously described to extract data [21]. At both hospitals, all 35-weeks’ gestation infants are observed for a minimum of 48 hours prior to discharge, and an appointment is scheduled with a medical provider within 1–3 days of discharge. At DRH, all infants < 36 weeks GA or < 2 kg birth weight (BW) are admitted to their level II NICU per unit protocol; while at DUH, 35-weeks’ gestation infants ≥2 kg BW are admitted to MB unless clinical concern. We included all infants 35 0/7 to 35 6/7 weeks’ gestation with BW ≥2 kg and LOS ≤40 days that survived to discharge home. Infants < 2 kg were excluded because these infants are automatically admitted to a higher level of care at both locations. Infants transferred to another hospital prior to discharge were excluded because of incomplete data and care at multiple hospitals confounding outcome interpretation. A LOS of 40 days was chosen as the cutoff to include the majority of infants in our 35-weeks’ gestation population (95%, mean + 2 standard deviations), but to exclude those with abnormally long LOS [19, 22–24]. We collected the following data: BW, GA at birth, sex, race, ethnicity, mode of delivery, discharge unit, insurance type, exposure to antenatal steroids (ANS) and magnesium sulfate, respiratory support, phototherapy, medications, nasogastric tube (NGT) placement, type of feeds, LOS and laboratory data. In addition, we collected data on ED visits and hospital readmissions in the DUHS within 30 days of discharge. The Duke University Institutional Review Board declared this study exempt from full review. STROBE guidelines for cohort studies were followed.
Variables
Hospital location
The infants’ hospital location (hospital and unit) was assigned based on the location at the time of discharge. For DUH, this included the level IV NICU, MB and general pediatrics units. Infants discharged from the cardiac unit were excluded because these were infants with complex congenital anomalies that required prolonged care. Infants are transferred to general pediatrics from MB or the NICU when a longer LOS is anticipated (antibiotic course, phototherapy, neonatal abstinence syndrome), when a higher level of care is no longer needed and/or for census or other administrative reasons. For DRH, this included the level II NICU and MB unit. Some infants were admitted to one unit and transferred to another unit within the same hospital prior to discharge (higher or lower level of care). Infants transferred between hospitals were excluded. For analyses, all 35-weeks’ gestation infants at DUH were compared to all 35-weeks’ gestation infants at DRH, regardless of unit(s) of care within that hospital.
Race and ethnicity
Race was self-reported as White/Caucasian, African American/Black, Asian, or other, which included those that reported 2 or more races, declined or answered ‘other’. Ethnicity was self-reported as not Hispanic/Latino, Hispanic/Latino (Hispanic Mexican, Hispanic other, Hispanic or Latino) or other (unavailable, declined). These were combined and reported as ‘White, non-Hispanic’, ‘Black, non-Hispanic’, ‘Hispanic’, ‘Asian, non-Hispanic’ and ‘other’. Other included 2 or more races, declined or answered ‘other’ for either race or ethnicity.
Health insurance
Private health insurance included managed care and commercial insurance. Public health insurance included Medicaid, Medicaid pending and other government insurance.
Respiratory support
The highest level of support used was reported. In increasing level of support these included: oxyhood, nasal cannula, continuous positive airway pressure and mechanical ventilation. Brief use of blow by oxygen or bag-valve mask at delivery was not included.
Bilirubin
For maximum total bilirubin (TB), the highest total serum bilirubin (TsB) level from the birth hospitalization was used. For infants who did not have a TsB recorded, the highest value for transcutaneous bilirubin (TcB) was used. All units within DUHS have the same hyperbilirubinemia protocol –all infants have a bilirubin level checked at least once prior to discharge, and phototherapy is initiated and discontinued at the providers’ discretion based on guidelines from the American Academy of Pediatrics [25]. In 2017, DUHS implemented the routine use of TcB, with a TsB collected only if the TcB was above a certain threshold (TcB ≥7 or ≥10 at 24 and 48 hours of life, respectively).
Medications
Maternal medications (magnesium and ANS) for the year prior to delivery and infant medications for the birth hospitalization were reported. Magnesium exposure was defined as any maternal intravenous (IV) magnesium sulfate. ANS was defined as injection of betamethasone. Surfactant included intratracheal administration of Poractant alfa, Calfactant or Beractant. Antibiotics included any oral or IV antibiotic.
Feeding
Feeding type was defined as the type of milk the infant received on the day of discharge, including maternal breast milk (MBM), donor breast milk (DBM) and/or formula. DBM was available to infants in the NICUs at either location during the entire study period, and available to all infants starting in 2018. Lactation consults are automatically placed at both locations for all “at risk” infants, which includes late preterm infants. Therefore, all infants in this study would have had a lactation consult placed.
ED visit and readmission
ED visit was defined as any visit to a DUHS ED within 30 days of discharge. Hospital readmission included admissions from the ED as well as urgent and emergent direct admissions to DUHS within 30 days of discharge. Admissions could be for observation (anticipated stay < 24 hours) or inpatient status. Non-ED visits, such as primary care visits (well or sick) and routine (planned) admissions (usually for procedures), were not included. Infants could have more than one ED visit and/or readmission. The date of the encounter, reason for visit, LOS (if applicable), medication use, laboratory data and phototherapy use were recorded.
Statistical analysis
Continuous variables were presented as medians and categorical variables as counts. We compared categorical variables using the Fisher’s exact test and continuous variables using Chi square. Because of the relatively small number of infants and events in our study, we were unable to conduct multivariate analyses for most outcomes (except breast milk); to perform logistic regression with only 3 factors would require an N of 638 [26]. Statistical analyses were performed using JMP Pro 15.0 (SAS Institute Inc., Cary, NC). All statistical tests were 2-sided and P values < 0.05 were considered statistically significant. There were no adjustments for multiple comparisons, except in multivariate analyses.
Results
Study population
We identified 726 infants of 35-weeks’ gestation born at DUH and DRH between 2014 and 2019 (Fig. 1). One hundred and thirty-five infants were excluded: 86 had BW < 2 kg, 23 had LOS > 40 days, 1 was discharged from a pediatric cardiac unit, 2 were discharged to an outside hospital, 4 did not survive to discharge and 19 were transferred between DUH and DRH (4 to DUH and 15 to DRH). This left 591 infants in the final analysis, 462 discharged from DUH and 129 discharged from DRH. As expected, the majority of infants at DUH were discharged from the MB unit (338/462, 73.2% MB; 104/462, 22.5% NICU; 20/462, 4.3% general pediatrics), while the majority of infants at DRH were discharged from the NICU (113/129, 87.6% NICU). There were no differences in BW, sex, race/ethnicity or exposure to ANS between infants discharged from DUH and DRH (Table 1). Infants discharged from DUH were more likely to be delivered by cesarean section (C-section), exposed to magnesium and have public insurance or self-pay.
Flow diagram of infants included/excluded in the analysis.
Demographic and baseline data for 35-weeks’ gestation infants discharged from Duke Regional Hospital (DRH) and Duke University Hospital (DUH)
Comparisons were made using Fisher’s exact test or Chi Square where appropriate, P values < 0.05 were considered significant. Abbreviation: IQR –25% and 75% interquartile.
Infants discharged from DRH were more likely to have medical interventions including: NGT placement (57/129, 44.2% DRH vs 72/462, 15.6% DUH; p < 0.0001), phototherapy (76/129, 58.9% DRH vs 101/462, 21.9% DUH; p < 0.0001), caffeine (14/129, 10.9% DRH vs 3/462, < 1% DUH; p < 0.0001), blood culture (82/129, 63.6% DRH vs 196/462, 42.5% DUH; p < 0.0001) and antibiotics (70/129, 54.3% DRH vs 94/462, 20.3% DUH; p < 0.0001) (Table 2). There was no difference in the number of infants that required respiratory support or surfactant (Table 2). The maximum TB was higher at DRH than DUH, but there was no difference in the number of direct antiglobulin (DAT) positive infants (data not shown). One blood culture was positive but was determined to be a contaminant (Staphylococcus epidermidis) and the infant was not treated with antibiotics.
Outcomes and interventions for 35-weeks’ gestation infants discharged from Duke Regional Hospital (DRH) and Duke University Hospital (DUH)
Outcomes and interventions for 35-weeks’ gestation infants discharged from Duke Regional Hospital (DRH) and Duke University Hospital (DUH)
Comparisons were made using Fisher’s Exact test or Chi Square where appropriate. P values < 0.05 was considered significant. Abbreviations: NGT – nasogastric tube, BM – breast milk (maternal or donor), DBM – donor breast milk, MBM –maternal breast milk, TB – total bilirubin (serum or transcutaneous), mg/DL – milligrams per deciliter, IQR – 25% and 75% interquartile range, OH –oxyhood, NC – nasal cannula, CPAP –continuous positive airway pressure, LOS – length of stay, d –days, ED – emergency department. a10 infants included twice because of multiple ED visits/readmissions (5 had 1 of each, 3 had 2 readmissions, 3 had 2 ED visits, no readmissions).
Infants discharged from DRH were more likely to be fed exclusively MBM at discharge (68/129, 52.7% DRH vs 150/462, 32.5% DUH; p = 0.0002) (Table 2). In a multivariate logistic regression analysis, exclusive MBM at discharge was independently associated with insurance type, self-reported race/ethnicity and location (DRH/DUH). Public insurance, Black, non-Hispanic and DUH were associated with decreased exclusive MBM feeding at discharge.
Finally, the median LOS at DRH was 4 days longer than at DUH (7 days DRH vs 3 days DUH; p < 0.0001). These results did not differ significantly if the inclusion criteria were changed to exclude infants with a LOS > 14 days (instead of 40) or if the 19 infants transferred between DRH and DUH were included. In addition, significantly more infants discharged from DRH had a LOS > 14 days compared to those discharged from DUH (23/129, 17.8% DRH vs. 19/462, 4.0% DUH, p < 0.0001). The most common reasons for continued hospitalization were feeding difficulties and monitoring for events and were similar between DUH and DRH. LOS > 14 days was associated with exclusive maternal breast milk feeding at discharge (p = 0.008).
There was no significant difference in the number of ED visits that did not result in hospital readmission in the first 30 days after initial discharge between infants discharged from DRH and DUH (DRH 6/129, 4.7% vs DUH 31/462, 6.7%; p = 0.54) (Table 2). However, infants discharged from DUH were significantly more likely to be readmitted within 30 days post-discharge (DRH 2/129, 1.6% vs DUH 36/462, 7.8%; p = 0.008). Ten infants (all DUH) had multiple ED visits and/or readmissions and each event was included –5 infants had 1 of each, 3 infants had 2 readmissions, no ED visits and 2 infants had 2 ED visits, no readmissions. The results did not change significantly if only 1 event per infant was included or if the 19 infants transferred between DUH and DRH were included. Only one infant with a LOS > 14 days required readmission (DUH) and five had ED visits within 30 days of discharge (4 DUH, 1 DRH). The top 4 reasons for readmission were hyperbilirubinemia, respiratory concerns (included tachypnea, upper respiratory infection), failure to thrive/poor oral intake, and temperature instability (Table 3). There were no known deaths during this period for infants discharged from either hospital. With an absolute risk reduction of 6%, 16 infants (16.1) would need to be admitted automatically to the NICU to prevent one readmission within 30 days of discharge.
Emergency department visits and hospital readmission to the Duke University Health System within 30 days of discharge for 35-weeks’ gestation infants discharged from Duke Regional Hospital (DRH) and Duke University Hospital (DUH)
Emergency department visits and hospital readmission to the Duke University Health System within 30 days of discharge for 35-weeks’ gestation infants discharged from Duke Regional Hospital (DRH) and Duke University Hospital (DUH)
Abbreviations –ED –emergency department; FTT –failure to thrive; BRUE –brief resolved unexplained event; CSF –cerebrospinal fluid; PCR –polymerase chain reaction, IQR –25% and 75% interquartile range, d –days a10 infants included twice because of multiple ED visits/readmissions (5 had 1 of each, 3 had 2 readmissions, 3 had 2 ED visits, no readmissions). bone infant with Escherichia coli bacteremia and enterovirus meningitis 22 days after discharge.
In this retrospective study of over 700 infants, we found that newborns discharged from DRH, where 35-weeks’ gestation infants are automatically admitted to their level II NICU, were more likely to receive medical interventions (phototherapy, caffeine, antibiotics, blood culture, NGT placement) and had an increased LOS by 4 days compared to infants discharged from DUH, where infants are automatically admitted to the MB unit, unless clinical concern. However, infants discharged from DUH were more likely to be readmitted to the hospital and less likely to receive exclusive MBM at discharge. We calculated a “number needed to admit” of 16 to prevent one readmission.
The readmission rate within 30 days of discharge (7.8%) was significantly higher for DUH than DRH, but was consistent with reported rates from other hospitals, which ranges from 1.8 to 15.8%, with the majority reporting between 6.5 to 9% [27, 28]. While the readmission rate after discharge for DRH was much lower than the reported range, the LOS was longer than the typically reported 3 to 6 days [19, 22–24]. Previous studies have also reported an association between LOS and readmission in LP infants. Escobar et al found an association between LOS < 96h and an increased risk of readmission for LP infants [28] and Kuzniewicz et al found a LOS > 3d was protective against readmission for LP infants [27]. Interestingly, another study also found that introduction of a level II NICU increased the LOS for 35-weeks’ gestation infants, but they did not look at readmission rates [24].
Although infants at DRH were less likely to be readmitted to the hospital, they were exposed to more medical interventions during the initial hospitalization, including NGT placement, caffeine, phototherapy, blood culture and antibiotics, with the rates at DRH (for all except blood culture) at least 2-fold higher than that at DUH (Table 2) and above typically reported ranges [19, 29]. There were no differences in respiratory support needed, surfactant use or positive blood cultures, suggesting infants at DRH were not sicker. Infants at DRH did have higher median TB level, but this likely reflects the longer LOS and detection of the peak bilirubin level (bilirubin peaks day 3–7) rather than a difference in severity of hyperbilirubinemia. The high rate of interventions at DRH suggest some therapies may have been overused. In a NICU, infants are often assumed to be “sick until proven well”, while the opposite is true in MB units. It is possible 35-weeks’ gestation infants admitted to the NICU were managed more like preterm infants that need interventions, rather than healthy term infants, lowering the threshold for therapies and increasing LOS. Interestingly, the number of infants receiving exclusive MBM at discharge was higher for DRH than DUH, despite maternal-infant separation. This is at least partially explained by the differences in patient populations (insurance, race/ethnicity) and the longer LOS at DRH, which may allow for increased time for mothers to establish breast milk production, in addition to providing additional opportunities for consultation and education from lactation services, but there could be other unmeasured factors contributing as well.
There are several limitations to this study. First, it is a retrospective study and relied on charting and billing data. Because of the limitations of these data, we compared 35-weeks’ gestation infants from DUH and DRH as a whole, rather than comparing healthy 35-weeks’ gestation infants at the two locations for several reasons. First, it would have been difficult to accurately determine “healthy” based on the data available. As an example, a well-appearing 35-weeks’ gestation infant might be admitted per protocol to the DRH NICU and subsequently diagnosed with hypoglycemia, respiratory distress or tachypnea, but these may or may not have reached a threshold that would have required transfer to a higher level of care at DUH. Second, infants could transfer between multiple units throughout a hospitalization (for medical or administrative reasons) which would complicate the analysis and categorization. There were also several transfers of infants between DUH and DRH, which could bias the LOS data. However, transfers were < 3% (21/726) of the cohort and equal between DUH and DRH (by percent).
Another limitation is that we were only able to capture ED visits, readmissions and deaths within the DUHS, so it is possible that all the relevant encounters were not captured. However, there are very few children’s hospitals in the Raleigh-Durham area, so for the vast majority of infants delivered at either hospital, infants seen in an outside ED would be transferred to DUHS ED if they required admission. Therefore, while, some ED visits may have been missed, it is unlikely that a significant number of hospital readmissions were missed. In addition, any missed readmissions should be spread equally between those discharged from DRH and DUH.
Finally, despite the close proximity, there were significant differences identified between the patient populations at DRH and DUH and there are likely other unmeasured differences. Infants in DUH were more likely to have public insurance, suggesting a lower resource population. Public insurance/no insurance has been associated with increased rates of readmission and ED utilization in infants and older children [31, 32]. Infants at DUH were also more likely be delivered by C-section and exposed to magnesium and, although DUH has roughly 2 times the number of births as DRH, it had almost 4 times as many 35-weeks’ gestation infants. These factors suggest a sicker maternal population - the majority of high-risk deliveries take place at DUH under the care of Maternal Fetal Medicine specialists. However, there were no differences in BW, ANS or respiratory support at the two hospitals, suggesting no major differences in the infant populations at birth. Infants at DUH had fewer interventions and a shorter LOS, also not reflective of a sicker infant population; however, we cannot exclude an impact of these factors on readmission. We did not have data available on other potential confounders, including lactation support, maternal education and social support. Additionally, because of the small number of events and power limitations, we were unable to perform a multivariate analysis to determine the impact (if any) of these differences on readmission.
Our findings suggest that the optimal initial location for managing 35-weeks’ gestation infants after birth remains unclear and highlight the difficulties in caring for LP infants. Larger studies are needed to address this question, and it will be important to track balancing measures, such as LOS, readmission rate, deaths, etc for any changes made to admission policies. DRH implemented a protocol of admitting all 35-weeks’ gestations infants to the NICU because of the high rate of readmissions. While this was effective and the rate of readmission for infants discharged from DRH is significantly lower than for that of DUH, the number needed to admit to prevent a readmission is 16 and the median LOS for readmission was only 2 days. A more targeted admission strategy that aims to address the most common reasons for readmission (hyperbilirubinemia, feeding difficulty) might be more effective. Until an effective strategy is identified, for hospitals that automatically admit 35-weeks’ gestation infants to the NICU, practices that might increase LOS such as caffeine for self-limited events or overtreatment of hyperbilirubinemia and hypoglycemia need to be examined carefully. On the other hand, for hospitals that choose to admit these infants to MB units, general pediatricians should remain vigilant for red flags: Poor feeding, especially for breast fed infants; hyperbilirubinemia, especially when DAT is positive; borderline blood glucoses and temperature instability. Delaying discharge should strongly be considered for infants with even subtle signs of these conditions. General pediatricians should also discuss post discharge complications associated with LP infants with the families. In addition, physicians should consider not discharging 35-weeks’ gestation infants before 48–72 hours of age because this does not allow enough time to properly evaluate infants for the most common causes of hospital readmission, including feeding difficulties and hyperbilirubinemia.
Funding
No funding was secured for this study.
Conflict of interest
The authors have no conflicts of interest to disclose.