Trends in morbidities of late preterm infants in the neonatal intensive care unit

To characterize demographics and trends in length of stay (LOS), morbidities, and mortality in late preterm infants. Cohort study of infants born between 34 0/7 and 36 6/7 weeks gestation between 1999 and 2018 without major congenital anomalies at Pediatrix Medical Group neonatal intensive care units (NICUs). 307,967 infants from 410 NICUs met inclusion criteria. The median (25th–75th percentile) LOS was 11 (8–16) days in the entire period. Postmenstrual age (PMA) at discharge increased during the cohort for all gestational ages (p < 0.001). There was a decrease in invasive ventilation, receipt of phototherapy, and reflux medications observed (p < 0.001). In this large cohort, given 20 years of time for medical advancement, there was no significant improvement in the LOS of late preterm infants. All infants had an increased PMA at discharge, despite multiple practice changes that were observed.


INTRODUCTION
The late preterm infant, born between 34 0/7 and 36 6/7 weeks, is an important focus of research and advancement in neonatology [1].While late preterm infants experience less mortality than those born prior to 34 weeks, they are still at risk of significant morbidities [2,3].This can be distressing for parents as they are often anxious to take home what they perceive is just a small term infant [4].
The prevalence of late preterm deliveries has continued to rise over the past few years, increasing the number of admissions in this already large neonatal intensive care unit (NICU) population [5].In 2019, late preterm births accounted for 7.5% of all deliveries in the United States, up from 6.8% in 2014 [5,6].The increase in late preterm births is thought to be due in part to increasing maternal age and use of assisted reproductive technology [7].The rising volume of preterm infants places strain on hospital systems, increases the number of hospital days, and results in increased healthcare costs.The cost associated with preterm birth can be as much as 5 times higher than infants born at term [8].Common problems seen in late preterm infants include: feeding difficulties, respiratory distress, hypoglycemia, hyperbilirubinemia, and temperature instability [4,8,9].In one prior retrospective cohort study including 235 late preterm infants, these disorders were severe enough to require intervention, therefore prolonging stay, in roughly 30% of infants [4].
The purpose of this study was to evaluate whether length of hospital stay and morbidities decreased for late preterm infants over a 20-year time period.

Study design
We performed a cohort study of infants discharged between 1999-2018 at Pediatrix Medical Group NICUs.The Pediatrix database includes infants discharged from over 400 NICUs in the United States and Puerto Rico.Information from daily notes is extracted, de-identified, and compiled into the Pediatrix BabySteps Clinical Data Warehouse [10].The available extracted information includes maternal and infant demographics, laboratory results, including blood, urine, and spinal fluid culture results, medications, diagnoses, and other clinical data.
We excluded infants with major congenital anomalies, outborn infants, or those transferred to a higher level of care.We divided infants into 2 eras (1999-2008 and 2009-2018) based on year of hospital discharge and evaluated trends in length of stay (LOS).Primary outcomes of interest included number of admissions over time, LOS, post-menstrual age (PMA) at discharge, the presence of diagnoses including hypoglycemia, hyperbilirubinemia, sepsis, receipt of respiratory support, and death before discharge.

Definitions
We defined late preterm infants as those born between 34 0/7 weeks and 36 6/7 weeks.We defined large for gestational age as an infant with a weight greater than or equal to the 90 th percentile for gestational age and small for gestational age as an infant with a weight less than or equal to the 10 th percentile for gestational age using the Olsen curves [11].We defined hypoglycemia as any blood glucose less than 45 mg/dL.We defined respiratory distress syndrome (RDS) by either documentation in clinician notes or the receipt of any surfactant.We defined thrombocytopenia as platelet count less than 50,000/microL.We defined early onset sepsis as a positive blood culture within the first three days of age, excluding skin contaminants [including non-speciated streptococci, Bacillus sp., Micrococcus sp., gram-positive rods (not including Listeria sp.), Lactobacillus sp., Stomatococcus sp., Propionibacterium sp., Corynebacterium sp., Bacteroides sp., and diptheroids sp.] [12,13].We defined late onset sepsis as a positive blood culture after 3 days of age, excluding the same skin contaminants, and meningitis as any positive cerebrospinal fluid culture, also excluding common skin contaminants.We defined urinary tract infection as any positive urine culture.We defined pneumonia by clinical documentation.We defined the use of reflux medications as any receipt of proton pump inhibitors or H2 blockers.We defined respiratory support as the receipt of nasal cannula, oxyhood, continuous positive airway pressure (CPAP), noninvasive positive-pressure ventilation (NIPPV), conventional mechanical ventilation, or high frequency ventilation (jet or oscillator).We defined necrotizing enterocolitis (NEC) as the diagnosis of either surgical or medical NEC, as documented in clinician notes.We considered all of the above complications or interventions to be morbidities.

Statistical methods
We used the chi-square test to compare demographics and morbidities between the two eras.The chi-square test was also used to compare morbidities between the gestational ages (GAs), 34 weeks, 35 weeks, and 36 weeks.Descriptive statistics such as frequencies and percentages were calculated for categorical variables, and median with interquartile ranges for continuous variables.Wilcoxon rank sum was used to determine significance for continuous variables such as LOS and PMA at discharge.Trend analyses were performed using the Jonckheere-Terpstra test for trend and the Cochran-Armitage test for trend.P-values < 0.05 were considered to be significant.All analyses were conducted using Stata 17.0 (College Station, TX).RESULTS 307,967 infants from 410 NICUs met the inclusion criteria, including 110,993 infants in the first era (1999-2008) and 196,974 in the second (2009-2018), 29% of all infants studied.While there was an overall increase in the absolute number of late preterm infants over the study period, it was driven by an overall increase in the number of NICUs and infants for whom Pediatrix provided care (Fig. 1).This increase in 34-, 35-, and 36-week infants was statistically significant (p < 0.001, Jonckheere-Terpstra test for trend).During the study period the participating NICUs changed year to year, with 102 sites in 1999 and 279 sites in 2018.The relative percent of NICU patients that were late preterm remained constant between 26-31% of all NICU admissions.Among late preterm infants, the median (25 th -75 th percentile) for GA was 35 (34)(35)(36) weeks.The majority of infants included were born at 34 weeks (Fig. 1).The median birth weight was 2355 (2070-2671) grams and the majority of infants were male, 55%.There was an increase in the percentage of Black infants in the second era, from 15 to 19% (p < 0.001, Table 1).
The median (25 th -75 th percentile) LOS for the entire study period was 11 days (8-16), with an increase from 10 days to 11 days in the later era (Table 1, p < 0.001).A statistically significant increase in LOS was observed for both the 34-week and 35-week infants (Fig. 2, p < 0.001, Jonckheere-Terpstra test for trend).PMA at discharge increased over time for all GAs (p < 0.001, Table 1, Fig. 2).The median PMA at discharge was 36-37 weeks for the entire study period (Table 1).
Mortality was low throughout the study period, 0.2% in the first era and 0.14% in the second (Table 1).The most common morbidities experienced by infants were receipt of phototherapy, receipt of respiratory support, and RDS, and all were more likely in the 34-week infants (Tables 2, 3).The incidence of hypoglycemia was consistent across GAs at 7% (p = 0.002) (Table 3).Serious infections such as sepsis were uncommon for all GAs, at 0.1-0.4% (Table 3).
Between the two eras, receipt of antenatal steroids increased from 23 to 34% (Table 1, p < 0.001).Additional analysis demonstrated a statistically significant trend of increasing steroid use during the entire study period (p < 0.001, Cochran-Armitage test for trend).When divided by year, 2017 and 2018 demonstrated the highest use at up to 54 and 57% respectively (p < 0.001).The percent of infants receiving surfactant was consistent among all gestational ages at about 10% (Table 3).Surfactant use decreased between the two eras, going from 12 to 9% (p < 0.001, Table 2).While overall receipt of respiratory support remained consistent between the two eras, there was a significant increase in less invasive ventilation in era 2 (2009-2018).The receipt of HFNC increased from 7 to 14% and NIPPV increased tenfold (0.2 to 2%, p < 0.001, Table 2).Fewer infants received either conventional or high frequency ventilation in the more recent era (Table 2).
The use of anti-reflux medications decreased significantly from 3 to 1% between the eras (p < 0.001) (Table 2).The duration of reflux medication use was 3 days (1, 6).There was no significant difference in use across GAs (Table 3, p = 0.106).Caffeine use stayed stable at 4% across eras (Table 2, p < 0.001).The median duration of caffeine for the cohort was 3 days (1,8).Phototherapy use decreased from 46 to 39% between the two eras (Table 2).

DISCUSSION
In this large cohort of infants across the United States, we found that the length of hospital admission for late preterm infants has increased over the past twenty years.The median LOS increased by one day for the full cohort, with a significant increase in LOS observed in the 34-week and 35-week infants.The most common morbidities were the receipt of respiratory support, receipt of phototherapy, RDS, and hypoglycemia, similar to previous findings [4,14,15].As would be expected, hyperbilirubinemia and RDS were more prevalent in the 34-week infants [9].We also noted important practice changes over the 20-year period, including a decrease in invasive ventilation, increase in antenatal steroid use, decrease in phototherapy use, and decrease in reflux medication use.
Multiple studies have evaluated hospitalizations in this group but with a large variety in definitions surrounding LOS and inclusion criteria, making comparison somewhat difficult [4,[14][15][16][17][18][19].At some centers, 35-36-week infants may be admitted to a full-term nursery or intermediate level nursery and are included in the analysis as well.Previous work in a Toronto area study reported a LOS for infants admitted to the NICU at 8 days, which was slightly lower than our finding of 11 days, with 45% of their late preterm population requiring NICU care [14].Another study that included intermediate nursery patients reported an overall LOS of about 6 days [4].A similarly sized study from the Vermont Oxford Network only commented on the frequency of short stays (under 3 days), which was 10% in the late preterm group [18].Over a shorter time interval, 2005 to 2018, another Vermont Oxford Network study found an increased LOS in very preterm infants [20].They were unable to explain precise reason for the increased hospital stay [20].Our observed LOS is likely in part due to our inclusion criteria, as we included infants requiring a NICU stay who may be sicker at baseline than infants with less comorbidities who qualify for a lower level of care.When previously measured in national and international single center retrospective trials, the percent of late preterm infants that required the NICU for at least part of their stay had a large amount of variation, from 20 to 55% [4,14,21,22].Information about infants admitted to other units would be helpful in accurately determining LOS for the age group as a whole, as they may discharge with their mothers at 2-3 days [15].
The trend towards increasing LOS in our study was unexpected, given the improvements and technological advancements in the NICU over the time period.Our finding may be in part due to hesitation to discharge infants too early due to the risk of readmission, which is historically higher in the late preterm cohort overall [3].Prior studies have shown that the most common causes for readmission in the late preterm age group are hyperbilirubinemia and feeding difficulties [23].Because of this, late preterm infants are not discharged until they are consistently demonstrating weight gain with feeds, which is a different bar than term infants.This hesitation to discharge may be somewhat unwarranted, with recent studies demonstrating no correlation between longer stays and reduced readmission [23,24].Another  contributing factor could be improvements in monitoring technology, where small changes in heart or oxygen saturations are more easily captured, leading to additional periods of observation [25].We found that all GAs of infants in the cohort were discharged at about the same postmenstrual age, around 36-37 weeks.This is consistent with a prior study from 2021 in which a similar population was evaluated, and aligns with the expected GA for feeding maturity [26][27][28].Though there does not appear to be a significant shift in the population, there was an increase in the number of centers participating throughout the study.Centers that were added in the later period may have had differing protocols for late preterm infants, causing them to stay longer.
Several important changes in practice occurred over the past 20 years for this cohort of infants.One was a statistically significant increase in the number of infants that received antenatal steroids (from 23% in era 1 to 34% in era 2, p < 0.001).The increase in steroid use in the more recent era may be reflective of trials such as the Antenatal Late Preterm Steroids (ALPS) trial, which were occurring during the study period and demonstrated improvement in respiratory outcomes in the late preterm group with steroid administration [29].The American College of Obstetricians and Gynecologists also reflected this data in 2017 with their committee opinion that 34 0/7 -36 6/7 week mothers at risk of preterm birth in the next 7 days should receive steroids [30].Our study found an increase in steroid use after this recommendation, in both 2017 and 2018.This increase in steroid use over the entire study period could also have contributed to the observed increase in non-invasive ventilation, and a subsequent decrease in invasive ventilation in our study.These trends may also represent practice changes towards minimizing invasive ventilation [31].Similarly, the decrease in surfactant use in our study may be due to improved respiratory outcomes with increasing steroids, or from the additional barrier to administration in non-intubated infants.
Our study also demonstrated a decrease in acid reflux medication use, particularly in H2 blockers.There has been emerging evidence of the lack of efficacy of acid suppression treatment, and the risks associated with acid suppression, which may have influenced this change [32][33][34].We also noted a decrease in the use of phototherapy, likely influenced by consensus statements on safe bilirubin levels in this cohort, and the introduction of tools such as the Stanford Premie BiliRecs [35,36].
A significant strength of this study was the large sample size, which was able to be measured consistently over 20 years from a variety of centers, making a good representative population.However, there were multiple limitations to this study.Due to the method of data collection, we were unable to study issues related to feeding.Invasive devices like nasogastric and orogastric tubes were not recorded for the cohort, and percent of oral feeds were not able to be collected.We were unable to determine if feeding volumes were limiting hospital discharge or if an infant was discharged home on tube feeds.We were also unable to identify which infants had temperature instability or were on apnea or bradycardia watches.As previously mentioned, this cohort does not capture the infants that were admitted to other units in the hospital outside of the NICU, such as the mother-baby unit.This limits the ability to extrapolate data to the age group, since those infants are likely healthier and discharge earlier.The Pediatrix Clinical Data Warehouse does not have universal protocols for participating NICUs.The dataset does not capture institutional variation in protocols, including differences in admission criteria, discharge criteria, or goal discharge weight, which may have contributed to differences in LOS.We were unable to determine if protocols changed during the study as well.With the increase in number of NICUs during the study, there may have been a shift in LOS due to an increasing volume of NICUs with differing protocols.Similarly, NICU level is not captured a site.This study also does capture information regarding hospital readmissions, a common concern with early discharge.In our study there was a slight increase in infants identified as Black, but given the limitations of the dataset we were not able to determine the racial makeup of the geographic areas surrounding the NICUs.More information may help clarify if there were changes in access to care during the study.
To determine the cause of increased LOS, further data would need to be collected and analyzed [37].Individual hospital guidelines for discharge and, if possible, listed reasons by physicians for an individual infant's delayed discharge would be helpful for analysis.Explicit reasons for delay could help outline if there are other factors such as social concerns or required monitoring periods that could impact discharge at specific sites.In addition, data should be collected surrounding safety after discharge, including re-admission and infant death rates.This information could help guide future interventions that could safely shorten LOS for late preterm infants.
In conclusion, in this study of late preterm infants over a 20-year period, there was no decrease in the LOS.We were able to show when infants discharged in this time period leave the NICU and the major problems that they continue to face.Further research is required to determine why infants are staying longer, and if there are contributing factors not captured in this data set such as feeding or monitoring.There continues to be room for improvement to safely discharge these infants earlier and to decrease the burden of their stay on the medical system and their families.

Fig. 1
Fig. 1 Discharges Per Year.Number of infants per year by gestational age.

Fig. 2
Fig. 2 Length of Hospitalization by Year.Length of hospitalization per year by gestational age.
a p-value calculated by chi-squared test.