In this study, we compared birth GA, morbidities, mortality, and LOS between singletonand multiple gestation births among neonatal discharges from 2000 through 2016 utilizing the KID, which is a large population-based administrative database compiled by the HCUP of AHRQ. This report is the largest one of its kind, representing > 4000 hospitals and > 22 million discharge records over the last two decades. Our analysis controlled for baseline characteristics and previously identified variables that could affect the outcome, such as IUGR or SGA status. We believe these results are widely generalizable and could be of benefit with perinatal counseling in the case of multiple gestation pregnancies.
In our study, the percentages of premature babies in each GA category from 24 to 34 weeks were significantly higher among triplet and higher-order births compared to twins, which, in turn, were significantly higher than the singleton. The mean GA-expressed as mean (SD)- for singleton was significantly higher at 38.30 (2.21) weeks of GA compared to twin (36.39(4.21)) and triplets or higher-order multiples (32.72(4.14)). Thismean GA in our study was similar to another large study from the United States, which reported mean GA as 39 weeks in singleton, 35.8 weeks in twin, and 32.5 weeks in triplet births . Similarly, in each BW category below 2500 g, the representation of twin, triplet, or higher-order multiple births was significantly higher than singleton (Table 1). IUGR or SGA babies were significantly higher among all multiple gestation births compared to singleton. This data underlines the previous reports suggesting that the incidence of preterm and LBW deliveries increases with multiple gestation births [17, 18, 29].
Our data shows that the rate of short-term respiratory morbidity like RDS was significantly lower in all multiple gestation births, whereas pulmonary hemorrhage was significantly higher in all multiple gestation births compared to singleton. We also found a significantly higher incidence of BPD among triplet or higher-order multiple births compared to singleton and a similar trend in twin births when compared to singleton, but it was not statistically significant. Wadhawan et al.  looked at the short-term and long-term outcomes of more than 13,000 extremely LBW babies (birth weight 401–1000 g) born out of multiple births in participating centers of the Neonatal Research Network between 1996 and 2005. They reported a significantly higher need for surfactant therapy in twins and triplet births compared to singleton and also a higher need for mechanical ventilation in triplet compared to singleton births. They also had reported a higher incidence of BPD among twin births compared to triplet. These authors did not adjust for BW, GA or other confounding characteristics while reporting their short-term clinical outcomes, which could account for the difference in their outcomes compared to our results. Garg et al.  have published the perinatal characteristics and neonatal outcome data of preterm singleton, twin, and triplet births at 22–31 weeks’ gestation from Australia during the years 1994–2005 after adjusting for birth weight percentile, GA and other population-based characteristics. In this study, twins were more likely to have hyaline membrane disease compared to singletons. Surprisingly they have reported a lower rate of BPD in twins compared to the other groups. Neonatal practices and outcomes vary across the world. The neonatal practices in Australia in 1990’s and early 2000’s may be different from the United States in the years 2000–2016, which could explain the difference in outcomes. In addition, adherence to better practices and bundles, including the use of antenatal steroids, exogenous surfactant administration, early nasal continuous positive airway pressure (NCPAP), and other lung-protective strategies during mechanical ventilation, have improved the respiratory morbidities of preterm neonates leading to a difference in the outcomes reported during this study period between 2000 and 2016 [31,32,33,34,35].
Intraventricular hemorrhage and periventricular Leukomalacia
In our study, the incidence of any IVH was significantly lower in twin compared to singleton births, whereas the incidence of severe IVH (grades 3 and 4) was significantly higher in twin births compared to singleton. The difference in the incidence of any IVH or severe IVH was not statistically significant in triplet or higher-order multiple births compared to singleton. Kaufman and colleagues  reported an increased incidence of mild IVH in triplet birthscompared to singleton. Garg et al.  have reported similar severe IVH rates between singleton, twin, and triplet births, whereas Yee and group  reported a lower risk of IVH in triplet infants with BW < or = 1250 g, compared to singletonsand twinpatientswith similar BW and GA. The difference in outcomes between these studies could be related to the variation in the populations studied or the changes in the management practices within different neonatal units across the world over time.
The incidence of PVL was significantly lower in triplet or higher-order multiple birthscompared to singleton, whereas there was no difference in PVL incidence when comparing twin gestation births to singleton. This was in contrast to Resch et al. , who reported a significantly higher incidence of PVL in twin and triplet births, which may be explained due to this study reporting rates of PVL per pregnancy as opposed to per discharge record in our study. In addition, their sample size was relatively small. The protective effects of antenatal steroids in preventing IVH is well documented. Higher antenatal steroid coverage may be one of the reasons for lower IVH and PVL in triplet or higher-order multiple births. Over the last several years, changes in the neonatal practices, including better ventilation strategies [33, 39, 40] and adherence to various IVH prevention bundles [41, 42], are shown to be associated with an improvement in the incidence of IVH and PVL among preterm patients.
Sepsis and necrotizing enterocolitis
In our study, the incidence of bacterial sepsis was significantly lower in twin gestation births compared to singleton. While other studies have reported similar rates of sepsis in multiple gestation births compared to singleton [43,44,45], the large sample size and adjustment for baseline characteristics may have allowed for this study to detect differences not previously noted.
The incidence of NEC was not significantly different between any of the three groups, although the severe NEC was marginally higher in the triplet or higher-order multiple groups. Similar to our study, many of the previous studies have reported comparable rates of NEC among twins and triplets [20, 37, 44, 45]. Similar to other care practices in neonatology, during this study period, changes in neonatal feeding practices and central line care bundles may have significantly modified the neonatal outcomes such as NEC and sepsis [34, 35, 46,47,48,49,50,51,52].
Retinopathy of prematurity
In our study, we found an increased rate of all stages of ROP in triplet or higher-order gestation births compared to singleton, although we did not find any significant difference in the incidence of severe ROP between the groups. Like our findings, Kaufman and colleagues  also have reported increased incidence of ROP in triplet births, but they also found an increased incidence of severe ROP in triplet compared to singleton births. Some other authors [45, 53] have also reported a significantly higher rate of advanced ROP (stages II-III) in singleton compared to multiple gestation births. Consistent with our findings, Garg et al.  have also reported a similar incidence of severe ROP among singleton, twin, and triplet births. There has been an overall decline in the incidence of severe ROP in recent years [54, 55], which may have contributed to the inability to detect differences in severe ROP in our study. The incidence of stage 3, 4, or 5 ROP was very low in our study (< 0.40% in all groups). Improvements in neonatal practices including better oxygen saturation targeting which minimize repeat episodes of alternating hypoxia and hyperoxiamay have led to a change in the incidence of severe ROP requiring intervention [34, 35, 54, 56, 57].
We found a higher incidence of mortality for twin and triplet births compared to singleton among term patients, even after adjusting for baseline and hospital characteristics. In the preterm group, our analysis showed that adjusted odds of mortality were higher for triplet and higher-order births compared to singleton. However, the adjusted odds of mortality were similar for twin and singleton births. Heino et al. analyzed the neonatal death from the Euro-Peristat project, which included 5 million births from 29 countries and reported a pooled relative risk of 7.0 (95% Cl 6.1–8.0) for neonatal mortality among multiple gestation births compared to singleton . Martin et al. reported that mortality was higher for multiple gestation births compared with singleton when analyzing the birth data in the US for 1980–97 . Shinwell et al. also have reported an increased risk of death among triplet births on their analysis on Israel’s national very low birth weight (VLBW) infant database .
Several other reports suggested no significant increased risk of mortality among multiple births. In a retrospective cohort study, which was matched for GA, sex and country of birth, Shah et al. analyzed the outcomes of a total of 6079 triplets and 18,232 singletons of 24 to 32 weeks’ gestation or 500 to 1499 g . This study reported no significant difference in the primary outcome between triplets and singletons. Similarly, Garite et al.  reported no difference in mortality between the preterm singleton, twin or triplet births at 23 to 35 weeks of gestation.
In contrast to these reports, Russell et al., using National Center for Health Statistics data of the United States from 1980 to 1999, showed that VLBW and moderately LBW multiple gestation infants had a lower mortality rate than for singletons in similar BW categories . Jacquemyn et al. also have reported lower perinatal mortality in twin births compared to singleton . The use of different statistical methods among studies may account for variations in results.
Length of stay
We found that the median LOS was significantly longer in multiple gestation births compared to the singleton overall and also within each GA category. Only a very few studies with a relatively smaller sample size have compared the LOS between singleton and multiple gestation births, including twin, triplet and higher-order multiples. Vachharajani et al. , reported that twin and triplet births had a longer LOS compared to singleton patients born at 34 weeks, but the LOS for twin and triplet births was comparable to that of singleton at 35 and 36 weeks. A few studies did not find any difference between singleton and multiple gestation births regarding the LOS. Yee and colleagues have reported no difference in the mean LOS for twin and triplet births compared with singleton . They had > 1700 babies in their analysis and included patients < 1250 g. Qiu et al.  and Maayan-Metzger et al.  also found no difference in the duration of NICU stay between the singleton and multiple gestation births.
When we analyze the data from a large population-based data set, such as KID, demographic variables like the distribution of sex, race, primary insurance or payer, and the type of hospital may be significantly different among the groups. In addition, there are wide variations in neonatal practices, including prenatal care, resuscitation guidelines, ventilation strategies, and nutritional management among the newborn nurseries across the country. These differences in socio-demographic parameters and patient care variations may have played a significant role in variable outcomes..
This study has limitations due to the retrospective use of an administrative database. The findings heavily rely on the accurate reporting of the ICD codes and other variables. There are additional limitations in the case of transfers since discharge records are not linked in KID. We accounted for this limitation by excluding discharge records with indicators for “transfer in” or “transfer out.” In doing so, it is possible that the LOS reported was inadvertently shorter since the sicker patients would be more likely to undergo transfer, and the LOS from the multiple hospitalizations were not combined in the database. It is also worth noting that the gestational age ICD codes used for analysis in this dataset cover a period of 2 weeks prompting GAbreakdown in two-week categories. So the data has to be interpreted accordingly and not used for individual gestational ages. In addition, KID does not include neurodevelopmental outcomes; hence they were not analyzed. Furthermore, evidence exists to reflect the evolution of neonatal care over the years, including the years analyzed in this study between 2000 and 2016. This study could not account for advancements in clinical practice which could change the outcome of these patients over time.