Neonatal intensive care admission for term neonates and subsequent childhood mortality: a retrospective linkage study

Neonatal intensive care unit (NICU) admission entails immediate risk for morbidity and mortality for the newborns and emotional burden for their families, at a high cost to the health care system. The majority of studies have focused primarily on admission and risk factors for preterm NICU admission and long-term follow-up. In a time trend analysis of 38 units in the USA between 2007 and 2012, the overall NICU admission rate increased by 23% after adjusting for maternal and neonatal characteristics, while the admitted neonates were increasingly likely to be at term [1]. Mortality and long-term morbidity are major concerns for infants at NICU, while death among term neonates during the NICU hospitalization is considered a common occurrence, and its reported incidence is 8–15% [2, 3].

To our knowledge, no previous studies have examined long-term childhood mortality for this group of neonates. Thus, the aim of this study was to study the association between the NICU admission of term neonates and the risk of long-term childhood mortality.

During the study period, 206,509 births were recorded at the Shaare Zedek Medical Center. After the application of study exclusion criteria, we included 192,527 (93.23%) term singleton live births in the study (91,697 women). During a median follow-up of 7.2 years, 311 childhood deaths were recorded; 47.9% (n = 149) of those who died had records of a NICU admission as compared with 2.7% (n = 5143) of the survivors (P < 0.001) (Fig. 1). Detailed neonatal mortality for the newborns admitted/not admitted to NICU by follow-up time period until death can be found at Additional file 1: Table S2.

An analysis for predicting mortality among term neonates in relation to NICU admission was performed. The median [IQR] follow-up time was 7.2 [3.9–10.7] years. A detailed comparison between the characteristics of the decedents and the survivors is presented in Table 1.

A multivariable Cox proportional hazards model for the mortality risk, adjusted to delivery and neonatal characteristics in this cohort of neonates revealed that NICU admission was the most prominent risk factor associated with mortality HR = 19.72, 95% CI [14.66, 26.53] P < 0.001. After excluding 6486 births (3.4%) with records of congenital anomaly (chromosomal or genetic syndrome), the model retained a similar significant magnitude of association between NICU admission and mortality HR = 19.79 95% CI [11.71–33.45]; P < 0.001. Other significant risk factors were 5′ Apgar ≤ 7 (HR = 4.40, 95% CI [2.08, 9.30] P < 0.001), Arab population group (HR = 2.85, 95% CI [1.74, 4.66] P < 0.001) and SGA (HR = 2.20, 95% CI [1.30, 3.80], P < 0.001) (Table 2).

We further performed an analysis to compare NICU and no-NICU-admitted neonates for childhood mortality risk according to the time of death during the follow-up period. Neonatal NICU admission was significantly and consistently associated with an increased rate of mortality at any time during follow-up (Fig. 2). The neonatal NICU admission after birth was revealed to be strongly associated with neonatal mortality risk at 7 to 27 days (OR = 431.65, 95% CI [132.8, 1402.1] P < 0.001); the risk correlation was pronounced for mortality in the first 7 days after birth (OR = 364.39, 95% CI [145.3, 913.3] P < 0.001); however, remained significantly high at the later period of follow up; specifically, after the age 4 years HR, 7.1 [3.0; 17.0], p < 0.001. The cumulative incidence of mortality in the NICU as compared to the no-NICU admission groups is described by Kaplan–Meier curves, with significant divergence among the groups; while the no-NICU groups stabilized after 4 years of follow up, the mortality risk in the NICU admission group continued to rise through the entire follow-up period of 15 years; p (log-rank) < 0.001 (Fig. 3A). The separate analyses of each of the main population sub-groups showed a similar pattern of mortality risks as the general population; nevertheless, the long-term risk was significantly more evident for the Arab population as compared to the Jewish population. Notably, albeit the higher overall risk for long-term mortality; in the Arab subpopulation group this risk was less related to the NICU admission event as compared to the Jewish population (HR = 13.23 [7.13; 24.5] and HR = 21.69 [15.45; 30.45], respectively) (Fig. 3B).

We further assessed mortality risk during follow-up in association with the diagnoses at the initial NICU admission. Notably, we found that mortality at specific periods was strongly associated with specific causes of NICU admission e.g. mortality at 0 to 7 days was mainly associated with brain and respiratory causes, while mortality at 28 days–6 months was mostly associated with meconium aspiration, infections, and jaundice (Fig. 4). In order to evaluate the significance of such diagnoses separated from other background diseases, we first used the prototype of “neonatal jaundice.” Overall for the study population, we identified 8,295 neonates with a diagnosis of “jaundice”; for 6656 (80.2%) jaundice was the only registered neonatal diagnosis, among which 92 (1.4%) were admitted to NICU. None of these neonates died during the follow-up period. Next, we evaluated the long-term mortality risk with a model that included the diagnoses grouped as disorders with a potentially high severity association strength and those with a low, potentially transient association. We found that in a model adjusted for the potential severity of the disease, the NICU admission event remained an independent and significant risk factor for long-term mortality as compared to the non-admitted neonates; HR = 8.21 95% CI [5.43; 12.4] P < 0.001.

In this study, we show that admission to NICU with any diagnosis is a significant risk indicator of long-term childhood mortality for term neonates.

The study population included 192,527 term neonates, with a 2.75% rate of NICU admission following delivery. Compared to other large-scale studies conducted [8‐11], this admission rate for NICU lies within the lower range; in another study, the risk for NICU admission was 4.1%, and risk factors for admission included non-citizen status, low or no health insurance coverage, and premature rupture of membranes [12]. We suggest that the difference in NICU admission rate may be explained mainly by our study being performed at a single center, with a homogenous population covered by a national health care plan, no out-of-hospital transfers, and a highly trained neonatology service present in the delivery room. In addition, different exclusion criteria and different hospitalization protocols (e.g., induction of preterm premature rupture of membranes prior to term, at 35 weeks’ gestation) could partially explain the difference. However, due to the large population in our study, we were able to comprehensively assess pre-birth information such as maternal background, vital signs and blood tests at admission for birth, characteristics of labor, and neonatal characteristics.

As previously noted, death is considered a common occurrence among term neonates during the NICU hospitalization, with a reported incidence of 8–15% [2, 3]. The existing scoring systems such as SNAPPE-II (Score for Neonatal Acute Physiology with Perinatal Extension-II) serve as a tool to predict early mortality during the NICU hospitalization [3, 13]. Yet, to the best of our knowledge, no study has assessed the association between NICU admission for term neonates and long-term mortality also after hospital discharge. Overall, among the 0.16% of the cohort who died during the 15-year follow-up period, half had records of neonatal NICU admission, a significant and intriguing high rate as compared to neonates not admitted to the NICU; we found no other similar study to compare our rates to and consequently our study may serve as a comparison base for future research.

The Cox proportional hazards model suggested that neonates admitted to the NICU had a 20-fold increased risk for death, not only immediately subsequent to the admission but throughout the entire 15-year follow-up period. Interestingly, the association of this magnitude remained evident after adjustment for gestational age at birth and exclusion of congenital malformations, and chromosomal and genetic syndromes, which are considered well-known leading causes for NICU admission and early mortality [14, 15]. This finding reveals an additional significance for the Arab population who presented an increased risk of post-NICU-related mortality, however, less related to the NICU admission event itself as compared to the Jewish population. beyond the consanguineous marriages and low termination of pregnancy due to antenatal detected fetal abnormalities, we postulate that the Arab population’s higher mortality following NICU admission might be due to the poor post-hospital discharge follow-up and care; both because of the higher maternal rate of postpartum depression, parents’ perception of the importance of neonatal follow-up and the peripheral residence areas with low access to health care, lack of safe spaces and a higher rate of accidental deaths, especially for sick children discharged from NICU [16‐20]. Nevertheless, for both populations, the risk was significantly higher for those admitted to the NICU.

Others have previously described risk factors for early mortality [3, 21], such as low Apgar score, ethnic/population groups, and SGA neonates which were significant also in the present study, at a much lower magnitude with only a two-fold increased death risk. Markedly, the cumulative mortality rate in our study continued to increase throughout the entire follow-up period for those admitted to the NICU; in contrast to a stabilization of the risk observed after 4 years of follow-up among those not admitted to the NICU. It is difficult and beyond the scope of the present study to theorize over an explanatory hypothesis. However, we may postulate that neonates admitted to NICU suggest a role for genetic and epigenetic variations for susceptibility to stress, inflammation-induced damage [22‐24], and different maternal diets and supplements during pregnancy, especially for mothers with diabetes mellitus or hypertensive disorders [25‐27].

While the effects of NICU admission indications and diagnoses on early mortality have been previously discussed [28], the association with late childhood mortality is less well known, especially with respect to term neonates. The most common reasons for a term infant to be admitted to the NICU after birth are temperature instability, hypoglycemia, respiratory distress, and hyperbilirubinemia [29]. Our population showed a similar pattern for diagnoses at the time of the NICU admission. Interestingly, neonates admitted to NICU with potentially early detectable and treatable diagnoses like jaundice and common metabolic complications without any background of metabolic or other diseases (dehydration and electrolytes disorders, hypoglycemia, infant to a diabetic mother) exhibited likewise a sustained long-term childhood mortality risk, at least until 4 years of age. Importantly, hyperbilirubinemia and hypoglycemia in NICU-admitted term neonates are complications that are not overlooked and carefully observed; medical management is sufficient and well tolerated [30]. However, the exact association of each diagnosis at NICU admission on childhood mortality is beyond the scope of the present study; the majority of the neonates had records of one or more diagnoses and the inherent difficulty of record study does not allow us to dissect between the primary or the secondary diagnoses. For example, diagnoses such as jaundice and electrolyte disorders may not stand as an isolated diagnosis for NICU admission. Therefore, jaundice alone has little or no association with the long-term mortality risk for the NICU admitted neonates. Furthermore, it may be obvious to the clinician intuition that the cause and the severity of the condition that determined the neonatal admission to NICU is probably the link to the long-term mortality rather than the event of the NICU admission itself. This is a question that is difficult to resolve in a study based on medical records such as ours. However, in order to evaluate the NICU admission event itself as an independent risk for long-term mortality we approached it by using models fitted for the potential severity of the recorded diagnoses. Indeed, the independent magnitude of the NICU admission risk was reduced two folds yet remained significantly high compared to the non-admitted neonates.

Altogether, we postulate that the short-term mortality increased risk for the neonates admitted to NICU may be attributed mostly to the underlying disease, the long-term mortality risk related to the NICU admission event is enthralling and further studies are required to fully explore into this pattern of long-term mortality risk and its pathogenesis.

NICU admission event for term neonates, probably independent of the diagnosis at admission, is a remarkable event with a significant association with the risk for long-term childhood mortality. Neonates at term admitted to NICU may benefit from long-term focused health care and risk-reducing interventions.