Objective Prematurity and other risk factors are associated with the development of intraventricular haemorrhage (IVH) in newborns with respiratory distress syndrome (RDS). Conversely, further analysis can determine what characteristics might be associated with a decreased risk of IVH.
Study design By using International Classification of Diseases, Ninth Revision, Clinical Modification codes from data obtained from the Nationwide Inpatient Sample of the Healthcare Cost and Utilization Project from 2000 to 2009, we identified a large number of cases of RDS. Multivariable logistic regression analysis identified potential variables associated with decreased risk of IVH.
Result Our cohort included 194 621 neonates with RDS, of whom 20 386 (10.5%) developed IVH. Variables associated with decreased risk of both all grades of IVH and severe IVH only included infant of diabetic mother (IDM) status (OR 0.62 (0.54 to 0.70), p<0.001; OR 0.56 (0.42 to 0.74), p<0.001), Trisomy 21 (OR 0.45 (0.30 to 0.69), p<0.001; OR 0.38 (0.16 to 0.93), p=0.034), maternal hypertension (OR 0.62 (0.53 to 0.72), p<0.001; OR 0.28 (0.18 to 0.43), p<0.001), caesarean birth (OR 0.79 (0.74 to 0.84), p<0.001; OR 0.83 (0.73 to 0.94), p<0.001) and, consistent with prior studies, female gender (OR 0.85 (0.82 to 0.88), p<0.001; OR 0.76 (0.72 to 0.80), p<0.001). Polycythaemia (OR 0.67 (0.49 to 0.92), p=0.013; OR 0.79 (0.43 to 1.45), p=0.449) and hypothermia (OR 0.86 (0.75 to 0.99), p=0.039; OR 1.01 (0.81 to 1.28), p=0.903) were associated with lower risk of all IVH but not severe IVH only.
Conclusions Previous associations with IVH such as lower birth weight were confirmed. However, infants in whom new variables such as IDM status were present were less likely to develop all IVH grades. Further analysis of these potential protective variables is necessary to better understand the pathophysiology of IVH.
Statistics from Altmetric.com
What is already known on this topic
Intraventricular haemorrhage (IVH) is a major cause of morbidity and mortality in the premature newborn with respiratory distress syndrome (RDS).
While numerous studies have reported risk factors associated with the development of IVH in this population, fewer studies have identified variables associated with decreased risk of IVH.
What this study adds
This study examined a relatively large cohort of premature newborns with RDS obtained from a diverse national database.
While previously established protective factors were confirmed, maternal diabetic mother status and neonatal hypothermia were identified as novel variables that are associated with decreased risk of IVH.
Intraventricular haemorrhage (IVH), one of the most significant causes of morbidity and mortality in the premature newborn,1 was first staged radiologically by Papile et al.2 Multiple studies have reported an inverse relationship between incidence of IVH and factors such as birth weight (BW) and gestational age (GA).1 ,3–6 Increasing severity of illness among premature newborns is also associated with higher IVH rates.7–11 While improvements in neonatal care have led to decreasing incidence of IVH among individual neonatal centres,11 increasing survival rates in extremely premature neonates have concurrently led to increased numbers of IVH cases.8 A history of IVH, even if mild (grades I and II), also remains a cause of worse neurodevelopmental outcomes among neonatal intensive care unit graduates with haemorrhages compared with those without.12 While examining the incidence of IVH, numerous studies have identified reproducible risk factors that are not necessarily markers of severity of illness such as male gender and transport after birth but nevertheless contribute to its occurrence.10 ,13–15 Fewer studies have identified variables associated with decreased risk for IVH such as antenatal steroid use, pregnancy-induced hypertension, magnesium sulfate therapy, antenatal tocolysis and postnatal administration of indomethacin and/or surfactant.10 ,15 ,16 Analysis of these as well as other potential variables in a large sample size of neonates with respiratory distress syndrome (RDS) would lead to a better understanding of why certain neonatal populations have lower rates of all grades of IVH.
Materials/subjects and methods
We conducted a retrospective cross-sectional study of a large sample of neonates with RDS using data obtained from the Nationwide Inpatient Sample (NIS) for the years 2000–2009, sponsored by the Agency for Healthcare Research and Quality. The NIS database is maintained by the US Department of Health and Human Services as a part of the Healthcare Cost and Utilization Project and is the largest all-payer inpatient care database in the USA representing 20% of all hospitalisations. The database contains data elements such as patient demographics, hospital characteristics, primary and secondary diagnoses, procedures, hospital charges, length of stay and discharge disposition. Detailed information regarding the content and the methods of the NIS is available at http://www.hcup-us.ahrq.gov/nisoverview.jsp.
International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) codes were used to identify neonates with RDS as well as other clinical and demographic factors. Details of these ICD-9-CM codes are listed in online supplementary table S1. Professional coders entered data in every contributing hospital. Oversight/error correction is provided by the NIS database, which is available at https://www.hcup-us.ahrq.gov/db/quality.jsp#procedures. Exclusion criteria were admission ages >30 days and cases transferred out to another hospital to prevent double counting of the same cases. Neonates were divided into groups based on their GA and BW. BW was classified by ranges and was therefore analysed as categorical data:
GA (weeks): (1) ≤28, (2) 29–32, (3) 33–36, (4) ≥37
BW (g): (1) ≤999, (2) 1000–1499, (3) 1500–2499, (4) ≥2500
The outcome of interest was development of any grade IVH. IVH cases were divided into mild (grades I and II) and severe (grades III and IV) with comparison of demographic and clinical factors associated with severe IVH versus mild or no IVH as well as with any IVH versus no IVH. As the prior literature has shown conflicting evidence regarding impact of mild IVH on outcomes,17–20 while severe IVH has been universally shown to be associated with worse outcomes compared with both no IVH and mild IVH, we decided to explore the associations of severe IVH compared with the combined group of no IVH and mild IVH with various predictors in this study. Tables 1 and 2 outline these demographic and clinical features as well as characteristics of the discharging hospitals.
Comparisons between categorical variables were made using the χ2 test. Unadjusted ORs were obtained by univariable logistic regression using one predictor variable and one outcome variable. Adjusted analyses were calculated by using multivariable logistic regression. Two separate models were created using all IVH and severe IVH as outcomes variables. The following covariates were included in both regression models based on significant univariable associations: Trisomy 21, infant of diabetic mother (IDM) status, maternal hypertension, polycythaemia, caesarean birth, hypothermia, patent ductus arteriosus (PDA), neonatal sepsis, chest tube insertion, endotracheal intubation, umbilical vein catheterisation and twin pregnancy. BW, gender, race and transfer status were also controlled for in regression models. We did not use the grades of IVH as an ordinal outcome variable in the regression models. The two outcomes variables IVH versus no IVH and severe IVH versus no/mild IVH were used as binary outcome variables in the models. Therefore, assumptions of proportionality were not applicable in our analyses and were not analysed. Additional analyses included patients with known BW of <1500 g to adequately control for the BW group with the highest risk of IVH and are included in the online supplementary tables S3 and S4. Statistical significance was set at p<0.05 a priori. Significant numbers of data were missing for variables like race/ethnicity (25% missing data), BW, where missing data were coded as ‘birth weight unspecified’ (18% missing data), and GA, where missing data were coded as ‘gestational age unspecified’ (40% missing data). For other demographic variables such as gender, negligible cases (<0.1% cases) had missing data, which is illustrated in figure 1. As the diagnoses and procedures were defined by the presence or absence of related ICD-9 codes in the discharge abstracts, these variables had no missing data. Therefore, to address missing data and prevent exclusion of cases from regression models, race/ethnicity and missing BW were coded with a separate category of ‘missing race/ethnicity’ and ‘missing birth weight’, respectively. As there is a strong collinearity between BW and GA, we included only BW in the regression models as GA had significantly more missing data. The Statistical Package for the Social Sciences V.17.0 (SPSS, Chicago, Illinois, USA) was used for all data analyses.
Figure 2 describes the study population of 194 621 neonates with RDS, which included 20 386 (10.5%) who developed IVH. Severe IVH occurred in 5566 (2.9%) neonates representing 27.3% of all cases of IVH in this cohort. Consistent with previous reports, tables 1 and 2 show higher mortality rates in IVH versus no IVH, and in severe IVH versus mild/no IVH groups. In addition, patients with IVH were more likely to be lower in GA and BW, transferred from another hospital and to be male compared with those with no IVH. Multivariable logistic regression showed the lower the BW, the higher risk of IVH (see online supplementary files 3 and 4), which examine the high-risk group of BW <1500 g.
On univariable analysis, variables associated with lower rate of any degree of IVH (table 3) as well as severe IVH (table 4) were IDM, maternal hypertension, polycythaemia and Trisomy 21. Variables associated with increased rate of all IVH grades (table 3) and severe IVH (table 4) were hypothermia, PDA, necrotising enterocolitis, neonatal sepsis, chest tube insertion, endotracheal intubation/ventilation, umbilical venous catheterisation, pneumothorax (independent of chest tube insertion) and pulmonary haemorrhage. On multivariable logistic regression analysis, the perinatal factors of IDM status, maternal hypertension, Trisomy 21 and caesarean birth were independently associated with lower risk of both all grades of IVH and severe IVH (see online supplementary files 5 and 6). However, polycythaemia and hypothermia, the latter of which was associated with increased risk for all and severe IVH in univariable analysis, were associated with lower risk of all IVH but not severe IVH.
Since early death is a potential competing outcome with IVH, a logistic regression using death or IVH as outcome was conducted. Multiple variables such as lower BW and markers of severity of illness (ie, neonatal sepsis) have an increased risk of IVH or death (see online supplementary table S2).
The goal of this study was to determine variables associated with all IVH and severe IVH in a large sample size of neonates with RDS. The incidence of IVH is lower in this study than previously reported studies of smaller sample sizes.1 ,5 ,21 The denominator used for analysis was all patients with RDS. Some cases of IVH may have been missed in patients who did not survive to undergo ultrasonography screening, and this database does not contain information regarding whether ultrasounds were performed. It is important to note that this study only includes neonates with a diagnosis of RDS, as opposed to a study population including all premature newborns. Additionally, while this study confirmed several well-established variables associated with an increased rate of IVH, such as lower GA and BW, sepsis, pneumothorax, umbilical line placement, neonatal transport, PDA, intubation and mechanical ventilation and male gender,8 it also underscored the existence of potential variables associated with a decreased IVH rate. Both GA and BW have been categorised, and due to the binary type of these variables, these data appear as histograms to better understand the relationship with outcome (figures 3 and 4). We confirmed the findings of prior smaller studies, which identified maternal hypertension as protective.16 ,22 Cortisol-induced in utero fetal maturation leading to reduction in inotrope usage and other markers of severity of illness may be the mechanism responsible for this protective effect.8 ,10 In addition, maternal hypertension may lead to preterm delivery to avoid progression from pre-eclampsia to eclampsia. Although premature, these fetuses are likely to be delivered before exposure to preterm labour and ultimately chorioamnionitis, a known risk factor for IVH due to the resultant inflammatory cytokine cascade.22–24 Tocolysis of pre-eclamptic mothers with magnesium sulfate has also been shown to be an independent protective variable against IVH through an unclear mechanism.4 ,10 ,16
Study of mode of delivery and its impact on incidence of IVH have yielded conflicting results in the literature. Despite concerns for increased intracranial pressure causing IVH during passage through the birth canal, studies have shown favourable effects on the incidences of IVH in premature newborns delivered by both vaginal and surgical routes.7 ,25 ,26 In this report, multivariable analysis of a large study population found caesarean birth to be associated with a lower rate of any IVH and severe IVH (table 5).
A new variable associated with decreased IVH identified in this study population is maternal diabetic mother (DM). In contrast to prior studies that demonstrated no effect of maternal DM status on the incidence of IVH,27 our analysis reveals that the infant of a DM has a significantly lower rate of mild and severe IVH. Near infrared spectroscopy analysis has shown that increased cerebral blood flow (CBF), associated with conditions like hypoglycaemia, is associated with IVH.28 Therefore, protection may begin in utero for the potentially hyperglycaemic fetus. Once delivered, postnatal hypoglycaemia in the IDM is aggressively avoided, thereby avoiding an increase in CBF, with its inherent risk of IVH. In addition, we noted a reduced incidence of IVH associated with Trisomy 21, which, to our knowledge, has not been previously reported. We speculate that the relative polycythaemia seen in Trisomy 21 may be responsible for this protective effect. Despite Trisomy 21 being associated with a lower risk of IVH, multivariable logistic regression analysis revealed an increased risk for the combination outcome of IVH or death. Although the database does not contain BW or GA age for each individual Trisomy 21 patient, we speculate that this finding may suggest that Trisomy 21 has a lower IVH rate because the syndrome is typically seen in term and late preterm newborns, whom are not at increased risk of IVH. Death is likely higher in these patients due to congenital heart disease and other medical issues common in this population.
Another novel variable identified in this study was hypothermia. Previous studies of smaller cohorts showed that hypothermia was associated with an increased risk of IVH.29 ,30 Our study, however, showed that hypothermia at birth reduced the risk of IVH. Although a substantial literature exists on the neurodevelopmental benefits of therapeutic hypothermia in the asphyxiated term newborn,31 no prospective trials have evaluated the effects of this condition on neurological sequelae in preterm newborns. It is certainly feasible that hypothermia in the preterm population might yield neurodevelopmental benefits by reducing the risk of IVH. An additional variable that is associated with a lower rate of IVH was neonatal polycythaemia. A growing body of data has revealed benefits of delayed cord clamping, including a reduction in the incidence of IVH.32 ,33 We speculate that delayed cord clamping and resultant relative polycythaemia may lead to hyperviscosity in the cerebral vasculature, which might reduce CBF and prevent IVH.
The major strength of this study is attributable to the relatively large study population derived from a large and diverse national database. However, limitations of this study are its retrospective cross-sectional design and the inability to control few potential confounding variables such as administration of antenatal steroids, and magnesium sulfate and post-natal administration of indomethacin. Missing data not provided by the contributing hospitals is another limitation. However, >99.9% of all cases (194 557 out of 194 621) were included in the logistic regression. Only 64 patients with missing data regarding gender were excluded from the model.
The database also does not capture certain clinical and biochemical information that is not reported via ICD-9 codes, such as values of PaCO2, serum sodium, Apgar scores and timing of cranial ultrasounds. Potential lack of ultrasonography in some patients due to early death may limit the validity of the results (especially low-grade IVH) because some patients may have late IVH, especially low grade. Data were entered in database by professional billing coders. Approximately one-fourth of all included cases did not have data regarding race/ethnicity due to the decision not to report them in some states. Although the missing race/ethnicity data are expected to be randomly distributed with regard to IVH, variations in IVH rates among different races/ethnicities should still be interpreted with caution. Although potentially protective in their own right, conditions like hypertension and DM may result in other alterations in prenatal and postnatal care that may be responsible for the reduced incidence of IVH. The potentially protective effects of these and the other factors identified must be verified prospectively in future studies.
The authors gratefully acknowledge the maintenance of the NIS database.
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
Competing interests None declared.
Ethics approval The Institutional Review Boards of both Staten Island University Hospital and the SUNY Downstate College of Medicine granted approval for data analysis without requiring consent from the de-identified, publicly accessible data available in the NIS.
Provenance and peer review Not commissioned; externally peer reviewed.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.