Article Text
PDF
Abstract
Objective (1) To determine which antepartum and/or intrapartum factors are associated with the need for advanced neonatal resuscitation (ANR) at birth in infants with gestational age (GA) ≥34 weeks. (2) To develop a risk score for the need for ANR in neonates with GA ≥34 weeks.
Design Prospective multicentre, case–control study. In total, 16 centres participated in this study: 10 in Argentina, 1 in Chile, 3 in Brazil and 2 in the USA.
Results A case–control study conducted from December 2011 to April 2013. Of a total of 61 593 births, 58 429 were reported as an GA ≥34 weeks, and of these, only 219 (0.37%) received ANR. After excluding 23 cases, 196 cases and 784 consecutive birth controls were included in the analysis. The final model was generated with three antepartum and seven intrapartum factors, which correctly classified 88.9% of the observations. The area under the receiver operating characteristic (AROC) performed to evaluate discrimination was 0.88, 95% CI 0.62 to 0.91. The AROC performed for external validity testing of the model in the validation sample was 0.87 with 95% CI 0.58 to 0.92.
Conclusions We identified 10 risk factors significantly associated with the need for ANR in newborns ≥34 weeks. We developed a validated risk score that allows the identification of newborns at higher risk of need for ANR. Using this tool, the presence of specialised personnel in the delivery room may be designated more appropriately.
- Resuscitation
- Neonatology
Statistics from Altmetric.com
What is already known on this topic?
Little is known about factors predicting the need for advanced neonatal resuscitation (ANR) in newborns ≥34 weeks gestational age. These represent approximately 95% of births.
What this study adds?
This study provides an objective risk assessment for deliveries that require high level of expertise in ANR.
Introduction
The intrauterine to extrauterine transition is a major physiological hurdle for every newborn. While the majority of newborns manage this transition without medical intervention, approximately 5% require assistance at birth and it is estimated (with variation among developed and low-income and middle-income countries) that approximately 2% will require intubation to support respiratory function, and 0.1–0.2% will require chest compressions and/or epinephrine for survival.1 ,2 Successful advanced resuscitation requires not only the knowledge and skill to provide manual ventilation, endotracheal intubation, vascular access and drug administration but also the ability to work as a team.3
The ability to accurately predict the need for advanced resuscitation would be extremely useful both for the provision of care to the vulnerable patient who needs it and for the conservation of expensive resources for those who do not.4 Since the ability to provide positive pressure ventilation is required for all deliveries, the most important question is whether or not the need for advanced life support can be predicted prior to delivery.3
Because little information about the factors associated with the need of advanced resuscitation is available in the scientific literature, we undertook the present study to determine which maternal antepartum (AP) and/or intrapartum (IP) factors are associated with the need for advanced resuscitation at birth in infants with gestational age (GA) ≥34 weeks and to develop a risk score for the need for advanced neonatal resuscitation (ANR) in neonates with GA ≥34 weeks based on those factors.
Patients and methods
Study design
Due to low frequency of ANR in the target population, for which we were looking for risk association, we performed a multicentre, case–control study. In order to increase the accuracy of the data, we chose a prospective study design.
Study population
Study participants were obtained at 16 centres (all level III or IV hospitals): 10 in Argentina, 1 in Chile, 3 in Brazil and 2 in the USA. (A complete list of centres is presented in online supplementary appendix I).
Supplementary appendix 1
Cases
Newborns≥34 weeks GA receiving ANR and free of major congenital malformations were considered the cases. Unexpected major congenital malformations, identified during the delivery process, were excluded. Since the primary goal of the current study was to provide information to help providers to identify and allocate, in advance, a medical team appropriate for the level of risk of each delivery, we did not include preterm babies <34 weeks as well as those with prenatally diagnosed major malformations because two or more highly qualified healthcare providers are routinely assigned to these deliveries in most units.5 For the purpose of this study, ANR was defined as endotracheal intubation or chest compression and/or medication.
GA was determined by the best available method, starting by early ultrasound, followed by menstrual history and, finally, physical exam estimation.
Controls
Four controls for each case were selected among newborns ≥34 weeks GA who did not receive any type of resuscitation procedures during birth and who were delivered at the same centre immediately after the index case. When the case was part of a multiple pregnancy, siblings were excluded as control candidates due to their shared AP and IP factors.
The research protocol and informed consents were approved by the IRB/EC of each participating centre. Two centres in the USA and one in Chile obtained an IRB authorisation for a waiver of the consent since de-identified data from the clinical charts were obtained.
Data management
We took the list of risk variables described by Neonatal Resuscitation Program American Accademy of Pediatrics (NRP-AAP) as predictive of the need for resuscitation.3 This is a rational list of variables that come mostly from consensus of expert opinion, but there is a lack of evidence-based list. We speculate that the babies needing advanced resuscitation will share at least some of these variables.
To avoid losses and errors of exclusion and inclusion criteria, the principal investigator and subinvestigator at each centre were responsible for identifying all the cases and matched controls included in this study. Central coordinators of the study conducted random audits among centres to confirm the patient registration forms of the study with birth statistics.
Parents of all study enrolees provided written, informed consent before maternal discharge in order to include their infants' information into this study.
We classified all the variables associated with the need for advanced resuscitation into AP and IP groups. In addition, the type of resuscitation, the interventions implemented in the delivery room (DR) and the level of training of the resuscitation team, as well as the number of people involved were analysed. We defined significant IP bleeding as profuse vaginal bleeding source that produces haemodynamic disorders and for clinical chorioamnionitis (CLC) we followed Tita and Andrews publication.6
Sample size calculation and statistical analysis
Sample size
For a case/control rate of 1:4, an estimated prevalence of risk factors of 10%, a power of 80% and a significance level of 5%, 172 cases and 688 controls (total 860 newborns) were needed.
Data analysis
The analysis was planned in three sequential stages: (1) the evaluation of the risk factors as determinants of ANR, (2) the development of a predictive model and (3) the validation of the model. The first stage in data analysis was to evaluate the risk factors of ANR, which were included in this study (see table 1). These risk factors were derived from the NRP-AAP list of variables associated with the need for neonatal resuscitation (see table 3). Analysis of these variables began with simple statistics (χ2, Fischer's exact test, Shapiro, etc.) to determine whether or not each of these risk factors was significantly associated with ANR. A univariate logistical regression analysis was then performed on each factor to assess its value as a predictor of the event of interest (ANR). A final determination of the appropriateness of including those risk factors that were predictive of ANR in the creation of a predictive model (stage 2 analysis) was done by evaluating each risk factor along with its variation coefficient using the Wald and likelihood ratio tests. In stage 2, those predictor variables found to be appropriate for inclusion were used to develop a predictive model for ANR. The data used to build this model came from those patients enrolled in the 10 study centres located in Argentina. Data obtained at the remaining centres were later used to validate this model (analysis stage 3). Those predictor variables deemed appropriate, based on univariate and bivariate analysis of these variables done in stage 1, were included in the predictive model. To create this model, adjusted multivariate analysis was subsequently performed introducing the variables in the model one by one to establish a correlation between the variable of interest (ANR) and emergency caesarean section (ECS), entered first based on the univariate analysis, followed by the other predictor variables introduced in the model in order to determine which were confounders and/or effect modifiers. Once the final model was identified, results were presented as OR with a 95% CI. A significance level of 5% had been determined for all effects. Once created, the validity of this predictive model was assessed using the data collected from study centres in Chile, Brazil and the USA. Discrimination and calibration were measured in this test sample in order to assess the external validation of the model. The third state of data analysis involved the validation of the predictive model created in stage 2. This was done by calculating the area under the receiver operating characteristic plot (AROC) to assess discrimination. Calibration was evaluated using the Hosmer–Lemeshow goodness of fit test. Stata V.11 statistical software was used for these analyses (StataCorp, College Station, Texas, USA). Finally, in stage 4, the predictor variables developed from the predictive model were used to create an ANR weighed risk score. The model's weighted risk score was calculated by adding the coefficient of the predictors and then dividing by their common denominator. The results of this equation are then transformed into a score associated with each variable, so that the clinician can easily estimate the relative risk of each delivery for its likely need of ANR.
Characteristics of study population
Results
Between December 2011 and April 2013, 61 593 deliveries occurred in the 16 participating centres, 3164 of them were <34 GA (5%). Out of the remaining 58 429 births, 220 newborns (0.37%) met the inclusion criteria. After excluding 24 cases (21 did not provide informed consent, 1 unexpected major malformation and 2 due to insufficient DR data), 196 cases were included in the analysis with the 784 corresponding controls (figure 1).
Flow chart: Description of included and excluded patients. ANR, advanced neonatal resuscitation; GA, gestational age.
Participating centres were stratified into two groups based on the number of deliveries performed each year (≤3000 deliveries vs >3000 deliveries ear). Analyses of these groups found no statistical difference in the incidence of ANR: 7 centres with >3000 deliveries a year performed 157 ANRs (mean=0.33, SD=0 .05, 95% CI 0.19 to 0.46) and the 9 centres performing <3000 delivering performed 39 ANRs (mean=0.22, SD=0.03, 95% CI 0.14 to 0.30). The number of cases from each site is available in online supplementary appendix II.
Supplementary appendix 2
GA was similar between groups (table 1). From the cases, six newborns (3%) died in the DR; 180 (91.8%) were admitted in the neonatal intensive care unit (NICU) and 10 (5.1%) to the intermediate care nursery. Only eight (1%) newborns in the control group presented with an Apgar score ≤5 at 1 min, and there were no scores ≤5 at 5 min after birth in this group. From the control group, 22 newborns (2.8%) were admitted to intermediate care nursery and 761 (97%) went to rooming-in after birth. None of this group died (table 1).
For the 196 cases, a team composed an average of 3.2 people (SD=0.8) participated in the resuscitations, at least one of which was a neonatologist. Table 2 describes the certifications and credentials of team members and the interventions in each case.
Characteristics of resuscitation team and interventions of 196 cases
Univariate analysis
Table 3 shows the OR and 95% CI of each of the variables currently suggested by the NRP-AAP to identify the deliveries needing advanced resuscitation.
Variables associated with the need for ANR
Among maternal factors, we found eight variables significantly associated with the need for ANR: no prenatal care, GA 34–37 weeks, IP maternal fever, CLC, rupture of membranes >18 hours, pregnancy-induced hypertension, gestational diabetes (GD) and pre-eclampsia (table 3).
Among fetal factors, intrauterine growth restriction (IUGR) and ANR were statistically significant associations (table 3).
Between factors described by the NRP-AAP as predictive of ANR and the occurrence of ANR, we found 10 variables of statistical significant association: requirement of forceps or vacuum delivery (FD), general anaesthesia (GETA), ECS, fetal bradycardia (FB), meconium stained amniotic fluid (MSAF), purulent amniotic fluid, abruption placentae (ABP), placenta previa, magnesium sulfate treatment and significant IP bleeding (table 3).
Multivariate analysis
Regression analysis of maternal and delivery-related risk factors revealed two sets of risk factors associated with ANR (table 4).
Multivariable logistic analysis: antepartum and intrapartum factors
We selected three variables (GA 34–37 weeks, IUGR, GD) and used them to correctly classify 81.85% of study participants as either needing or not needing ANR with a sensitivity of 11.73%, a specificity of 99.38%, a positive predictive value of 82.61% and negative predictive value of 81.83%, using multivariate analysis. The AROC was performed to evaluate discrimination, this being 0.574 with a p value <0.009 (figure 2).
Multivariate logistic model with only antepartum factors. Ten centres in Argentina. Cases=162/controls=648. Area under the receiver operating characteristic (AROC) curve=0.574—Prob>χ2=0.0096.
We finally identified a selection of 10 variables (three AP and seven IP factors) that correctly classified 88.89% of the observations. This model has a sensitivity of 72.22%, a specificity of 93.06% with a positive predictive value and negative predictive value of 72.22% and 93.03%, respectively. The AROC was performed to evaluate discrimination, this being 0.86 with a p value <0.00.1 (figure 3). This resulted in a probability of Prob>χ2=0.4431. There was no statistical difference between observed and expected need of ANR.
Multivariate logistic model with antepartum factors+intrapartum factors. Ten centres in Argentina. Cases=162/controls=648. Area under the receiver operating characteristic (AROC) curve=0.88—Prob>χ2≤0.0001, 95% CI 0.62 to 0.91.
To test the model, we decided to use a validation sample. In this sample, the AROC was 0.85 with a p value <0.001 (figure 4). We evaluate the model calibration again with the Hosmer–Lemeshow test. This resulted in a probability of Prob>χ2=0.4431 since no difference was found between the observed and expected.
Validation—multivariate logistic model with antepartum factors+intrapartum factors. Six centres (Brazil, Chile, USA). Cases=34/controls=136. Area under the receiver operating characteristic (AROC) curve=0.87—Prob>χ2≤0.001, 95% CI 0.58 to 0.92.
ANR risk
We developed an overall ANR risk assessment based on multivariate logistic model with three AP variables and seven IP variables (table 4). In this analysis, each variable has a statistical weight calculated by the multivariate logistic model. Evaluating each pregnancy for the presence or absence of AP or IP factors returns a rate of the risk for the need of ANR (table 5).
Individual risk assignable to each variable to ANR
Based on these data, we created a risk factor calculator (available for download at: http://www.oumedicine.com/pediatrics/department-sections/neonatal-perinatal-medicine/links).
For example, the presence of at least two of the listed IP factors exceeds the 70% of chance for the need of ANR; and the presence of two AP and one IP risk factors included in this list exceeds the 90% of chance for the need of ANR (table 6).
Composite risk assignable to combine variables to ANR
Discussion
In this large multicentre case–control study of babies ≥34 weeks, GA we were able to identify 10 variables significantly associated with the need for ANR. These variables included three AP (34–37 weeks GA, IUGR and GD) and seven IP factors (MSAF, CLC, ABP, FB, forceps delivery (FD), GETA and ECS). Using these, we were able to construct a validated risk score as a tool to calculate the probability of the need for ANR in each case.
Several guidelines recommend having at least two people trained in full resuscitation skills in cases of high-risk pregnancies, and that number can be increased to 3–4 people with different degrees of skills acting as a team for each baby.3 ,7
The results of our study will help to reduce the impact on resource allocation by better defining which deliveries are likely to need ANR, thus allowing hospitals to better allocate and use both personnel and expensive resources.
This study did not include babies under 34 weeks GA. The term prematurity as we used it here refers only to babies between 34 and 37 weeks GA, commonly referred to as late preterm babies in published literature. Our study found these late preterm infants to be at higher risk for ANR. In a previously published study, de Almeida et al8 compared the resuscitative procedures in this specific population and obtained similar results.
In the present study, IUGR and GD remarkably increased risk for ANR. In a large series of newborns classified by birth weight percentiles, McIntire et al9 found a significant increase in the rate of intubation, proportion of babies with Apgar scores ≤3 at 5 min and umbilical-artery blood pH ≤7 in babies born with birth weights less than the third percentile group. In a recently published study, Reif et al10 report a significantly lower arterial pH among children born to diabetic mothers compared with controls.
MSAF has been recognised for >30 years as associated with respiratory depression at birth.11 In a recently published study reviewing 1 year's deliveries in a single centre, Afjeh et al12 found a 0.65% rate of intubation for resuscitation and, in accordance with our findings, MSAF and CLC were significantly associated with ANR in the multiple regression analysis.
Placental abruption increases the risk of neonatal asphyxia and need for expanded neonatal support. The risk factors for this condition were recently explored in a cohort study over 10 years by Boisramé et al,13 who found a significant association with Apgar scores ≤5 at 5 min, umbilical pH ≤7 and neonatal resuscitation.
We decided to control for FB in order to increase the homogeneity in the diagnoses among centres. In spite of the controversy surrounding cardiotocographic interpretation, there is consensus about the low risk associated with normal cardiotocographic tracing and on the association of late decelerations and bradycardia with unsatisfactory neurological outcomes.14–16
We found that forceps and vacuum-assisted deliveries were associated with an increase in risk for ANR. This is consistent with the identification of a significant level of risk for neonatal encephalopathy recently published by Walsh et al17 in the analysis of a cohort of 76 810 term singleton neonates. Our findings regarding the increase in the risk associated with GETA as well as with ECS are also consistent with other studies addressing these conditions.4 ,18 ,19
To our knowledge, this is the only prospective study addressing the predictive factors, both antenatal and IP, for ANR. Similar studies have seldom been undertaken. Aziz et al reported a single-centre analysis of their prospectively collected data in the DR including both preterm and term infants with some coincidences (IUGR, maternal diabetes, MSAF, GETA, non-reassuring fetal heart rate, FD) and several differences in the variables considered as predictors for neonatal resuscitation compared with our findings. These differences between their and our findings could be attributable to (1) the definition of the cases (they included those babies requiring positive pressure ventilation (PPV) or continuous positive airway pressure (CPAP), and/or intubation for suctioning leading to an inclusion of 3564 out of 5691); (2) the different sample size (we screened 61 593 deliveries) and, (3) the fact that our study only recruited babies ≥34 weeks while Aziz's study included infants of all GA.4
Our study is not free of limitations. We did not match cases and controls by birth weight, choosing instead the immediately consecutive delivery that makes the populations comparable in terms of time and the assisting team. Nevertheless, the mean birth weight and GA were comparable between groups. We did not include those babies requiring only bag and mask ventilation since the current recommendation is to have in every delivery a person capable of initiating resuscitation including bag and mask ventilation.
Unlike many US and European hospitals, it is standard practice for a neonatologist to be present at all deliveries at all non-US centres participating in this study. We collected only the professional qualifications of the primary health provider for the cases and assumed that a similar team was present at the control deliveries in the non-US centres, but the professional qualifications of all attending personnel were not collected.
The main objective of this study was to predict ANR in order to anticipate the need of a DR team with skills above the above-mentioned baseline.
Conclusions
We evaluated risk factors associated with the need for ANR in newborns ≥34 weeks.
Those statistically significant were GA 34–37 weeks, IUGR, GD, MSAF, FB, ABP, ECS, CLC, GETA and forceps-assisted deliveries.
We developed a validated risk score that may improve the ability to predict the need for ANR. This new tool allows the identification of newborns at higher risk of need for ANR; in these cases, the need for specialised personnel in the DR may be anticipated.
References
Footnotes
Collaborators Jorge Tavosnaska, Carolina Olmo Herrera, Maria Lidia Failla, Valeria Quispe, Rocio Decena, Mariela Torres, Adriana Cazón, Lorna Andreussi, Carola Capelli, Lucrecia Rosales, Guillermo Colantonio, Ramón Pio Larcade, Héctor Sexer, Natalia Davase, Maria Lidia Failla, Liliana Roldan, Lory Lewis, Cristina Pinedo, Adriana Castro, Ana López, Guillermo Sudini, Ana Yessica López, Sandra Barón, Jorge Fabres, Paulina Toso, Miriam Faune Perez, F Martin, Ligia Maria Souza, Sippo Rugolo, Fabiana Enjamio, Macarena Paz Trigo.
Contributors JPB, ES, AA, ME and FA conceptualised and designed the study, drafted the initial manuscript, and approved the final manuscript as submitted. JPB, AA, AF, DD and FS carried out the initial analyses, reviewed and revised the manuscript, and approved the final manuscript as submitted. RG and MFBdA designed the data collection instruments, and coordinated and supervised data collection at four (San Pablo ANR network) of the 16 sites, critically reviewed the manuscript and approved the final manuscript as submitted. GA, MV, DA and GP coordinated and supervised data collection in their sites, critically reviewed the manuscript and approved the final manuscript as submitted.
Funding This study was funded by FUNDASAMIN (Fundación para la Salud Materno Infantil) and IECS (Instituto de Efectividad Clínica y Sanitaria), Buenos Aires, Argentina.
Competing interests None declared.
Patient consent Parental/guardian consent obtained.
Ethics approval The research protocol and informed consents were approved by the IRB/EC of each participating centre.
Provenance and peer review Not commissioned; externally peer reviewed.
Linked Articles
- Fantoms