Background The aim of this study was to determine if the National Institute of Child Health and Human Development (NICHD) calculator, designed to predict mortality or neurosensory disability in infants 22–25 weeks’ gestation, was valid for contemporary Australian infants.
Method Outcome data at 2 years of age for 114 infants who were liveborn in Victoria, Australia, in 2005, between 22 and 25 completed weeks’ gestation, weighing 401–1000 g at birth, and free of lethal anomalies, were entered into the NICHD online calculator. Predicted outcomes were then compared with the actual outcomes.
Results Of the 114 infants, 99 (87%) were inborn and 15 (13%) were outborn. The overall prediction of death for inborn infants was 47.1% compared with the actual death rate to 2 years of age of 49.5%. The area under the curve (AUC) was 0.803 (95% CI 0.718 to 0.888; p<0.001) for mortality, comparable with the AUC for the NICHD study (AUC: 0.753; 95% CI 0.737 to 0.769; p<0.001). The accuracy for predicting death was not as precise for outborn infants (AUC: 0.643; 95% CI 0.337 to 0.949; p=0.36). The calculator overestimated the combined outcome of death or survival with major disability at 72.0%, compared with an actual rate of 60.5%.
Conclusions The NICHD outcome estimator was helpful in predicting mortality for inborn infants, 22–25 weeks’ gestation, but was less precise for outborn infants. It overestimated the combined outcome of death or major disability in infants born in Victoria, Australia, in 2005.
- Outcomes research
Statistics from Altmetric.com
What is already known on this topic
Gestational age is often used as the basis for predicting the likelihood that an extremely preterm infant will not survive, even if intensive care is offered.
The NICHD have shown that the combination of gestational age with four additional factors improves the ability to accurately predict mortality and major morbidity in infants born between 22 and 25 weeks’ gestation.
What this study adds
The NICHD calculator is an accurate tool for predicting mortality in extremely preterm infants who are inborn in Victoria, Australia.
The NICHD calculator may overestimate the risk of adverse neurodevelopmental outcome in 22–25-week gestation infants, born in Victoria, Australia.
Perinatal healthcare providers need access to accurate and current outcome data to counsel parents facing the birth of an extremely premature infant. Parents will want to know their infant's risk of mortality, as well as the risk of long-term serious neurodevelopmental morbidity if their infant survives. It is essential that gestation-specific mortality and morbidity data can be presented to parents in such a way that they are able to make informed decisions regarding the management of the birth and subsequent care of their infant.1
Misconceptions regarding an infant's potential for survival may influence the management of labour and delivery, including decision making regarding transfer of the woman to a perinatal centre prior to birth.1–3 Furthermore, studies have shown that many perinatal healthcare providers underestimate survival and overestimate disability rates in extremely premature infants.2–4 To ensure optimal obstetric management and appropriate counselling of parents, clinicians need simple and accurate methods for estimating an individual infant's chance of death or survival with major disability.
Researchers at the National Institute of Child Health and Human Development (NICHD) in the USA have developed an online ‘outcome estimator’.5 The outcome estimator was developed prospectively by a team of researchers from the Neonatal Research Network (NRN) at the NICHD, to predict the likelihood of a favourable outcome if an infant born between 22 and 25 completed weeks’ gestation was offered intensive care. The researchers at the NRN developed a five-factor model that uses birth weight, gender, plurality and maternal antenatal corticosteroid exposure in addition to gestational age, to generate possible outcomes. Compared with using gestational age alone, the NRN researchers found that the five-factor model resulted in a higher degree of accuracy when estimating an extremely preterm infant's potential for a favourable or unfavourable outcome.6
The aim of our study was to determine if the NICHD outcome estimator was valid for contemporary infants born in Victoria, Australia, including those who were outborn. If shown to accurately predict mortality and morbidity in infants born in Victoria, local perinatal healthcare providers could use the outcome estimator to guide care around an extremely preterm birth, and to assist with counselling the parents.
The NICHD outcome estimator in http://www.nichd.nih.gov/neonatalestimates is publicly available via the NICHD web site.5 Five criteria are required to generate a tabulated list of potential outcomes for an individual infant. These criteria are: gestational age (22–25 completed weeks), gender, birth weight (401–1000 g), plurality (singleton or multiple), and any maternal antenatal corticosteroid exposure within 7 days of the birth (yes or no). In the original study, strict inclusion and exclusion criteria were applied. The outcome estimator will not generate an outcome for an infant with a birth weight <401 g, as very few would be offered intensive care below this birth weight. Infants with a birth weight >1000 g or above the 97th centile are excluded on the basis that gestational age may have been underestimated. Infants with a major congenital anomaly are also excluded.6
The possible outcomes generated by the calculator are based on the outcomes of 4446 extremely premature infants inborn at one of the 19 tertiary perinatal centres in the NRN hospitals in the USA between 1998 and 2003.6 ,7 Among the infants in the study cohort, 3702 (83%) received intensive care in the form of mechanical ventilation. The outcomes for all the infants and the outcomes for those who received mechanical ventilation were analysed using a logistic mixed model. These form the basis of the NICHD outcome estimator.5
Having entered all five criteria for an individual infant into the calculator, the possible outcomes are provided in two tables: one for all infants, and a second only for those who receive intensive care. Six possible outcomes are generated by the calculator. These are: survival, survival without profound neurodevelopmental impairment, survival without moderate to severe neurodevelopmental impairment, death, death or profound neurodevelopmental impairment and death or moderate to severe neurodevelopmental impairment. Each possible outcome is calculated as a percentage risk.
We used data from the 2005 Victorian Infant Collaborative Study (VICS) to test the accuracy of the NICHD outcome estimator for contemporary infants in Victoria. Of the 382 infants in the 2005 VICS cohort who were liveborn at <28 weeks’ gestation or weighing <1000 g, birth details, mortality and morbidity data were available for 114 infants who also met all five criteria for the NICHD outcome estimator. Survivors were assessed at 24 months of corrected age by a team of psychologists and paediatricians who were blinded to the child's perinatal details. Development was assessed with the Bayley Scales of Infant and Toddler Development (Bayley-III),8 with the children's scores compared with the mean and SD of contemporaneous term controls who were matched for socioeconomic variables.9 ,10 By comparing extremely preterm children to randomly selected term controls, we have shown that the proportions of extremely preterm survivors with developmental delay have been similar, irrespective of which version of the Bayley Scales have been used.9 ,10 Major disability was defined as any of the following. Developmental delay: (Cognitive or Language Composite Scores <−2 SD relative to controls); moderate or severe cerebral palsy: (Gross Motor Function Classification system, Level>1), blindness (visual acuity worse than 6/60 in the better eye) or deafness (requiring hearing aids or worse). Our criteria for major disability are identical to the NICHD criteria for moderate to severe neurodevelopmental impairment for vision, hearing and cerebral palsy, but differed for developmental delay where the NICHD used cut-offs of <70 for either the Mental or Psychomotor scales of the Bayley-II, which would be more than 2 SD below the mean based on the expected norms. Profound disability was defined as either cognitive or language score <−3 SD relative to controls, or a Gross Motor Function Classification System level of 4 or 5. The reported outcomes for the VICS study were death before 2 years of corrected age, death or major disability, death or profound disability and survival without major disability. The VICS 2005 cohort study was approved by the human research ethics committee at the Royal Women's Hospital, Melbourne, Australia.
We entered the five factors into the online calculator for each eligible infant in the VICS cohort. A quarter of the infants were double-entered into the NICHD calculator by a second investigator to check for consistency of data entry. The calculator generated the risk of six possible outcomes for all 114 infants. The predicted outcome was then compared with the actual outcome for each infant in the VICS 2005 cohort.
Performance of the NICHD calculator to predict outcome was assessed by constructing a receiver operating characteristic (ROC) curve for mortality in all infants to 2 years of age, and calculating the area under the curve (AUC) and its 95% CIs. Separate ROC curves were generated for mortality in inborn infants to 2 years of age, and for mortality in outborn infants to 2 years of age. The predicted rates of the combined outcomes of death or survival with major disability, death or survival with profound disability, and survival free of major disability from the NICHD calculator were also compared with the actual rates in inborn infants in Victoria. Data were analysed by SPSS V.17.0 for Windows (SPPS Inc, Chicago, Illinois, USA).
The baseline characteristics of the 114 eligible infants we entered into the NICHD online calculator are shown in table 1.
The accuracy of the NICHD calculator for predicting mortality for inborn infants, 22–25 weeks’ gestation and free of lethal malformations was excellent (figure 1). In the VICS 2005 cohort, 49 (49.5%) of the 99 inborn infants died, compared with the overall predicted mortality rate generated by the NICHD calculator of 47.1%. The AUC was 0.803 (95% CI 0.718 to 0.888; p<0.001). This was higher than the AUC for the NICHD study (AUC=0.753; 95% CI 0.737 to 0.769; p<0.001). However, the accuracy for predicting mortality was not as precise for outborn infants, (figure 1) with an AUC of 0.643 (95% CI 0.337 to 0.949; p=0.36).
The follow-up rate for survivors was 100%. The calculator overestimated the combined outcome of death or survival with major disability in infants in Victoria, at 24 months of corrected age. The predicted rate was 72%, compared with an actual rate of 60.5% (69/114) in infants in Victoria, overall. The proportions of live births who survived without major disability in the VICS 2005 cohort were 39.4% (39/99) of inborn infants, and 40.0% (6/15) of outborn infants. The predicted rate of death or profound impairment was 59%, compared with an actual rate of 52.6% (60/114) in VICS infants overall. The proportions of live births who either died or had profound disability were 53.5% (53/99) of inborn infants, and 46.7% (7/15) of outborn infants. We have not compared the rates of survival free of major disability or profound disability with similar data from the NICHD study, as the criteria for the outcomes were not identical in the two studies.
We have shown that the NICHD outcome estimator is a valid tool to predict the risk of mortality in infants born in a tertiary perinatal centre between 22 and 25 completed weeks’ gestation in Victoria, Australia. The close correlation between the predicted mortality rate and the actual mortality rate demonstrated that the NICHD outcome estimator could be used with confidence by staff in perinatal units in Victoria to calculate an individual infant's chance of death or survival with major disability, despite the observation that there were only 99 inborn infants in the Victorian cohort compared with 4446 in the NICHD study.
One potential limitation of the NICHD outcome estimator is that the calculations are based on the outcomes of infants who were all inborn in a tertiary perinatal centre with a neonatal intensive care unit (NICU), and the NICHD acknowledge that the estimated outcomes generated by the online calculator may not be generalisable to infants born in non-tertiary centres (outborn infants).
We found that the NICHD outcome estimator was not as accurate for predicting mortality in outborn infants in Victoria. However, there were only 15 outborn infants (13%) in the VICS 2005 cohort with all five criteria required for the NICHD calculator. With such a small sample size, we cannot make any generalisations regarding the sensitivity and specificity of the outcome estimator for outborn babies.
Recognising that survival cannot be accurately predicted using gestational age alone, researchers at the NICHD developed the outcome estimator by combining gestational age with four other factors proven to affect survival: namely birth weight, gender, plurality and maternal antenatal corticosteroid exposure.6 All five factors can be quickly assessed at birth and in some instances, before birth, to estimate an infant's risk of death, or survival with major disability. The web-based tool is publicly accessible via the NICHD web site and is quick and simple to use.5
The NICHD outcome estimator has been shown to be the most accurate tool available to clinicians in tertiary hospitals in the USA to estimate the likelihood that an extremely preterm inborn infant, free of lethal malformations, will survive if intensive care is provided, and to estimate the chance of neurodevelopmental impairment in survivors.6 ,11
Traditional methods of predicting an infants’ chance of survival, such as assessing the infant's condition at birth, letting the infant ‘declare’ himself/herself following resuscitation or reliance on the 1-min and 5-min Apgar scores, have all been shown to be prone to error.12 ,13
A large study by Lee et al11 reported that the NICHD outcome estimator was accurate for estimating survival to discharge in both inborn and outborn infants. This study reported on 4691 infants who were born in one of 128 hospitals in California; 33% were inborn at a tertiary perinatal centre, and 67% were outborn. Of the outborn infants, 8% were outborn in an intermediate NICU (equivalent to a Level 2 nursery in Victoria), 52% in a community hospital (equivalent to a Level 1 nursery in Victoria), and 7% of the birth hospitals were unclassified. Lee et al reported that the NICHD outcome estimator more accurately predicted survival to discharge than using gestational age alone in both inborn and outborn infants.11 The authors stated that, based on these results, clinicians in non-tertiary centres can ‘rely on the outcome estimates generated by the NICHD estimator’ (Lee et al, 2010, p. e648). Based on the results reported by Lee et al, we felt it was relevant to attempt to validate these findings for outborn infants in Victoria, Australia.
Although the NICHD model predicts death and/or survival with neurodevelopmental impairment to 18–22 months’ corrected age, the study by Lee et al did not report outcomes beyond hospital discharge. Our study looked at both mortality and neurodevelopmental impairment in outborn infants to 2 years of age, whereas Lee et al, only reported mortality to the time of discharge from hospital. While our results were not consistent with those of Lee et al, we acknowledge that our results were limited by our sample size.
We also wanted to determine if the NICHD outcome estimator would accurately predict the number of infants, 22–25 weeks’ gestation, who would die or survive with a major disability. The risk of the combined outcome of death and/or severe disability for the infants born in Victoria was calculated to be 72%. The actual rate in infants from the VICS 2005 cohort was 60.5%. This combined rate of death, and/or severe disability for infants in Victoria, was similar to that reported in New South Wales between 1998 and 2001, so we were confident that our results were a true reflection of Australian norms in 2005.1 ,14 It is unlikely that differences in healthcare between Australia and the USA are the explanation for the differences in long-term outcome between countries; rather it is differences in how the subjects have been identified. As our study was population-based, capturing all 22–25-week gestation infants in the State of Victoria, we were not reporting on a skewed sample of subjects at higher risk of adverse long-term outcomes as might have been seen if only infants from selected NICUs had been reported, which is the case with the NICHD study.
The use of the NICHD calculator in clinical practice in Victoria
In a tertiary perinatal centre, parents will usually have the opportunity for counselling by a neonatologist who has ready access to outcome data from the hospital's NICU. In a non-tertiary hospital, the role of counselling the parents may fall to a clinician with limited experience in the care of extremely preterm infants.
Mortality and morbidity data specifically for outborn infants are not readily available. NICU outcome data are more accessible, but quoting the survival and long-term disability rates of inborn premature infants to the parents of an outborn premature infant is not accurate, or appropriate. Having internet access to a reliable tool to generate possible outcomes for extremely preterm outborn infants would therefore be extremely valuable for perinatal healthcare providers in non-tertiary hospitals in Victoria.
The NICHD outcome estimator provides clinicians with a valid and reliable tool that can be used before, or at birth, to predict the risk of death for inborn infants, 22–25 completed weeks’ gestation, born in a geographical area outside the USA. This information can be used to guide clinicians and parents to make informed decisions regarding the appropriateness of withholding or providing intensive care to infants born at the borderline of viability. While we are yet to demonstrate the validity of the tool for predicting mortality in outborn infants, the NICHD outcome estimator does provide clinicians in non-tertiary hospitals with more accurate estimates of outcome than using gestational age alone.
The NICHD calculator also provides an estimate of long-term neurodevelopmental outcome, ranging from intact survival through to moderate to severe, or profound disability. As the rate of serious neurodevelopmental impairment is overestimated in infants in Victoria by the NICHD outcome estimator, caution is needed if using the calculated outcome data to counsel parents regarding these risks. We recommend that local data be used for this purpose.
RAB is the recipient of the Felix Meyer Faculty Research Scholarship from The University of Melbourne, and receives funding from the Clinical Centre for Research Excellence in Newborn Medicine (National Health and Medical Research Council (NHMRC) Grant No. 546519). JAD is the recipient of a National Health and Medical Research Council (NHMRC) postdoctoral fellowship. PGD and LWD hold an Australian NHMRC Programme Grant No. 606789. Also funded in part by NHMRC Project Grant No. 454413, and the Department of Innovation, Industry and Regional Development (Victorian Government).
Participants: (in alphabetical order) Peter J Anderson, Catherine Callanan, Elizabeth Carse, Margaret P Charlton, Mary-Ann Davey, Noni Davis, Lex W Doyle, Julianne Duff, Cinzia de Luca, Marie Hayes, Esther Hutchinson, Elaine Kelly, Marion McDonald, Gillian Opie, Gehan Roberts, Michael Stewart, Andrew Watkins, Amanda Williamson & Heather Woods.
Contributors RAB reviewed the literature, designed the study, entered the data provided by LWD and wrote the manuscript. LWD designed the study, provided the VICS data, analysed the data and edited the manuscript. JAD designed the study, checked the data entry and edited the manuscript. PGD designed the study and edited the manuscript.
Financial disclosure None.
Competing interests None.
Ethics approval The cohort study was approved by the Human Research Ethics Committee at The Royal Women's Hospital.
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.