If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
Reprints: Haresh Kirpalani, BM, MSc, Division of Neonatology, Children's Hospital of Philadelphia, 34th St. and Civic Center Blvd., Philadelphia, PA 19104
The objective of the study was to assess whether recent data reporting survival of preterm infants introduce a bias from the use of varying denominators.
Study Design
We performed a systematic review of hospital survival of infants less than 1000 g or less than 28 weeks. Included publications specified the denominator used to calculate survival rates.
Results
Of 111 eligible publications only 51 (46%) specified the denominators used to calculate survival rates: 6 used all births, 25 used live births, and 20 used neonatal intensive care unit admissions. Overall rates of survival to hospital discharge ranged widely: from 26.5% to 87.8%. Mean survival varied significantly by denominator: 45.0% (±11.6) using a denominator of all births, 60.7% (±13.2) using live births, or 71.6% (±12.1) using used neonatal intensive care unit admissions (P ≤ .009 or less for each of 3 comparisons).
Conclusion
Variations in reported rates of survival to discharge for extremely low-birthweight (<1000 g) and extremely low-gestational-age (<28 weeks) infants reflect in part a denominator bias that dramatically affects reported data.
Advances in perinatal care, such as the use of antenatal steroids, surfactant therapy, and delivery room temperature control, have improved the rates of survival to discharge for extremely low-birthweight (ELBW) and extremely low-gestational-age (ELGA) infants.
Part of this variation might be due to center differences with respect to antenatal care and choices in active resuscitation or variations in baseline risk because of population differences.
Moreover, additional variation may arise from differences in definitions. For example, different definitions of still and live births in many developed countries affect overall national mortality rates.
Further variation stems from how centers calculate and report survival rates because several different denominators can be used in the calculation. These differences lead to bias in reported survival rates of extremely premature infants, often referred to as a selection bias or a denominator bias.
A systematic bias in reporting of survival rates could dramatically affect understanding of families at the time of counseling about potential survival of their premature infant and prevents accurate comparisons between centers. In a systematic review, we evaluated whether the degree of variation in survival rates in publications over the period 2000-2010 might reflect a denominator bias.
Materials and Methods
Primary outcome
Our primary outcome was the survival rates of ELBW and ELGA infants calculated with different denominators. We defined survival as the proportion of infants surviving until hospital discharge. The denominators used in these studies were: all births including stillbirths, live births, or neonatal intensive care unit (NICU) admissions.
Search for studies
The MEDLINE, EMBASE, PubMed, and Cochrane Library databases were searched using a combination of the following subject headings (MeSH) and free text (text word): infant, premature; or infant, very low birthweight; or infant, extremely low birthweight and neurodevelopmental impairment; or neurodevelopmental outcome; or developmental outcome; or survival; or infant mortality. No language restriction was applied. All potentially relevant titles and abstracts were retrieved and assessed for eligibility by 2 independent observers, whose disagreements were resolved by consensus. The reference lists of relevant articles were reviewed, and relevant citations were retrieved if they had not been obtained in the primary search. Reference lists of reviews, editorials, commentaries, and letters were also reviewed and retrieved if relevant.
Eligibility criteria
Publications were selected for inclusion if they: (1) were published between January 2000 and June 2010; (2) included susceptible infants born after January 1990, either with birthweight less than 1000 g or gestational age less than 28 weeks; (3) reported survival to hospital discharge; (4) specified the population denominator used to calculate survival rates: all births, live births, or NICU admissions.
Data extraction
For each publication, 2 independent searchers extracted the reported numerator (number of infants surviving until hospital discharge) and the reported denominator. This denominator was the total number of infants born (including stillbirths), the total number of infants born alive, or the total number of infants admitted to the NICU. Publications were grouped by the denominator used to report survival. Publications that reported survival outcomes by more than 1 denominator were included only once by randomly selecting 1 of the denominators to avoid dependence between the groups.
The period during which the infants were born was also extracted. Finally, we planned to perform adjusted analyses for which we collected potentially confounding baseline population characteristics such as maternal age, race, education, marital status, infant sex, country of birth, outborn/inborn status, Score for Neonatal Acute Physiology (SNAP) or Clinical Risk Index for Babies (CRIB) severity of illness scores, and the use of antenatal steroids. Disagreements were resolved by consensus.
Statistical analysis
One-way analysis of variation was used to test the hypothesis that estimates of survival differ by the denominator used to calculate survival rates. Analyses were completed using STATA/IC 10.0 software (Stata Corp, College Station, TX). P < .05 was considered statistically significant.
We performed subgroup analysis of differences in survival by overall birthweight less than 1000 g only and by gestational age less than 28 weeks only. A Bonferroni correction for multiple comparisons was used in these subgroup analyses, making a P < .008 statistically significant.
Results
Eligible studies
Two searchers achieved very good agreement on the inclusion and exclusions of studies (κ = 0.8) and on the extraction of data (κ = 0.85). All disagreements were resolved.
We identified 369 potentially relevant articles. Of these, 111 publications reported single-center or multicenter outcomes of infants less than 1000 g or less than 28 weeks born after 1990. Only 51 publications could be included because they reported both survival rates and the denominator used to calculate survival.
These studies included a total of 75,322 infants. Six publications (11.8% of all eligible studies) reported outcomes as a percentage of all births (Table 1). They represented 10,130 infants (13.4%). Twenty-five publications (49.0% of eligible studies) reported outcomes as a percentage of all live births (Table 2). They represented 49,289 infants (65.4%). Finally, 20 studies (39.2% of eligible studies) reported survival rates of infants admitted to the NICU (Table 3). They represented the remaining 15,903 infants for whom outcomes were available (21.1%).
TABLE 1Characteristics of studies reporting outcome of all births
NICHD Neonatal Research Network Follow-Up Study Neurodevelopmental outcomes of extremely low birth weight infants <32 weeks' gestation between 1993 and 1998.
EPIPAGE Nord-Pas-de-Calais Study Group Factors associated with neurodevelopmental outcome at 2 years after very preterm birth: the population-based Nord-Pas-de-Calais EPIPAGE Cohort.
National Institute of Child Health and Human Development Neonatal Research Network Changes in neurodevelopmental outcomes at 18 to 22 months' corrected age among infants of less than 25 weeks' gestational age born in 1993-1999.
Group tDES The Danish national study in infants with extremely low gestational age and birthweight (the ETFOL study): respiratory morbidity and outcome.
Vermont Oxford Network ELBW Infant Follow-Up Study Group Neurodevelopmental outcome of extremely low birth weight infants from the Vermont Oxford Network: 1998-2003.
Mortatlity and neurologic, mental and psychomotor development at 2 years in infants born less than 27 weeks' gestation: the Leiden Follow-Up Project on Prematurity.
National Institute of Child Health and Human Development Neonatal Research Network Intensive care for extreme prematurity—moving beyond gestational age.
Neurodevelopmental and functional outcomes of extremely low birth weight infants in the National Institute of Child Health and Human Development Neonatal Research Network, 1993-1994.
Outcome of extremely premature infants at early school age: health-related quality of life and neurosensory, cognitive, and behavioral outcomes in a population-based sample in northern Germany.
Changes in mortality, morbidity and early neurodevelopmental outcomes among extremely-low-birth-weight (ELBW) infants born in the early and recent post-surfactant era.
Beneficial effects of breast milk in the neonatal intensive care unit on the developmental outcome of extremely low birth weight infants at 18 months of age.
Only six studies (11.8% of eligible studies) reported their data using more than 1 denominator: 3 reported survival as a percentage of all births and all live births, 2 reported survival as a percentage of live births and NICU admissions, and 1 reported data with all 3 denominators. Data for each of these studies were included in the main analysis only once.
Survival to hospital discharge
There was a marked variation in survival rates. If the choice of denominator is ignored, overall survival to hospital discharge ranged from 26.5% to 87.8%. The range becomes narrower when the data are grouped by denominator. Using all births as a denominator, survival ranged from 28.2% to 61.0%, with a mean survival of 45.0% (±11.6). However, if the denominator was all live births, survival ranged from 26.5% to 81.8%, with a higher mean survival of 60.7% (±13.2). In contrast, using NICU admission as the denominator, survival ranged from 45.1% to 87.8%, with a mean of 71.6% (±12.1). The differences between the 3 group means are statistically significant (Figure). Survival rates were significantly higher because the denominator used a more narrow representation of the patient population (NICU admissions greater than all live births greater than all births). Differences in survival could not be attributed to differences in year of birth between the 3 groups (P = .83).
The lower and upper margins of each box represent the 25th and 75th percentile. The line intersecting each box represents the mean. The 2 tails of each box represent the SD.
Guillen. Bias in survival outcomes. Am J Obstet Gynecol 2011.
When only infants with birth weight less than 1000 g were included in the analysis, a significant difference in survival rates was found between those that used a denominator that included all births vs the denominator of NICU admissions (P = .002). When the reports specifying data for infants less than 28 weeks were analyzed, a significant difference in survival rates again was found between denominators that included all births and NICU admissions (P = .001). It was not possible to analyze survival rates between groups using narrower birthweight or gestational age intervals because not enough studies reported their data using this breakdown. Because of inconsistent and incomplete reporting of important maternal and infant characteristics, it was impossible to adjust our analyses for these potentially confounding factors. There was also inconsistent and incomplete reporting on the inclusion or exclusion of infants with major malformations; however, most publications excluded infants with major malformations from their overall survival rate.
Comment
This systematic review illustrates that the choice of population denominator leads to significant variation in reported survival rates of extremely low-birthweight (<1000 g) and extremely low-gestational-age (<28 weeks) infants. This potential bias was previously noted by Evans and Levene
in reports from 1978 to 1998. Our data show that in the era 2000-2010, wide differences in reporting still exist, which complicate the interpretation of variations in regional or hospital outcomes.
It is not surprising that survival rates are a function of how they are calculated. If survival rates are reported using a denominator of admission to the NICU, the outcome of infants who died during labor or who could not be resuscitated is ignored, consequently overestimating the survival rates. Evans and Levene
first labeled this phenomenon as a potential selection bias and showed that survival rates can be exaggerated by as much as 100% at 23 weeks' gestation when only a denominator of NICU admissions is used. Evans and Levene
argued that direct comparisons of neonatal outcomes could not be made in the setting of unclear definitions. Despite these concerns, marked variation in the denominators used to calculate survival persists into the era of this current study: 2000-2010.
The use of a specific denominator is important because each provides different, clinically useful information. We suggest that the single unspecified term, survival rate, should not be used to describe all of them without qualification. Each denominator addresses a different clinical question that can be useful for physicians who counsel parents at different time points. When the denominator is all births, it describes the chances of an ELBW infant surviving the delivery. Survival calculated with the denominator of live births describes the probability that a baby born alive will survive the initial resuscitation.
Together these survival statistics give a more complete picture and are likely best used in the antenatal period of counseling. When the denominator used is that of all NICU admissions, the derived survival statistic describes the probability that an infant who survived the initial resuscitation and was admitted to the NICU will leave the hospital. This statistic is best used to inform parents of the likely survival of their infant once in the NICU. Between each of these time points is a hurdle the infant must overcome. Each denominator gives clinicians and parents an idea of prognosis at each hurdle.
If clinicians are to use published outcome data to counsel parents about how to proceed when they are facing delivery of an ELBW or ELGA infant, accurate and unbiased outcome data are necessary. Such distinctions are also important for epidemiological considerations and for center comparisons. Comparisons between centers cannot legitimately be made without clear identification of the intake population. Yet the literature does contain comparison of survival rates between centers despite differences in the denominator used by each center.
It is relevant then that 60 of the 111 eligible publications in this review did not clearly report the denominator they used.
The strength of this study is that we demonstrate the magnitude of a continuing lack of clarity in current reports of neonatal survival outcomes and show that survival rate, a nonspecific single term, is really a range of rates that are derived from different denominators. Furthermore, we highlight the fact that less than 50% of the studies reviewed reported the denominator that was used; hence, there is no way of knowing what type of survival was described. Currently also, population-based cohorts that may aim to describe similar data vary in how they define their populations, whether by gestational age or birthweight. Because gestational age rather than birthweight better reflects the adaptation to a postnatal environment, this is also potential for lack of clarity. We suggest that obstetric journals and newborn journals adopt consistent and standard reporting guidelines.
To have a complete understanding of the data that are reported, each denominator should be clearly defined. Instead of using the term, survival rate, indiscriminately, perhaps a different, longer term should be used for each population. For example, one could use the term, survival rate of babies admitted to the hospital, for a cohort to differentiate it from rates that include only inborn patients.
The limitations of our study arise from the sparse published information on the underlying denominators and inconsistency in reporting of baseline variables, making adjustments for known risk factors impossible. Nonetheless, more than half of the studies identified in our initial search could not be included because they did not clearly specify a denominator.
Wide variations in survival within and between each denominator group were seen, likely reflecting differences between cohorts that cannot be accounted for by a denominator. Such variation has been previously been well described by Horbar and coworkers
using the data from the Vermont Oxford Network. Possible causes include variations in baseline risk in different populations; variation in use of effective antenatal therapies such as corticosteroids and postnatal therapies such as surfactant; differences in use of therapies not known to be effective; variations in withdrawal of care practices or nonescalation of care practices in ELBW infants; or even inadequate recording or improper classification of births.
Baseline information about published cohorts was not consistently reported. Without these data, comparisons using potential confounders could not be made and adjusted analyses could not be performed. This again suggests that journals publishing newborn survival papers should standardize reporting criteria for outcome studies. Even then, cross-country comparisons may be difficult.
We included publications from a wide variety of countries in this analysis, although most cohorts originated in the United States. Not only does the choice of denominator remain important in this context, but also 2 other issues arise. First, in both Europe and the United States, there are different definitions for birth registration and classification of stillbirth that will also affect the statistics reported.
Moreover, different attitudes toward resuscitation at the lower gestation limits will make the direct comparisons of survival rates between countries difficult.
EPIPAGE Nord-Pas-de-Calais Study Group Factors associated with neurodevelopmental outcome at 2 years after very preterm birth: the population-based Nord-Pas-de-Calais EPIPAGE Cohort.
Differences in the denominator used to define survival remain an important source of variation in reported survival rates of extremely low-birthweight infants. Physicians involved in counseling parents facing delivery of an ELBW infant must be aware of the presence of this potential denominator bias in published reports to avoid giving parents overestimated rates of survival. Finally, to minimize potential for bias and confusion, journal editors and reviewers should consider that denominators used to calculate study results are clearly described or subscribe to a standardized reporting practice.
References
El-Metwally D.
Vohr B.
Tucker R.
Survival and neonatal morbidity at the limits of viability in the mid-1990s.
Mortatlity and neurologic, mental and psychomotor development at 2 years in infants born less than 27 weeks' gestation: the Leiden Follow-Up Project on Prematurity.
Neurodevelopmental and functional outcomes of extremely low birth weight infants in the National Institute of Child Health and Human Development Neonatal Research Network, 1993-1994.
Outcome of extremely premature infants at early school age: health-related quality of life and neurosensory, cognitive, and behavioral outcomes in a population-based sample in northern Germany.
Changes in mortality, morbidity and early neurodevelopmental outcomes among extremely-low-birth-weight (ELBW) infants born in the early and recent post-surfactant era.
Beneficial effects of breast milk in the neonatal intensive care unit on the developmental outcome of extremely low birth weight infants at 18 months of age.
Cite this article as: Guillen Ú, DeMauro S, Ma L, et al. Survival rates in extremely low birthweight infants depend on the denominator: avoiding potential for bias by specifying denominators. Am J Obstet Gynecol 2011;205:329.e1-7.
00660-0&id=gr1.jpg){kind=link}