 |
|
 |
|
 |
 |
|
 |
 |
 |
 |
 |
 |
 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
,
* Department of Pediatrics
Center for the Evaluative Clinical Sciences and Department of Community and Family Medicine, Dartmouth Medical School, Hanover, New Hampshire
Institute for Clinical Evaluative Sciences, and Department of Health Policy, Management and Evaluation, Faculty of Medicine, University of Toronto, Toronto, Ontario, Canada
 |
ABSTRACT |
|---|
 TOP ABSTRACT METHODSRESULTSDISCUSSIONREFERENCES |
|---|
Methods. Linking the United States 1998 singleton birth cohort (N = 3.8 million) with county and health care characteristics, we conducted a small area analysis of LBW across 246 regions of neonatal health services. We measured observed rates and then used a multivariable, hierarchical model to estimate adjusted LBW rates by regions. We then stratified these rates by race for the 208 regions with adequate sample size.
Results. Observed LBW rates varied across regions from 3.8 to 10.6 per 100 live births (interquartile range: 5.0–6.8 [25th–75th percentile]; median: 5.9). After controlling for known maternal and area risk factors, 67 (27.0%) regions had rates significantly below and 98 (39.8%) regions had rates significantly higher than the national rate of 6.0 per 100 live births. Although black mothers were more likely to give birth to an LBW newborn, regional adjusted rates still varied >3-fold within both black and nonblack subgroups.
Conclusions. After controlling for known maternal and area risk factors, LBW rates markedly varied across US regions of neonatal health services for both black and nonblack mothers. Additional analyses of these regions may provide opportunities for regional accountability in pregnancy outcomes, LBW research, and targeted improvement interventions, especially in high-risk populations.
Key Words: infant low birth weight • small area analysis • health services research
Abbreviations: LBW, low birth weight • NICR, Neonatal Intensive Care Region • NSA, Neonatal Intensive Care Region
Despite improvements in perinatal outcomes over the past decades, low birth weight (LBW), defined as a birth weight <2500 g, remains a major problem in the United States. LBW is the most prevalent and dominant risk factor for infant mortality and childhood developmental disorders, making it an important target for improvement efforts. Research has shown that LBW results from interactions between the biological determinants of the mother and the fetus; the parent’s social milieu; and the effectiveness of medical care during the periconceptual, prenatal, and perinatal periods, but these interactions are complex and poorly understood. Furthermore, public health initiatives aimed at reducing LBW rates have been largely unsuccessful, and national rates have actually increased over the past decade, despite the goal of the Healthy People 2010 of 5.0 per 100 live births, a 30% reduction.1–6
Although LBW is a national problem, its burden falls unevenly across communities and their perinatal populations.7–15 For example, black infants have LBW rates twice that of nonblack infants, and this persistent difference is believed to be 1 factor that is fundamental to longstanding racial disparities in child health.2,16 Much less is known about the contribution of place, such as the region of maternal residence and health care delivery, to newborn outcomes. Research on regional rates of LBW serves 3 purposes. First, it describes the extent of variation in rates of LBW. Second, research can identify regions that have significantly high or low adjusted LBW rates. Regions with adjusted rates lower than observed are performing better than expected and may serve as benchmarks for others to emulate. Regions that do relatively poorly would be places to direct additional research and prevention efforts. Finally, research on area variation in LBW rates can examine the interaction between individual characteristics, such as race, and community and health system factors.
Although the identification of high-risk populations has always been central to LBW prevention, there have been few studies of regional variation in LBW rates using the US birth cohort,17–21 and none that has used regions that reflect neonatal health care delivery. In this study, we link the US singleton cohort to constructed areas of neonatal health care to measure the extent of regional variation in LBW rates.22 We compute regional observed rates, and, using a multivariable, hierarchical model, we estimate regional adjusted rates. Any remaining variation in adjusted rates cannot be explained by measurable individual and community variables. To clarify further the contribution of region of maternal residence, we stratify the analyses by maternal race to compare the relative influence of these factors.
|
METHODS |
|---|
|
TOPABSTRACTMETHODSRESULTSDISCUSSIONREFERENCES |
|---|
Overview of Units of Analysis and Observation
This study describes variation at a regional level while controlling for individual and subregional characteristics in the statistical models. The unit of analysis was the maternal–newborn dyad, and the exposure of interest was the region of neonatal services, called the Neonatal Intensive Care Region (NICR). This is referred to as multilevel modeling.24,25 By including individual and region-level variables, these models enable (1) interpretation of patient- and region-level factors, (2) optimal control for confounding at all levels, and (3) inference to the individual patient. The lowest level of geography was the county, the geographic identifier present on each maternal–newborn record. The next level was Newborn Service Areas (NSAs; N = 1601), which are county aggregates, and the highest level was NICRs (N = 246), which are NSA aggregates. Area characteristics were measured at the county level, with the exception of the supply of obstetricians, which was measured at the level of the NSA. Observed and adjusted LBW rates were summarized and presented by NICRs.
Definition of Regions of Neonatal Health Services
Using traditional methods of small area analysis, we used the NSA (n = 1601) and NICRs (n = 246) as units of observation for measuring neonatal health services and the occurrence of LBW. (The definitional methods are described in detail elsewhere.26,27) NSAs and NICRs, aggregates of contiguous counties, delineate relatively discrete areas of perinatal care and neonatal health care resources. These service areas are large enough to make stable comparisons but small enough for planning, and they do exhibit large variations in outcomes. NSAs are county groupings whose expectant mothers gave birth primarily in hospitals within that area. NICRs are NSA groupings whose very low birth weight infants (<1500 g) received most of their health care from NICUs within that area. For this analysis, the NICRs had a wide range in number of births (minimum: 1322; median: 10149; maximum: 154201) and LBW births (minimum: 54; median: 591; maximum: 8038), as well as in the number of LBW births to black mothers (minimum: 0; median: 129; maximum: 3573).
Individual Characteristics
Models included individual characteristics that were shown in previous research to influence LBW. Variables from Vital Statistics included maternal age (<20 years, 20–34, and 
35), race (black, white, or other), Hispanic ethnicity (yes, no, or unknown), completed high school (yes, no, or unknown), marital status (married or single), live birth order (first born, subsequent, or unknown), weight gain (
21 pounds, >21, or unknown), maternal birth place (foreign, United States, or unknown), tobacco and alcohol use (yes, no, or unknown), and timing of prenatal care (first trimester, later/none, or unknown). We defined maternal medical risk factors as being 1 or more of the following: cardiac disease, acute or chronic lung disease, diabetes, hydramnios or oligohydramnios, hemoglobinopathy, chronic or pregnancy-associated hypertension, previous preterm delivery, renal disease, Rh sensitization, or "other medical risk factors."4,26,28–33 Use of assisted reproductive technology is not recorded on vital records, making it unavailable for this study.34 However, we excluded infants from multiple births, significantly reducing the contribution of assisted reproductive technology in our study population.
Regional Characteristics
County-level characteristics reflect community risks and have been shown in previous studies to be associated with LBW.5,7,8,10–15,35 We used county median household income from the 2000 decennial Census and metropolitan statistical area designation as a proxy for county urban or rural status.36 We generated contextual variables from Vital Statistics for the county-specific percentage of teenaged and smoking mothers.23 Finally, we used county elevation data from the Area Resource File,37 categorized as 3000 ft, 3001 to 6500 ft, and >6500 ft, except for counties in Alaska and Hawaii and cities within Virginia. For Virginian cities, we assigned the value of the contiguous county. For Alaska and Hawaii, we used county airport elevation.38
Greater physician supply and access may influence birth weight.39 We developed regional physician supply measures from the 1996 American Medical Association and the American Osteopathic Association Masterfiles, which represent a census of US physicians irrespective of association membership.40 We defined adult primary care physician supply at the county level as the sum of internists and family practitioners divided by the total population. Rates of obstetricians per women aged 15 to 44 were calculated at the NSA level. We also included for each county the presence or absence of a maternal–fetal specialist.40
Statistical Analyses
We calculated overall crude rates by regions of neonatal health services (NICRs) with the number of LBW births as the numerator and the number of live births as the denominator. We also stratified by black and nonblack mothers, excluding regions with imprecise rates as a result of small numbers of black births (<100 births to black mothers; N = 38 regions).
To calculate regional adjusted rates, we used multivariable logistic regression analyses with individual birth as the unit of analysis.24,25,41 We first developed a model that included relevant individual- and area-level predictors (odds ratios and confidence intervals are presented in the Appendix Table). Seven variables of theoretical interest were excluded from this model because they were either collinear with median household income (county rates of child poverty, unemployment rates) or not influential in both bivariate and multivariable models (number of hospitals with obstetric services, neonatologists per birth, pediatricians per child 0–17 years, and the presence of a Title X or a fertility clinic). We adjusted for the effect of nesting of LBW infants within counties in this initial model by using overdispersed binary logistic models, clustered by county.24 The net effect of the clustering was to increase the width of the confidence intervals by 
2.1%. We used these variables to develop a second model to compute adjusted NICR rates by including indicator variables for each of the 246 NICRs and excluding the intercept term; all covariates were centered around the study population means so that the adjusted LBW rates would reflect an average member of the study population. For transforming the NICR-specific regression coefficients to the original scale, they were exponentiated and then calibrated so as to have the same overall mean as the crude NICR rates. We considered NICRs to have significantly higher or lower rates than the national mean when the 95% confidence interval for the adjusted rate did not include the national singleton rate of 6.0 per 100 live births. Analyses were performed using SAS version 8.2 (Cary, NC).42 This study was granted exempt status from the Dartmouth College institutional review board.
|
|
RESULTS |
|---|
|
TOPABSTRACTMETHODSRESULTSDISCUSSIONREFERENCES |
|---|
21 pounds during pregnancy (10.9), smoked tobacco (10.5), or drank alcohol (12.9). Not surprising, LBW was frequent in premature newborns (<37 weeks’ gestation, 36.7 per 100 live births).
|
|
|
Regional Variation in Adjusted LBW Rates
After controlling for known individual and regional characteristics using a multilevel model, LBW rates still varied widely across regions (Fig 2). Adjusted regional rates revealed a >3-fold difference, ranging from 3.4 to 11.1 per 100 live births (interquartile range: 5.3–6.9 [25th–75th percentile]; median: 6.2). The adjusted rates were significantly below the national rate in 67 (27.0%) regions and significantly higher in 98 (39.8%) regions. Many of the 67 regions with adjusted rates lower than the national rate were concentrated in the Pacific and Northeast regions. Conversely, most of the southeastern, south-central, and western areas had adjusted rates higher than the national rate.
|
|
DISCUSSION |
|---|
|
TOPABSTRACTMETHODSRESULTSDISCUSSIONREFERENCES |
|---|
There are 2 possible explanations for the regional variation in the adjusted rates found in this study. The first is that our models incompletely controlled for the underlying biological and social risks of LBW within regions. For example, populations may have different rates of maternal nutrition, periodontal disease, or bacterial vaginosis, perhaps contributing to differences in regional rates.5,43,44 As long as the biological and social causes of LBW remain inadequately understood, they will, of course, be inadequately measured.29 For these to explain the variation that we found, the risk factors that were omitted from the models would need to be strong predictors of LBW and be uncorrelated with the included variables. Although possible, this seems unlikely. Nevertheless, the identification of additional risk factors would be an important advancement in perinatology and would create new opportunities for improvement in newborn outcomes.
A second possible explanation of the variation in adjusted rates is that a region’s social and medical systems respond differently to the biological and social risk of its population. For example, more effective social welfare or health programs in some regions may lead to healthier women of childbearing years, even in regions with elevated individual risk.3 Differential access to contraception may lead to variations in the occurrence of unintended pregnancies, a factor that is known to be associated with LBW and other poor birth outcomes.45 Furthermore, the organization and the delivery of health services, including perinatal transport or the interaction of primary, secondary, and tertiary care for mothers and their newborns, may play important roles in regional rates of LBW. Some regions of the country with higher adjusted LBW rates, such as the Southeast, are also known to have higher rates of adult conditions, including stroke and kidney disease, warranting additional investigation.46 It is interesting that there are many regions in close proximity, even within states, that have low rates of LBW adjacent to regions with high LBW rates. These area pairs may be particularly useful for community studies of the health care provided to women of childbearing age.
The disparities in regional rates between black and nonblack infants are considerable and reflect consistent, unexplained elevated health risks for black infants. Racial disparities in perinatal health outcomes are well documented, although the relative contributions of biological, socioeconomic, and health care factors remain unknown.4,47–49 Irrespective of biological and social risk, several studies reveal health care utilization differentials between black and nonblack individuals.12,16,30,50,51 In addition to these disparities, this study shows that there are large variations of LBW rates within racial groups across these neonatal health care regions, suggesting that region of residence and perinatal care are as important as race, even if the mechanism is poorly understood.
The method that we use for studying regions provides an additional mechanism for national accountability in newborn outcomes, an unmet goal in perinatal policy.52 The 1976 document "Toward Improving the Outcome of Pregnancy" provided a conceptual framework to improve neonatal outcomes by endorsing the present hospital-based regionalized system of care. A second committee in 1993 (Toward Improving the Outcome of Pregnancy II) expanded their recommendations to emphasize the significance of preconception and perinatal care and accountability of outcomes.53,54 To date, no national process of measurement that provides regional, population-based, risk-adjusted perinatal outcomes has been implemented. Recent studies have used hospitals as the unit of analysis to document variation in neonatal mortality rates.55–58 Although these studies provide valuable hospital-specific outcomes, they lack, by design, a population-based analysis of the region where the prenatal and perinatal care is delivered. Other studies have compared LBW rates across institutions, counties, states, and specific regions, but they have not controlled for many known risk factors, and all but 1 lack a national perspective to appreciate fully the range of variation.2,17–21,56,59,60 Furthermore, whereas health policy decisions in the United States are often made at the state or national level, health service delivery is more local. The 246 health regions that were used in this study are sufficiently large to allow for stable comparisons of LBW rates, unlike most counties, yet are still small enough to reveal important variations and thus can be useful in regional profiling and in planning regional perinatal improvement efforts.61
There are several limitations to our study. Our study used a single-year cohort of newborns; future studies are needed to establish the stability of these findings over time. Also, information obtained from vital statistics, although encompassing the entire birth cohort, varies in quality and may vary by region. In addition, LBW is caused by many perinatal disorders, most commonly premature birth. Our findings may not apply to more specific causes of LBW. Furthermore, LBW is sometimes viewed as an outcome itself, although its primary value is as an easily measured characteristic associated with neonatal mortality and childhood morbidities. Repeating this study with neonatal or infant mortality as the dependent variable is technically feasible but would require multiple years of data for stable rates. Some have also argued that the LBW definition of <2500 g is arbitrary and advocate for population-specific birth weight standards.62–64 Although this is likely to be a more precise measurement to compare different groups across populations, our analyses are primarily concerned with regional performance within populations.
This study demonstrates that although the risks of LBW remain incompletely understood, the region of maternal residence and perinatal care are strongly associated with LBW rates. This method of estimating adjusted area rates of LBW offers a tool for initiating national accountability of LBW rates at the level of health services delivery. Identifying health service regions and communities with high or low LBW rates provides a new direction for future research on unknown and unmeasured risk and assists targeted efforts of interventions in a field in which progress has been elusive. Until efforts are made to understand and intervene in the causes of regional variation in rates of LBW, there is little chance that the United States will achieve the LBW goals of 5.0 per 100 live births of Healthy People 2010.
|
ACKNOWLEDGMENTS |
|---|
We are deeply grateful for the thoughtful insights of George A. Little, MD, on review of this article, as well as for the broader views given by Lorraine Klerman, PhD.
|
FOOTNOTES |
|---|
Address correspondence to Lindsay A. Thompson, MD, MS, 622 West 168th St, VC4-402D, Columbia University College of Physicians and Surgeons, New York, NY 10032. E-mail: lt2132{at}columbia.edu
No conflict of interest declared.
|
REFERENCES |
|---|
|
TOPABSTRACTMETHODSRESULTSDISCUSSIONREFERENCES |
|---|
This article has been cited by other articles:
 |
|
 |
|
  D. Fradin, S. Heath, J. Lepercq, M. Lathrop, and P. Bougneres Identification of Distinct Quantitative Trait Loci Affecting Length or Weight Variability at Birth in Humans J. Clin. Endocrinol. Metab., October 1, 2006; 91(10): 4164 - 4170. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
 |
||
|
||
Home | My Pediatrics | Journal Information | Current Issue | Past Issues | Subscriptions & Services | Contact Us Click here for faster international access © 2005 American Academy of Pediatrics. All rights reserved. |
||
|
||
