Children born at very low birth weight (VLBW) (<1500 g) are at increased risk for neurodevelopmental difficulties including low IQ and poor reading and math skills.1- 2 There is growing interest in examining other potential sequelae of being born at VLBW, such as additional health, education, and social consequences. One outcome of interest is social function, which includes communication, problem solving, play skills, and peer interaction. Such skills lay the foundation for school readiness.3- 4
Previous studies5- 7 suggest that lower family socioeconomic status is associated with lower social function and school readiness in VLBW children. The socioeconomic context of children's lives extends beyond the family to their neighborhood. Previous research8 suggests that low-birth-weight (LBW) (<2500 g) children living in poor neighborhoods have lower cognition at, but not before, age 3 years than LBW counterparts living in more advantaged neighborhoods. However, it has not been examined whether neighborhood disadvantage is associated with social function among children born at VLBW. Yet children born at VLBW may be especially sensitive to the physical, service, social, or economic characteristics of their neighborhood because they are at higher risk for health and developmental difficulties than other LBW children.
The availability of a statewide cohort of children born at VLBW allowed us to examine the relationship between neighborhood disadvantage and social function at ages 2 and 3 years among children born at VLBW. We hypothesized that living in a disadvantaged neighborhood is associated with worse social function at ages 2 and 3 years.
In addition to social function, health-related quality of life (HRQoL) is increasingly recognized as an important metric of the health and well-being of children, especially those at risk for developmental difficulties.9 Health-related quality of life is generally agreed to be a summary measure of multiple dimensions of physical, social, and psychological health10 and, for children, includes items such as not being able to keep up with peers, getting teased by other children, missing day care because of illness, and physical inability to pick up toys. Not only do children born at VLBW demonstrate lower HRQoL than their normal-birth-weight peers,11- 12 but their HRQoL and social function scores are correlated.13
Thus, a second aim of this study was to examine how the relationship between VLBW children's HRQoL and social function relates to neighborhood disadvantage. Children's HRQoL and social function are overlapping but separate constructs,7,14 where HRQoL captures not only function, but problems generated in the child's interaction with his or her surroundings. It is unclear how social function and HRQoL relate temporally13 and there remains a debate in the field regarding this issue.15 Therefore, we do not impose a temporal association between HRQoL and social function but rather aim to describe influence of neighborhood disadvantage on the correlation between social function and HRQoL. Living in a disadvantaged neighborhood may hamper children's HRQoL (because of adverse physical, biological, or social exposures), which may also negatively interfere with opportunities for optimal learning and social interactions with peers (social function). In addition, the association between HRQoL and social function may be different for children with different levels of neighborhood disadvantage (ie, neighborhood disadvantage is an effect modifier). That is, children with low HRQoL may be particularly vulnerable to the ill effects of living in a disadvantaged neighborhood.
The Newborn Lung Project is a regional cohort of infants born at VLBW in 2003 and 2004 and hospitalized in 1 of 16 newborn intensive care units in Wisconsin or near the state border. The original study recruited and collected clinical data on 1479 infants between July 1, 2003, and December 31, 2004. Of the 1479, 1296 survived to hospital discharge and 993 provided follow-up contact information and permission to be contacted. By age 2 years, 14 of the 993 died, leaving 979 to be contacted. Of the 979, 719 children had parent interview data collected when the child was between the ages of 2 and 3 years, or more specifically, 24 to 43 months. Of the 719, 50 observations were excluded because of missing data on social function for 26 (because of lack of a Spanish version of outcome measure, described later), maternal education for 2, and income for 22. The remaining 669 included only 13 Hispanic and 30 Asian, Hawaiian, or multiracial children. Because of these small numbers, we restrict our analyses to white and black non-Hispanic children (N = 626).
To address potential bias introduced by differential follow-up from birth, we used medical and sociodemographic covariates measured at birth (described later) to calculate weights reflecting propensity for follow-up and compare results with and without weights.
Social function at ages 2 and 3 years was measured using the Pediatric Evaluation of Disability Inventory,16 a functional assessment appropriate for children aged 6 months to 7 years. The social function domain contains 65 items scored by parents dichotomously as able to perform the task in most situations (1) or unable to perform the task in most situations (0). Raw scores were converted to a normalized standardized score (mean [SD] = 50 ). Reported reliability ranges from 0.74 to 0.96 across all subscales.17- 18 The Pediatric Evaluation of Disability Inventory discriminates between children with and without disabilities.19
Maternal education was categorized as less than high school, high school or equivalent, beyond high school, or college degree. Severity of neonatal morbidity was measured using the Score for Neonatal Acute Physiology II,20 an index ranging from 0 to 115 that comprises 6 physiologically based items (pH, urine output, blood pressure, temperature, oxygen requirement, and seizures) with higher scores reflecting more severe morbidity. Child's race and ethnicity were grouped as non-Hispanic white or non-Hispanic black (going forward, referred to as white and black). Concurrent annual family income was categorized as less than $10 000, $10 000 to less than $30 000, $30 000 to $60 000, and greater than $60 000. Birth weight was measured in grams. Number of siblings in the home was categorized as 0, 1, 2, or 3 or more. Maternal employment status was dichotomized as employed or unemployed. We also included sex of the child, family structure (child lives with 1 vs 2 biological parents), maternal age at birth (in years), and child's age (in months) (Table 1).
Children's HRQoL at ages 2 and 3 years was measured using the Pediatric Quality of Life Inventory.21 The Pediatric Quality of Life Inventory contains 21 items in 4 domains: physical, emotional, social, and school/day care. Parents are asked to rate, on a 4-point scale (0 = never a problem to 4 = almost always a problem), their perception, in the last month, of how difficult each task was for their child to complete. The Pediatric Quality of Life Inventory discriminates between children with and without chronic conditions and has a Chronbach α of .70.22 A total score (range 0-100) reflects the summation of reverse-coded responses across all subscales. Higher scores indicate higher HRQoL. We categorized HRQoL scores: (1) low if the score was 1 sample SD or more below the mean; (2) high if the score was 1 sample SD or more above the mean; or (3) otherwise middle.
A neighborhood disadvantage index was created using principal component analysis of 5 census tract sociodemographic variables23 (Table 2): percentage of families in poverty, percentage of households with income higher than the state median, percentage of women with bachelor's degree or more, percentage of single mothers, and percentage of mothers of young children unemployed. Principal component analysis24 is a data reduction technique that determines how to combine variables into a single score that captures as much as possible of the overall variability in all the variables and has been previously used in perinatal epidemiologic research.25 Specifically, the 5 census variables were standardized (after reverse coding percentage of women with bachelor's degree or more and percentage of households with income higher than the state median) following standard procedure. An overall neighborhood disadvantage score (mean [SD] = 0 ; α = .86) was created as an average of items weighted by the item loadings (whose elements measure the strength of the relationship between the variable and principal component). The linear index was then split into tertiles, and children were classified as living in either disadvantaged (highest third), middle advantage (middle third), or advantaged (lowest third) neighborhoods.
Our cross-sectional analysis estimated associations using a mixed linear regression model with a random intercept. This allowed us to adjust for nonindependence arising from the fact that children in the same neighborhood share measured and additional unmeasured neighborhood characteristics.26 The multilevel models included 444 neighborhoods (intraclass correlation coefficient = 0.64). In the first model, the unadjusted influence of neighborhood disadvantage on social function was examined. The second model additionally included child and family sociodemographic and health covariates. The third model additionally included children's HRQoL, and a final model included interaction terms between categories of neighborhood disadvantage and HRQoL (to test for effect modification). All continuous variables were centered at their mean. For each model, we report the estimated mean difference in social function and its 95% confidence interval associated with each covariate.
All analyses were conducted in SAS version 9.13.27 The institutional review board at University of Wisconsin–Madison and all participating centers approved this study. Parents of the study children provided written informed consent to participate.
Of the full sample (N = 626), 221 children were categorized as living in a disadvantaged neighborhood and 238, in advantaged neighborhoods (Table 1). Table 1 shows that mean social function scores were lowest in disadvantaged (mean [SD] = 43.8 [13.7]) and highest in advantaged (mean [SD] = 49.6 [13.7]) neighborhoods. High HRQoL score was 6 percentage points more prevalent in advantaged neighborhoods (76 of 238; 32%) than in disadvantaged neighborhoods (57 of 221; 26%).
Inclusion of propensity weights did not notably change estimates and we present the unweighted results. In adjusted analyses (Table 3), model 2 reveals that children living in disadvantaged neighborhoods had a 4.6-point lower social function score (mean difference, −4.60; 95% confidence interval, −8.4 to −0.8) than children living in the advantaged neighborhoods, even after adjusting for multiple individual- and family-level covariates.
Adding HRQoL (model 3) demonstrated that children with the lowest HRQoL had, on average, a 12-point lower social function score (mean difference, −12.28; 95% confidence interval, −15.3 to −9.3) than children with the highest HRQoL. Moreover, when including HRQoL, the mean differences for neighborhood disadvantage were no longer statistically significant (P = .08, test of joint significance), suggesting that both have common neighborhood influences.
Model 4 showed that neighborhood disadvantage was a statistically significant effect modifier (P = .04, test of joint significance) of the association between HRQoL and social function, suggesting that the association between HRQoL and social function is strongest for children living in disadvantaged neighborhoods.
Thus, according to models 2 and 3, 2- and 3-year-old VLBW children living in disadvantaged neighborhoods have worse HRQoL regardless of their social function or, conversely, have worse social function regardless of their HRQoL. For children with both a disadvantaged neighborhood and low HRQoL, social function is particularly low.
We found that VLBW 2- and 3-year-olds living in disadvantaged neighborhoods have lower social function than their peers living in less disadvantaged neighborhoods, even after adjusting for a host of child and family sociodemographic and health covariates. This finding differs from a previous study8 that found no neighborhood effect on IQ among slightly higher-birth-weight LBW children younger than 3 years. Our findings indicate that VLBW 2- and 3-year-olds may not be shielded from the negative influences of neighborhood disadvantage. For example, families with young children in disadvantaged neighborhoods may not have adequate access to and quality of neighborhood resources (eg, community health clinics, parks, libraries, community center playgroups, and day care/preschool) that could foster not only cognitive, but also social communication skills. Parents living in disadvantaged neighborhoods may avoid using neighborhood amenities (eg, parks and libraries) because of safety concerns. Future research should explore these potential mechanisms.
Another finding is that in VLBW 2- and 3-year-olds HRQoL and social function are closely related. The association between neighborhood disadvantage and social function seems to operate, at least in part, through mechanisms also captured in children's HRQoL. That is, living in a disadvantaged neighborhood is negatively associated with VLBW children's HRQoL, which is directly associated with social function. Theoretical models28- 29 suggest that disadvantaged neighborhoods can enhance the association between social function and children's developmental problems arising from their surroundings. Our findings are consistent with this interpretation, though longitudinal analyses will be needed in future research to confirm this chain of events.
Conversely, living in an advantaged neighborhood somewhat buffers (ie, modifies the effect of) the association between HRQoL and social function. The results of our study disentangle the associations between neighborhood disadvantage, HRQoL, and social function for children born at VLBW and may help guide future intervention studies for this population.
We acknowledge multiple limitations of this study. First, this is a cross-sectional analysis, which precludes making causal inferences. We cannot disentangle the direction of the relationship between HRQoL and social function. In fact, HRQoL and social function likely interact dynamically over the life course. Future research should use longitudinal analyses with multiple measures of HRQoL and social function to capture these associations over time.
There is no agreed-on measure of neighborhood disadvantage for early childhood development. A limitation of the index approach is that, by integrating all covariates into 1 measure, one cannot estimate the individual effects of each. However, most neighborhood researchers30 advocate for index measures to more fully capture the complex context of the neighborhood and to address the multicollinearity of separate neighborhood characteristics. Moreover, because families are not randomly assigned to neighborhoods, measures of neighborhood disadvantage may capture unmeasured family characteristics.
A review of pediatric HRQoL measures22 suggests a lack of well-defined, psychometrically sound tools. However, among the available measures, the Pediatric Quality of Life Inventory is reported to have excellent reliability, be sensitive to changes preintervention and postintervention, and be easily administered. In addition, HRQoL for children younger than 4 years is collected by parent report. This could introduce bias if parents differentially report HRQoL in ways that also relate to characteristics of where families live. Moreover, in this study, social function and HRQoL are biased in that they are both collected by parent report. Future follow-up studies should also include child-reported measures of HRQoL and social function assessments by clinicians.
About one-third of the original cohort was lost to follow-up. This has the potential to introduce bias if the children who were followed up differ from children not followed up on characteristics of interest. However, our analyses with and without the propensity score weights revealed no differences, which supports that minimal bias is introduced by differential follow-up.
Finally, the association between HRQoL and social function may indicate that they are measuring the same construct, that is, children's functional performance. If this were true, we might expect consistent associations between HRQoL and children's functional skills across varying functional subscales. However, the correlation between HRQoL and function is stronger for motor function than social function among toddlers born at VLBW.31 Health-related quality of life and motor function seem to capture similar skill sets (eg, locomotion, mobility, bathing, and carrying objects)32 and parents perceive motor function difficulties as problematic. Conversely, HRQoL and social function seem to capture related, yet not overlapping, constructs, and parents seem less likely to perceive social function difficulties as problematic.33 Thus, the weaker correlation between HRQoL and social function may suggest that the HRQoL has more of a contextual component, that is, it captures not only how limitations affect what the child can do, but also how other children react to these limitations.31,33
Given the emphasis in the field of disability research to include measures of function (ie, rather than impairments) and HRQoL, the results of our study contribute to the relatively sparse literature examining socioeconomic, health and well-being, and neighborhood correlates to social function in 2- and 3-year-olds born at VLBW. Second, we used multilevel methods to appropriately account for the clustering of children with neighborhoods and propensity score weighting to address potential biases introduced by missing data.
We found a significant association between living in a disadvantaged neighborhood and poorer social function among 2- and 3-year-olds born at VLBW. Moreover, the relationship between children's social function and HRQoL is different across levels of neighborhood disadvantage. That is, the ill effects of living in a disadvantaged neighborhood are particularly bad for children with lower HRQoL.
Correspondence: Beth Marie McManus, PT, MPH, ScD, University of Wisconsin–Madison, School of Medicine and Public Health, 707 WARF, 610 Walnut St, Madison, WI 53726 (firstname.lastname@example.org).
Accepted for Publication: July 29, 2010.
Published Online: October 4, 2010. doi:10.1001/archpediatrics.2010.185
Author Contributions: Dr McManus had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. Study concept and design: McManus, Albanese, Sadek-Badawi, and Palta. Acquisition of data: Albanese, Sadek-Badawi, and Palta. Analysis and interpretation of data: McManus, Robert, and Palta. Drafting of the manuscript: McManus, Robert, Albanese, and Palta. Critical revision of the manuscript for important intellectual content: McManus, Robert, Sadek-Badawi, and Palta. Statistical analysis: McManus and Palta. Obtained funding: Albanese, Sadek-Badawi, and Palta. Administrative, technical, and material support: Robert, Albanese, and Sadek-Badawi. Study supervision: Robert.
Financial Disclosure: None reported.
Funding/Support: This research was supported by grant HL038149 from the National Heart, Lung, and Blood Institute.
Additional Contributions: Dr McManus thanks the Robert Wood Johnson Health Foundation Health and Society Scholars program at UW-Madison for its support. This research would not have been possible without the dedicated efforts of Lisa Loder, Kathleen Madden, and Tanya Watson, who conducted parent interviews.
Country-Specific Mortality and Growth Failure in Infancy and Yound Children and
Association With Material Stature
Use interactive graphics and maps to view and sort country-specific infant and early
dhildhood mortality and growth failure data and their association with maternal
Thank you for submitting a comment on this article. It will be reviewed by JAMA Pediatrics editors. You will be notified when your comment has been published. Comments should not exceed 500 words of text and 10 references.
Do not submit personal medical questions or information that could identify a specific patient, questions about a particular case, or general inquiries to an author. Only content that has not been published, posted, or submitted elsewhere should be submitted. By submitting this Comment, you and any coauthors transfer copyright to the journal if your Comment is posted.
* = Required Field
Disclosure of Any Conflicts of Interest*
Indicate all relevant conflicts of interest of each author below, including all relevant financial interests, activities, and relationships within the past 3 years including, but not limited to, employment, affiliation, grants or funding, consultancies, honoraria or payment, speakers’ bureaus, stock ownership or options, expert testimony, royalties, donation of medical equipment, or patents planned, pending, or issued. If all authors have none, check "No potential conflicts or relevant financial interests" in the box below. Please also indicate any funding received in support of this work. The information will be posted with your response.
Register and get free email Table of Contents alerts, saved searches, PowerPoint downloads, CME quizzes, and more
Subscribe for full-text access to content from 1998 forward and a host of useful features
Activate your current subscription (AMA members and current subscribers)
Purchase Online Access to this article for 24 hours
Some tools below are only available to our subscribers or users with an online account.
Download citation file:
Web of Science® Times Cited: 1
Customize your page view by dragging & repositioning the boxes below.
Users' Guides to the Medical Literature
Table 9.2-2 Refuted Evidence From Studies of Physiologic or Surrogate Endpoints
All results at
and access these and other features:
Enter your username and email address. We'll send you a link to reset your password.
Enter your username and email address. We'll send instructions on how to reset your password to the email address we have on record.
Athens and Shibboleth are access management services that provide single sign-on to protected resources. They replace the multiple user names and passwords necessary to access subscription-based content with a single user name and password that can be entered once per session. It operates independently of a user's location or IP address. If your institution uses Athens or Shibboleth authentication, please contact your site administrator to receive your user name and password.