March 2017, VOLUME139 /ISSUE 3

Socioeconomic Attainment of Extremely Low Birth Weight Survivors: The Role of Early Cognition

  1. Kathleen G. Dobson, MSca,
  2. Mark A. Ferro, PhDb,
  3. Michael H. Boyle, PhDc,
  4. Louis A. Schmidt, PhDd,
  5. Saroj Saigal, MD, FRCP (C)e, and
  6. Ryan J. Van Lieshout, MD, PhD, FRCP(C)c
  1. aDepartments of Clinical Epidemiology and Biostatistics,
  2. cPsychiatry and Behavioural Neurosciences,
  3. dPsychology, Neuroscience, and Behaviour, and
  4. ePediatrics, McMaster University, Hamilton, Ontario, Canada; and
  5. bSchool of Public Health and Health Systems, University of Waterloo, Waterloo, Ontario, Canada
  1. Ms Dobson conceived the study and its design, analyzed and interpreted the data, and drafted the initial manuscript; Drs Ferro and Boyle helped to conceive the study and its design, aided in the analysis and interpretation of the data, and reviewed and revised the manuscript; Dr Saigal started the cohort, helped to conceive the study and its design, selected the data collection instruments, supervised data collection, aided in the analysis and interpretation of the data, and reviewed and revised the manuscript; Dr Schmidt helped to conceive the study and its design, selected the data collection instruments, supervised data collection, aided in the analysis and interpretation of the data, and reviewed and revised the manuscript; Dr Van Lieshout helped to conceive the study and its design, selected the data collection instruments, aided in the analysis and interpretation of the data, and reviewed and revised the manuscript; all authors participated in the conceptualization and design of the study, interpretation of data, revising the manuscript critically for important intellectual content; and all authors approved the final manuscript as submitted.

  • Ms Dobson's current affiliation is the Dalla Lana School of Public Health at the University of Toronto, Toronto, Ontario, Canada.


OBJECTIVES: To determine: (1) if childhood cognitive and academic abilities mediate the association between being born at extremely low birth weight (ELBW) and socioeconomic attainment at age 29 to 36 years; (2) which cognitive abilities (IQ, verbal abilities, fluid intelligence, mathematical abilities, or academic achievement) most strongly mediate this association; and (3) if the mediating role of cognition is different in ELBW survivors with significant neurosensory impairment (NSI).

METHODS: A prospective, longitudinal cohort of 100 Canadian ELBW survivors born between 1977 and 1982 and 89 normal birth weight comparison participants were used to examine the mediating role of childhood cognition by using 5 cognitive mediators assessed at age 8 years (overall IQ, verbal IQ, performance IQ, quantitative ability, and academic achievement) on socioeconomic attainment at adulthood. Socioeconomic attainment was defined as personal annual earnings and full-time employment assessed via self-report at age 29 to 36 years.

RESULTS: Mediation models revealed that childhood cognition mediated the association between ELBW status and income attainment, with mathematical abilities and overall IQ each accounting for 26% of the direct effect. Mediated effects were not statistically significant in full-time employment models. For both outcomes, the mediating effect of cognition was stronger for ELBW survivors with NSI.

CONCLUSIONS: Childhood cognitive abilities partially mediate associations between ELBW status and adult income attainment. Early life cognition is a critical predictor of socioeconomic attainment in ELBW survivors, particularly in those born with NSI. Interventions aimed at enhancing early cognition in ELBW survivors may help optimize their later socioeconomic attainment.

  • Abbreviations:
    confidence interval
    extremely low birth weight
    normal birth weight
    neurosensory impairment
    socioeconomic status
    Wechsler Intelligence Scale for Children-Revised
    Wide Range Achievement Test-Revised
  • What’s Known on This Subject:

    Little is known about the socioeconomic attainment of extremely low birth weight (ELBW) survivors or its determinants. Research suggests that enhanced childhood cognitive function is associated with greater adult socioeconomic attainment; however, this has never been examined in ELBW survivors.

    What This Study Adds:

    Childhood intelligence accounted for 26% of the association between ELBW status and personal earnings at age 30. Mathematical abilities, academic achievement, and verbal abilities also mediated this association. Mediation by childhood cognition was particularly strong among survivors with neurosensory impairment.

    More extremely low birth weight (ELBW; <1000 g) survivors are living into adulthood than ever before.1 However, very little is known about their socioeconomic attainment in adulthood, a critical marker of success and an important determinant of health.2 This information is crucial at an individual level and at a population level to model human capital, a function of the current health and its depreciation rate of the labor force in developed countries.3 As roughly 9% of all infants are born preterm,4 a significant proportion of the labor force will face perinatal adversities.

    Studies utilizing European registry data suggest that very low birth weight (<1500 g)5 and preterm individuals are at greater risk of lower earnings and unfavorable employment outcomes in adulthood6,7; however, the mechanisms linking perinatal adversity to lower socioeconomic attainment remain unclear. Research in general population samples suggests that enhanced early cognitive functioning is positively associated with educational attainment, income attainment, and labor stability in adulthood.810 Cognitive dysfunction is among the most common disabilities experienced by ELBW survivors,11 with up to 50% of extremely premature survivors suffering from some form of neurodevelopmental disability (ie, cerebral palsy or intellectual disability).12

    Only 1 study to date has examined the mediating role of cognitive abilities on socioeconomic outcomes in high-risk pediatric survivors. Basten et al13 reported that early cognitive abilities partially mediated the association between preterm birth (28 to 38 weeks’ gestation) and wealth at age 40. However, it is unclear how many ELBW survivors (generally born before 28 weeks’ gestation with greater perinatal risk) were included in their study. Secondly, their aggregated wealth estimate (consisting of family income, social class, housing tenure, employment status, and self-perceived financial situation) may be dependent on multiple contextual factors such as sex, marital status, or if the individual has children. Additionally, this may inadvertently obscure the intricacies of how different cognitive abilities influence individual socioeconomic outcomes.

    Using data from the oldest, longitudinal cohort of ELBW survivors, the current study explored 3 objectives to determine: (1) if childhood cognitive and academic abilities at age 8 mediated associations between being born at ELBW and socioeconomic outcomes (personal income attainment and full-time employment) at age 29 to 36; (2) which cognitive or academic ability (IQ, verbal abilities, fluid intelligence, mathematical abilities, or overall academic achievement) most strongly mediated this association; and (3) if the mediation pathway for ELBW survivors differed for those born with or without serious neurosensory impairment (NSI).



    The study cohort consisted of 100 ELBW survivors and 89 normal birth weight (NBW; ≥2500 g) comparison individuals born in central west Ontario, Canada. Between 1977 and 1982, 397 ELBW survivors were recruited at birth; 179 survived to hospital discharge (Fig 1).14 At 8 years, 143 ELBW survivors participated in cognitive assessments. At age 22 to 26 years, 142 survivors participated in sociodemographic and health assessments15; of these, 100 (70%) participated and provided socioeconomic information at age 29 to 36 years.1

    FIGURE 1

    Participant flowchart. This figure briefly highlights participation of ELBW survivors and NBW comparisons.

    At age 8, 145 NBW individuals group-matched for age, sex, and parental socioeconomic status (SES) were recruited from the Hamilton Public School System. At age 22 to 26, 133 NBW individuals participated in data collection and were eligible to participate in the latest data assessment. Of these 133 NBW individuals, 89 (67%) provided data on socioeconomic outcomes at ages 29 to 36. Written informed consent was received from all participants in adulthood and from their parents during childhood. This study received ethics approval from the McMaster University Health Sciences Research Ethics Board.

    Mediator: Childhood Cognitive Abilities

    Childhood cognitive and academic abilities were assessed at age 8 years; assessment procedures have been previously described.16 We examined overall intelligence (IQ), verbal abilities, fluid intelligence, mathematical abilities, and academic achievement as mediators, as they have been previously studied as mediators of adult wealth in preterm and general population samples.8,13 All cognitive tests have adequate psychometric properties.1722

    IQ, verbal abilities, and fluid intelligence were assessed by using the Wechsler Intelligence Scale for Children-Revised (WISC-R). The WISC-R consists of 10 subtests, each with a mean of 10 and SD of 3. Combining these subtests creates a performance IQ, an assessment of fluid intelligence evaluating visuospatial abilities, alertness to detail, and processing speed; and a verbal IQ, an assessment of reading, verbal, and language abilities.23 These 2 IQ scores are then combined to calculate an overall IQ (mean = 100, SD = 15), estimating overall intelligence.23

    Mathematical abilities were assessed by using the Wide Range Achievement Test-Revised (WRAT-R) arithmetic subscale. The subscale comprises questions pertaining to counting, reading numerical symbols, and solving mathematical problems through numerical computation; it has a mean of 100 and SD of 15.21,24

    The Woodcock–Johnson Psychoeducational Battery was used to assess academic achievement. Three subscales of reading (letter–word identification, work attack, and passage comprehension) were administered. The subscales are combined into a single score (mean = 100, SD = 15).25

    Outcomes: Personal Income Attainment and Full-Time Employment

    We examined annual personal income and full-time employment as socioeconomic attainment outcomes. These 2 variables were assessed by using standardized questions from the Ontario Child Health Study questionnaires.26 Personal annual income was assessed by summing the amount of Canadian income earned over the past 12 months from the following: wages and salaries before deductions, self-employment, employment-insurance benefits, child benefits at provincial and federal levels, social assistance, child/spousal support, and any other income sources such as dividends, interests, capital gains, and gratuities. The median total income in Canada in 2013 was approximately $32 020.27 Full-time employment was assessed by asking participants if they had been employed 30 hours or more per week for the past 12 months.


    We included variables associated with poor cognitive function, low birth weight, socioeconomic outcomes, and attrition as covariates within mediation models. These variables included childhood SES, sex, number of years of education completed, age, marital status, and dependent status.7,28 Childhood SES was assessed via reports by participants’ parents at age 8 by using the Hollingshead 2-factor index of social position.29 This index has 5 levels, where 1 indicates the highest SES level and 5 indicates the lowest SES level. Educational attainment was self-reported and calculated by summing the years of education each cohort member had successfully completed at the time of testing. Marital status was self-reported and defined as those who were married or living common-law with a partner versus those who were not. Dependent status was self-reported and defined as having at least 1 child versus no children.

    Statistical Analysis

    Statistical analyses were performed by using SAS version 9.3 (SAS Institute, Cary, NC). Descriptive statistics were examined by using independent sample t tests for continuous variables and χ2 tests for categorical variables. To determine the influence of attrition in our cohort, we compared participants and nonparticipants on variables used to match ELBW and NBW participants at age 8, and childhood cognitive abilities.

    To explore the mediating role of childhood cognitive and academic abilities on socioeconomic outcomes (objectives 1 and 2), we performed the product of coefficient mediation method.30 We performed single mediator bias-corrected bootstrap models (n = 10 000) to examine the mediating influence of each childhood cognitive factor (overall IQ, verbal IQ, performance IQ, mathematical abilities, and academic achievement) separately (Fig 2).30,31 In each model, the independent variable was birth weight status, and the dependent variable was either annual personal income or full-time employment status.

    FIGURE 2

    Single mediator model. This figure highlights the mediation models used for objectives 1 and 2.

    To address our first objective, the mediated effect was calculated by multiplying the a and b coefficients (indirect effect, ab) and computing its 95% confidence interval (CI) to determine statistical significance. To address our second objective, we qualitatively compared the magnitude of each mediated effect for significant models by calculating the proportion mediated for each model.30

    To attempt to address the influence of missing data in our models, we performed a multiple imputation analysis creating 10 imputed data sets. The average value for each continuous predictor from these 10 data sets was used to replace missing data in subsequent analyses. As the results from these mediation analyses did not statistically differ, we report results only for participants with complete data.

    When exploring associations between cognitive abilities and socioeconomic attainment in ELBW survivors, it is critical to account for this population’s high rate of NSI. Therefore, for our third objective we performed mediation analyses by using a multicategorical independent variable (Fig 3) and compared the magnitude of mediated effects in ELBW survivors with and without NSI.32 NSI were diagnosed at age 3 by developmental pediatricians and were defined as follows: cerebral palsy, mental retardation, blindness, deafness, or microcephaly.

    FIGURE 3

    Multicategorical independent variable mediator model. This figure highlights the mediation models used for objective 3.


    Sample Characteristics

    Table 1 contains the characteristics of the study participants. No differences were seen in current age, sex, childhood SES, or total years of education between ELBW and NBW participants. At age 8, ELBW survivors scored significantly lower in all cognitive domains compared with NBW participants (P < .001). ELBW participants reported annual personal incomes approximately $20 000 lower than NBW participants (P < .001). Nearly 77% of NBW participants and 62% of ELBW participants reported having full-time employment over the past year (P = .042). Twenty-six ELBW participants and 1 NBW participant had an NSI.

    TABLE 1

    Participant Characteristics

    When comparing the demographic characteristics of participants and nonparticipants (Table 2), male sex was a predictor of attrition in both ELBW and NBW groups. In ELBW survivors, lower childhood SES was a predictor of attrition. Nonparticipants generally had lower scores on childhood cognitive assessments compared with participants. It was found that cognitive scores in both participants and nonparticipants followed the same trend in that higher socioeconomic participants generally had higher cognition scores compared with those in lower socioeconomic strata, suggesting that nonparticipants were missing at random.

    TABLE 2

    Demographic Factors of Participants and Nonparticipants at Current Sweep

    Objective 1 and 2: Personal Income Mediation Models

    Of the 189 participants, 157 (83%) had complete data on all variables and were included in mediation models. In each mediation model, ELBW participants scored significantly lower on all cognitive assessments at age 8 compared with NBW participants (a pathway, P < .01). Higher scores on all early life cognitive measures were associated with higher reported personal income at age 30 (b pathway, P < .01), except for the WISC-R Performance IQ (P = .10).

    Significant indirect effects (ab) were seen in all models except for the WISC-R Performance IQ (Table 3), supporting the presence of mediation. Our results indicate that ELBW children with lower scores on these cognitive and academic measures in childhood reported lower annual incomes in adulthood. However, because ELBW status was still a significant predictor of income attainment in our models, cognitive abilities only partially mediated the association between ELBW status and income attainment.

    TABLE 3

    Bootstrapped Direct and Indirect Effects of Childhood Cognition and ELBW on Income Attainment and Full-Time Employment

    In addressing our second objective (Table 3), the largest mediated effects were seen by overall IQ, accounting for 26% (ab = –$4516, 95% CI: –$9810 to –$730) of the total effect (c = –$17 210) in ELBW survivors, and mathematical abilities, also accounting for 26% (ab = $–4424, 95% CI: –$9012 to –$1403) of the total effect in ELBW survivors. The smallest significant mediated effect was seen with the Woodcock–Johnson academic achievement score, accounting for 16% of the total effect between ELBW status and adult earnings.

    Full-Time Employment Mediation Models

    Of the 189 participants, 151 (80%) had complete data and were used in models. Our models suggest that childhood cognitive and academic abilities may not mediate the association between ELBW status and full-time employment (Table 3), because the indirect effect (ab) was not statistically significant in any model. When performing logistic regressions of full time on employment ELBW status, cognitive mediators, and covariates, the most significant predictor of full-time employment was sex. Compared with men, the odds of full-time employment were 3 times lower for women.

    Objective 3

    To address our third objective, we performed mediation models by using a multicategorical independent variable separating ELBW survivors with and without NSI (Table 4). Our results suggest that the mediating role of cognitive abilities on annual income attainment may be stronger in ELBW survivors with NSI because this group had larger indirect effects in every model. For example, overall IQ accounted for 24% (ab = –$4023) of the total effect (c = –$16 871) in ELBW survivors without NSI, but accounted for 37% (ab = –$6828) of the total effect (c = –$18 664) in ELBW survivors with NSI. This trend was seen with every cognitive mediator. For ELBW survivors without NSI, indirect effects for overall IQ, verbal IQ, and mathematical abilities were statistically significant. For ELBW survivors with NSI, indirect effects for overall IQ, mathematical abilities, and academic abilities were statistically significant. When exploring the mediating role of childhood cognition on full-time employment, the indirect effect estimates were not statistically significant.

    TABLE 4

    Mediation Effects for NSI Subgroup Models


    Our study suggests that childhood cognition partially mediated the association between being born at ELBW and socioeconomic attainment in the fourth decade of life. Single mediator models suggest that mathematical abilities and overall IQ had the strongest influence on this link, each accounting for 26% of the association between being born at ELBW and personal income attainment. However, given the large yet nonsignificant mediated effects, our results are inconclusive as to whether cognitive abilities mediate the association between ELBW status and employment status. In posthoc power analyses, it was observed that personal income models were sufficiently powered (β = .80), but full-time employment models were not (β = .52), limiting the ability to identify significant effects. Additionally, perhaps the ELBW survivors and NBW participants who reported having children were not working full time to support their family in the home. Therefore, this association warrants further study in larger samples.

    The direction and magnitude of our results align with those previously reported by Basten et al.13 In their study, they report a medium-sized effect estimate for mathematical abilities (accounting for ∼35% to 36% of the total effect) and a small effect for general IQ (accounting for ∼4% to 19% of the total effect).13 Our study extends the results of their important work by suggesting that this association is upheld in the most vulnerable infants surviving preterm birth, and that their other comorbidities (ie, NSI) strongly influence this link.

    In addressing our third objective, we found that the cognitive mediation pathway for personal income attainment was stronger in ELBW survivors with NSI. Many individuals with NSI report it is difficult to gain and maintain employment.33 According to the 2010 Survey of Americans With Disabilities, many individuals with NSI cite that gaining employment is challenging because they cannot perform, receive accommodation, or find employment in their desired field; they may face discrimination at their job; or having employment would lead to a loss of their government benefits.34

    There are multiple reasons why cognitive abilities may influence employment and income attainment in ELBW survivors. Perhaps their poorer cognitive and academic abilities result in ELBW survivors being less integrated into their educational and occupational environments, resulting in fewer opportunities for socioeconomic gains.35 For example, educators of ELBW survivors with impaired verbal abilities (ie, reading and public speaking) may not encourage their ELBW students to partake in extracurricular activities such as athletics, arts, or student councils. This may result in smaller social networks and a lost opportunity for these individuals to gain networking skills and socializing skills in early life that may be beneficial in starting their career.36

    Secondly, due to their impairments in mathematical abilities, ELBW individuals may not be pursuing advanced education or quantitatively focused careers, such as those in mathematics, engineering, or sciences, those typically known to have higher entrance incomes.37 Of their current educational attainment, only 48 ELBW individuals reported completing college or university and 8 reported having a graduate or professional degree.1 Compared with ELBW participants (57.2%), more NBW participants (63.6%) reported completing college/university or a graduate degree.1

    Lastly, ELBW survivors are at increased risk for many other impairments, including difficulties with social skills and psychiatric illness.38,39 It may be that their multiple comorbidities place ELBW survivors at greater disadvantage in adulthood, which may be associated with reduced productivity or increased absenteeism,40 resulting in fewer opportunities for upward social mobility.

    Our study is not without limitations. Our self-reported socioeconomic outcomes may have been subject to reporting bias. We attempted to minimize this by asking participants clear questions about multiple sources of their annual income and providing explicit definitions of full-time employment. Secondly, cohort attrition has ensued over the past 30 years, limiting sample size and statistical power. Because nonparticipants had lower childhood SES and childhood cognition, this could potentially have led to an underestimate of our effect estimates. We attempted to minimize any bias caused by attrition by adjusting for factors associated with differential attrition over time (ie, male sex, childhood SES).

    Thirdly, as the comparison group was recruited from a primary school setting, it is possible that children with severe disabilities may be less common in the NBW control group, which may have led our results to be overestimated. A fourth limitation of our study is that we only examined single mediation models and our mediated effect estimates could only be examined qualitatively, which did not take into account measurement error that may have occurred. Lastly, advances in neonatal care and differences in socioeconomic climates for ELBW infants born after our cohort may reduce the generalizability of our results. However, evidence suggests that later generations of ELBW survivors also suffer from impaired cognitive abilities41,42; therefore, our results may be useful for ELBW survivors of all ages.


    This is the first study to provide evidence that early cognition can have a lasting impression on socioeconomic outcomes in ELBW survivors. Our findings suggest that childhood cognition, particularly overall intelligence and mathematical abilities, partially mediated the association between being born at ELBW and income attainment in adulthood, and that this association may be stronger in survivors born with NSI. Research should explore the mediating influence of childhood cognitive abilities in larger, more contemporary samples of preterm survivors. Clinical and public health interventions aimed at enhancing early cognitive functioning may be important to ensure that infants who face perinatal adversity can lead the healthiest and most productive lives possible.


    The authors thank the many participants and their families for their continued participation over the years, and Kimberly Day for her help with data preparation.


      • Accepted December 1, 2016.
    • Address correspondence to Kathleen G. Dobson, MSc, Women’s Health Concerns Clinic, St. Joseph’s Healthcare Hamilton, West 5th Campus, Room C142, 100 West 5th St, Hamilton, ON, Canada L8N 3K7. E-mail: kathleen.dobson{at}
    • FINANCIAL DISCLOSURE: The authors have indicated they have no financial relationships relevant to this article to disclose.

    • FUNDING: This work was supported by a Canadian Institutes of Health Research Team Grant (grant TMH 103145); the Canadian Institutes of Health Research (grant MOP42536); the National Institute of Child Health and Human Development (grant 1-RO1HD40219); and the Hospital for Sick Children Foundation (grant ESPM85-201).

    • POTENTIAL CONFLICT OF INTEREST: The authors have indicated they have no potential conflicts of interest to disclose.