Objective. To determine the relationship between perinatal and sociodemographic factors in low birth weight and sick infants hospitalized at regional neonatal intensive care units (NICUs) and subsequent educational disabilities.
Method. NICU graduates born between 1980 and 1987 at nine statewide regionalized level III centers were located in Florida elementary schools (kindergarten through third grade) during academic year 1992–1993 (n = 9943). Educational disability was operationalized as placement into eight mutually exclusive types of special education (SE) classifications determined by statewide standardized eligibility criteria: physically impaired, sensory impaired (SI), profoundly mentally handicapped, trainable mentally handicapped, educable mentally handicapped, specific learning disabilities, emotionally handicapped, and speech and language impaired (SLI). Logistic regression was used to estimate the odds of placement in SE for selected perinatal and sociodemographic variables.
Results. Placement into SE ranged from .8% for SI to 9.9% for SLI. Placement was related to four perinatal factors (birth weight, transport, medical conditions [congenital anomalies, seizures or intraventricular hemorrhage] and ventilation), and five sociodemographic factors (child's sex, mother's marital status, mother's race, mother's educational level, and family income). Perinatal factors primarily were associated with placement in physically impaired, SI, profoundly mentally handicapped, and trainable mentally handicapped. Perinatal and sociodemographic factors both were associated with placement in educable mentally handicapped and specific learning disabilities whereas sociodemographic factors primarily were associated with placement in emotionally handicapped and SLI.
Conclusions. Educational disabilities of NICU graduates are influenced differently by perinatal and sociodemographic variables. Researchers must take into account both sets of these variables to ascertain the long-term risk of educational disability for NICU graduates. Birth weight alone should not be used to assess NICU morbidity outcomes.
- birth weight
- child development
- educational status
- longitudinal studies
- low birth weight
- intensive care units
- learning disorders
- risk factors
- logistic models
- socioeconomic factors
- NICU =
- neonatal intensive care unit •
- RPICC =
- Regional Perinatal Intensive Care Center •
- DOE =
- Department of Education •
- SE =
- special education •
- PI =
- physically impaired •
- SI =
- sensory impaired •
- PMH =
- profoundly mentally handicapped •
- TMH =
- trainable mentally handicapped •
- EMH =
- educable mentally handicapped •
- SLD =
- specific learning disabilities •
- EH =
- emotionally handicapped •
- SLI =
- speech and language impaired •
- BWG =
- birth weight group
Improving survival among extremely low (<1000 g) and very low (<1500 g) birth weight infants has raised questions about subsequent morbidity.1-3 The cognitive and behavioral performance of school-aged children who received treatment in neonatal intensive care units (NICUs) has come under closer scrutiny as physicians increasingly have been asked to take into account the long-term consequences of extending care to infants at the threshold of viability.4-6 Follow-up studies of NICU graduates have used tests of intelligence, physical coordination, or neurologic functioning as markers of school readiness.7-11 Results on these standardized tests as well as teacher and parent reports have been used to contrast differences among birth weight groups within NICU samples.12-19 A number of studies have looked at the prevalence of learning disabilities among very low birth weight infants followed into the early grades of elementary school.20-24Some researchers have observed a negative impact of low birth weight (<2500 g) on subsequent learning capacity.12-15 25 26However, these studies have for the most part been based on relatively small sample sizes that do not permit the investigation of a large number of factors collectively. Hence, no consensus has emerged about the long-range effects of early NICU experience on school performance. In this study we investigated the assignments at school age of NICU graduates to programs for children with educational disabilities. This outcome is of particular interest because these assignments determine the facilities made available to children and have a major impact on their subsequent development.27
Educational outcome has been reported to be associated with medical conditions in the perinatal period (eg, congenital anomalies, intraventricular hemorrhages, seizures, and ventilation)28-31 and certain sociodemographic variables (eg, income, race, sex, maternal education, and marital status).27 32-34 We conducted this prospective follow-up study to determine the educational disabilities of extremely low (<1000 g), very low, and low birth weight infants as well as sick term infants who required regionalized perinatal intensive care. We examined the relationship of birth weight, medical conditions requiring intensive care, and sociodemographic risk factors to placement in specific types of special education (SE) classes. Our statewide sample was large enough to conduct statistical analyses that control for these factors when assessing the independent effects of birth weight, medical conditions, and sociodemographic variables on specific educational disabilities.
Between 1980 and 1987 Florida's Children's Medical Services' Regional Perinatal Intensive Care Centers (RPICC) program was the major provider of level III neonatal intensive care in the state.35-37 In this study, children born between September 1, 1980, and August 31, 1987, and treated at RPICC NICUs in Ft Lauderdale, Gainesville, Jacksonville, Miami, Orlando, Pensacola, St Petersburg, Tampa, and West Palm Beach were followed into the state of Florida school system. During the study period, the RPICC program operated under a set of statewide rules and practice parameters derived from system-wide clinical monitoring and use of International Classification of Diseases, 9th Revision (ICD-9) diagnostic codes. None of the centers during this period were involved in research protocols evaluating the use of surfactant or betamethasone. By aggregating across centers, it became possible to achieve cell sizes large enough to conduct analyses of relatively infrequent birth weight categories and rare SE placements. Center-to-center variability was accounted for, at least partially, by including in our multivariable modeling process a large set of predictor variables which are prominent in defining a center's case mix and commonly thought to potentially influence educational outcomes.
Records for 24 698 children who were treated at RPICCs during this period and who did not die before discharge were available to be linked to the Florida Department of Education (DOE) student database for the academic year 1992 to 1993. The DOE database contains the academic performance records of all students in Florida public schools. Sixty-one percent of the RPICC records (n = 15 096) were matched to DOE records. Forty percent of the records (n = 9943) had complete information on all variables considered in the analysis. Twenty-one percent (n = 5153) had incomplete information on one or more variables and therefore could not be included in the analysis. The remaining 39% (n = 9602) were not linked. A comparison of the three groups showed minimal differences in the percentage distribution of infants into categories of predictor variables (see Table 1).
The perinatal predictor variables considered were birth weight, transport status, medical conditions, and ventilation. Five birth weight categories were defined: 500 to 749 g, 750 to 999 g, 1000 to 1499 g, 1500 to 2499 g, and ≥2500 g. Children were designated as having a missing transport status if their hospital of birth was unavailable; they were considered inborn if they were born in one of the RPICCs; and they were considered to be transported if they were moved to an RPICC after birth. Medical conditions included all children who had a congenital anomaly [ICD9s 740–759.9, excluding patent ductus arteriosus (747.0)]), an intraventricular hemorrhage (772.1), or seizure (779.0, 779.2). The sociodemographic predictor variables considered were child's sex, mother's race, mother's marital status, mother's education, and family income at time of birth. Mother's age is highly confounded with marital status and maternal education and therefore was not included. Grade of child was also included in our analysis as a control variable.
SE consisted of assignment to any one of eight specific, mutually exclusive classroom categories designed to serve children with an educational disability. Assignment was determined by the child's primary exceptionality which identified the disability requiring the greatest allocation of personnel resources (in cases in which more than one disability was diagnosed). SE categories included: 1) physically impaired (PI): severe skeletal or neuromuscular condition adversely affecting educational performance, comprises 0.3% of Florida's kindergarten through third grade population in 1992 to 1993; 2) sensory impaired (SI): deaf, blind, hard of hearing, partially sighted, 0.1%; 3) profoundly mentally handicapped (PMH): IQ < 25, 0.1%; 4) trainable mentally handicapped (TMH): IQ between 25 and 54, 0.2%; 5) educable mentally handicapped (EMH): IQ between 55 and 69, 0.9%; 6) specific learning disabilities (SLD): psychologic processing disorders marked by difficulties in the acquisition and use of language, reading, writing, or math, 3.8%; 7) emotionally handicapped (EH): condition resulting in persistent and maladaptive behaviors, 0.8%; and 8) speech and language impaired (SLI): disorders of language, articulation, fluency, or voice, 7.8%). A ninth category, “other SE,” included all other SE subtypes (eg, the low-occurrence classification, traumatic brain injury, 0.4%) except gifted.
Procedures for determining eligibility and classification criteria of primary exceptionality are standardized throughout the state, dictated by Florida Board of Education Rules38 and monitored by the Florida Bureau of Student Services and Exceptional Education.39 These definitions and eligibility criteria which classify children into SE types are comprehensive and require extensive multidisciplinary evaluation procedures by qualified professionals using widely accepted assessment tools and methods.
Separate analyses were completed for each of the eight SE outcomes described above. Because of the rarity of some SE conditions studied, it was not feasible to estimate and test interactions in these analyses. Therefore, main effects models were fit. The categorical modeling procedure in SAS was used to estimate and test model parameters by maximum likelihood methods.40 This multivariate modeling process controlled for other variables in the model when assessing the effect of each predictor variable.
The analysis for each SE subtype was done conditional on no other SE. That is, the outcome variable in each analysis had two categories: 1) placement in the specific subtype of SE; and 2) no SE placement of any type. Analyses were performed on the subsample of children who fell into these two categories only, for each SE subtype. No SE was taken as the reference category in each analysis. Odds ratios, with associated confidence intervals, were used to measure the effect of levels of each predictor variable in relation to an appropriate reference level. All tests were performed at the .05 significance level.
Placement in Specific SE Types
Table 2 provides a description of the study sample by giving its distribution into levels of predictor variables (ie, percentages in the “total” column) and the distribution of these subgroups into the various types of SE placements (ie, percentages across SE type columns within each predictor row). A total of 2613 of the 9943 children in this NICU sample (26.3%) required SE. The percentage of placement in specific SE types ranges from .8% in SI to 9.9% in SLI. The percentage of placement in SE by birth weight group (BWG) ranged from 39.8% (BWG, 500–749 g) to 23.9% (BWG, >2499 g), a difference of 15.9%.
Results from analyses of the relationship of perinatal and sociodemographic variables to specific types of SE placement are given in Table 3. Odds ratios describing the effects of perinatal and sociodemographic factors on each educational outcome variable are given therein. Conceptually, these analyses hold all other variables constant while looking at the independent contribution of each variable. Each type of SE was associated with different predictor variables.
PI was significantly and adversely associated with the perinatal factors of birth weight (500–749 g, 750–999 g, 1000–1499 g) and medical conditions. Being inborn or black race had a protective effect.
SI was significantly and adversely associated with two perinatal factors only: birth weight (500–749 g, 750–999 g) and medical conditions.
PMH (IQ <25) was significantly and adversely associated with two perinatal factors, medical conditions, and ventilation, and one sociodemographic factor, mother's education greater than high school. Black race was protective.
TMH (IQ 25–54) was associated with medical conditions only.
EMH (IQ 55–69) was associated with birth weight (500–749 g, 750–999 g, 1000–1499 g), medical conditions, ventilation, male sex, black race, and mother's education less than high school. Being inborn or mother's education greater than high school was protective.
SLD was associated with two perinatal factors, birth weight (750–999 g) and medical conditions and three sociodemographic factors, male sex, mother's education less than high school, and family income $4000–$11 999. Black race was protective.
EH was associated with the sociodemographic factors male sex, unmarried mother, family income (<$4000, $4000–$11 999). Being inborn was protective.
SLI was associated with male sex and family income <$4000. Protective factors were inborn, black race, other (primarily Hispanic) race, or mother's education greater than high school.
In summary, PI, SI, PMH, and TMH were significantly influenced primarily by perinatal factors. In contrast, sociodemographic and perinatal factors both played a role in EMH and SLD, whereas EH and SLI were influenced almost exclusively by sociodemographic factors. Males were at greater risk for placement in SE categories affected primarily by sociodemographic factors but did not differ from females in categories affected primarily by perinatal factors.
With increasing survival among low birth weight infants have come questions about long-term outcome for NICU survivors.41-48 It is also important to study the period before the widespread use of maternal corticosteroids and surfactant as a benchmark for estimating improvements in outcome. Calls for further research to assess the quality of life and functional capacities of infants treated in NICUs continue to appear.49 50 One important measure of outcome for NICU graduates is assignment at school age to programs for children with educational disabilities. This study prospectively followed a cohort of NICU graduates born between 1980 and 1987 into Florida elementary schools (kindergarten through third grade) during the academic year 1992 to 1993.
Our descriptive statistics suggested a birth weight effect on SE placement: 39.8% of infants weighing 500 to 750 g required SE compared with 23.9% of infants weighing ≥2500 g. However, it was unclear how much of this difference might be attributable to the corelationship of birth weight with medical conditions, ventilation, or sociodemographic factors. To assess the unconfounded effect of birth weight, we used multivariate modeling techniques to estimate the effect of each perinatal and sociodemographic factor when controlling for the others.
The results of our analyses by SE subtype were consistent with the hypothesis that certain SE outcomes result from structural damage before NICU discharge and others from cumulative environmental factors. Four SE outcomes (PI, SI, PMH, and TMH; ie, 6.2% of the sample) were significantly affected by perinatal factors and, with two exceptions (race effect on PI and maternal education effect on PMH), not by sociodemographic factors. In contrast, SE outcomes that might be expected to be environmentally induced (EH and SLI; ie, 11.6% of the sample) were significantly influenced by several of the sociodemographic factors but by only one perinatal factor, transport. Further study is needed to determine why being inborn had a protective effect on four SE outcomes.
In an apparent contrast to several recent studies that have reported a greater likelihood of behavioral and emotional problems among children born at lower birth weights,12 17 51 placement in EH classes for this NICU sample was not associated with any birth weight category but rather with the sociodemographic factors of sex, mother's marital status, and income. This discrepancy may be attributable to the fact that earlier studies compared behavior problems in low birth weight children to normal birth weight, non-NICU children, whereas the reference group for this study was NICU children ≥2500 g.
EMH and SLD (ie, 8.3% of the sample) were affected by a combination of perinatal and sociodemographic factors. The least severe category of mental retardation, EMH, may result from either less serious structural damage than that involved in PMH or TMH or adverse environmental factors that are more likely to occur in families of low socioeconomic status.52 SLD also was associated with both perinatal and sociodemographic variables. Note that SLD is considered a broad category encompassing both children with learning problems attributable to neurophysiologic factors as well as those attributable to social-behavioral factors.53
The odds ratios provide more detailed support for the hypothesis that certain SE outcomes result from structural damage whereas others arise from cumulative environmental factors, primarily poverty. Very low birth weight was associated with PI, SI, and EMH only. This finding was consistent with previous research that the only clear-cut association of NICU hospitalization with educational outcome was restricted to sensory or physical impairments.54 The odds ratios also indicate that medical conditions (intraventricular hemorrhage or seizures or congenital anomalies) were significantly associated more frequently with SE subtypes than was low birth weight or very low birth weight. In this NICU population, perinatal variables generally were more strongly associated with SE outcomes than were sociodemographic factors. Among sociodemographic factors, males were at an increased risk for placement in four SE categories (EMH, EH, SLD, and SLI), a trend also reported in the general SE population.55
In Florida, more than 95% of all children with disabilities are enrolled in public schools, and public schools serve 90% of the kindergarten through third grade school-age population. Florida follows United States Department of Education established regulations for defining and identifying students with disabilities as recognized in the federal Individuals with Disabilities Education Act (1991). In turn, Florida's Department of Education provides rules and regulations that guide 67 county school districts in the identification of students with disabilities. Nationwide, the ways that state departments of education operationalize definitions of disabilities are quite similar; 94% of the states, including Florida, use comparable criteria to determine eligibility.56 Placement procedures for the more severe disabilities (PI, SI, PMH, TMH, EMH) are typically consistent across the states.57 Even for SLD, which has provoked the most controversy regarding definition and identification,5871% of the states, including Florida, use the exact or slightly modified language of the United States Department of Education in their regulations.56
Bias resulting from selective attrition or deletion from the study sample because of incomplete records is an issue in virtually all large-scale, long-term follow-up studies of this type. In univariate analyses, the effect of such selection or incompleteness can confound the assessment of explanatory factors on outcomes. The multivariate modeling technique used in this study reduced concerns about such confounding. Primary reasons for the loss-to-follow-up in this study were child's unresolved name change, non-SE child attended private school, and child moved out-of-state.
At least 80% of the level III intensive care beds in Florida were in the RPICC during the 8-year study period. Variations in case-mix and treatment practices across the nine centers were partially controlled by the multivariable modeling technique. However, other variables not included in the models may also contribute to variability in case-mix across centers. Infants born or transported to RPICCs during this period tended to represent the most serious cases. Therefore the RPICC sample is likely to have a higher incidence of SE than the entire Florida NICU population.
Our regionalized NICU sample, comprised of nine centers and nearly 10 000 infants, was large enough to explore the independent effects of perinatal and sociodemographic factors on rare SE outcomes. An exception was the separate effects of congenital anomalies, seizures, and intraventricular hemorrhage on placement in specific educational disability categories; these rare perinatal events required grouping. To determine the effects of these events without grouping, national pooling of data sets would be necessary. It would then be feasible to separate congenital anomalies from a combined intraventricular hemorrhage-seizure category to elucidate their respective effects on PMH and TMH. In future studies, it would be possible to assess the intervening effects of such medical variables as restricted postnatal growth, failure to thrive, anemia, chronic illness, and accidents on SE outcomes. Similarly, an analysis of the impact of a range of family support, early intervention, and preschool programs on NICU populations would be critically important.
This study demonstrates that different SE outcomes for NICU graduates are influenced differently by perinatal and sociodemographic variables. Birth weight should not be used as the sole predictor of outcome. To adequately assess the risk of subsequent morbidity, clinicians and researchers must consider both perinatal and sociodemographic risk factors and their effects on specific SE subtype outcomes.
This research was supported in part by grants from the United States Department of Education, Office of Special Education Programs (H024J30014); Florida Department of Education, Bureau of Information and Assessment Services (011–92000–60001); Florida Department of Health and Rehabilitative Services, Children's Medical Services (MQ620), and the University of Florida (UF), Office of Research, Technology, and Graduate Education.
We wish to acknowledge the cooperation and assistance of the medical directors and their staffs at Florida's RPICC hospitals during the period covered by this study: Dr Gregor Alexander, Orlando Regional Medical Center, Orlando; Dr Luis Arango, St Mary's Hospital, West Palm Beach; Dr Eduardo Bancalari, Jackson Memorial Medical Center, Miami; Dr Richard Bucciarelli, Shands Hospital, Gainesville: Dr Thomas Chiu, University Hospital, Jacksonville; Dr John Curran, Tampa General Hospital, Tampa; Dr Gregor Melnick, Broward General Hospital, Ft Lauderdale; Dr Robert Sosa, All Children's/Bayfront Center, St Petersburg; and Dr Edward Westmark, Sacred Heart Hospital, Pensacola.
We also wish to thank John Brady, Lavan Dukes, Tom Fisher, and Janice Smith-Dann of the Florida Department of Education; Dr William Ausbon, Dr Michael Cupoli, Janet Evans, Susan Folks, and Robert Furlough of Children's Medical Services.
We are also indebted to Zhanying Bai, Somnath Sarkar, Yuanshan Sun, and Chi-Hse Teng of the UF Division of Biostatistics for input in the early stages of developing our merging strategy; to Danny Deluca, Stefan Lupkiewicz, and Bill Pfeifer, UF Division of Computer Science, for data management; to Jaquelyn Liss Resnick, UF Counseling Center, for editorial assistance; and to Matt Davidson and Steven Priest, UF Department of Pediatrics, for manuscript preparation.
- Received October 31, 1997.
- Accepted March 2, 1998.
Reprint requests to (M.B.R.) Department of Pediatrics, College of Medicine, University of Florida, PO Box 100296, Gainesville, FL 32610-0296.
This work was completed while Shanti V. Gomatam was a postdoctoral fellow at the University of Florida.
- Hack M,
- Klein NK,
- Taylor HG
- La Pine TR, Jackson JC, Bennett FC
- Bhushan V,
- Paneth N,
- Kiely JL
- American Academy of Pediatrics Committee on Fetus and Newborn, American College of Obstetricians and Gynecologists Committee on Obstetric Practice
- Johnson A
- Marlow N,
- Roberts L,
- Cooke R
- McCormick MC,
- Workman-Daniels K,
- Brooks-Gunn J
- Halsey CL,
- Collin MF,
- Anderson CL
- Klebanov PK,
- Brooks-Gunn J,
- McCormick MC
- Ross G,
- Lipper EG,
- Auld PA
- Hocutt AM
- Hille ETM,
- Den-Ouden AL,
- Bauer L,
- van den Oudenrijn C,
- Brand R,
- Verloove-Vanhorick SP
- ↵Florida State Board of Education Administrative Rules. Special Programs 1: Chapter 6A-6.03011–6A-6.03025
- ↵Florida Statutes. Public Education, General Provisions, Definitions: Chapter 228:041(18)(19)
- ↵SAS Institute. SAS/STAT User's Guide: Version 6. 4th ed. Cary, NC: SAS Institute; 1990
- Lee KS,
- Kim BI,
- Khoshnood B,
- et al.
- Collin MF,
- Halsey CL,
- Anderson CL
- Escobar GJ,
- Littenberg B,
- Petitti DB
- McCormick MC
- ↵Harry B. The Disproportionate Representation of Minority Students in Special Education: Theories and Recommendations. Washington, DC: US Department of Education, Office of Special Education and Rehabilitative Services; 1994
- Resnick MB,
- Roth J,
- Ariet M,
- et al.
- ↵Marder C, D'Amico R. How Well Are Youth With Disabilities Really Doing? A Comparison of Youth With Disabilities and Youth in General. Menlo Park, CA: SRI International; 1992
- ↵Ysseldyke JE, Thurlow ML, eds. Educational Outcomes for Students With Disabilities. New York, NY: Haworth Press; 1994
- Reschly DJ
- Copyright © 1998 American Academy of Pediatrics