Effects of Prenatal Cocaine Exposure on Special Education in School-Aged Children

Measures

Medical characteristics were collected at birth (Table 1).

Statistical Methods

One-way analysis of variance and χ² were used for continuous and categorical variables, respectively. Logistic regression was used to examine the effect of PCE on 7-year special education outcomes, including the presence of an IEP, SE, SEC, SS, and SLS, controlling for covariates. Covariates included a priori were gender, low birth weight (LBW) (<1500 g), and study site. Additional characteristics were examined in preliminary analyses as candidate covariates (Tables 1 and 2). Candidate covariates that were correlated with PCE and the outcome measures (P ≤ .10) were included in the logistic regression analysis. Measures that met this criterion included small for gestational age, defined as gender-specific weight <10th percentile for gestational age; ethnicity (white versus nonwhite); low SES at 7 years (Hollingshead Index of Social Position level 5); any primary caregiver change (≥1); poor quality of the home environment as assessed by using the HOME total averaged at 10 months and 5.5 years and recoded into a dichotomous variable (lowest 20% cutoff); Medicaid insurance status; prenatal tobacco use; low caregiver IQ (<85); caregiver marital status at 1 month; maternal age at birth (in years); caregiver depression as indicated by average Beck Depression Inventory assessed at 4 months, 30 months, and 5.5 years and recoded into a dichotomous variable (score of >17, indicating moderate to severe depression); and any child abuse.

Regressions for each outcome variable used stepwise elimination of the covariates that made the least contribution to the models and had the least effect on other parameters. The eliminated covariates were then added back to the model in a stepwise manner to test for confounding effects. All covariates in the final models were required to have significant contribution (P < .10) to the outcomes except for the 3 selected a priori (gender, LBW, and study site).

Two sets of logistic regression analyses were performed: with and without low child IQ (score of <85 on the Wechsler Intelligence Scale for Children III) as a covariate. The IEP evaluation process involves a multidisciplinary approach³⁵ and often uses several diagnostic tools. Because IQ scores are a predictor of educational achievement and are often reviewed when assessing a student for the necessity of an IEP and other special education services, we thought that it was important to investigate the interactions of the hypothesized cocaine effects with low child IQ on special education.

Retention

Of the 1388 infants (658 PCE/opiate exposed and 730 comparison children) seen at the 1-month visit, 1023 had 7-year school outcome data. Among the 1023 children, 79 were opiate users and 1 had incomplete data and were not included in this study, resulting in a final sample of 943 children. Compared with children who were not included in the study, included children were more likely to be black, have lower SES, be from mothers with lower IQ, and have less prenatal tobacco exposure (any or heavy; Table 1). In addition, those with 7-year data had a higher percentage of caregivers who had postnatal alcohol (any or heavy) and marijuana use (any or heavy). Included children were also more likely to have a low HOME score and to have lived in environments with domestic violence. There were no significant differences in newborn medical characteristics or child IQ between the 2 groups.

Sample Description According to Prenatal Cocaine Exposure

Children in this study with PCE were more likely than comparison children to be born to families with low SES at 1 month; come from single-caregiver households; be on Medicaid; and have mothers who had less education, lower IQ, and no prenatal care and were older (Table 2). They were also more likely to have prenatal exposure to alcohol (any and heavy), tobacco (any and heavy), and marijuana (any and heavy). They also were exposed to more postnatal cocaine (any and heavy), alcohol (any and heavy), tobacco (any and heavy), and marijuana (any and heavy); had greater probability of primary caretaker change and low SES; and had a greater likelihood of living in a home with domestic violence. Children with PCE were also more likely to be small for gestational age, less likely to be the first born child, and more likely to have lower IQ.

Of the 267 mothers who reported prenatal cocaine use, 47.9% reported use during all trimesters, 14.6% during the first and second trimesters only, and 14.6% during the first trimester only. Overall reported use declined progressively over the trimesters (81.3%, 75.4%, and 66.7% in the first, second, and third trimesters, respectively) as did heavy use (23.1%, 16.8%, and 11.8%, respectively), whereas some use remained relatively constant (31.7%, 33.9%, and 32.4%, respectively). These trends are similar to those reported previously.³⁰ Of 92 women who reported heavy use in the first trimester, 22.8% continued heavy use in the first 2 trimesters and 32.6% continued heavy use throughout all 3 trimesters. Of 126 who reported some use during the first trimester, 50% continued throughout all 3 trimesters.

Special Education Outcomes

In unadjusted logistic regression analysis, children with PCE (16.5%) were more likely than children in the comparison group (11%) to have an IEP (odds ratio: 1.599 [95% confidence interval: 1.091–2.344]; P = .016). There were no effects of PCE on SEs, SSs, SECs, or SLSs.

Complete data for each outcome model controlling for covariates were available for 737 to 916 children. Logistic regression analyses with low child IQ as a covariate (Table 3) showed that children with PCE were more likely to have an IEP. Increased likelihood of an IEP was also related to male gender, LBW, white race, low child IQ, and site. No PCE effect was found on SEs, SSs, SECs, and SLSs. LBW and low child IQ were consistently related to the presence of all outcomes. Site was related to all outcomes except SECs. White race was associated with more SLSs.

In logistic regression analyses without low child IQ as a covariate (Table 3), PCE increased the likelihood of SSs but not IEPs, SEs, SECs, or SLSs. Having an IEP was related to male gender, LBW, white race, low caregiver IQ, low HOME score, and changes in caregiver. SEs were predicted by male gender, LBW, low HOME score, and site. Along with PCE, SSs were related to male gender, LBW, low HOME score, and site. SECs were predicted by LBW and low caregiver IQ. Increased SLSs were related to male gender, LBW, low HOME score, and site.

Our preliminary, univariate analyses did not find any correlations between the levels of prenatal cocaine use and any special education outcomes. We further tested the effect of heavy prenatal cocaine by controlling for the effects of heavy prenatal tobacco, alcohol, and marijuana use, as well as the dosage-response effects of these drugs (number of cigarettes per day, number of absolute alcohol drinks per day, and number of marijuana joints per day); however, no significant effects of heavy cocaine use were found.

PCE and Special Education Outcomes

We found that PCE increased the risk for receiving an IEP and SS with adjustment for covariates. Our findings compliment previously reported deficits in intelligence^7,14,36–39 and component academic skills, including learning disabilities, poor sustained attention, visual-motor integration and visuospatial memory, and more disorganized and less abstract thinking in this population^{5,36,38,40–46} and suggest that these deficits generalize to the school setting. The rate of services received through IEPs for both those with PCE (16.5%) and comparison children (11%) was significantly higher than the national average of 6.8% in 2003 for children in kindergarten through third grade,⁴⁷ reflecting the high-risk nature of this sample and their need for continued educational services.

The cocaine effect on IEPs was observed with IQ in the model in addition to other covariates, indicating that the IEP referral was not simply attributable to lower IQ in children with PCE. The cocaine effect on SSs was observed without but not with IQ in the model, suggesting that in this case, IQ could mediate the effects of PCE on SSs. Conversely, children with PCE were not more likely to receive SE, SEC, or SLS services. This could suggest that the effects of PCE on school function are not in specific academic domains. These children may need more special education services in general but not any specific type of service. We had expected more SLS referrals for the PCE children on the basis of previous work^14,18–24; however, within the IEP group, 95.3% of the PCE group and 91.4% of comparison children also had SLSs, indicating that SLSs are a major area of recognized resource necessity in special education for this population regardless of PCE.

Effects of Other Covariates

LBW was a consistent predictor of all special education outcomes in all models. This supports previous work suggesting that LBW is associated with multiple, adverse neurodevelopmental outcomes, including increased need for special education.⁴⁸ Similarly, that low child IQ was a predictor of all outcome models is to be expected, because IQ is often used in the assessment of special education. In addition, in all models, the inclusion of IQ explained a greater percentage of the variance and significantly changed the contributions of the other covariates (Table 3). Factors such as caregiver IQ, HOME score, and caregiver change were statistically significant only when IQ was excluded from the models, possibly suggesting complex interactions among intelligence, environmental, and genetic factors.^49,50 For example, a problematic, highly stressed home environment can adversely affect IQ scores, and low child IQ scores could mediate the effects of poor home functioning on special education outcomes.

We also found that boys were more likely to receive an IEP than girls, supporting previous reports.⁴⁷ Male gender also increased the likelihood of receiving SEs, SSs, and SLSs when IQ was not included in the models. This could be explained by the significantly greater percentage of boys (56.6%) versus girls (43.4%) in the low child IQ group (P = .03).

Our study also indicates that white race increases the likelihood of having an IEP and SLS, independent of other factors, including low SES and low child IQ. On closer inspection, we found that white students were more likely to receive an IEP at the Detroit site than nonwhite students (29% versus 11.9%; P = .007). This finding may also help to explain site effects (Table 3). None of the other sites had statistically significant differences in rates of IEPs or SLSs in white versus nonwhite samples. Enrollment in private versus public school was not a factor in IEPs or SLSs, because only 3 students with IEPs and 4 with SLSs were enrolled in private school. That race plays a role in determining special education services either as a proxy for other factors or as a causal factor requires additional investigation.

Study Limitations

There are several limitations to this study. First, the data are based on school record review. Although our staff was trained in how to abstract the data, we cannot verify the accuracy of records. Second, the data were gathered from several schools in 4 different states that may have had different thresholds for enrollment in special education. Although federal legislation mandates that each school must provide appropriate services for each child with a disability,³⁵ states and school systems have flexibility about information that they require in an IEP. There is state-to-state variation on documentation of standards and assessments on IEP forms, including differences between the states included in this study.⁵¹ These forms are a primary source of information to guide decisions during IEP team meetings. Ultimately, a group of qualified professionals and the parents look at the child's evaluation results and together decide whether the child is a “child with a disability,” as defined by Individuals With Disabilities Education Act,⁵² which is not a standardized process.

We also recognize that the special education outcomes are correlated; however, we believed that it was important to distinguish specific services because they have different educational implications for each child. Another possible limitation is that mothers were recruited at delivery and could represent a different population than women who are recruited during pregnancy. Although that they were asked to report their drug use over the entire pregnancy could be affected by recall bias, there is evidence that postnatal self-report measures of maternal cocaine use are as effective as antenatal measures in predicting neurobehavioral outcomes.⁵³ Finally, it is possible that we underestimated cocaine effects by adjusting for variables such as low IQ and LBW that are on the “causal pathway” between cocaine and the special education outcomes⁵⁴; however, we believed that it would be too difficult to interpret special education findings without considering these factors.

Policy Implications

These findings have policy implications. There were 4112052 births in the United States in 2004⁵⁵; 3.9% of pregnant mothers reported illicit drug use in 2004–2005 in the past month.⁵⁶ Although estimates vary, even a conservative estimate of 45000 children born with PCE per year would suggest, on the basis of our study, that 12.7% will receive an IEP. Enrollment in special education in the United States costs an additional $8080 per year per student⁵⁷ compared with nonenrolled students. The risk attributable to cocaine for an IEP is 1.79 − 1.00 = 0.79 (Table 3). Thus, additional cost per year for special education services as a result of PCE alone would total 45000 cocaine births per year × 12.7/100 (baseline IEP rate) × 0.79 (excess risk attributable to cocaine) × $5918 (additional cost per child) = $26718882. This yearly cost would then have to be multiplied by the number of years the child receives special education services in school. “Investing” in early intervention might not only relieve long-term suffering in these children but also be cost-effective.

Illicit drugs, such as cocaine, may not be the only prenatal exposures that lead to an increased use of special education resources. Legal substances such as tobacco and alcohol, which were controlled in our study, may have effects on these outcomes. Findings of special education use in children with prenatal alcohol exposure but not fetal alcohol syndrome (comparable to our study children) have varied.^58–60 We found no studies that examined the effects of prenatal tobacco use on special education outcomes. These are important areas for future investigation.