Early educational intervention for very low birth weight infants: Results from the infant health and development program

Early educational intervention for very low birth weight infants: Results from the Infant Health and Development Program Marie C. M c C o r m i c k , MD, ScD, C e c e l i a M c C a r t o n , MD, J a m e s monascia, PhD, a n d J e a n n e Brooks-Gunn, PhD From the Department of Maternal and Child Health, Harvard School of Public Health, and the Joint Program in Neonatology, Harvard Medical School, Boston~ Massachusetts; the Depait~ merit of Pediatrics, Albert Einstein College of Medicine, Bronx, New York; the Department of Biostatistics,Johns HopkinsSchool of Hygiene and Public Health, Baltimore, Maryland; and the Center for the Study of Young Children and Families, Department of Psychology, Teachers College, and Department of Pediatrics, College of Physiciansand Surgeons, Columbia University, New York, New York

Objective: To e x a m i n e the effect of early e d u c a t i o n a l intervention after discharge from the hospital on the health and d e v e l o p m e n t a l status of very low birth weight (___1500 gm) infants. Design: Randomized, controlled trial, with post hoc analysis. Setting: Eight sites, heterogeneous for s o c i o d e m o g r a p h i c a n d health care use. Participants: Infants (N = 280) born weighing <1500 gm and selected for the Infant Health and Development Program. Eligibility was limited primarily by geographic distance from the d a y care center. One third were randomly assigned to the intervention (INT) group and two thirds to f o l l o w - u p o n l y . Interventions: All children received intensive pediatric and d e v e l o p m e n t a l surv e i l l a n c e . The INT group received home visits and center-based e d u c a t i o n a l interventions until 36 months of a g e (corrected for gestatlonal age when final assessments were completed). Outcomes: Cog nitive d e v e l o p m e n t (Stanford-Binet Intelligence Scale), behavioral c o m p e t e n c e (Achebach Child Behavior Checklist), and health status (in: dexes summarizing reported morbidity, the Functional Status II [R] Scale, and General Health Ratings Index). Results: Cognitive d e v e l o p m e n t scores were 7.2 points higher (p = 0.002) in the INT group, after adjustment for baseline differences in site, s o c i o d e m o g r a p h i c characteristics, and neonatal morbidity, and were 9.4 points higher (p <0.0003) when the 29 children with significant cerebral palsy were removed. No differences in behavior, serious morbidity, functional status, or health rating were found overall. The infants in the INT group who weighed <1000 gm at birth had significantly lower behavior problem scores but no differences on other outcomes. All children in the INT group had slightly higher rates of less serious morbidity.

Supported by the Robert Wood Johnson Foundation, Maternal and Child Health Bureau (MCJ-360593-01-02-0), National Institute of Child Health and Human Development (HD 27344 01A1), and Pew Charitable Trusts to the Longitudinal Study Office, Albert Einstein College of Medicine, and the Data Coordinating Center, Johns Hopkins School of Hygiene and Public Health. For the names of members of the Infant Health and Development Program, see the listing at the end of the text.

Submitted for publication March 15, 1993; accepted May 27, 1993. Reprint requests: Marie C. McCormick MD, ScD, Professor and Chair, Department of Maternal and Child Health, Harvard School of Public Health, 677 Huntington Ave., Boston, MA 02115. Copyright | 1993 by Mosby-Year Book, Inc. 0022-3476/93/$1.00 + .10 9/20/48950 527

528

M c C o r m i c k et al.

The Journal of Pediatrics October 1993

Conclusion: The advantage conferred by being in the INT group, as previously reported for heavier infants, extends to very low birth weight children, supporting the use of early intervention in this group. (J PEDIATR1993;123:527-33) Early intervention programs have beneficial effects on the health and development of preschool children. 1 Most of the empiric evidence, however, involves single-site studies focused on samples limited by socioeconomic status or specific health or developmental problems. 24 The extension of this experience to other localities and more heterogeneous groups is not always straightforward. For example, one group of children identified as a target for early intervention programs under recent legislation in many states consists of very low birth weight (_<_1500 gm) children. 5 This selection reflects the accumulated evidence that these children have an increased risk of a number of developmental problems of varying severity.6-s Much of the evidence supporting early intervention in this group is derived from interventions conducted in the neonatal intensive care unit before the infant goes home, 9 and not from the home- or center-based programs more generally available in the community. Several reports 8, 10-12 have documented substantial gradients in cognitive functioning among VLBW children by the socioeconomic status of their families, suggesting the potential for altering their outcomes by postdischarge, early-childhood intervention programs. There is accumulating evidence that early intervention programs relying solely on physical therapy services may have limited benefits, a3, 14 but no studies have examined the efficacy of broader, multidisciplinary early interventions for VLBW children. The research that appears to address this question most directly emerges from the Infant Health and Development Program. Previous reports from this multisite, randomized trial of early educational intervention for low birth weight, premature infants document substantial enhancement in cognitive development and modest reductions in reported behavior problems for those receiving the intervention. 1517 These .gains have been achieved with only modest increases in less serious morbidity, 15 despite early and intensive reliance on a day-care center model. The degree of participation in the intervention and its influence on cognitive and behavioral outcomes have been examined, 18 as have the effects of an intervention that included 2 years of day-care on the use of child health care services 19 and subsequent maternal participation in the workforce, educational attainment, and childbearing.2~ Moreover, the intervention enhances cognitive development for all but the children with the best-educated mothers (i.e., those mothers with more than a high school education). 21 In both the primary analysis 15 and those by race/ethnicity and maternal education, 21 however, the effect of the intervention on cognitive development is substantially less for the lower birth weight

infants (in this case, ___2000 gin) than in the heavier infants (2001 to 2500 gm), suggesting that even smaller infants such as VLBW infants may not achieve much benefit from early intervention programs based on this model. To address this question, we performed a post hoc analysis of the effect of the IHDP on VLBW children. METHODS

The methods of this stuc[y have been reported previously and will be summarized here briefly. The study was approved by all participating institutions. Sample r Children eligible for the entire study consisted of all infants born in participating hospitals who weighed --<2500 gm at birth and had completed ___37 weeks of gestation.15, 16, 22 In addition, the families of the children had to live within a specified distance from the intervention site and be deemed able to deliver the intervention in English. Children with major anomalies or other specified conditions that were better assisted in other programs wer e excluded. Recruitment occurred differentially within two specified strata (>2000 gm and ~2000 gin), with the intent of having two thirds of the sample in the lower birth weight group at each site. The sample in this report included all VLBW children selected for the study in the lower birth weight group plus twin siblings who met original study criteria. One twin was CP ELBW INT VLBW

Cerebral palsy Extremely low birth weight ( ~ 1000 gin) Intervention [group] Very low birth weight (___1500 gm)

randomly selected to be in the primary analysis group and was randomly assigned to treatment status as noted below. Treatment status of the remaining twin sibling was yoked to that of the selected twin (i.e., if the twin in the primary analysis group was assigned to the intervention, his or her twin also received the intervention). There were 256 VLBW children in the primary analysis group and 24 twin siblings. Intervention, These children were then randomly assigned to a group, within a site and birth weight stratum, by using a weighted randomization procedure, 23, 24 one third of the children being assigned to the intervention group and two thirds to follow-up only. The intervention, described in detail elsewhere, 18 consisted of three major modalities: 1. Home visits weekly in the first year, and every other week thereafter through 36 months of age corrected for duration of gestation 2. Educational center-based care beginning when the

The Journal o f Pediatrics Volume 123, Number 4

child was 1 year of age corrected for duration of gestation, and available 5 days a week, 8 hours a day, through 36 months of age 3. Parent support groups available concurrently with the center-based care Both child development and parent-focused activities were guided by established curricula25, 26 throughout the duration of the program regardless of modality. The intervention strategy 18 had two major components. One was designed to assist parents to understand their child's development and to learn specific ways to foster cognitive, physical, and social development through a series of activities linked to the child's developmental level. The second component focused on enhancing the parents' own skills in coping with problems encountered in caring for their child. The program was initiated in the home visits during the first year and continued during the next 2 years with the combination of center-based and home-based care. Finally, developmentally appropriate play materials were also provided. All children in both groups received routine health and developmental assessments on a specified schedule. Local site staff were free to make referrals for any additional health or social services considered needed at any assessment point. Assessments. All children were assessed at 4-month intervals in the first year and 6-month intervals thereafter, according to an established schedule, within 2 weeks of the assessment dates. The information for this report was obtained at 36 months of age (corrected for duration of gestation). The primary outcome measures consisted of the following: 1. Cognitivedevelopment:Stanford-Binet InteUigenceScale (1972 norms)27 2. Behavioral competence: Child Behavior Checklist for ages 2 or 3 years28'29; higher scores indicated more behavior problems 3. Health status: three measures a. Mother's report of child morbidity: the total number of all health conditions and illnessesreported by the mother during the entire 36-month period across all assessments b. Functional impact of health status: Functional Status II (R) Scale30,3~; higher scores indicated better health c. Maternal rating of child health: General Health Ratings Index32; higher scores were consistent with better health Covariates consisted of measures of socioeconomic status and neonatal morbidity measured before the randomization process. They included: birth weight, gender, length of neonatal stay adjusted for birth weight (Neonatal Health Index),33 maternal educational attainment, maternal age and maternal race (black, Hispanic, other [mostly white, with Asian being the other category]), and study site. Because the intervention might be less effective for children with cerebral palsy, for whom it was not designed, the children were further classified as to the presence of CP on the basis of the 36-month physical examination, as de-

M c C o r m i c k et al.

529

scribed elsewhere. 34 Using a specified examination and trained nurse practitioners, we first classified each child's condition as normal, suspect, or abnormal for a general clinical impression of neuromuscular status based on the assessment; if the classification was abnormal, a formal diagnosis was given, including CP. Analyses were then conducted for the total group and for those without CP. Analysis. Baseline characteristics of the infants in the INT and follow-up groups were checked for comparability (overall and within birth weight groups) by using two-sample t statistics for measured characteristics and simple chisquare statistics for categorical characteristics. 35 Because the prerandomization covariates were significantly associated with one or more outcomes, is the approach in this analysis was to estimate the value of each outcome as a function of these covariates. A multiple linear regression model was developed for expressing each outcome as a function of treatment (intervention vs follow-up) and of seven baseline covariates: site (eight categories); birth weight (in kilograms); gender; the Neonatal Health Index; maternal education level (three categories: did not complete high school, completed high school, and beyond high school); maternal age (years); plurality (twin, singleton); and maternal race. Residuals for the model were examined for outliers and overall fit. 36 The coefficient of the treatment term in the models measures the difference in mean response for the intervention group versus the follow-up group, with control for the seven covariates. Adjusted mean values (i.e., population marginal means) 37 for each group were computed as the expected response for a child with covariate values set at the mean for the total sample; thus the difference between the adjusted values equals the regression coefficient for treatment. Similarly, regression models within each birth weight subgroup were developed with birth weight eliminated from the covariate set. 38 All analyses were done with the Statistical Analysis Systems computer software. 39 RESULTS

Characteristics of sample. As seen in Table I, the sample of VLBW infants was heterogeneous but weighted by indicators of socioeconomic disadvantage; more than half were black, and almost half had mothers who had not completed high school. Of the total, cerebral palsy was diagnosed in 10% at the 36-month examination. No difference by any of these variables was seen by treatment group status. The characteristics of the smaller children (extremely low birth weight, _<1000 gm) were similar to those of the VLBW group in terms of the percentage of female infants but had a higher percentage of black infants. The median maternal age and the percentage of mothers who had completed high school were similar to those of the entire VLBW group. The mean Neonatal Health Index score was lower

530

M c C o r m i c k et al.

The Journal of Pediatrics October 1993

T a b l e i. Characteristics of V L B W infants Birth weight group VLBW (_<1500 gm) INT (n ~ 98)

%Male 51.0 %Distribution by race/ethaicity Black 65.3 Hispanic 11.2 Other (mostly white) 23.5 Maternal age (yr) 24 Median (range) (14-41.) %Distribution by maternal education High school graduation 32.6 %Twlns 22.4 Neonatal Health Index 99.0 Mean (+_SD) (_+ 18.4) Birth Weight (gm) 1215 Median (range) (640-1500) %CP diagnosed at 36 mo 10.2

FU Only (n = 182)

49.0 57.7 10.4 31.9 25 (14-43) 41.2 28.0 30.8 14.3 98.8 (_+ 16.5) 1210 (540-1500) 10.4

ELBW ( ~ I 0 0 0 gm) p

INI" (n = 23)

FU Only (n = 57)

p

0.74

43.5

47.4

0.75

65.2 8.7 26. l 26 (14-38)

71.9 10.5 17.5 26 (14-43)

34.8 39.1 26.1 0.0 97.4 ( _+14.1) 920 (640-1000) 8.7

38.6 29.8 31.6 14.0 96.0 ( _+17.1) 900 (540-1000} 15.8

0.33 0.86

0.89 0.08 0.93 0.15 0.95

0.68 0.85

0.72 0.06 0.72 0.24 0.4

FU, Follow-up group.

for the E L B W group than for all V L B W infants, and the percentage with CP was slightly higher (13.8). The major difference, however, was in the percentage of twins; 17.1% of all V L B W children but only 10.0% of the E L B W children were twins. No significant differences between I N T and follow-up groups were seen among the E L B W children for any of these characteristics, although there were no twins in the I N T group and 14.0% in the follow-up group (p = 0.06). Outcomes. The V L B W I N T group averaged substantially higher IQ scores, with a difference of more than seven IQ points in favor of the I N T group (Table !I). When children with CP were removed from the analysis, the advantage incurred by the I N T group increased. There was no significant difference in behavior problem scores. Likewise, no significant difference between the two groups was seen on any of the health status measures except for the reported morbidity, for which'the I N T gioUp averaged higher scores. In the context of the results with the other health status measures, this finding is ]n{erpreted as a higher rate of less serious, relatively brief illnesses in the I N T group. When similar analyses were conducted for the E L B W group, t h e advantage in IQ seen was similar in magnitude to that for the overall V L B W group, but the sample size was too small to allow confident conclusions. For those without CP, the behavior problem score was also significantly lower in the I N T group. Consistent with the overall results, the E L B W group also had higher morbidity rates, but only in the number of reported conditions.

DISCUSSION W e conclude that early educational intervention improved the performance of V L B W children on measures of cognitive development by the end of the intervention: the difference m IQ scores was similar to that previously reported for all infants whose birth weight was _<2000 gin. t5 This advantage was greater with the climination from the analysis of children with CP. who may not have been able to benefit from an intervention of this type. These results are consistent With those previously reported for V L B W children 15 in Which a significant difference between the I N T and follow-up groups in the proportion with IQs <85 was found, but not in the proportion with IQs <70. The latter group is heavily weighted by those wi~h CP. The results are also similar to those for the entire study 15 in terms of the observed differences, by treatment group, in the health status measures. No differences in serious morbidity, the functional impact of health problems, or maternaI assessment 0f child health werc encountered. An increase in reporting of less serious, brief morbidity was seen in the I N T group, and we have weviously documented the association betwecn this increase and (1) entry into the day care center and (2) a modest increase in the number of physicians' visits. 19, 4o These results are reassuring in view of the general increased risk of morbidity among V L B W children,7, 41 the attendant ample use of medical care]" 41-44 and the increased risk of illness associated with center-based day care. 45- 46 The magnitude of difference in behavior problem scores

The Journal o f Pediatrics Volume 123, Number 4

M c C o r m i c k et al.

53 1

Table II. Adjusted* m e a n scores a n d differences for V L B W children at 3 years of gestation-corrected age Birth weight group VLBW (_<1500 gm) Total Total INT FU only Mean IQ (Stanford Binet Scale) Age 3 yr INT FU only Difference (INT vs FU only) p Value Mean score: Child Behavior Checklist (age 3 yr) INT FU only Difference (INT vs FU only) p Value Mean score: mother's report of child morbidity (ages 0-3 yrt) INT FU only Difference (INT vs FU only) p Value Mean score: General Health Ratings Index INT FU only Difference (INT vs FU only) p Value Mean score: Functional Status lI (R) Scale INT FU only Difference (INT vs FU only) p Value

98 182

Excluding CP

ELBW (~
Excluding CP

88 163

23 57

21 48

87.5 80.3 7.2 0.002

91.5 81.6 9.9 <0.0003

87.2 80.1 7.1 0.17

93.1 84.8 8.3 0.07

46.4 48.1 -1.7 0.51

45.1 48.5 -3.4 0.21

45.1 51.9 -6.8 0.10

43.9 54.0 -10.1 0.02

9.49 8.01 1.48 0.01

9.27 7.85 1.42 0.02

10.52 8.62 1.90 0.06

10.56 8.39 2.17 0.03

25.6 26.3 -0.7 0.34

26.5 26.7 -0.2 0.77

24.1 24.9 -0.8 0.45

25.3 25.7 -0.4 0.69

0.97 0.96 0.01 0.31

0.90 0.94 -0.04 0.08

0.93 0.93 0.00 0.78

0.95 0.95 -0.000 0.96

FU, Follow-up group. *Generalized Estimating Equations (GEE) regression and population marginal means with covariates fixed at their average value in the total sample37'3s were used to adjust means and differences for site, gender, plurality, race/ethnicity, maternal education, maternal age, and Neonatal Health Index, and to account for within twin correlations. tReported morbidity is the sum of all reported illnesses,outpatient surgeries, hospitalizations, and injuries from birth to age 3 years.

was similar to t h a t reported for the overall study 15 a n d substantially greater for children who weighed _< 1000 g m at b i r t h a n d in whom a statistically significant decrease was found a m o n g those without CP. T h e failure to detect a significant difference for this outcome for the overall V L B W group m a y reflect statistical power; a difference of similar m a g n i t u d e achieved statistical significance for the entire group of children who weighed --<2000 gm at birth. ~5 In interpreting these results, we m u s t consider w h e t h e r the effect of the intervention on IQ scores could, in part, represent improved test-taking abilities, especially for those in bilingual homes. In addition, with the exception of the IQ tests, the outcome measures are dependent on m a t e r n a l re-

port and were not m a d e by observers u n a w a r e of study status. Less likely to influence the results are cohort attrition and crossover effects, because these were low. 15, 16 Also unlikely to influence the results are baseline factors considered in the process of randomization. Previous results15, 16 have shown t h a t the groups are well balanced for m a t e r n a l age and education, gender, b i r t h weight, and race/ethnicity. Moreover, the high retention rate at 36 m o n t h s (92.7%) minimizes biases resulting from attrition. A l t h o u g h less educated m o t h e r s were more likely to drop out of the study, the effect was similar in the I N T and follow-up groups. Differential retention within t r e a t m e n t groups tended to bias the I N T group with higher-risk

532

M c C o r m i c k et al.

infants (lower birth weight, younger mothers), but the effects were small. 16 In addition, these analyses represent a post hoc analysis on a subgroup not originally designated in the original design for independent analysis. In line with critiques and suggestions for such analyses, 47 the designation of the subgroup uses a prerandomization variable (birth weight). W e have, however, not corrected the p value for multiple testing. The interpretation of the effects on cognitive development are also constrained by the characteristics of the measure used, the Stanford-Binet Intelligence Scale, Third Edition. Several tests were reviewed, and this test was selected because of its extensive use with well-established predictive validity and comparability with outcome measures used in other early-intervention studies. Alternative tests shared the disadvantage of potentially outdated norms or less well established validity. 48, 49 Thus the significance of our results in terms of later school performance remains to be established. Further, the IQ scores among those in the I N T group were in the low average range, suggesting that the intervention did not completely offset the biologic and socioeconomic disadvantages of many of these children. Finally, assessment of the generalizability of the results must consider both the nature of the sample and the nature of the intervention. The sample is heterogeneous but is heavily weighted toward disadvantaged children. Results from the overall study suggest that little advantage of the intervention is seen for children of better-educated mothers, 21 and these findings may also pertain to V L B W children, as suggested by more descriptive reports, s, 10 In addition, the intervention program is very intensive, Is and the potential effects of less intensive and perhaps less costly programs require further assessment. Despite these reservations, these results support the effect of early educational intervention in improving the outcomes of V L B W children. If the effects prove as durable as those reported for disadvantaged, healthy children, 2 such early intervention may reduce the proportion of similar children currently experiencing school failure 11 and may be particularly important for children with both biologic and environmental v u l n e r a b i l i t i e s - - V L B W children born to disadvantaged families. Members of the Infant Health and Development Program include the following: Longitudinal Study Office: Cecelia McCarton, MD (Albert Einstein College of Medicine, Bronx, N.Y.), J. Brooks-Gunn, PhD, (Columbia University, New York, N.Y.), and David Scott, PhD (Yale University School of Medicine, New Haven, Conn.); Data Coordinating Center: James Tonaseia, PhD, and Curtis L. Meinert, PhD (Johns Hopkins School of Hygiene and Public Health, Baltimore, Md.); participating universities: University of Arkansas for Medical Sciences (Little Rock); Albert

The Journal o f Pediatrics October 1993

Einstein College of Medicine (Bronx, N.Y.), Harvard Medical School (Boston, Mass.), University of Miami School of Medicine (Miami, Fla.), University of Pennsylvania School of Medicine (Philadelphia), University of Texas Health Science Center at Dallas, University of Washington School of Medicine (Seattle), and Yale University School of Medicine (New Haven, Conn.); site directors: Charles R. Bauer, MD (Miami), Judy Bernbaum, MD (Pennsylvania), Patrick H. Casey, MD (Arkansas), Cecelia M. MeCarton, MD (Einstein), Marie McCormick, MD (Harvard), Clifford J. Sells, MD, and Forrest C. Bennett, MD (Washington), David T. Scott, PhD (Yale), and Jon E. Tyson, MD, and Mark Swanson, MD (Texas); and the Executive Committee: Cecelia McCarton, chair (Einstein), J. Brooks-Gunn, PhD, co-chair (Columbia University), Patrick H. Casey (Arkansas), David T. Scott, PhD (Yale), James Tonascia, PhD (Johns Hopkins), Curtis L. Meinert, PhD (Johns Hopkins), and Sam Shapiro, special consultant to the Longitudinal Study Office and ex-officio member of the Executive Committee (Johns Hopkins). REFERENCES

1. Meisels S J, Shonkoff JP, eds. Handbook of early childhood intervention. Cambridge: Cambridge University Press, 1990. 2. Darlington RB, Royce JM, Snipper AS, Murray HW, Lazar I. Preschool programs and later school competence of children from low-income families. Science 1980;208:202-4. 3. Shonkoff JP, Hauser-Cram P. Early intervention for disabled infants and their families: a quantitative analysis. Pediatrics 1987;80:650-8. 4. Olds DL, Kitzman H. Can home visitation improve the health of women and children at environmental risk? Pediatrics 1990;86:108-16. 5. Meisels S J, Wasik BA. Who should be served? identifying children in need of early intervention. In: Meisels S J, Shonkoff JP, eds. Handbook of early childhood intervention. Cambridge: Cambridge University Press, 1990. 6. Hoy EA, Bell JM, Sykes DH. Very low birth weight: a longterm developmental impairment? Int J Behav Dev 1988;11:3767. 7. McCormick MC. The contribution of low birthweight to infant mortality and childhood morbidity. N Engl J Med 1985;312:8290. 8. McCormick MC, Brooks-Gunn J, Workman-Daniels K, Turner J, Peckham GJ. The health and developmental status of very low-birth-weight children at school age. JAMA 1992; 267:2204-8. 9. Bennett FC. Neurodevelopmental outcome in low-birthweight infants: the role of developmental intervention. Clin Crit Care Med 1988;12:221-49. 10. Hunt JV, Cooper BAB, Tooley WH. Very low birth weight infants at 8 and 11 years of age: role of neonatal illness and family status. Pediatrics 1988;82:596-603. 11. McCormick MC, Gortmaker SL, Sobol AM. Very low birth weight children: behavior problems and school difficulties in a national sample. J PEDIATR 1990;117:687-93. 12. Saigal S, Szatmari P, Rosenbaum P, Campbell D, King S. Cognitive abilities and school performance of extremely low birth weight children and matched term control children at age 8 years: a regional study. J PEDIATR 1991;118:751-60. 13. Tirosh E, Rubino S. Physiotherapy for children with cerebral palsy. Am J Dis Child 1989;143:552-5.

The Journal o f Pediatrics Volume 123, Number 4

14. Turbull JD. Early intervention for children with or at risk of cerebral palsy. Am J Dis Child 1993;147:54-9. 15. Infant Health and Development Program. Enhancing the outcomes of low-birth-weight, premature infants: a multi-site, randomized trial. JAMA 1990;263:3035-42. 16. Constantine WL, Haynes CW, Spiker D, Kendall-Tackett K, Constantine NA. Recruitment and retention in a clinical trial for low birth weight, premature infants. J Dev Behav Pediatr 1993;14:1-7. 17. Brooks-Gunn J, Liaw F, Klebanov PK. Effects of early intervention on low birth weight, preterm infants: what aspects of cognitive functioning are enhanced? J PED~ATR 1992; 120:350-9. 18. Ramey CT, Bryant DM, Wasik BH, Sparling J J, Fendt KH, LaVange LM. The Infant Health and Development Program for low birth weight, premature infants: program elements, family participation, and child intelligence. Pediatrics 1992; 89:454-66. 19. McCormick MC, Bfooks-Gunn J, Shapiro S, Benasich AA, Black G, Gross RT. Health care use among young children in day care: results of a randomized trial of early intervention. JAMA 1991;265:2212-7. 20. Brooks-Gunn J, McCormick MC, Shapiro S, Benasich A, Black G. Effects of early intervention on maternal employment, public assistance and health insurance. Am J Public Health (in press). 21. Brooks-Gunn J, Gross RT, Kraemer HC, Spiker D, Shapiro S. Enhancing the cognitive outcomes of low birth weight, premature infants: for whom is the intervention most effective? Pediatrics 1992;89:1209-15. 22. ConstantineNA, Kraemer HC, Kendall-Tackett KA, Bennett FC, Tyson JE, Gross RT. Use of physical and neurologic observations in assessment of gestational age in low birth weight infants. J PEDIATR 1987;110:921-8. 23. Efron B. Forcing a sequential experiment to be balanced. Biometrika 1971;58:403-17. 24. Pocock S J, Simon R. Sequential treatment assignment with balancing for prognostic factors in the controlled clinical trial. Biometrics 1975;31:103-15. 25. Sparling J, Lewis I, Ramey CT, Wasik BH, Bryant DM, LaVange LM. Partners, a curriculum to help premature, low birth-weight infants get off to a good start. Top Early Child Spec Educ 1991;11:36-55. 26. Wasik BH, Bryant DM, Lyons DC. Home visiting. Newbury Park: Sage Publications, 1990. 27. Terman LM, Merrill MA. Stanford-Binet Intelligence Scale: manual for the third revision, Form L-M. Boston: HoughtonMifflin, 1973. 28. Achenbach TM, Edelbrock CS, Howell CT. Empirically based assessment of the behavioral/emotional problems of 2- and 3-year-old children. J Abnorm Child Psyehol 1987; 15:629-50. 29. McConaughay SH, Achenbaeh TM. Practical guide for the Child Behavior Checklist and related materials. Burlington: University of Vermont Department of Psychiatry, 1988. 30. Stein REK, Jessop DJ. A noncategorical approach to chronic childhood illness. Public Health Rep 1982;16:354-62. 31. Stein REK, Jessop DJ. Functional Status II (R): a measure of child health status. Med Care 1990;28:1041-55.

M c C o r m i c k et al.

533

32. Eisen M, Donald CA, Ware JE, Brook RH. Conceptualization and measurement of health for children in the Health Insurance Study. Santa Monica, California: The Rand Corp., 1980. (Rand publication series R-2313-HEW.) 33. Scott DT, Bauer CR, Kraemer HC, Tyson J. A neonatal health index for preterm infants [Abstract]. Pediatr Res 1989;25(pt 2):263A. 34. McCarton C, Bennett FC, Tonascia J, et al. Neurological status at 36 months of age. In: Gross RT, ed. Infant Health and Development Program. Palo Alto, California: Stanford University Press (in press). 35. Armitage P, Berry G. Statistical methods in medical research. 2nd ed. Oxford: Blackwell Scientific Publications, 1987. 36. McCullagh P, Nelder JA. Generalized linear models. 2rid ed. London: Chapman & Hall, 1989. 37. Searle SR, Speed FM, Miliken GA. Population marginal means in the linear model: an alternative to least squares means. The American Statistician, 1980;34:216-21. 38. Liang KY, Zeger SL. Longitudinal data analysis using generalized linear models. Biometrika 1986;73:13-22. 39. SAS statistics; version 6 ed. Cary, North Carolina: SAS Institute, 1985. 40. McCormick MC, Brooks-Gunn J, Shapiro S, Benasich AA, Black G, Gross RT. Secondary effects: health care use in IHDP. In: Gross RT, ed. Infant Health and Development Program. Palo Alto, California: Stanford University Press (in press). 41. Hack M, Blanche C, Rivers A, Fanaroff AA. The very low birth weight infant: the broader spectrum of morbidity during infancy and childhood. J Dev Behav Pediatr 1983;4:243-9. 42. McCormick MC, Shapiro S, Starfield BH. Rehospitalization in the first year of life for high-risk survivors. Pediatrics 1980;66:991-9. 43. Shankaran S, Cohen SN, Linver M, Zonia S. Medical care costs of high-risk infants after neonatal intensive care: a controlled study. Pediatrics 1988;81:372-8. 44. McCormick MC, Bernbaum JC, Eisenberg JM, Kustra SL, Finnegan E. Costs incurred by parents of very low birth weight infants after initial neonatal hospitalization. Pediatrics 1991; 88:533-41. 45. Haskins R, Kotch J. Day care and illness: evidence, costs and public policy. Pediatrics 1986;77:951-82. 46. Bell DM, Gleiber DW, Mercer AH, et al. Illness associated with day care: a study of incidence and cost. Am J Public Health 1989;104:101-3. 47. Yusuf S, Wittes J, Probstfield J, Tyroler HA. Analysis and interpretation of treatment effects in subgroups of patients in randomized clinical trials. JAMA 1991;266:93-8. 48. Scott DT. Rationale for selection of measures: cognitive development. In: Gross RT, ed. The Infant Health and Development Program for Low Birth Weight, Premature Infants. Palo Alto, California: Stanford University Press (in press). 49. Brooks-Gunn J. Identifying the vulnerable young child. In: Rogers DE, Ginzberg E, eds. Improving the life chances of children at risk. In: Proceedings of the Cornell University Medical College Sixth Conference on Health Policy. Boulder, Colorado: Westview Press, 1990.