American Journal of Epidemiology

Copyright 1997 by The Johns Hopkins University School of Hygiene and Public Health
All rights reserved

Vol. 146, No. 9

Printed in U.S.A.

Relation between Very Low Birth Weight and Developmental Delay among
Preschool Children Without Disabilities

Diana E. Schendel, 1 Joseph W. Stockbauer,2 Howard J. Hoffman,3 Allen A. Herman,4 Cynthia J. Berg,5 and
Wayne F. Schramm 2

birth weight; child development; infant, very low birth weight

It is widely recognized that very low birth weight

infants (VLBW: < 1,500 g) are at greater risk for
neonatal morbidity and mortality, long-term morbidity, and developmental disability than infants with
higher birth weight (1-4), although typically more

than 50 percent of VLBW survivors born in recent

years lack frank disabling conditions, such as cerebral
palsy, deafness, or blindness (5-10). However, even
VLBW children without disabling conditions are at
risk of developing a variety of cognitive, motor, and
behavioral problems (6, 11, 12), although many of
these problems may not be detected until school age
(13). Before school age, nondisabled VLBW children
may manifest impaired development as a failure to
achieve developmental milestones at the expected age.
For the large proportion of VLBW children who otherwise appear well it may be important for health care
providers to be alert to the magnitude of the risk for
developmental delay (DELAY) and the need for developmental screening and possible intervention (cf.
reference 14). To the authors' knowledge, however,
there have been no population-based estimates of the
risk for developmental delay among nondisabled
VLBW preschool children relative to their higher
birth weight peers, particularly in more recent cohorts
of VLBW children that may have benefited from
advances in neonatal treatment and technology.
VLBW infants are not a homogeneous group with
respect to gestational age and growth status (15, 16).
They include: 1) a small number of premature infants
born at less than 23 weeks gestation; 2) a majority (70

Received for publication July 8,1996, and accepted for publication June 26, 1997.
Abbreviations: AGA, appropriate-for-gestational age; DELAY,
developmental delay; Denver II, Denver Developmental Screening
Test II; MLBW, moderately low birth weight; MMIHS, Missouri Maternal and Infant Health Survey; NBW, normal birth weight; SGA,
small-for-gestational age; VLBW, very low birth weight.
1
Developmental Disabilities Branch, Division of Birth Defects
and Developmental Disabilities, National Center For Environmental
Health, Centers for Disease Control and Prevention, Atlanta, GA.
2
Bureau of Health Data Analysis, State Center for Health Statistics, Missouri Department of Health.
3
Epidemiology, Statistics and Data Systems Branch, Office of
the Director, National Institute on Deafness and Other Communication Disorders, National Institutes of Health, Bethesda, MD.
4
Epidemiology Branch, Division of Epidemiology, Statistics and
Prevention Research, National Institute of Child Health and Human
Development, National Institutes of Health, Bethesda, MD.
5
Pregnancy and Infant Health Branch, Division of Reproductive
Health, National Center for Chronic Disease Prevention and Health
Promotion, Centers for Disease Control and Prevention, Atlanta,
GA.
Reprint requests to Dr. Diana E. Schendel, Developmental Disabilities Branch, Division of Birth Defects and Developmental Disabilities, Centers for Disease Control and Prevention, 4770 Buford
Highway, N.E., Mailstop F15, Chamblee, GA 30341-3724.

740

Downloaded from http://aje.oxfordjournals.org/ by guest on March 20, 2015

The authors examined the relation between very low birth weight (VLBW: <1,500 g) and possible developmental delay (DELAY) in the absence of frank developmental disability among young children. The prevalence of DELAY in a population-based cohort (Missouri resident births born from December 1989 through
March 1991) of singleton VLBW children (n = 367) was compared with the prevalence of DELAY among both
moderately low birth weight (MLBW: 1,500-2,499 g; n = 553) and normal birth weight (NBW: >2,500 g; n =
555) singleton control children. DELAY was defined by nine measures of performance on the Denver
Developmental Screening Test II at a median adjusted age of 15 months (range: 9-34 months). Subjects were
asymptomatic for disabling conditions at developmental follow-up. Apparently well VLBW children were
consistently at greater risk for both moderate and severe measures of DELAY and for DELAY across four
functional areas than were either the MLBW (adjusted odds ratios: 1.4-2.7) or NBW children (adjusted odds
ratios: 2.1-6.3). The greatest prevalence of DELAY tended to be among appropriate-for-gestational age VLBW
children who were also the most premature. This study supports developmental follow-up of nondisabled
VLBW children because of the significantly elevated risk for DELAY among apparently normal infants. Am J
Epidemiol 1997; 146:740-9.

Very Low Birth Weight and Developmental Delay

MATERIALS AND METHODS

Study population

The MMIHS study population has been described in

detail (20). Briefly, it consisted of all births among
Missouri residents that occurred during a 16-month
period from December 1, 1989 through March 31,
1991. Case infants included all VLBW singleton livebirths in the study population identified either from
state birth certificate files or from the delivery room
entry logs of five major urban hospitals in Missouri
that provide services for inner city residents or for
women at high risk during their pregnancies. Hospitalbased identification for the five urban hospitals was
designed to improve study response rates among
mothers who delivered at these particular facilities by
Am J Epidemiol

Vol. 146, No. 9, 1997

obtaining consent and completing a maternal questionnaire prior to discharge; mothers of infants who were
born elsewhere and identified from the birth certificate
files completed the questionnaire by mail.
Both MLBW and NBW control infants were drawn
from singleton livebirths in the MMIHS study population in approximately a 1:1 case-to-control ratio.
Among infants who were not born in one of the five
urban hospitals, controls were randomly selected from
the birth certificate files on the basis of frequency
matching with cases by maternal race (black, nonblack), age (<20 years, 20-24 years, ^25 years), and
residence (residence in St. Louis City, St. Louis
County, Kansas City, or Jackson County vs. residence
elsewhere in Missouri). For the urban hospital sample,
once a case infant was identified from delivery room
logs, a control was identified by choosing the next
infant recorded in the log with the appropriate birth
weight and whose mother fit in the same race-age
stratum as the case infant's mother.
The sample population for this analysis was derived
from all singleton case and control children enrolled in
the MMIHS who completed the Denver II [i.e., survivors who, at the time of developmental follow-up, had
no frank physical or other limitations, specifically,
cerebral palsy or other orthopedic problem (n = 11),
chronic health condition (n = 4), nonresponsive (n =
3), homeless (n = 2), Down's syndrome (n = 1), blind
{n = 1), unspecified brain injury (n 1), and who
were not lost to follow-up as a result of adoption,
refusal to participate, or no reply]. All participants
were tested between the adjusted age (i.e., chronological age adjusted for prematurity if a child was aged
<2 years and was born > 2 weeks prematurely) of 9
and 34 months; about 50 percent of the children in
each birth weight group had been tested by the adjusted age of 15 months.
The Denver II was administered primarily by local
county health department staff in a child's home (preferred) or other location (e.g., the Women, Infants, and
Children program office). All staff who administered
the test were trained by nurses from the Missouri
Department of Health who had undergone a training
program established by developers of the Denver II.
Outcome definition

The Denver II is designed to screen children for

possible developmental delay by comparing a child's
performance on an array of tasks with that of other
children of the same adjusted age. Nine outcomes
indicating DELAY from the Denver II were used in
this analysis. Eight of the outcomes were based on two
measures of performance in each of four domains
(personal-social, language, fine motor-adaptive skills,

Downloaded from http://aje.oxfordjournals.org/ by guest on March 20, 2015

percent or more) of infants born between 23 and 30

weeks gestation, most with birth weights appropriatefor-gestational age (AGA); and 3) some infants (25
percent or less) born at more than 30 weeks gestation,
70 percent or more of whom are small-for-gestational
age (SGA). In comparison with SGA VLBW infants,
the typically more premature AGA VLBW infants are
more likely to experience certain adverse perinatal
conditions, such as low Apgar scores, intraventricular
hemorrhage, and artificial ventilation, and to have
poor neurologic outcome, such as cerebral palsy (15,
17). It remains to be seen whether the risk for DELAY
among nondisabled VLBW children would also differ
on the basis of gestational age and growth status at
birth.
The purpose of this study was to assess the prevalence of DELAY in a recent population-based cohort
of young singleton VLBW children and to compare it
with the prevalence of DELAY in two groups of
singleton control children whose birth weights were
higher: moderately low birth weight infants (MLBW:
1,500-2,499 g) and normal birth weight infants
(NBW: >2,500 g). Children in all three birth weight
groups were apparently well and lacked frank physical
limitations or developmental disabilities at the time of
developmental follow-up. Information concerning
DELAY was based on the children's performance on
the Denver Developmental Screening Test II (Denver
II), the 1990 revision of the original Denver Developmental Screening Test (18, 19), which has been widely
used in the United States and abroad and can be
administered feasibly to large samples. All data were
obtained from the National Institute of Child Health
and Human Development, Missouri Maternal and Infant Health Survey (MMIHS), a population-based
study of the prenatal risk factors and health care needs
of VLBW infants.

741

742

Schendel et al.

Growth status at birth

For this analysis, birth weight and gestational age

(assigned by the physician) were obtained from the
birth certificate. SGA was defined as birth weight
below the 10th percentile of birth weight for gestational age, by sex and race, on the basis of singleton
births in Missouri from 1989 through 1994, and AGA
was defined as birth weight at or above the 10th
percentile.
Covariates

Information on risk factors for DELAY were obtained from the maternal questionnaire and the child's
birth certificate. These factors included sex of the
child and maternal age, race, residence, education
level, and marital and Medicaid status at birth of the
child, and smoking and alcohol use. Although maternal race, age, and residence were used as matching
variables in selecting control children, it was decided
to include these variables as covariates in this analysis,
because, in terms of these risk factors, the distribution
of the Denver II sample population differed slightly
from that of the original set of case and control infants.
Smoking and alcohol use were measured by two vari-

ables on the basis of self-reporting by the mothers:

whether the mother 1) smoked (any amount) during
the index pregnancy and/or 2) drank alcoholic beverages (any amount) during the index pregnancy. The
prevalence of smoking or alcohol use during pregnancy as reported by mothers in the MMIHS was
comparable to the prevalence of antepartal tobacco or
alcohol exposure detected by urine cotinine sampling
and hospital chart review, respectively, among mothers of low birth weight infants in the Missouri 1993
Perinatal Substance Abuse Study (21).
Analytic techniques

Odds ratios were used to estimate the relative risk

for DELAY when comparing the VLBW children with
either the MLBW or NBW children; adjusted odds
ratios and their 95 percent confidence intervals were
calculated for each outcome variable using unconditional logistic regression to control for multiple confounders. Each multivariable regression model included at least eight risk factors as covariates.
RESULTS

Table 1 presents, by birth weight group, sample

population sizes achieved at each step in the derivation
of the MMIHS Denver II sample. About 75 percent of
all potential subjects in each birth weight group were
actually enrolled in the MMIHS. During the interval
between enrollment and follow-up for the Denver II,
many of the VLBW children died (n 250). Among
the survivors, about 70 percent of both case and control children completed the Denver II.
Among all potential MMIHS subjects or all enrolled
survivors and those who completed the Denver II,
differences in the proportion of subjects in each birth
weight group at each covariate level were small (0.15.5 percent). Children in the Denver II sample population had slightly lower proportions of mothers who
were 1) in their teens, 2) had not completed high
school, 3) were African American, 4) not married, 5)
were receiving Medicaid, 6) lived in rural areas, or 7)
smoked during pregnancy.
Table 2 presents, by birth weight group, the distribution of risk factors for DELAY among children in
the Denver II sample. Among mothers of VLBW
infants, a significantly smaller proportion were married when their children were born than were mothers
of NBW infants, and a significantly smaller proportion
of mothers of VLBW infants were receiving Medicaid
at the time of their children's birth than were the
mothers of MLBW infants.
For children in the Denver II sample, table 3 shows
the mean birth weight and gestational age of each birth
Am J Epidemiol

Vol. 146, No. 9, 1997

Downloaded from http://aje.oxfordjournals.org/ by guest on March 20, 2015

and gross motor skills). One of the two domainspecific measures was whether a child failed one or
more tasks in each domain for which 75-90 percent of
children of the same adjusted age would pass (denoted
as receiving one or more caution scores in a given
domain); the other measure was whether a child failed
one or more tasks in each domain for which at least 90
percent of children of the same adjusted age would
pass (denoted as receiving one or more delay scores in
a given domain). The ninth outcome, denoted as overall test performance, was based on the total number of
caution and/or delay scores received across all domains and was categorized as follows: 1) questionablereceived two or more caution scores and/or a
maximum of one delay score, 2) abnormalreceived
two or more delay scores, 3) normalreceived a maximum of one caution score, and 4) untestablerefused
to perform one or more tasks.
The nine outcomes reflect two basic levels of
DELAY: 1) a moderate degree of DELAY that was
generally represented by a questionable overall test
performance, plus the four domain-specific outcomes
for children who received one or more caution scores
in a given domain, and 2) a severe degree of DELAY
that was represented by an abnormal overall test performance, plus the four domain-specific outcomes for
children who received one or more delay scores in a
given domain.

Very Low Birth Weight and Developmental Delay

743

TABLE 1. Derivation of the MMIHS Denver II sample* of births to Missouri residents from December 1,
1989 through March 31,1991
VLBWf
Potential MMIHS subjects*
Enrolled MMIHS subjects
Survivors at time of follow-up among
enrolled subjects
Survivors with Denver II follow-up
Loss to follow-up for Denver II among
survivors due to
No reply/refusal
Adoption
Physical/other limitation

MLBWt

NBWt

No.

No.

No.

1,013
780

77H

1,061
800

75H

1,040
798

77H

530
367

68#
69**

787
553

98#
70**

793
555

99#
70**

147
3
13

220
5
9

234
3
1

weight group as well as the number of children who

were born either SGA or AGA. Of all VLBW,
MLBW, and NBW children, 24 percent, 46 percent,
and 7 percent, respectively, had been SGA infants.
Because a much larger proportion (99 percent) of
VLBW children were born prematurely (i.e., at <37
weeks gestation) than either the proportion (61 percent) of MLBW or the proportion (5 percent) of NBW
children, table 3 also shows the growth status of these
children, by birth weight group. Of the VLBW,
MLBW, and NBW children who were born prematurely, 24 percent, 15 percent, and 0 percent, respectively, were SGA.
Table 4 presents the distribution of children classified in each developmental delay outcome category by
birth weight group. The proportions of children with
each outcome tended to increase with decreasing birth
weight such that the proportions of VLBW children
with these outcomes (718 percent) were typically one
and a half to two times greater than the corresponding
proportions for the MLBW group (3-12 percent) and
two to four times greater than the proportions for the
NBW group (2-10 percent). These differences in outcome distribution were significant (p ^ 0.05), except
for the differences between the VLBW and MLBW
groups in the proportions of children who received 1)
one or more delays in the language domain, 2) one or
more cautions in the fine motor-adaptive domain, or 3)
one or more delays in the fine motor-adaptive domain.
In addition, in comparison with MLBW or NBW
children, there was a significantly lower proportion of
VLBW children with no cautions or delays.
Am J Epidemiol

Vol. 146, No. 9, 1997

Table 5 shows the crude distributions of children by

DELAY category for SGA and AGA children. These
data were limited to VLBW and MLBW children who
were born prematurely. Regardless of growth status at
birth, VLBW children who were born prematurely
typically had a higher prevalence of each measure of
DELAY than did MLBW children who were born
prematurely. These birth weight group differences
were significant for four of the outcome measures
among the AGA children but, possibly because of
small sample sizes, the birth weight group differences
were significant in only one outcome comparison
among the *SGA children. In terms of the overall test
performance measure, those children with the least
favorable prognosis (i.e., the highest prevalence of
DELAY) were VLBW and AGA; these children, on
average, were the most premature, although of similar
birth weight than their older, growth retarded VLBW
SGA peers. Those children with the most favorable
developmental prognosis (i.e., the lowest prevalence
of DELAY) were MLBW and AGA; these children,
on average, were born about 2.5 weeks later in gestation than the VLBW SGA children and had a higher
mean birth weight than their older, but growth retarded
MLBW SGA peers. Within either the VLBW or
MLBW groups, however, the differences in DELAY
prevalence between SGA and AGA children across the
different outcome measures typically were small.
Table 6 presents the adjusted odds ratios and 95
percent confidence intervals derived from the multivariable analysis of each of the DELAY outcomes.
Sample sizes were not adequate to stratify these anal-

Downloaded from http://aje.oxfordjournals.org/ by guest on March 20, 2015

* MMIHS Denver II sample, children enrolled in the Missouri Maternal and Infant Health Survey who
completed the Denver Developmental Screening Test II.
t VLBW, very low birth weight (<1,500 g); MLBW, moderately low birth weight (1,500-2,499 g); NBW, normal
birth weight (>2,500 g).
$ Singleton births originally identified as case or control infants.
Subjects whose mothers returned maternal survey questionnaire.
H Percent of potential subjects.
# Percent of enrolled subjects.
** Percent of all survivors at follow-up.

744

Schendel et al.
TABLE 2. Distribution of risk factors for developmental delay for children in the MMIHS Denver II
samplef of births to Missouri residents from December 1,1989 through March 31 ,1991
Risk
factor

NBWt
(n=555)

MLBWt
(n = 553)

No.

No.

No.

180
187

49.1
50.9

253
300

45.8
54.2

290
265

52.2
47.8

86
245
36

23.4
66.8
9.8

119
394
40

21.5
71.3
7.2

131
385
39

23.6
69.4
7.0

105
257

29.0
71.0

158
388

28.9
71.1

148
403

26.9
73.1

130
226

36.5
63.5

205
343

37.4
62.6

221
325

40.5
59.5

181
185

49.5
50.5

269
284

48.6
51.4

310
241

56.3
43.7

132
220

37.5
62.5

267
277

49.1
50.9

236
309

43.3
56.7

131
39
197

35.7
10.6
53.7

175
42
336

31.6
7.6
60.8

207
35
313

37.3
6.3
56.4

137
219

38.5
61.5

241
300

44.6
55.4

185
366

33.6
66.4

93
260

26.4
73.6

150
395

27.5
72.5

149
398

27.2
72.8

* VLBW vs. NBW, p = 0.05; ** VLBW vs. MLBW, p = 0.001.

t MMIHS Denver II sample, children enrolled in the Missouri Maternal and Infant Health Survey who completed
the Denver Developmental Screening Test II.
t VLBW, very low birth weight (<1,500 g); MLBW, moderately low birth weight (1,500-2,499 g); NBW, normal
birth weight (22,500 g).

yses on growth status. The odds ratios associated with

the comparison of VLBW with NBW children (2.16.3; all statistically significant) were consistently
larger than the corresponding values from the comparison of VLBW children with MLBW children (1.42.7). Notably, the odds ratios associated with measures
of a severe degree of DELAY (i.e., abnormal overall
test performance or, within each domain, receiving
one or more delay scores) were larger than the corresponding measures of a moderate degree of DELAY
(i.e., questionable overall test performance or, within
each domain, receiving one or more caution scores);
the odds ratios associated with the gross motor domain
also tended to be larger than corresponding measures
in the other three domains.
The adjusted age range at developmental follow-up
was broad (9-34 months). Crude odds ratios (small
sample sizes did not permit a multivariable analysis)
based on a subset of children to whom the Denver II
was administered by an adjusted age of 15 months

(about 50 percent of the Denver II sample) did not

differ appreciably from those based on the children
representing the entire adjusted age range of the full
Denver II sample.
DISCUSSION

The design of this study permitted a comparison of

the prevalence of possible developmental delay among
a recent population-based cohort of VLBW children
and two groups of children whose birth weights were
higher. Approximately 25 percent of the VLBW children had either a questionable or abnormal overall test
performance on the Denver H These data clearly
showed that, even among apparently well children,
there was a consistently higher risk for all measures of
DELAY among VLBW infants than either MLBW or
NBW infants; the greatest differences (adjusted odds
ratio range: 2-6) were observed between the VLBW
and NBW children. These differences remained after
Am J Epidemiol

Vol. 146, No. 9, 1997

Downloaded from http://aje.oxfordjournals.org/ by guest on March 20, 2015

Sex
Male
Female
Maternal age (years)
<20
20-34
235
Maternal education
<High school
>High school
Maternal race
Black
Nonblack
Married at birth*
Yes
No
Medicaid client**
Yes
No
Maternal residence
Urban
Suburban
Rural
Smoke any trimester
Yes
No
Alcohol any trimester
Yes
No

VLBW*
(n = 367)

Very Low Birth Weight and Developmental Delay

745

TABLE 3. Measures of fetal growth and maturity for children in the MMIHS Denver II sample* of births
to Missouri residents from December 1,1989 through March 31,1991
Fetal

VLBWt

MLBWt

NBWt

Mean

SDt

Mean

SD

Mean

SD

Birth weight (g)

1,088

(268)

2,184

(267)

3,414

(485)

Gestationai age (weeks)

28.4

(" = 553)
(2.8)

39.4

measure

(n = 367)
(3.0)

35.6

(n = 365)
Growth status at birth

(n = 555)

(n = 546)

No.

No.

All children
SGA*
AGA*

353
86
267

24.4
75.6

541
248
293

45.8
54.2

Children born prematurely

SGA
AGA

350
83
267

23.7
76.3

332
50
282

(1.5)
(n = 543)

15.1
84.9

No.

534
36

6.7
93.3

498
27
0

0.0

27

100.0

TABLE 4. Crude distribution of outcome measures indicating developmental delay for children in the
MMIHS Denver II samplet of births to Missouri residents from December 1,1989 through March 31,1991
Outcome
measure

MLBWJ
(n = 553)

VLBWt
(n = 367)
No.

No.

NBWt
(n = 555)
%

No.

Overall performance
Questionable*
Abnormal**
Normal
0 caution/delay***
1 caution only
Untestable

64
40

17.4
10.9

65
32

11.7
5.8

44
16

7.9
2.9

204
57
2

55.6
15.5
0.5

384
69

69.4
12.5
0.5

412
82
1

74.2
14.8

65
15

11.8
2.7

53
12

0.2

Domain-specific performance
Personal-social
>1 cautions*
1 delays**
Language
21 cautions*
1 delays****
Fine motor-adaptive
>1 cautions*****
21 delays*****
Gross motor
21 cautions***
>1 delays***

64
26

17.5

62
32

17.0

66
32

12.0
5.8

47
22

8.5

8.8

44
29

12.0
7.9

48
29

8.7
5.3

33
12

6.0
2.2

64

17.5
10.7

49
22

8.9
4.0

22
10

4.0
1.8

39

7.1

9.6
2.2

4.0

* VLBW vs. NBW, p <, 0.001 and VLBW vs. MLBW, p <, 0.05; ** VLBW vs. NBW, p <, 0.001 and VLBW vs.
MLBW, p <, 0.01; ** VLBW vs. NBW, p <, 0.001 and VLBW vs. MLBW, p <, 0.001; **** VLBW vs. NBW, p <; 0.01;
***** VLBW vs. NBW, p 5S 0.001.
t MMIHS Denver II sample, children enrolled in the Missouri Maternal and Infant Health Survey wtio completed
the Denver Developmental Screening Test II.
t VLBW, very low birth weight (<1,500 g); MLBW, moderately low birth weight (1,500-2,499 g); NBW, normal
birth weight (>2,500 g).
Testable children (VLBW, n = 365; MLBW, n = 550; NBW, n = 554).

Am J Epidemiol

Vol. 146, No. 9, 1997

Downloaded from http://aje.oxfordjournals.org/ by guest on March 20, 2015

* MMIHS Denver II sample, children enrolled in the Missouri Maternal and Infant Health Survey who completed
the Denver Developmental Screening Test II.
t VLBW, very low birth weight (<1,500 g); MLBW, moderately low birth weight (1,500-2,499 g); NBW, normal
birth weight (>2,500 g).
$ SD, standard deviation; SGA, small-for-gestational age (<10th percentile of birth weight for gestationai age,
by sex and race); AGA, appropriate-for-gestational age (10 percentile of birth weight for gestationai age, by sex
and race).

746

Schendel et al.
TABLE 5. Crude distribution by growth status at birth of outcome measures indicating developmental
delay for children in the MMIHS Denver II sample* of births to Missouri residents from December 1,1989
through March 31,1991, VLBW* and MLBW* children born prematurely only
Outcome
measure

Birth weight, mean (g)

Gestationai age, mean (weeks)
Overall performance (%)
Questionable*
Abnormal
Normal
21 caution/delay*
21 caution only
Untestable

AGA
MLBW
(n=50)

VLBW
(n = 267)

MLBW
(n = 282)

1,080
31.1

1,993
35.4

1,094
27.6

2,114
33.6

8.4
13.3

16.0
4.0

21.0
10.1

11.7
6.4

57.8
20.5
0.0

68.0
12.0
0.0

54.7
13.5
0.8

69.2
12.4
0.4

24.1
7.2

14.0
4.0

16.2
7.5

12.5
2.1

14.5
8.4

8.0
2.0

17.4
9.1

12.1
5.3

12.1
10.8

2.0
6.0

12.1
7.6

8.9
6.1

19.3
7.2

10.0
2.0

17.4
11.7

8.9
3.9

* AGA, VLBW vs. MLBW, p <, 0.01; **SGA, VLBW vs. MLBW, p < 0.05.
t MMIHS Denver II sample, children enrolled in the Missouri Maternal and Infant Health Survey who completed
the Denver Developmental Screening Test II.
% VLBW, very low birth weight (<1,500 g); MLBW, moderately low birth weight (1,500-2,499 g).
SGA, small-for-gestational age (<10th percentile of birth weight for gestationai age, by sex and race); AGA,
appropriate-for-gestational age (210th percentile of birth weight for gestationai age, by sex and race).
H Testable children (VLBW: SGA, n = 83; AGA, n = 265. MLBW: SGA, n = 50; AGA, n = 281).

controlling for the potential confounding effects of

sex, selected demographic and socioeconomic characteristics of the mother, and smoking and alcohol use of
the mother during pregnancy.
After restricting the comparison to VLBW and
MLBW children who were born prematurely, it was
also found that VLBW children typically had a greater
prevalence of DELAY than MLBW children, regardless of growth status at birth. The VLBW AGA children (who actually were the most premature) tended to
have the greatest prevalence of DELAY, but differences in DELAY prevalence by outcome between
VLBW AGA and SGA children were small.
The Denver II protocol provides a score to reflect
overall test performance as well as subscores for performance in each of four functional areas. For each of
the comparisons of VLBW children with MLBW or
NBW children, the greater risk for DELAY among
VLBW children in this sample was apparent in measures of both moderate and severe levels of DELAY,
but particularly marked at the severe levels. Furthermore, VLBW children manifested DELAY in all areas

of function, with the greatest risk for DELAY associated with the gross motor domain. This is consistent
with previous reports of poor motor functioning or
"motor clumsiness" in VLBW children without major
neurodevelopmental impairment (6, 9, 12), but may
indicate a greater risk for delay in acquiring gross
motor skills than fine motor skills among very young,
preschool-aged VLBW children.
This analysis drew upon a case-control study of
liveborn infants that included an entire 16-month cohort of VLBW infants born in Missouri. Despite the
relatively large follow-up sample size, one limitation
of the study may be that the Denver II sample may not
have been representative of the larger population from
which it was drawn. For selection bias to have an
impact on the study results, however, the bias would
have had to have been related to both birth weight and
developmental status (22). Mothers of VLBW children
who suspected or were aware of delayed development
in their children may have been particularly motivated
to participate in the follow-up study, thereby artificially increasing the prevalence of delay in the VLBW
Am J Epidemiol

Vol. 146, No. 9, 1997

Downloaded from http://aje.oxfordjournals.org/ by guest on March 20, 2015

Domain-specific performance (%)H

Personal-social
21 cautions
21 delays*
Language
21 cautions
21 delays
Fine motor-adaptive
21 cautions**
21 delays
Gross motor
21 cautions*
21 delays*

SGA
VLBW
(n = 83)

Very Low Birth Weight and Developmental Delay

747

TABLE 6. Adjusted odds ratios (95% confidence intervals) for outcome measures indicating
developmental delay for children in the MMIHS Denver II sample* of births to Missouri residents from
December 1,1989 through March 31,1991
VLBWf (n = 320) vs.
Outcome
measure

Overall performance
Questionable
Abnormal

NBWt
(n = 524)

OR*

95% CI*

OR

1.66
2.02

1.09-2.51
1.18-3.45

2.74
4.81

1.74-4.31
2.51-9.23

1.64
2.74

1.09-2.48
1.36-5.53

2.12
3.21

1.38-3.24
1.54-6.68

1.41
1.79

0.93-2.12
1.04-3.09

2.16
2.97

1.39-3.37
1.61-5.47

1.42
1.60

0.88-2.28
0.90-2.84

2.10
4.88

1.26-3.50
2.34-10.20

2.16

1.39-3.34
1.38-4.68

4.95
6.26

2.89-8.47
2.87-13.65

2.54

95% Cl

* MMIHS Denver II sample, children enrolled in the Missouri Maternal and Infant Health Survey who completed
the Denver Developmental Screening Test II.
t VLBW, very low birth weight (<1,500 g); MLBW, moderately low birth weight (1,500-2,499 g); NBW, normal
birth weight (>2,500 g).
t OR, odds ratio; Cl, confidence interval.
In calculating the adjusted odds ratio, residence was dropped from the multivariable regression model.

group relative to the other birth weight groups. Because all children to be tested with the Denver II had
to appear well and have no known physical limitations, however, it seems unlikely that selection bias
with respect to developmental status, alone or in combination with birth weight status, could have been
large.
Alternatively, bias in the follow-up sample could
have occurred with respect to other risk factors for
DELAY. However, estimates of the differences in the
distribution of risk factors for DELAY between the
larger pools of potential subjects or enrolled survivors
and the final Denver II sample suggest that the demographic and socioeconomic characteristics of children
in the samples were comparable, although children in
the Denver II sample tended to be less disadvantaged
socioeconomically. The results of this analysis possibly reflect a population that, overall, is at no greater
risk and perhaps is at a somewhat lower risk of
DELAY than is the larger population from which it
was drawn.
There was little apparent confounding by the nine
risk factors for DELAY that were included in the
multivariable analysis. In part, this was due to the
striking homogeneous distribution of risk factors
across the birth weight groups. In terms of the effects
of smoking and alcohol use, the variables we used
Am J Epidemiol

Vol. 146, No. 9, 1997

distinguished users from nonusers, but could not adjust for the effects of differences in the pattern of use
on the risk for DELAY. Finally, because the Denver II
sample was quite young as a whole (median adjusted
age, 15 months), the influence of socioeconomic factors on development may not have been as strong as
has been observed in older preschool- and school-aged
children (10, 23).
Prior to administering the Denver II, the tester must
determine the adjusted age of the child, a determination that is based on the child's gestational age. Thus,
by virtue of being aware of a child's prematurity, a
tester is not completely unaware of the child's birth
weight status. This knowledge could bias a tester's
judgment of the child's performance. It is possible that
the MMIHS testers (local county health department
staff trained by nurses from the Missouri Department
of Health) could have had a negative bias toward the
performance of children born prematurely, thereby
artificially increasing the prevalence of delay in the
VLBW group relative to the other birth weight groups.
On the other hand, the fact that all children, including
those born prematurely, were presumably well and
without any overt physical impairment could have had
the opposite effect.
In results of a study by Glascoe et al. (24) to assess
the accuracy of the Denver n, the screening test was

Downloaded from http://aje.oxfordjournals.org/ by guest on March 20, 2015

Domain-specific performance
Personal-social
1 cautions
1 delays
Language
>1 cautions
>1 delays
Fine motor-adaptive
>1 cautions
>1 delays
Gross motor
1 cautions
>1 delays

MLBWt
(n = 512)

748

Schendel et al.

ACKNOWLEDGMENTS

This work was supported in part by the National Institute

of Child Health and Human Development contract No.
NO1-HD-6-2916 and the Centers for Disease Control and
Prevention (CDC).
The authors thank Dr. Vicky Howell, Roberta Boley,
Linda Tellman, Meschelle Lairmore, Ruth Kliethermes, Dr.
Robert Metzger, Fern Lemaster, Jean Kitchen, Jeanne
McDonald, and Philip d'Almada and the staffs of local,

county, and city health departments in Missouri for their

assistance in this study. We also thank Drs. Pierre Decoufle'
and Marshalyn Yeargin-Allsopp for their thoughtful critiques of the manuscript.

REFERENCES
1. McCormick MC. The contribution of low birth weight to
infant mortality and childhood morbidity. N Engl J Med
1985;312:82-90.
2. McCormick MC, Brooks-Gunn J, Workman-Daniels K, et al.
The health and developmental status of very low-birth-weight
children at school age. JAMA 1992;267:2204-8.
3. Hack M, Weissman B, Breslau N, et al. Health of very low
birth weight children during their first eight years. J Pediatr
1993;122:887-92.
4. Hagberg B, Hagberg G, Olow I. The changing panorama of
cerebral palsy in Sweden. VI. Prevalence and origin during the
birth year period 1983-1986. Acta Paediatr 1993;82:387-93.
5. Aylward GP, Pfeiffer SI, Wright A, et al. Outcome studies of
low birth weight infants published in the last decade: a metaanalysis. J Pediatr 1989;115:515-20.
6. Saigal S, Szatmari P, Rosenbaum P, et al. Intellectual and
functional status at school entry of children who weighed 1000
grams or less at birth: a regional perspective of births in the
1980s. J Pediatr 1990; 116:409-16.
7. Escobar GJ, Littenberg B, Pettiti D. Outcome among surviving very low birthweight infants: a meta-analysis. Arch Dis
Child 1991;66:204-ll.
8. Veen S, Ens-Dokkum MH, Schreuder AM, et al. Impairments,
disabilities, and handicaps of very preterm and very-lowbirthweight infants at five years of age. Lancet 1991;338:
33-6.
9. Scottish Low Birthweight Study Group. The Scottish Low
Birthweight Study: I. Survival, growth, neuromotor and sensory impairment. Arch Dis Child 1992,67:675-81.
10. Weisglas-Kuperus N, Baerts W, Smrkovsky M, et al. Effects
of biological and social factors on the cognitive development
of very low birthweight children. Pediatrics 1993;92:658-65.
11. McCormick MC, Gortmaker ST, Sobol AM. Very low birth
weight children: behavior problems and school difficulty in a
national sample. J Pediatr 1990;l 17:687-93.
12. Sommerfelt K, Ellertsen B, Markestad T. Personality and
behavior in eight-year-old, non-handicapped children with
birth weight under 1500 g. Acta Paediatr 1993;82:723-8.
13. den Ouden AL, Hille ETM, Bauer L, et al. School performance in very preterm children. Lancet 1993;342:550-l.
14. Committee on Children with Disabilities. Screening infants
and young children for developmental disabilities. Pediatrics
1994;93:863-5.
15. Arnold CC, Kramer MS, Hobbs CA, et al. Very low birth
weight: a problematic cohort for epidemiologic studies of very
small or immature neonates. Am J Epidemiol 1991;134:
604-13.
16. Berg CJ, Zupan J, d'Almada PJ, et al. Gestational age and
intrauterine growth retardation among white and black very
low birthweight infants: a population-based cohort study. Paediatr Perinat Epidemiol 1994;8:53-61.
17. Veelken N, Stollhof K, Claussen M. Development and perinatal risk factors of very low-birth-weight infants. Small versus appropriate for gestational age. Neuropediatrics 1992;23:
102-7.
18. Frankenburg WK, Dodds J, Archer P, et al. Denver II technical manual. Denver, CO: Denver Developmental Materials
Inc, 1990.
19. Frankenburg WK, Dodds J, Archer P, et al. The Denver II: a
major revision and restandardization of the Denver Developmental Screening Test. Pediatrics 1992;89:91-7.
Am J Epidemiol Vol. 146, No. 9, 1997

Downloaded from http://aje.oxfordjournals.org/ by guest on March 20, 2015

criticized for identifying many children with abnormal

or questionable test results. These children were not
assigned an adverse developmental diagnosis on the
basis of their concurrent performance on a battery of
more specific developmental tests that were designed
to detect such conditions as mental retardation,
speech-language impairments, and learning disabilities. In the Glascoe et al. study, Denver II sensitivity
was fairly high (83 percent), but specificity (43 percent) and predictive value positive (23 percent) were
limited. The authors concluded that, unlike the original
Denver Developmental Screening Test (25), the Denver II identifies most children with developmental
problems, including those with subtle developmental
difficulties (24).
The level of accuracy of the Denver II should affect
the internal validity of the results for this study only if
the sensitivity and specificity levels of the test are
different for different birth weight groups. The Denver
II is designed to measure the acquisition of skills with
age and includes a correction for prematurity. To the
extent that the differences in birth weight in this study
were linked to differences in gestational age at birth,
these differences were accounted for by the correction.
Thus, there is no reason to suspect that the relative
differences in the performance on the Denver II observed between the birth weight groups are not valid;
however, the results may overestimate the absolute
prevalence of DELAY for all children that might be
obtained by more specific developmental tests.
A more important question is whether the level of
DELAY observed here is associated with academic
performance or with problems in development among
children at later ages. Although the ability to predict
school performance by a preschool screening tool may
be problematic (26), poor performance on preschool
screening tests has been associated with a significantly
increased risk for school-age problems (26, 27). If the
latter is true, then this study would support the perception that developmental surveillance of apparently
well VLBW children by their health care providers
may be warranted, thereby providing the opportunity
for early developmental intervention.

Very Low Birth Weight and Developmental Delay

20. Schramm WF, Stockbauer JW, Hoffman HJ. Exercise, employment, other daily activities, and adverse pregnancy outcomes. Am J Epidemiol 1996;143:211-18.
21. Dempsey ME, Schlechte T, Stockbauer JW, et al. Prevalence
and implications of perinatal substance use in Missouri. Mo
Med 1996;93:292-9.
22. Kelsey JL, Thompson WD, Evans AS. Methods in observational epidemiology. New York: Oxford University Press, 1980.
23. Lee H, Barratt MS. Cognitive development of preterm low
birth weight children at 5 to 8 years old. J Dev Behav Pediatr
1993; 14:242-9.
24. Glascoe FP, Byrne KE, Ashford LG, et al. Accuracy of the

749

Denver-H in developmental screening. Pediatrics 1992;89:

1221-5.
25. Meisels SJ. Can developmental screening tests identify children who are developmentally at risk? Pediatrics 1989;83:
578-85.
26. Lindahl E, Michelsson K, Dormer M. Prediction of early
school-age problems by a preschool neurodevelopmental
examination of children at risk neonatally. Dev Med Child
Neurol 1988;30:723-34.
27. Drillen CM, Pickering RM, Drummond MB. Predictive value
of screening for different areas of development. Dev Med
Child Neurol 1988;30:294-305.

Downloaded from http://aje.oxfordjournals.org/ by guest on March 20, 2015

Am J Epidemiol

Vol. 146, No. 9, 1997