Download Assessing Denver Developmental Screening & Stanford-Binet Tests for Preschool Intelligence and more Lecture notes Pathology in PDF only on Docsity!
Evaluation of 298 preschool children
The Denver Developmental^ Screening^ Test
Compared With the Stanford-Binet Test
JOHN A. GRANT, M.D., M.P.H., and ALAN M.^ GITTELSOHN,^ Ph.D.
SYSTEMATIC screening of
patient populations^ to^ facili-
tate clinical services^ had^ its^ prac-
tical beginning with^ the^ Kaiser-
Permanente Clinic^ (1)^ in^ the
early 1950's.^ This^ approach^ had
not been^ seriously^ applied^ to
pediatric age^ groups,^ however,
until Frankenburg and Dodds^ (2) introduced the Denver Develop- mental Screening Test^ (DDST) in 1967. In recent years, this^ test has become widely used^ by a
Dr. Grant is^ chief^ physician^ and
deputy director, Community^ Pedi-
atric Center,^ University^ of^ Mary-
land School of Medicine, and
former chief,^ school^ health^ sec-
tion, division of maternal and
child health, Maryland State^ De-
partment of Health. Dr.^ Gittel-
sohn is professor of biostatistics,
Johns Hopkins University School
of Hygiene and Public^ Health.
Tearsheet requests to^ John^ A.
Grant, M.D.,^ Department^ of
Pediatrics, University of^ Mary-
land School^ of^ Medicine^ (attn:
Mrs. Mantik), Redwood^ and
Greene Sts., Baltimore,^ Md.
variety of professional and non-
professional workers in pediatric
clinics, at maternal and child
health conferences, and^ in^ chil-
dren and youth projects. In their early publication, Frankenburg and Dodds com-
pared the performance of^ the
Denver Developmental^ Screen-
ing Test^ in a^ study^ population^ of 18 children to the Revised Yale
Developmental Schedule (3).
They found high correlations but
did not examine the performance
of the test in more explicit ad-
ministrative terms of pathology
discovered or pathology missed.
We show how^ the perform-
ance of the Denver Developmental
Screening Test was evaluated for
a specific population and relate
this kind of evaluation to the
needs of the service program ad-
ministrator.
Method and Materials
The test was given to^ 590 chil-
dren, 31/2^ to^ 4½/2^ years^ of^ age,
during an epidemiologic study of
congenital rubella^ (the^ parent
study) in the summer of 1968
(4). The^ children^ were^ chosen
to represent the population of^ a semirural Maryland county;^ they came from^ a^ wide^ variety^ of socioeconomic strata,^ as^ meas- ured by a two-index (education
and occupation) Hollingshead
scale (table 1).
The primary goal of the parent
study was to determine the^ prev-
alence of congenital rubella^ in^ the
sample population by^ means^ of^ a
five-part screening battery. The
Denver Developmental Screening
Test was one part of the battery.
The test was administered, under
close supervision, by four spe-
cially trained technicians and
scored according to the manual
provided by Frankenburg and
Dodds in 1968. Toward the^ end
of the parent study the^ staff^ be-
came concerned about the field
performance of the test^ because
of an apparently high overrefer-
ral rate, which raised parallel
concern about underreferral-a
more serious error^ in^ view^ of the
primary study goal. We therefore
decided to^ evaluate^ the^ perform-
ance of this screening test^ in^ at-
tempting to correctly identify the
intelligence of^ a^ typical group^ of
May 1972, Vol. 87, No.^5
preschool children^ as^ measured by a standard test.
Planning, funding, and relocat-
ing subjects from the parent study consumed 12 months-longer than anticipated. Samples of 160 DDST "normals," 75 DDST "questionables," and 63 DDST "abnormals" were located and given the Stanford-Binet Test of Intelligence by a team of four ex- perienced (^) psychologists, on leave from the Baltimore City schools,
under the direction of Lewis
Armistead, Ph.D., Maryland State
Department of Health. The dis-
tribution of 298 IQ scores is de-
tailed in table 2. It is slightly
skewed toward the higher score
intervals, reflecting the distribu- tion of socioeconomic status in
the study population.
Since the data needed to calcu-
late the Hollingshead scale (table
1) were not available for either
the county under study or for the
State as a whole, the percentage
distribution of Hollingshead's
original study group is given as
a frame of reference. As would
be expected in^ comparing an
urban industrial community with
a semirural county, a^ shift is seen
in the distribution of the semi-
rural county from class IV to
Table 2. Distribution of 298
actual IQ scores in this study as compared with expected nor- mal distribution
Expected Actual Stanford-Binet percent percent IQ scores distri- distri- bution bution
0-67........... 68-83.......... 84-99.......... 100-115........ 116-131........ 132 +..........
Total..... 100.0 100.
class II, thus giving reassurance that a representative sample was
being studied.
DDST normals, abnormals,^ and
questionables were defined, ac-
cording to Frankenburg^ and
Dodds' manual, as^ follows:
DDST normal: no items failed below the child's age level; at
least one item, dissected by
the age line, passed.
DDST abnormal: at least two
items failed below the child's
age level.
DDST questionable: all others.
A child's total DDST rating
was the lowest rating in any one
of the four sections: gross motor,
fine motor, language, or personal-
Table 1. Range and distribution of^ socioeconomic status^ of^590 study
subjects, as measured by two-index (education and^ occupation)
Hollingshead scale
Percent Score Percent (^) distribution, Social class range, distribution, Hollings- this study this study head's New Haven community 1
I, professional and^ management ......... 11-13^ 3.1^ 2. II, business^ and^ semiprofessional ........ 14-30^ 14.8^ 9. III, clerical^ and^ sales^ .................. 31-49^ 21.9^ 18. IV, skilled workers ................... 50-64 37.8 48. V. unskilled^ workers^ and^ unemployed.... 65-77^ 22.4^ 20. Total .100.0^ 100.
(^1) See text above. SOURCE: Reference 5.
social. Tasks that were refused
were considered questionable.
To analyze the ability of the
Denver Developmental Screening
Test to predict intelligence as
measured by the Stanford-Binet
test, we dichotomized the results
of both reference and screening
tests as follows:
Stanford-Binet Test of Intelli-
gence: Positive =^0 to 83 Negative 84 Denver Developmental Screening Test:
Positive =^ questionable and
abnormal
Negative =^ normal
Table 3 shows the screening
test-reference test two-by-two
contingency table, from which
vlarious evaluative indices may be
calculated (6, 7). The indices of
sensitivity, specificity, underrefer-
ral, overreferral, and^ tetrachoric
r (cosine-pi approximation)
selected for^ this^ study^ were
calculated as follows: Sensitiv-
ity (^) = a+c specificity (^) b (^) +d'
underreferal = (^) c+d'overrefer- c + d ral (^) = (^) a b' and (^) tetrachoric r
+ b
(cosine-pi approximation) =
Section a of table 3 indicates
true positives, section b false
positives, section c false negatives,
and section d true negatives. Sen-
sitivity indicates the ability of a
screening test to detect a true
positive result out of all diseased
persons. Specificity indicates the
ability of^ a^ screening test to^ de-
tect a true negative result out of
all nondiseased persons.
Underreferral is the ratio of
persons screened^ negative^ with
disease to total persons not re-
474 Health Services Reports
tions contribute little or nothing
to correlation of the test with the
Stanford-Binet IQ test. The fine
motor and^ language sections in-
dividually show higher correla-
tion, but a combination of these
two sections is superior to the
test as a whole. This result is not
surprising because the Stanford-
Binet test relies heavily on fine motor and language functions.
Two important limitations of
this (^) evaluation should be noted:
(a) the screening and the refer-
ence tests were (^) separated by a
period of 12 months and (b)
some DDST tasks (for example,
gross motor and personal-social) were not designed to be predic-
tors of intelligence quotients, as
measured by the Stanford-Binet
test. Even though the screening
tests are age specific, the variance
would be expected to be greater
in a given group of children at an
earlier age. Also, some mothers
may have schooled their children
in several of the fine motor and
language tasks of the test that
are derived from the Stanford-
Binet test.
These limitations, in addition
to low r (^) values in the gross motor and (^) personal-social sec- tions would help to explain (^) the
relative superiority of combining
the fine motor and language sec-
tions as a predictor of a Stanford-
Binet intelligence score. Further,
to obviate the fact that we have
related this screening test,to only
one measure of intelligence (the
Stanford-Binet test), a followup
study of this cohort to examine
other parameters of social, emo-
tional, educational, and neuro-
logical function is planned for
REFERENCES (1) Collen, M. F., and Linden, C.: Screening in a group practice prepaid medical care plan. J
Chronic Dis 2: 400-408, Octo- ber 1955. (2) Frankenburg, W. F., and Dodds, J. B.: The Denver Developmen- tal Screening Test. J Pediatr 71: 181-191, August 1967. (3) Gesell, A., and Amatruda, C. S.: Developmental (^) diagnosis. Ed. 2. Harper & Row, New York, 1958. (4) Grant, J. A., and (^) Gittelsohn, A. M.: Estimating educational casualty from the (^1964) rubella epidemic in Maryland. Md State Med J. In press. (5) Hollingshead, A. B., and Red- lich, F. C.: Social class and men- tal illness. John Wiley & Sons, Inc., New York, 1958. (6) Thorner, R. M., and Remein, Q. R.: (^) Principles and procedures in the (^) evaluation of screening for (^) disease. PHS Publication No. 846, Public Health (^) Monogr No.
- U.S. Government (^) Printing Office, Washington, D.C., 1961. (7) Blum, H. L., Peters, H. (^) B., and Bettman, J. W.: The Orinda Study: Vision screening for ele- mentary schools. University of California Press, Los Angeles,
GRANT, JOHN A. (University of Maryland School of Medicine), and GITTELSOHN, ALAN M.:
The Denver Developmental Screening Test compared with the Stanford-Binet test: Evaluation
of 298 preschool children. Health Services Reports, Vol. 87, May 1972, pp 473-476.
The (^) Denver Developmental Screening Test (DDST) was evaluated by comparing the DDST performance of 298 (^) children with their perform- ance under the Stanford-Binet Test of Intelligence. The classic epidemiologic profile, including sensi- tivity (.788), specificity (.578), underreferral (.044), overreferral (.812) and tetrachoric r (.581) was used. Under these circumstances (testing of a dis- creet population at a specific point in time) the Denver Developmental Screening Test was shown to be a useful procedure for identifying preschool
children (^) with low Stanford-Binet intelligence scores, but it had a strong tendency to overrefer. Limitations of this evaluation were a 12-month interval between screening and reference tests
and the use of a broad spectrum test like the
Denver Developmental Screening Test to predict
a single parameter of development-intelligence
as measured by the Stanford-Binet test. The eval- uation does, however, reflect performance of the test under actual field conditions, (^) providing the administrator with valuable information for pro- gram control.
(^476) Health Services Reports
