7/30/2019 jama 264_8_036

1/5

Prevalence of Reading Disabilityin Boys and GirlsResults of the Connecticut Longitudinal StudySally E. Shaywitz, MD; Bennett A. Shaywitz, MD; Jack M. Fletcher, PhD; Michael D. Escobar, PhD

We hypothesized that results of previous investigations indicating an increasedprevalence of reading disability in boys compared with girls reflected a bias insubject selection. In an epidemiologic sample of 215 girls and 199 boys, weidentified two groups of reading-disabled children: research identified and schoolidentified. Results indicated no significant differences in the prevalence ofreading disability in research-identified boys compared with research-identifiedgirls in either second (17 [8.7%] of 196 boys; 15 [6.9%] of 216 girls) or third grade(18 [9.0%] of 199 boys; 13 [6.0%] of 215 girls). In contrast, school identificationresulted in the classification of 27 (13.6%) of 198 boys and seven (3.2%) of 216girls in second grade and 20 (10.0%) of 199 boys and nine (4.2%) of 215 girls in

third grade. Our data indicate that school-identified samplesare

almost unavoid-ably subject to a referral bias and that reports of an increased prevalence ofreading disability in boys may reflect this bias in ascertainment. These findingscaution against relying solely on schools for identification of reading-disabledchildren.

(JAMA. 1990;264:998-1002)

LEARNING disabilities rank amongthe most common neurobehavioralchildhood disorders.1 Of the learningdisabilities, reading disability (RD), thefocus ofthis report, represents the most

frequently studied. Although RDwas

initially described and characterized byphysicians,2,3 only recently have physicians become actively involved in theidentification and management of reading-disabled children.

Federal legislation mandating thatlearning-disabled children receive special-education services and that school

systems implement procedures to identify children eligible for such servicesresulted in a marked increase in theidentification of reading-disabled children.4 Profound shifts in the provision of

healthcare

services tochildren have

coincided with this significant change inpublic policy. As acute, life-threateningillnesses lessen as a cause of morbidityin children, physiciansparticularlypediatricians, family physicians, andpdiatrie neurologistsincreasingly focu s o n and accept responsibility formore chronic handicapping conditions.MFor many physicians, the most dramatic shift has been in the increased demand for services related to school

problems, particularly learning disabilities.9 As demand increases for their in

volvement in the assessment of learning-disabled children, physicians needto become more knowledgeable about

learning disabilities, particularly aboutidentifying and managing children withRD.

The diagnosis of RD presents a set ofcircumstances that differs from thatusually encountered in medical practice. While biologically based, RD ismanifest and expressed within the context of the classroom so that its identification often depends on educational processes and procedures. Since most

target populationsof

reading-disabledchildren are selected on the basis of prior identification by their schools, we examined the possible sample bias resulting from the use of school-identifiedsamples for studies of RD. We hypothesized that results of previous investigations that indicated an increased prevalence of RD in boys compared with girlsreflected a bias in subject selection rather than a gender difference.

We proposed to investigate the implications of sample selection on studies ofRD by using the epidemiologic sample

provided bythe Connecticut

Longitudinal Study (CLS). The availability of thisepidemiologic sample offered the opportunity to examine the prevalence of RDin boys and girls in (1) a subset ofthe CLS sample meeting criteria for anabilityachievement discrepancy (research identified), and (2) a subset of theCLS sample identified by their schoolsas reading disabled and entitled to receive special-education services (schoolidentified), in theory based on the sameabilityachievement discrepancy definition of RD.

Our strategy was first to use the CLSsample to determine the prevalence ofRD in boys and girls as defined by (1)

From the Departments of Pediatrics (Drs S. E. andB. A. Shaywitz), Neurology (Dr B. A. Shaywitz), ChildStudy Center (Drs S. E. and B. A. Shaywitz), and Epide-miology and Public Health (Dr Escobar), Yale UniversitySchool of Medicine, New Haven, Conn; and the Depart-ment of Pediatrics, University of Texas at Austin (Dr

Fletcher).Reprint requests to Department of Pediatrics, YaleUniversity School of Medicine, PO Box 3333, New Ha-ven, CT 06510-8064 (Dr S. E. Shaywitz).

wnloaded From: http://jama.jamanetwork.com/ by a Hebrew University of Jerusalem User on 09/02/2013

7/30/2019 jama 264_8_036

2/5

objective criteria of an ability-achievement discrepancy and (2) school identification as reading disabled. We then examined gender differences within theoverall CLS sample on a series of variables assessing cognitive ability, academic function, and behavior. We nextexplored the main effects of both groupand gender on these dependent measures in the research-identified and the

school-identified groups. Finally, wecompared research identified and schoolidentified to determine which factorsdifferentiated the two classifications.

SUBJECTS AND METHODS

Sample

The target population for the CLSwere children attending Connecticutpublic kindergarten during the 1983-1984 school year. Children included inthe study were selected by a two-stageprobability sample. Administratively,the state is stratified into six regionaleducational areas comprising 146 townsand nine regional districts, or 155 primary sampling units. Within each regional area, a systematic sample of apair of towns was selected with probability proportional to size based on 1981kindergarten enrollments. The secondstage of sampling consisted of the selection of two kindergarten classes withinthe school system of each town; eachclass within a town had equal probability of selection. For each town, two random numbers selected the two classes,and a total of 24 classes were selected in12 towns.

Subjects

All children in the selected classeswere invited to participate in the study.Exclusionary criteria were limited tosignificant sensory impairment or serious psychiatric difficulty and to Englishnot being the primary language; thisresulted in the exclusion of one blindchild. Four hundred forty-five childrenrepresenting 96.5% ofthe selected sam

ple participated, including 235 girls(53%) and 210 boys (47%). The sampleincluded 375 white (84.3%), 50 black(11.2%), four Asian (0.9%), nine Hispanic (2.0%), and seven children whoserace was unknown (1.6%).

The sample has been followed longitudinally from entry into kindergarten,and these results reflect data obtainedthrough third grade. As part of the ongoing protocol, the children receive individualized assessments yearly.Teachers complete rating forms, andschools indicate which children were

identified through routine proceduresas meeting criteria for RD. Age at thetime of testing in third grade was

8.70.42 years (mean SD), with arange of 7.9 to 10.5 years.

Materials

The children were assessed on a series of child- and teacher-based measures encompassing academic, cognitive, and behavioral domains. Abilitywas assessed by the Wechsler Intelligence Scales for Children (WISC-R)10and achievement with the reading andmathematics subtests of the Woodcock-Johnson Psycho-Educational Battery,Part II (W-J).n Classroom performancewas assessed by the Multigrade Inventory for Teachers (MIT),12 composed of aseries of six empirically derived scales:attention, activity, language, dexterity, behavior, and academics. All scoresrange from 0 to 5, with higher scoresindicating poorer performance. At theend of the academic year, informationwas obtained from each child's school,indicating which children were designated as learning disabled in readingand receiving special-educationservices.

Criterion Measures

The objective finding of an ability-achievement discrepancy representsthe standard for the diagnosis of an RD.Such a discrepancy-based definition isconsistent with the notion of specific RDin which poor reading skills are not accounted for by the level of general intelligence and is currently incorporatedinto both

federal guidelinesand state

operational definitions. In this study,reflecting current standards, RD is defined by the presence of a discrepancybetween predicted reading achievement (based on intellectual ability) andactual reading achievement. In the usual practice, before children can be testedto determine whether a discrepancy ispresent, they first must be referred fortesting. The availability of an epidemiologic sample made it possible to examinechildren without having to rely onschool identification procedures..

Childrenare

classifiedas

havingan

RD by two criteria: (1) research identified and (2) school identified. The research-identified classification is basedon the demonstration of an ability(WISC-R, Full Scale IQ)-achievement(W-J Reading Cluster) discrepancy, using a regression equation developed onthe CLS population. Use of a regressionequation takes into account regressionto the mean effects that occur when thecorrelation between two measures(ability and achievement) is not perfect.Children scoring 1.5 SDs or more below

their predicted reading achievementwere classified as research-identifiedreading disabled. Children were classi-

fied as school-identified reading disabled if they were identified by theirschool systems as reading disabled andeligible to receive special-education services under this designation. Childrenwhose Full Scale IQ was below 80 wereexcluded from the analysis (n= 13). Theresearch-identified and school-identified designations were determined forsecond and third grade.Statistical Analysis

The regression equation was used todetermine prevalence rates for RD.Gender differences in prevalence rateswere tested with the \2 test. Gender differences on ability (WISC-R),achievement (W-J), and behavioral(MIT) measures in the overall samplewere analyzed by unpaired t tests. Effects of group and gender on these behavioral, ability, and achievement measures were analyzed using a 2x2(research-identified or school-identifiedgroup x gender) analysis of variance.To examine the possible influence ofteacher perceptions on identification asreading disabled, we used conditionallogistic regressions.18,14 This procedureaccounts for the matched pairs by applying a logistic regression to the subjectswho are differentially classified by research-identified and school-identifiedcriteria. Statistical significance for allmeasures was defined as a two-tailed Pvalue of less than .05.

RESULTS

Prevalence

The results of the classification process for research-identified and school-identified RD for second and thirdgrades are presented in Table 1. Nosignificant differences in the prevalenceof RD in research-identified boys compared with research-identified girls ineither grade were noted. In contrast,school identification resulted in the classification of significantly more boyscompared with girls in each of thesegrades.

Characteristics of RD

Teacher (MIT)- and child (WISC-Rand W-J)-based measures were used toinvestigate possible gender differencesin the overall sample. These measureswere then used to characterize differences between research-identified compared with non-research-identified,and school-identified compared withnon-school-identified groups. The results reflect teacher- and child-based information obtained in third grade. In

the overall CLS sample, boys receivedsignificantly poorer ratings from theirteachers on each scale of the MIT, in-

wnloaded From: http://jama.jamanetwork.com/ by a Hebrew University of Jerusalem User on 09/02/2013

7/30/2019 jama 264_8_036

3/5

Table 1 .Prevalence of Reading Disability in Research-Identified (RI) and System-Identified (SI) Boys andGirls

Reading Disabled, No. (%)

Non-RI RI Total Non-SI SI Total

Second gradeBoys 179 (91.3) 17 (8.7)* 196 171 (86.4) 27 (13.6)tGirls 201 (93.1) 15 (6.9) 216 209 (96.8) 7 (3.2) 216

Third gradeBoys 181 (91.0) 18 (9.0)t 199 179 (90.0) 20 (10.0) 199Girls 202 (94.0) 13 (6.0) 215 206 (95.8) 9 (4.2) 215

*x'=0.429, not significant, boys compared with girls.tx2 = 14.8, P< .0001, boys compared with girls.tx'=1 -34, not significant, boys compared with girls.X2=5.5, P = .02, boys compared with girls.

Table 2.Mean Scores on Teacher- and Child-Based Measures of Boys and Girls in the ConnecticutLongitudinal Study

Source (Measure) ScaleBoys

(n = 199)Girls

(n=215) t StatisticTeacher (Multigrade Inventory for Teachers*) Activity

Behavior

Attention

Fine motor

LanguageAcademics

1.63 1.01t1.21 0.53t

1.76 1.27t1.23 0.73t1.12 0.94$1.52 1.21t

7.03

5.40

5.31

4.95

2.34

3.95

Child (Wechsler Intelligence Scales forChildren) Verbal IQ

Performance IQ 111.6

Full Scale IQ 112.0

108.5

109.4

109.8

1.04

1.69

Child (Woodcock-Johnson Psycho-EducationalBattery, Part II) Reading cluster 104.0

Mathematics cluster 105.5

104.3

107.1

-0.25

-1.08

'Higher values indicate poorer performance.tP=.0001.tPs.02.

Table 3Mean Teacher Ratings of Research-Identified (RI) Compared With Non-RI and System-Identified(SI) Compared With Non-SI Boys and Girls

Measures:MIT*

Scales

RI Non-RI

Boys(n = 18)

Girls

(n= 13)Boys

(n = 181)Girls

(n=202)

Test Statistic!F(1,411)

LD Gender

Activity 1.61 0.99 2.21 48.21Behavior 1.17 1.04 1.22 0.50 0.05 5.79HAttention 2.25 1.25 8.931 26.87Fine motor 0.72 8.531 23.21 Language 1.58 1.02 1.07 0.93 5.26II 4.9111Academicst 2.40 1.44 1.43 1.19 6.67H 19.42

Non-SI

Boys(n=20)

Girls

(n=9)Boys

(n = 179)Girls

(n = 206)

Activity 2.00 1.57 1.59 1.00 6.971 45.30

Behavior 2.03 1.28 0.50 12.10# 25.63Attention 2.34 1.93 1.69 13.52# 24.33Fine motor 1.90 1.07 0.71 10.141 21.18Language 1.33 1.50 1.09 0.92 5.38H 4.27HAcademics 2.14 1.97 1.45 1.16 23.08 12.03#

MIT indicates Multigrade Inventory for Teachers; higher scores indicate poorer performance.tLD, Gender refer to the F test for main effects on LD group (RI or SI) and gender.^Significant group x gender interaction (F=4.86, P = .03). No other interactions reached significance.Ps.0O01.HPs.05.IPs.01.#Ps.001.

eluding those assessing activity, behav

ior, attention, fine-motor skills, and academic performance (Table 2). Nosignificant gender differences in ability

were evident, as measured by Verbal

IQ, Performance IQ, and Full ScaleIQ, or in achievement (reading ormathematics).

Reading-Disabled vsNon-Reading-DisabledBoys and Girls

Results of teacher ratings of research-identified compared with non-research-identified children and ofschool-identified compared with non-school-identified children are shown inTable 3. Comparisons of teacher ratingsof both research-identified vs non-re

search-identified children and of school-identified vs non-school-identified children demonstrated significant maineffects for gender for all scales; boyswere perceived as performing morepoorly in each domain assessed. Groupeffects for research identified vsnon-research identified were significant for all scales except activity andbehavior, whereas group comparisonsbetween school-identified and non-school-identified children indicated significant group effects for all the MITscales, including those assessing behavior and activity.

Comparison of ability and achievement scores for research-identified vsnon-research-identified children indicated no significant gender differencesin overall ability (Verbal IQ, Performance IQ, and Full Scale IQ) or inachievement (reading or mathematics)(Table 4). Although no significant groupdifferences were noted in Verbal IQ,Performance IQ, and Full Scale IQ ofresearch-identified compared with non-research-identified children, research-identified children scored

significantlylower on achievement measures, including reading and mathematics. In comparisons of school-identified vs non-school-identified children, no positivemain effects for gender on Verbal IQ,Performance IQ, or Full Scale IQ werenoted. There were significant main effects for group, with school-identifiedchildren scoring lower on Verbal andFull Scale IQ. Significant group, but notgender differences, were apparent onthe reading and mathematics tests.

The results of the conditional logistic

regressionare shown in Table 5 and

indicate a significant behavior effectdistinguishing research identified andschool identified. Examination of thesedata indicated a natural break betweenchildren who scored 0 and those childrenwho scored above 0 on the behaviorscale. We then compared research identified and school identified on whetherthey scored at or above 0 on the behavior scale (Table 6). Children who wereidentified by their schools as readingdisabled but who did not meet objectivecriteria for RD were significantly more

likely to exhibit behavioral difficultiesthan children who satisfied objectivecriteria for RD but who were not identi-

wnloaded From: http://jama.jamanetwork.com/ by a Hebrew University of Jerusalem User on 09/02/2013

7/30/2019 jama 264_8_036

4/5

Table 4.Mean Scores of Research-Identified (RI) Compared With Non-RI and System-Identified (SI)Compared With Non-SI Boys and Girls on Measures of Ability and Achievement

RI Non-RI Test Statistic*i- -1 i-"-1 P(1,411)

Boys Girls Boys Girls

7/30/2019 jama 264_8_036

5/5

tivity and behavioral problems asamong the most prevalent characteristics associated with learning disability.21,22 Our data are also consistent withthe findings of Ysseldyke and colleagues23'26 that factors not related to adiscrepancy between a child's abilityand achievement influence school identification and placement for learningdisability. We find that only 13 (45%) of29

school-identified children meetdis

crepancy criteria for RD based on research-identified criteria. Our data indicate further that behavioral problemssignificantly differentiate reading-disabled children who are and are not identified by their schools. These findingsare supported by an accumulating bodyof literature22,27"33 suggesting that teachers' perceptions of what constitutes inappropriate behavior enter into thedecision and that, in particular, overac-tivity and behavioral difficulties arelikely to be disruptive to a classroom

and to influence decisions regardingsuch children.31'33 Similar findings thatindicate behavioral disturbances increase the likelihood of identification of

boys have also been found for attention-deficit disorder.34

The results of this study indicatingthat the source of the sample may profoundly influence the results emergingfrom studies of RD have significant implications for both research and clinicalpractice. Our data indicate that school-identified samples are almost unavoidably subject to a referral bias and that

reports in the literature of an increasedprevalence of RD in boys may reflectthis bias in ascertainment. Caution is

suggested in generalizing from researchstudies based on school-identified populations.

The conceptualization of child healththat drives the provision of health careservices to children is changing. A narrow definition of health focused primarily on the preservation of life is expanding to a much broader conceptualizationofhealth that incorporates the quality oflife. As a result, providers of health careto children assume responsibility for thehealth and welfare of their patients inthe broadest sense. Reading disability,in terms ofthe large number of childrenaffected and the high degree of morbidity imposed by the disorder, representsa significant challenge to the health andwelfare of our nation's children.

Physicians' awareness of the bias inreports of increased prevalence of RD inboys increases their effectiveness inboth identifying and managing childrenwith school failure. The results of this

study encourage thephysician's

active

participation in and provide guidelinesfor the evaluation of patients with po-

tential RD. Our data indicate that a history of language difficulties, fine-motorproblems, and inattention should raiseconcern about a possible reading problem. The absence of behavioral difficulties is not necessarily an indication thatthe child does not have RD. In practicalterms our data indicate that physiciansshould not rely solely on schools for theidentification of RD in their patients.These data should

encourage physiciansto inquire directly about academic functioning, particularly in girls. If anyquestion arises that the child is experiencing academic difficulty, even if notcorroborated by the school, the physician should request tests of ability andachievement.

Physicians caring for children acceptas their responsibility the fullest understanding of the factors affecting growthand development. As such, physiciansare in a unique position to assume aleadership role in identifying and help

ing to manage children with school-related problems. Just as the physicianhas undertaken responsibility for understanding the interrelationships between the child and family as significantfactors that influence the child's developmental progress, the physician nowadds a new dimensionunderstanding the child's relationship with theschoolas critical to providing optimalhealth care services to children.

This study was supported by grants POl HD21888 and P50 HD 25802 from the National Institute of Child Health and Human Development, by

grant GOO-8535118 from the US Office of Education, and by a contract from the Connecticut StateDepartment of Education.

The technical assistance ofDeborah Lessne, MA,in data gathering and of V. R. Towle, MPhil, andJohn Holahan, PhD, in statistical analysis is gratefully acknowledged.

References

1. The Implementation of the Education of theHandicapped Act. Washington, DC: US Dept ofEducation; 1987. Ninth Annual Report to Con-gress.2. Hinshelwood J. Word-blindness and visualmemory. Lancet. 1895;2:1564-1570.3. Morgan WP. A case of congenital word blind-ness. BMJ. 1896;2:1378.

4. Education for All Handicapped Children Act,Pub L No. 94-142. S.6, 94th Congress, 1st Sess(1975). Report No. 94-168.5. Council on Scientific Affairs. Dyslexia. JAMA.1989;261:2236-2239.6. Chilton LA. School health experience before andafter completion of pediatric training. Pediatrics.1979;63:565-568.7. Dworkin PH, Woodrum DT, Brooks KS, Mar-shall R. Pediatric-based assessment: children withschool problems. J Sch Health. 1981;51:325-329.8. Rogers DE, Blendon RJ, Hearn RP. Some ob-servations on pediatrics: its past, present, and fu-ture. Pediatrics. 1981;68(suppl):776-784.9. Burnett RD, Bell LS. Projecting pediatric prac-tice patterns: a survey by the American Academyof Pediatrics Committee on Manpower. Pediatrics.

1978;62:625-679.10. Wechsler D. Manual for the Wechsler Intelli-gence Scale for Children\p=m-\Revised.New York,

NY: Psychological Corp; 1974.11. Woodcock RW, Johnson MB. Woodcock-John-son Psycho-Educational Battery: Part Two (Testsof Achievement). Boston, Mass: Teaching Re-sources; 1977.12. Shaywitz SE. Early Recognition of Educa-tional Vulnerability\p=m-\EREV. Hartford: Connecti-cut State Dept of Education; 1986. Technicalreport.13. Cox DR. The Analysis of Binary Data. Lon-don, England: Methuen & Co Ltd; 1970.14. Holford, TR. The analysis of pair-matchedcase-control studies: a multivariate approach. Bio-

metrics. 1978;34:665-672.15. Critchley M. The Dyslexic Child. 2nd ed.Springfield, Ill: Charles C Thomas Publishers;1970.16. Finucci JM, Childs B. Are there really moredyslexic boys than girls? In: Ansara A, GeshwindN, Galaburda A, Albert M, Gartrell N, eds. GenderDifferences in Dyslexia. Towson, Md: Orton Dys-lexia Society; 1981:1-9.17. Hallgren B. Specific dyslexia: A clinical andgenetic study. Acta Psychiatr Neuro Scand.1950;65(suppl):179-189.18. Vogler GP, DeFries JC, Decker SN. Familyhistory as an indicator of risk for reading disability.JLearning Disabilities. 1985;18:419-421.19. Rie ED, Rie HE. Reading deficits and intellec-tual patterns among children with neurocognitive

dysfunctions. Intelligence. 1979;3:383-389.20. Phipps P. The LD learner is often a boy\p=m-\why?Academic Therapy. 1982;17:425-430.21. Keogh BK, Major-Kingsley SM, Omori-Gor-don HP, Reid H. A System ofMarker Variablesforthe Field ofLearning Disabilities. New York, NY:Syracuse University Press; 1982.22. Keogh BK. Individual differences in tempera-ment\p=m-\acontribution to the personal, social, andeducational competence of learning-disabled chil-dren. In: McKinney JD, Feagans L, eds. CurrentTopics in Learning Disabilities. Norwood, NJ: Ab-lex Publishing Corp; 1983:33-55.23. Ysseldyke JE, Algozzine B. Diagnostic classifi-cation decisions as a function of referred informa-tion. JSpecial Educ. 1981;15:429-435.24. Ysseldyke JE, Algozzine B, Richey L. Judge-ment under

uncertainty:how

manychildren are

handicapped? Except Child. 1982;48:6-10.25. Ysseldyke JE,Algozzine B, Richey L, GradenJ. Declaring students eligible for learning-disabil-ity services: why bother with the data? LearningDisability Q. 1982;5:37-44.26. Ysseldyke JE, Algozzine B, Thurlow ML, eds.Placement Team Decision-Making: A Naturalis-tic Investigation. Minneapolis: University of Min-nesota Institute for Research on Learning Disabil-ities; 1980. Research Report No. 41.27. Deshler DD, Shumaker JB. Social skills oflearning-disabled adolescents: characteristics andintervention. Topics in Learning and LearningDisabilities. 1983;3:15-23.28. Keogh BK, Pullis ME, Caldwell J. A short formof the Teacher Temperament Questionnaire. JEduc Measurement. 1982;19:323-329.29. Pullis ME, Caldwell J. The influence of chil-dren's temperament characteristics on teachers'decision strategies. Am Educ Res J. 1982.30. Kavale K, Alper A, Purcell L. Behavior disor-ders, reading disorders, and teacher perceptions.Except Child. 1981;28:114-118.31. Sandoval J, Lambert NM. Hyperactive andlearning-disabled children: who gets help? J Spe-cial Educ. 1984-1985;18:495-503.32. Robbins RC, Mercer JR, Meyers CE. Theschool as a selecting-labelingsystem. J School Psy-chol. 1967;5:270-279.33. De Stefano MA, Gesten EL, Cowen EL.Teachers' views of the treatability of children'sschool adjustment problems. J Special Educ.1977;11:275-280.34. Berry CA, Shaywitz SE, Shaywitz BA. Girls

with attention-deficit disorder: a silent minority? areport on cognitive and behavioral characteristics.Pediatrics. 1985;76:801-809.