Am J Psychiatry 155:6, June 1998

Special Article

ROBERTS, ATTKISSON,
PREVALENCE OF PSYCHOPATHOLOGY
AND ROSENBLATT

Prevalence of Psychopathology Among Children and Adolescents

Robert E. Roberts, Ph.D., C. Clifford Attkisson, Ph.D., and Abram Rosenblatt, Ph.D.

Objective: This study was done to update and expand information given in recent re-
views, provide a more systematic critique of past research, identify current research trends and
issues, and explore possible strategies for future research in child psychiatric epidemiology.
Method: The authors identified and reviewed 52 studies done over the past four decades that
attempted to estimate the overall prevalence of child and adolescent psychiatric disorders.
Results: About as many studies have been published since 1980 as were published before.
Sample sizes ranged from 58 to 8,462; most were in the 500–1,000 range. Studies were carried
out in over 20 countries, most frequently the United States and the United Kingdom. Subjects’
ages ranged from 1 to 18 years. Rutter’s criteria were the most frequently used for case defi-
nition; more recent studies were more likely to use DSM criteria. The most frequently used
interview was the Rutter schedule. The most common time frame for calculating prevalence
was the present, followed by 6 months and 1 year. Prevalence estimates of psychopathology
ranged from approximately 1% to nearly 51% (mean=15.8%). Median rates were 8% for
preschoolers, 12% for preadolescents, 15% for adolescents, and 18% in studies including
wider age ranges. Conclusions: The evidence is less informative than expected because of
several problems that continue to plague research on child and adolescent disorders. These
involve sampling, case ascertainment, case definition, and data analyses and presentation.
Progress in understanding the epidemiology of child disorders will largely depend on whether
future research successfully meets these challenges.
(Am J Psychiatry 1998; 155:715–725)

U nderstanding the prevalence of psychiatric disor-

ders among children and adolescents is an essen-
tial component of a sound public policy for the provi-
tion about the prevalence of psychopathology among
children and youths. A review of published studies can
help shape the direction of current and future research
sion of mental health and other services. Given the and provide the baseline context for interpreting new
current rapid and extensive changes in social policy and information as additional data become available.
service delivery, such as welfare reform and adoption of Our goal was to provide a synthesis of epidemiologic
managed care, it is timely to review published informa- studies that have tried to estimate the prevalence of
childhood disorders. To do so, we pursued a strategy
Received Sept. 23, 1996; revisions received June 24 and Oct. 10,
that was inclusive temporally and geographically, that
1997; accepted Dec. 1, 1997. From the School of Public Health, Uni- focused on clinically meaningful definitions of disor-
versity of Texas Health Science Center, Houston, and the Department ders, and that interpreted the body of empirical data in
of Psychiatry, University of California, San Francisco. Address reprint terms of extant criteria for conducting epidemiologic
requests to Dr. Roberts, School of Public Health, University of Texas
Health Science Center, P.O. Box 20186, Houston, TX 77225. research on childhood disorders. We limited studies to
Supported in part by NIMH grant MH-46122 (Dr. Attkisson), those that attempted to estimate the prevalence of over-
grant MH-51687 (Dr. Roberts), and grant MH-43694 from the Cen- all psychiatric disorders in children and adolescents.
ter for Mental Health Services Research (Drs. Roberts and Attkisson);
and by evaluation research contracts from the California State De-
partment of Mental Health (89-70225, 90-70195, 91-71106, 92-
72090, 92-72347, 93-73346, 94-74252, 94-74285, and 95-75217 HISTORICAL CONTEXT
(Drs. Rosenblatt and Attkisson).
The authors thank Harold Baize, Nancy Mills, Juliana Ortegon,
Susan De Magri, and Sue Tico for ongoing contributions to their re- Although Rutter (1) identified the research carried
search and scholarship. out in the early 1950s by Lapouse and Monk (2) as the

Am J Psychiatry 155:6, June 1998 715

PREVALENCE OF PSYCHOPATHOLOGY

first large-scale epidemiologic study in child psychiatry, from 294 to 8,462 (mean=1,769, median=1,127) in the
others (3, 4) have pointed out that community surveys first stage and 74 to 1,015 (mean=291, median=233) in
of child mental health problems were conducted at least the second stage. Mean prevalences were 15.0% in sin-
a quarter-century earlier. Links (4) cited Long’s (5) as gle-stage studies and 17.5% in two-stage studies. Over-
the earliest community survey of child mental health, all, the mean prevalence was 15.8% (median=13.7%,
while Gould et al. (3) cited Wickman’s (6) as the earliest mode=12.0%).
school-based survey of child maladjustment. Regard- Prevalence rates varied from approximately 1% to
less of how one defines the origins of such research, it almost 51%. The ages of the subjects varied substan-
is clear that descriptive epidemiologic research on chil- tially across studies. Accordingly, we grouped samples
dren has a relatively long history (3, 4). The number of into four broad categories: preschoolers (ages 1 to 5 or
epidemiologic studies has steadily increased, with a 6 years), preadolescents (ages 6 to 12 or 13 years), ado-
concomitant need for periodic analysis and synthesis to lescents (ages 12 or 13 years and older), and samples
assess progress in our understanding of the epidemiol- including wider age ranges. The 10 preschool samples
ogy of child and adolescent disorders. had a mean prevalence of 10.2% (median=8.3%, range=
Since the early 1980s, several literature reviews have 3.6%–24%). The 21 preadolescent samples had a mean
been published which assessed studies that investigated prevalence of 13.2% (median=12.2%, range=1.4%–
clinical psychiatric disorders and that used systematic 30.7%), and the 12 adolescent samples had a mean
strategies to minimize the impact of sampling and non- prevalence of 16.5% (median=15.0%, range=6.2%–
sampling errors in epidemiologic studies. The strategies 41.3%). The 14 samples that included multiple age
addressing the latter have involved the use of stand- groups had a mean prevalence of 21.9% (median=
ardized diagnostic criteria (e.g., DSM-III) to reduce cri- 18.4%, range=7.4%–50.6%).
terion variance and the use of structured diagnostic in- These rates were generated by diverse methods of
terview schedules (e.g., the Diagnostic Interview data collection (case ascertainment) and equally diverse
Schedule for Children [7]) to reduce information vari- methods of diagnosis (case definition). By far the most
ance. For example, Costello (8) reviewed five studies popular case ascertainment procedure was some vari-
and Brandenburg et al. (9) reviewed eight studies pub- ation of the Rutter interview schedules or question-
lished in the 1980s (10). Not surprisingly, there was naires (13) (19 studies). In addition, 12 other studies
considerable overlap, albeit not complete, between the utilized an unspecified psychiatric interview for case as-
two sets of studies reviewed. Although these more re- certainment. Similarly, Rutter’s classification proce-
cent research efforts reflected considerable diversity in dure was the one most frequently used for case defini-
geographic location and research methods, they also re- tion (17 studies). DSM-III and DSM-III-R were also
flected increasing sophistication in diagnostic proce- used frequently (15 studies), particularly in the more
dures and study design. These studies, relying more on recent studies.
structured clinical assessments and explicit diagnostic Did the methods of case ascertainment and case defi-
criteria, have generated more homogeneous results than nition make a difference? The answer is yes. There was
earlier efforts. This newer generation of studies, in ad- considerable variation in prevalence across all methods.
dition to providing estimates of the prevalence of clini- For example, Rutter procedures yielded prevalence
cal disorder, have confirmed the finding from earlier rates clustering around 12%. The Schedule for Affec-
studies that disorders of childhood and adolescence are tive Disorders and Schizophrenia for School-Age Chil-
relatively common. dren (K-SADS) yielded rates in the 14% range, and the
Diagnostic Interview Schedule for Children yielded
prevalence rates in the 20%–25% range. With respect
RESULTS OF STUDIES REVIEWED to case definition, DSM-III and DSM-III-R criteria
generated similar prevalence rates of 19%–23% and
We identified 52 separate studies, reported in 47 20%–22%, respectively, while clinical opinion yielded
sources, that were designed to estimate the overall rates of 10%–14%. However, the various procedures
prevalence of psychiatric disorders among children and yielded prevalence rates that varied by 10%–15% or
adolescents. The studies are summarized in table 1. more across studies, with substantial overlap in ranges
They were conducted over a period of nearly 40 years, across the various procedures. Thus, it is possible to
beginning in the 1950s. The samples came from over 20 find studies that used a particular method of case ascer-
countries; the United Kingdom and the United States tainment and case definition which found both higher
were the most frequent sites, with six and 13 studies, and lower prevalences of psychiatric disorder than a
respectively. However, studies were carried out in sites study that used other procedures.
in Europe, Asia, Africa, and South America. Sample One of the perennial questions in mental health is
sizes ranged from 58 to 8,462 (mean=1,201, median= whether prevalence rates of psychopathology are
831). Thirty-three of the 52 studies used single-stage changing over time—in particular, whether they are in-
designs, in which all study subjects received some type creasing. To examine this question, we grouped studies
of psychiatric assessment; sample sizes ranged from into those conducted in 1970 or earlier, in 1971–1980,
58 to 2,679 (mean=898, median=756). Nineteen stud- in 1981–1990, and after 1990. The mean prevalence in
ies had two-stage designs, with sample sizes ranging studies undertaken in 1970 or earlier was 15.4%. The

716 Am J Psychiatry 155:6, June 1998

 ROBERTS, ATTKISSON, AND ROSENBLATT

TABLE 1. Prevalence and Case Ascertainment of Psychiatric Disorders Among Children and Adolescents in 52 Studies
Subjects’ Number
Method of Case Age of Time Method of Case
Study Ascertainmenta (years) Subjectsb Informants Frame Definition Prevalence (%)c
Goldfarb, 1963 (11) Psychiatric interview 3rd–5th 514 Child, mother, Current Pupil Adjustment 12.8d
graders school Rating Scale
records
Krupinski et al., Clinical history taken Children, 997 Child, family Current Psychiatric diagnosis 7.5d
1967 (12) through structured adoles- members,
and semistructured cents others
interviews
Rutter et al., 1970 I: Rutter Question- 10, 11 2,199 Child, parent, Current Rutter classification 5.4 (unadjusted);
(13) naires (scales A and [286] teacher with severity 6.8e
B); II: Rutter Child criterion
and Parent Inter-
views, teacher and
parent question-
naires
Werner et al., 1971 Parent interview, 9–11 1,012 Child, parent, Current Clinical opinion (o): 26.4; (s): 13.0
(14) teacher question- teacher,
naire, health and others
school records
Bjornesson, 1974 Maternal interview 5–15 1,100 Mother Current Clinical opinion 11.8
(15) and symptom Evaluation of symp- 18.8
checklist tom loading
Leslie, 1974 (16) I: Teacher and parent 13–14 807 Child, parent, 1 year Rutter classification (o): 17.2e; (s): 4.4e
questionnaires; II: [141] teacher with severity
Rutter Child and criterion
Parent Interviews
Rutter et al., 1975 I: Rutter Question- Mother, 1 year Rutter classification
(17) naire (scale B2); II: teacher with severity
Rutter Parent Inter- criterion
view
Inner London 10 1,689 25.4e
borough [265]
Isle of Wight 10 1,279 12.0e
[211]
Richman et al., I: Behavior Screening 3 705 Child, parent Current Clinical opinion (o): 7.4d; (s): 1.1e
1975 (18) Questionnaire; II: [200]
parent interview,
child observation
Minde, 1975 (19) Rutter Questionnaire 7–15 577 Teacher Current Rutter classification 18.0
(scale B)
Kastrup, 1976 (20) Parent(s) interview, 5–6 175 Child, Current Clinical opinion 8.0
child interview, parent(s)
medical record
review
Zimmerman-Tansel- Rutter Questionnaire 6–13 418 Teacher Current Rutter classification 9.3d
la et al., 1978 (21) (scale B)
Earls and Richman, Behavior Screening 3 58 Parent Current Behavior Screening 15.5
1980 (22) Questionnaire Questionnaire scale
Earls, 1980 (23) Behavior Screening 3 100 Mother Current Behavior Screening
Questionnaire Questionnaire scale
Score ≥11 11.0
Score ≥10 24.0
Score adjusted for 16.5e
sensitivity and
specificity
Earls, 1980 (24) Behavior Screening 3 85 Father Current Behavior Screening
Questionnaire Questionnaire scale
Score ≥11 3.6
Score ≥10 8.3
Connell et al., 1982 I: Rutter Question- 10–11 779 Child, parent, Current Rutter classification 14.1
(25) naires (scales A and [176] teacher
B); II: Rutter
Interview
McGee et al., 1984 Rutter Questionnaires 7 951 Parent, Current Rutter classification (p): 17.3; (t): 8.9;
(26) (scales A and B) teacher (p) and (t): 5.5;
(p) or (t): 30.7

Am J Psychiatry 155:6, June 1998 717

PREVALENCE OF PSYCHOPATHOLOGY

TABLE 1 (continued)
Subjects’ Number
Method of Case Age of Time Method of Case
Study Ascertainmenta (years ) Subjectsb Informants Frame Definition Prevalence (%)c
Vikan, 1985 (28) I: Rutter Question- 10 1,510 Child, parent, Current Rutter classification 5.0e
naires (scales A and [139] teacher
B, abridged); II:
Rutter Interview
Verhulst et al., I: Child Behavior 8, 11 334 Child, parent, 6 DSM-III criteria (o): 26.0e; (s): 7.0e
1985 (29) Checklist, parent [116] teacher months
and teacher ver-
sions; II: semistruc-
tured parent inter-
view, child assess-
ment
Offord et al., 1987 Four scales based on 4–16 2,679 Parent, teach- 6 DSM-III criteria and 18.1e
(30); Boyle et al., the Child Behavior er, adoles- months Rutter severity
1987 (31) Checklist and DSM- cent (12–16 criterion
III age group)
Kashani et al., 1987 Diagnostic Interview 14–16 150 Adolescent, Current DSM-III criteria 41.3
(32) for Children and parent With impaired func- 18.7
Adolescents, child tioning and need
and parent versions for treatment
Weyerer et al., 1988 Standardized psychi- 3–14 358 Child, mother 3 Rutter multiaxial 3 months: 18.4; 1
(33) atric examination months, scheme, ICD-9 year: 20.7
with child and 1 year
parent
Costello et al., 1988 I: Child Behavior 7–11 789 Child, parent 1 year DSM-III criteria (c) or (p): 22.0e; (c):
(34) Checklist; II: Diag- [300] 13.8e; (p): 11.8e
nostic Interview
Schedule for Chil-
dren, child and
parent versions
Bird et al., 1988 (35) I: Child Behavior 4–16 777 Child, 6 DSM-III criteria 49.5e
Checklist, parent [386] mother, months With CGAS score 17.9e
and teacher versions; teacher <61
II: Diagnostic Inter- With CGAS score 16.0e
view Schedule for 61–70
Children, child and
parent versions; Chil-
dren’s Global Assess-
ment Scale (CGAS)
Larson et al., 1988 Child Behavior 3 756 Parent Current Child Behavior Check- 11.1
(36) Checklist list deviant behav-
ior syndrome score
Moilanen et al., Rutter Questionnaire 8 987 Parent Current Rutter classification 11.8f
1988 (37) (scale A2)
Rutter Questionnaire 1,033 Teacher Current Rutter classification 11.7f
(scale B2)
Child Depression 406 Child Current Child Depression 17.7f
Inventory Inventory
Velez et al., 1989 Diagnostic Interview 9–18 776 Child, parent Current DSM-III-R with 17.7d
(38) Schedule for Chil- (usually severity criteria
dren, child and mother)
parent versions
Wang et al., 1989 Rutter Questionnaire 7–14 2,432 Teacher Current Rutter classification 8.3
(39) (scale B)
Matsuura et al., Rutter Questionnaire 6–12 1,860 Teacher Current Rutter classification 3.0
1989 (40) (scale B)
Esser et al., 1990 Structured interview 8 216 Parent 6 ICD-9, Rutter (o): 16.2; (s): 4.2
(41) patterned after months severity criterion
Rutter Interview
Ekblad, 1990 (42) Rutter Questionnaire 11–13 248 Mother 1 year Rutter classification 17.3
(scale A)
Rutter Questionnaire 266 Teacher 1 year Rutter classification 8.6
(scale B)
Sawyer et al., 1990 Child Behavior 10–11 279 Mother 6 Child Behavior
(43) Checklist months Checklist
North American 21.0
scoring
Australian scoring 9.2

718 Am J Psychiatry 155:6, June 1998

 ROBERTS, ATTKISSON, AND ROSENBLATT

TABLE 1 (continued)
Subjects’ Number
Method of Case Age of Time Method of Case
Study Ascertainmenta (years) Subjectsb Informants Frame Definition Prevalence (%)c
Sawyer et al., 1990 Child Behavior 14–15 249 Mother 6 Child Behavior
(43) (continued) Checklist months Checklist
North American
scoring 14.6
Australian scoring 8.5
Luk et al., 1991 (44) I: Preschool Behavior 36–48 855 Child, parent, Current Clinical opinion with (o): 5.58d; (s): 0.75e
Checklist, Behavior months [233] teacher severity criteria
Screening Question-
naire; II: semistruc-
tured parent inter-
view, child interview
(modified Rutter
Interview), teacher
interview
Koot and Verhulst, Child Behavior 2–3 421 Parent 6 Child Behavior 7.8
1991 (45) Checklist months Checklist syn-
drome score
Garrison et al., I: Center for Epidemi- 12–15 3,283 Adolescent, 1 year DSM-III criteria and (c) or (p): 19.9d; (c):
1992 (46) ologic Studies De- [488] mother CGAS score <61 14.1d; (p): 7.8d
pression Scale; II:
Schedule for Affec-
tive Disorders and
Schizophrenia for
School-Age Children
(K-SADS), CGAS
Bergeron et al., Diagnostic Interview 6–14 139 Child, adoles- 6 DSM-III-R criteria
1992 (47) Schedule for Chil- cent, parent months Ages 6–11 (p): 19.1; (c): 13.2
dren-2, parent and Ages 12–14 (p): 12.7; (c): 15.5
child versions; Dom- With impairment
inic Picture-Based in at least one
Questionnaire for area
children aged 6 to Ages 6–11 (p): 13.2
11 Ages 12–14 (p): 2.8; (c): 8.5
Fergusson et al., Child and mother 15 961–986 Child, mother Current DSM-III-R criteria
1993 (48) interviews (number Optimal informant (c) and (p): 27.3
varied method
with test) Latent class method (c) and (p): 25.7
Single informant (c): 22.1; (p): 13.0
Kasmini et al., 1993 I: Research Question- 1–15 507 Child, parent 1 year Rutter multiaxial 6.1
(49) naire for Children; [74] scheme, ICD-9
II: parent and child
semistructured psy-
chiatric interviews
Matsuura et al., Rutter Questionnaires 6–12 Parent, Current Rutter classification
1993 (50) (scales A and B) teacher
Tokyo 2,638 (p): 12.0; (t): 3.9;
(p) and (t): 1.4
Beijing 2,432 (p): 7.0; (t): 8.3; (p)
and (t): 2.1
Seoul 1,975 (p): 19.1; (t): 14.1;
(p) and (t): 4.5
Lewinsohn et al., K-SADS with com- 14–18 1,710 Adolescent Current, DSM-III-R criteria Point: 9.59;
1993 (51) bined features from lifetime lifetime: 37.08
the epidemiologic
and present episode
versions
Morita et al., 1993 I: Rutter Question- 12–15 1,992 Child, parent, Current Rutter classification 15.0d
(52) naires (scales A and [613] teacher with severity
B); II: Rutter Child criterion
Interview
Stallard, 1993 (53) Behavior Checklist 3 1,170 Parent Current Behavior Checklist 10.0
criteria
Fombonne, 1994 I: Child Behavior 8–11 2,158 Parent, 3 Rutter multiaxial (o): 12.4e; (s): 5.9e
(54) Checklist, Rutter [217] teacher months scheme, ICD-9
Questionnaire (scale
B2); II: Rutter Parent
Interview, CGAS

Am J Psychiatry 155:6, June 1998 719

PREVALENCE OF PSYCHOPATHOLOGY

TABLE 1 (continued)
Subjects’ Number
Method of Case Age of Time Method of Case
Study Ascertainmenta (years) Subjectsb Informants Frame Definition Prevalence (%)c
Gomez-Beneyto et I: Child Behavior 8, 11, 15 1,127 Child, mother Current DSM-III-R criteria (o): 21.7e; (s): 4.4e
al., 1994 (55) Checklist; II: K- [320]
SADS; Global
Assessment of
Functioning scale
Burns et al., 1995 I: a parent question- 9, 11, 13 3,896 Child, parent 3 DSM-III-R criteria 20.3e
(56); Costello et naire; II: Child and [1,015] months “Core” diagnoses 12.1e
al., 1996 (57) Adolescent Psychi- With impairment 11.1e
atric Assessment
Jensen et al., 1995 I: Child Behavior 6–17 294 Child, parent 6 DSM-III-R criteria (p): 24.4e; (c): 13.6e;
(58) Checklist; II: Diag- [104] (usually months (p) or (c): 40.8e
nostic Interview mother) With impairment (p): 18.6e; (c) 8.8e;
Schedule for Chil- (p) or (c): 26.3e
dren, child and With service need (p): 7.7e; (c): 7.0e;
parent versions (p) or (c): 15.8e
Shaffer et al., 1996 Diagnostic Interview 9–17 1,285 Child, parent, Current DSM-III-R criteria (c): 32.2; (p): 30.3;
(59) Schedule for combined (c) and (p): 50.6
Children-2.3 With CGAS score (c): 15.3; (p): 12.1;
≤70 (c) and (p): 24.7
With CGAS score (c): 7.2; (p): 6.5; (c)
≤60 and (p): 12.8
With CGAS score (c): 3.4; (p): 3.2; (c)
≤50 and (p): 6.2
With diagnosis-speci- (c): 19.6; (p): 19.2;
fic impairment (c) and (p): 32.8
criteria (DSIC)
With DSIC and (c): 12.3; (p): 10.2;
CGAS score ≤70 (c) and (p): 20.9
With DSIC and (c): 6.1; (p): 5.5; (c)
CGAS score ≤60 and (p): 11.5
With DSIC and (c): 2.8; (p): 2.7; (c)
CGAS score ≤50 and (p): 5.4
aScreening instruments are listed when the information was used in case determination or for adjustment of prevalence rate; they are distinguished
from more intensive measures by a Roman numeral I.
bNumbers in brackets are for second stages, when applicable and available.
cAbbreviations used: (o)=total or overall prevalence; (s)=severe case rate; (t)=teacher; (p)=parent; (c)=child or adolescent.
dCalculated/estimated from available information.
ePrevalence rate adjusted for cases missed or for nonrespondents.
fOnly total prevalence is reported here; study cites rates for several geographic areas.

mean prevalence was 14.1% in studies in 1971–1980 symptom criteria and those who met symptom criteria
and 13.8% in studies spanning 1981–1990. For studies and had some degree of functional impairment. The
carried out after 1990, the mean prevalence for youths prevalence rates adjusted for impairment were typically
meeting symptom criteria was 26% (range=12.1%– less (sometimes much less) than one-half the prevalence
50.6% for child, parent, and combined reports). How- rates based only on meeting symptom criteria. How-
ever, three studies (48, 56, 59) constitute special cases: ever, from the data of the 23 studies that did adjust for
unlike most previous studies, they used both the child impairment, it is very clear that there is little consensus
and a parent as the informants, and two of them (48, on how to assess the degree of impairment in children
59) reported prevalences based on both separate and or adolescents who meet or exceed diagnostic criteria.
combined information from the informants. Two of A variety of impairment measures were used. The two
these three studies (56, 59) also reported prevalence most frequently used were need for treatment and im-
rates adjusted for severity or impairment. Therefore, pairment scores derived from some scale, typically the
the studies conducted after 1990 are difficult to com- Children’s Global Assessment Scale. Upon review of
pare with earlier studies. Excluding these three studies, these studies, it is clear that incorporating impairment
there appears to be no trend for increasing prevalence into diagnostic algorithms substantially affects preva-
among studies carried out since the early 1950s. lence rates. What is not clear is what the resulting
There is growing consensus concerning the use of im- prevalences mean, as illustrated by two recent articles
pairment criteria in determining “caseness” (58), and by Jensen and his colleagues (58) and Shaffer and his
we examined that as well. Twenty-three studies either colleagues (59).
presented only prevalence rates adjusted for impair- In the Jensen et al. study, the prevalence of any disor-
ment or presented prevalence rates for subjects who met der based on the child interview was 13.6%. This rate

720 Am J Psychiatry 155:6, June 1998

 ROBERTS, ATTKISSON, AND ROSENBLATT

dropped to 8.8% if there was impairment in at least one small in many studies. For example, a prevalence rate
life domain (home, school, peers) and to 7.0% if there of 12% would yield only 120 cases in a sample of
was any indication of need for mental health treatment. 1,000. In the second stage of two-stage studies, the av-
In the Shaffer et al. study, the prevalence of criterion erage sample size was only 291, which would yield only
symptoms based on child interview was 32.2%, and the 35 cases of clinical disorder at a prevalence of 12%. The
prevalence for symptom criteria with impairment in most obvious disadvantage of small sample sizes is de-
one or more life domains was 19.6%. When Children’s creased precision in estimates of prevalence and in esti-
Global Assessment Scale scores also were incorporated, mates of the relative contributions (associations) of pu-
the rates dropped even more dramatically—to 2.8% for tative risk factors.
the child meeting symptom criteria with impairment in
at least one of the three life domains and a Children’s Case Ascertainment
Global Assessment Scale score of 50 or less. Shaffer and
colleagues selected as their “true” prevalence rate one At this juncture, there are basically two types of
that combined 1) meeting symptom criteria, 2) having standardized interviews for research purposes: struc-
impairment in one or more life domains, and 3) having tured (such as the Diagnostic Interview Schedule for
a Children’s Global Assessment Scale score of 70 or Children) and semistructured (such as the K-SADS [57,
less. This rate was 10.2% for the parent version of the 64]). From an epidemiologic perspective, Edelbrock
Diagnostic Interview Schedule for Children, 12.3% for and Costello (65) made a provocative observation
child version of the Diagnostic Interview Schedule for about the impact of the two types of interviews. They
Children, and 20.9% for combined interviews. While argued that the semistructured interviews require more
this procedure appears reasonable, there are no data clinical inference (and expertise), focus on specificity,
indicating its criterion or predictive validity. and use higher diagnostic thresholds; hence, fewer chil-
dren meet criteria. Structured interviews focus on sen-
sitivity and use lower diagnostic thresholds; hence,
CENTRAL RESEARCH CHALLENGES more children should meet criteria. What this distinc-
tion means for estimates of prevalence is that when the
As is apparent from the studies summarized in table latter interviews are used, prevalences should be higher,
1, the body of evidence from descriptive epidemiologic and with use of the former, prevalences should be
studies is less informative than might otherwise be the lower, all other things being equal. The three studies
case because of several problems that continue to based on the K-SADS had prevalences around 14%,
plague research on child and adolescent disorders. whereas those that used the Diagnostic Interview
These problems, which we designated as central re- Schedule for Children had prevalences clustering
search challenges, involve sampling, case ascertain- around 21%–25%. Given the lack of studies using both
ment, case definition, and data analyses and presen- strategies, the epidemiologic implications of these two
tation (60, 61). Progress in our understanding of the alternate strategies remain an empirical question.
epidemiology of child psychiatric disorders will depend A number of writers have argued that estimates of
in large part on whether future research successfully prevalence can be obtained with greater accuracy and
meets these challenges. We now examine each of these. less cost by using a multistage design (66, 67). The field
appears to be moving in this direction, as two-stage
Sampling procedures are being used increasingly in epidemiologic
studies (for example, reference 9). We found that 19 of
Two key problems in sampling relate to represen- the 52 studies used a two-stage design. However, as
tativeness and sample size. Representativeness has been noted above, the average sample size in the second stage
problematic from two perspectives: 1) only a portion of was only about 300. Beyond the issue of sample size,
the studies to date have actually used probability sam- the viability of multistage strategies depends in large
pling designs, and 2) the samples studied do not repre- part on the efficiency of the first-stage assessment, or
sent, even when taken as a group, the diversity of the the screener (1, 68, 69). Thus far, the efficiency of most
child and adolescent population generally. Most studies screeners for childhood psychopathology leaves much
have focused on either a narrow age range (middle to be desired (70).
school, high school) or a specific age (age 3, age 8, age Most interview schedules, whether structured or semi-
11, etc.), so we cannot determine how prevalence structured, are designed to collect data on child psycho-
changes or does not change over the lifespan of child- pathology from both the parent and the child (65, 71).
hood. This is a critical issue, since we are unable to as- We found that 26 of the 52 studies collected data from
certain whether there are developmental thresholds for the child and at least one other informant (usually, a
risk of disorder that might be indicated by changes in parent or teacher). A major issue is that there are no
prevalence by age (62). The larger issue, of course, is the agreed-upon decision rules on how best to use informa-
external validity of our results (63). tion from multiple informants to make diagnoses in or-
Sample size was almost without exception small (the der to estimate prevalence in epidemiologic studies. The
median size was 831). This means that the actual num- problem is that no two studies use the same decision
ber of cases of clinical disorder identified was also quite rules in deriving estimates of the prevalence of caseness.

Am J Psychiatry 155:6, June 1998 721

PREVALENCE OF PSYCHOPATHOLOGY

This lack of comparability of prevalence rates across ment would help resolve the question of comparability
studies led Costello (60) to argue for the importance of of clinical and community cases.
reporting diagnostic rates separately by informant. The Assessment of severity in the studies reviewed varied
rationale for such a strategy is twofold. First, the deri- substantially. Only 16 of the 52 studies reported preva-
vation of the prevalence is more readily apparent, more lence rates adjusted for severity as well as crude preva-
directly interpretable, and, at least theoretically, more lence rates. An additional seven studies incorporated se-
reproducible. Second, given the low concordance verity into their case definitions, while four used
among different informants (65, 71, 72), it provides multiple “caseness” scores. Clearly, there is little con-
separate estimates based on the sources of the diagnos- sensus on how to operationalize severity of disorders.
tic data; this could be extremely important from an epi- Equally clear is the substantial impact on prevalence
demiologic perspective. The critical epidemiologic rates when severity criteria and need for treatment are
question is whether different sources yield different introduced into case definition; prevalence rates are
prevalences, different natural histories (incidence, dura- greatly reduced, sometimes by a factor of three or more.
tion), and different risk factor profiles. The available (In a number of cases, prevalences adjusted for severity
data, albeit meager, suggest that prevalences do differ of impairment and/or need for treatment are in the
by source (58, 59). In this group of studies, those with range of 4%–8%.) From the perspective of planning
a child informant and one other informant reported a mental health services, the policy implications of preva-
prevalence of 20%; in those with a parent informant lences of 4%–8% compared to prevalences of 18%–
only, the prevalence was 14%; with a teacher informant 22% (60) are profound.
only, 9%; and with a child informant only, 15%.
Whether natural history or associated risk factors also Data Analyses and Presentation
differ, and in what ways, is essentially unknown.
There is an additional consideration which, although
Case Definition seldom discussed, has impeded progress in psychiatric
epidemiology: the lack of common analytic techniques
With the exception of a few studies that used Rutter and uniform modes of presenting data (60, 61). There
criteria (17, 25, 28, 73), the overwhelming majority of are still no uniform modes of presenting data from epi-
recent studies have used DSM criteria to define case- demiologic studies of child and adolescent disorders.
ness. This accounts, in part, for the narrower range of Prevalence rates may be either point or period preva-
prevalence rates reported. But definition of a case in- lences. If the latter, the referent period may vary, but
volves more than just application of diagnostic criteria typically it is 1 month, 6 months, 1 year, or lifetime.
to ascertain the presence of a psychiatric disorder. It Thirty-two studies reported current prevalence (median=
also involves the severity of the disorder, in terms of 12.0%, mean=14.9%), another 10 studies reported 6-
either functional impairment or perceived need for month prevalence (median=15.9%, mean=19.4%), and
mental health services (9). The concern about severity another eight reported 1-year prevalence (median=14.1%,
emanates in part from concern about whether commu- mean=15.1%).
nity or epidemiologic “cases” are cases in the same For estimates of prevalence to be most useful, one
sense as the cases of children brought to clinical settings must know the precision of the estimate. That is, one
(58–60). Part of this concern no doubt stems from the must know the tolerable error or the size of the interval
fact that only a small minority of children and adoles- around the estimate with a specified degree of confi-
cents diagnosed in community surveys have had any dence (80). Given the small sample sizes in most child
contact with mental health professionals (74–76). studies, the estimates no doubt have large confidence
There is growing concern (58, 59, 77) about the va- intervals at the 95% level. More recent publications
lidity of a diagnostic nomenclature that identifies one- have begun to provide such information (46, 51, 58).
fourth to one-third or more children and adolescents as Comparability across studies also would be enhanced
meeting criteria for one or more clinical psychiatric dis- and cumulative evidence from different studies facili-
orders. The degree of inclusiveness is not only counter- tated if more child psychiatric epidemiologists em-
intuitive but calls into serious question the usefulness of ployed analytic techniques commonly used in other ar-
DSM nomenclature in its present form. Future research eas of epidemiologic research. For example, in the
should focus more on assessing severity of symptoms as presentation of prevalence rates for subgroups based on
well as functional impairment and need for treatment. age, gender, or socioeconomic status, calculation and
The available data (58, 59, 78, 79) indicate that a sub- reporting of odds ratios could indicate the relative risk
stantial proportion of individuals who meet symptom of disorder in specific subgroups, such as middle ado-
criteria for a DSM diagnosis appear to be functioning lescents, males, and lower-status youths.
adequately in their lives. This suggests that caseness is
best determined by the presence of both symptoms and Strategies for the Next Phase of Research
impairment. If a common strategy could be adopted, it
would constitute an important step. In this regard, re- Few would probably disagree with the conclusions by
search comparing treated and community samples in researchers such as Earls (81) and Rutter (1) that we
terms of phenomenology, severity, and need for treat- now know a good deal about child disorders—cer-

722 Am J Psychiatry 155:6, June 1998

 ROBERTS, ATTKISSON, AND ROSENBLATT

tainly, a great deal more than we did even a decade ago. of addressing the eight criteria outlined. That study, the
Few also would disagree that we still have much to Great Smoky Mountains Study of Youth, is a large
learn. From an epidemiologic perspective there is, in community-based, prospective study of the incidence
fact, a myriad of the most fundamental scientific ques- and prevalence of DSM-III-R psychiatric disorders
tions for which we have few or no empirical data. These among 9- , 11- , and 13-year-olds. The study has a two-
basic questions focus on issues of incidence, prevalence, stage design, with 3,896 youths screened at baseline
natural history, and etiology of psychiatric disorders in and then 1,015 assessed with a diagnostic interview.
nonpatient or community populations. Of these, we Parents also are interviewed. All subjects in the second
note eight in particular. stage are followed up annually. In addition, there is an
1. The available data on prevalence are quite limited. extensive array of putative risk factors assessed, among
For example, there are almost no data on the prevalence them age, gender, socioeconomic status, rural or urban
of clinical disorders among diverse ethnic minorities, residence, physical health, child development, family
different socioeconomic strata, and rural compared burden of the child’s mental health problems, maternal
with urban populations. depression, and family psychiatric history. A key fea-
2. From the perspective of prevention and treatment ture is linkage of the epidemiologic data with extensive
as well as epidemiology, understanding the natural his- data on utilization of mental health services.
tory of child and adolescent disorders is critical, yet Great Smoky Mountains Study procedures include
there are essentially no data on incidence, duration, and assessment of both diagnostic status and functional im-
recurrence in community populations. pairment (in three domains—school, home, peers).
3. Comorbidity is increasingly recognized as a key Thus far, prevalences have been reported (56) for
phenomenological feature of psychiatric disorders youths who met no diagnostic criteria and had no func-
among children and adults (82, 83), yet there are basi- tional impairment (63.7%), who met diagnostic criteria
cally no community-based epidemiologic data on the but were not functionally impaired (9.1%), who did not
prevalence, incidence, and natural history of comorbid meet diagnostic criteria but had sufficient symptoms to
disorders in children. be impaired (16.1%), and who met both diagnostic cri-
4. A key to understanding child disorders is under- teria and functional impairment criteria (11.1%). Over-
standing the role of developmental factors (76), but at all, 20.3% of the youths met diagnostic criteria on the
present there are few data on the role of development basis of combined parent and child reports. At baseline,
in the manifestation of psychiatric problems, because the diagnostic interview sample (N=1,015) was large
there are few data from epidemiologic studies examin- enough to permit stable estimates of most major DSM-
ing the relation between developmental milestones or III-R diagnoses and to examine the role of major risk
stages and clinical psychiatric syndromes. factors, with the exception of biological factors, which
5. Ultimately, the goal of child psychiatric epidemiol- were not assessed. Major limitations are the absence of
ogy is to explain the etiology of mental disorders. A children under 9 years of age, the limited age range at
necessary requisite is data from prospective, longitudi- baseline (ages 9, 11, and 13 years), and a sample size
nal studies assessing the roles of multiple risk factors that on successive follow-ups will not be large enough
drawn from both psychosocial and biological domains to estimate the risk of many specific disorders or the
in specific disorders and the specificity of effects of these effects of some risk factors. Still, the study stands out
factors. Few such data are available. for its many features designed to yield maximal epi-
6. In addition to examination of the role of risk fac- demiologic knowledge.
tors in etiology, there also is a need for community- Where do we go from here? As Rutter (1) has noted,
based epidemiologic studies aimed at understanding child psychiatric epidemiology indeed has made consid-
the factors affecting duration and recurrence of child erable progress in the 30 years since the landmark Isle of
disorders, in particular, factors that affect help seek- Wight study began. But, as we suggest in the preceding
ing, including use of mental health and general medical section, research on the epidemiology of child and ado-
services (76). lescent psychiatric disorder is very much a journey in
7. To date the role of biological factors has been little progress. The central research challenges discussed pro-
studied and is poorly understood vis-à-vis child and vide a point of departure for the next segment of the jour-
adolescent disorders. Fuller knowledge of the etiology ney, as well as a map for at least some alternative routes
of such disorders will require inclusion of biological that may make the trip more productive and more infor-
and genetic variables in our conceptual models (84–86). mative for policy and mental health services. The desti-
8. It is also essential to understand the relation be- nation is resolution of the research questions outlined
tween the presence or absence of psychiatric disorder above, questions that define the limits of our current
and the utilization of mental health and other health knowledge of the epidemiology of child disorders.
and human services, and how the presence of a diagno-
sis and the level of impairment can be combined to pro-
vide estimates of the need for services. REFERENCES
How do the studies reviewed fare in terms of address- 1. Rutter M: Isle of Wight revisited: twenty-five years of child psy-
ing these research issues? Poorly, overall. In our opin- chiatric epidemiology. J Am Acad Child Adolesc Psychiatry
ion, only one study (56, 57) does even a reasonable job 1989; 28:633–653

Am J Psychiatry 155:6, June 1998 723

PREVALENCE OF PSYCHOPATHOLOGY

2. Lapouse R, Monk MA: An epidemiologic study of behavior lation of seven-year-old children: prevalence, stability and types
characteristics in children. Am J Public Health 1958; 48:1134– of disorder: a research report. J Child Psychol Psychiatry 1984;
1144 25:251–259
3. Gould MS, Wunsch-Hitzig R, Dohrenwend BP: Formulation of 27. Rahim SI, Cederblad M: Effects of rapid urbanisation on child
hypotheses about the prevalence, treatment, and prognostic sig- behaviour and health in a part of Khartoum, Sudan. J Child Psy-
nificance of psychiatric disorders in children in the United States, chol Psychiatry 1984; 25:629–641
in Mental Illness in the United States: Epidemiologic Estimates. 28. Vikan A: Psychiatric epidemiology in a sample of 1510 ten-year-
Edited by Dohrenwend BP, Dohrenwend BS, Gould MS, Link B, old children, I: prevalence. J Child Psychol Psychiatry 1985; 26:
Neugebauer R, Wunsch-Hitzig R. New York, Praeger, 1980, pp 55–75
9–45 29. Verhulst FC, Berden GFMG, Sanders-Woudstra JAR: Mental
4. Links PS: Community surveys of the prevalence of childhood health in Dutch children, II: the prevalence of psychiatric disor-
psychiatric disorders: a review. Child Dev 1983; 54:531–548 der and relationship between measures. Acta Psychiatr Scand
5. Long A: Parents’ reports of undesirable behavior in children. Suppl 1985; 324:1–45
Child Dev 1941; 12:43–62 30. Offord DR, Boyle MH, Szatmari P, Rae-Grant NI, Links PS,
6. Wickman EK: Children’s Behavior and Teachers’ Attitudes. New Cadman DT, Byles JA, Crawford JW, Blum HM, Byrne C,
York, Commonwealth Fund, 1928 Thomas H, Woodward CA: Ontario Child Health Study, II: six-
7. Shaffer D (ed): National Institute of Mental Health Diagnostic month prevalence of disorder and rates of service utilization.
Interview Schedule for Children, version IV. New York, Colum- Arch Gen Psychiatry 1987; 44:832–836
bia University, 1996 31. Boyle MH, Offord DR, Hofmann HG, Catlin GP, Byles JA, Cad-
8. Costello EJ: Developments in child psychiatric epidemiology: an man DT, Crawford JW, Links PS, Rae-Grant NI, Szatmari P:
epidemiologic study of behavior characteristics in children. J Am Ontario Child Health Study, I: methodology. Arch Gen Psychia-
Acad Child Adolesc Psychiatry 1989; 28:836–841 try 1987; 44:826–831
9. Brandenburg NA, Friedman RM, Silver SE: The epidemiology of 32. Kashani JH, Beck NC, Hoeper EW, Fallahi C, Corcoran CM,
childhood psychiatric disorders: prevalence findings from recent McAllister JA, Rosenberg TK, Reid JC: Psychiatric disorders in
studies. J Am Acad Child Adolesc Psychiatry 1990; 29:76–83 a community sample of adolescents. Am J Psychiatry 1987; 144:
10. Verhulst FC: A review of community studies, in The Epidemiol- 584–589; correction, 144:1114
ogy of Child and Adolescent Psychopathology. Edited by Ver- 33. Weyerer S, Castell R, Biener A, Artner K, Dilling H: Prevalence
hulst FC, Koot HM. Oxford, England, Oxford University Press, and treatment of psychiatric disorders in 3–14-year-old children:
1995, pp 146–177 results of a representative field study in the small rural town re-
11. Goldfarb A: Teacher ratings in psychiatric case-finding, I: meth- gion of Traunstein, Upper Bavaria. Acta Psychiatr Scand 1988;
odological considerations. Am J Public Health 1963; 53:1919– 77:290–296
1927 34. Costello EJ, Costello AJ, Edelbrock C, Burns B, Dulcan M, Brent
12. Krupinski J, Baikie AG, Stoller A, Graves J, O’Day DM, Polke P: D, Janiszewski S: Psychiatric disorders in pediatric primary care.
A community health survey of Heyfield, Victoria. Med J Aust Arch Gen Psychiatry 1988; 45:1107–1116
1967; 1204–1211 35. Bird HR, Canino G, Rubio-Stipec M, Gould MS, Ribera J, Ses-
13. Rutter M, Tizard J, Whitmore K: Education, Health and Behav- man M, Woodbury M, Huertas-Goldman S, Pagan A, Sanchez-
ior: Psychological and Medical Study of Childhood Develop- Lacay A, Moscoso M: Estimates of the prevalence of childhood
ment. London, Longman Group, 1970 maladjustment in a community survey in Puerto Rico: the use of
14. Werner EE, Bierman JM, French FE: The Children of Kauai: A combined measures. Arch Gen Psychiatry 1988; 45:1120–1126
Longitudinal Study From the Prenatal Period to Age Ten. Hono- 36. Larson CP, Pless IB, Miettinen O: Preschool behavior disorders:
lulu, University of Hawaii Press, 1971 their prevalence in relation to determinants. J Pediatr 1988; 113:
15. Bjornesson S: Epidemiological investigation of mental disorders 278–285
of children in Reykjavik, Iceland. Scand J Psychol 1974; 15:244– 37. Moilanen I, Almqvist F, Piha J, Rasanen E, Tamminen T: An
254 epidemiological survey of child psychiatry in Finland. Nordisk
16. Leslie SA: Psychiatric disorder in the young adolescents of an Psykiatrisk Tidsskrift 1988; 42:9–11
industrial town. Br J Psychiatry 1974; 125:113–124 38. Velez CN, Johnson J, Cohen P: A longitudinal analysis of se-
17. Rutter M, Cox A, Tupling C, Berger M, Yule W: Attainment and lected risk factors for childhood psychopathology. J Am Acad
adjustment in two geographical areas, I: the prevalence of psy- Child Adolesc Psychiatry 1989; 28:861–864
chiatric disorder. Br J Psychiatry 1975; 126:493–509 39. Wang Y-F, Shen Y-C, Gu B-M, Jia M-X, Zhang AL: An epi-
18. Richman N, Stevenson JE, Graham PJ: Prevalence of behaviour demiological study of behaviour problems in school children in
problems in 3-year-old children: an epidemiological study in a urban areas of Beijing. J Child Psychol Psychiatry 1989; 30:907–
London borough. J Child Psychol Psychiatry 1975; 16:277–287 912
19. Minde KK: Psychological problems in Ugandan school children: a 40. Matsuura M, Okubo Y, Kato M, Kojima T, Takahashi R, Asai
controlled evaluation. J Child Psychol Psychiatry 1975; 16:49–59 K, Asai T, Endol T, Yamada S, Nakane A, Kimura K, Suzuki M:
20. Kastrup M: Psychic disorders among pre-school children in a An epidemiological investigation of emotional and behavioral
geographically delimited area of Aarhus county, Denmark: an problems in primary school children in Japan. Soc Psychiatry
epidemiological study. Acta Psychiatr Scand 1976; 54:29–42 Psychiatr Epidemiol 1989; 24:17–22
21. Zimmerman-Tansella C, Minghetti S, Tacconi A, Tansella M: 41. Esser G, Schmidt MH, Woerner W: Epidemiology and course of
The Children’s Behaviour Questionnaire for completion by psychiatric disorders in school-age children—results of a longi-
teachers in an Italian sample: preliminary results. J Child Psychol tudinal study. J Child Psychol Psychiatry 1990; 31:243–263
Psychiatry 1978; 19:167–173 42. Ekblad S: The Children’s Behavior Questionnaire for completion
22. Earls F, Richman N: The prevalence of behaviour problems in by parents and teachers in a Chinese sample. J Child Psychol
three-year-old children of West Indian-born parents. J Child Psy- Psychiatry 1990; 31:775–791
chol Psychiatry 1980; 21:99–106 43. Sawyer MG, Sarris A, Baghurst PA, Cornish CA, Kalucy RS: The
23. Earls F: Prevalence of behavior problems in 3-year-old children: prevalence of emotional and behavioral disorders and patterns
a cross-national replication. Arch Gen Psychiatry 1980; 37: of service utilization in children and adolescents. Aust NZ J Psy-
1153–1157 chiatry 1990; 24:323–330
24. Earls F: The prevalence of behavior problems in 3-year-old chil- 44. Luk SL, Leung PWL, Bacon-Shrone J, Chung SY, Lee PWH,
dren: comparison of the reports of fathers and mothers. J Am Chen S, Ng R, Lieh-Mak F, Ko L, Yeung CY: Behavior disorder
Acad Child Psychiatry 1980; 19:439–452 in pre-school children in Hong Kong: a two-stage epidemio-
25. Connell HM, Irvine L, Rodney J: Psychiatric disorder in Queens- logical study. Br J Psychiatry 1991; 158:213–221
land primary schoolchildren. Aust Paediatr J 1982; 18:177–180 45. Koot HM, Verhulst FC: Prevalence of problem behavior in Dutch
26. McGee RM, Silva PA, Williams S: Behavior problems in a popu- children aged 2–3. Acta Psychiatr Scand Suppl 1991; 367:1–37

724 Am J Psychiatry 155:6, June 1998

 ROBERTS, ATTKISSON, AND ROSENBLATT

46. Garrison CZ, Addy CL, Jackson KL, McKeown RE, Waller JL: ogy. Edited by Rutter M, Tuma AH, Lann IS. New York, Guil-
Major depressive disorder and dysthymia in young adolescents. ford Press, 1988, pp 87–112
Am J Epidemiol 1992; 135:792–802 66. Shrout PE, Skodol AE, Dohrenwend BP: A two-stage approach
47. Bergeron L, Valla JP, Breton JJ: Pilot study for the Quebec Child for case identification and diagnosis: first-stage instruments, in
Mental Health Survey, part I: measurement of prevalence esti- Mental Disorders in the Community: Progress and Challenge:
mates among six to 14 year olds. Can J Psychiatry 1992; 37:374–385 Proceedings of the American Psychopathological Association,
48. Fergusson DM, Horwood LJ, Lynskey MT: Prevalence and co- 42nd ed. Edited by Barrett JE, Rose RM. New York, Guilford
morbidity of DSM-III-R diagnoses in a birth cohort of 15 year Press, 1986, pp 286–303
olds. J Am Acad Child Adolesc Psychiatry 1993; 32:1127–1134 67. Kraemer HC; Thiemann S: How Many Subjects? Statistical
49. Kasmini K, Kyaw O, Krishnaswamy S, Ramli H, Hassan S: A Power Analysis in Research. Beverly Hills, Calif, Sage Publica-
prevalence survey of mental disorders among children in a rural tions, 1987
Malaysian village. Acta Psychiatr Scand 1993; 87:253–257 68. Murphy JM: Depression screening instruments: history and is-
50. Matsuura M, Okubo Y, Kojima T, Takahashi R, Wang Y-F, sues, in Depression in Primary Care: Screening and Detection.
Shen YC, Lee CK: A cross-national prevalence study of children Edited by Attkisson CC, Zich JM. New York, Routledge, 1990,
with emotional and behavioral problems—a WHO collaborative pp 65–83
study in the Western Pacific Region. J Child Psychol Psychiatry 69. Shrout PE: Statistical design of screening procedures. Ibid, pp
1993; 34:307–315 84–97
51. Lewinsohn PM, Hops H, Roberts RE, Seeley JR, Andrews J: 70. Roberts RE, Lewinsohn PM, Seeley JR: Screening for adolescent
Adolescent psychopathology, I: prevalence and incidence of de- depression: a comparison of depression scales. J Am Acad Child
pression and other DSM-III-R disorders in high school students. Adolesc Psychiatry 1991; 30:58–66
J Abnorm Psychol 1993; 102:133–144 71. Gutterman EM, O’Brien JD, Young JG: Structured diagnostic
52. Morita H, Suzuki M, Suzuki S, Kamoshita S: Psychiatric disor- interviews for children and adolescents. J Am Acad Child Ado-
ders in Japanese secondary school children. J Child Psychol Psy- lesc Psychiatry 1987; 26:621–630
chiatry 1993; 34:317–332 72. Richters JE: Depressed mothers and informants about their chil-
53. Stallard P: The behaviour of 3-year-old children: prevalence and dren: a critical review of the evidence of distortion. Psychol Bull
parental perception of problem behaviour: a research note. J 1992; 112:485–499
Child Psychol Psychiatry 1993; 34:413–421 73. Connell HM, Irvine L, Rodney J: The prevalence of psychiatric
54. Fombonne E: The Chartres Study, I: prevalence of psychiatric disorder in rural school children. Aust NZ J Psychiatry 1982;
disorders among French school-aged children. Br J Psychiatry 16:43–46
1994; 164:69–79 74. Costello EJ: Primary care pediatrics and child psychopathology:
55. Gomez-Beneyto M, Bonet A, Catala MA, Puche E, Vila V: Preva- a review of diagnostic, treatment and referral practices. Pediat-
lence of mental disorders among children in Valencia, Spain. rics 1986; 78:1044–1051
Acta Psychiatr Scand 1994; 89:352–357 75. Burns BJ: Mental health service use by adolescents in the 1970s
56. Burns BJ, Costello EF, Angold A, Tweed D, Stangl D, Farmer and 1980s. J Am Acad Child Adolesc Psychiatry 1991; 30:144–150
EMZ, Erkanli A: Children’s mental health service use across 76. Costello EJ, Angold A: Toward a developmental epidemiology of
service sectors. Health Affairs 1995; 14(3):147–159 the disruptive behavior disorders. Developmental Psychopathol-
57. Costello EJ, Angold A, Burns BJ, Stangl D, Tweed D, Erkanli A, ogy 1993; 5:91–101
Worthman CM: The Great Smoky Mountains Study of Youth: 77. Jensen PS, Koretz D, Locke BZ, Schneider S, Radke-Yarrow M,
goals, design, methods, and the prevalence of DSM-III-R disor- Richters JE, Rumsey JM: Child and adolescent psychopathology
ders. Arch Gen Psychiatry 1996; 53:1129–1136 research: problems and prospects for the 1990s. J Abnorm Child
58. Jensen PS, Watanabe HK, Richters JE, Cortes R, Roper M, Liu Psychol 1993; 21:551–580
S: Prevalence of mental disorder in military children and adoles- 78. Costello EJ, Tweed D: A Review of Recent Empirical Studies
cents: findings from a two-stage community survey. J Am Acad Linking the Prevalence of Functional Impairment With That of
Child Adolesc Psychiatry 1995; 34:1514–1524 Emotional and Behavioral Illness or Disorder in Children and
59. Shaffer D, Fisher P, Dulcan MK, Davies M, Piacentini J, Schwab- Adolescents. Washington, DC, Center for Mental Health Serv-
Stone ME, Lahey BB, Bourdon K, Jensen PS, Bird HR, Canino ices, 1994
G, Regier DA: The NIMH Diagnostic Interview Schedule for 79. Canino G, Bird HR, Shrout PE, Rubio-Stipec M, Bravo M, Mar-
Children version 2.3 (DISC-2.3): description acceptability, tinez R, Sesman M, Guevara LM: The prevalence of specific psy-
prevalence rates, and performance in the MECA study. J Am chiatric disorders in Puerto Rico. Arch Gen Psychiatry 1987; 44:
Acad Child Adolesc Psychiatry 1996; 35:865–877 127–133
60. Costello EJ: Developments in child psychiatric epidemiology. J 80. Henry GT: Practical Sampling: Applied Social Research Methods
Am Acad Child Adolesc Psychiatry 1989; 28:836–841 Series, vol 21. Newbury Park, Calif, Sage Publications, 1990
61. Jablensky A: Epidemiologic surveys of mental health of geo- 81. Earls F: Epidemiology and child psychiatry: entering the second
graphically defined populations in Europe, in Community Sur- phase. Am J Orthopsychiatry 1989; 59:279–283
veys of Psychiatric Disorders. Edited by Weissman MM, Myers 82. Brady EU, Kendall PC: Comorbidity of anxiety and depression
JK, Ross CE. New Brunswick, NJ, Rutgers University Press, in children and adolescents. Psychol Bull 1992; 111:244–255
1986, pp 257–313 83. Bird HR, Gould MS, Staghessa BM: Patterns of diagnostic co-
62. Angold A, Costello EJ: Developing a developmental epidemiol- morbidity in a community sample of children aged 9 though 16
ogy, in Rochester Symposium on Developmental Psychopathol- years. J Am Acad Child Adolesc Psychiatry 1993; 32:361–368
ogy, vol 3: Models and Integrations. Edited by Cicchetti D, Toth 84. Cohen S, Kessler RC, Gordon LU: Strategies for measuring stress
SL. Rochester, NY, University of Rochester Press, 1991, pp 75–96 in studies of psychiatric and physical disorders, in Measuring
63. Cook TD; Campbell DT: Quasi-Experimentation: Design and Analy- Stress: A Guide for Health and Social Scientists. Edited by Cohen
sis Issues for Field Settings. Boston, Houghton Mifflin, 1979 S, Kessler RC, Gordon LU. New York, Oxford University Press,
64. Lahey BB, Flagg EW, Bird HR, Schwab-Stone ME, Canino G, 1995, pp 3–28
Dulcan MK, Leaf PJ, Davies M, Brogan D, Bourdon K, Horwitz 85. Roger D: Emotion control, coping strategies, and adaptive be-
SM, Rubio-Stipec M, Freeman DH, Lichtman JH, Shaffer D, havior, in Stress and Emotion: Anxiety, Anger, and Curiosity, vol
Goodman SH, Narrow WE, Weissman MM, Kandel DB, Jensen 15. Edited by Spielberger CD, Sarason IG, Brebner JMT, Green-
PS, Richters JE, Regier DA: The NIMH methods for the epidemi- glass E, Laungani P, O’Roark AM. Washington, DC, Taylor &
ology of child and adolescent mental disorders (MECA) study: Francis, 1995, pp 255–264
background and methodology. J Am Acad Child Adolesc Psy- 86. Uchino BN, Cacioppo JT, Kiecolt-Glaser JK: The relationship
chiatry 1996; 35:855–864 between social support and physiological processes: a review
65. Edelbrock C, Costello AJ: Structured psychiatric interviews for with emphasis on underlying mechanisms and implications for
children, in Assessment and Diagnosis in Child Psychopathol- health. Psychol Bull 1996; 119:488–531

Am J Psychiatry 155:6, June 1998 725