The Identification of Psychiatric Illness by Primary
The Effect of Patient Gender
PAUL D. CLEARY, PhD, BARBARA J. BURNS, PhD, GREGORY R. NYCZ
Objective: This study tested several hypotheses about why
women are more likely than men to have psychiatric dis-
orders noted by their primary care physicians.
Design: Patients were screened for mental disorders using
the General Health Questionnaire. A stratified sample was
assessed using the Schedule for Affective Disorders and
Schizophrenia. Information on utilization and identifica-
tion of mental health problems was abstracted from the
Setting: The study was conducted at a multispeciaity group
practice in a semirural area of Wisconsin.
Patients: Study participants consisted of a stratified prob-
ability sample of 247patients seeking primary care.
Results: Patients with a psychiatric illness who were rela-
tively frequent users of the clinic were most likely to be
identified by a physician as having a mental health prob-
lem. When psychiatric illness and utilization rates were
statistically controlled, men and women had comparable
Key words: psychiatric illness; gender, diagnosis; mental
illness. J GEN INTERN MED 1990; 5:355-360.
IN RECENT YEARS, attention increasingly has focused on
the role of primary care physicians in the recognition
and treatment of mental health problems. 1-3 There are
several reasons for this attention to the interface be-
tween primary care and mental health services. Many
studies, conducted in different settings and using dif-
ferent methodologies, have demonstrated that a dispro-
portionate number of patients receiving general medi-
cal care have relatively serious mental health problems,
and that patients with mental health problems tend to
use medical services frequently. 4-~1 In fact, primary
health care physicians have been described as compris-
ing a de facto mental health services system. 6 Despite
the importance of mental health problems in primary
care settings, the available data indicate that many pa-
tients with psychiatric diagnoses are not recognized as
having a mental health problem), s, 12-14
It is not clear at this point whether treatment or
referral by a primary care provider is the most efficient
Received from the Department of Health Care Policy, Harvard
Medical School, Boston, Massachusetts (PDC), Duke University, Dur-
ham, North Carolina, (BJB), and Marshfield Medical Foundation,
Marshfield, Wisconsin (GRN).
Supported in part by a contract (DBE-77-0071) and grant from
the National Institute of Mental Health (MH-33940) and a grant from
the Robert Wood Johnson Foundation.
Address correspondence and reprint requests to Dr. Cleary: De-
partment of Health Care Policy, Harvard Medical School, 25 Shattuck
Street, Parcel B, 1st Floor, Boston, MA 02115.
or efficacious way of treating patients with mental
health problems. Most agree, however, that in order to
provide appropriate care, primary care physicians
should be aware of whether or not a patient is suffering
from a mental disorder. Understanding the factors un-
derlying the recognition of mental health problems is
an important step in improving the treatment of pa-
tients with such problems.
A predictor of physician identification of psychiat-
ric problems is gender; women are much more likely to
be identified as having such problems. 4, 12, ts, 16 There
are few data on whether the higher identification rates
for women reflect more accurate diagnosis or simply
more frequent labeling of women's problems as psychi-
atric. We have analyzed physician identification of
mental health problems and examined the extents to
which frequency of physician visits, number of self-
reported symptoms, and patient gender are related to
the accuracy of physician assessments of psychiatric
Compared with men, women more frequently re-
port psychiatric symptoms, are more likely to see them-
selves as having emotional problems, are more likely to
seek help from general physicians and psychiatrists, are
more likely to receive psychiatric treatment in outpa-
tient clinics, general hospitals, and mental hospitals,
and are more likely to receive medications prescribed
for the treatment of mental illness. 17, is Some have in-
terpreted these various findings as indicating that
women have higher rates of mental disorders than men.
An alternative explanation is that, for any particular
level of distress or psychiatric illness, women are more
likely to seek professional help for emotional problems
and to be recognized as having problems by their doc-
tors because they are more likely to make their feelings
or problems known. 19 It also may be that physicians
have more opportunity to interact with women patients
and discuss their personal and emotional problems be-
cause women have higher utilization rates than men. 2°
Both men and women frequently seek medical care
for vague and diffuse symptoms. Such complaints are
often not easily described by a specific diagnosis, and
they may be labeled psychogenic if the patient visits the
doctor frequently. Since women have higher utilization
rates, they may be more likely to have such complaints
Clearyeta/., PATIENT GENDER AND IDENTIFICATION OF PSYCHIATRIC ILLNESS
labeled mental health problems. It may also be that, for
a given level of distress, women are more likely to be
perceived as having mental illness because they are
viewed as weaker, more emotional, and less able to deal
with stress than men. 2t, 22
In this study we examined the hypotheses that fe-
male patients are more likely to be identified by pri-
mary care providers as having psychiatric disorders be-
cause: 1) women have more psychiatric illness than
men, 2) women are more likelyto make their problems
evident to a physician, 3) women are more likely to be
labeled as having mental health problems because they
are high users of medical services, 4) women have more
opportunity for discussing psychosocial problems with
their physicians because they have higher utilization
rates than men, and 5) physicians have a stereotype of
women that biases their assessment.
METHODS AND PROCEDURES
The study was conducted at a 175-physician multi-
specialty group practice located in a semirural area in
central Wisconsin. At the time of the study, the town
had a population of approximately 17,000, with about
50,000 persons in the surrounding area. A central
clinic provided primary health care services for all resi-
dents in this immediate area through both a prepaid
group practice plan and a fee-for-service arrangement.
The clinic had primary, specialty medical, and mental
health care services physically and administratively
Patients and Methods
A total of l, 327 consecutive primary care patients,
aged 18 years and older, who resided in the study area
and who used the central clinic for primary care (family
practice, general internal medicine, pediatrics, or im-
mediate care), were contacted and asked to take part in
the study. A total of 1,072 patients (81%) agreed to
participate and were screened for the presence of a
mental disorder using the self-administered, 30-item
General Health Questionnaire (GHQ23). The GHQ has
been used in several studies of primary care popula-
tions and has been shown to be predictive of psychiatric
Based on these screening results, a disproportion-
ate stratified random subsample of 350 patients was
selected. Patients with high GHQ scores were oversam-
pied to increase the number of patients with psychiat-
ric disorders in the sample for other analyses. Seventy-
one percent of these patients, 192 with GHQ scores of
4 or above and 55 with lower scores, completed a com-
prehensive psychiatric interview, the Schedule for Af-
fective Disorders and Schizophrenia-Lifetime version
(SADS-L). The SADS-L is a clinical psychiatric interview
that systematically evaluates the possibility of specific
diagnoses.25, 26 These interviews were conducted by a
psychiatrist and two psychiatric social workers, and the
results were not revealed to the primary care providers.
In psychiatric studies of community residents or of
ambulatory medical patients, it is important to identify
symptoms that are related to underlying medical ill-
ness. A previous study of the relationship between
SADS-L diagnosis and medical diagnosis in this popula-
tion revealed a significant association between psychi-
atric disorders and genitourinary disorders (for female
patients) and gastrointestinal disorders (for male pa-
tients).27 The psychiatric examiner reviewed the medi-
cal records of all patients for whom a mental disorder
diagnosis might have been falsely given as a result of an
associated physical disorder, but no case of a spurious
physical cause of a psychiatric diagnosis was found.
Comparisons of participants and non-participants
indicated that those who refused to complete the GHQ
were slightly older, more likely to have had a diagnosis
of mental disorder in the preceding year, likely to have
had fewer physician visits in the past year, and less
likely to be members of a prepaid plan. Among those
completing the GHQ, patients who did not complete
the SADS-L were slightly younger, were less likely to
have had a diagnosis of mental disorder in the past year,
and had slightly lower GHQ scores than did those who
did complete the SADS-L.
Patients participating in the SADS-L interview were
weighted by the inverse of their original probabilities
of selection and by nonresponse factors so that the anal-
ysis sample had characteristics similar to those of the
1,072 patients enrolled in the study. Weights also were
used to adjust for the fact that because patients were
selected during a brief period, the original sample had a
greater proportion of high users than if the sampling
had taken place over an entire year. The data analyzed
for this paper consist of a weighted sample of 233 pa-
tients" who completed both the GHQ and the SADS-L
interview. The weighted sample did not differ from the
group of all eligible patients with respect to age,
gender, method of payment, prevalence of diagnosed
mental disorders, or health care utilization.
Utilization rates for the six-month period follow-
ing administration of the GHQ and SADS-L were used as
a measure of exposure to a physician, because that pe-
"For ease of comparison, all statistical tests were calculated on
tables produced from the weighted data without taking into account
the effect of weighting on the distribution of the test statistics.
Readers should be aware of this limitation when interpreting signifi-
cance levels. All analyses were also done using unweighted data, and
none of the substantive conclusions differed.
All the physicians in the study saw more than one patient. Pa-
tients were sampled consecutively and not by physician, however, so
the sample is not a cluster sample. Inferences are made about the
experiences of patients, not physician behaviors.
JOURNAL OF GENERAL INTERNAL MEDICINE, Volume 5 (July/August), t990
riod corresponds to the period for which the medical
charts were reviewed for indications of identification.
On the basis of a systematic review of all subjects' medi-
cal records, a dichotomous variable was constructed
indicating whether there was any indication in the pa-
tient's chart of diagnosed psychiatric illness, psychiat-
ric symptoms (as evaluated by a board-certified psychi-
atrist), psychotropic drug medication, referral to a
mental health specialist, or psychotherapy at any time
during this period.
The identification of psychiatric illness is de-
scribed by calculating the proportions of different sub-
sets of patients for whom there was any indication in
their medical records of diagnosed psychiatric illness,
symptoms, treatment, or referral during the six months
prior to the index visit. To evaluate the relationship
between the factors hypothesized to influence identifi-
cation, we developed a multivariate logistic model. 28
The independent variables analyzed were patients'
SADS-L results (no diagnosis, some diagnosis), GHQ
results (score of less than 4, score of 4 or more),~-
number of visits to the clinic in the six months follow-
ing study entry (1, 2, 3-4, 5+), and gender (male,
The data on physician identification were consist-
ent with other findings in the literature; women were
more likely than men to be recorded as having a mental
health problem. Review of the medical charts indicated
that for 22.3% (28) of the women and 11.9% (13) of
the men, a physician had indicated the existence of
some type of mental health problem (Table 1). Consist-
ent with our first hypothesis, more women (32.5%)
than men (21.5%) were judged to be currently psychi-
atrically ill on the basis of the SADS-L interview. The
most prevalent diagnoses (using Research Diagnostic
Criteria and the SADS-L evaluation) were major depres-
sion (14 patients), phobic disorder (13 patients), in-
termittent depression (12 patients), labile personality
(9 patients), minor depression (12 patients), cyclothy-
mic personality (5 patients), and generalizedanxiety
(4 patients). As expected, the presence of a psychiatric
diagnosis was related to its being identified by the phy-
sician. The relationship between the diagnosis and the
identification was weak, however, suggesting that
other factors are important in the identification pro-
~fAlthough responses to the GHQ can be scored to yield a con-
tinuous score, it is usually used as a dichotomy. Thus, persons report-
ing four or more symptoms are classified as having a "positive" score
and those reporting fewer than four symptoms are classified as having
a "negative" score. We follow that convention in the analyses pre-
Percentages of Patients Identified by Their Physicians as Having Mental
Health Problems by Patient Characteristics
Results of psychiatric interview
No current illness
Negative (O- 3)
Number of visits
Identified as Having
Mental Health Problem
cess. Among patients with no psychiatric diagnosis, the
proportion identified as having such a diagnosis was
14.6%, whereas of those with a psychiatric diagnosis,
25.5% were identified as having a problem. Also con-
sistent with our hypotheses, patients with high GHQ
scores were more likely to be identified as having a
mental health problem.
There was a strong relationship between the num-
ber of visits in the preceding six months and identifica-
tion of a mental health problem, with the rates of iden-
tification varying from 11% among those with only one
visit to 35% among those with five or more visits. To
investigate further the relationship between visit rates,
diagnosed psychiatric illness, and physician identifica-
tion of psychiatric illness, we calculated the probabili-
ties of being so identified for patients with and without
psychiatric illness within each category of visit rate
The data in Table 2 indicate that identification
rates were positively related to utilization only among
those patients who were psychiatrically ill. In general,
Percentages of Patients Identified by Their Physicians as Having
Mental Health Problems
Number of Visits
2 1 3 or 4 S or More
No psychiatric illness 14%
Some psychiatric illness 0%
*Number of patients in parentheses.
Cleary eta/., PATIENT GENDER AND IDENTIFICATION OF PSYCHIATRIC ILLNESS
Logit Coefficients for Model of Physician Identification of
Mental Health Problems
(p) Independent Variable Coefficient
Results of psychiatric interview
No current illness
Negative (0- 3)
Number of visits
Illness- visit interaction
III X 1 visit
111X 2 visits
II1X 3-4 visits
III X 5+ visits
these data support the differential-exposure hypothesis
and not the labeling hypothesis. If the labeling hypoth-
esis had been correct, there would have been a rela-
tionship between utilization and identification even
among patients who were not psychiatrically ill, which
was not the case.
The multivariate model that was developed to test
more rigorously the study hypotheses included gender,
the presence or absence of SADS-L-defined psychiatric
illness, whether the patient had a high or a low GHQ
score, utilization in the previous six months, and the
interaction between psychiatric illness and utilization
as independent variables. That model fit the data well
(likelihood-ratio chi-square ---- 24.27;
freedom = 22; p = 0.333). The coefficients for the re-
lationships between the independent variables and
identification are presented in Table 3.
In the multivariate model, the strongest predictor
of identification of mental health problems was the
interaction between visit rate and psychiatric illness,
providing support for the hypothesis that identification
rates are related to the presence of psychiatric illness,
especially among patients with more visits. Controlling
for that relationship, GHQ scores have almost no rela-
tionship to identification when other predictors are
controlled. Thus, the hypothesis that women are more
likely to express their problems and that this tendency
helps account for the gender difference in physician-
identified mental illness was not supported.
Controlling for psychiatric illness, utilization
rates, and GHQ scores, gender was not significantly
related to identification rates (p = 0.06). The unad-
justed identification rates for men and women were
11.9% and 22.3%, respectively (Table 1). When the
model presented in Table 3 was re-estimated without
the term relating gender to identification, the pre-
dicted rates for men and women were similar (17.0%
and 18.3%). That small difference in identification
rates may have been due to a number of unmeasured
factors in addition to bias.
The results of this study indicate that patients with
psychiatric illnesses who have high utilization rates are
most likely to be identified by their primary care physi-
cians as having psychiatric illnesses. Controlling for
other predictors, self-reported problems, as measured
by the GHQ, are not statistically significant predictors
of such physician identification.
A possible explanation for the absence of a rela-
tionship between GHQ scales and identification when
other variables were controlled is that the effects of
GHQ scores are mediated by utilization patterns; pa-
tients with negative GHQ scores had an average of 3.93
visits, while those with positive scores had an average
of 4.54 visits. Another explanation for the lack of a
relationship is that the GHQ score may be a poor proxy
for propensity to express problems and is strongly re-
lated to psychiatric illness. However, the zero-order
association between GHQ scores and identification was
not statistically significant (p----0.28). Further work
with better measures of propensity to express problems
is needed to test more rigorously the relationship be-
tween thetendency to express emotional problems and
physician identification of such problems.
It frequently has been argued that physicians are
"biased" by a patient's gender, but these data do not
support that hypothesis. Women may have a tendency
to be expressive in the physician's presence that is not
reflected by GHQ scores. However, the failure of var-
ious investigations to explain gender differences on the
basis of response tendencies leads us to doubt this ex-
planation. 17 It is also possible that the preponderance
of women among clinic users and the tendency for dis-
tressed persons to use clinic facilities more frequently
might lead physicians to misjudge the probability, in
general, of women's having mental health problems.
Stereotypes affect perception, memory, and behav-
ior.22, 29 It is not clear how stereotypes affect the per-
ceptions and responses of physicians to patients, but
this is an important area for further investigation.
The fact that women use medical facilities more
frequently than do men appears to result in somewhat
more accurate recognition of psychiatric problems. For
example, if SADS-L results are used as the criteria, the
accuracy (proportion of correct positives and correct
JOURNAL OF GENERAL INTERNAL MEDICINE. Volume 5 (July~August), 1990
negatives) for women is 3 5.7%, whereas for men it is
26.2%. However, recognition rates for both men and
women are low, and women have higher false-positive
rates (19.0%) than do men (10.7%).
The overall low rate of recognition of psychiatric
disorders is noteworthy. Other studies have found that
primary care providers correctly recognized between
about a third and a half of depressed patients, t3, t4, 30, 3t
with one study finding a recognition rate as high as
73%. 32 The lower identification rates in our study may
be due to a number of factors, but one probable expla-
nation is that medical records do not fully reflect pri-
mary care providers' knowledge about the psychiatric
status of patients. Physicians may not record psychiatric
diagnoses because they are not thought to be relevant
for the medical management of the patient or because
of the sensitive nature of psychiatric diagnoses. This
may be especially true since many of the disorders in
primary care settings are somatoform disorders that fre-
quently are not easily described. We attempted to ad-
dress this problem by defining identification as any
mention of psychiatric symptoms, psychotropic drug
prescriptions, psychotherapy, referral to a mental
health professional, or a psychiatric diagnosis. Never-
theless, it may be that these markers do not fully cap-
ture physicians' recognition and response to psychiat-
The results presented here are important because
they represent one of the few situations in which it has
been possible to control carefully for psychiatric ill-
ness and to examine the independent effects of psychi-
atric illness, self-reported problems, utilization pat-
terns, and patient gender on physician recognition of
psychiatric illness. The measure of identification used
may underrepresent physician "recognition,"
these data provide useful insights into the factors re-
lated to the identification process.
It is clear that physicians are responding to some-
thing other than strictly defined "psychiatric illness."
The relatively low recognition rates for those with psy-
chiatric illness and the relatively high identification
rates for those without psychiatric illness indicate that
physicians probably respond to a variety of factors,
such as decreased social f~anctioning and patient dys-
phoria, that are not necessarily directly related to psy-
Physician responses to patients depend on a com-
plicated and involved series of perceptions and judg-
ments based on clinical experience and the character-
istics of a particular patient. Awareness of these
influences and how they operate would be useful in
sensitizing physicians and patients to their importance
and perhaps improving the quality of the physician-
patient interaction and subsequent outcome. Possible
clinical consequences of the physician identification
patterns in these data are that some women without
psychiatric problems may be receiving inappropriate
attention, while men with psychiatric problems are less
likely to receive needed psychological attention from
their primary care physicians. Perhaps the most impor-
tant finding is the strength of the relationship between
the recognition of psychiatric problems and visit fre-
quency. Unfortunately, the data are not detailed
enough to determine the causal ordering of recognition
and utilization. They suggest, however, that efforts to
improve physician recognition should focus on the in-
frequent users of care as well as on the "over-users."
Darrell A. Regier, Irving D. Goldberg, Edwin W. Hoeper, and Larry G.
Kessler played central roles in the design and conduct of the study
described. The authors thank AHan Horwitz and Ron Kessler for help-
ful comments on an earlier draft.
1. Borns JF, Howes MJ, Devins NP, Rosenberg R, Livingston WW.
Primary health care providers' recognition and diagnosis of men-
tal disorders in their patients. Gen Hosp Psychiat. 1988;
2. Kamerow DB, Pincus HA, MacDonald DI. Alcohol abuse, other
drug abuse, and mental disorders in medical practice. Preva-
lence, costs, recognition, and treatment. JAMA. 1986;
3. Schulberg HC, Burns BJ. Mental disorders in primary care: epi-
demiologic, diagnostic, and treatment research directions. Gen
Hosp Psychiat. 1988;10:79-87.
4. Locke BZ, Gardner EA. Psychiatric disorders among patients of
general practitioners and internists. Public Health Rep. 1969;
5. Eastwood MR. The relation between physical and mental illness.
Toronto: University of Toronto Press, 1975.
6. Regier DA, Goldberg ID, Taube CA. The de facto U.S. mental
health services system: a public health perspective. Arch Gen
7. Hankin JR, Oktay JS. Mental disorder and primary medical care:
an analytical review of the literature. National Institute of Mental
Health Series D. Washington, DC 20402: Superintendent of Doc-
uments, U.S. Government Printing Office, 1979;78-661.
8. Hoeper EW, Nycz GR, Cleary PD, Regier DA, Goldberg ID. Esti-
mated prevalence of RDC mental disorder in primary medical
care. IntJ Mental Health. 1979;8:6-15.
9. Shepherd M, Cooper B, Brown AC, Kalton GW. Psychiatric illness
in general practice. Oxford: Oxford University Press, 1966.
10. Gardner EA" Em°ti°nal dis°rders in medical practice" Ann Intern
1 I. Hankin JR, Steinwachs DM, Burns BJ, Goldberg ID, Regier DA,
Hoeper EW. The utilization of medical care services by persons
with mental disorders. Arch Gen Psychiat. 1982;39:225-31.
12. Gerber PD, Barrett J, Barrett J, Manheimer E, Whiting R, Smith R.
Recognition of depression by internists in primary care: a com-
parison of internist and" gold standard" psychiatric assessments.
J Gen Intern Med. 1989;4:7-13.
13. Goldberg DP. Training family physicians in mental health skills:
implications of recent research findings. In Mental health ser-
vices in general health care. Washington, DC: National Academy
of Sciences, 1979; Vol I.
14. Nielsen AC, Williams TA. Depression in ambulatory medical pa-
tients. Prevalence by self-report questionnaire and recognition
by nonpsychiatric physicians. Arch Gen Psychiat. 1980;
15~ Locke BZ, Krantz C, Kramer M. Psychiatric need and demand in a
prepaid group practice program. Am J Public Health. 1966;
16. Rosen BM, Locke BZ, Goldberg IC, Babigian HM. Identification
360 Download full-text
C/eary etaL, PATIENT GENDER AND IDENTIFICATION OF PSYCHIATRIC ILLNESS
of emotional disturbance in patients seen in general medical
clinics. Hosp Community Psychiat. 1972;23:364-70.
17. Cleary PD. Gender differences in stress related outcomes. In
Barnett RC, Biener L, Baruch GK, eds. Gender and stress. New
York: The Free Press, 1987.
18. Cleary PD. The need and demand for mental health services. In
Taube CA, Mechanic D, Hohmann A, eds. The future of mental
health services research. DHHS Pub. No. (ADM)89-1600. Wash-
ington, DC: Superintendent of Documents, U.S. Government
Printing Office, 1989; 161-84.
19. Phillips D, segal B. Sexual status and psychiatric symptoms. Am
Sociol Rev. 1969;34:58-72.
20. Cleary PD, Mechanic D, GreenleyJR. Sex differences in medical
care utilization: an empirical investigation. J Health Soc Behav.
21. Seller RH, Blascovich J, Lenkei E. Influence of stereotypes in the
diagnosis of depression by family practice residents. J Faro Pract.
22. Colameco S, Becker LA, Simpson M. Sexbias in the assessment of
patient complaints. J Fam Pract. 1983; 16:1117-21.
23. Goldberg DP. The detection of psychiatric illness by question-
naire. London: Oxford University Press, 1972.
24. Cleary PD, Goldberg ID, Kessler LG, Nycz GR. Screening for
mental disorder among primary care patients: usefulness of the
general health questionnaire. Arch Gen Psychiat. 1982;
25. Endicott J, Spitzer RL. A diagnostic interview: the schedule for
affective disorders and schizophrenia. Arch Gen Psychiat.
26. Spitzer RL, Endicott J, Robins L. Research diagnostic criteria:
rationale and reliability. Arch Gen Psychiat. 1968;35:773-82.
27. Kessler LG, Tessler RC, Nycz GR. Co-occurrence of psychiatric
and medical morbidity in primary care. J Faro Pract.
28. Bishop YMM, Fienberg SE, Holland PW. Discrete multivariate
analysis: theory and practice. Cambridge, MA: MIT Press, 1975.
29. McCauley C, Stitt CL, Segal M. Stereotyping: from prejudice to
prediction. Psychol Bull. 1980;87:195-208.
30. German PS, Shapiro S, Skinner EA, et al. Detection and manage-
ment of mental health problems of older patients by primary care
providers. JAMA. 1987;257:489-93.
31. VonKorff M, Shapiro S, Burke J, et al. Anxiety and depression in a
primary care clinic. Arch Gen Psychiat. 1987;44:152-6.
32. Jones L, Badger L, Ficken R, Leper J, Anderson R. Inside the
mental health network. Examining mental health care delivery
of primary care physicians. Gen Hosp Psychiat. 1987;9:287-93.
When the time came for me to begin research I had not a clue how to start. -- P. B.
MEDAWAR, Advice to a Young Scientist, Harper & Row, New York, 1979