Beruflich Dokumente
Kultur Dokumente
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted
digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about
JSTOR, please contact support@jstor.org.
Sage Publications, Inc., American Sociological Association are collaborating with JSTOR to digitize,
preserve and extend access to American Sociological Review
This content downloaded from 138.26.87.12 on Wed, 18 May 2016 17:55:11 UTC
All use subject to http://about.jstor.org/terms
PETER A. LACHENBRUCH
University of California, Los Angeles
A sociological model for the mental health consequences of social organization is distinguished from a sociomedical model for the social etiology of particular disorders. Both
models use stress to explain associations between social placement and disorder. These
models are not interchangeable, despite apparent similarities, but researchers frequently
apply the sociomedical model to sociological questions. Discrepancies between models are
illustrated with survey data collected from a community sample of adults. We demonstrate
that gender differences in the impact of stress are disorder-specific and do not indicate
general differences between women and men in susceptibility to stress.
countered difficult, harsh, or traumatic life conditions. Elevated rates of disorder also were attrib-
This content downloaded from 138.26.87.12 on Wed, 18 May 2016 17:55:11 UTC
All use subject to http://about.jstor.org/terms
sequence of social organization, i.e., as an interserved that the various structural arrangements
tive insofar as these persons do not have the disorder being investigated.4 This classification,
however, is misleading from the perspective of
of stress.
stress.
order as a proxy for mental health status compromises the assessment of the consequences of stress
Sociomedical Paradigms
We argue that the sociomedical paradigm is wellsuited for identifying etiologic factors for particular disorders, but is inherently inadequate for
identifying the mental health consequences of
social organization. Our central critique of the
empirical stress literature is that the impact of stress
for psychological functioning overall. Specifically, etiologic estimates for particular disorders
are biased estimates of the overall impact of stress.
Moreover, this misclassification is nonrandom:
the proportion misclassified as not disordered
increases as stress exposure increases precisely
because these are stress-related disorders. Con-
ticular types of disorder because persons with symptoms of other types of stress-related disorder are classified as asymptomatic, i.e., unaffected by exposure
to stress.
This content downloaded from 138.26.87.12 on Wed, 18 May 2016 17:55:11 UTC
All use subject to http://about.jstor.org/terms
sion of the misclassification issue, with the added complication that subgroups of the population
may vary in the rate of occurrence of the various
stress-related disorders included in the "nondisordered" category. Group comparisons of stressreactivity for a single disorder are predicated on
The slippage between the two paradigms is illustrated by the case of gender. Sex role theory attributes higher rates of disorder among women
digm in stress research has limited conceptualizations of stressful life experience. The treatment of stress as an independent variable has
produced a measurement strategy designed to
purge any potential contamination by disorder
from the assessment of stress. First, research has
concentrated on acute events because chronic
difficulties are likely to be influenced by the individual's psychological state insofar as the person is an active participant in those difficulties
connections between gender and stress that produce more disorder among women than among
men. Applied research, however, implicitly has
adopted the sociomedical, disorder-specific paradigm, and focused upon symptoms of depression and/or anxiety, commonly referred to as
nonspecific psychological distress (Dohrenwend,
Shrout, Egri, and Mendelsohn 1980). The label of
nonspecific psychological distress, however, is a
misnomer because numerous possible manifestations ofdistress, e.g., substance abuse/dependence,
are omitted.
This content downloaded from 138.26.87.12 on Wed, 18 May 2016 17:55:11 UTC
All use subject to http://about.jstor.org/terms
Forsythe 1983).
Measures
a highly structured instrument designed for administration by lay interviewers (Robins, Helzer,
four sites.
design (census blocks and households), with random selection of adults (18 years or older) within
households. A 68 percent response rate yielded a
This content downloaded from 138.26.87.12 on Wed, 18 May 2016 17:55:11 UTC
All use subject to http://about.jstor.org/terms
170
AMERICAN
SOCIOLOGICAL
REVIEW
Generalized anxiety disorder, while not included in the version of the DIS used in this
debate among sociologists, psychiatrists, and epidemiologists. These measurement issues, however, are extraneous to the issue of model mis-
configuration of these analytic categories is immaterial to the central issues examined here. Any
categorization that falls short of adequately representing all stress-related disorders would suffice to illustrate the slippage between theoretical
and empirical models.
We define disorder using both diagnostic criteria (presence/absence of disorder) and symptom counts. This dual approach is used because
the recent emphasis in psychiatric epidemiology
is on diagnosable disorders, while most previous
stress research concerned symptomatology.
Moreover, debate continues about whether diagnosable disorders or symptoms are the most appropriate object of inquiry (Mirowsky and Ross
1989b). The two operationalizations are similar
curring to the individual, recent research suggests that events occurring to significant others
may be especially important to women (Kessler
imum number of symptoms necessary for a positive diagnostic status. Diagnostic ascertainment
This content downloaded from 138.26.87.12 on Wed, 18 May 2016 17:55:11 UTC
All use subject to http://about.jstor.org/terms
(.6) (.9)
ditions -having difficulty affording food, clothing, medical care, and furniture and paying bills
(Pearlin, Lieberman, Menaghan, and Mullan
1981). This count ranges from zero to five with a
mean of 1.40, a standard deviation of 1.67, and
internal consistency reliability of .81. Because
comprehensive assessment of psychiatric disorder
episode
(.2) (.2)
Panic
disorder
.1
.5
(.1) (.2)
Phobia
2.6
8.4
(.4) (.7)
Obsessive-compulsive .5 .3
.1
time was devoted to potential psychosocial antecedents of disorder. Consequently, our stress
measures are not ideal and are not an extensive
.0
(.0) (.1)
(.2) (.2)
Somatization
.0
.0
(.0) (.0)
Any substance-use disorder: 9.2 2.4
(.8) (.5)
Analysis
We compare disorder-specific models with models for all types of disorder combined. The disorder-specific models are: the presence/ absence of
any affective or anxiety disorder and the presence/absence of any substance-use disorder; the
combined model considers the presence/ absence
of any disorder. Parallel disorder-specific models
are estimated for symptom counts.
(.7) (.4)
Drug abuse/dependence 1.8 .7
(.4) (.2)
Other disordersb .3 .2
(.1) (.1)
All
disorders
14.3
14.7
(.9) (1.0)
Maximum weighted number of cases 1,472 1,632
a Multiple diagnoses are possible.
b "Other disorders" refers to schizophrenia, schizophreniform disorder, and antisocial personality.
Note: Standard errors in parentheses.
RESULTS
This content downloaded from 138.26.87.12 on Wed, 18 May 2016 17:55:11 UTC
All use subject to http://about.jstor.org/terms
Female
1.55*
1.86*
.23***
.17***
.71*
.79
Ethnicity:a
(years)
Income
.99
(000)
Employed
Married
.99
.98**
.86
.50*
.98***
.98**
.86
.50*
.98***
.99
.97
.72
.99
.99**
.99**
.99**
.99**
.98
.73
.91
.58**
.92
.59**
Mexican-American
Other
1.02*
Events-self
1.16*
1.02
1.02*
1.13
1.02
.99
.99
.99
1.13
1.22*
.99
1.01
1.02
1.19***
1.01
1.02
1.25**
Female
Female
Female
x
x
x
events-self
1.05
events-other
financial
Intercept
.11
strain
.09
.21
-.71**
.83*
1.05
.91
1.19
1.16
.23
.30
.92
1.08
.28
Note: Model II differs from Model I by the inclusion of gender-by-stressor interaction terms.
pressive disorder episodes and phobia among females. Recent affective or anxiety disorders and
activity is unlikely to account for the gender distributions of both affective or anxiety disorders
and substance-use disorders because these gen-
disorder - affective or anxiety disorder and substance-use disorder - and for any psychiatric
disorder. The gender-by-stressor interaction terms
for particular disorders. These terms signify genderdifferences in overall stress-reactivity only for
the model for all disorders. Logistic regression
This content downloaded from 138.26.87.12 on Wed, 18 May 2016 17:55:11 UTC
All use subject to http://about.jstor.org/terms
female continues to be associated with an increased risk of disorder when other characteristics and stress-exposure are held constant. This
gender difference is greater among the married
than the unmarried. Mexican-Americans and
those of other ethnic backgrounds are at lower
risk than non-Hispanic whites. Increased income
is associated with a decreased risk of affective or
anxiety disorder, except among those of other
ethnic backgrounds. All three stressors are independently associated with an increased risk of
affective or anxiety disorder (Model I). Adding
the three gender-by-stressor interactions (Model
II) alters these associations. Negative life events
occurring to someone important to the respondent
financial strain stressor attains statistical significance for women (el = 1.23; p < .05), but not men
ders: events-self (el = 1.14; p > .05), events-other (el = 1.40; p < .001), and financial strain
(el = 1.14; p > .05). The financial strain and
events-self coefficients, however, are of similar
magnitude for women and men; the difference in
p> .05).
This content downloaded from 138.26.87.12 on Wed, 18 May 2016 17:55:11 UTC
All use subject to http://about.jstor.org/terms
Age x ethnicity:a
Female
.048
-.049**
events-self
pact of gender is muted considerably. Being female slightly decreases the risk of a disorder, a
gender difference that is smaller among the married than among the unmarried. Increasing age
Female
.012
-.020
events-other
Female
.064*
-.004
financial
strain
This content downloaded from 138.26.87.12 on Wed, 18 May 2016 17:55:11 UTC
All use subject to http://about.jstor.org/terms
en only.
are replicated. Financial strain, however, is significantly related to symptoms among women
(B =.01l;p <.01), but not among men (B = .010;
significance in the combined analysis, this gender difference must be viewed with skepticism.
These separate analyses, then, demonstrate that
events to others exert similar effects for women
This content downloaded from 138.26.87.12 on Wed, 18 May 2016 17:55:11 UTC
All use subject to http://about.jstor.org/terms
fluence typical and extreme patterns of symptoms in the same manner. We would add that
(Kessler and McLeod 1984). Alternative explanations for this unexpected finding include: the
contains a high proportion of Mexican-Americans; the specific nature of the types of events
gender differences in stress-reactivity were replicated for the composite category of all disorders. Indeed, events occurring to others appear to
be more important for men than women for affective or anxiety disorder, but more important
for women than men for substance-use disorder.
Gender contingencies also do not generalize from
one type of symptom count to the other. Gender
differences in the impact of stress are disorderspecific and do not indicate general differences
between men and women in stress-reactivity.
toms that correspond to clinically defined entities, typically more severe forms of disorder.
This content downloaded from 138.26.87.12 on Wed, 18 May 2016 17:55:11 UTC
All use subject to http://about.jstor.org/terms
CAROLYN M. RU-TER is a Ph.D. candidate in Biostatistics at the UCLA School of PublicHealth with a minor
in Psychiatric Epidemiology. Her research interests
include logistic and log-linear models; her dissertation is concerned with spatial statistics.
PETER A. LACHENBRUCH is Professor of Biostatistics,
UCLA School of Public Health. His recent work is
concerned with small sample properties, sample size
determination, and discriminant analysis.
REFERENCES
This content downloaded from 138.26.87.12 on Wed, 18 May 2016 17:55:11 UTC
All use subject to http://about.jstor.org/terms
dine de Gruyter.
havior 30:11-25.
41:949-58.
ior 27:161-78.
30:241-56.
Turner, R. J. and Samuel Noh. 1983. "Class and Psychological Vulnerability Among Women: The Significance of Social Support and Personal Control."
havior 22:337-56.
Review 55:209-23.
This content downloaded from 138.26.87.12 on Wed, 18 May 2016 17:55:11 UTC
All use subject to http://about.jstor.org/terms