Social Problems, Vol. 54, Issue 4, pp. 504–522, ISSN 0037-7791, electronic ISSN 1533-8533. © 2007 by Society for the Study of
Social Problems, Inc. All rights reserved. Please direct all requests for permission to photocopy or reproduce article content
through the University of California Press’s Rights and Permissions website at www.ucpress- journals.com/reprintinfo/asp.
Mental Illness Careers in an Era of Change
Eliza K. Pavalko, Indiana University, Bloomington
Courtenay M. Harding, Boston University
Bernice A. Pescosolido, Indiana University, Bloomington
The concept of illness career conceptualizes the illness experience as a dynamic pathway that takes shape
over time as the person interacts with the treatment system. In this article we use data from 238 persons treated
in Vermont State Hospital during the 1950s to evaluate several fundamental career assumptions and to illustrate
how different predictors are contextualized by the career. We find that, among persons hospitalized for severe
mental illness prior to deinstitutionalization, social status was more strongly associated with the length of the
initial hospitalization than illness or community characteristics. We also find evidence of accumulation of hospi-
tal experience, with the pace of later hospitalizations shaped by earlier ones. Institutional reform was successful
redirecting hospitalization careers and illness characteristics become more salient for the pace of hospitalization
in this post-reform era. While mental health researchers have debated the relative influence of social and illness
characteristics for predicting hospitalization, our findings suggest that the specific influence of any of these char-
acteristics depends on where person is in their own illness career as well as where that career is located in histor-
ical time. Keywords: deinstitutionalization, careers, mental illness, historical change, life course.
A half century ago Erving Goffman (1961) painted a dramatic picture of the illness career
for persons with mental illness and the role hospitals played in shaping that career. Through
processes of mortification and the establishment of barriers to the outside world, he observed
that patients became institutionalized and unable to function outside the hospital walls
(Goffman 1961). The larger significance of Goffman’s observation was his attention to the
ways that organizational structures like total institutions shape individual lives. His focus on
the dual-sidedness of careers that link the individual experience with institutional structures
suggests careers can provide a valuable tool for understanding the interplay between individ-
uals and treatment systems.
While Goffman’s insights have been valuable for understanding the experience of mental
illness and how those experiences are shaped by the treatment system, our understanding of
career variation, particularly the factors shaping the direction those careers will take, remains
in its infancy. The handful of long-term longitudinal studies following persons diagnosed
with severe and persistent mental illness have documented wide variation in outcomes, and
there is substantial long-term variation even among those diagnosed with schizophrenia
(Harding 1988). What remain relatively unknown are the factors that distinguish these path-
ways and how their influence may change across the illness career.
Understanding this variation in mental illness careers is further complicated because
careers unfold over time, and during that time, the institutions shaping the career are also
This research was supported in part by research grants from the National Institute of Mental Health (R24–
MH51669 and K02–MH42655) to the third author and from the National Institute on Aging (AG11564) to the first author.
The authors wish to thank Shiva Gautam and Camilla Saulsbury for their assistance in accessing the data and Scott Long,
Jane McLeod, and Brian Powell for comments on earlier drafts. Direct all correspondence to: Eliza Pavalko, Depart-
ment of Sociology, 744 Ballantine Hall, 1020 East Kirkwood Ave., Indiana University, Bloomington, IN 47405-7103.
Mental Illness Careers
changing. Since the time of Goffman’s observations at St. Elizabeth’s Hospital, the mental
health treatment system has been transformed. Methods used to treat mental illness, the
structure of mental hospitals, and the role of those hospitals within the treatment delivery
system have all changed. While this transformation has been well documented by both soci-
ologists and historians (Brown 1985; Grob 1994; Pescosolido and Rubin 2000; Shorter 1997)
we know relatively little about how the illness careers of those moving through the system
during this time were shaped by these institutional shifts. An even greater unknown is the
extent to which having an established illness career when reforms were implemented made
that career less malleable to institutional change.
This article uses data from one of the rare long-term studies that followed persons insti-
tutionalized for mental illness during the 1950s to address these issues. Because of the com-
plexity involved in assessing both individual and institutional change, we focus on a single
dimension of the illness career—the pace of moves in and out of the hospital.1 Our general
purpose is to evaluate several basic assumptions of the career concept and to assess how fac-
tors influencing the pace of hospitalization—in this case the length of hospitalization and
length of time between hospitalizations—are contextualized by where they fall in the career.
We thus focus on three specific empirical questions. First, what is the relative influence of ill-
ness characteristics (type and severity of the illness), status characteristics (gender, social class
background), and the community setting in which one resided prior to admission (rural,
lived alone) at different points in the illness career? Second, is the pace of movement in and
out of the hospital cumulative over time, such that longer stays or frequent readmissions
early in the career alter the length of hospitalizations later on? Finally, to what extent does
institutional reform alter the career, particularly among those who had established a history
of mental health treatment prior to institutional reforms?
A Life Course Perspective on Illness Careers
One of Goffman’s insights was his theoretical attention to the illness career, conceptual-
ized as a multilevel, dynamic path linking individuals and institutions. Goffman’s interest was
in the subjective experience of the career and changes in self concept, but the career concept
has been applied to a wide array of topics including family careers (Moen 2003), unpaid work
careers (Pavalko and Woodbury 2000), delinquent careers (Sampson and Laub 1993; Shaw
1930) as well as institutional careers (Abbott 1993; Abbott and Hrycak 1990; Rosenfeld
1992). The precise meaning of career varies across these applications, but most share a con-
ceptualization of the career as a sequence of connected statuses and roles and a temporal pro-
cess that takes shape over time. A basic assumption is that the temporal, cumulative
dimensions of the illness career may help contextualize mental health treatment (Aneshensel
1999; Gove 2004; Strauss, Hafez, Lieberman, and Harding 1985). Because earlier and later
events in the career are connected, some factors may have a direct effect on the pace of hos-
pitalization early in the career, but have only an indirect effect later in the career. In other
words, factors that set the career in motion may differ from those that maintain it later on.
The temporal and cumulative nature of careers also suggests that experiences earlier in the
career will also shape later events.
A life course perspective shares interest in the career as a dynamic pathway through a
series of roles, but also emphasizes the importance of a changing historical context for shap-
ing careers (Elder, Johnson, and Crosnoe 2003; Marshall and Mueller 2003). In the mental
health system, this transformation includes both a change in the structure of hospitals and in
1. This definition of illness career differs from Goffman’s focus on the subjective experience of the career and
changes in self concept. It also differs from the concept of “illness course” used in the psychiatric literature, which refers
to the biological unfolding of the disease (Ciompi 1980; Harding 1988).
the treatment system in which hospitals are imbedded. The first half of the twentieth century
marked a period of relative stability in the system of care for persons with mental disorders
and the state mental hospital served as the core location for the treatment of persons thought
to be mentally ill, as well as a wide range of other “deviants” who had nowhere else to go.
Beginning in the late 1950s, an initial deinstitutionalization of patients from the hospital
became a broad “transfer of care” (Brown 1985; Gronfein 1985; Mechanic 1989) that pro-
duced a complex set of organizations (e.g., acute and long-term hospitals, day treatment,
treatment teams, groups homes, nursing homes, community mental health centers, alterna-
tive care, and peer-operated systems), all engaged in the care of persons with mental disor-
ders. Persons entering the mental health system in the second half, and particularly the final
decades of the twentieth century, encountered a complex system that includes a maze of
organizations and programs. While we know that the illness careers of those who entered the
mental health system after this transformation look dramatically different than those who
were treated prior to this change, we do not have a clear picture of how the illness careers of
those already in the system were impacted by this change.
Factors Influencing the Length of Stay
A key assumption of career perspectives is the career contextualizes the experience of
mental illness. Factors influencing the pace of hospitalization at early points in the career
may differ from those affecting later admissions or returns. Our first step in the analysis will
be to examine this assumption by comparing the influence of various factors (severity of ill-
ness, social status, and community context) on the pace of movement in and out of the hos-
pital at different points in the illness career. We thus draw on, and hope to inform, a long
history of debate within the mental health literature about the influence of various character-
istics on illness outcomes. For example, within sociology there has been a long-standing
debate over the relative influence of illness characteristics (type and severity of the illness)
versus social location (race, class, gender) on the future prospects of persons treated for men-
tal disorders (e.g., Gove 1980; Scheff 1966). While our general expectations about how each
set of characteristics relate to the pace of movement in and out of the hospital are discussed
below, our primary interest is whether these characteristics influence the pace of hospitaliza-
tion in different ways at key points across the illness career.
Severity of Illness
There is little question that there is wide variation in the functioning of persons with
severe mental illness, and we would thus expect that those with more severe forms of illness,
particularly various subtypes of schizophrenia, poorer functioning at discharge, and earlier
age of onset would have longer hospitalizations and less time out between hospitalizations
than those with less debilitating forms of illness. While age of onset is one indicator of sever-
ity of illness, in this pre-nursing home era mental hospitals were also used to care for persons
with dementia, and in these cases, those who were older at first hospitalization may actually
have longer stays.
Sociological perspectives on mental illness have also demonstrated that one’s position in
the social structure shapes their treatment experiences. Scores of studies have documented
variation in rates of mental illness and by gender, race, and class, and differential diagnosis or
reaction to persons depending on their social status (Hollingshead and Redlich 1958; Link
and Phelan 2001; Loring and Powell 1988). Following the consistent finding that persons of
Mental Illness Careers
lower status are more likely to be hospitalized than those with higher status, we generally
expect similar findings in this study. There is some evidence, however, that because Vermont
was a rural state, working class men may have been discharged sooner if they were needed to
work as farm laborers (Kincheloe and Hunt 1989).
Finally, the communities in which people live are important for their health and health
care. Research ranging from Emile Durkheim’s (1951) early interest in social integration and
suicide to examinations of social networks, mental health, and pathways to care (Horwitz,
Tessler, Fisher, and Gamache 1992; Peek and Lin 1999; Pescosolido 1991, 1996) suggest that
persons who are more integrated in families and communities are generally in better health
and use health care differently than those who are isolated. In the case of severe mental ill-
ness, families and communities may serve both as sources of support and sources of rejection
and conflict (McGrew, Wright, and Pescosolido 1999). Persons living alone may be kept in
the hospital longer because of fears that they will have less family support after release, but
once discharged, they may remain out of the hospital for longer periods of time because they
have less monitoring by family. We would also expect that those from smaller communities
may be returned to the hospital more quickly because of the stigma associated with mental
hospitalization and the greater recognition of persons as former mental patients in small
Accumulation of Illness Experience across the Illness Career
A second assumption of life course and other perspectives interested in careers is that
careers are cumulative, and thus, that earlier events will influence later pathways. While
individual or community characteristics are likely to shape hospitalization early in the career,
earlier events in the career may shape later outcomes. This view of careers is consistent with
labeling theory. In its earliest form, labeling theory suggested that responses to deviant
behavior (especially whether a label is applied) play a major role in whether a person will
become a career deviant or whether that episode will be written off as a one-time anomaly
(Scheff 1966). Efforts to “treat” or “reform” the individual reinforce the label, thus producing
a downward spiral of increasing mental illness. While many aspects of labeling theory have
since been revised (Link et al. 1989), and the downward spiral of mental illness is by no
means a certainty (Gove 2004), the power of labels and stigma to shape the illness experience
remains significant (Link and Phelan 2001; Martin, Pescosolido, and Tuch 2000). The frame-
work of the illness career can further inform our understanding of labeling processes by
exploring the extent to which the length and pace of earlier hospitalizations influences the
length of later stays.
Institutional Change and the Illness Career
Both Goffman’s view of the career as two sided, shaped by both individual and institu-
tion, and the life course emphasis on careers unfolding in the context of historical change
suggest that institutional change in the mental health system should affect illness careers.
However, the ways in which illness careers are affected by institutional change, particularly
for persons with established illness careers when changes are implemented, remains largely
unexplored. In assessing the impact of institutional change on illness careers, we are particu-
larly interested in two questions. First, to what extent do the characteristics that influenced
the pace of movement in and out of the hospital in the pre-reform period continue to shape
the length of time persons spend in or out of the hospital during and after the reforms? In
prior work, Walter Gove (1980) suggested that medical criteria may have taken on increasing
importance after the adoption of new treatment methods and reforms of the mental health
system. From this perspective we should see greater distinction in the pace of hospitalization
after hospital reforms. However, it is unclear how these changes intersect with accumulated
experiences in the illness career. Our second question is whether hospitalization experiences
accumulated prior to institutional reform continue to impact post-reform careers, or whether
the pace of moves in and out of the hospital are rewritten with the implementation of major
Data and Methods
The Historical Context and Case: Vermont State Hospital
Dramatic changes in the mental health system provide an excellent case for investigating
the impact of the changing system as individual patients move in, out, and through mental
health treatment. We use data from a long-term longitudinal study of patients who were hos-
pitalized at Vermont State Hospital and were enrolled into a model rehabilitation program
initiated at that hospital in the late 1950s. In addition to providing a rare source of long-term
longitudinal data for patients who were hospitalized during the first wave of institutional
change, the model rehabilitation program provides a concrete example of institutional
change that is consistent with reforms implemented during the shift from hospital to commu-
nity treatment. Before describing these data, we first overview changes in the institutional
context at Vermont State Hospital.
Changes in the treatment of persons thought to have a mental illness in Vermont in the
mid-twentieth century preceded those beginning in many states during this era. In the first
half of the twentieth century, the mental hospital was the mental health system, and, like all
states, Vermont struggled to meet the needs of a growing and increasingly diverse population
housed in mental hospitals (Chittick, Brooks, Irons, and Deane 1961). An influx of federal
dollars for community mental health centers in the 1960s initiated a multifaceted mental
health system, of which the state asylum was only a part. The late 1960s and 1970s were
marked by massive downsizing of hospitals and, in many cases, the transinstitutionalization
of inhabitants to more specialized facilities such as nursing homes and board and care homes,
which were largely financed by federal rather than state dollars (Brown 1985; Gronfein
1985). While Vermont also deinstitutionalized and transinstitutionalized patients from the
hospital, unlike many states that made this move before setting up community services, Ver-
mont began experimenting with community programs relatively early and their programs
became models for other mental health systems attempting to put greater emphasis on com-
munity mental health.
The timeline in Figure 1 illustrates these historical developments in Vermont and traces
their development at three levels—individual, hospital, and the state mental health system.
At the system level, Vermont was a leader in mental health treatment, with relatively early
drug treatment, emphasis on community treatment, rehabilitation, and a coordinated mental
health system (DeSisto et al. 1995, 1998; Kincheloe and Hunt 1989). For instance, as early as
1939, Vermont established guidance clinics for children, and structures such as these placed
them in a better position to establish community mental health centers (DeSisto et al. 1995).
This dynamic also translates to changes at the organizational (hospital) level. Vermont
State Hospital faced a growing inpatient population through the early 1960s. The hospital
had begun developing innovative rehabilitation and treatment programs in the early 1950s,
including reducing overcrowding, improving the working conditions, and training of staff,
Mental Illness Careers
Figure 1 • Timeline of Individual, Hospital, and Mental Health System Events, Vermont 1930 to 1980
Source: Adapted from DeSisto et al. 1991
1955 – 1961
1968: 95 percent
state MH budget to
I-------I I—I I—I I---I I-I I-----I I-I
I—I I—I I-----I I-I
Vermont State Hospital
Mental Health System Change
1980: 29 percent
state MH budget to
1939: 7 child MH clinics
funded by state
1964: 9 MH clinics
1972: 222 board
1963: State given $520K
to construct CMHCs
1978: 558 board homes; 6
rehab (halfway) houses
1956–1958: 3 rehab
1948: Federal $ allotted
to states for CMH
1945: ECT treatment
and increasing the number of professional staff (Chittick et al. 1961). However, the shift from
custodial, asylum-based care to rehabilitative, hospital, and community care was clearly
marked at Vermont State Hospital by the introduction of a model rehabilitation program in
1955 (see shaded portion in Figure 1). Although designed by hospital staff, this program
embraced the ideals of the community mental health movement. Resources were provided
for staff to work intensively with patients in the hospital, but the program also provided
intensive support and rehabilitation for program members once they reentered the commu-
nity after discharge. The program was very successful, and ten years after the start of the
model program 70 percent of the former patients remained out of the hospital (Deane and
Brooks 1967; Harding et al. 1987a, 1987b).
As shown in Figure 1, these program changes are also reflected in the illness careers of
individuals entering the hospital during this era. The hospitalization histories of six persons
who were involved in the model program are illustrated at the bottom of Figure 1. While all
were involved in the model reform program, they encountered it at different points in their
illness careers. The dashes between brackets mark periods in the hospital and the blank areas
in each row indicate periods outside of the hospital. The hospitalization when the patient
entered the model program is indicated by bold type. The top three cases are persons who
had been hospitalized at least once prior to the model program, while the bottom three illus-
trate illness careers of those who were placed in the model program during their first hospi-
talization. For both groups, the patterns of hospital entry and exit differ prior to 1960 than
afterwards. Post-1960 hospitalizations are shorter and more sporadic. But, consistent with
concerns about the depopulation of hospitals creating a revolving door of entry and exit
(Segal, Baumohl, and Johnson 1977; Wegner 1990), when multiple hospitalizations occur,
they tend to occur in rapid succession.
The Vermont Longitudinal Study (VLS) of persons with severe mental illness began in
1955 when clinicians from Vermont State Hospital and Vermont’s Vocational Rehabilitation
Division established a new program to rehabilitate severely mentally ill patients (Chittick
et al. 1961; Harding et al. 1987a). Earlier reforms and treatments, including early trials with
phenothiazines, had reduced populations of the patients most responsive to treatment, leav-
ing the hospital with a smaller number of the sickest patients who had not initially responded
to the new drug therapies (Chittick et al. 1961). These sicker, less responsive patients were
the population from which the VLS was drawn.
Between January 1955 and December 1960, 269 patients from the back wards were
selected for participation in the rehabilitation program and thus became part of the VLS
(Chittick et al. 1961; Harding et al. 1987a). The rehabilitation program used group and drug
therapies and worked intensely and collaboratively with patients on social and vocational
skills for several years inside the hospital. Extensive community rehabilitation continued for
patients after discharge from the hospital. Efforts were made to blur boundaries between the
hospital and the community, for example, using former patients as discussion group leaders
and peer support and making efforts to include patients as active collaborators in their own
treatment (Chittick et al. 1961).
Patients were selected for the study based on hospital staff referrals. The two primary cri-
teria for referral were that the patient was “chronically mentally ill” and the staff thought the
patient had potential for improvement (Chittick et al. 1961). The study defined chronicity as
having been disabled by their illness for at least one year prior to their entry into the study
combined with poor responses to initial drug treatments. At the time of admission into
the study, subjects averaged 16 years of illness, 6 years of continuous hospitalization, and
10 years of total disability. They were between 16 and 65 years of age with an average age of
40, and had experienced a median of two hospitalizations. Some patients had experienced as
Mental Illness Careers
many as ten hospitalizations when they entered the rehabilitation program and generally had
little contact with friends and relatives outside the hospital (Chittick et al. 1961; Harding et al.
Courtenay Harding and her colleagues (1987a, 1987b) collected admission and discharge
dates for each episode from the first hospitalization to 1982 or death, including information
on the rare case of hospitalizations at other institutions. These hospital records serve as the
basis for our analyses. Of the original 269 patients in the study, 238 have complete informa-
tion on all variables and are used in our analyses.2
Analytic Method and Variables
We focus specifically on the hospitalization career in this study, defined by the pace of
movement between hospital and community, or more specifically, variation in the amount of
time spent in the hospital during each stay, the number of hospitalizations, and the amount
of time spent in and out of Vermont State Hospital. We are particularly interested in the char-
acteristics that predict the duration of key hospitalizations and the length of time patients
remain in the community between hospitalizations. We focus on two critical points in the ill-
ness career—the first and second hospital stays and the hospitalization when persons entered
the model program (hereafter referred to as the model or index hospitalization), which
implemented the major institutional changes underway at that time.
All of the VLS participants were placed in the model program, but it fell at different
points in their illness careers. We split the sample into two groups because the relevant
dependent variables differ for the two groups and because the comparison allows us to
explore how institutional change influences the career while taking into account where that
change falls in the career. The first group consists of 45 percent of the sample (n = 107) who
had an established illness career (i.e., those who had one or more hospital stays) prior to
being placed in model program. The second group was enrolled in the model program during
their first hospital stay. Our main focus is on the group of 107 who were hospitalized prior to
the start of the model program because they provide information on the illness career both
before and after the model program. For this group we analyze five measures reflecting the
pace of hospitalization: length of first admission, time in community between first and second
admission, length of second admission, length of model admission, and time in community
after discharge from the model admission.
We also analyze length of model admission and time in community after model dis-
charge for those who were placed in the model program during their first admission. While
analyses of this second group are not directly comparable to VLS subjects who had prior hos-
pitalizations, it does allow us to explore whether predictors of the pace of hospital admission
and discharge differ from those who had more established illness careers when they encoun-
tered institutional reform.
All dependent variables are based on the amount of time (in days) until an event (i.e.,
either hospital discharge or hospital readmission) occurs. Cox proportional hazards models
are used to estimate these effects (Allison 1995; Yamaguchi 1991) and are estimated with
SAS PHREG (SAS 1991). One advantage of proportional hazards models is that they allow us
to assume time dependence, but they do not require that we specify its form, which, in our
analyses, is likely to vary across events.3 In proportional hazards models, the dependent vari-
able is the log of the hazard function, which is the unobserved instantaneous risk that an
2. Demographic information reported in this study differs from that reported by Harding and colleagues (1987a,
1987b) because we retain persons diagnosed as having an organic disorder.
3. One assumption of proportional hazards models is that relationships between covariates do not change over
time. We considered the possibility that this might not be the case across diagnostic group. Models stratified by diagnos-
tic group yielded results that were consistent with analyses shown (Allison 1995; Yamaguchi 1991).
event will occur at time t, given that it has not already occurred (Allison 1995; Yamaguchi
1991). For ease of interpretation, tables show hazard ratios calculated as eb. Positive coeffi-
cients, or increased log of the hazard function, produce a hazard ratio greater than 1.0. Nega-
tive coefficients produce a hazard ratio less than 1.0.
These models allow for the possibility of censored events, which is salient for the final
outcome in our analyses—duration in the community after discharge from the model pro-
gram. Among the group who had hospitalizations before entering the model program, 35
percent were censored because they did not re-enter the hospital between discharge from the
model program and 1984 or death, while 39 percent of those who entered the model pro-
gram during their first hospitalization were censored.4
Descriptive statistics for the dependent and independent variables are shown in Table 1.
The first four columns show the variable descriptions, mean, and standard deviation for all
Table 1 • Descriptive Statistics of Indicators of the Pace of Hospitalization and Covariates, Vermont
Longitudinal Study (n = 238)
VariablesMeanS.D. Min. Max.
Mean, First is
(n = 107)
is Model a
(n = 131)
Length 1st stay
Length 1st to 2nd
Length 2nd stay (n = 51)
Total admits before model
Length model stay
Time to readmit after model
— 1.971.36 81.97
Age 1st admit
GAF 1st discharge
Status prior to 1st admission
Community characteristics prior to 1st admission
.11 .3201 .07 .16
Cohort of entry
Source: Vermont Longitudinal Study (Chittick et al. 1961; Harding et al. 1987a)
aBolded numbers indicate significant differences between the two groups, based on t-tests and chi-square statistics
(p < .05; two-tailed tests).
4. A small number of cases died after discharge from the model program but before readmission. Alternative models
specifying death as a competing event to rehospitalization were also estimated; results are similar.
Mental Illness Careers
variables. The final two columns divide the means according to the two separate samples
considered in our analyses. The first, and primary sample for our analyses, consists of persons
who were hospitalized at least once prior to their entry into the model program; the second
are those who entered the model program during their first hospitalization. The bolded num-
bers indicate significant differences between the two groups.
Descriptive statistics for the dependent variables are shown under the heading for illness
career because variables such as length of first hospitalization serve as dependent variables in
initial models but are indicators of the prior illness career in later models. The mean length of
first hospitalization for patients who were hospitalized prior to their entry into the model
program was 374 days and the mean length of time out of the hospital between first and sec-
ond hospitalization was 1,084 days. There was, however, wide variation in the length of first
stay (8 to 4,019 days), and even more so in the amount of time persons remained in the com-
munity between their first and second hospitalization (1 to 7,525 days) and the length of sec-
ond hospitalizations (30 to 4,901 days). Patients averaged slightly less than 2 hospitalizations
prior to starting the model program, but some patients had as many as eight prior hospitaliza-
tions. On average, the hospitalization when subjects were placed in the model program
tended to be much longer than the first hospitalization (2,513 days versus 374 days), partially
reflecting the intensive rehabilitation efforts of the model program. The mean length of the
model hospitalization, the time out of the hospital after model hospitalization discharge, and
percent of those readmitted after model hospitalization discharge did not differ for the two
groups. Thus, for the data where we have comparable information, the pace of transitions
between hospital and community did not differ for these two groups.
Four additional groups of independent variables correspond to categories of likely predic-
tors influencing the illness career. Measures of illness characteristics include diagnostic category,
global assessment of functioning at the time of first discharge, and age at first admission.5
While those who have later onset of mental illness are often thought to have less severe
forms of illness, we note that some of those in the VLS were as old as 60, which may reflect
dementia or persons with few alternatives for care in this pre-nursing home era.
Diagnostic categories are based on DSM-III definitions, created from a blind review of
each patient’s chart stripped of all prior diagnoses (see Harding et al. 1987b for full details).
Just over half (53 percent) of the sample had symptoms characteristic of schizophrenia or
atypical psychoses and 19 percent had symptoms consistent with serious affective disorders.
The 16 percent of patients categorized as “other” include developmental disorders and alco-
holism. Chi-square statistics indicate that the distribution of disorders does differ between the
two samples, with those admitted into the model program during their first hospitalization
more likely to have psychotic symptoms or developmental disorders. As further evidence that
those in the VLS were more severely ill than the general hospital population, during this time
period less than half of patients in the hospital as a whole had psychotic disorders, compared
to 53 percent with psychotic disorders in the VLS (Kincheloe and Hunt 1989).
Because levels of functioning vary widely within and between these broad diagnostic
categories, we also include several indicators of the patient’s level of functioning. In particu-
lar, we use scores based on the Global Assessment of Functioning scale (GAF), an often-used
measure of functioning in the mental health literature designed to estimate the overall level
of psychological and social functioning (Endicott et al. 1976).6 This scale can range from 0 to
100, with higher scores indicating better functioning. We use the GAF score derived from
5. There is increasing evidence that psychiatric illnesses, and especially schizophrenia, change over time and have
their own “illness course” (Ciompi 1980; Harding 1988). Ideally we would be able to treat illness characteristics as time-
dependent variables, but doing so is beyond the scope of this study.
6. We expect that these assessments were used, in part, to support clinician argument that the patient was ready
for discharge. Thus, we view the scores as more valuable as indicators of relative functioning within the sample rather
than an absolute level of functioning.
chart review one week prior to discharge from the first hospitalization.7 GAF at the time of
first discharge was lower for those who were hospitalized prior to the model program than
for those who were first hospitalized during the model program. However, a comparison of
GAF at model program discharge for the two groups (analyses not shown) indicated that the
GAF scores were similar between the two groups at model program discharge. Age at first
admission is also included as a proxy for age of illness onset, which is correlated with illness
severity (Pfeiffer, O'Malley, and Shott 1996).8
Status characteristics include gender and parental social class. In the VLS, social class was
measured by the Hollingshead scale (Hollingshead and Redlich 1958). We collapse this scale
to contrast patients from working class backgrounds to those from middle and upper class
backgrounds because preliminary analyses indicated that this dichotomy reflected the most
significant relationships between social class and hospitalization. Social class is based on
parental social class prior to first admission to avoid the influence of hospitalization on the
individual’s subsequent social class. We do not control for race/ethnicity because there is no
racial variation, reflecting the racial homogeneity of Vermont at that time. The sample
includes slightly more women than men and 41 percent of the sample came from working
class backgrounds. Chi-square tests indicate that gender and class distributions of the two
subsamples are not significantly different.
The community is indicated by size of community (rural versus small towns and urban
areas) where the subject lived prior to first admission and whether or not the subject lived
alone prior to first admission.9 While we cannot assume that persons necessarily returned to
the same family and community arrangements after discharge, these provide broad compari-
sons of the types of community resources from which patients might draw. Only a small per-
centage of the sample (7 percent) came from households in heavily rural areas and only
11 percent of the sample lived alone prior to first hospitalization. Chi-square tests indicate
that those who entered the model program during their first hospitalization were more likely
to have lived alone prior to first admission but the two samples did not differ in the size of
community from which they came.
Information on the prior illness career is applicable for all models except those estimating
the length of first hospitalization. In our main analyses, length of the first hospitalization is
included in subsequent models. The total number of prior admissions is included as an inde-
pendent variable in models estimating the length of the model hospitalization.10
Our design requires that we consider the overlapping influences of aging (or illness
career duration), cohort of entry, and period changes. One advantage of the VLS is that we
have information to directly measure many of these temporal processes (Glenn 2003). Rather
than defining the period of institutional change by the same time period for everyone, we are
able to incorporate information on the exact time of entry and exit from the model program.
Illness career characteristics allow us to consider the wide variation in patterns of hospitaliza-
tion after first entry rather than relying on a single measure of duration from first entry to
model program. However, cohort of entry is strongly associated with several measures of the
illness career, and we thus include cohort of entry as a control.
7. Additional models were estimated for those hospitalized prior to the model program, including a second GAF
measure, reflecting functioning prior to discharge from the model program. It was not related to subsequent readmis-
sion but was highly correlated with GAF from first discharge, and thus is not included in the models shown.
8. Initial models also included a measure of reason for first admission (danger to self or others, threat, odd behav-
ior, etc.) based on review of patient charts, but these categories were not associated with the timing of any of the transi-
tions and thus are not included in models shown.
9. The vast majority of those not living alone prior to their first admission were living in a nuclear family struc-
ture, thus not allowing meaningful comparisons across the broader range of potential living arrangements.
10. We also constructed several other measures of the illness career including the total length of time the person
had been hospitalized and the average time admitted (combining information on total time and number of admissions).
These alternative measures were not significant in any models and thus are not shown.
Mental Illness Careers
Cohort of entry is measured by the time period of first hospitalization. The periods corre-
spond to the broad eras of treatment philosophy at the hospital, contrasting the era of custo-
dial care (1922 to 1944) to the early reform period (1945 to 1954) and the period when the
model program was established (1955 to 1961).11 Approximately one-fifth of the sample was
hospitalized for the first time before 1945, and another 40 percent were first hospitalized
between 1945 and 1954. Although cohort of entry is strongly correlated with whether or not
the person was initiated into the model program during their first hospitalization, these two
sources of information are not entirely redundant. Twelve percent of those who were initi-
ated into the model program during their first hospitalization entered the hospital prior to the
Second World War and 14 percent of those who had prior hospitalizations started their hospi-
talization in 1955 or later.
Factors Influencing the Hospitalization Careers in the Pre-Reform Era
Models in Table 2 assess the pace of hospitalization across the illness career for patients
hospitalized before the initiation of the model program. The first three dependent variables
(length of first hospitalization, time out of the hospital between first and second hospitaliza-
tion, and length of second hospitalization) represent the illness career prior to major institu-
tional reform. The length of the hospitalization during the model program represents the
point in the illness career when institutional change was implemented. Time out of the hos-
pital after discharge from the model program reflects the illness career in the immediate post-
In the pre-reform era, the early stages of the hospitalization career are not clearly differ-
entiated by diagnostic category or level of functioning. The one illness characteristic that does
explain variation in the length of the first hospitalization is age when first admitted to the
hospital. Persons who were older at the time of their first admission have a lower hazard of
discharge and thus are hospitalized longer. However, the direction of that effect is contrary to
a medical interpretation of the influence of age at onset, which would predict that a later age
of onset would be associated with a milder form of the disease. This effect may reflect the use
of the state mental hospital for custodial care of disabled older persons in this pre-nursing
home era. Given that patients were as old as 60 at first admission and the long average length
of first admission, some patients were potentially quite elderly when being considered for dis-
charge. Overall, the relatively weak influence of illness characteristics in this pre-reform
period indicates that the type and severity of the illness had little impact on the pace of moves
in and out of the hospital at this time.
Status characteristics are influential in these pre-reform careers, as both gender and
social class are associated with variation in the length of first hospitalization. As expected,
men are released from the hospital sooner than women, with men having a 69 percent
greater hazard of discharge. However, contrary to our expectations, we find that persons
from working class backgrounds were released sooner than middle and upper class patients.
One likely reason for this pattern is that, in this heavily rural state, working class men were
sometimes discharged to provide seasonal farm labor. It is important to note that these sta-
tuses only directly influence the length of the first hospitalization but not the subsequent
pace of movement in and out of the hospital. Finally, community characteristics prior to first
admission (e.g., living alone and community size) are unrelated to hazard of discharge or
time between first and second hospitalizations in this era.
11. This periodization is based on historical accounts provided by Chittick et al. (1961) and Kincheloe and Hunt
One of the strongest influences on the pace of movement in and out of the hospital prior
to reform is the time period when the first hospitalization occurred. This finding is consistent
with historical changes in the philosophy of the institution in different eras; it also reflects
and partially controls for effects of the sample design. However, sample design should play
less of a role in contrasts between the 1922–44 and the 1945–54 cohort and should have no
effect on contrasts in length of model program hospitalization. Consistent with the historical
record, compared to the 1922–44 cohort, the 1945–54 cohort had shorter periods between
first and second hospitalization and shorter hospitalizations during the model program.
The variables at the bottom of the table represent the cumulative influence of the prior
illness career on the later pace of hospitalization. Consistent with labeling theory and with
Goffman’s (1961) concerns about the institutionalization of patients, longer initial hospital-
izations in the pre-reform era are associated with a faster return for a second hospitalization.
Although the magnitude of influence for each additional day is minute, the wide variation in
length of the first stay suggests that this effect may be noteworthy when accumulated over
Table 2 • Hazards Ratios for the Pace of Hospitalization across the Illness Career, Patients Admitted
Prior to Model Program (n = 107)
(n = 51)
Age 1st admit
GAF at 1st discharge
Status characteristics prior to 1st admission
Community characteristics prior to 1st admission
Cohort of entry1
Length 1st stay(days)
Total admits before model
Length model stay (days)
−2 log likelihood (without covariates)
−2 log likelihood (with covariates)
0 00 0
Source: Vermont Longitudinal Study (Chittick et al. 1961; Harding et al. 1987a)
*p < .05 **p < .01 ***p < .001 *(two-tailed tests)
aThe comparison group for diagnosis is psychotic disorders, for gender is female, social class is upper and middle class,
for farm is urban areas and small towns, for living alone is living with family or roommates, and for cohort of entry is
Mental Illness Careers
many days. We also assume that estimates of effects of accumulation in the illness career are
conservative in this sample because we are assessing variation among the sickest patients.
The power of the first hospitalization for the subsequent pace of movement in and out of
the hospital continues through subsequent hospitalizations. Model 3 shows estimates of fac-
tors influencing the length of second hospitalization for the subset of patients who had two or
more hospitalizations prior to entering the model program. In Model 3, the length of first
hospitalization is the only characteristic among all those considered in previous models to be
associated with hazard of discharge from the second hospitalization.12
Factors Influencing Hospitalization Careers during and After Reform
The last two columns in Table 2 show variation in the pace of hospitalization for these
same patients during and after program reforms. Characteristics predicting the hazard of
readmission after discharge from the model program are substantially different than those
associated with readmission after the first hospitalization (the fifth versus second columns),
thus suggesting that the institutional reforms reconfigured the factors associated with the
pace of movement in and out of the hospital. Consistent with increasing institutional differ-
entiation in care, and particularly the institutional separation of the treatment of develop-
mental disorders from mental illness in the post-reform era, we find differentiation in the
hazard of readmission for those with developmental and other types of disorders. Most
importantly, compared to those diagnosed with one of the subtypes of schizophrenia, those
with developmental or other disorders have a lower hazard of readmission after discharge,
and thus remain out of the hospital longer.
Persons with higher levels of functioning (GAF) after their first discharge also had a
reduced hazard of readmission after participating in the model program. For each one-point
increase in functioning on the GAF scale, the hazard of readmission is reduced by 2 percent.
Finally, after reform, community and living situation prior to first admission become salient
for differentiating subsequent readmission. Both influences suggest that persons who came
from more isolated living environments had a dramatically increased risk of readmission after
reforms were implemented.
These last two models also suggest that institutional reforms were powerful enough to
reshape established illness careers. The prior illness career is associated with the pace of
movement in and out of the hospital up to and including the model program. However, indi-
cators of the prior illness career are not associated with readmission after the model program.
Likewise, cohort of entry, which was related to the pace of hospital admission and discharge
through the model program, is not associated with hospital readmission after the model pro-
gram. Although difficult to fully disentangle the changes associated with institutional change
from those occurring as the illness career progresses, the overall pattern of results suggests a
process of accumulation or progression in the illness career prior to the reforms, but also the
potential for institutional reform to erase these accumulating influences.
Factors Influencing the Illness Careers of Persons First Hospitalized
during the Model Program
Models in Table 3 show the characteristics influencing the pace of discharge and read-
mission for patients who entered the model program during their first hospitalization and
thus allow us to explore the impact of institutional reform occurring early in the illness
12. Only 51 of the 107 persons in this subsample had a second hospitalization before entering the model program.
In additional analyses we reestimated model 1 with this smaller sample and results were generally consistent with those
found for the larger sample.
career.13 While not directly comparable to results in the prior table because of different model
specifications and sample differences, broad comparisons to the prior table are suggestive. We
are particularly interested in whether illness characteristics differentiate among these patients
in the reform and post-reform period even though this reform falls earlier in their illness
career. We find that they are, and indeed, among patients who entered the model program
during their first hospitalization, diagnostic categories are the only statistically significant fac-
tors differentiating length of hospitalization during the model program. Patients with symp-
toms consistent with affective disorders or developmental disorders have a much greater
hazard of discharge from the model program, and thus are discharged earlier than those with
schizophrenia or atypical psychoses. Once discharged, these symptoms appear to be sub-
sumed by the influence of level of functioning at discharge. As with the prior models, status
characteristics and community are also associated with the pace of discharge and readmission
for this group, although the direction of these relationships are less consistent than was found
for those who started their illness career prior to the model program. We also find that the
length of hospitalization during the model program has little impact on the pace of readmis-
sion after the model program. As with the prior group, post reform readmission is not well
predicted by this set of variables.
Table 3 • Hazards Ratios for the Pace of Hospitalization across the Illness Career, Patients
Admitted during Model Program (n = 131)
Readmission after Discharge
from Model Program
Age 1st admit
GAF at discharge
Status characteristics prior to 1st admission
Working class a
Community characteristics prior to 1st admission
Length model stay (days) 1.00
−2 log likelihood (without covariates)
−2 log likelihood (with covariates)
Source: Vermont Longitudinal Study (Chittick et al. 1961; Harding et al. 1987a)
*p < .05 **p < .01 ***p < .001 *(two-tailed tests)
aThe comparison group for diagnosis is psychotic disorders, for gender is female, social class is upper and
middle class, for farm is urban areas and small towns and for living alone is living with family or room-
13. Cohort of entry is not included in these models because it is redundant with length of index hospitalization for
Mental Illness Careers
Discussion and Conclusions
Over the past half century, sociologists have examined the many ways that social
factors—ranging from social statuses such as race, class, and gender to social interaction and
stigma—shape the experiences of persons with severe and persistent mental illness. Our anal-
yses seek to illuminate how these influences are contextualized within the individual’s illness
career and how attention to the illness career may help us better understand these influ-
ences. We find that, among persons hospitalized for severe mental illness prior to deinstitu-
tionalization and the move to community mental health, social status (gender and
socioeconomic status) was more strongly associated with the length of the initial hospitaliza-
tion than illness characteristics. However, the pace of subsequent moves in and out of the
hospital, including the time spent in the community after the first discharge and the length of
the second hospitalization, were increasingly shaped by the earlier career. In other words, the
flow of hospitalization careers in this era were very consistent with concerns raised by label-
ing theorists of that era—that the severity of the illness had less to do with how long or how
often persons were hospitalized than one’s social status and that once a person was hospital-
ized, the illness career developed a momentum of it’s own (Scheff 1966).
However, the other part of this story is that these illness careers are imbedded in histori-
cally specific institutional structures, and a change in the treatment program was successful
rewriting these careers. While earlier research had demonstrated that this program was effec-
tive in allowing these former patients to lead far more integrated lives than was thought pos-
sible, there continued to be wide variation in functioning among this group (Harding et al.
1987b). By focusing on the illness career before and after this institutional change, we can
explore the extent of this success in more detail. We found that the reforms did not simply
raise the independence of all former patients, but actually leveled the influence of the prior
career history and introduced a new set of characteristics that differentiated the pace of sub-
Moreover, the characteristics most important in shaping the post-reform pace of hospi-
talization are consistent with the availability of new treatments that made the differentiation
of mental illnesses by diagnosis and level of functioning more meaningful. The appearance of
these influences after the adoption of the model program is particularly noteworthy because
they were based on characteristics in place prior to the start of the first hospitalization. While
mental health researchers have vigorously debated the relative influence of social status,
severity, and type of illness and labeling processes (see, for example Gove 1980; Scheff 1966),
our findings support elements of each of these perspectives, but suggest that the specific
influence of any of these characteristics depends on where person is in their own illness
career as well as where that career is located in historical time.
These findings are important for rounding out the historical record of deinstitutionaliza-
tion, but we feel that our efforts to assess the unfolding of the illness career in the context of
changing institutions also have relevance to today’s mental health system. Contemporary
mental illness careers and institutional structures look vastly different than those of even a
few decades ago, but our findings suggest that evaluations of the success of new programs
may benefit from attention to where that program falls in the illness career. In addition to
evaluating whether a program improved the average level of functioning, program success
might also be indicated if a program interrupted or altered the prior flow of the illness
career. Attention to the illness career could also assess whether model programs are more
successful when implemented at some points of the illness career. Indeed, while beyond the
scope of this article, future research might benefit from exploring whether the success of
model programs varies depending on timing of where that program is implemented in the
Our attention to illness careers comes during a time of resurgence in interest in illness
careers, and more generally in the person-centered longitudinal view offered by a career
perspective (Aneshensel 1999; Gove 2004; Pavalko 1997). Many studies of careers take a
holistic view of the career, comparing broad differences across a set of career typologies. Our
analyses take a stochastic approach to each transition in the career, allowing attention to
turning points and accumulation of effects at specific points in the career. This approach also
allows closer attention to institutional change as it intersects with career transitions.
Although the VLS offers a unique opportunity to assess variation in how hospitalization
careers unfold over a long period of time, there are several limitations in what these data can
tell us about illness careers. First, it is important to note that the VLS sample is homogenous
and not representative of persons with mental illness, even during that historical era. In addi-
tion to the uniqueness of Vermont as a small, relatively homogenous and rural state, persons
selected for the VLS were purposefully different than the broader population of persons hos-
pitalized at that time. The VLS selected the sickest patients—those who had not responded to
the first wave of new antipsychotic medications and who had been delegated to the back
wards of the hospital. This sample homogeneity means that our findings are very conserva-
tive and identify factors that stand out as differentiating characteristics among a homogenous
group. For example, although we find that social status and accumulation are significant in
differentiating among persons in the VLS, we suspect that these factors would be more
important in a sample where there was broader variation. Likewise, illness and community
characteristics, which had only limited effects on the illness career in this sample, may have
much stronger influences if assessed in a more heterogeneous state or a population of
patients with a wider range of illness experiences. Our focus on the hospitalization careers of
the sickest group of patients also is certain to provide a view of mental illness that is substan-
tially different than if we assessed careers of persons with less disabling illnesses (Gove 2004).
This study is also limited in its ability to fully differentiate the illness career from histori-
cal change. For example, we interpret the statistical significance of the GAF in predicting hos-
pital readmissions in the post-reform period as indicating that illness characteristics become a
better predictor of readmission after program reforms. An alternative explanation is that ill-
ness characteristics are relatively unimportant in the early stages of the career but become
more influential later in the illness career. While it is difficult to imagine such a scenario,
there is nothing in our data that will allow us to empirically distinguish between these two
scenarios. Our analysis of a second cohort of patients who entered the model reform program
during their first hospitalization does shed some light on this issue and supports our initial
interpretation, but these two cohorts are not directly comparable and still do not allow us to
fully distinguish between these different interpretations.
Finally, our focus on hospitalization careers, and specifically the pace of movement
between hospital and community, provides information on just one dimension of the illness
career. Although the concrete nature of this measure is a benefit for assessing patterns of
change during an era when definitions of mental illness changed drastically, changes in self-
concept, symptomatology, or ability to interact with others are not well reflected in move-
ment in and out of the hospital. More broadly, the illness career is just one of the many
careers experienced by the lives of these former patients. Illness careers intersect, and in
many cases compete with family, work, and community careers, and a focus on only one of
these dimensions almost certainly understates the complexity people’s lives (Moen 2003).
While these limitations suggest caution when interpreting these findings, they do not
diminish the potential value of the illness career for helping us better understand this impor-
tant historical era, for contextualizing the influence of different factors on the pace of hospi-
talization, and for exploring the interplay between individuals and institutions. Whether the
focus is on health care, education, prisons, or employment, social scientists are often inter-
ested in how single institutions affect individuals at a given point in time. While a valuable
first-step, cross-sectional designs can overstate the stability of institutions and relationships
and even longitudinal research may be limited if it only focuses on change either at the
individual or institutional level, without assessing the intersection between the two. The
Mental Illness Careers
dual-sidedness of the career offers an opportunity to study the intersection of individual and
society as they unfold over time.
Abbott, Andrew. 1993. “The Sociology of Work and Occupations.” American Review of Sociology 19:187–209.
Abbott, Andrew and Alexandra Hrycak. 1990. “Measuring Resemblence in Sequence Data: An Optimal
Matching Analysis of Musicians' Careers.” American Journal of Sociology 96:144–85.
Allison, Paul D. 1995. Survival Analysis Using the SAS System: A Practical Guide. Cary, NC: SAS Institute, Inc.
Aneshensel, Carol S. 1999. “Mental Illness as a Career: Sociological Perspectives.” Pp. 585–603 in Hand-
book of the Sociology of Mental Health, edited by C. S. Aneshensel and J. C. Phelan. New York: Kluwer
Brown, Phil. 1985. The Transfer of Care: Psychiatric Deinstitutionalization and Its Aftermath. New York: Rout-
Chittick, Rupert A., George W. Brooks, Francis S. Irons, and William N. Deane. 1961. The Vermont Story:
Rehabilitation of Chronic Schizophrenic Patients. Burlington, VT: Queen City Printers, Inc.
Ciompi, Luc. 1980. “Catamnestic Long-Term Study of the Course of Life and Aging Schizophrenics.”
Schizophrenia Bulletin 6: 606–18.
Deane, William N. and George W. Brooks. 1967. Five Year Follow-Up of Chronic Hospitalized Patients. Water-
bury, VT: Vermont State Hospital.
DeSisto, Michael J., Courtenay M. Harding, Millard A. Howard, and George W. Brooks. 1991. Perspectives
on Rural Mental Health: A Comparison of Mental Health System Policy and Program Development in Maine
and Vermont. Augusta, ME: Kennebec Press.
DeSisto, Michael J., Courtenay M. Harding, Rodney V. McCormick, Takamaru Ashikaga, and George W.
Brooks. 1995. Perspectives on Rural Mental Health: A Comparison of Mental Health System Policy and Pro-
gram Development in Maine and Vermont. Augusta, ME: Kennebec Press.
——. 1998. “The Maine and Vermont Three Decade Studies of Serious Mental Illness: Longitudinal
Course Comparisons.” Pp. 331–48 in Historical and Geographical Influences on Psychopathology, edited
by P. Cohen, C. Slomkowski, and L. N. Robins: Lawrence Erlbaum.
Durkheim, Emile. 1951. Suicide. New York: Free Press.
Elder, Glen H., Jr., Monica Kirkpatrick Johnson, and Robert Crosnoe. 2003. “The Emergence and Devel-
opment of the Life Course.” Pp. 3–19 in Handbook of the Life Course, edited by J. T. Mortimer and M.
J. Shanahan. New York: Plenum.
Endicott, J., R. L. Spitzer, J. L. Fleiss, and J. Cohen. 1976. “The Global Assessment Scale: A Procedure for
Measuring Overall Severity of Psychiatric Disburbance.” General Archives of Psychiatry 33:766–71.
Glenn, Norval D. 2003. “Distinguishing Age, Period, and Cohort Effects.” Pp. 465–76 in Handbook of the Life
Course, edited by J. Mortimer and M. Shanahan. New York: Kluger Academic/Plenum Publishers.
Goffman, Erving. 1961. Asylums. Garden City, NJ: Anchor Books.
Gove, Walter R. 1980. “Labeling and Metal Illness: A Critique.” Pp. 53–109 in The Labeling Deviance: Eval-
uating a Perspective, 2nd edition, edited by Walter Gove. Beverly Hills, CA: Sage Publications.
——. 2004. “The Career of the Mentally Ill: An Integration of Psychiatric, Labeling/Social Construction,
and Lay Perspectives.” Journal of Health and Social Behavior 45:357–75.
Grob, G. 1994. The Mad among Us: A History of the Care of America's Mentally Ill. Cambridge, MA: Harvard
Gronfein, William. 1985. “Incentives and Intentions in Mental Health Policy: A Comparison of the Med-
icaid and Community Mental Health Programs.” Journal of Health and Social Behavior 26:192–206.
Harding, Courtenay M. 1988. “Course Types in Schizophrenia: An Analysis of European and American
Studies.” Schizophrenia Bulletin 14:633–43.
Harding, Courtenay M., George W. Brooks, Takamaru Ashikaga, John S. Strauss, and A. Breier. 1987a.
“The Vermont Longitudinal Study of Persons with Severe Mental Illness, I: Methodology, Study
Sample, and Overall Status 32 Years Later.” American Journal of Psychiatry 144:718–26.
——. 1987b. “The Vermont Longitudinal Study of Persons With Severe Mental Illness, II: Long-Term
Outcome of Subjects Who Retrospectively Met DSM-III Criteria for Schizophrenia.” American Jour-
nal of Psychiatry 144:727–735.
Hollingshead, August B. and Fredrick C. Redlich. 1958. Social Class and Mental Illness: A Community Study.
New York: John Wiley.
522 Download full-text
Horwitz, Allan V., Richard C. Tessler, Gene A. Fisher, and Gail M. Gamache. 1992. “The Role of Adult
Siblings in Providing Social Support to the Severely Metally Ill.” Journal of Marriage and the Family
Kincheloe, Marcia R. and Herbert G. Hunt. 1989. Empty Beds: A History of Vermont State Hospital. Barre, VT:
Link, Bruce G., Francis T. Cullen, Elmer Struening, Patrick E. Shrout, and Bruce P. Dohrenwend. 1989.
“A Modified Labeling Theory Approach to Mental Disorders: An Empirical Assessment.” American
Sociological Review 54:400–23.
Link, Bruce G. and Jo C. Phelan. 2001. “Conceptualizing Stigma.” Annual Review of Sociology 27:363–85.
Loring, Marti and Brian Powell. 1988. “Gender, Race, and DSM-III: A Study of the Objectivity of Psychi-
atric Diagostic Behavior.” Journal of Health and Social Behavior 29:1–22.
Marshall, Victor W. and Margaret M. Mueller. 2003. “Theoretical Roots of the Life Course Perspective.”
Pp. 3–32 in Social Dynamics of the Life Course: Transitions, Institutions, and Interrelations, edited by W. R.
Heinz and V. W. Marshall. New York: Aldine De Gruyter.
Martin, Jack K., Bernice A. Pescosolido, and Steven A. Tuch. 2000. “Of Fear and Loathing: The Role of
'Disturbing Behavior,' Labels and Causal Attributions in Shaping Public Attitudes Toward Persons
with Mental Illness.” Journal of Health and Social Behavior 41:208–33.
McGrew, John H., Eric R. Wright, and Bernice A. Pescosolido. 1999. “Closing of a State Hospital: An Over-
view and Framework for a Case Study.” Journal of Behavioral Health Services and Research 26:236–45.
Mechanic, David. 1989. Mental Health and Social Policy. 3d ed. Englewood Cliffs, NJ: Prentice Hall.
Moen, Phyllis. 2003. It's About Time. Ithaca, NY: Cornell University Press.
Pavalko, Eliza K. 1997. “Beyond Trajectories: Multiple Concepts for Analyzing Long-Term Process.”
Pp. 129–47 in Studying Aging and Social Change: Conceptual and Methodological Issues, edited by M. A.
Hardy. Thousand Oaks, CA: Sage Publications.
Pavalko, Eliza K. and Shari Woodbury. 2000. “Social Roles as Process: Caregiving Careers and Women's
Health.” Journal of Health and Social Behavior 41:91–105.
Peek, M. Kristin and Nan Lin. 1999. “Age Differences in the Effects of Network Composition on Psycho-
logical Distress.” Social Science and Medicine 49:621–36.
Pescosolido, Bernice A. 1991. “Illness Careers ad Network Ties: A Conceptual Model of Utilization and
Compliance.” Advances in Medical Sociology 2:161–84.
——. 1996. “Bringing the “Community” Into Utilization Models: How Social Networks Link Individuals
to Changing Systems of Care.” Research in the Sociology of Health Care 13a:171–97.
Pescosolido, Bernice A. and Beth A. Rubin. 2000. “The Web of Group Affiliations Revisited: Social Life,
Postmodernism, and Sociology.” American Sociological Review 65:52–76.
Pfeiffer, Steven I., Dawn S. O'Malley, and Susan Shott. 1996. “Factors Associated with the Outcome of
Adults Treated in Psychiatric Hospitals: A Synthesis of Findings.” Psychiatric Services 47:263–69.
Rosenfeld, Rachel A. 1992. “Job Mobility and Career Processes.” Annual Review of Sociology 18:39–61.
Sampson, Robert J. and John H. Laub. 1993. Crime in the Making: Pathways and Turning Points Through Life.
Cambridge, MA: Harvard University Press.
SAS Institute. 1991. “SAS Technical Report P-217, SAS/STAT Software: The PHREG Procedure, Version
6.” Cary, NC: SAS Institute.
Scheff, Thomas J. 1966. Being Mentally Ill: A Sociological Theory. Chicago, IL: Aldine.
Segal, Steven P., Jim Baumohl, and Elsie Johnson. 1977. “Falling Through the Cracks: Mental Disorder
and Social Margin in a Young Vagrant Population.” Social Problems 24:385–400.
Shaw, Clifford R. 1930. The Jack-Roller: A Delinquent Boy's Own Story. Chicago: University of Chicago Press.
Shorter, Edward. 1997. A History of Psychiatry: From the Era of the Asylum to the Age of Prozac. New York:
John Wiley & Sons.
Strauss, John S., H. Hafez, P. Lieberman, and Courtenay M. Harding. 1985. “The Course of Psychiatric
Disorder, III: Longitudinal Principles.” American Journal of Psychiatry 142:289–96.
Wegner, Eldon L. 1990. “Deinstitutionalization and Community-Based Care for the Chronic Mentally
Ill.” Research in Community and Mental Health 6:295–323.
Yamaguchi, Kazuo. 1991. Event History Analysis. Newbury Park, CA: Sage Publications.