Administration and Policy in Mental Health
Vol. 25, No. 2, November 1997
THE ASSOCIATION BETWEEN PROGRAM
CHARACTERISTICS AND SERVICE DELIVERY IN
ASSERTIVE COMMUNITY TREATMENT
John H. McGrew and Gary R. Bond
ABSTRACT: The authors describe the relationship between service intensity and staffing,
organizational, client, and site characteristics in 19 programs based on the Thresholds
Bridge adaptation of the assertive community treatment (ACT) model. Pearson correlations
were examined between 14 program characteristics and intensity of ACT services. Several
staffing and organizational attributes were related to service intensity: larger team size,
shared caseloads, greater supervisor involvement in direct client services, and assignment of
primary responsibility for the client to the team. The potential facilitating relationship between several aspects of team operation and intensive services is discussed as are implications for local implementation of ACT.
A critical question facing program planners is how to organize services
for differing treatments and client groups to both maximize client benefits
and minimize costs. Largely in response to this issue, managed care and
related service provision schemes, e.g., capitation and utilization review,
attempt to maintain benefits while increasing service efficiency, thus, providing more care at less cost. This increased efficiency is often achieved
through changes in practice parameters and the organization of care. Although used extensively for physical illness, the mental health arena is one
of the last to begin to come under the umbrella of managed care (Feldman & Goldman, 1993). Mental illness, however, poses a special challenge
John McGrew, Ph.D., is Associate Professor of Psychology and Gary Bond, Ph.D., is Professor of Psychology and Director of the Doctoral Program in Clinical Rehabilitation Psychology, both at Indiana
University-Purdue University in Indianapolis (IUPUI).
This study was supported in part by NIMH Research Scientist Development Award K02MH00842 to
Gary Bond, and by NIMH Research Infrastructure Support Program Grant R24MH51669-02 supporting John McGrew. Computer services were provided by the IUPUI Computer Facilities.
Address for correspondence: John McGrew, Ph.D., Department of Psychology, Purdue University
School of Science, 402 N. Blackford, LD124, Indianapolis, IN 46202.
175
© 1997 Human Sciences Press, Inc.
176
Administration and Policy in Mental Health
for providers. Mental health care is often perceived as too expensive, too
open-ended, often ineffective, and with poorly specified outcomes. Of
those with mental illness, the most expensive and most difficult to treat are
persons with severe mental illness (SMI).
As with any medical population, program planners must face two issues
when choosing how to provide care for those with SMI: (1) how much
service do they need, and (2) how best to provide it. With respect to the
latter question, case management (CM), especially intensive CM, is often
advocated as one of the best methods for coordinating and providing care
to those with SMI. However, there are many different varieties of case management (Solomon, 1992). Moreover, case management models differ in
several important ways that affect the organization and planning of services, for example, how services are delivered (team CM vs. individual CM,
out-of-office vs. in-office visits), client-to-staff ratios, caseload sizes, and
staffing (primary use of medical specialities such as nursing and psychiatry
as opposed to use of bachelor level or master's level mental health workers) (Brekke & Test, 1992; Reinke & Greenley, 1986). In addition, different
case management approaches likely are not equivalently effective.
Assertive community treatment (ACT) is the best known and most carefully researched intensive case management model available to program
planners. Based on the Training in Community Living (TCL) program
developed by Stein and Test (1980) in Madison, WI, ACT has been widely
disseminated as a model of mental health services for persons with SMI.
Recent reviews generally have concluded that ACT is effective in reducing
hospital use for persons with SMI (Bond, McGrew, Fekete, 1995; Levine,
Toro, & Perkins, 1993; Olfson, 1990; Solomon, 1992; Taube, Morlock,
Burns, & Santos, 1990; Test, 1992). ACT uses a low client-to-staff ratio, and
a team approach to provide intensive, continuous support for revolving
door clients for as long as they need treatment. Although ACT appears to
be cost-effective in reducing expensive hospital use (Bond 1984; Bond,
Miller, Krumweid, & Ward, 1988; Bond et al., 1990; Hoult, Rosen, & Reynolds, 1984; Knapp et al., 1994; Weisbrod, Test, & Stein, 1980), a concern of
program planners is that features of the ACT model may be either
unnecessary or overly expensive to achieving adequate client outcomes
(Goldberg, 1991).
For example, some observers have argued that individual caseloads may
be more efficient than the team approach used in ACT, saving time required for intra-team communication. The additional time required for
meetings and other intra-team communication, by taking away from available clinical time, may reduce or reverse any potential increase in overall
services resulting from sharing caseloads (Rapp, 1993; Reinke & Greenley,
1986). The expectation in the ACT model of daily team meetings is a
concrete example of potential distraction from direct service. With this in
John H. McGrew and Gary R. Bond
177
mind, state administrators in Illinois recently mandated that nine new
ACT teams starting up in the state meet twice weekly, and not daily (Fekete, Smith, Salyers, Bond, Giffort, & Hampton 1996).
In contrast, experts in ACT have rated daily team meetings as critical for
intra-team communication (McGrew & Bond, 1995). Moreover, some research has suggested that team approaches, such as ACT, reduce turnover
and discontinuity of services (Bond et al., 1991) and forestall burnout
(Boyer & Bond, 1992). Burnout, in turn, contributes to decreased productivity (Maslach, 1978).
In fact, the usefulness of most features of ACT to service provision and
maximizing productivity is unknown. Research to date has focused mainly
on outcomes, not on process variables. Basic links within the service delivery system remain largely unexplored. For example, there is little
documentation of normal variations in ACT service delivery, nor is there
information about the relationship of service delivery to staffing and organizational features.
An active ACT team leader ensures high quality services.
When Intagliata and Baker (1983) reviewed factors affecting case management services for persons with SMI, they were able to identify only four
major empirical studies. Since their review, there have been few additional
studies of factors affecting services in case management, generally. Moreover, to our knowledge, there have been no published studies evaluating
quantitatively the factors affecting service delivery in ACT programs. Thus,
although there have been a number of studies examining process variables
since Intagliata and Baker (1983), none have the level of detail necessary
to assist managed care planning. For example, most have examined only
one ACT site (Brekke & Test, 1987, 1992; Reinke & Greenley, 1986) or have
limited their examination to the content and form of services and have not
explicitly investigated the relationship of services to organizational factors
(Bond, Miller, Krumweid, & Ward, 1988; Brekke & Test, 1987). Those who
have looked at program characteristics affecting service provision have
been limited to a case study methodology (Reinke & Greenley, 1986).
Part of the problem is that studies to date have examined a very limited
number of ACT sites. Thus, researchers have been unable to present quantitative information on the strength of any observed relationships with services. In fact, few studies have examined organizational variables, partly
because the unit of analysis is the program, requiring multiple sites before
quantitative analyses are possible. Those that have used multiple sites have
looked across a range of heterogeneous programs (Brekke & Test, 1992),
rather than within a specific program model. In the absence of precise
data on actual service utilization by targeted groups or for differing site
178
Administration and Policy in Mental Health
conditions, planners often rely on "conventional wisdom" (Harris & Bergman, 1990). However, as Harris and Bergman also demonstrated, conventional wisdom (e.g., clients with a history of chronicity should require
more services) is often incorrect.
Intagliata and Baker (1983) delineated four major factors affecting case
management services: case manager characteristics, client characteristics,
organizational factors/work environment, and broader service network.
These same factors, also can be applied to ACT programs. Each of these
factors can be viewed as directly affecting service provision (both quality
and quantity). Service quality and quantity, in turn, along with direct links
from client characteristics and characteristics of the broader service system
network, can be viewed as affecting client outcomes. What organizational
and staffing factors are optimal for the delivery of ACT services? How do
client and site characteristics affect service delivery?
Broadly speaking, the relationship of program elements to services can
be viewed as either (1) facilitating productivity, (2) detracting from productivity, or (3) unrelated to productivity. The current study examines
quantitatively the relationship between services and organizational and client factors within a sample of 19 programs based on the ACT model. The
study builds on the earlier work of McGrew, Bond, Dietzen, and Salyers
(1994), who examined variations in ACT program characteristics and their
effects on hospital use, within a sample of 18 programs. Seventeen specific
program characteristics measuring fidelity to ACT (largely limited to Intagliata and Baker's organizational features) were examined. Programs rated
higher in fidelity reported better outcomes, although individual elements
related differentially to outcomes. In the current study, to provide a more
fine-grained and ACT-specific classification, the organizational elements
were further subdivided into three factors: ACT team operation (e.g., daily
team meetings, degree of sharing of caseloads, coordinator involvement
with direct services, 24-hour availability, designation of team as primary
therapist), ACT team staffing (e.g., nurse on team, psychiatrist availability
to the team), and organizational resources (e.g., team size, caseload size,
client-to-staff ratio). Although we did not have specific hypotheses for most
elements, in general we expected elements indexing ACT team operation
to be most strongly related to delivery of services.
In addition to organizational elements, another important set of concerns for program planners are client characteristics, e.g., caseload mix. It
has been argued repeatedly in the literature that certain types of clients
will demand more frequent contact (e.g., substance users, persons with
history of chronicity). For example, in other treatment models, number
and severity of problems (Clark, Landis, & Fisher, 1990) and comorbidity
of schizophrenia and substance abuse have been associated with increased
service utilization (Bartels, Teague, Drake, Clark, Bush, & Noordsy, 1993);
John H. McGrew and Gary R. Bond
179
whereas chronicity has not been related to service use (Harris & Bergman,
1990). In the current study, we tested whether similar results obtain for
ACT services.
A final purpose of the current study was to document normative variations in service delivery across ACT programs. Sechrest, West, Phillips,
Redner, and Yeaton (1979) suggest the examination of variation across
existing programs as one method to begin to establish program standards.
For example, types of services (e.g., office visits) exhibiting large interprogram variations in frequency may be less central to the ACT model than
those with low variability. The current study, then, expands on the earlier
work of McGrew et al. (1994) to examine: (a) site variations in service
delivery, and (b) associations between organizational, staffing, client, and
site characteristics and service delivery patterns.
METHOD
Samples
Starting in 1986, Thresholds Bridge has served as the model for a series
of ACT replications in Indiana and elsewhere (Bond et al., 1988, 1990,
1991; McGrew, Bond, Dietzen, McKasson, & Miller, 1995; Witheridge &
Dincin, 1985). The sample for the current study consisted of 19 programs,
servicing 557 clients, either developed by Thresholds Bridge or developed
elsewhere and intended as a direct replication of the Bridge, and evaluated in a series of studies over a 10-year period by the second author and
his colleagues. The 19 programs include the 18 sites included in an earlier
study of fidelity to ACT (McGrew et al., 1994), plus one additional site.
Readers are referred to the earlier study for detailed information on study
samples, complete citations, and brief descriptions of the programs (McGrew et al., 1994). Across the 18 programs examined in the earlier study,
the mean reduction in days hospitalized during treatment compared to
baseline and the fidelity scores obtained suggested that these programs
overall were both reasonably faithful to the model and had a moderate
impact on reducing hospital use for persons with serious mental illness.
Client Characteristics
In all programs, clients met the state's criteria for serious mental illness
and additional criteria indicating high utilization of psychiatric hospitals
and/or other emergency services. Some programs also had specific additional criteria relating to homelessness, substance abuse, and/or failure to
use traditional aftercare. Details on clients characteristics are provided in
McGrew et al. (1994).
180
Administration and Policy in Mental Health
Program/Site Characteristics
Ten site characteristics identified by an expert sample as critical to ACT
(McGrew & Bond, 1995) were examined. Two program characteristics
used in the earlier study were not included (ACT located at separate site
from parent organization, and use of time-limited services), because they
were rated as unimportant to the model by the expert sample. In addition,
the six service-related variables used as predictors in the previous study
were included as dependent variables in the current study (see below).
The following site characteristics were measured as continuous variables:
(1) Client-to-staff ratio; (2) Team size; (3) Psychiatrist availability to team: the
percent of ideal available, estimated retrospectively by the clinical directors
of the ACT programs or by research or clinical consultants to the programs; (4) Caseload size; (5) Shared caseloads: the degree to which all team
members had contact with all clients on a regular basis, based on consultants' retrospective subjective judgments (0-100%), unless otherwise specified, the remaining variables were rated dichotomously as either present
(1) or absent (0); (6) Nurse on team: at least a 3/4-time nurse on the team;
(7) Team as primary therapist: the outreach team has primary clinical and
record-keeping (e.g., treatment plans) responsibility for the client; (8)
Daily team meetings; (9) Coordinator provides direct client service: the supervisor
devoted at least half-time to client contacts, either conjointly, during supervision of team members, or individually; (10) 24-hour availability: team provided 24-hour direct access to the team; where access to the team was triaged
through the CMHC emergency 24-hour on call service, scores of .5 were
assigned. Details on ratings for the individual variables are described in
McGrew et al. (1994).
Four additional site variables were also included describing client caseload characteristics (i.e., mean frequency of previous hospitalizations, percent schizophrenic, percent with poor attendance at aftercare, and percent homeless).
Service Delivery
Frequency (number of contacts), duration (cumulative number of hours
of contacts), location, and recipient of contacts with clients and collaterals
during the first 6 months of treatment were documented by ACT case
managers at each site through the use of service log sheets. Brekke (1987)
analyzed a similar log sheet used in the TCL program, on which the log
sheets used in these programs were based, and reported good interobserver agreement for frequency and categorization of contacts. The following types of contacts were coded: In vivo contacts: face-to-face contacts with
clients in their homes and in community settings (e.g., restaurants); Office
contacts: face-to-face contacts with clients in ACT team's office; Collateral
John H. McGrew and Gary R. Bond
181
contacts: face-to-face or phone contacts with persons or agencies on behalf
of client (e.g., family members, landlords, employers, social service agency
staff, etc.); Phone contacts: phone contacts with client; Hospital contacts: faceto-face contacts with client in hospital; and Total contacts: sum of all aforementioned contacts. Hospital contacts, however, were excluded from the
analyses. Hospital contacts were available for only 11 programs.
RESULTS
Site Comparisons
Table 1 displays the means, standard deviations and ranges for each of
the variables. In general, there was moderate variability across the sample
in program characteristics. For example, across sites the percent of caseload with a diagnosis of schizophrenia ranged between 30% and 89% and
the team size ranged between 2 and 7.5 full time equivalents. There was
substantial variability across the sites in service intensity. Using the ratio of
maximum to minimum values across the sites as an index, number of
phone (max/min = 43.5), collateral (max/min = 529.0), and office contacts
39.1 vs. 0) showed the greatest site variability. Most striking was the number of sites with zero or near-zero values for these services categories. Number of in vivo contacts (max/min = 6.31) and total contacts (max/min =
7.97) showed considerable, although relatively less, variability. Similarly,
using the coefficient of variation as an index of variability, office (1.18)
and phone contacts (1.20) were more than twice as variable, compared to
total (.58) and in vivo contacts (.58) (Table 1). Between-site variability in
hours of contacts was consistently smaller than was variability in frequency
of contacts.
Association Between Site Variables and Service Delivery
Description of Analyses. To reduce the dimensionality of the correlational
analyses, only number of total contacts was examined. In our previous
study (McGrew et al., 1994), frequency, but not hours, of contacts was
related to improved client outcomes. Also, experts in the model tended to
emphasize the importance of frequency over hours of contacts (McGrew &
Bond, 1995). In addition, total number of contacts was used because it is
the best single overall measure of number of contacts and it correlates
highly with the other types of contacts. An anticipated increase in the
number of in vivo visits for programs with a high percentage of homeless
clients was examined separately. For exploratory purposes, correlations
also are displayed for number of in vivo, office, collateral and phone contacts (Table 2). The pattern of correlations was similar using both hours
182
Administration and Policy in Mental Health
TABLE 1
Means, Standard Deviations, Coefficients of Variation, and
Ranges for Program Characteristics and Service Variables
Maximum
Coefficient
of Variation
11.5
.0
.1
72.6
39.1
52.9
.58
1.18
1.09
9.40
1.96
.8
.8
34.8
6.5
1.20
.71
59.59
26.98
6.52
3.54
34.47
12.79
7.81
2.85
18.2
10.8
.0
.0
145.1
60.6
31.1
12.2
.58
.47
1.20
.81
1.32
.93
.2
3.6
.70
1.62
1.08
.4
3.3
.67
39.87
30.00
17.66
15.83
18.7
15
82.4
80
.44
.53
10.42
4.16
.79
2.17
1.87
.42
7
2.0
0
15
7.5
1
.21
.45
.53
.62
.68
.35
.48
.10
0
.53
.56
.51
.27
.42
.18
88.63
Std Dev Minimum
Variable
Mean
In vivo visits
Office visits
Collateral visits
(n =18)
Phone contacts
Hospital visits
(n =11)
Total visits
27.67
10.15
12.30
16.10
12.01
13.45
7.86
2.75
Hours-in vivo visits
Hours-office visits
Hours-collateral
visits
Hours-phone contacts
Hours-hospital visits
(n = 10)
Hours-total visits
Caseload size
(n = 18)
Caseload ratio
Team size
Daily team meetings
Shared caseloads
Supervisor involvment
Primary therapist
Psychiatrist availability
Nurse on team
24-hour on call
Days hospitalizedprior (n = 18)
Percent schizophrenic (n = 18)
Percent homeless
(n = 15)
Percent attend aftercare (n = 18)
1.00
1
.56
.71
0
.10
1
.99
.96
.48
.51
.34
43.38
0
0
37.8
1
1.00
200.8
1.21
1.89
.49
54.39
16.32
30
89
.30
12.87
26.27
0
100
61.67
28.57
9
100
2.04
.46
Note: Service variables are the raw number reported for the first six months during treatment. N = 19
unless otherwise specified.
183
John H. McGrew and Gary R. Bond
and frequency of service contacts, although the size of the correlations
tended to be smaller when using hours of contacts.
Correlational Results. Three of the five variables indexing ACT team operation were related to service delivery (Table 2). Direct clinical contact with
clients by the clinical supervisor, greater sharing of caseloads, and the assignment of primary therapist role to the ACT team were positively related
to number of total contacts. Daily team meetings were not significantly
related to services.
One of three variables indexing organizational resources was related to
services (Table 2). Programs with larger teams reported greater frequency
of total contacts. Client-to-staff ratio and caseload size were unrelated to
services.
One of two ACT team staffing variables was related to service delivery.
TABLE 2
Correlation Between Frequency of Services and Program Characteristics
Type of Contact
Program variables
Caseload size
Client/staff ratio
Team size
Daily team meetings
Shared caseloads
Supervisor involvement
Primary therapist
Psychiatrist availability
Nurse on team
24-hour on call
Caseload variables
Days hospitalized
(n = 18)
Schizophrenic (%)
(n = 18)
Homeless (%)
(n = 15)
Attend aftercare (%)
(n = 18)
In vivo
Office
Collateral
Phone
Total
.32
-.31
.66*
.21
.60**
.37
.04
-.13
-.01
.39
.40
.41
.19
-.16*
.50*
.38
.53*
.29
.17
-.20
.47*
.27
.47*
.34
.29
-.30
.58*
.44
72***
.51*
.50*
.13
.02
.05
.36
.05
.17
.23
.65**
.47*
.11
.34
-.05
.02
.07
-.17
.05
.54*
.49*
.21
.41
-.07
-.41
.30
.56*
.00
.35
.37
-.05
.70**
-.29
.30
- .59**
-.00
-.53*
Note: N = 19 unless otherwise specified. *p<.05. **p<.01. ***p<.001.
.17
-.47
-.25
.35
-.59*
184
Administration and Policy in Mental Health
Greater psychiatrist availability to the team was related to a higher frequency of total visits. Programs with a nurse on the team did not record an
increased frequency of contacts.
One of the four variables indexing caseload characteristics was related to
total services (Table 2). Teams serving caseloads with a higher percentage
with poor records of attendance at aftercare reported a decreased frequency
of total contacts. Site variations in percent of caseload with a history of
multiple hospitalizations, in percent homeless, and in percent diagnosed
with schizophrenia did not predict total services. Finally, teams serving
caseloads with a higher percentage homeless showed a greater frequency
of in vivo contacts.
DISCUSSION
As mentioned earlier, Sechrest and colleagues (1979) suggested examining the variation in existing programs to establish standards of implementation. Within our sample of 19 Bridge-adaptations of ACT, there was substantial between-site variability in service delivery. The largest variability
was in number of office, collateral, and phone contacts. For example, at
some sites there was virtually no client contact in any of these service domains. These results suggest that frequency of services in these areas is
subject to considerable local adaptation. Alternatively, implementation
standards for these services areas may be underspecified. For example,
relatively little has been written concerning guidelines for frequency of
phone contacts. In addition, experts have rated office contacts as neither
important nor unimportant to ACT (McGrew & Bond, 1995). There was
substantial variability, but less so, in number of in vivo and total contacts.
These results suggest that expectations for frequency of services in these
areas is more standardized, at least according to normal site variation.
Several client and program characteristics were related to service delivery. The most theoretically interesting findings concerned the association
between measures of team operation and service delivery. As was the case
in our earlier study predicting reduced hospital use (McGrew et al., 1994),
the predictor of increased frequency of total services with the largest correlation coefficient value was the degree to which programs shared caseloads. Thus, shared caseloads is emerging as an important aspect of the
ACT team. Several factors that have been associated with team treatment
could partially explain this result. As already discussed, the decreased
burnout associated with ACT (Boyer & Bond, 1992) could help maintain
higher levels of team motivation and productivity. Also, as Reinke and
Greenley (1986) observed, ACT "staff are interchangeable" and "can share
duties and distribute workloads more easily than staff of other models."
Teams, for example, may be better able to adjust flexibly to temporary or
John H. McGrew and Gary R. Bond
185
permanent changes in personnel that could lead to episodes of missed or
reduced care. Future research should more precisely identify the specific
components of shared caseloads and assess the potential mechanisms influencing service delivery (e.g., decreased burnout). Global ratings, such
as those used in the current study, fail to identify potentially important
program differences in the implementation of shared caseloads. For example, proponents of the ACT model disagree whether shared caseloads imply equal responsibility for providing direct services to all clients or for planning treatment for all clients (McGrew & Bond, 1995).
The assignment of primary therapist role to the team was another aspect
of team operation related to increased frequency of total services. Teams
not assigned the role of primary therapist may be less able to and, therefore, less likely to, take on ultimate responsibility for the client, as prescribed by the model (McGrew & Bond, 1995). This diffusion of responsibility may result in decreased client services compared to programs that
are assigned the primary therapist role. Moreover, lack of primary responsibility may lead to perceptions of decreased autonomy in working with the
client. Graham (1980) noted that case managers with greater discretion
and autonomy saw their clients more frequently.
Mindless adherence to a model program may be neither necessary nor appropriate.
A third aspect of team operation related to increased services was the
presence on the team of a clinical director who provides direct client services. Team leadership may be an important understudied aspect of the
team. We speculate that the example of the leader may provide a standard
against which other team members can both emulate and measure their
performance. The clinically active team leader also may be better informed
and better able to "direct" care and ensure high intensity services. In addition, high quality and consistent supervision of case managers has been
associated with greater work motivation (Graham, 1980).
Somewhat surprisingly, a fourth aspect of team operation, daily team
meetings, was unrelated to services. As mentioned earlier, although daily
team meetings are an integral part of the ACT model (McGrew & Bond,
1995), some administrators and clinicians have argued that team meetings
are inefficient. It is therefore surprising that there was a positive (although
nonsignificant) correlation between daily team meetings and intensity of
service. We speculate that the shared decision making and decreased isolation Test (1979) identified as advantages important for maintaining energy
and creativity in working with chronic clients, counterbalance any expenditure of time or energy required by the daily meetings.
One organizational resources variable was related to services: larger
teams provided more total services (i.e., teams of 5-7 provided more services than teams of 2-4). As discussed earlier, there is controversy among
186
Administration and Policy in Mental Health
experts concerning the optimal size for ACT teams. Although none of the
teams in our sample of programs were as large as the teams advocated by
some experts, i.e., 10 or more team members (McGrew & Bond, 1995),
within our sample of programs, smaller teams tended to provide fewer
services. We speculate that very small teams may be less efficient, leaving
less time for direct client services, as opposed to administrative or other
non-clinical services.
Contrary to expectations, lower client-to-staff ratio was unrelated to increased services. This null finding for client-to-staff ratio was probably due
to restriction of range in the sample (7:1 to 15:1).
Psychiatrist availability was related to increased total services. However,
the reason for this was unclear. Psychiatrist availability, alone among the
program variables, also was related to increased frequency of office visits
(perhaps because psychiatrists are more likely to see clients in the office).
Moreover, the correlation with office visits was somewhat larger than the
one with total services. When a partial correlation was calculated holding
office visits constant, the correlation between psychiatrist availability and
total services was no longer significant (r=.37, p>.10). Thus, it is possible
that the original simple correlation was spurious.
Only one caseload characteristic was significantly associated with total
services. Programs serving caseloads with a higher percentage with poor
attendance at aftercare also provided fewer total services. That is, the poor
history of service contact characteristic of these clients also was found for
ACT services, despite the fact that ACT often specifically targets these difficult to serve clients. ACT teams may need to provide greater than usual
outreach to ensure adequate services for these individuals. Also, consistent
with previous research (Harris & Bergman, 1990), sites serving caseloads
with a higher frequency of previous hospitalizations did not report greater
frequency of services. Similarly, percent of caseload with a diagnosis of
schizophrenia was unrelated to services. At the program level, then, neither diagnosis nor number of previous hospitalizations appear to be good
indicators of services. At the individual level too, their ability to predict
client outcomes appears weak (Arns & Linney, 1995).
As with most studies, there are several caveats. Individual correlations
should be interpreted cautiously. Many correlations were examined, and
some may be significant by chance. Furthermore, sites were not implemented concurrently nor necessarily with the same goals (e.g., one site was
intended to treat persons who were homeless) and variations in site, program, client, and therapist characteristics were not systematically manipulated. In addition, our data cannot disentangle whether differences in program characteristics influenced service delivery or vice versa, or whether
some third variable underlies both. For example, both service intensity and
the site variables, chosen because they were expert-rated as critical ACT
John H. McGrew and Gary R. Bond
187
ingredients, can be viewed as measures of ACT fidelity. Thus, consistency
in implementation of both organizational and service variables could underlie the obtained correlations. That is, teams that are faithful to the ACT
model may tend to be faithful across-the-board. However, in our earlier
study, when intercorrelations were examined between three scales measuring different aspects of fidelity to ACT, only one was significant (McGrew
et al., 1994). Finally, we were limited by the small sample size. Statistical
power was low, effectively precluding the confident use of multivariate analyses. Moreover, because the programs represented only Bridge-adaptations of ACT, findings may not generalize to all ACT programs.
In addition to the methodological concerns outlined above, the current
study only examined the quantity of services provided. Quality of services
was not examined, which, it could be argued, may be a more important
index of service provision. Future research could profitably examine both
subjective (e.g., client satisfaction with services) and objective measures of
service performance (e.g., number of attempts required to contact case
manager). Independent ratings of service conformity to program standards also should be examined (e.g., measures of therapeutic rapport, adherence to practice parameters).
Implications for Mental Health Administration
Mental health administrators must both identify potential program
models to serve specific service needs and adapt those models to the conditions and needs of the local mental health system, e.g., taking into account limitations and variations in funding, clients, setting, and personnel.
Mindless adherence to a model program may be neither necessary nor
appropriate. When making decisions about how to implement/adapt a
model locally, however, program developers may find little help in the evaluation literature. Partly, this is because the goal of most evaluation research is to examine program effectiveness. Only infrequently do evaluations answer questions of how programs achieved effectiveness, or which
program ingredients are critical.
The current study forms part of a general effort by researchers in the
field to begin to look inside the "black box" of ACT, i.e., to identify the
active ingredients of ACT. This effort is in its infancy. Moreover, findings
from any single study must be viewed cautiously. Nevertheless, in the absence of more definitive research, program administrators may find the
current results helpful in judging the relative importance of particular
ACT elements for service delivery. For example, shared caseloads emerged
as the strongest predictor of both reduction in hospital days, in our earlier
study (McGrew et al., 1994), and total services, in the current study. Program designers, then, rather than being suspicious about the putatively
negative effects of team treatment on productivity, likely should be cau-
188
Administration and Policy in Mental Health
tious about not including shared caseloads in any replication of ACT. Moreover, elements indexing team operation, generally (e.g., team leader who
also provides treatment, clinical responsibility assigned to the team), appear to be particularly important both to provision of intensive services
and to reducing hospital use (McGrew et al., 1994).
Although we hope these results prove useful to program planners as a
preliminary template, we encourage further research to verify findings
from the current study and to test directly the hypotheses generated. The
current study serves as only a tentative first step in beginning to understand service delivery patterns in ACT and their relationship to individual
elements of program implementation.
REFERENCES
Arns, P.G., & Linney.J.A. (1995). Relating functional skills of severely mentally ill clients to subjective
and societal benefits. Psychiatric Research, 46, 260-265.
Bartels, S.J., Teague, G.B., Drake, R.E., Clark, R.E., Bush, P., & Noordsy, D.L. (1993). Substance abuse
in schizophrenia: Service utilization and costs. Journal of Nervous and Mental Disorder, 181, 227-232.
Bond, G.R. (1984). An economic analysis of psychosocial rehabilitation. Hospital & Community Psychiatry, 35, 356-362.
Bond, G.R., Miller, L.D., Krumwied, R.D., & Ward, R.S. (1988). Assertive case management in three
CMHCs: A controlled study. Hospital & Community Psychiatry, 39, 411-418.
Bond, G.R., McDonel, E.C., Miller, L.D., & Pensec, M. (1991). Assertive community treatment and
reference groups: An evaluation of their effectiveness for young adults with serious mental illness
and substance abuse problems. Psychosocial Rehabilitation Journal, 15(2), 31-43.
Bond, G.R., McGrew, J.H., & Fekete, D.M. (1995). Assertive outreach for frequent users of psychiatric
hospitals: A meta-analysis. Journal of Mental Health Administration, 22, 4-16.
Bond, G.R., Pensec, M., Dietzen, L., McCafferty, D., Giemza, R., & Sipple, H.W. (1991). Intensive case
management for frequent users of psychiatric hospitals in a large city: A comparison of team and
individual caseloads. Psychosocial Rehabilitation Journal, 15(1), 90-98.
Bond, G.R., Witheridge, T.F., Dincin.J., Wasmer, D., Webb, J., & De Graaf-Kaser, R. (1990). Assertive
community treatment for frequent users of psychiatric hospitals in a large city: A controlled study.
American Journal of Community Psychology, 18, 865-891.
Boyer, S., & Bond, G.R. (1992). A comparison of assertive community treatment and traditional case
management on burnout and job satisfaction. Outlook 2(2), 13-15.
Brekke, J.S. (1987). The model-guided method for monitoring program implementation. Evaluation
Review, 11, 281-299.
Brekke, J.S., & Test, M.A. (1987). An empirical analysis of services delivered in a model community
support program. Psychosocial Rehabilitation Journal, 10(4), 51-61.
Brekke, J.S., & Test, M.A. (1992). A model for measuring the implementation of community support
programs: Results from three sites. Community Mental Health Journal, 28, 227—247.
Clark, K.A., Landis, D., & Fisher, G. (1990). The relationship of client characteristics to case management service provision: Implications for successful system implementation. Evaluation and Program
Planning, 13, 221-229.
Fekete, D.M., Smith, M.E., Salyers, M., Bond, G.R., Giffort, D., & Hampton, B. (1996, February). Issues
in state-wide implementation of assertive community treatment. Presentation at the annual conference of the National Association of State Mental Health Program Directors, Alexandria, VA.
Feldman, S., & Goldman, W. (1993). Managed mental health care: Editors notes. New Directions for
Mental Health Services, 59, 1-4.
Graham, K. (1980). The work activities and work related attitudes of case management personnel in
New York State Office of Mental Health community support systems. Unpublished dissertation, Albany, New York.
Goldberg, D. (1991). Cost-effectiveness studies in the treatment of schizophrenia: A review. Schizophrenia Bulletin, 17, 453-459.
John H. McGrew and Gary R. Bond
189
Harris, M., & Bergman, H.C. (1990). Misconceptions about use of case management services by the
chronic mentally ill: A utilization analysis. Hospital & Community Psychiatry, 39, 1276-1280.
Hoult, J., Rosen, A., & Reynolds, I. (1984). Community oriented treatment compared to psychiatric
hospital oriented treatment. Social Science and Medicine, 18, 1005-1010.
Intagliata, J., & Baker, F. (1983). Factors affecting case management services for the chronically mentally ill. Administration in Mental Health, 11, 75-91.
Knapp, M., Beecham, V., Koutsogeorgopoulou, V., Hallam, A. Fenyo, A., Marks, I., Conolly, J., Audini,
B., Muijen, M. (1994). Service use and costs of home-based versus hospital-based care for people
with serious mental illness. British Journal of Psychiatry, 165, 195-203.
Levine, M., Toro, P.A., & Perkins, D.V. (1993). Social and community interventions. Annual Review of
Psychology, 44, 525-558.
Maslach, C. (1978). Job burnout: How people cope. Public Welfare, 36, 56-58.
McGrew, J.H., & Bond, G.R. (1995). Critical ingredients of assertive community treatment: Judgments
of the experts. Journal of Mental Health Administration 22(2), 113-125.
McGrew.J.H., Bond, G.R., Dietzen, L., McKasson, M., & Miller, L. (1995). A multi-site study of assertive
community treatment. Psychiatric Services, 46, 696-701.
McGrew.J.H., Bond, G.R., Dietzen, L., & Salyers, M. (1994). Measuring the fidelity of implementation
of a mental health program model. Journal of Consulting and Clinical Psychology, 62, 670-678.
Olfson, M. (1990). Assertive community treatment: An evaluation of the experimental evidence. Hospital & Community Psychiatry, 41, 634-641.
Rapp, C.A. (1993). Client-centered performance management for rehabilitation and mental health
services. In R.W. Flexer & P.L. Solomon (Eds.), Psychiatric Rehabilitation in Practice (pp. 173-192).
Boston, MA: Andover Medical Publishers.
Reinke, B., & Greenley, J.R. (1986). Organizational analysis of three community support program
models. Hospital & Community Psychiatry, 37, 624-629.
Sechrest, L., West, R.G., Phillips, M.A., Redner, R., & Yeaton, W. (1979). Some neglected problems in
evaluation and research: Strength and integrity of treatments. Evaluation Studies Review Annual, 4,
15-35.
Solomon, P. (1992). The efficacy of case management services for severely disabled clients. Community
Mental Health Journal, 28, 163-180.
Stein, L.I., & Test, M.A. (1980). An alternative to mental hospital treatment. 1: Conceptual model,
treatment program, and clinical evaluation. Archives of General Psychiatry, 37, 392-397.
Taube, C.A., Morlock, L., Burns, B., & Santos, A. (1990). New directions in research on assertive
community treatment. Hospital & Community Psychiatry, 41, 642-647.
Test, M.A. (1979). Continuity of care in community treatment. In L. Stein (Ed.), Community support
systems for the long-term patient (pp. 15-24). San Francisco, CA: Jossey-Bass, Inc.
Test, M.A. (1992). Training in Community Living. In R.P. Liberman (Ed.), Handbook of psychiatric rehabilitation (p. 153-170). New York: MacMillan.
Weisbrod, B., Test, M., & Stein, L. (1980). An alternative to mental hospital treatment. II: Economic
cost-benefit analysis. Archives of General Psychiatry, 37, 400-405.
Witheridge, T.F., & Dincin, J. (1985). The Bridge: An assertive outreach program in an urban setting.
New Directions for Mental Health Services, 26, 65-76.