A systematic review of the effect of community treatment orders on service use.

Soc Psychiatry Psychiatr Epidemiol (2014) 49:651–663 DOI 10.1007/s00127-013-0781-0

ORIGINAL PAPER

A systematic review of the effect of community treatment orders on service use Daniel Maughan • Andrew Molodynski Jorun Rugka˚sa • Tom Burns

•

Received: 9 April 2013 / Accepted: 4 October 2013 / Published online: 18 October 2013 Springer-Verlag Berlin Heidelberg 2013

Abstract Purpose The evidence regarding community treatment order effectiveness has been conflicting. This systematic review aims to bring up to date the review performed by Churchill and colleagues in 2005 by assessing and interpreting evidence of CTO effectiveness defined by admission rates, number of inpatient days, community service use, and medication adherence published since 2006. Method Databases were searched to obtain relevant studies published from January 2006 to March 2013. Results 18 studies including one randomised controlled trial were included. There remains lack of evidence from randomised and non-randomised studies that CTOs are associated with or affected by admission rates, number of inpatient days or community service use. The most recent and largest RCT is included in this review and found no significant impact on admission rate (RR = 1.0, 95 % CI 0.75–1.33) or number of days in hospital (IR = 0.90, 95 % CI 0.65–1.26). Results from the two largest longitudinal datasets included in this review do not concur. Studies using the New York dataset found that CTOs were associated with reduced admission rates and inpatient days, while studies using the Victoria dataset generally found

D. Maughan A. Molodynski J. Rugka˚sa T. Burns Oxford Health NHS Foundation Trust, Oxford, UK D. Maughan (&) A. Molodynski J. Rugka˚sa T. Burns Social Psychiatry Group, Department of Psychiatry, Warneford Hospital, University of Oxford, Warneford Lane, Oxford OX3 7JX, UK e-mail: [email protected] J. Rugka˚sa Health Services Research Unit, Akershus University Hospital, Lørenskog, Norway

that they were associated with increased admission rates and inpatient days. Conclusion There is now robust evidence in the literature that CTOs have no significant effects on hospitalisation and other service use outcomes. Non-randomised studies continue to report conflicting results. Distinguishing between CTO recall and revocation and different patterns of community contact is needed in future research to ensure differentiation between CTO process and outcome. Keywords Community treatment orders Outpatient commitment Hospitalisation CTO Community psychiatry

Introduction Community treatment orders (CTOs) are legal regimes that give clinicians the authority to supervise patients with severe mental illness in the community and to bring them to hospital if necessary. First introduced in some states in the USA more than three decades ago [1], similar regimes now exist in approximately 70 different jurisdictions worldwide (only in Western industrialised countries). Early research on CTOs appeared to show that they reduced rates of hospital admission but subsequent studies have failed to confirm that finding [2]. This review has based its search methods on the systematic review by Churchill and colleagues published in 2005. They concluded overall that ‘there is very little evidence to suggest that CTOs are associated with any positive outcomes’ [2] (see Fig. 1 for details). Many of the studies included in Churchill and colleague’s review had significant methodological failings according to the authors and neither of the two randomised controlled trials (RCTs) of CTOs included demonstrated

123

652 Fig. 1 Summary of findings from Churchill and colleague’s systematic review that corresponds to outcomes investigated in this review

Soc Psychiatry Psychiatr Epidemiol (2014) 49:651–663 Churchill and colleagues’ findings [2] Admission rates Neither RCT found statistically significant differences between the CTO and control groups in terms of hospital admissions. Two large cohort studies indicated that CTO patients were more likely to be readmitted to hospital. Inpatient days No study found any differences between CTO and control groups in admission length. Community service use Neither RCT found statistically significant differences between CTO and control groups contact with services and service intensity. Non-randomised studies suggested that CTO patients were more likely to use services more frequently than other patients. Medication adherence Neither RCT found statistically significant differences between the CTO and control groups in terms of adherence with treatment. Non-randomised studies provided inconsistent evidence.

reductions in hospitalisation rates or length of stay [3, 4]. A Cochrane review of the two RCTs concluded that they ‘provide no significant evidence that CTOs affect healthservice use’ [5]. Authors of narrative reviews, however, have drawn more positive conclusions about the value of CTOs [6–10]. This may reflect the different weight given to non-RCT-level evidence. Even so, review articles agree that because of weaknesses in other study designs, RCT evidence is needed to determine the CTO effectiveness [2, 5]. Churchill and colleagues [2] performed a systematic review of the literature on CTOs from 1966 to 2005 that included 72 studies. They included all available studies pertaining to CTOs, providing a thorough analysis of the evidence. Since then there have been several studies examining CTO effectiveness in different jurisdictions. This evidence has not yet been systematically reviewed. Studies include one recent RCT and outcome data from two large longitudinal datasets. Our review examines articles published from 2006 to 2013 regarding the effects of CTOs upon admission rates, number of inpatient days, community service use, and medication adherence. We aim to provide an analysis of current evidence for CTOs that will bring up to date the review performed by Churchill and colleagues [2] and provide evidence for clinical practice and future research. The search terms and strategy used here are those used by Churchill and colleagues, [2] to ensure the same methodological approach.

Method Inclusion criteria A range of outcomes has been investigated in studies of CTOs including social functioning and problem behaviours

123

[11], arrests [12], and homelessness [13]. We limited this review to the following outcomes: admission rates, number of inpatient days, community service use, and medication adherence. This review thus focuses on outcomes relating to inpatient service use as these are commonly the primary outcome measures used in CTO research [10], and CTOs are often legislated specifically to reduce inpatient service use [14]. We also focused on outcomes relating to use of services in the community (community service use, medication adherence) to provide a comprehensive analysis of all mental health services that may be used by patients. We decided on the following inclusion criteria: publication from January 2006 to March 2013 inclusive; English language; a quantitative, experimental or epidemiological design with reference to the specified outcomes. Search strategy Articles were identified from a systematic search of electronic databases. These comprised PsycINFO Medline and EMBASE. The list of search terms is a replication of those used by Churchill and colleagues’ systematic review [2] and consisted of: • • • • • • • •

Community treatment orders CTO Mandatory outpatient Involuntary outpatient Outpatient commitment Involuntary commitment IOT Assisted outpatient treatment

Abstracts of generated articles were reviewed to determine as to which studies related to CTOs. One author (DM) reviewed those full articles that related to CTOs to identify articles that met the inclusion criteria. Articles meeting the inclusion criteria were then hand searched by one author

Soc Psychiatry Psychiatr Epidemiol (2014) 49:651–663

653

here as the Victoria dataset. This is an administrative database of all clinical contacts with mental health patients in the public sector in the State of Victoria, Australia. Between 1990 and 2000 it recorded 16,216 CTOs. Three studies [20–22] analysed the 8-year outcome data from the New York State Office of Mental Health, referred to here as the New York dataset. This is also an administrative database that collected data on all clinical contacts for mental health services in the State of New York. Between 1999 and 2006 it recorded 8,752 CTOs. Results table Fig. 2 Flow chart showing selection of included studies in the systematic review

(DM) to find references to studies potentially missed by the search of library database.

We discussed the results below according to outcome measure. Within each section, studies are then discussed according to the dataset used. The intention is to achieve clarity about the findings from different datasets for the various outcome measures. Numerical data and statistical tests can be found in Table 1.

Data extraction Admission rates We present data in the same format as used in the articles reviewed. We did not perform any re-analysis or metaanalysis. Articles that re-analysed data taken from a previously performed study were recorded as ‘secondary’ evidence to enable greater clarity about data source. To maximise the available evidence, no articles were excluded on the basis of quality.

Results The search generated 1,786 articles, most of which related to a cardiology term ‘chronic-total-occlusion’ also abbreviated as CTO. The review of abstracts revealed 78 articles related to community treatment orders. Full texts of these articles were then assessed for eligibility. Of the 78 articles, 52 did not include the relevant outcome measures and were thus excluded. Two articles were excluded as they were not published in English. A further six articles did not have an experimental or epidemiological design. The remaining 18 articles were included. Hand searching the references of the included articles did not identify any further articles (Fig. 2). These 18 articles used data from 11 data sources (see Table 1). One study reported a RCT. This type of experimental study provides the strongest level of evidence available. Nine of the 18 articles reported controlled before-and-after studies, six reported uncontrolled beforeand-after studies, and two reported epidemiological studies. Two large datasets were analysed and reported in several studies included (see Table 1). Five studies [15–19] analysed the Victoria Psychiatric Case Register, referred to

Some studies use the term ‘readmission rates’ whereas others use ‘admission rates’. Legislation between jurisdictions varies, with some requiring admission prior to CTO initiation and others not. The term admission is used here as the two terms are used interchangeably in the literature. Twelve studies reported on admission rates. The only RCT included in this review found no statistically significant differences between groups for admission rate, time to admission, or number of patients subjected to repeated admissions [14]. Burns et al.’s RCT was conducted over a 12-month period with 333 patients included, making it the largest RCT investigating CTOs to date. It is a multisite study randomizing patients proposed by their clinicians for CTO. To comply with European legislation, at the point of randomization those in the control group were allocated to a form of therapeutic pass (i.e. remaining on their involuntary detention but allowed to be away from the hospital). This so called ‘Section 17 leave’ is short lived and rapidly transferred to voluntary care by the clinician without any necessary additional statutory procedures. Seven studies found that CTOs were statistically significantly associated with a reduced rate of admission [17, 19, 21, 22, 25, 28, 30]. Four of these studies found further reductions in rates of admission if the CTO lasted longer than 6 months [19, 21, 22, 25]. Four studies found that CTOs were significantly associated with an increased rate of admission [15, 16, 24, 27]. Swartz et al.’s [21] study analysed the New York dataset and concluded that CTOs were associated with reduced admission rates, both during the first and second

123

123

Outcome measures

Study description and level of evidence (LoE)

Segal and Burgess [18], Australia


No. of admissions


Admission length

Community service use

No. of admissions Admission length


Admission length

Admission rate

Burgess et al. [15], Australia

UBA study. Investigated the effect of community versus hospital-initiated CTOs. Compared outcomes in a hospital-initiated orders only group, a community-initiated orders only group, and a group with both community and hospital-initiated orders. Groups were not matched. LoE = Primary

CBA study. Analysed the effects of CTOs implemented within the first month of entering mental health services (early use). The first group was hospitalised and discharged on CTO. The second group was hospitalised and discharged to voluntary status. Both these groups were discharged within 30 days of entering the mental health system. The third group was placed on CTO after 30 days of entering the mental health system. Groups were not matched. LoE = Primary

Epidemiological study. Compared the population discharged on CTO to the population discharge to voluntary status. This study excluded those initiated on a CTO in the community. Level of evidence (LoE) = Primary CBA study. 2 control groups used and aggregated for final analysis; a group matched for age, gender and diagnosis and a group randomly drawn from those discharged to voluntary status. LoE = Primary

Studies that used the Victoria dataset

Study, country

Table 1 Characteristics and results of included studies

Community-initiated CTOs associated with reduced admission length compared with hospital-initiated CTOs

Early use of CTOs associated with reduced number of admissions, reduced admission length. No significant association with community service use

CTOs associated with increased number of admissions, increased admission length. No significant association with community service use

CTOs associated with increased admission rate

Summary of finding(s)

CTO: 8,879

no CTO: 16,094

CTO: 8,879

no CTO: 16,094

CTO: 8,879

no CTO: 112,211

CTO: 16,216

Number in study

Community-initiated orders significantly associated with reduced subsequent inpatient days (mean = 35.97, SD = 198.33) compared with individuals in the hospital-initiated order only group (mean = 88.88, SD = 246.06) and the combined group (mean = 267.01, SD = 398.03)

Those put on CTO within 30 days of entry to mental health system were associated with reduced number of admissions (CTO \30 days: mean = 2.3, SD = 2.6; CTO [30 days: mean = 4.7, SD = 4.6; no CTO: mean = 5.1, SD = 5.4); reduced inpatient days (CTO \30 days: mean = 39.3, SD = 110.3; CTO [30 days: mean = 86.1, SD = 116.1; no CTO: mean = 159.9, SD = 328.3); reduced inpatient days per admission (CTO \30 days: mean = 16.6, SD = 25.3; CTO [30 days: mean = 25.7, SD = 47.2; no CTO: mean = 46.5, SD = 122.0) and reduced community care episodes (CTO \30 days: mean = 3.4, SD = 2.8; CTO [30 days: mean = 6.0, SD = 4.5; no CTO: mean = 7.0, SD = 5.2)

CTO use associated with an increased risk of admission (hazard ratio 1.27; 95 % CI 1.24–1.30 p \ 0.001). Risk of admission reduced with increasing number of admissions (admission 1: HR 1.43; 95 % CI 1.36–1.51, admission 3: HR 1.08; 95 % CI 1.02–1.15, admission 5: HR 0.99; 95 % CI 0.92–1.07, admission 10?: HR 0.82; 95 % CI 0.79–0.85) CTO use associated with more inpatient days compared with control (131.1 v. 77.0 days, respectively), increased number of admissions (CTO = 4.7, SD = 5.1; control = 2.1, SD = 2.7) and increased community treatment days (CTO = 203.9, SD = 232.0; control = 86.9, SD = 154.8). CTO use associated with shorter admissions (b = -8.32, SE = 2.55, p \ 0.001), and fewer inpatient days per 100 days in care (b = -7.76, SE = 0.49, p \ 0.001). In UBA analysis CTO patients spent fewer days per month as an inpatient on average post-CTO compared with pre-CTO (pre-CTO: 6.4 ± 9.1 days, post-CTO: 1.8 ± 4.1 days, p \ 0.001)

Outcomes from studies: relevant to inclusion criteria

654 Soc Psychiatry Psychiatr Epidemiol (2014) 49:651–663

No. of admissions


CBA study. Compared the effect of CTOs that lasted at least 6 months. Controls were a diagnostically matched, treatment-compliant group with periods of voluntary outpatient care that lasted at least 6 months. LoE = Primary


Van Dorn et al. [22], USA

No. of admissions

Swartz et al. [21], USA

Admission rate

MPR

MPR


Admission length

MPR

Busch et al. [20], USA

UBA study. Compared medication possession rates and admission rates of those receiving oral or depot antipsychotic or both. LoE = Primary

This CBA study presented the main findings from the 8-year outcome data from the New York State dataset. Controls were not matched, but were those receiving ACT without CTO. LoE = Primary

CBA study. Compared medication possession rates between three groups; those on CTO (with or without ACT), those on ACT without CTO, and those without either intervention. Controls were matched according to dates of treatment. LoE = Primary

Studies that used the New York dataset


Admission length

Outcome measures

Study, country

Table 1 continued

CTO period associated with reduced admission rate and increased MPR

CTOs associated with reduced number of admissions, reduced admission length, increased community service use and increased MPR

CTOs associated with a greater increase in MPR

CTO: 591

CTOs lasting [6 months associated with reduced number of admissions. No significant association with admission length or community service use

CTO: 3,576

no CTO: 2,025

CTO: 3,576

no CTO: 4,915

CTO: 2,847

no CTO: 591

Number in study


CTO use with ACT associated with increased MPR (OR 1.05, 95 % CI 1.00–1.10). If CTO[6 months, increased MPR sustained without ACT (OR = 1.19, 95 % CI 1.10–1.30). If CTO lasted \6 months with continued ACT, MPR significantly reduced, over and above those with no current or past CTO and no current ACT (OR 0.77, 95 % CI 0.61–0.96). CTO without ACT associated with reduced admission rates compared to those with ACT alone (\6 months: OR 0.78, 95 % CI 0.70–0.86;[6 months: OR 0.68, 95 % CI 0.60–0.76). Reduced admission rate sustained post-CTO, if CTO [6 months (OR 0.59, 95 % CI 0.51–0.69)

CTO use associated with reduced admission (\6 months: OR = 0.77, 95 % CI 0.72–0.82; \1 year OR = 0.59, 95 %CI 0.54–0.65), increased percentage of those receiving C80 % MPR (\6 months: OR 1.47, 95 % CI 1.40–1.55;\1 year OR 1.88, 95 % CI 1.75–2.01), and reduced days hospitalised (\6 months: OR 0.80, 95 % CI 0.78–0.82; \1 year OR 0.84, 95 % CI 0.81–0.86). Addition of a CTO to ACT significantly reduced likelihood of admission beyond reductions with ACT alone (OR 0.43, 95 % CI 0.29–0.62) and significantly increased service use (OR 2.13, 95 % CI 1.45–3.12)

Predicted probability of an MPR C80 % increased over time (CTO group increased by 31–40 %, AOT increased by 15–22 %, and neither treatment group increased by 8–19 %)

CTO group had reduced inpatient days during 1-year period post-CTO compared with 1-year period pre-CTO, (preCTO = 56.4 ± 82.5 days; post-CTO = 19.6 ± 40.4 days). Control group also had reduced inpatient days (pre = 37.2 days, post = 10.4 days). Each day of community treatment on CTO decreases the chance of hospitalisation in the post-period for the CTO group by 3.2 % over those in the comparison group (v2 = 425.70, df = 12, p \ 0.001). CTO group associated with an increased number of community treatment days (preCTO = 27.5 ± 40.2 days, post-CTO = 41.1 ± 41.1 days), control group also had increased number of community treatment days (pre = 13.3 ± 42.0 days, post = 18.8 ± 24.7 days)


Soc Psychiatry Psychiatr Epidemiol (2014) 49:651–663 655

123

123

Outcome measures

Hunt et al. [25], Canada

Burns et al. [14], UK

Kisely et al. [24], Australia and Canada

Kisely et al. [23], Australia


Admission length

No. of admissions


Admission length

Time to readmission

Admission rate


Admission length

Admission rate


Admission length

Individual studies

Study, country

Table 1 continued

CBA study examining the effects of CTO in addition to ACT in Toronto. All those on ACT were included in the study with those not on CTO used as controls. LoE = Primary

Non-blinded 12-month parallel arm RCT in the UK. 333 patients analysed, 167 were discharged from involuntary hospitalisation to CTO, 169 were discharged to treatment as usual, which was Section 17. This is a short-term measure to test patients’ stability before discharging them to voluntary status (median duration: 8 days). Eligible patients were those detained for inpatient treatment, aged 18–65 years, diagnosed with psychosis, considered for CTO by their clinical team and able to give informed consent. Data was obtained from medical notes and structured patient interviews at baseline and 12 months. LoE = Primary

Epidemiological study. Compared a population on CTOs from a previously published CBA analysis of CTOs in western Australiaa with a control group in Nova Scotia, Canada, where CTOs were not available. Population samples were matched for age, gender, and diagnosis. LoE = Secondary

CBA study examining for change in CTO effectiveness over the 11 years of use in Western Australia. Comparisons made between all those on CTO between 1997 and 2008 and an equal number of controls matched on age, sex, discharge date and diagnosis. LoE = Primary


CTOs associated with a larger reduction in number of admissions and admission length and reduced community service use

CTOs not associated with admission rate, time to readmission or admission length

CTOs associated with reduced admission rate, reduced number of admissions lasting [100 days and increased community service use

CTOs associated with reduced admission length and increased community service use


no CTO: 92

CTO: 224

no CTO: 169

CTO: 167

no CTO: 265

CTO: 274

no CTO: 2,958

CTO: 2,958

Number in study

CTO group associated with significantly greater reductions in average number of hospital days (\6 months: CTO = 39; control = 27: p \ 0.007; [6 months: CTO = 44; control = 25: p \ 0.001). CTOs lasting [6 months associated with a larger reduction in number of admissions (CTO = 1.3, SD = 0.96; control = 0.9, SD = 1.71, p \ 0.009). Control group remained longer with assertive community management (Control = 519.5 days; CTO = 306 days)

At 12 months there was no significant difference in admission rate 59/166, (36 %) CTO patients versus 60/167, (36 %) controls, adjusted relative risk 1.0 (95 % CI 0.75–1.33) and no significant difference in time to admission from randomisation or from initial hospital discharge (median 295 days for CTO patients versus 241 for controls p = 0.755). There were no significant differences in the number of days in hospital (median 41.5 days for CTO patients versus 48 days for controls, incident rate 0.90, 95 % CI 0.65–1.26) and in the number of patients subjected to repeated admissions (10 CTO patients and 18 controls). Randomised compulsory outpatient supervision differed significantly between the two groups (median 183 days for CTO patients versus 12 days for controls, p \ 0.001) alongside the duration of total time under legal compulsion (median 255 days for CTO patients versus 102 days for controls (Wilcoxon rank-sum p \ 0.001). No significant difference found between groups for community service contacts per month [CTO mean = 3.0 (SD = 3.1) median = 2.1 (IQR = 0.4–4.4); no CTO mean = 3.9 (SD = 5.9) median = 2.2 (IQR = 0.8–4.7)]

CTO group associated with fewer admissions lasting 100 days or longer in the year post-CTO (adjusted OR = 0.02). CTO group was associated with more frequent and earlier admission on survival analysis—no data given. Outpatient contacts reported as significantly greater for the CTO group—no data given

CTO use associated with a mean decrease of 5 inpatient days (in the first year of CTO) when compared to controls (b = -5.23, SE = 1.60, t = -3.26, p \ 0.001). CTO use was associated with an increase of 8 days in outpatient contacts (in the first year of CTO) when compared to controls (b = 8.31, SE = 1.17, t = 7.11, p \ 0.001)



No. of admissions Admission length

No. of admissions

Zanni et al. [27], USA

Muirhead et al. [28], Australia

O’Brien et al. [29], Canada

Admission length

Segal et al. [26], Australia



Admission length

Outcome measures

Study, country

Table 1 continued

UBA study investigating community engagement and use of supportive housing. All patients on CTOs from two psychiatric hospitals in Ontario during a 3-year period were included. LoE = Primary

UBA study examining the effect of CTOs on 2 groups of patients with schizophrenia. One group treated with oral (n = 31), the other with depot medication (n = 63). All patients on CTO between 1996 and 1999 across 3 mental health services in Melbourne were included. LoE = Primary

CBA study by re-analysing an existing data set collected from 1995 to 1997. All patients at St Elizabeth Hospital, District of Columbia, that were put on CTO during these 2 years were included. Controls were voluntary outpatients from this hospital with at least one prior admission. LoE = Secondary

CBA study examining the effects of CTO. This was a secondary analysis of a study that had included 228 subjects and 228 controls. This secondary analysis included 129 subjects and 117 controls (matched on diagnosis, sex, age and race) as only those hospitalised before-and-after CTO were included. LoE = Secondary


CTO period associated with increased community service use

CTO period associated with reduced number of admissions, reduced admission length and reduced community service use

CTOs associated with increased number of admissions and increased admission length

CTOs associated with reduced admission length


CTO: 84

CTO: 94

CTO: 115 no CTO: 78

no CTO: 117

CTO: 129

Number in study

CTO period associated with an increase in patients receiving ACT [pre-CTO = 7/84 (8.3 %) post-CTO = 19/84 (15.4 %)]

80 % increase in community-based services (t = 8.19; p \ 0.01)

CTO group significantly associated with reduced number of admissions (Mann–Whitney Z = -5.986, p \ 0.001), reduced total inpatient days (Mann–Whitney Z = -6.696, p \ 0.001), reduced number of crisis team referrals (Mann–Whitney Z = -2.456, p = 0.014) and reduced number of other relapses, defined by increase in medication or frequency of contact (Mann–Whitney Z = -2.187, p = 0.029). Number of admissions significantly reduced for both the oral subgroup and depot subgroup during the year following CTO (Mann–Whitney Z = -3.947, p \ 0.001, Mann–Whitney Z = -4.556, p \ 0.001, respectively), and total length of hospital admissions also reduced for both the oral subgroup (Mann–Whitney Z = -4.055, p \ 0.001) and depot subgroup during the year following CTO (Mann–Whitney Z = 5.330, p \ 0.001)

Before-and-after analysis demonstrated that CTO period was associated with reduced number of admissions per year (preCTO = 0.81; post-CTO = 0.57, p \ 0.002) and reduced inpatient days per year (pre-CTO = 82.55; post-CTO = 56.23, p \ 0.061)

CTO group significantly associated with increased number of admissions compared to controls (CTO = 4.25, SD = 3.80; control = 1.64, SD = 1.44, p \ 0.0001), increased admission length (days) (CTO = 437, SD = 496.77; control = 87, SD = 190.29, p \ 0.0001), increased average number of days per admission (CTO = 162, SD = 267.04; control = 45, SD = 75.08, p \ 0.0001)

CTO group associated with a mean reduction of 19.16 days per admission (SE = 9.0, t = 2.13, df = 1, p = 0.034). Mean number of inpatient days decreased during 1-year period after CTO implementation (pre-CTO = 46.3 days; post-CTO = 26 days)


Soc Psychiatry Psychiatr Epidemiol (2014) 49:651–663 657

123

123

No. of admissions

UBA investigating the first 50 cases on CTO in Florida after they were introduced in 2005. LoE = Primary

UBA study examining the effects of CTOs over different time periods. All patients given a CTO at two hospitals in Quebec between 1998 and 2007 were included. LoE = Primary


CTO period associated with reduced number of admissions

CTO period associated with reduced number of admissions. No association with admission length


CTO: 50

CTO: 72

Number in study

CTO period associated with a reduced number of total admissions (pre-CTO = 46 range = 1–7; post-CTO = 41 range = 1–13)

This study analysed four time periods; early, pre-index, index and post-index. The index period commenced at beginning of the admission when the first CTO was given. The post-index period commenced at readmission post-CTO. The pre-index period commenced at admission prior to the index admission and the early period was the 5 years prior to the beginning of the pre-index admission. The CTO period (index period) was significantly associated with a decrease in the median number of admissions per year (early = 1.34; pre-index = 1.54; index = 0.52; postindex = 0.75, Kruskal–Wallis X2 = 38.80, df = 3, p \ 0.001). The CTO period (index period) was significantly associated with an increase in median days to admission (early = 164.5; preindex = 193; index = 662; post-index = 420, Wilcoxon test X2 = 28.26, df = 3, p \ 0.001). No statistically significant difference in median hospital days per year (early = 40.4; preindex = 56.4; index = 41.4; post-index = 39.6, Kruskal–Wallis X2 = 3.42, df = 3, p \ 0.33)


a

For reference of this study see [32]

Studies are sorted initially due to those studies analysing the same dataset and secondly according to study size CBA controlled before-and-after study, UBA uncontrolled before-and-after study, ACT assertive community treatment, MPR medication possession ratio, b regression coefficient

Christy et al. [31], USA

No. of admissions

Nakhost et al. [30], Canada

Admission length

Outcome measures

Study, country

Table 1 continued



6-month periods on CTO. The addition of a CTO to Assertive Community Treatment (ACT) was found to further reduce the likelihood of admission when compared with ACT alone but only if the CTO lasted longer than 6 months [21]. The study by Van Dorn et al. also used the New York dataset and compared CTO patients to voluntary patients on ACT, and showed that reduced admission rates were sustained during the 6 months after the CTO had ended [22]. By contrast, two studies using the Victoria dataset found that CTOs were significantly associated with an increased likelihood of admission: one CBA study by Segal and Burgess [16] and the large epidemiological analysis by Burgess et al. [15]. Other findings from the Victoria dataset included a CBA study, which found that being placed on a CTO within 30 days of entering the mental health system was significantly associated with reduced subsequent admissions [17]. Another CBA study by Segal, which used the Victoria dataset, found that being on a CTO for longer than 6 months significantly decreased the risk of admission as compared to controls [19]. Hunt et al.’s CBA study from Toronto found that both CTO and control groups had reduced number of admissions. However, if CTOs lasted longer than 6 months there was a greater reduction of admissions in the CTO group when compared with controls. There was no significant difference if CTOs lasted less than 6 months [25]. Kisely et al.’s [24] epidemiological study from Nova Scotia and Western Australia stated that the CTO population had more frequent and earlier admissions in a survival analysis, although they reported no figures for this particular outcome in their study. This study selected patients from different countries with different mental health services and may as a result have introduced significant confounding factors. They also reported difficulties in matching controls. Three UBA studies reported a reduction in admissions during the CTO period [28, 30, 31] although their generalisability is limited due to the lack of control groups. In addition, one of these UBA studies did not report significance values [31]. Length of stay and total number of inpatient days ‘Length of stay’ refers to the duration of individual hospital admissions, while the ‘total number of inpatient days’ defines the number of inpatient days during a specified time period. Either one or both of these two outcomes were reported on in 13 studies (see Table 1). Seven studies reported a significant association between CTOs and a reduced number of inpatient days [17, 18, 21, 23, 25, 26, 28] and two studies reported a significant association between CTOs and an increased number of inpatient days [16, 27]. Three studies reported no association of

659

significance [14, 19, 30] including the only randomised trial. Other study types provide conflicting evidence and also experienced difficulty in obtaining accurate control groups. Swartz et al.’s [21] CBA analysis of the New York dataset reported a significant association between CTO use and a reduced number of inpatient days regardless of length of order. By contrast, Segal and Burgess’s [16] CBA study compared the Victoria dataset with matched controls and observed a significant increase in total inpatient days for the CTO group. Other findings from the Victoria dataset included one CBA study, which showed that those who received only community-initiated CTOs had significantly fewer inpatient days than those with hospital-initiated CTOs [18]. Positive prognostic factors reported to be associated with community-initiated orders (being older, female, more likely to have affective disorder) could have confounded this analysis. Another CBA study that used the Victoria dataset found those placed on CTOs within 30 days of first service contact had fewer inpatient days per admission and overall when compared with controls [17]. The large CBA study performed by Kisely et al. [23] showed that CTO use was associated with a mean decrease in number of inpatient days in the first year of CTO, when compared to controls. In Hunt et al.’s [25] CBA study, both CTO and control groups had reduced average number of inpatient days but there were significantly greater reductions for the CTO group when compared with controls. Muirhead et al. [28] showed that the number of inpatient days significantly reduced during the CTO year, when compared with the year pre-CTO. This study had no control group and other factors potentially reduce its generalisability such as large numbers of exclusions, problematic allocation procedures, and differences in medication between subgroups. Segal et al.’s [26] secondary analysis showed that CTOs were significantly associated with a reduced number of inpatient days overall. The non-significant findings of the original study conflict with the significant findings of this secondary analysis, which used only 50 % the original sample. Also, this study was unable to control for the number of prior admissions, which is a significant risk factor for future admissions [33]. Zanni and Stavis’s [27] CBA analysis in the District of Columbia reported that CTO use was associated with an increased length of admission. In this study, there were significant amounts of missing data, a randomly selected and unmatched control group, and no record of how subjects were recruited. Kisely et al.’s [24] epidemiological study showed that those subject to CTO were less likely to have prolonged hospital stays of over 100 days (the original study found an overall increase in total inpatient days).

123

660

Community service use Community service use refers to the number of service contacts a patient has in a given time frame. It is generally used as a measure of community engagement in CTO research. In one study included here, there is reference to whether these contacts were emergency (referred to as Crisis team referrals) or routine. All other studies simply record number of service contacts. Four of the 10 studies that investigated this outcome reported that CTOs were associated with increased service use [21, 23, 24, 29] and two studies reported that they were associated with decreased service use [25, 28]. Of the four studies that reported no significant association between CTOs and community service use, three used the Victoria dataset [16, 17, 19]. Burns et al.’s [14] RCT found no statistically significant association between CTOs and number of community contacts per month. Swartz et al.’s [21] CBA analysis of the New York dataset demonstrated that addition of a CTO to ACT was significantly associated with increased service use over and above increases seen with ACT alone, but only if the CTO lasted longer than 6 months. Kisely et al’s [23] large CBA analysis in Western Australia showed that CTO use was significantly associated with a mean increase in outpatient contacts in the first year of CTO, when compared to controls. O’Brien et al.’s [29] UBA study in Canada reported that 80 % of those on CTO had increased their use of community-based services and that CTO use was associated with increased use of ACT. Patients on CTO were, however, prioritised for ACT in this service. In contrast, Muirhead et al.’s [28] UBA study demonstrated significant reductions in crisis team referrals during the CTO period. Hunt et al.’s [25] CBA analysis reported that those on CTO were less likely to experience ACT when compared with those not on CTO although no significance values were given. Medication adherence The three studies that investigated medication adherence all reported that CTO use was associated with increased rates of adherence [20–22]. All three based their analysis on the New York dataset and used a proxy measure for adherence called the medication possession ratio (MPR). MPR is a dichotomous variable constructed from pharmacy records. A ‘high possession month’ equates to patients picking up at least 80 % of prescribed medication from the pharmacy and a ‘low possession month’ is when less than 80 % is picked up. Administration of depot was coded as 100 % MPR for the given period. One of the studies concluded that CTO use with ACT was significantly associated with an increased MPR [21].

123


Another CBA study found that after discharge of a CTO lasting less than 6 months, patients then had a significantly reduced MPR when compared with those who had never received a CTO or ACT. In contrast, if the CTO was in place for over 6 months, an increased MPR was sustained even after the CTO and ACT was stopped [22]. Another CBA study compared those receiving CTO without ACT, those receiving ACT without CTO, and those without either intervention [20]. For all three groups, the predicted probability of MPR C80 % increased over time. The authors noted that potential false negative results regarding substance misuse might have affected results.

Discussion The literature on CTOs published since 2005 has provided conflicting results. This finding is in keeping with Churchill and colleague’s review of the evidence up to and including 2005 [2]. In each of the outcomes investigated here (other than adherence rate), there has been no consensus of evidence in the studies which considered that CTOs are associated with any particular outcome. The results from the UK-based RCT [14] correspond with results from the two RCTs from the US [3, 4], that there is no evidence that CTOs affect admission rates or number of inpatient days. In addition, Burns et al.’s [14] RCT found that CTOs have no significant effects on community service use. These three RCTs are from different jurisdictions with different mental health systems and different CTO legislation. Despite this, their results regarding CTO effectiveness concur. These are the only three randomised controlled studies that have reviewed CTO use and the congruity of their results is significant. Results from the two largest longitudinal datasets do not concur. While studies using the New York dataset found that CTOs were associated with reduced admission rates and inpatient days, studies using the Victoria dataset generally found that CTOs were associated with increased admission rates and inpatient days. Differences between available mental health services and limitations in study design may have contributed to these contrasting findings [2]. Differences in CTO legislation between jurisdictions may also have contributed. For example, some types of CTO provide a form of conditional discharge, while others are more flexible and allow CTO initiation from the community [10]. Several studies report associations between outcome and the duration of CTOs. CTOs that last longer than 6 months were reported to have improved outcomes over and above CTOs that last less than 6 months [19, 21, 22, 25]. This finding may be due to an inclusion effect. In most jurisdictions, CTOs last a maximum of 6 months [2], after


which time a further application is required for the CTO to continue. It is possible that those responding well to the CTO intervention were those that were maintained on CTO after 6 months. A further consistent finding between studies was that CTOs are associated with increased rates of medication possession. This outcome was, however, only investigated in three studies and all used the same dataset [20–22]. None of the articles included in this review differentiated between the two potential types of admission that can occur under a CTO. A ‘recall’ or ‘breach’ admission is when a patient is admitted to hospital for assessment and/or to enable administration of medication and is then discharged within a specified time frame to remain on the CTO. The other type of admission (referred to as ‘recall then revocation’ in the UK, or ‘recall then admission’ in other jurisdictions such as Victoria or New South Wales) is when a patient is taken to hospital to receive medication but is then required to stay in hospital due to a deterioration in mental health. This is an important distinction as using ‘recalls’ to ensure medication adherence is part of the standard process of a CTO. In such a case, a very short admission may enable maintenance of mental health through continued adherence. Recall may, therefore, be considered part of the process of a CTO, whereas revocation (that is, the termination of the CTO) may be considered an outcome of a CTO. Despite the fact that the majority of CTO studies focus on rates of admission [9], the literature to date does not distinguish between types of admission. Admission rate has traditionally been considered to be a reliable proxy measure for relapse [34], but we suggest that where recalls cannot be distinguished from revocations it is problematic to use this as a proxy measure for relapse and, therefore, as a measure of CTO effectiveness. A similar lack of clear categorization is evident in the recording of community service contacts. Although the level of community support available is likely to vary significantly between services, an increased number of contacts is often used as a proxy measure for improved engagement. As patients under CTO often have a history of poor engagement, regular contact with the community team is one of the legal conditions routinely attached to a CTO. An increased frequency of regular contacts may, therefore, reflect the process of the CTO. Some types of community contacts, such as crisis contacts or periods of intense short-notice contact, may reflect relapse and be considered an outcome. Measures that simply record the frequency of community contacts may not, therefore, be a valid measure of CTO effectiveness. Recording the pattern and nature of contacts could potentially provide more specific information. Such details would include frequency and type of contact, the service involved, and whether the

661

contacts are planned or emergencies. Types of community contact were differentiated in only one of the studies reviewed here [30]. This study found that CTOs were associated with reductions in both crisis and routine community contacts. In the wider literature on treatment effectiveness, pattern of community contacts is rarely used as an outcome measure. The only types of study that have measured patterns of community contacts are studies comparing costs of community care and inpatient care [35]. Outcomes of this nature should allow a more accurate and sophisticated understanding of mental health service use and patient needs. As community services continue to develop and become increasingly complex this is likely to become increasingly important. Limitations of included studies Only one of the studies randomised its sample. A number of studies had difficulty matching intervention and control groups. Other studies used administrative datasets. This type of data collection has the recognized problem of reduced validity [36]. The necessarily lengthy follow-up periods in some studies, although useful given the longterm nature of the CTO intervention, introduce confounding factors such as changes in clinical practice and service design as well as the natural fluctuation of illness severity. These first two issues may be reflected in the fluctuating numbers of CTO referrals over time, within and between regions [37]. The difficulty in measuring the effects of a CTO intervention without a control group is demonstrated by the conflicting results from within two datasets where the case–control component and the before-and-after component found CTOs associated with increased and decreased inpatient days, respectively [16, 27]. This observation undermines the generalisability of the uncontrolled before-and-after studies as it brings to light the difficulty of over-interpreting results without a control group for reference. Three studies included in this review are secondary analyses of existing datasets [24, 26, 27]; results from these studies should be considered carefully. The use of the MPR as a proxy measure for medication adherence may limit the generalisability of the findings from the three studies that analysed this outcome. Methods of prescribing may change following CTO intervention, with more patients being prescribed depot [38]. As administration of depot was coded as 100 % MPR for the given time frame in these three studies, increased MPR may reflect an increased number of patients on depot. There was no reference made to the number of patients on depot or oral medication in any of the relevant publications. The three studies that reported on this outcome all shared the same data set and any conclusions must be considered with that in mind.

123

662

Limitations of this review We limited the scope of this review to specific clinical outcomes and, therefore it does not include studies of how CTOs may affect other important aspects such as quality of life, social functioning, or offending behaviour. A further limitation includes the restricted time frame of the literature search between 2006 and 2013. This review aims to bring up to date the review by Churchill and colleagues [2] and although more studies were published during the period covered in their review, their conclusions are very similar to ours. This review is descriptive in nature and has not attempted to further analyse data from the included studies.

Conclusions There is a current lack of differentiation between CTO process and outcome in the literature. Studies investigating CTO effectiveness should distinguish between the different types of admission and between the type and pattern of community contacts to accurately determine the effects of CTOs upon mental health services and those who use them. This review analyzes the accumulating evidence regarding clinical outcome of community treatment order use from 2006 to 2013. A broad range of evidence is included: a RCT, controlled and uncontrolled before-andafter studies, and epidemiological studies. Our findings concur with the systematic review performed by Churchill and colleagues [2] and the Cochrane review [5]. Both of these reviews analysed studies up to and including 2005 and found no significant evidence for CTOs association with hospital or community service use outcomes. Research in this area is challenging. In addition to methodological challenges, frequent changes to mental health services and the fluctuating symptoms experienced by those with mental illness make it difficult to provide rigorous evidence for CTO effectiveness. There are also many individual patient factors that can affect CTO outcomes, for example, a recently published study suggests that CTO use varies according to type of medication administration [39]. In a situation with conflicting evidence from controlled and uncontrolled before-and-after studies, RCTlevel evidence has been called for to further determine the effects of CTOs [5] and assist clinicians to make evidencebased decisions. This review includes the only recent RCT on CTOs. This has not been included in previous reviews. While other types of studies have provided conflicting evidence, three RCTs performed in different jurisdictions have all found that CTOs do not reduce the rate of admission to hospital. Given the consistency of the RCT

123


evidence and the sustained inconsistency of non-randomised studies, there is now a strong level of evidence that CTOs have no significant effect on hospitalisation outcomes or community service use. Conflict of interest All authors declare that they have no conflict of interest

References 1. Geller J (2006) The evolution of outpatient commitment in the USA: from conundrum to quagmire. Int J Law Psychiatry 29:234–248 2. Churchill R, Owen G, Hotopf M, Singh S (2007) International experiences of using community treatment orders. Institute of Psychiatry, London. http://www.iop.kcl.ac.uk/news/downloads/ final2ctoreport8march07.pdf 3. Steadman H, Gounis K, Dennis D, Hopper K, Roche B, Swartz M et al (2001) Assessing the New York City involuntary outpatient commitment pilot program. Psychiatr Serv 52(3):330–336 4. Swartz MS, Swanson JW, Wagner HR, Burns BJ, Hiday VA, Borum R (1999) Can involuntary outpatient commitment reduce hospital recidivism?: Findings from a randomized trial with severely mentally ill individuals. Am J Psychiatry 156(12):1968–1975 5. Kisely S, Campbell LA, Preston N (2005) Compulsory community and involuntary outpatient treatment for people with severe mental disorders. Cochrane Database Syst Rev 3:CD004408 6. Ridgely S, Borum R, Pertila J (2001) The effectiveness of involuntary outpatient treatment. Empirical evidence and the experience of 8 States. RAND, California 7. O’Reilly RL (2001) Does involuntary outpatient treatment work? Psychiatr Bull 25:3714 8. Bindman J (2002) Involuntary outpatient treatment in England and Wales. Curr Opin Psychiatry 15:595–598 9. Swartz M, Swanson J (2004) Involuntary outpatient commitment, community treatment orders, and assisted out-patient treatment: what’s in the data? Can J Psychiatry 49:585–591 10. Dawson J (2005) Community treatment orders: international comparisons. The Law Foundation, New Zealand 11. Ingram G, Muirhead D, Harvey C (2009) Effectiveness of community treatment orders for treatment of schizophrenia with oral or depot antipsychotic medication: changes in problem behaviours and social functioning. Aust N Z J Psychiatry 43(11):1077–1083 12. Link B, Epperson M, Perron B, Castille D, Yang L (2011) Arrest outcomes associated with outpatient commitment in New York State. Psychiatr Serv 62(5):504–508 13. Compton S, Swanson J, Wagner H, Swartz M, Burns B, Elbogen E (2003) Involuntary outpatient commitment and homelessness in persons with severe mental illness. Ment Health Serv Res 5(1):27–38 14. Burns T, Rugka˚sa J, Molodynski A, Dawson J, Yeeles K, Vazquez-Montes M, Voysey M, Sinclair J, Priebe S (2013) Community treatment orders for patients with psychosis: a randomised controlled trial. Lancet 381. doi:10.1016/S0140-6736(13)60107510.1016/S0140-6736(13)60107-5 15. Burgess P, Bindman J, Leese M, Henderson C, Szmukler G (2006) Do community treatment orders for mental illness reduce readmission to hospital? An epidemiological study. Soc Psychiatry Psychiatr Epidemiol 41:574–579 16. Segal S, Burgess P (2006) Conditional release, a less restrictive alternative to hospitalization? Psychiatr Serv 57:1600–1606

Soc Psychiatry Psychiatr Epidemiol (2014) 49:651–663 17. Segal S, Burgess P (2006) Factors in the selection of patients for conditional release from their first psychiatric hospitalization. Psychiatr Serv 57:1614–1622 18. Segal S, Burgess P (2008) Use of community treatment orders to prevent psychiatric hospitalization. Aust N Z J Psychiatry 42:732–739 19. Segal S, Burgess P (2006) The utility of extended outpatient civil commitment. Int J Law Psychiatry 29(6):525–534 20. Busch A, Wilder C, Van Dorn R, Swartz M, Swanson J (2010) Changes in guideline-recommended medication possession after implementing Kendra’s Law in New York. Psychiatr Serv 61:1000–1005 21. Swartz M, Wilder C, Swanson J, Van Dorn R, Robbins P, Steadman H et al (2010) Assessing outcomes for consumers in New York’s assisted outpatient treatment program. Psychiatr Serv 61:976–981 22. Van Dorn R, Swanson J, Swartz M, Wilder C, Moser L, Gilbert A et al (2010) Continuing medication and hospitalization outcomes after assisted outpatient treatment in New York. Psychiatr Serv 61:982–987 23. Kisely S, Preston N, Xiao J, Lawrence D, Louise S, Crowe E, Segal S (2013) An eleven-year evaluation of the effect of community treatment orders on changes in mental health service use. J Psychiatr Res 47(5):650–656 24. Kisely S, Campbell LA, Preston NJ, Xiao J (2006) Can epidemiological studies assist in the evaluation of community treatment orders? - The experience of Western Australia and Nova Scotia. Int J Law Psychiatry 29(6):507–515 25. Hunt A, da Silva A, Lurie S, Goldbloom D (2007) Community treatment orders in Toronto: the emerging data. Can J Psychiatry 52(10):647–655 26. Segal S, Preston N, Kisely S, Xiao J (2009) Conditional release in Western Australia: effect on hospital length of stay. Psychiatr Serv 60(1):94–99 27. Zanni G, Stavis P (2007) The effectiveness and ethical justification of psychiatric outpatient commitment. Am J Bioeth 7(11):31–41 28. Muirhead D, Harvey C, Ingram G (2006) Effectiveness of community treatment orders for treatment of schizophrenia with oral or depot antipsychotic medication: clinical outcomes. Aust N Z J Psychiatry 40:596–605

663 29. O’Brien A, Farrell S, Faulkner S (2009) Community treatment orders: beyond hospital utilization rates examining the association of community treatment orders with community engagement and supportive housing. Community Ment Health J 45:415–419 30. Nakhost A, Perry JC, Frank D (2012) Assessing the outcome of compulsory treatment orders on management of psychiatric patients at 2 McGill University-associated hospitals. Can J Psychiatry 57(6):359–365 31. Christy A, Petrila J, McCranie M, Lotts V (2009) Involuntary outpatient commitment in Florida: case information and provider experience and opinions. Int J Forensic Ment Health 8:122–130 32. Preston N, Kisely S, Xiao J (2002) Assessing the outcome of compulsory psychiatric treatment in the community: an epidemiological study in Western Australia. BMJ 324:1244–1246 33. Huntley D, Cho D, Christman J, Csernansky J (1998) Predicting length of stay in an acute psychiatric hospital. Psychiatr Serv 49(8):1049–1053 34. Burns T (2007) Hospitalisation as an outcome measure in schizophrenia. Br J Psychiatry 191(suppl. 50):37–41 35. Knapp M, Chisholm D, Leese M, Amaddeo F, Tansella M, Schene A et al (2002) EPSILON Study Group. Comparing patterns and costs of schizophrenia care in five European countries: the EPSILON study. Acta Psychiatr Scand 105(1):42–54 36. Baglio G, Sera F, Cardo S, Romanini E, Guasticchi G, Labek G, Torre M (2009) The validity of hospital administrative data for outcome measurement after hip replacement. Ital J Public Health 6(2):114–127 37. Robbins P, Keator K, Steadman H, Swanson J, Wilder C, Swartz M (2010) Assisted outpatient treatment in New York: regional differences in New York’s assisted outpatient treatment program. Psychiatr Serv 61(10):970–975 38. Evans R, Makala J, Humphreys M, Chitra R (2010) Supervised community treatment in Birmingham and Solihull: first 6 months. Psychiatrist 34:330–333 39. Patel M, Matonhodze J, Baig M, Taylor D, Szmukler D, David A (2013) Naturalistic outcomes of community treatment orders: antipsychotic long-acting injections versus oral medication. J Psychopharmacol 27:629

123

Use of community treatment orders in an inner-London assertive outreach service.

Factors associated with the use of community treatment orders.

Canadian Studies on the Effectiveness of Community Treatment Orders.

Residence Conditions on Community Treatment Orders.

The epistemic challenges of CTOs: Commentary on . . . Community treatment orders.

Effectiveness of Community Treatment Orders: The International Evidence.

Community treatment orders in the UK 5 years on: a repeat national survey of psychiatrists.

Doubts About the Clinical Effectiveness of Community Treatment Orders.

Community treatment orders: background and implications of the OCTET trial.

A systematic review of community based hepatitis C treatment.

Community treatment orders in a Swedish county--applied as intended?

A systematic review of the effect of daily panty liner use on the vulvovaginal environment.

Community treatment orders: current evidence and the implications.

The OCTET trial, community treatment orders and evidence-based practice.

The effect of treatment timing on the management of facial fractures: a systematic review.

The effect of a geriatric evaluation on treatment decisions for older cancer patients--a systematic review.

Examining the use of metaphors to understand the experience of community treatment orders for patients and mental health workers.

Impact of Community-Based DOT on Tuberculosis Treatment Outcomes: A Systematic Review and Meta-Analysis.

Mental Health Service Use Among Immigrants in the United States: A Systematic Review.

Compassion Fatigue among Healthcare, Emergency and Community Service Workers: A Systematic Review.

Systematic review on the use of omalizumab for the treatment of asthmatic children and adolescents.

Effect of chin-cup treatment on the temporomandibular joint: a systematic review.

The effect of cross-sex hormonal treatment on gender dysphoria individuals' mental health: a systematic review.

A 10-year service evaluation of an assertive community treatment team: trends in hospital bed use.