Abstract
Mobile emergency services are considered to be an integral part of an effective mental health service. Their use is widespread, and despite a lack of evaluative evidence, it is commonly thought that mobile emergency services reduce hospital admissions [1, 2]. Several research methodologies have been employed to demonstrate that mobile emergency services influence hospitalization rates, including the comparison of hospitalization rates before and after the introduction of mobile emergency services, examining the effects of a mobile emergency treatment group only, and comparing hospitalization rates in geographical catchment areas with and without mobile emergency services. Such studies have identified the clinical and social characteristics of users of psychiatric emergency services [3–6] the factors influencing decision to admit to hospital [7, 8] and the impact of emergency services on hospitalization [9–13]. However, few studies have compared the effectiveness of mobile services with hospital-based emergency services in their capacity to reduce hospitalization. This study compared hospitalization rates between a mobile community-based component and a hospital-based component of an assessment and emergency service in suburban Adelaide.
Two of the Australian National Mental Health Strategy objectives for crisis services are [1]: to prevent unnecessary or inappropriate hospital admission; and [2] to facilitate referral to regional community-based services [14]. Mental health services within South Australia have attempted to meet both of these objectives. Assessment and crisis teams provide integrated 24 hour emergency response services that act as a ‘gateway’ into each region's mental health treatment services. The current study setting is the Western Regional assessment and crisis intervention service that comprises a mobile community-based site and a hospital-based site. Both sites provide assessment, crisis intervention and referral services, and are part of an integrated regional mental health service system which provides a range of communitybased and inpatient treatment services. As the need for emergency services is often determined by the availability of other services within the mental health service system, this study provided an opportunity to control for such variation as both mobile and hospital-based components had access to the same service system. Both sites facilitate referral to other service components and admission to the inpatient unit. The service aims to provide treatment in the least restrictive environment and to reduce the number of hospitalizations. Past studies have shown that admission to inpatient services is often inappropriate and preventable [15, 16]. In addition to comparing hospitalization rates, this study also aimed to identify the characteristics of mobile-service users who were more likely to be hospitalized.
Methods
Setting
The Western regional assessment and crisis intervention service (WACIS) has been in operation for approximately five years and serves a population of approximately 200 000 people in the western metropolitan region of Adelaide. The service consists of two components, a community-based mobile component that operates from 08.00 to 22.30 hours daily, and a hospital-based component which is part of the emergency department of the local 360-bed public general hospital which operates 24 hours a day. The hospital includes a 40-bed acute psychiatric inpatient unit. Consumers may access the WACIS service either by telephone, or presenting in person. A single emergency phone number will connect consumers to the community team component of WACIS between 08.00 and 22.30 hours, and to a combined regional triage service (CRT, serving all metropolitan areas) between 22.30 and 08.00 hours. Consumers may present in person at either the community site of WACIS between 09.00 am and 17.00 hours, or to the emergency department of the local hospital 24 hours a day.
The mobile team component comprises 13 staff including a full-time psychiatrist, three social workers, 9 mental health nurses and a halftime clinical psychologist rostered over 7 days. The team provides assessment, crisis intervention, community-based short-term treatment, home visiting, telephone triage service and facilitates hospital admission and referral. Treatment decisions, such as hospitalization, are made through consultation within the multidisciplinary staff of the team. The hospital-based component comprises a rostered psychiatric registrar who responds to requests from the emergency department to attend mental health contacts and a mental health nurse for 3 afternoons a week. The service provides assessment, crisis intervention, facilitates hospital admission and referral, and has access to the mobile team services. Treatment decisions are usually made singularly by the psychiatric registrar or in consultation with the community-based team. Both service components communicate regularly and have access to other mental health services in the region.
Design
The present study compared outcomes for consumers assessed by the community-based mobile component with those assessed by the hospital-based component of the assessment and crisis intervention service. A three-month continuous cohort of all new contacts for assessment at the mobile and hospital-based settings was included in the study.
Sample
Although some consumers walk into the emergency services, most contacts are referred via telephone triage. Thus face-to-face contacts are most frequently made with consumers deemed serious enough to warrant a face-to-face assessment. A total of 514 consumers had face to-face contact with the mobile (n = 304) and hospital-based (n = 210) emergency services during the study period. As emergency referrals specifically requesting hospital admission (e.g. police referrals) tend to present directly to hospital emergency departments [4, 8] and have a greater likelihood of hospitalization, all involuntary hospital admissions were excluded from the study. Following the exclusion of all involuntary hospitalizations the sample was reduced to 461, with 298 contacts with the mobile service (female 47.5%, mean age of 38.3, male mean age of 33.8) and 163 contacts with the hospital-based service (female 50%, mean age of 37.8 years, male mean age of 36.2 years).
Measures
The Health of the Nation Outcome Scales (HoNOS) [17] was completed on all new face-to-face contacts with the mobile and hospitalbased services. The HoNOS is a clinician rated assessment applicable to all psychiatric diagnoses, used to rate consumers in 12 symptom and problem areas. Each item is rated on a 5 point scale from 0 (no problem within the rating period) to 4 (severe to very severe), with a score of 2 506 A COMPARISON IN HOSPITALIZATION RATES or above indicating clinical significance. The HoNOS has been shown to have adequate psychometric properties and is suitable for use in a range of mental health treatment settings [18]. The HoNOS was completed by clinicians at each site within 24 hours of completing their assessment. All clinical staff are trained in the administration of the rating scale, as the HoNOS is part of the service-wide outcome measurement protocol and is completed on all new face-to-face contacts with the assessment and emergency services. The referral outcomes for all face-to-face contacts were documented as: no further contact required; short-term treatment by the emergency team; admission to hospital; referral to a community mental health team for case management; referral to another community agency; and referral to general medical practitioner or private psychiatrist. For the purposes of this study, referral outcomes were recoded as a dichotomous variable (admitted, not admitted). Provisional primary diagnostic, demographic details and time of contact was also obtained.
Results
Equivalence of consumer characteristics at each assessment site
There were no differences between hospital and communityassessed samples in the gender distribution (hospital-assessed 50% female, community-assessed 47.47% female, χ 2 (1) = 0.27, ns), mean age (F (1,446) = 1.29 ns), the proportion receiving case management (χ 2 (3) = 0.69 ns) nor type of primary provisional diagnosis (χ 2 (6) = 8.92 ns), although a greater number of additional diagnoses were given to consumers assessed in the community compared to the number given in hospital (t(459) = − 4.9, p < 0.001). Face-to-face contacts in both mobile and hospital-based emergency services tended to have more serious mental health disorders, with 58% of mobile community service contacts and 52% of hospital-based contacts assigned a provisional primary diagnosis of either a psychotic disorder (such as schizophrenia) or a major affective disorder [19].
Since some authors have queried the ability of total HoNOS scores to differentiate between consumer groups [18], comparisons between assessment sites used single item scores. Prior to determining assessment outcome, there were differences in HoNOS scores of consumers presenting for assessment at the hospital site compared to those presenting at the community site as indicated by Hotelling T 2 (F (12,433) = 7.27, p < 0.001). Scheffe test posthoc comparisons (p < 0.001) of HoNOS item scores revealed community-assessed patients were reported to have more severe cognitive problems and ‘other problems’, while hospital-assessed patients were reported to have more problems with occupation and activities of daily living.
Outcomes of emergency assessments
A total of 22.13% (n = 102) of voluntary face-to-face emergency contacts were admitted to inpatient services, 89% (n = 91) of these were admitted between the hours of 09.00 and 21.00, or within the hours of the mobile team's operation. There was a substantial difference in the admission rates between the two assessment sites (χ 2 (1) = 42.98, p < 0.001), with 43% (n = 64) of hospital-based contacts and 13% (n = 38) of mobile service contacts admitted to hospital.
Hospital admission as a function of consumer characteristics and assessment-site
A hierarchical logistic regression analysis was performed to predict assessment outcome (i.e. whether the consumer was admitted to hospital or not admitted). Table 1 lists each predictor in order of entry and χ 2 tests for the contribution each predictor makes as it is added to those entered previously. Neither age, nor gender improved the prediction of assessment outcome from a constant-only model (χ 2 (2) = 1.09, ns). However, principal diagnosis and the HoNOS items significantly improved the prediction of assessment outcome (χ 2 (13) = 97.44, p < 0.001). A model containing age, gender, principal diagnosis, and HoNOS items (– 2∗log-likelihood = 363.37, χ 2 (15) = 98.54, p < 0.001) accounted for approximately 31% of the variance in assessment outcome. Assessment-site substantially improved the prediction of assessment outcome, even after age, gender, principal diagnosis, and HoNOS item scores had been entered (log-likelihood = − 184.66, χ 2 (7) = 48.81, p < 0.001). Assessment-site uniquely contributed approximately 9% of the variance accounted for in assessment outcome. The model combining age, gender, principal diagnosis, HoNOS and assessment site accounted for approximately 40% of the variance in assessment outcome.
Logistic regression analysis of assessment outcome as a function of individual characteristics and assessment site
To further test the adequacy of the combined model, post hoc classification of consumers based on the odds ratio for the combined model including assessment site (OR = 10.85) were conducted. The combined model containing consumer characteristics and assessment site correctly classified 93.58% of those who were not admitted to hospital, and 42.4% of those who were admitted to hospital, yielding an overall correct classification rate of 81.7%. Clearly, the combination of age, gender, diagnosis, HoNOS and assessment-site was more successful at accounting for who would not be admitted to hospital than who would be admitted.
Clinical characteristics of consumers admitted to hospital
There were significant differences in HoNOS scores between consumers admitted to hospital and those not admitted (Hotelling T 2 = 104.887, F (12,433) = 8.5241 p < 0.001). Those admitted to hospital had higher scores for aggression/agitation (t(450) = 5.26, p < 0.001), non-accidental self-injury (t(448) = 3.71, p < 0.001), hallucinations and delusions (t(44) = 5.55, p < 0.001), problems with living conditions (t(450) = 3.97, p < 0.001) and problems with occupation and activities (t(450) = 5.53, p < 0.001). In addition, the distribution of principal provisional diagnoses varied marginally between consumers hospitalized and those not admitted (χ 2 (6) = 13.18, p < 0.05) mainly due to a greater proportion of hospitalized consumers receiving diagnoses of psychotic disorders (28.7% vs 18.6% those not admitted) and major affective disorders (41.6% vs 33.5% those not admitted).
Discussion
The present study found that contacts with a regional psychiatric assessment and crisis intervention service were more than three times as likely to be hospitalized if assessed by the hospital-based component of the emergency service than if assessed by the mobile communitybased component, regardless of presenting illness symptom acuity. This result is consistent with claims that mobile crisis services are associated with reductions in the rate of hospital admissions following initial contact.
However, whether mobile services are effective in impacting overall inpatient utilization is unclear from the current study. Several consumer characteristics may affect staff decisions to admit to hospital. For one, consumers with significant comorbid medical and psychiatric diagnoses may be more likely to be admitted irrespective of assessment site. Although the present study suggests that the presence of comorbid psychiatric diagnoses did not increase the likelihood of admission (since community-site staff gave more secondary diagnoses than hospital staff), comorbid medical conditions were not assessed in the present study and should be in future investigations. The consumer's level of contact with a service (whether they are presenting for the first time or are a ‘revolving-door’ client) may also influence whether they are admitted, although it is not clear whether level of contact exerts consistent effects on admission rates. Geller [8] found that while mobile emergency services were successful in preventing first admissions, they might not affect the rate of recidivist admissions or overall inpatient bed utilization. Furthermore, whether the consumer has accommodation, friends or social supports to assist during crisis may influence admission decisions. Consistent with this, the present results showed that consumers who had problems with living conditions and problems with occupation and activities were more likely to be hospitalized. However, future studies may need to include additional specific measures of accommodation type and level of social support to properly ascertain their influence on admission rates. Future studies could also look at the impact of individuals’ coping skills. For example, Schnyder [8] found that individuals’ coping strategies contributed more to the target variance in predicting hospitalization than did diagnosis.
Among the clinical characteristics of contacts more likely to be admitted to inpatient services were those being diagnosed with either schizophrenia or major affective disorder at the time of the assessment, and experiencing problems with hallucinations and delusions or deliberate self-injury. These factors have been found to be significant predictors of hospitalization in previous studies [7, 9]. For example, Guo et al.'s [9] study of contacts with an emergency service that used the HoNOS to rate the severity of problems, found that people with diagnoses of schizophrenia and major affective disorder, and who had primary problems with hallucinations and delusions and suicidal behaviour were more than twice as likely to be hospitalized than other consumers.
While clinical characteristics were an important factor in predicting hospital admission in the current study, non-clinical factors may also influence hospital admission rates. Although past studies have shown that hospitalization usually occurs within two days of initial presentation to a crisis service [9], future research could track contacts for several days following initial presentation to gain a more accurate assessment of their outcome. It has also been suggested from past research that a greater proportion of contacts using emergency services is hospitalized during the night-time period [8]. Given that the mobile team operated between 08.30 and 22.00 hours, all admissions to hospital out of these hours were facilitated through the hospital-based emergency service. As 89% of all admissions occurred between 09.00 and 21.00 hours, or during the operational hours of the mobile team, most of these contacts had access to the mobile service.
An area worthy of further investigation is the identification of the characteristics of a mobile service that contribute to the reduction of hospitalization rates. As the composition and capacity of the community-based mobile service differed significantly from the hospitalbased service, one cannot conclude that the location of the service alone was important in reducing hospital 508 A COMPARISON IN HOSPITALIZATION RATES admissions. The community-based service included a multidisciplinary, specialized team that provided an assessment and follow-up home visiting service, and had greater familiarity with other components of the regional mental health system. The hospital-based service often comprised a single clinician operating in relative isolation, without the support of a specialized team. Mobile team staff report that home-visiting consumers in their own environment enables them to more accurately assess the strengths and available resources that the consumer can use to deal with their current problem situation. Their capacity for follow-up home visits to provide treatment and support, their specialized team approach and access to community-based services appear to be important factors in diverting more emergency contacts to community-based treatment. However, other factors may also be responsible for the difference in admission rates between assessment sites. While the impact of service location alone on hospitalization rate was not investigated in this study, it should be considered in future research. The context in which assessments occur, and the relative proximity of inpatient facilities in hospital-based emergency settings, may influence clinical decision-making.
This study supports the view that mobile crisis services increase access for consumers to community-based services and reduce the need for admission to inpatient services. However, given the limitations of the resources available to the hospital-based emergency service, it was not surprising that hospitalization rates differed so significantly. Although consumers and relatives have reported greater satisfaction with mobile services [2, 10], both services appear necessary components of an integrated psychiatric emergency service. Some studies have reported that mobile services see a qualitatively different population than hospital-based services [5]. However, this study found little difference in the diagnostic, symptom acuity or demographic characteristics between the two groups. The hospital-based component provides access to consumers in a mainstreamed general hospital setting, whilst the mobile service is available to those preferring to be seen in their home setting, and those who are unwilling or have difficulty accessing hospitalbased emergency services. Past studies have shown that a significant proportion of individuals seen by a mobile service were more likely to have refused to go to a hospital-based emergency service and did not realize they needed help [5]. A mobile capacity and specialized team approach appear to be the crucial elements of an effective psychiatric emergency service if consumers are to access treatment in the least restrictive environment and avoid hospitalization.