by Sjoerd Sytema, Philip Burgess, and Michele Tansella Abstract

Transcription

1 Does Community Care Decrease Length of Stay and Risk of Rehospitalization in New Patients With Schizophrenia Disorders? A Comparative Case Register Study in Groningen, the Netherlands; Victoria, Australia; and, Italy by Sjoerd Sytema, Philip Burgess, and Michele Tansella Abstract This was a long-term foilowup study of defined cohorts of new patients who met ICD-0 criteria for schizophrenia and related disorders. Three mental health service systems with psychiatric case registers were compared. In two areas (Victoria, Australia, and, Italy), a comprehensive system of community-based psychiatric care was developed, with a substantial reduction of hospital beds (to 0.27 beds/,000 people). In the third area (Groningen, the Netherlands), despite the presence of community psychiatric services, mental health care was still mainly hospital based (.6 beds/,000 people). Two hypotheses were tested: () the length of stay is shorter in a community-based system than in a hospital-based system; and (2) the risk of rehospitalization is independent of the characteristics of the mental health system, so that risk is expected to be similar in areas with communitybased and hospital-based systems. The first hypothesis was confirmed. The risk of rehospitalization, however, was lower in Victoria than in the two other areas, which might show the potential impact of community care in reducing rehospitalization. Keywords: Schizophrenia, psychiatric case register, readmission, community mental health services, longitudinal studies. Schizophrenia Bulletin, 28(2):273-28,2002. The relationship between community-based psychiatric care and use of hospital beds has been studied in various ways. Early outcome studies conducted in patients with schizophrenia followed long term after their first admission have been summarized by Harding (988). The main finding was the heterogeneity in long-term outcome. More recently, there have been some large-scale longitudinal studies, also mostly conducted with schizophrenia patients, that used data from inpatient psychiatric case registers to investigate the probability of rehospitalization (Eaton et al. 992a, 992fc; Haro et al. 994; Mortensen and Eaton 994; Korkeila et al. 998). These studies found that younger age coincides with higher readmission rates, but this effect seems to disappear when all readmissions over many s of foilowup are considered. Moreover, variables related to the patient's psychiatric history (number of previous admissions, duration of the latest discharge, duration of the last admission) predicted readmission risk. No relationships with service system variables were included in any of these studies. A large U.K. study by the Team for the Assessment of Psychiatric Services investigated the closure of two mental hospitals. Of 523 discharged long-stay patients who survived the 5- foilowup period, a third were readmitted at least once, but about 90 percent were still living in the community at the end of foilowup, mainly in staffed residential homes (Trieman et al. 999). Others studies compared rehospitalization rates of chronic mentally ill patients in different service systems. For example, the Nordic Comparative Study on Sectorized Psychiatry involved a - foilowup of - incident cases in seven psychiatric sectors in four Nordic countries (Oiesvold et al. 2000). The relationship between age of onset and rehospitalization risk was confirmed. This relationship was found both across participating sectors and across diagnostic groups. Readmission risk showed considerable variation between sectors but Send reprint requests to Dr. S. Sytema, Department of Psychiatry, University Hospital Groningen, P.O. Box 3000, 9700RB Groningen, The Netherlands; s.sytema@med.rug.nl. 273 Downloaded from on 2 January 208

2 Schizophrenia Bulletin, Vol. 28, No. 2, 2002 S. Sytema et al. was unrelated to available beds. The followup period, however, was very short, at no more than. Fisher et al. (992) compared readmission rates in five regions in Massachusetts. One of the regions clearly differed from the others in having a higher level of community facilities and a lower funding of inpatient care. However, no differences in readmission rates were found. Comparative studies of the Groningen (the Netherlands) case register with both the (Italy) (Sytema et al. 997) and the Victoria (Australia) case registers (Sytema and Burgess 999) investigated continuity of care and readmission risk for schizophrenia and related disorders. A higher level of continuity of care was found in the community-based systems in Victoria and, compared with the more hospital-oriented service system in Groningen. The readmission risk after the index admission was not significantly different. The aim of this study, conducted in cohorts of new patients with schizophrenia and related disorders from three different mental health service systems, was to compare the length of hospitalization with the frequency. The three service systems have a comparative supply of community-based psychiatric care, but Groningen has far more mental hospital beds and therefore a more hospitalbased orientation. The systems were selected because of the availability of long-term, comparable, and comprehensive case registers suitable for research purposes. The case registers used in this study included both inpatient and out- and day-patient contacts, as compared with inpatient registers, in which only inpatient episodes are included. So, in the present study, inpatient care was assessed in relation to outpatient care and day care provided in the same period of time. Length of stay depends on not only the number of beds available in the area but also the possibility to refer the patients discharged from the hospital to community services for further care. This followup is particularly important for patients with schizophrenia and related disorders, who need continuity of care over prolonged periods of time (Moreno-Kustner et al. 200). The first hypothesis of this study was that patients admitted to the hospital in areas with a highly community-based system, as compared with patients admitted in areas with a more hospital-based system, would be discharged earlier, to be further treated in community settings. It has been suggested that a relapse of psychosis or suicidal behavior requires hospitalization in most cases, irrespective of the type of mental health system in the area (Fisher et al. 992). However, there is some evidence that a special type of community care namely, assertive community treatment (an intensive team-based approach for the severely mentally ill) reduces readmission risk compared to standard community care and compared to hospital-based rehabilitation services (Marshall and Lockwood 200). But another model of community care namely, case management (a means of coordinating care for severely mentally ill patients in the community) seems to increase admission rates to psychiatric hospitals (Marshall et al. 200). The present study was related to routine daily practice, which is measured by the actual number of contacts provided irrespective of type or model of community care. We expected that the decision to rehospitalize a patient with schizophrenia would be more closely related to the clinical course of the illness than to other service level variables. The second hypothesis of this study was therefore that schizophrenia patients' risk of being rehospitalized would be similar in community-based and in hospital-based systems of care and would be independent of the characteristics of the total number of beds available in the area. Methods and Material Mental Health Systems in the Case Register Areas. The Groningen register covers a mental health region in the Netherlands with 460,000 inhabitants. In this country, the total number of psychiatric beds decreased only slightly from 84 to 6 per 00,000 between 986 and 998. More than 90 percent of these beds are located in mental hospitals. In 998, 56 percent of total service expenditure was used by these general mental hospitals (which also include outpatient and day-patient facilities). Service expenditure by nonhospital services (which include outpatient and day-patient facilities located both in and outside mental hospitals, general-hospital psychiatric units, and sheltered homes in the community) was about 36 percent of total expenditures. These proportions were unchanged from 994 levels (Have ten et al. 996; Ministerie van Volksgezondheid, Welzijn en Sport 2000). The Victoria register covers the state of Victoria, which has 4.5 million inhabitants. In 992, all states and territories of Australia adopted the National Mental Health Strategy, which gives directions for reform of mental health services (Commonwealth of Australia 996; Mental Health Branch 997). The key features of that strategy are a change from centralized services to more accessible local services. In Victoria, the total number of beds was reduced from 42.3 to 26.6 per 00,000 between 992 and 996. In , psychiatric hospitals used 54.9 percent of total service expenditure, which decreased to 25.2 percent in On the other hand, in this 4- period, the proportion of service expenditure by nonhospital services increased from 3.8 percent to 54.2 percent. These indicators illustrate the major shift from hospital-based to community-based care in Victoria within a relatively short period of time. 274 Downloaded from on 2 January 208

3 Influence of Community Care in New Patients With Schizophrenia Schizophrenia Bulletin, Vol. 28, No. 2, 2002 The register covers the southern region of Verona, which has 75,000 inhabitants. Reform of Italian psychiatry accelerated after a law was passed by the parliament in 978. This law prescribed the development of community-based services for defined catchment areas. These services were to include day care and outpatient facilities and sheltered homes in the community for chronic patients, as well as small general-hospital psychiatric units (with no more than 5 beds) and were intended to ensure continuity of care for all residents in the area. Admissions to state mental hospitals have been forbidden since 982. In the new structure was successfully implemented, although this is not representative of Italy as a whole (Tansella et al. 987). The main service in the area is the Community Psychiatric Service, in which over 80 percent of those receiving psychiatric care are treated. It includes a community mental health center providing day care, rehabilitation and home visits, a general-hospital psychiatric unit with 5 beds (0.20 per,000 residents), sheltered apartments, outpatient services, and a 24-hour crisis intervention service. Multidisciplinary teams (coordinated by a psychiatrist) work both in community settings and in the hospital. There are three such teams; each is responsible for a subsector of the area. The same consultant psychiatrist directs both the hospital unit and the outpatient and community services. Every effort is made to ensure prompt followup and community care for all discharged patients, as well as availability of hospital beds for those treated in the community who need to be admitted (Tansella 99; Tansella and Micciolo 998). Sample. Patients with a lifetime clinical diagnosis of "schizophrenia and related disorders" (comprising the ICD-0 codes F20, F2, F22, F23, F24, F25, and F29) were selected. The cohorts were then restricted to those patients first registered in the psychiatric case registers in the s in the Groningen and Victoria case registers and in the s in, reflecting the maximum time period for which comparable data were available. The patients were followed from their first contact until the end of 996 in and until the end of 997 in the other two areas. Only those patients with at least one episode of hospitalization in this observation period were included. Procedure. To investigate differences among the three areas, Cox's regression (Cox 972) was used to analyze the length of stay (number of days between admission and discharge) and the time until readmission (number of days between discharge and the next admission) on the combined data sets. If an admission did not end before the end of observation, the number of days between admission and the end of observation was used to calculate length of stay. Such episodes were considered as censored observations in the Cox's regression analysis on length of stay. If there was no further readmission within the observation period, the number of days from last discharge until the end of observation was used. These episodes were considered as censored observations in the Cox's regression analysis of readmission risk. The coefficients resulting from Cox's regression analysis are hazard ratios (HRs). The first category from each variable is the reference category. An HR greater than means that the risk of experiencing the terminating event (in our case, a discharge or a readmission) is higher than the reference category. In standard Cox's regression, all observations included in the analysis are considered independent. However, in our case, the observations clustered by patient, as each patient generated a variable number of hospitalizations. Therefore, multiple episodes per patient should not be considered as independent from each other. Not taking into account dependence between the observations does not affect the estimate of the coefficients, but their variance will be biased. We used the method of Lin and Wei (989) implemented in the statistical software package Stata, Release 5 (997), which takes this clustering effect into account in estimating the variance of the estimated coefficients. Moreover, because readmission risk might be different after each adjacent readmission, we controlled for the number of previous admissions. We further studied readmission risk and its interrelationships with other care consumption indexes on each of the three data sets separately. Cox's regression was applied to analyze the time until readmission in the way stated above. Results Table provides an overview of resources available for the three cohorts in this study. In Victoria, beds in private hospitals (0.07/00,000) were not included, as these services do not provide data to the case register. These hospitals, however, play a relatively limited role in the treatment of psychoses. Some beds of the private hospitals in (see the table notes for more information) that provide data to the case register were included. The data show that the provision of community services was at a comparable level in the three areas, but the supply of beds was by far greater in the Groningen register area. Table 2 shows the age distribution of the cohorts. The Victoria cohort had the highest proportion of patients in the youngest age group (0-25 s). The cohort had a relatively high proportion of patients above 45 s of age. There was no significant difference in mean age between the Groningen and cohorts. The mean age of the Victoria cohort, however, Downloaded from on 2 January

4 Schizophrenia Bulletin, Vol. 28, No. 2, 2002 S. Sytema et al. Table. Public mental health resources in Groningen, the Netherlands; Victoria, Australia; and South- Verona, Italy, case register areas Population (994) Psychiatric beds available In mental hospital In general hospital Beds in residential homes 5 Clinical staff Total Inpatient Outpatient Groningen Victoria ,000.60/,000.40/, /, /,000 62/00,000 32/00,000 30/ ,500, /, /,000 0./, /,000 83/00,000 52/00,000 3/00,000 75, /, /, /, /,000 80/ Estimates based on national figures; excludes specialized services for children and for dependency and psychogeriatric nursing homes. 2 Estimates , public mental health services only. 3 Estimates 995. Includes psychiatric beds in public and private hospitals. 4 In the city of Verona, apart from 5 beds in a general hospital psychiatric unit (0.20/,000) dedicated to residents, other psychiatric beds (total 220) are available in private hospitals where patients from a larger area in northeastern Italy can be admitted at the expense of the National Health Service. During the study period, /,000 of these private beds were occupied each day by South- Verona residents. Admissions to these beds are reported to the Psychiatric Case Register and were included in the analyses. 5 Staffed residential homes in the community. 6 Only 4 ward nurses worked for inpatients only (8.7/00,000). Most staff ( psychiatrists, 5 psychologists, 2 social workers, 3 health visitors 28/00,000) worked for both in- and outpatients (single-staff module). Moreover, there were 3 community nurses, 0 hostel support care providers, and 2 hostel counselors (33.3/00,000) working for outpatients only. 7 Including the Regional Institutes of Outpatient Care and outpatient services in mental hospitals and general hospitals. Table 2. Age distribution of the three cohorts Age >45 Total Mean age (SD) n (6.4) Groningen % % male n 3,585 2,693,487 2,32 0, (7.4) Victoria % % male n % (6.0) % male Note. SD = standard deviation. One-way analysis of variance: Groningen vs. Victoria (p = 0.00); Groningen vs. (p 0.309, not significant); Victoria vs. (p = 0.005) was significantly lower than the mean age of the other areas' cohorts, although this difference was only 2 s between the Groningen and Victoria cohorts and 3.4 s between the and Victoria cohorts. Table 2 also shows the proportion of males in each age category. In each cohort, the proportion of males was highest in the youngest age category and decreased with older age. After about 45 s of age, females outnumbered males. This pattern is in accordance with the wellknown fact that the age of onset of schizophrenia is generally earlier for males than for females. The total proportion of males was about equal in Groningen and South- Verona (49%) but was significantly higher in Victoria (57%). This is partly explained by the higher proportion of patients in Victoria's youngest age category, a category dominated by male patients. But, as the table shows, the proportion of males in the Victoria cohort was higher in each age category compared to the corresponding proportion in the other areas, which also partly explains the difference. A logistic regression analysis (data not shown) 276 Downloaded from on 2 January 208

5 Influence of Community Care in New Patients With Schizophrenia Schizophrenia Bulletin, Vol. 28, No. 2, 2002 with the male/female ratio as the dependent variable and age and area as the explanatory variables revealed that both factors contribute significantly to the difference. Service Utilization. Table 3 shows a number of service utilization indicators. The total number of days of inpatient treatment per and per patient was considerably higher in the Groningen area than in the other two areas. The mean and median number of readmissions were about the same in Groningen and. The mean number of readmissions was significantly lower in Victoria, although the median was slightly higher. Day treatment was at the lowest level in Victoria. The number of outpatient contacts, on the other hand, was highest in Victoria. Length of Stay and Risk of Readmission to the Hospital. Table 4 shows the differences between the areas in the length of stay (risk of discharge) and in the time from discharge to readmission (risk of readmission), using Groningen as the reference category. Victoria and had a significantly shorter length of stay (higher risk of discharge) than Groningen. According to the second dependent variable, the time from discharge to readmission, no difference was found between Groningen and, but the time to readmission was significantly longer in Victoria (lower risk of readmission). The relationships between readmission risk and age at first contact and between readmission risk and other service utilization characteristics were further analyzed separately for each of the three cohorts. Table 5 shows a decreasing risk of readmission for increasing age categories in Victoria and but not in Groningen. In all areas, a lower readmission rate was associated with a longer duration of the last admission and with a longer episode of time between the last two admissions. Having day-patient care during the followup period was associated with an increased readmission risk in all areas. The same held for patients with the highest number of outpatient contacts, except for the Groningen patients. There was a small but significant tendency for the risk of readmission to decrease after each additional admission, controlled for the total number of admissions during followup. Without controlling for the latter variable, the opposite would have been found. In this we followed Eaton et al. (992a) in dealing with the effect of each adjacent admission on readmission risk. This was potentially biased by the fact that patients with higher numbers of admissions had to squeeze them into the same time interval and therefore by definition would have increased readmission risks. Adding the total number of admissions during followup into the analysis could neutralize this bias. Table 3. Care utilization in the three areas Number of days in inpatient treatment Number of days in day-patient treatment Number of outpatient contacts Number of readmissions (since discharge first admission) after Mean/ (SD) 62 (88) 5 (25) 0 (6) 0.45 (0.98) Groningen Median/ Mean/ (SD) 9 (5) 3 (7) 6 (4) 0.29 (0.48) Note ANOVA = analysis of variance; SD = standard deviation. Victoria Median/ Mean/ (SD) 6 (0) 7 (22) 7 (0) 0.44 (0.84) Median/ Significance of difference (ANOVA) Groningen vs. Victoria, Groningen vs. Groningen vs. Victoria, Victoria vs. Groningen vs. Victoria, Victoria vs. Groningen vs. Victoria, Victoria vs. 277 Downloaded from on 2 January 208

6 Schizophrenia Bulletin, Vol. 28, No. 2, 2002 S. Sytema et al. Table 4. Risk of discharge from the hospital and risk of readmission to the hospital on the combined Groningen, Victoria, and data sets, adjusted for the number of previous admissions (Cox's regression) Area Groningen Victoria Number of observations Percentage of censored observations Risk of discharge (time from admission to discharge), hazard ratio (95% Cl).49(.4-.57) 2.8( ) 3,60 Risk of readmission (time from discharge to readmission), hazard ratio (95% Cl) 0.86( ).00 (0.9-.0) 3, Note. Cl = confidence interval. A hazard ratio greater than means that the risk of experiencing the terminating event (the discharge from or the readmission to the hospital) is higher than the reference category (i.e., in Groningen). "Revolving Door" and "New Long-Stay" Patients. "Revolving door" patients were those admitted at least four times by the end of the observation period, with a mean number of at least two admissions per. Twenty patients in Groningen (2.%), 9 patients in Victoria (0.9%), and 7 patients in (3.3%) fulfilled these criteria. "New long-stay" patients were those who at the end of observation had been continuously in the hospital for at least 2 s. In Groningen 35 (3.7%) and in Victoria 72 (0.7%) new long-stay patients were found, and none were found in. In Groningen, another 36 patients (3.8%) were living in staffed residential homes in the community at the end of the observation period, 2 (.3%) of them for at least 2 s. In Victoria, 57 patients (0.6%) stayed in such a home at the end of observation, 22 of them (0.2%) for more than 2 s. In, where sheltered accommodation is generally provided in hostels for only relatively short periods of time and every effort is made to transfer patients to apartments where they can live with some support and supervision, only two patients (%) had stayed in these facilities for more than 2 s at the end of observation. Discussion One of the major objections to community-based psychiatric care has been that prolonged hospitalization is replaced with more frequent rehospitalization and by a revolving door pattern of hospital care that benefits neither patients nor their families. Such problems have arisen mainly in underdeveloped community care systems where there is a lack of continuity of care and of liaison between hospital and community services (Bachrach 996). On the other hand, there is some evidence that an intensive type of community care in other words, assertive community treatment is able to reduce rehospitalization (Marshall et al. 200). The three psychiatric service systems we compared in this study do not differ in their level of care in the community, but one area (Groningen) has a far greater supply of beds. Mental health care development in the Netherlands is unusual in that complementary community services have been established without concomitant reduction of the number of psychiatric beds in the hospital. By contrast, in Victoria and, alternative community-based facilities ensuring liaison with hospital facilities and continuity of care have been established, with a parallel and substantial reduction of hospital beds. In addition, in, Italian legal changes since 982 have prohibited any admissions to mental hospitals, so only general hospital and private hospital beds are available for admissions, both voluntary and compulsory. The present study focused on new patients with schizophrenia disorders who were entering the psychiatric case register for the first time, rather than on long-stay patients moving to community settings. Patients at their first psychiatric contact are a more homogeneous group with less severe illness compared with longer-term patients, whose further pattern of care with services may be biased by their previous psychiatric history. The results of this study cannot therefore be extrapolated to all patients with schizophrenia. A further limitation of this study should be kept in mind in interpreting the results: Severity of illness could not be controlled for. The diagnoses were case register clinical diagnoses made by different clinicians, and dieir reliability may be questioned. However, a multicenter 278 Downloaded from on 2 January 208

7 Influence of Community Care in New Patients With Schizophrenia Schizophrenia Bulletin, Vol. 28, No. 2, 2002 Table 5. Risk of readmission to the hospital of patients with at least one readmission and with at least 90 days of observation after the last discharge (Cox's proportional hazard models) Age at first contact >45 Duration of last admission (quartiles) Duration of previous episode (i.e., time between admissions [quartiles]) Day-patient care during followup No Yes Mean number of ly outpatient contacts (quartiles) Total number of hospitalizations during followup Each additional admission Total number of readmissions Percentage of censored observations ns Groningen Hazard ratio (95% Cl) 0.90( ) 0.70 ( ) 0.57 ( ) 0.99( ) 0.93( ) 0.78 ( ).30(.0-.52) ns.08(.07-.0) 0.96 ( ), Note. Cl = confidence interval; ns = not significant. Victoria Hazard ratio (95% Cl) 0.89 ( ) 0.82 ( ) 0.74 ( ) 0.96( ) 0.85 ( ) 0.74 ( ) 0.96(0.9-.0) 0.84 ( ) 0.72 ( ). (.06-.6).50 ( ).9 ( ) 2.20 ( ).08( ) 0.96 ( ) 7, Hazard ratio (95% Cl) 0.7 ( ) 0.59 ( ) 0.68 ( ) 0.76 ( ) 0.64 ( ) 0.67 ( ).07( ) 0.87(0.68-.) 0.72 ( ).30(.0-.79).6( ).24( ).50( ).02(.0-.02).00( ) case register study (which included both the Groningen and registers) has shown a satisfactory interrater agreement of this grouped diagnostic category (kappa = 0.79) (Sytema et al. 989). Our first hypothesis was that the duration of admission to the hospital is a service system-related phenomenon, longer stay being associated with a more hospitaloriented treatment philosophy, as indicated by a relatively high number of hospital beds available and by a less structured liaison with the psychiatric facilities in the community. This hypothesis was confirmed: The length of stay of each admission, as expected, was longer in Groningen, where more beds are available. Nonetheless, only a very few patients exhibited a long-stay pattern of care in traditional mental hospitals. Most of these patients were transferred to houses for small groups within the same mental hospital site. We further hypothesized that the risk of readmission would not be different in service systems with a strong focus on community care (Victoria and ), compared with a largely hospital-based system (Groningen). This hypothesis was partly confirmed. We did not find a significant difference in risk of readmission between and Groningen, but the readmission risk was lower in Victoria. In our previous study, we did not find a difference between Groningen and Victoria, but that study was restricted to readmission risk after one index admission (Sytema and Burgess 999). The lower readmis- 279 Downloaded from on 2 January 208

8 Schizophrenia Bulletin, Vol. 28, No. 2, 2002 S. Sytema et al. sion risk in Victoria might be explained by the more intensive community care that the Victoria patients received. They consumed about twice the amount of contacts compared with the Groningen and cohorts. Because the readmission risk was not found to be lower in the area with the hospital-based system (Groningen), we may conclude that highly community-based and deinstitutionalized mental health systems, such as those in Victoria and, do not create a higher risk of a revolving door pattern of hospital utilization. In fact, in the two community-based services, only a small proportion of patients, varying from percent (Victoria) to 3 percent (South- Verona) were revolving door patients. We found in each area that longer admissions are associated with a lower risk of readmission, which accords with previous studies (e.g., Appleby et al. 993; Mortensen and Eaton 994). One might conclude from this that short admissions should be avoided in order to lower the risk of readmission. However, this conclusion is too simplistic. It is more likely that different patterns of hospital utilization reflect different courses of illness. Many studies have shown that the risk of readmission increases with more previous admissions (Thornicroft et al. 992; Appleby et al. 993; Haro et al. 994; Hoffmann 994; Mortensen and Eaton 994; Daniels et al. 998; Doering et al. 998). We found the same result if we did not control for the total number of admissions during followup, which none of these earlier studies did. But when we did control for it, a significant but very small opposite effect emerged, consistent with but considerably weaker than reported by Eaton et al. (992a). The interpretation of this finding is that patients with many admissions have a higher readmission risk than patients with few admissions. However, within each of these groups, this risk showed a tendency to decrease rather than to increase after each additional admission. The finding in some studies that younger age of onset was associated with a higher risk of readmission was partially confirmed. A negative relationship between older age and readmission risk was found in Victoria and in. Eaton and colleagues found a strong relationship with age when they studied rehospitalization risk after the first discharge (Eaton et al. 9926) but not when they included all rehospitalizations (Eaton et al. 992a). The same trend was evident in the study of Mortensen and Eaton (994), where the effect of age disappeared after the fifth discharge. Our study found that a higher readmission risk was associated with a higher use of outpatient care in two areas (Victoria and ) and with the use of day-patient care in all areas. In other words, high users of inpatient facilities tended also to be high users of community services. So there was no evidence that a high risk of readmission was associated with a low uptake of community resources. Finally, in all three areas, the new long-stay pattern of care had almost disappeared, but that did not lead to a substantial number of revolving door patients. References Appleby, L.; Desai, P.N.; Luchins, D.J.; Gibbons, R.D.; and Hedeker, D.R. Length of stay and recidivism in schizophrenia: A study of public psychiatric hospital patients. American Journal of Psychiatry, 50:72-76, 993. Bachrach, L.L. Deinstitutionalisation: Promises, problems, and prospects. In: Knudsen, H.C., and Thornicroft, G., eds. Mental Health Service Evaluation. Cambridge, U.K.: Cambridge University Press, 996. pp Commonwealth of Australia. National Mental Health Report 995. Third Annual Report: Changes in Australia's Mental Health Services in Year Three of the National Mental Health Strategy. Canberra, Australia: Commonwealth Department of Health and Family Services, 996. Cox, D.R. Regression models and life tables. Journal of the Royal Statistical Society, 34:87-202, 972. Daniels, B.A.; Kirkby, K.C.; Hay, D.A.; Mowry, B.J.; and Jones, I.H. Predictability of rehospitalization over 5 s for schizophrenia, bipolar disorder and depression. Australian and New Zealand Journal of Psychiatry, 32:28-286, 998. Doering, S.; Muller, E.; Kopcke, W.; Pietzcker, A.; Gaebel, W.; Linden, M.; Muller, P.; Muller-Spahn, F.; Tegeler, J.; and Schussler, G. Predictors of relapse and rehospitalization in schizophrenia and schizoaffective disorder. Schizophrenia Bulletin, 24(l):87-98, 998. Eaton, W.W.; Bilker, W.; Haro, J.M.; Herrman, H.; Mortensen, P.B.; Freeman, H.; and Burgess, P. Long-term course of hospitalization for schizophrenia: II. Change with passage of time. Schizophrenia Bulletin, 8(2):229-24, 992a. Eaton, W.W.; Mortensen, P.B.; Herrman, H.; Freeman, H.; Bilker, W.; Burgess, P.; and Wooff, K. Long-term course of hospitalization for schizophrenia: I. Risk for rehospitalization. Schizophrenia Bulletin, 8(2):27-228, Fisher, W.H.; Geller, J.L.; Altaffer, F.; and Bennett, M.B. The relationship between community resources and state hospital recidivism. American Journal of Psychiatry, 49: , 992. Harding, CM. Course types in schizophrenia: An analysis of European and American studies. Schizophrenia Bulletin, 4(4): , 988. Haro, J.M.; Eaton, W.W.; Bilker, W.B.; and Mortensen, P.B. Predictability of rehospitalization for schizophrenia. 280 Downloaded from on 2 January 208

9 Influence of Community Care in New Patients With Schizophrenia Schizophrenia Bulletin, Vol. 28, No. 2, 2002 European Archives of Psychiatry and Clinical Neuroscience, 244:24-246, 994. Have ten, M.L.; Lorsheyd, J.J.G.; Bijl, R.V.; and Osterhun, P. Jaarboek geestelijke gezondheidszorg 996/997. Utrecht, The Netherlands: De Tijdstroom/NcGv/NZi, 996. Hoffmann, H. Age and other factors relevant to the rehospitalization of schizophrenic outpatients. Acta Psychiatrica Scandinavica, 89:205-20, 994. Korkeila, J.A.; Lehtinen, V.; Tuori, T.; and Helenius, H. Frequently hospitalized psychiatric patients: A study of predictive factors. Social Psychiatry and Psychiatric Epidemiology, 33: , 998. Lin, D.Y., and Wei, L.J. The robust inference for the Cox proportional hazards model. Journal of the American Statistical Association, 84: , 989. Marshall, M.; Gray, A.; Lockwood, A.; and Green, R. Case management for people with severe mental disorders (Cochrane review). In: The Cochrane Library, issue 3. Oxford, U.K.: Update Software, 200. Marshall, M., and Lockwood, A. Assertive community treatment for people with severe mental disorders (Cochrane review). In: The Cochrane Library, issue 3. Oxford, U.K.: Update Software, 200. Mental Health Branch. National Mental Health Report 996: Fourth Annual Report, Changes in Australia's Mental Health Services Under the National Mental Health Strategy Canberra, Australia: Commonwealth Department of Health and Family Services, 997. Ministry of Health and Sports. Document on Care Facilities 200. The Hague, the Netherlands: Sdu Publishers, Moreno-Kiistner, B.; Torres Gonzales, F.; and Luna del Castillo, J.D. Analysis of patterns of mental health care with three different approaches (cross-sectional, longitudinal and dynamic). Epidemiologia e Psichiatria Sociale, 0:82-89,200. Mortensen, P.B., and Eaton, W.W. Predictors for readmission risk in schizophrenia. Psychological Medicine, 24: , 994. Oiesvold, T.; Saarento, O.; Sytema, S.; Vinding, H.; Gostas, G.; Lonnerberg, O.; Muus, S.; Sandlund, M.; and Hansson, L. Predictors for readmission risk of new patients: The Nordic Comparative Study on Sectorized Psychiatry. Acta Psychiatrica Scandinavica, 0: ,2000. Stata Corp. Stata Statistical Software: Release 5.0. College Station, TX: Stata Corporation, 997. Sytema, S., and Burgess, P. Continuity of care and readmission in two service systems: A comparative Victoria and Groningen case-register study. Acta Scandinavica, 00:22-29, 999. Psychiatrica Sytema, S.; Giel, R.; ten Horn, G.H.; Balestrieri, M.; and Davies, N. The reliability of diagnostic coding in psychiatric case registers. Psychological Medicine, 9: , 989. Sytema, S.; Micciolo, R.; and Tansella, M. Continuity of care for patients with schizophrenia and related disorders: A comparative and Groningen caseregister study. Psychological Medicine, 27: , 997. Tansella, M., ed. Community-Based Psychiatry: Long- Term Patterns of Care in. Psychological Medicine Monograph Supplement 9. Cambridge, U.K.: Cambridge University Press, 99. Tansella, M., and Micciolo, R. Unplanned first contact as a predictor of future intensive use of mental health services. Social Psychiatry and Psychiatric Epidemiology, 33:74-80, 998. Thornicroft, G.; Gosh, C; and Dayton, D. Readmission to hospital for long-term psychiatric patients after discharge to the community. British Medical Journal, 305: , 992. Trieman, N.; Leff, J.; and Gloves, G. Outcome of long stay psychiatric patients resettled in the community: Prospective cohort study. British Medical Journal, 39:3-6,999. Acknowledgments We are most grateful to Giuliano Meneghelli (South- Verona Psychiatric Case Register) for collecting service utilization data for. This study was partially supported by the Fondazione Cassa di Risparmio di Verona Vicenza Belluno e Ancona, Progetto Sanita' , Grant "The Role of Social Factors on Onset, Course and Outcome of Mental Disorders: An Epidemiological Approach" to Professor Michele Tansella. The Authors Sjoerd Sytema, Ph.D., is Senior Research Sociologist, Department of Psychiatry, University of Groningen, Groningen, the Netherlands; Philip Burgess, Ph.D., is Associate Professor, Policy and Analysis Group, Mental Health Research Institute of Victoria, Victoria, Australia; Michele Tansella, Ph.D., is Professor of Psychiatry, Department of Medicine and Public Health, Section of Psychiatry, University of Verona, Verona, Italy. 28 Downloaded from on 2 January 208

10 Downloaded from on 2 January 208