The cost of relapse in patients with schizophrenia in the European SOHO (Schizophrenia Outpatient Health Outcomes) study

Progress in Neuro-Psychopharmacology & Biological Psychiatry 33 (2009) 835–841

Contents lists available at ScienceDirect

Progress in Neuro-Psychopharmacology & Biological Psychiatry j o u r n a l h o m e p a g e : w w w. e l s ev i e r. c o m / l o c a t e / p n p

The cost of relapse in patients with schizophrenia in the European SOHO (Schizophrenia Outpatient Health Outcomes) study Jihyung Hong a,b,⁎, Frank Windmeijer c, Diego Novick b, Josep Maria Haro d, Jacqueline Brown b a

LSE Health, London School of Economics, London, UK European Health Outcomes Research, Eli Lilly and Company, Windlesham, Surrey, UK Department of Economics, University of Bristol, Bristol, UK d Sant Joan de Deu-SSM, CIBERSAM, Sant Boi, Barcelona, Spain b c

a r t i c l e

i n f o

Article history: Received 1 January 2009 Received in revised form 27 March 2009 Accepted 29 March 2009 Available online 5 April 2009 Keywords: Cost Relapse Schizophrenia

a b s t r a c t Background: Relapse in schizophrenia is one of the greatest burdens of the illness. Aims: To estimate the costs associated with relapse in a pan-European naturalistic setting. Method: The SOHO study is a 3-year, prospective, observational study of 10,972 outpatients with schizophrenia across 10 European countries. The cost of resource use (inpatient stay, day care, psychiatrist visits and medication) for those who ever relapsed in three years was compared to those who never relapsed. One-year costs for both groups were also compared for a more stringent comparison. The analyses were adjusted for patient characteristics and took account of non-normality of the cost data by using a log-link function. UK unit costs were applied to resource use. The analysis was repeated after multiple imputation for missing data. Results: Costs incurred by patients who ever relapsed (£14,055) during three years were almost double to those incurred by patients who never relapsed (£7417). 61% of the cost difference was accounted for by hospital stay. The impact of relapse was even greater in the 1-year cost comparison. Results from the additional analysis with imputed missing data remained largely consistent. Conclusions: Our findings confirm the significant economic burden of relapse, and show such costs were mainly due to hospital stay. Nevertheless, the use of UK unit costs requires caution when interpreting this costing in the context of a specific country, as resource use and their associated costs will differ by country. © 2009 Elsevier Inc. All rights reserved.

1. Introduction Schizophrenia is a chronic mental illness with serious physical, social and economic consequences (Buckley, 1998; de Sena et al., 2003). The annual economic burden of schizophrenia was estimated to be £6.7 billion in 2004/05 in England, of which the direct cost of treatment and care accounted for about 30% (£2 billion) (Mangalore and Knapp, 2007). The burden of indirect costs — mainly because of lost productivity due to unemployment, absence from work and premature mortality of patients — was shown to be even greater, amounting to nearly £4.7 billion. A review of the international cost-ofillness studies also revealed (Knapp et al., 2004) an equally sizable economic burden of the illness.

Abbreviations: CGI, Clinical Global Impression; CGI-SCH, Clinical Global ImpressionSchizophrenia; HCHS, Hospital and Community Health Services; NHS, National Health System; PSSRU, Personal Social Services Research Unit; SOHO, Schizophrenia Outpatient Health Outcome; TFR, Trust Financial Returns. ⁎ Corresponding author. LSE Health, London School of Economics, Houghton Street, London, WC2A 2AE, UK. Tel.: +44 20 7955 6840; fax: +44 20 7955 6803. E-mail address: [email protected] (J. Hong). 0278-5846/$ – see front matter © 2009 Elsevier Inc. All rights reserved. doi:10.1016/j.pnpbp.2009.03.034

Much of this cost burden can be attributed to the consequences of relapse (Weiden and Olfson, 1995). The majority of patients with schizophrenia typically experience acute symptomatic relapses alternated with durations of full or partial remission over a period of many years (de Sena et al., 2003). Specialist psychiatric hospital admission and targeted treatments are often required during periods of acute relapse, which leads to significant resource demands on health care and social care systems (Beard et al., 2006). In addition, frequent relapses can cause progressive functional deterioration, worsening treatment response and clinical prognosis (Csernansky et al., 2002; Pompili et al., 2007), which would in turn increase both direct and indirect costs of the illness. Just a decade ago in the United States, Weiden and Olfson (1995) estimated the cost of relapse to be about $2 billion (at 1993 price) just for hospital readmission. Rather surprisingly however, there is only limited information on the actual costs of relapse in Europe. Many of cost estimates related to relapse have been drawn from economic evaluations of antipsychotics based on various assumptions. The large prospective pan-European Schizophrenia Outpatient Health Outcomes (SOHO) study provides a unique opportunity to estimate the direct costs associated with relapse over a 3-year follow-up in routine clinical practice in Europe.

836

J. Hong et al. / Progress in Neuro-Psychopharmacology & Biological Psychiatry 33 (2009) 835–841

2. Methods 2.1. Patient recruitment The SOHO study was an international prospective observational study that was conducted in 10 European countries. One thousand and ninety-six investigators participated, who were psychiatrists working mostly in public (46.9%) or combined public and private (37.2%) practices. A total of 10,972 patients were enrolled. Recruitment by country was highest in Germany (n = 3449 patients), Italy (n = 3016 patients), Spain (n = 2053 patients) and France (n = 964 patients). The remaining 1490 patients were enrolled in Denmark (n = 38 patients), Greece (n = 766 patients), Ireland/the UK (n = 344 patients), the Netherlands (n = 177 patients), and Portugal (n = 175 patients). The study was approved in all countries either at the site, regional, or national level, depending on the country and local regulations. Patient consent followed country regulations. All patients gave at least oral informed consent and written informed consent was obtained in Denmark, Italy, Portugal, Spain, Ireland and the UK. Participating psychiatrists offered enrolment to patients who met the following entry criteria: initiating or changing antipsychotic medication for the treatment of schizophrenia (according to DSM-IV or ICD10 diagnostic criteria and/or clinical judgement); presenting within the normal course of care in the outpatient setting or in the hospital when admission was planned for the initiation of antipsychotic medication and discharge planned within 2 weeks; at least 18 years of age; and not participating in an interventional study. Patients were included regardless of the reason for treatment change and regardless of whether an antipsychotic drug was being initiated as a replacement for a previous medication, was an addition to existing treatment, or was being initiated for the first time or after a period of no treatment. As the original design of the study was to focus on the outcomes of patients treated with olanzapine compared with other antipsychotic treatments, the study was designed to provide two cohorts of patients of approximately equal size by oversampling patients starting olanzapine: (1) patients who initiated therapy with or changed to olanzapine; and (2) patients who initiated therapy with or changed to non-olanzapine antipsychotic. Enrolment was thus conducted in a systematic alternating order in most countries; the first patient was recruited into the olanzapine cohort, the next patient was recruited into the non-olanzapine cohort, etc. All patient care during the study period was at the discretion of the participating psychiatrist. In order not to alter prescription practices, no minimum number of patients was required to be enrolled by each participating investigator and the enrolment period was purposely long. There were no instructions regarding medication treatment in the study description. The participating psychiatrists could prescribe medication and change it at any time following their usual practice. Patients remained in the study regardless of any medication change after baseline. Data collection for the study occurred during visits within the normal course of health care. The routine outpatient visit at which patients were enrolled served as the time for baseline data collection. Subsequent data collection was targeted for 3, 6, 12, 18, 24, 30 and 36 months. For each data collection target, investigators were allowed to collect data up to one month before or after the target month. Patients who were not seen during the normal course of care within one assessment interval were not excluded from subsequent data collection. Clinical severity was assessed using a scale based on the Clinical Global Impression (CGI) (Guy, 1976), which evaluated positive, negative, cognitive, depressive and overall symptoms on the day of assessment (CGI-SCH) (Haro et al., 2003a,b). The CGI and CGI-SCH are physician-rated scales with values ranging from 1 (not ill) to 7 (among the most severely ill patients). Compliance was also assessed by the participating investigators using a single-item question rating scheme: (1) patient has not been prescribed antipsychotic medication

during the prior 4 weeks; (2) patient almost always complies/adhere to antipsychotic medication treatment; (3) patient complies/adheres to antipsychotic medication treatment about half of the time; and (4) or patient almost never complies/adheres to antipsychotic medication treatment. Further details about the design of the SOHO study have been provided elsewhere (Haro et al., 2003a,b, 2005). 2.2. Definition of relapse Relapse is usually defined as a significant increase in symptom severity, decrease in social functioning or change in the pattern of care such as hospitalisation (Lader, 1995; Robinson et al., 1999). In the present study, relapse was defined as an increase in the CGI-overall severity score; or having had a hospitalisation. To qualify for relapse, the increase in the CGI-overall severity score has to be at least 3 points when the minimum score for that patient during follow-up was 1 (not ill), at least 2 points when the minimum score was 2 (borderline ill) or 3 (mildly ill); or at least 1 point when the minimum score was 4 (moderately ill), 5 (markedly ill) or 6 (severely ill). By including hospital admission in the definition of relapse, the cost of relapse was going to be higher due to admission costs. Thus, a sensitivity analysis was conducted defining relapse only in terms of the increase in the CGI-overall severity score (i.e. no requirement for hospitalisation). 2.3. Resource use and associated costs The following resource uses were recorded at every visit: the number of schizophrenia-related admissions and inpatient days; the number of schizophrenia-related day hospital or day care centre days; and the number of schizophrenia-related outpatient consultations with a psychiatrist. In addition, the type and dose of antipsychotic drugs prescribed and the type of concomitant medications taken (anticholinergics, antidepressants, anxiolytics/hypnotics or mood stabilisers) were also obtained. UK unit costs were applied to the pan European resource use data. UK unit costs were used because of their quality and availability. Unit costs for an inpatient day, day hospital/day care visits and outpatient psychiatric consultation were taken from the Personal Social Services Research Unit (PSSRU) unit cost volumes (Netten and Curtis, 2002, 2003), which in turn were based on the Trust Financial Returns (TFR2) specialty and programme cost returns to the UK Department of Health by NHS Trusts. The costs referring to the mental health specialty were inflated to include an element of capital costs based on the schemata published by Netten and Curtis (2002) and inflated to 2005 price levels using the Hospital and Community Health Services (HCHS) Pay and Prices Index (Netten and Curtis, 2005). The cost per mg of each antipsychotic drug was estimated from the Monthly Index of Medical Specialties (MIMS, 2005) and the Chemist and Druggist Supplement (CMP Medica Ltd, 2005). Where an antipsychotic had more than one form of packaging, the relative retail market shares for the UK were used to weigh the average price per mg using the IMS Health MIDAS® database (IMS Health Inc., Q2/2005). Where patients received concomitant medications, the average daily cost of anticholinergics, antidepressants, anxiolytics/hypnotics, or mood stabilisers was taken from the IMS Health MIDAS® database and weighted by the retail market share in the UK also provided by the IMS Health MIDAS® database (IMS Health Inc., Q2/2005). We assumed that patients receiving clozapine underwent haematological monitoring (differential and white blood cell count) prior to treatment initiation, weekly for the first 18 weeks of treatment and fortnightly thereafter. Patients co-treated with the mood stabilisers lithium, valproate or carbamazepine were assumed to have their drug plasma levels monitored as well as a complete blood cell count taken 2.5 times per annum. The costs of the tests were taken from the Pharmaceutical Industry Costing Analysis System (PICAS®) database.

J. Hong et al. / Progress in Neuro-Psychopharmacology & Biological Psychiatry 33 (2009) 835–841

2.4. Statistical analysis Of the 10,972 patients enrolled in the study, 7112 patients (69.6%) completed the 3-year follow-up. Relapse was defined within the group of 5674 patients who had non-missing data for CGI-overall severity scores and hospitalisation. Further, costs of relapse were estimated for the 3045 patients who had non-missing data for all variables included in a cost model. Note that the patient demographic profile was largely similar between the study completers (N = 7112) and non-completers, as well as between those patients included (N = 3045) and excluded in the final cost analysis (data not shown). The total costs attributable to relapse were estimated using a Poisson regression to take into account the skewness of the distribution of the cost data (Manning and Mullahy, 2001). A modified Park test (Santos-Silva and Tenreyro, 2006) indicated that this specification was adequate relative to other types of distributions such as the Gamma or inverse-Gaussian. OLS regression analysis was employed to estimate inpatient costs however because those who never relapsed during the 3-year follow-up (according to our definition) would have ‘zero’ inpatient costs, making the Poisson model less suitable due to its multiplicative nature. Relapse status was included as a binary variable in the model, and other baseline factors were controlled for. The baseline factors considered were country, gender, age at first treatment for schizophrenia, time since first treatment, alcohol dependency, substance abuse, suicide attempts, hostility, overall CGI, positive CGI, negative CGI, depressive CGI, cognitive CGI, compliance, body mass index, social functioning, marital status, living independently, having paid employment, being socially active, extra-pyramidal symptoms, tardive dyskinesia, impotence, and use of concomitant medications (anticholinergics, antidepressants, anxiolytics, and mood stabilisers) before baseline. Model reduction was conducted by dropping from the model the non-significant variables at the 0.05 level. The percentage change in costs due to relapse was calculated from the model. The reference cost was first calculated by taking the average predicted cost for the sample as a whole assuming all patients had not relapsed. The incremental cost associated with relapse was then calculated by applying the predicted percentage increase on the reference cost.

837

The primary cost analysis had many observations lost due to missing data; more than a half of the 3-year study completers (4067 patients, 57%) had at least one missing data for any of the variables included in the model, and were dropped from the primary analysis. A sensitivity analysis was therefore performed with imputed data for the missing values. The missing values were imputed with a multiple imputation method (Little and Rubin, 1987; Schafer, 1997), using all the variables included in the model. The procedure generated 20 imputed datasets, creating equally plausible versions of the complete data. The results of the 20 datasets were then combined using Rubin's rules for scalar estimands (Rubin, 1987) to produce one set of estimates and standard errors that incorporate missing data uncertainty. 2.5. Sensitivity analysis: 1-year costs associated with relapse In the primary analysis, the cost of resource use for those who ever relapsed in three years was compared to those who never relapsed. However, this method would inevitably dilute the cost of the intensive resource use during the acute episode of relapse as these costs would be spread over 3 years. Therefore, as a sensitivity analysis, 1-year resource use costs of relapsers and non-relapsers were compared to provide a more stringent cost of relapse estimate. In addition, the patient sample was restricted to those who achieved a CGI-overall severity score equal to or less than 3, which is compatible with symptomatic remission. This restriction would help flag out nonrelapse patients who should not have been included for analysis as they had chronic severe symptoms (which is technically a permanent relapse). In this analysis, non-relapsers were thus defined as those who continued to have a CGI-overall score ≤3 until the end of the study. The definition of relapse was the same as that used for the primary 3-year analysis. In total, 4383 patients were potentially eligible for the analysis, of which 32.4% (n = 1421) relapsed. However, only 2997 patients were included in the cost analysis because they had non-missing values for all variables included in the model. For relapsers, the costs of resource use were taken from 6 months before to 6 months after the visit when the patients first met a definition of relapse. For non-relapsers, the costs of resource use were taken during the 1-year period from the visit where the patients first achieved a symptomatic remission. In addition, this analysis was

Table 1 Baseline clinical and socio-demographic characteristics of patients who never relapsed and ever relapsed during follow-up.

Age at first treatment contact for schizophreniaa,† Years since first treatment for schizophreniaa,‡ Maleb,‡ Body Mass Index (kg/m2) a,† Never treated before included in SOHOb,† Having spouse or partner at baselineb,‡ Living independentlyb,‡ Having paid employment at baselineb,† Having had social contacts in the four weeks prior to baselineb,† Suicide attempts in the six months before baselineb,† Being hostileb,† CGI-SCH overall severity scorea,† GGI-SCH positive symptoms scorea,† CGI-SCH negative symptoms scorea,† CGI-SCH depressive symptoms scorea,† CGI-SCH cognitive symptoms scorea,† Use of antidepressants upon enrollmentb,‡ Use of anxiolytics upon enrollmentb,† Use of mood stabilisers upon enrollmentb,† Extrapyramidal symptomsb,‡ Tardive dyskinesiab,‡ Alcohol abuse at baselineb,† Substance abuse at baselineb,†

All patients (N = 5674)

Patients never relapsed (N = 3275)

Patients ever relapsed (N =2399)

28.5 (10.1) 11.8 (10.9) 58.0 26.3 (4.8) 12.3 29.3 47.9 24.2 68.8 4.3 27.2 4.4 (1.0) 3.8 (1.4) 4.1 (1.2) 3.5 (1.3) 3.8 (1.3) 18.7 38.2 9.3 39.3 9.9 2.6 2.1

28.9 (10.4) 11.8 (11.0) 56.9 26.2 (4.6) 13.4 29.7 48.6 26.5 70.8 3.3 25.0 4.3 (1.0) 3.7 (1.4) 4.0 (1.2) 3.4 (1.3) 3.7 (1.3) 18.5 35.1 8.3 39.4 9.3 2.2 1.8

27.9 (9.8) 11.8 (10.8) 59.5 26.4 (5.0) 10.6 28.7 47.0 21.0 66.2 5.8 30.2 4.6 (1.0) 3.9 (1.4) 4.2 (1.3) 3.5 (1.3) 3.9 (1.3) 19.0 42.4 10.8 39.1 10.7 3.2 2.6

† Significant at p b 0.05, ‡Non significant at p = 0.05; similar results achieved using non-parametric tests. GGI = Clinical Global Impression. a Value given as mean (standard deviation); t-test employed. b Value given as %; chi-square test employed.

838

J. Hong et al. / Progress in Neuro-Psychopharmacology & Biological Psychiatry 33 (2009) 835–841

repeated excluding hospital admissions in the definition of relapse, and also with the imputed missing data using multiple imputation as in the primary analysis. All analyses were conducted using STATA SE/10 (Stata.com). 3. Results More than half the patients were men (58%), and the mean age of the population at first treatment contact for schizophrenia was 28.5 years (Table 1). Patients, on average, had a long duration of illness (mean time since first treatment was 11.8 years). At baseline, approximately one in four patients had paid employment and about one in three had a spouse or partner. Of the 5674 patients potentially eligible for the analysis, 2399 patients (42.3%) ever relapsed during the 3-year follow-up (relapsers). Compared with the patients who never relapsed (non-relapsers), relapsers had an earlier onset of schizophrenia and had a poorer level of social functioning at baseline (i.e. a lower frequency of paid employment and social contacts in previous 4 weeks). In addition, more relapsers (5.8%) had suicide attempts in the six months before baseline, compared to nonrelapsers (3.3%). Furthermore, relapsers were more likely than nonrelapsers to be using anxiolytics and mood stabilisers at baseline. The demographic profile was largely similar for the 3045 patients who had any non-missing data and were eventually included in the primary cost analysis (data not shown). 3.1. 3-year resource use and costs Service use over the 3-year period and costs per patient are summarised by relapse status in Table 2. Over the 3 years, relapsers incurred almost double the costs of non-relapsers. The average estimated cost for relapsers was £14,055 while that for non-relapsers was £7417. Much of the cost difference was accounted for by inpatient care. The average number of admissions among relapsers was 1.2, with an average of 20.4 inpatient days (by definition, non-relapsers were never hospitalised). This inpatient service use incurred costs of £4117 for relapsers, accounting for 61% of the total cost difference between relapsers and non-relapsers (Fig. 1a). The proportion of inpatient care costs on the total cost difference still remained high even when relapse was defined based on worsening CGI-overall score only (not admission). The inpatient care costs accounted for 59% of the total cost difference (Fig. 1b). The second contributor to the cost difference was day hospital/day care visits. Relapsers visited day hospital/day care at least 2.5 times more frequently than non-relapsers (41.2 vs 16.7 visits) over the 3year follow-up period (see Table 2). This incurred additional costs of Table 2 Mean 3-year service use and costs per patient among those who never relapsed and who ever relapsed during 3-year follow-up. Patients never relapsed (N = 2053)

Patients ever relapsed (N = 992)

Average usage (SD) Costs (£) Average usage (SD) Costs (£) Total costsa Inpatient care (no. of admission) Inpatient care (no. of days) Day hospital visit Psychiatric visit Antipsychotics Concomitant medications Totalb

– 0

7417 0

0

– 1.2 (2.1)

14 055 4117

Fig. 1. Proportion of cost component on the 3-year cost difference between patients who never relapsed and patients who ever relapsed during follow-up, depending on the definition of relapse.

£1753 for relapsers, accounting for approximately one-quarter of the total increments due to relapse (Fig. 1). However, there was relatively a small difference in costs between relapsers and non-relapsers due to antipsychotic use: relapsers (£4316) vs non-relapsers (£3813). Similarly, costs for psychiatric visits and concomitant medications were very similar between relapsers and non-relapsers. Both incurred only small costs for these services: £1326 vs £1174 for psychiatric visits, and £757 vs £582 for concomitant medication. Using multiple imputation methods, 6275 patients were retained and, of these, 39.5% ever relapsed during the 3-year period. The analysis with the imputed data provided similar results to the primary analysis, although the incremental costs associated with relapse were slightly higher (Table 3). Over the 3 years, the total costs incurred by relapsers and non-relapsers were £15,585 and £7487, respectively. A slight increase in the costs for relapsers, compared to the primary analysis, was led by inpatient care costs, which increased to £5831 (from £4,117) in this analysis. These cost estimates were similar to the Table 3 Mean 3-year costs per patient among those who never relapsed and ever relapsed during 3-year follow-up, with imputed missing data using multiple imputation. Costs for patients never relapsed Costs for patients ever relapsed (£) (N = 3798) (£) (N = 2477)

20.4 (39.3)

16.7 (73.0) 23.9 (19.6) – –

1803 1174 3813 582

41.2 (108.4) 27.2 (22.4) – –

3556 1326 4316 757

–

7373

–

14 072

SD = standard deviation. a The total cost estimated with the aggregated patient costs. b The sum of each cost component after they were separately estimated.

Total costsa Inpatient care Day hospital/care visit Psychiatric visit Antipsychotics Concomitant medications Totalb a b

7487 0 1950

15 585 5831 3500

1202 3702 568

1374 4136 742

7421

15 584

The total cost estimated with the aggregated patient costs. The sum of each cost component after they were separately estimated.

J. Hong et al. / Progress in Neuro-Psychopharmacology & Biological Psychiatry 33 (2009) 835–841 Table 4 Factors associated with total cost.

Relapsedb Femaleb Age at first contact Having a spouse/partnerb Suicide attemptsb Being hostileb The use of antidepressants upon the enrolmentb The use of anxiolytics upon the enrolmentb The use of mood stabilisers upon the enrolmentb CGI-SCH cognitive symptoms score Compliance (having been non-compliant vs)b Having been compliant Having not been prescribed antipsychotics Constant

Coefficient (β)

95% CI

Exp(β) − 1a

0.64 − 0.10 − 0.06 − 0.17 0.26 0.13 0.11

[0.56; 0.72] [− 0.18; − 0.02] [− 0.10; − 0.02] [− 0.24; − 0.10] [0.02; 0.50] [0.04; 0.22] [0.00; 0.22]

0.90 − 0.10 − 0.06 − 0.15 0.30 0.14 0.12

0.19

[0.10; 0.27]

0.20

0.15

[0.04; 0.26]

0.16

0.06

[0.03; 0.09]

0.06

[− 0.04; 0.16] [− 0.26; − 0.03]

0.06 − 0.13

0.06 − 0.14 8.92

[8.61; 9.23]

839

costs on the total cost difference between relapsers and non-relapsers dropped to almost a half of that in the primary analysis (14% vs 26%). However, as in the primary analysis, only a small difference was found between relapsers and non-relapsers on outpatient consultation visits costs, antipsychotic costs and concomitant medication costs. In addition, when the analysis was repeated with the imputed data, the results were very similar to those without imputation. As with the primary analysis, there was a slight increase in the costs for the relapsers attributable to an increase in inpatient costs. 4. Discussion 4.1. Direct costs of relapse

–

a

Proportion change in total costs following a one-unit change in the explanatory variable (relative to a reference group if the variable is categorical). b Categorical variables.

cost estimates when the component costs were separately estimated with their own non-missing subset data (not shown here), rather than with non-missing data for all the cost variables in the primary analysis. 3.2. Factors associated with total costs Table 4 provides the factors associated with total costs over the 3year follow-up. Relapse had the strongest impact on total costs, which led to a 90% increase in costs when all other variables were held constant in the model. The impact of relapse was even greater in the 1year sensitivity analysis (206.3%) (data not shown). Other baseline factors also had an effect on total costs, independently of relapse status. Costs were raised by patients who had previous suicide attempts (30%), who were hostile (14%), who had more severe cognitive symptoms (6% per unit of CGI cognitive score), and who took concomitant medications upon enrolment (12% for antidepressants, 20% for anxiolytics, and 16% for mood stabilisers). Total costs were, however, reduced by patients who were female (10%), older at the first contact for treatment (0.6% per year of age), and who had a spouse or partner (15%). 3.3. Sensitivity analysis: 1-year resource use and costs The economic impact of relapse was even greater in a sensitivity analysis, which was performed to capture the 1-year costs directly attributable to relapse. Of the 2997 patients included in the analysis, 815 patients were identified as relapsers. Table 5 summarises the 1-year service use costs per patient by relapse status. In this sensitivity analysis, the economic impact of relapse was, not surprisingly, higher than that found in the primary analysis. The costs for relapsers were approximately three times higher than that for non-relapsers in this analysis while they were only two times higher in the primary analysis. The 1year costs incurred by both relapsers and non-relapsers were £7270 and £2374, respectively, which were equivalent to a half of the 3-year costs for relapsers and one-third of the 3-year costs for non-relapsers. As expected in this sensitivity analysis, inpatient care costs accounted for an even higher proportion of the total cost difference between relapsers and non-relapsers. Regardless of whether relapse was defined based on an increase in CGI score, with or without inpatient admission, about 80% of the total increments associated with relapse were explained by inpatient care costs. Meanwhile, the proportion of day hospital/day care

Our study showed that the costs incurred by patients who ever relapsed (£14,055) over the 3-year follow-up were about double to those incurred by patients who never relapsed (£7417) during the same period. The economic impact of relapse was even greater in a sensitivity analysis, which was performed to capture the 1-year costs directly attributable to relapse. This analysis found three times higher costs for relapsers compared to non-relapsers (£7270 vs £2374). While our study provided evidence of the substantial impact of relapse, the burden of relapse is likely to be even greater due to other “hidden” or “indirect” costs. Substantial societal costs can arise due to lost productivity, increased family burden, elevated rates of suicide and other co-morbid conditions etc. Violence and crimes could also arise during an acute episode of relapse, which would in turn increase not only costs but also deepen the stigma towards mental illness. Given the scarcity of cost data however, no studies have estimated the indirect costs associated with relapse in schizophrenia. The existing few cost-of-illness studies have highlighted sizable indirect cost burden. Further research is thus required to encompass both direct medical costs as well as other indirect costs to assess the widespread impact of relapse across many parts of the society. 4.2. Comparison with previous research The impact of relapse in our analysis was smaller than that reported in a previous UK study conducted by Almond et al. (2004). To our knowledge, their study is the only study available in the UK on the costs of relapse, except for one other study which estimated the costs using a simulation model (Almond and O'Donnell, 2000). Almond et al. (2004) calculated the costs of relapse for a sample randomly drawn from psychiatric caseloads in urban and suburban areas of Leicester, UK. They found that the costs for relapsers (£8212) were over four times higher than those for non-relapsers (£1899) during the 6-month study period. This difference was mainly because the patients in the Almond et al. study (2004) tended to stay longer in hospital, compared with the SOHO sample. The average length of inpatient stay among Table 5 Sensitivity analysis: mean 1-year costs per patient by relapse status.

Total costsa Inpatient care Day hospital/ care visit Psychiatric visit Antipsychotics Concomitant medications Totalb a b

Costs estimated without missing data (£)

Costs estimated with imputed missing data (£)

Non-relapsers (N = 2182)

Relapsers (N = 815)

Non-relapser (N = 3196)

Relapser (N = 1343)

2374 0 522

7270 3937 1227

2389 0 559

7742 4323 1323

418 1236 185

508 1405 260

421 1207 184

517 1401 258

2360

7336

2371

7822

The total cost estimated with the aggregated patient costs. The sum of each cost component after they were separately estimated.

840

J. Hong et al. / Progress in Neuro-Psychopharmacology & Biological Psychiatry 33 (2009) 835–841

relapsers was 57.8 days over the 6-month period in the Almond et al. study (2004) (the number of admissions is not clear), while it was only 20.4 days for the average of 1.2 admissions over the 3-year period in the SOHO study. This discrepancy could partially be attributed to the difference in the organisation of mental health services in the countries from which patients were recruited. Although the closure of psychiatric hospitals has been observed across Western Europe over the past 30 years, the care patterns still vary to a great extent from country to country. For instance, the average length of stay for schizophrenic-related disorders in 2003 was higher in the UK (104.5 days) than in Germany (39.6), Italy (17.3), Spain (44.9) and France (5.2) (OECD, 2007) — the main countries participating in SOHO. In addition to the study by Almond et al. (2004) the only other European study identified is a Spanish retrospective study, which estimated the cost of hospitalisation (psychiatric ward) per relapse to be about €3421 with a mean hospital stay of 21.8 days (Bernardo et al., 2006; Peiro et al., 2004; Rejas et al., 2002). In the US health system, Weiden and Olfson (1995) were the first to provide the cost estimates related to relapse, and they estimated that the total costs of rehospitalisation within 2 years after first hospitalisation discharge were almost US$2 billion. 4.3. Key cost driver of relapse: hospitalisation Consistent with the existing literature (Almond et al., 2004; Weiden and Olfson, 1995), hospitalisation was the key cost driver for the costs associated with relapse. Although length of stay among SOHO patients was relatively short, hospitalisation accounted for about 61% of the total incremental costs attributable to relapse over the 3-year follow-up (and about 80% in the sensitivity analysis that was conducted for 1-year costs). This is similar to that reported in the Almond et al. study (2004), though a direct comparison is problematic due to differences in costing resources, study design and analytic methodology. Nevertheless, 79% of the costs incurred by relapsers during 6 months in their study were accounted for by hospitalisation. All other costing studies focused only on the overall costs of schizophrenia (or costs of hospitalisation itself), but reported results consistent with the finding that a high proportion of the total costs were attributable to hospitalisation (Knapp et al., 2004). In addition, our study showed that the use of day hospital/day care also accounted for more than one-quarter of the total incremental costs attributable to relapse. However, the cost incurred by antipsychotic use was only marginally different between the relapsers and non-relapsers. This suggests that the cost of antipsychotic treatments (which prevent relapse) would be offset by the savings associated with preventing relapse/hospitalisation. 4.4. Other factors associated with total costs Our results confirm that relapse has the strongest impact on the total cost of treating schizophrenia. This is mainly attributable to the use of inpatient care services. In addition, previous suicide attempts, being hostile and having worse cognitive symptoms at baseline were independently associated with more resource use and higher costs. Notably, the factors associated with higher costs were not the most characteristic symptoms of schizophrenia (positive or negative symptoms), but behavioural manifestations (hostility and suicide attempts) that could plausibly be their consequences. Cognitive impairment on the other hand could have led to more care and treatment presumably because of difficulties in daily functioning (e.g. activities of daily living, social integration, and having employment). Another noteworthy point is that non-adherence to medication at baseline was not associated with higher costs despite the wellestablished finding that non-adherence increases the risk of relapse and contributes significantly to costs. This lack of association is probably because the impact of non-adherence is already reflected in the relapse status. Furthermore, medication non-adherence in our

analysis was measured at baseline and thus its time-variant nature could have not been taken into account. The present analysis, therefore, may not have adequately assessed the impact of nonadherence on costs, given the fact that non-adherence behaviour could vary with medication switches over time, an event that was frequent and prevalent during the normal course of care in SOHO. It is, however, beyond the scope of the present study to examine the timevarying effect of non-adherence on costs since the focus of our investigation was on the total costs over a 3-year follow-up period. Among the factors associated with lower costs, being female, older age at first contact for treatment and having a spouse/partner emerged as significant predictors. These factors have consistently been reported to be predictors of better clinical outcomes (Davidson and McGlashan, 1997; Perkins et al., 2004), and thus probably led to less resource use over the follow-up period. Notably, having a spouse/ partner conferred the most positive impact on lowering costs. This could be attributed to the protective role of social support, though one cannot exclude the possibility that patients who needed less care and treatment (and thereby less costs) were more likely to be engaged in relationships or social networks. Another possible explanation is that spousers/partners could have played an important role as informal care providers, and thus patients could have returned home as early as possible and received their care at home. This is particularly likely in Mediterranean countries such as Italy, where relatively strong familyoriented solidarity still exists. 4.5. Relapse rate Our study showed that 42.3% of patients ever relapsed over the 3year follow-up period and the equivalent rate was 32.4% among those who achieved remission. It should be noted, however, that these were crude estimates. As the study focus was to estimate the cost attributable to relapse over the 3-year follow-up, we only included the 3-year study completers who had non-missing data, which led to the exclusion of more than half of the enrolled patients. Given the longitudinal nature of the SOHO study, a survival analysis would provide a more accurate estimate of the relapse rate, as it can handle the missing data while taking into account time to relapse (Gaston and Marshall, 2003). A Kaplan–Meier analysis showed that relapse occurred in about one-quarter of the SOHO patients who achieved remission, and the relapse risk was stable over the 3-year follow-up (Haro et al., 2006). Given that on average about 30% to 40% of patients relapse within a 1-year period even if they are receiving maintenance treatment (de Sena et al., 2003; Schooler, 2003; Üçok et al., 2006), this 3-year relapse rate in SOHO is rather low. This could be attributed to our sample of patients being recruited in the outpatient setting and thus possibly having milder illness. 4.6. Limitations Given the multinational and naturalistic nature of the SOHO study, several aspects are worth mentioning when interpreting the findings. Firstly, more than a half of the 7112 study completers (57%) had missing data for at least one variable, especially for costs variables. In order to lessen biases in the sample and to minimise the loss of degree of freedom, the costs were thus re-analysed with imputed data for missing values, using multiple imputation. Although these cost estimates were consistent with those without imputation, there was a slight increase in the inpatient care costs and thereby total costs for relapsers. Such estimates were similar to the cost estimates when the component costs were estimated separately with their own nonmissing subset data, rather than with non-missing data for all the cost variables. This may suggest that to some extent, missing data could contribute to underestimates of the costs of relapse because among relapsers the most severely ill and thus costly patients are more likely to have missing data. Consequently, as illustrated by the present study

J. Hong et al. / Progress in Neuro-Psychopharmacology & Biological Psychiatry 33 (2009) 835–841

findings, multiple imputation may be of value in overcoming such a concern. Secondly, the SOHO study recorded only the key resources used in the treatment of schizophrenia. Other relevant services such as nurse visits were excluded from the present analysis, thereby likely leading to an underestimation of costs associated with relapse. In addition, as SOHO patients were recruited in the outpatient setting, they were more likely to have a milder illness. This in turn may have also led to some underestimation of costs. Thirdly, costs of relapse across pan-European countries were analysed using UK unit costs as per the methods proposed by Willke et al. (1998). Although we could aggregate resource use and cost data across countries using other methods such as the one proposed by Raikou et al. (2000), this was not done because of the paucity of good quality cost data across participating countries. Nevertheless, the use of UK unit costs requires caution when interpreting this costing in the context of a specific country. Resource use and their costs, particularly for inpatient care, do differ by country, depending on how healthcare financing is structured as well as the incentives and constraints facing individual clinical staff. Country-specific resource utilisation patterns along with their associated costs would thus be of interest to stakeholders and policy makers. Therefore, future investigation could focus on country level variation. The key cost drivers were nonetheless presented here. Lastly, patients were only evaluated at the assessment points (every six months post-baseline). Relapse or deterioration that occurred between visits was only detected when hospitalisation occurred as a result. 5. Conclusion This prospective European SOHO study confirms the significant economic burden of relapse in schizophrenia even when considering direct costs only. Such costs were mainly dominated by hospital stay while the costs of antipsychotics use were only marginally different between the relapsers and non-relapsers. However, the use of UK unit costs requires caution when interpreting this costing in the context of a specific country, as resource use and their associated costs will differ by country. Acknowledgements The authors would like to thank Deirdre Elmhirst, PhD, for her help in editing the manuscript. The SOHO study was funded by Eli Lilly and Company Limited, Windlesham, Surrey, UK. References Almond S, O'Donnell O. Cost analysis of the treatment of schizophrenia in the UK: a comparison of olanzapine, risperidone and haloperidol. Pharmacoeconomics 2000;17:383–9. Almond S, Knapp M, Francois C, Toumi M, Brugha T. Relapse in schizophrenia: costs, clinical outcomes and quality of life. Br J Psychiatry 2004;184:346–51. Beard SM, Maciver F, Clouth J, Ruther E. A decision model to compare health care costs of olanzapine and risperidone treatment for schizophrenia in Germany. Eur J Health Econ 2006;7:165–72. Bernardo M, Ramon Azanza J, Rubio-Terres C, Rejas J. Cost-effectiveness analysis of schizophrenia relapse prevention: an economic evaluation of the ZEUS (Ziprasidone-Extended-Use-In-Schizophrenia) study in Spain. Clin Drug Investig 2006;26: 447–57.

841

Buckley PF. Treatment of schizophrenia: let's talk dollars and sense. Am J Manag Care 1998;4:369–83. CMP Medica Ltd. Chemist and Druggist Supplement. In; 2005. Csernansky JG, Mahmoud R, Brenner R. A comparison of risperidone and haloperidol for the prevention of relapse in patients with schizophrenia. N Engl J Med 2002;346:16–22. Davidson L, McGlashan TH. The varied outcomes of schizophrenia. Can J Psychiatry 1997;42:34–43. de Sena EP, Santos-Jesus R, Miranda-Scippa A, de Casto Quarantini L, de Oliveira IR. Relapse in patients with schizophrenia: a comparison between risperidone and haloperidol. Rev Bras Psiquiatr 2003;25:220–3. Gaston P, Marshall RW. Survival analysis is a better estimate of recurrent disc herniation. J Bone Joint Surg Br 2003;85:535–7. Guy W, editor. Clinical global impression; 1976. Haro JM, Edgell ET, Jones PB, Alonso J, Gavart S, Gregor KJ, et al. The European Schizophrenia Outpatient Health Outcomes (SOHO) study: rationale, methods and recruitment. Acta Psychiatr Scand 2003a;107:222–32. Haro JM, Edgell ET, Novick D, Alonso J, Kennedy L, Jones PB, et al. Effectiveness of antipsychotic treatment for schizophrenia: 6-month results of the Pan-European Schizophrenia Outpatient Health Outcomes (SOHO) study. Acta Psychiatr Scand 2005;111:220–31. Haro JM, Kamath SA, Ochoa S, Novick D, Rele K, Fargas A, et al. The Clinical Global Impression-Schizophrenia scale: a simple instrument to measure the diversity of symptoms present in schizophrenia. Acta Psychiatr Scand 2003b;107(Suppl 416):16–23. Haro JM, Novick D, Suarez D, Alonso J, Lépine JP, Ratcliffe M, et al. Remission and relapse in the outpatient care of schizophrenia. Three-year results from the Schizophrenia Outpatient Health Outcomes Study. J Clin Psychopharmacol 2006;26:571–8. IMS Health Inc. MIDAS®. Q2/2005. Knapp M, Mangalore R, Judit S. The global costs of schizophrenia. Schizophrenia Bulletin 2004;30:279–93. Lader M. What is relapse in schizophrenia? Int Clin Psychopharmacol 1995;9(Suppl 5):5–9. Little R, Rubin D. Statistical analysis with missing data. New York: John Wiley; 1987. Mangalore RK, Knapp M. Cost of schizophrenia in England. J Ment Health Policy Econ 2007;10:23–41. Manning WG, Mullahy J. Estimating log models: to transform or not to transform? Journal of Health Economics 2001;20:461–94. MIMS. London: Haymarket publications; 2005. Netten A, Curtis L. Unit costs of health and social care 2002. Canterbury: PSSRU University of Canterbury; 2002. Netten A, Curtis L. Unit costs of health and social care 2003. Canterbury: PSSRU University of Canterbury; 2003. Netten A, Curtis L. Unit costs of health and social care 2005. Canterbury: PSSRU University of Canterbury; 2005. OECD. OCED health data. In; 2007. Peiro S, Gomez G, Navarro M, Guadarrama I, Rejas J. Length of stay and antipsychotic treatment costs of patients with acute psychosis admitted to hospital in Spain. Description and associated factors. The Psychosp study. Soc Psychiatry Psychiatr Epidemiol 2004;39:507–13. Perkins D, Lieberman J, Gu H, Tohen M, McEvoy J, Green A, et al. Predictors of antipsychotic treatment response in patients with first-episode schizophrenia, schizoaffective and schizophreniform disorders. Br J Psychiatry 2004;185:18–24. Pompili M, Amador X, Girardi P, Harkavy-Friedman J, Harrow M, Kaplan K, et al. Suicide risk in schizophrenia: learning from the past to change the future. Ann Gen Psychiatry 2007;6:10. Raikou M, Briggs A, Gray A, et al. Centre-specific or average unit costs in multi-centre studies? Some theory and simulation. Health Econ 2000;9:191–8. Rejas J, Ortega P, Mackell J. Impacto y consecuencias económicas de los efectos secundarios de los fármacos antipsicóticos actuales. Actas Esp Psiquiatr 2002;30:52–66. Robinson D, Woerner MG, Alvir JMJ, et al. Predictors of relapse following response from a first episode of schizophrenia or schizoaffective disorder. Arch Gen Psychiatry 1999;56:241–7. Rubin D. Multiple imputation for nonresponse in surveys. New York: John Wiley & Sons; 1987. Santos-Silva JMC, Tenreyro S. The log of gravity. Rev Econ Stat 2006;88:641–58. Schafer J. Analysis of incomplete multivariate data. London: Chapman & Hall; 1997. Schooler NR. Relapse and rehospitalization: comparing oral and depot antipsychotics. J Clin Psychiatry 2003;64:14–7. Üçok A, Polat A, Çakır S, Genç A. One year outcome in first episode schizophrenia. Predictors of relapse. Eur Arch Psychiatry Clin Neurosci 2006;256:37–43. Weiden PJ, Olfson M. Cost of relapse in schizophrenia. Schizophr Bull 1995;21:419–29. Willke RJ, Glick HA, Polsky D, Schulman K. Estimating country-specific costeffectiveness from multinational clinical trials. Health Econ 1998;7:481–93.