Future employability, a new approach to cost-effectiveness analysis of antipsychotic therapy

Schizophrenia Research 63 (2003) 111 – 119 www.elsevier.com/locate/schres

Future employability, a new approach to cost-effectiveness analysis of antipsychotic therapy Rahul Ganguly a, L. Stephen Miller b, Bradley C. Martin a,* a

Department of Clinical and Administrative Sciences, College of Pharmacy, University of Georgia, 250 DW Brooks Drive, Athens, GA 30602, USA b Department of Psychology, University of Georgia, Athens, GA 30602, USA Received 17 April 2002; received in revised form 16 July 2002; accepted 22 July 2002

Abstract Objective: Schizophrenia patients pose a substantial burden in terms of indirect costs, much of which is attributable to loss of employment. We present a new approach to assess the cost-effectiveness (CE) of risperidone vs. haloperidol using employability as an outcome measure. Methods: A decision analytic CE model was developed to compare the two treatments over a 1-year period including all direct medical costs and the number of employable persons as a measure of effectiveness. A measure of executive functioning, the Wisconsin Card Sorting Test Category score (WCST-Cat score), was used as an intermediate endpoint from which employability was modeled. A Monte Carlo procedure, using WCST-Cat score sampling distributions from clinical trials, simulated the WCST-Cat score distribution for a cohort of 10,000 patients. A clinically stable patient, with a Positive and Negative Syndrome Scale (PANSS) score increase of at least 20% and a WCST-Cat score of z 3.5 was assumed to be employable. Sensitivity analysis was performed for key values. Results: The base case per-patient cost of risperidone and haloperidol was US$6422 and US$4989, respectively, and the number of employable persons was 3258 (32.58%) and 2517 (25.17%), respectively, which translates to an incremental CE ratio of US$19,609 per employable person for risperidone. Risperidone remained cost increasing and had higher number of employable persons over all the ranges used in the sensitivity analysis. The incremental CE ratio ranged from a high of US$1,000,000 to a low of US$3000 per employable persons when the rates of clinical stability for risperidone and haloperidol therapy were varied. Conclusion: Gains from earning rates for employed persons with schizophrenia, savings in informal caregiver costs and other human benefits could justify an incremental cost of US$19,609 for each additional employable person prescribed risperidone. D 2002 Elsevier Science B.V. All rights reserved. Keywords: Cost-effectiveness; Employment; Risperidone; Haloperidol

1. Introduction In 1993, the annual costs of Schizophrenia were estimated at around US$33 billion (Meltzer, 1999). * Corresponding author. Tel.: +1-706-542-5367; fax: +1-706542-5228. E-mail address: [email protected] (B.C. Martin).

Nearly half of this estimated annual cost is attributable to direct health care costs (US$17.5 billion for hospitalization and health care and US$0.5 billion for medication) and the rest to indirect costs (US$15 billion for morbidity, mortality, and lost productivity). Indirect costs, which pose a substantial financial burden in schizophrenia, are primarily due to lost productivity or the schizophrenia patients’ inability

0920-9964/02/$ - see front matter D 2002 Elsevier Science B.V. All rights reserved. doi:10.1016/S0920-9964(02)00377-8

112

R. Ganguly et al. / Schizophrenia Research 63 (2003) 111–119

to obtain and maintain stable and gainful employment (Sevy and Davidson, 1995). Besides reducing indirect costs, employment has other positive effects on the patient—improving his or her self-esteem, facilitating a more disciplined life, increased compliance with treatment and improved social re-integration (McGurk and Meltzer, 2000). Many social and community interventions are either devoted to improving employment skills (vocational training, rehabilitation, supported employment, and job clubs) or include a component aimed at improving the patients’ employment skills (social skills training). Cognitive impairment has been found to be a major contributor to the schizophrenia patients’ inability to obtain and maintain jobs (Sevy and Davidson, 1995). Corresponding evidence suggests that various forms of cognitive impairment may be of equal or greater importance than positive or negative symptoms in predicting functional outcomes such as work status, activities of daily living, community outcome, social problems and skill acquisition (Meltzer and McGurk, 1999; Meltzer et al., 1996; Keks, 1997). One such aspect of cognition, executive functioning, as measured by the Wisconsin Card Sorting Test (WCST) score has been found in studies to predict work behavior in schizophrenia patients (McGurk and Meltzer, 2000). This is of special relevance today as recent studies have shown that atypical antipsychotics improve cognitive functions in schizophrenia patients (Meltzer and McGurk, 1999). For example, risperidone has relatively consistent positive effects on working memory, executive functioning and attention whereas olanzapine improves verbal learning and fluency and executive function (Meltzer and McGurk, 1999). These cognitive functions have been found to be important predictors of work status (Meltzer, 1999). Conversely, most of the literature on conventional antipsychotic therapy report that higher levels of antipsychotic medications are associated with poorer performance on tests of cognition (Sweeney et al., 1991). Thus, from a theoretical standpoint, it can be hypothesized that due to their differential effects on cognitive functioning, use of either atypical or conventional antipsychotics could potentially influence the employability of the schizophrenia patient. A literature search revealed two studies (Hamilton et

al., 2000; Meyer et al., 2002) that investigate a relationship between antipsychotic use and employment status. The Hamilton study is a double-blind randomized controlled trial and reports a significant difference between olanzapine vs. haloperidol in clinician rated work status (15.1% vs. 5.3% employed). The Meyer et al. (2002) study is a nonrandomized cross-sectional naturalistic trial that compares atypical (risperidone or olanzapine) use and traditional antipsychotic use (haloperidol) and finds no significant difference in paid employment (40% vs. 35% employed). However, we did not find any trial that estimates the cost-effectiveness using ‘employment’ as a measure of effectiveness. In situations where clinical trial or observational data do not exist, decision modeling techniques have been found to be appropriate in estimating and comparing the effects of competing interventions (Torrance et al., 1996). A review of current literature on modeling studies of antipsychotics revealed that most studies rely on clinical endpoints (presence or absence of extra pyramidal syndrome, Brief Psychiatric Rating Scale, Positive and Negative Scores Scale) or measures of direct cost (hospital days) to compare the different drugs (Foster and Goa, 1998). We use decision-modeling techniques to compare the effect of an atypical antipsychotic, risperidone, and a conventional antipsychotic, haloperidol, on future employability of the schizophrenia patient. Using future employability or the ‘number of employable persons’ as our effectiveness measure and all direct medical costs as our cost measure, we calculated and compared the cost-effectiveness of these two treatments. The payer perspective was used for the analysis and the study time horizon was 1 year.

2. Methods We used a Monte Carlo simulation model to estimate the effect of the two different drugs on employability. First, two hypothetical cohorts of 10,000 recently diagnosed or hospital discharged outpatients with schizophrenia are selected and assigned to either risperidone or haloperidol treatment. Second, the model simulated the course of the disease for a 1-year period under each assigned alternative. The

R. Ganguly et al. / Schizophrenia Research 63 (2003) 111–119

113

Fig. 1. Simulation model of schizophrenia outpatients for a 1-year period.

cohorts were followed from a baseline outpatient visit through consecutive clinical states (Fig. 1) to the end of a year. The simulation procedure reported number of patients in each health state (stable, exacerbated and hospitalized) and generated their WCST Category (WCST-Cat) scores by random sampling from appropriate post treatment WCST-Cat score distributions assigned to the treatment alternatives. Individuals who were stable and had a WCST-Cat score of z 3.5 were assumed to be employable. The model parameters (e.g., health states, probabilities and costs) relied on clinical trials, meta analysis of clinical trials and sources of cost. The base case costs, number of employable persons, costs per employable person, incremental cost effectiveness ratios and results of sensitivity analysis of key variables are reported. 2.1. Model structure Our model is an extension of the Glazer and Ereshefsky (1996) model for schizophrenia outpatients (Fig. 1). Our study population consists of schizophrenia patients’ who were either treated in the outpatient setting or recently discharged from the hospital. We added terminal branches to the original model for WCST-Cat scores. This model was built using Data 3.5 software (TreeAge, 2002) and extends for a period of 1 year. Since the noncompliant branch follows the same clinical pathway as the compliant branch and haloperidol treatment follows the same pathway as risperidone those branches have been collapsed in Fig. 1 denoted by a ‘ + ’.

2.2. Model states and key assumptions The model states with their abbreviated operational definitions are reported in Table 1 and probability estimates for these states with their sources are reported in Table 2. These patients are assumed to be receiving their medication in an outpatient setting. Thus, the first major issue is compliance. Trial completion rates are used as proxy measures of compliance and drop out rates for non-compliance. The key parameters (Compliance rate, Stability rate after compliance) were obtained from Davies et al. (1998) and Peuskens (1995), which are a meta analysis of six trials and a single large trial (n>1000). Since extrapyramidal (EPS) side effects are a major component of

Table 1 Clinical definitions of the health states modeled Health state

Clinical definition

Compliant Non-compliant

Patient uses medication for all of 1 year Patient uses medication for half year. Sensitivity analysis was performed under varying assumptions ranging from medication use for zero days to half a year Stable Reduction of total PANSSa score of at least 20% from baseline Exacerbation Intermittent exacerbation of symptoms not severe enough to lead to hospitalization Hospitalization No change in PANSSa scores needing hospitalization WCST-Catb> = 3.5 Patients who are stable and have a WCST-Catb score of more than 3.5 were assumed to be employable and the rest were considered not employable a b

PANSS: Positive and negative syndrome scale. WCST: Wisconsin Card Sorting Test.

114

R. Ganguly et al. / Schizophrenia Research 63 (2003) 111–119

Table 2 Probability values used in the model Baseline

Rangea

Source

Risperidone Compliance Comply—stableb Comply—exacerbation Comply—hospitalization

0.772 0.79 0.105 0.105

0.62 – 0.93 0.60 – 0.90

Davies et al. (1998) Peuskens (1995) Assumption, Glazer and Ereshefsky (1996) Assumption, Glazer and Ereshefsky (1996)

Haloperidol Compliance Comply—stable Comply—exacerbation Comply—hospitalization

0.67 0.81 0.095 0.095

Noncompliant Stable Exacerbation Hospitalization Duration of drug use

0.167 0.333 0.50 90 days

a b

0.00 – 0.21

0.54 – 0.80 0.65 – 0.97

Davies et al. (1998) Peuskens (1995) Assumption, Glazer and Ereshefsky (1996) Assumption, Glazer and Ereshefsky (1996)

0.00 – 0.19

0.00 – 0.50 0.40 – 0.60 0.00 – 182.50 days

Assumption, Glazer and Ereshefsky (1996) Assumption, Glazer and Ereshefsky (1996) Davies et al. (1998), Glazer and Ereshefsky (1996) Expert opinion

Range used to perform sensitivity analysis. Probability value for being clinically stable for patients who comply with risperidone therapy.

antipsychotic therapy, the cost of EPS was built into the model. It was assumed that 10.8% of the compliant risperidone patients and 28.5% of the compliant haloperidol patients would have an EPS episode at 6 months. These percentages were obtained from the Davies et al. (1998) study. 2.3. Wisconsin Card Sorting Test (WCST) scores The Wisconsin Card Sorting Test (WCST) has been used extensively in schizophrenia patients to measure executive functioning (Goldman et al., 1991). Various studies have reported a significant relationship between these scores and occupational status or occupational performance of the patient (Lysaker et al., 1995; Jaeger and Douglas, 1992; McGurk and Meltzer, 2000). These findings indicate that executive functioning, as measured by WCST, has concurrent and predictive validity for social and occupational functioning (McGurk and Meltzer, 2000). Two subtypes of WCST, WCST Category measure (WCST-Cat score) and WCST Percent Perseveration have been commonly used in schizophrenia patients to measure cognitive functions. In most of these studies, the WCST Category measure was found to best discriminate the employed from the rest (Meltzer et al., 1996; McGurk and Meltzer, 2000; Lysaker et al., 1995). Thus, our model used the WCST-

Cat score as the intermediate outcome measure from which employability was modeled. In order to determine the number of employable persons, a WCST-Cat cut-off score to differentiate the ‘‘employable’’ from the ‘‘unemployable’’ group of people was needed. With the purpose of selecting any one cut-off we compared and contrasted published literature relating WCST-Cat score and employment (Meltzer et al., 1996; McGurk and Meltzer, 2000) and agewise normative WCST-Cat scores (Heaton, 1981) for the general population. Since our simulated patient cohort of risperidone patients was based on studies with a mean age of around 34 years (Rossi et al., 1997), we referred the normative scores for the general population aged < 40 years (Heaton, 1981). The lower limit (95% confidence interval (CI)) for normative WCST-Cat scores for this group was 3.6. Since this was similar to the score of 3.5 (upper limit of 95% CI for those unemployed) derived from the McGurk and Meltzer (2000) study relating WCSTCat scores and employment status in patients with schizophrenia, we chose to use 3.5 as a cut-off. We assumed that those patients who have a WCST-Cat scores of > = 3.5 and remain ‘‘stable’’ over the 1-year period have the ability to work at least part-time. We performed a sensitivity analysis varying this cut-off from 2.5 to 5.5.

R. Ganguly et al. / Schizophrenia Research 63 (2003) 111–119

In addition to the minimum WCST-cat score, it was assumed that only those patients who are stable would be employable. It must be noted that this may not always be true and some patients experiencing episodic exacerbation may be able to work between episodes of active symptoms. The hospitalized patients were also assumed to be in the unemployable group. 2.3.1. Treatment effects on WCST scores A literature search was performed to identify studies that report WCST-Cat score distributions following risperidone, haloperidol and no treatment by searching Medline from 1995 to 2000 and searching back references of articles identified. Criteria for selecting studies were: published in English, report WCST-Cat scores at baseline and after haloperidol or risperidone treatment, and had a minimum follow up time of 4 weeks. We identified five studies that studied executive functioning as an outcome of risperidone, however only one study met our selection criteria (Rossi et al., 1997). Most of the studies were excluded because they did not use WCST-Cat score as a measure of executive functioning or did not report baseline scores. The study meeting criteria included 30 patients with schizophrenia (20 men and 10 women, mean age 33.9 years) and reported both baseline WCST-Cat scores and scores after 4 weeks of risperidone therapy (mean dose 4.6 mg). Most of the literature on conventional antipsychotic therapy report that higher levels of antipsychotic medications are associated with poorer performance on tests of cognition (Sweeney et al., 1991). The Sweeney et al. (1991) study has reported a negative correlation between dose of chlorpromazine and/or benzotropine (rho = 0.38 and 0.56, rho = 0.26 and 0.20) and WCST-Cat scores at initial and follow-up assessment after a year. No study reported improvement in WCST-Cat scores. Thus, for the purpose of the study it was assumed that the patients in the haloperidol group would not show any improvement in WCST-Cat scores and maintain their baseline WCST-Cat scores. A sensitivity analysis was performed to permit possible decrements in WCST-Cat scores associated with haloperidol. Since we assume that the non-compliant group would consume little or no medication, we assumed that their WCST-Cat scores would remain constant over the 1-year period. It would be preferable to use a

115

specific score distribution for non-compliant patients but no such data was available in the literature. 2.4. Costs We considered all direct costs of treatment: drug therapy, clinic visits, case management, monitoring and management of moderate/severe side effects, and daily inpatient costs of rehospitalization. The various unit costs have been listed in Table 3. The unit measures for resource utilization were obtained from the Glazer and Ereshefsky (1996) study. The dosage for Antipsychotic therapy was estimated from the clinical trials. The baseline usage was estimated at 4 mg/day for risperidone and 10 mg/day for haloperidol. Rehospitalization costs were valued using the Georgia FY95 Department of Human Resources (1995) hospital and budget utilization report for State operated psychiatric hospitalizations by dividing the total costs (Direct care, Direct patient and other support costs) by total patient days (Days of Active Client Enrollment) to arrive at an average per-diem rate. Physician costs were valued using the National Physician Fee Schedule Relative Value File for the calendar year 1999 obtained from the Health Care Financing Administration (HCFA) currently known as Centers for Medicare and Medicaid Services (CMS) website (http://www.hcfa.gov/stats/pufiles.htm). The costs were calculated from the payment formula for the year 1999. We computed a national average by using a unit value for Geographic Practice Cost Index (GPCI) in the formula. A Relative Value Unit (RVU) of 0.64 was considered for calculating the base visit cost and 1.29 for a relapse or an Extrapyramidal Syndrome (EPS) episode. All costs were converted to 2001 US$ using the Medical Care component of the US city average annual Consumer Price Index (http://www.bls.gov/ cpi/home.htm). 2.5. Monte Carlo simulation procedure We used TreeAge Data 3.5 (1999) version to build the stochastic model. We built a hypothetical cohort of 10,000 schizophrenia patients for each arm. We assumed that the two cohorts have equivalent patient and environmental characteristics at baseline. Both

116

R. Ganguly et al. / Schizophrenia Research 63 (2003) 111–119

Table 3 Costs in 2001 US$ Type of cost

Unit costs (2001 US$)

Haloperidol, oral (10 mg/day)

0.05/day (Red Book, 1999)

Range

Risperidone (4 mg/day)

6.80/day (Red Book, 1999)

3 – 10

Rehospitalizations Physician visit Physician visit (post-exacerbation) Case management Anticholinergic use for EPS

442.15/day (DHR, 1995)a 54.00/visit (NPFS, 1999)b 86.4/visit (NPFS, 1999)b

221 – 660 27 – 81 43 – 129

24.4/h (Glazer and Ereshefsky, 1996) 0.03/day (Red Book, 1999)

Antidepressant

1.25/day (Red Book, 1999)

Comments Incremental costs of 15% and 30% considered for exacerbation and hospitalization Incremental costs of 15% and 30% considered for exacerbation and hospitalization 30 days at 6 months, and again 14 days after 9 months Regular visits Added visits

12 – 36 10.8% of risperidone patients and 28.5% of haloperidol patients use this daily after an EPS episode Averaged cost of Fluoxetine and Amitryptallinec—10% to 20% haloperidol group would be on antidepressantsd

a

Department of Human Resources hospitalization and budget utilization report. National fee schedule relative value scale. c Fluoxetine and Amitryptalline were found to be the most commonly used antidepressants and cost was calculated from the proportions using either drug. The proportions were obtained from Williams et al. (1999). d Davies et al. (1998). b

cohorts consisted of recently diagnosed or hospital discharged outpatients with schizophrenia with a mean age of around 34 years. Using compliance rates obtained from a metaanalysis (Davies et al., 1998), we calculated that 7720 patients out of 10,000 would be compliant and the rest (2280) non-compliant with risperidone. Similarly, 6700 patients would be compliant with haloperidol. We used the WCST-Cat score distribution post risperidone treatment, 3.50 (S.D. = 2.46) from the Rossi et al. (1997) study to generate simulated WCST-Cat score data for the risperidone-compliant patients (7720). The simulation procedure assumed the score distribution to be normal with a mean of 3.5 (S.D. = 2.46) and generated WCST-Cat scores through random sampling from this distribution. The WCSTCat score data before risperidone treatment or baseline WCST-Cat score, 3.00 (S.D. = 2.24) was used to generate simulated WCST-Cat scores for the risperidonenon-compliant, haloperidol compliant, and haloperidol non-compliant groups since our original assumption was that their scores would remain unchanged from baseline. The simulation procedure reported number of patients in each health state (stable, exacerbated and hospitalized) and their WCST-Cat scores. This data was further exported as a text file and analyzed using the SAS Version 8.02 (SAS, 2002) to determine the number of employable persons. To ensure replicability

of the study, a predictable random sequence seed value of 1000 was used for the Monte Carlo simulation procedure (TreeAge, 1999).

3. Results 3.1. Base case The expected per patient cost of risperidone treatment was US$6422, US$1453 more than treatment with haloperidol (Table 4). Treatment with risperidone resulted in more employable persons (3258 persons) as compared to haloperidol (2517 persons), which translated into an increased employability rate of 8% with risperidone. The incremental cost-effectiveness Table 4 Cost-effectiveness (CE) under baseline assumptions

Expected cost per patient Expected number of employable persons (n = 10,000) CE ratioa Incremental CE

Risperidone

Haloperidol

US$6442

US$4989

3258

2517

US$19,773 US$19,609

US$19,821

a CE ratio=(Expected cost per patient  10,000)/Expected no. of employable persons.

R. Ganguly et al. / Schizophrenia Research 63 (2003) 111–119 Table 5 Sensitivity analysis of selected cost and probability values Range Risperidone Compliance Comply—stable Non-comply— hospitalized Cost of risperidone Haloperidol Compliance Comply—stable Non-comply— hospitalized a

Incremental CEa 1999 US$

0.62 – 0.93 0.60 – 0.90 0.40 – 0.60

4902 – 88,465 5068 – 1,084,815 12,822 – 26,879

US$3 – 10

2940 – 30,713

0.54 – 0.80 0.65 – 0.97 0.40 – 0.60

4629 – 57,668 3268 – 81,355 10,191 – 30,626

117

Table 7 Sensitivity analysis varying WCST-Cata mean score after haloperidol therapy WCST-cata mean score

Risperidone (n = 10,000) employable persons

Haloperidol (n = 10,000) employable persons

Incremental CEb 2001 US$

2.0 2.3 2.5 2.8 3.0

3258 3258 3258 3258 3258

1641 1891 2063 2345 2517

8986 10,629 12,159 15,915 19,609

a b

WCST: Wisconsin Card Sorting Test. CE: Cost-effectiveness.

CE: Cost-effectiveness.

Sensitivity analysis was performed for all probability values, cost variables, WCST-Cat cut-off scores and WCST-Cat score distribution after haloperidol treatment. Only those cost and probability values that produced notable changes in results have been reported in Table 5. Risperidone remained cost increasing and had higher number of employable persons over all the ranges used for all variables in the sensitivity analysis. The probability of achieving clinical stability for compliant patients played a major role in influencing the results of the analysis followed by compliance rates with the drugs. The incremental CE ratio ranged from a

high of US$1,084,815 per employable person when the probability of being stable on risperidone was lowered to 60% and to a low of US$2940 per employable person when the cost of risperidone was lowered to US$3 per unit. The WCST-Cat score cut-off was varied from 2.5 to 5.5 and risperidone had higher number of employable persons over all the ranges of WCST-Cat cut-off scores. The incremental CE ratio exhibited a U-shaped pattern where it was lowest for the base-case scenario and increased when the WCSTCat cut-off scores were varied in either direction with the highest CE ratio of US$27,261 (cut-off: 5.5) (Table 6). The WCST-Cat score distribution after haloperidol treatment was varied shifting the mean from the baseline value of 3.0 (S.D. = 2.24) to a low of 2.0 keeping the standard deviation constant. The incremental CE reduced to US$8986 from the baseline value of US$19,609 (Table 7).

Table 6 Sensitivity analysis varying WCSTa employment cut-off scores

4. Discussion

ratio for risperidone was US$19,609 for each additional employable person. 3.2. Sensitivity analysis

a

WCST-Cat score cut-off

Risperidone (n = 10,000) employable persons

Haloperidol (n = 10,000) employable persons

Incremental CEb 2001 US$

z 2.5 z 3.0 z 3.5 z 4.0 z 4.5 z 5.0 z 5.5

4313 3791 3258 2722 2228 1764 1354

3594 3056 2517 2019 1548 1144 821

20,209 19,769 19,609 20,669 21,368 23,435 27,261

a b

WCST: Wisconsin Card Sorting Test. CE: Cost-effectiveness.

As far as we are aware, this is the first attempt to model employability of patients with schizophrenia. As a check of the validity of this new model we found employment rates derived from our model to be consistent with those published in the literature, i.e. around 25 – 30% (Birchwood and Jackson, 2001). Treatment with risperidone results in a higher number of employable persons compared with haloperidol therapy. This result is influenced by the fact that patients are; more compliant with risperidone therapy, more likely to be stable on the therapy, and have

118

R. Ganguly et al. / Schizophrenia Research 63 (2003) 111–119

higher levels of cognition and executive functioning (WCST-Cat scores). This result is not sensitive to variations in the model assumptions and risperidone remains more effective when the key assumptions are varied. Further evidence of model validity is obtained from the fact that the incremental cost of risperidone treatment, US$1453 compares closely with the incremental cost of US$1744 obtained in the Glazer and Ereshefsky (1996) study (result was converted to 2001 US$). Our results were consistent with other costeffectiveness studies that find risperidone to be more cost-effective than haloperidol using clinical or health care utilization effectiveness measures (Foster and Goa, 1998). The incremental cost effectiveness ratio of US$19,609 per employable person with risperidone therapy seems reasonable in the light of the fact that a typical full time employed patient earns around US$9500 per year (US$5.50 per hour, 36-h weeks) (Mcgurk and Meltzer, 2000) and part-time US$2520 per year (US$5.25 per hour, 10-h weeks) (Mcgurk and Meltzer, 2000). Since our ‘employable groups’ do not include unpaid work (prevocational training, volunteer work, sheltered workshop) (Meyer et al., 2002) and assumes at least part-time employment, we can assume that the employable subjects will be paid at least US$2520 per year. Furthermore, based on the Meyer et al. (2002) study, we could speculate that 21% of the ‘employable’ group will have competitive employment 12% agency-contracted individual placement, 46% agency-contracted group placement and 21% in agency-run business. All jobs would pay minimum wages and the agency-contracted ones will be through a contractual agreement between the mental health or rehabilitation agency and the employer. Besides the benefit of employment the employed patients would likely incur savings in informal caregiver time of over US$4296 per year (Arno et al., 1999). This would be in addition to the many human benefits on the patients’ self-esteem, sense of discipline and overall ability to reintegrate into society. Although recent literature on cost-effectiveness analyses recommends using Quality Adjusted Life Years (QALYs) or a utility value as a measure of change in health status (numerator) we believe that ‘employability’ is a reasonably accurate measure of

changes in health status. This is because one of the major components reflecting change in health status is ‘production output’ or changes in patients’ productivity that is well captured in the measure of employability used in our study. Also as noted above, employability partially captures the effects on informal caregiver time use, which is substantial in schizophrenia. Some limitations of this study need to be addressed here. First and foremost, the reader should be aware that we do not claim that ‘employable’ persons would be gainfully ‘employed.’ That would depend to a large extent on several other factors, for example education, age, government regulations and the patients’ own personal priorities, which we cannot control. We tried to measure the number of patients who may not be limited by a cognitive disability to find gainful employment. Second, although the results of this analysis favor risperidone therapy, it must be noted that the WCST-Cat scores, which have an important influence on the results, were obtained from a relatively small trial of risperidone with 30 patients. Further, it should be noted that the studies relating WCST-Cat scores and work status show a large amount of overlap in scores between the employed and unemployed groups. We have tried to take a best estimate cut-off matching these results with the normative scores and further test the robustness of this cut-off by performing sensitivity analyses. Also, it should be noted that our study used a modest time horizon of 1 year, studies that have evaluated the impact of risperidone over longer time horizons have determined that risperidone may be cost-saving or cost-neutral (Davies et al., 1998). Naturally, our results cannot be extended beyond the 1-year time frame.

5. Conclusion This study found that outpatient risperidone treatment may increase the number of employable persons when compared to treatment with haloperidol at a cost of US$19,609 per each additional employable person. Risperidone was found to increase the number of employable persons and cost more than haloperidol across all ranges used in the sensitivity analysis. Given the increased productivity, decreases in informal caregiver time, and increased quality of life

R. Ganguly et al. / Schizophrenia Research 63 (2003) 111–119

associated with employment, risperidone use appears to be justified.

Acknowledgements We would like to thank Dr. Herbert Y. Meltzer (Division of Psychopharmacology, Vanderbilt University School of Medicine, Nashville, TN) and Dr. Susan R. McGurk (Mount Sinai School of Medicine, New York, NY) for sharing a prepublished version of their research entitled ‘‘The role of cognition in vocational functioning in schizophrenia’’ with us.

References Arno, P.S., Levine, C., Memmott, M.M., 1999. The economic value of informal caregiving. Health Aff. (Millwood, Va.) JID 8303128 18, 182 – 188. Birchwood, M., Jackson, C., 2001. Schizophrenia. Psychology Press, East Sussex, UK, p. 110. Davies, A., Langley, P.C., Keks, N.A., Catts, S.V., Lambert, T., Schweitzer, I., 1998. Risperidone versus haloperidol: II. Costeffectiveness. Clin. Ther. JID - 7706726 20, 196 – 213. Department of Human Resources, 1995. Hospital and budget utilization report for state operated psychiatric hospitalizations for financial year 1995. Foster, R.H., Goa, K.L., 1998. Risperidone. A pharmacoeconomic review of its use in schizophrenia. Pharmacoeconomics JID 9212404 14, 97 – 133. Glazer, W.M., Ereshefsky, L., 1996. A pharmacoeconomic model of outpatient antipsychotic therapy in ‘revolving door’ schizophrenic patients. J. Clin. Psychiatry JID - 7801243 57, 337 – 345. Goldman, R.S., Axelrod, B.N., Tandon, R., Berent, S., 1991. Analysis of executive functioning in schizophrenics using the Wisconsin card sorting test. J. Nerv. Ment. Dis. JID - 0375402 179 (8), 507 – 508. Hamilton, S.H., Edgell, E.T., Revicki, D.A., Breier, A., 2000. Functional outcomes in schizophrenia: a comparison of olanzapine and haloperidol in a European sample. Int. Clin. Psychopharmacol. 15 (5), 245 – 255. Heaton, R.K., 1981. Wisconsin Card Sorting Test Manual. Psychological Assessment Resources, Odessa, FL. Jaeger, J., Douglas, E., 1992. Neuropsychiatric rehabilitation for persistent mental illness. Psychiatr. Q. JID - 0376465 63, 71 – 94.

119

Keks, N.A., 1997. Impact of newer antipsychotics on outcomes in schizophrenia. Clin. Ther. JID - 7706726 19, 148 – 158. Lysaker, P., Bell, M., Beam-Goulet, J., 1995. Wisconsin card sorting test and work performance in schizophrenia. Psychiatry Res. JID - 7911385 56 (1), 45 – 51. McGurk, S.R., Meltzer, H.Y., 2000. The role of cognition in vocational functioning in schizophrenia. Schizophr. Res. JID 8804207 45, 175 – 184. Meltzer, H.Y., 1999. Outcome in schizophrenia: beyond symptom reduction. J. Clin. Psychiatry JID - 7801243 60 (Suppl. 3), 3 – 7. Meltzer, H.Y., McGurk, S.R., 1999. The effects of clozapine, risperidone, and olanzapine on cognitive function in schizophrenia. Schizophr. Bull. JID - 0236760 25, 233 – 255. Meltzer, H.Y., Thompson, P.A., Lee, M.A., Ranjan, R., 1996. Neuropsychologic deficits in schizophrenia: relation to social function and effect of antipsychotic drug treatment. Neuropsychopharmacology JID - 8904907 14, 27S – 33S. Meyer, P.S., Bond, G.R., Tunis, S.L., McCoy, M.L., 2002. J. Clin. Psychiatry 63, 108 – 116. National Physician Fee Schedule Relative Value Scale, 1999. Obtained from the Health Care Financing administration (HCFA) website public use files, http://www.hcfa.gov/stats/pufiles.htm. Peuskens, J., 1995. Risperidone in the treatment of patients with chronic schizophrenia: a multi-national, multi-centre, doubleblind, parallel-group study versus haloperidol. Risperidone Study Group. Br. J. Psychiatry JID - 0342367 166, 712-726. Red Book, 1999. Drug Topics Red Book. Medical Economics, Montvale, NJ. Rossi, A., Mancini, F., Stratta, P., Mattei, P., Gismondi, R., Pozzi, F., Casacchia, M., 1997. Risperidone, negative symptoms and cognitive deficit in schizophrenia: an open study. Acta Psychiatr. Scand. JID - 0370364 95, 40 – 43. SAS Institute, SAS 8.02. (8.02), 2002. Cary, NC, USA, 2001. Sevy, S., Davidson, M., 1995. The cost of cognitive impairment in schizophrenia. Schizophr. Res. JID - 8804207 17, 1 – 3. Sweeney, J.A., Haas, G.L., Keilp, J.G., Long, M., 1991. Evaluation of the stability of neuropsychological functioning after acute episodes of schizophrenia: one-year follow-up study. Psychiatry Res. JID - 7911385 38, 63 – 76. Torrance, G.W., Siegel, J.E., Luce, B.R., 1996. Framing and designing the cost-effectiveness analysis. In: Cost-effectiveness in Health and Medicine. Oxford University Press, New York, pp. 54 – 79. TreeAge Software, Data 3.5. 2002. Williams, C.L., Johnstone, B.M., Kesterson, J.G., Javor, K.A., Schmetzer, A.D., 1999. Evaluation of antipsychotic and concomitant medication use patterns in patients with schizophrenia. Med. Care JID - 0230027 37 (4 Suppl Lilly), AS81 – AS86.