Matthew Gitlin, PharmD

The chronic nature of schizophrenia is a major contributor to overall costs of treatment, which are $33 to $65 billion a year in the United States alone.

 Current estimates of total direct costs per patient vary widely, from $3700 to $57,000 per year depending on insurance status, medications available, data source used, and comorbidities.

 The biggest contributor to the direct cost of treating schizophrenia was previously thought to be inpatient and outpatient services. However, drug costs now stake this claim due to the increased use of atypical antipsychotics. Previous guidelines suggested atypical antipsychotics as second-line treatment for use after typical antipsychotics. But current practice guidelines suggest atypical antipsychotics as first-line treatment, then switching to another antipsychotic if the response is inadequate. If there still is no response, typical antipsychotics or clozapine are recommended before treatment with combination drugs, including mood stabilizers.

 There is little information on drug costs within cost-of-illness studies and even less information for privately insured-patients, who make up 35% of all schizophrenia patients.

The objectives of this study were to determine schizophrenia-related direct costs and total direct costs, and to examine drug use patterns and costs. We also compared costs and treatment patterns between newly and previously diagnosed patients.

Three years of claims data were obtained from a large privately insured population in California from June 2001 through May 2004. The analysis utilized insurance eligibility claims, medical claims, prescription claims, and all claims from a mental health carve-out benefit plan.

A total of 1012 persons with schizophrenia were identified through the presence of at least 1 inpatient or 2 outpatient medical claims specific to a schizophrenia diagnosis. Schizophrenia diagnoses included the following 

International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) 

codes: 295.xx (schizophrenia disorders) and V11.0 (history of schizophrenia illness). The date of the first inpatient claim meeting these requirements or the second outpatient claim was  designated as the index or starting date of the treatment period.

Patients were between 18 and 65 years of age at the start of study period, had continuous insurance eligibility for a minimum of 24 months, and had a 12-month preindex period and 12-month evaluation period (referred to as treatment period) from the index date. This left a final sample size of 415 persons, which was further broken down into newly diagnosed patients (n = 109), defined as those with no history of schizophrenia-related treatment or resource use for 12 months prior to the index date or no history of schizophrenia (

code V11.0) throughout the entire study period. Remaining patients were designated as previously diagnosed with schizophrenia (n = 305).

“Cost allowance” was used, defined as the cost of the service negotiated between the insurer and provider, which is considered the closest to actual costs. The last 4 months of claims data were removed to account for data loss due to normal delays in  claims processing. The first 12 months of data were defined as the preindex period and analyzed separately for newly and  previously diagnosed groups to ensure that no unusual cost pattern occurred prior to the treatment period. For newly diagnosed patients, the preindex period was before the schizophrenia diagnosis date (defined based on 1 inpatient or 2 outpatient claims on different dates). Postindex treatment period lengths were capped at 12 months to assess equivalent time periods for both the preindex and treatment periods.

Resource use was categorized into inpatient, outpatient, prescription, long-term care, and other services, including laboratory, home healthcare, and ambulance. Schizophrenia-specific costs were determined based on the presence of a claim with an  

diagnosis for schizophrenia or the use of schizophrenia medications. Mood stabilizers were included as a  schizophrenia-related cost. Claims associated with all other 

 codes were classified as costs not specific to schizophrenia, as were all other medications. Total costs included both schizophrenia-specific and nonspecific costs, because it is likely that some nonspecific costs may have been schizophrenia specific but were not coded as such. 

Mean costs per patient per year were separately assessed by type of service and type of medication for total direct versus schizophrenia-specific costs. Total direct and schizophrenia-specific costs were compared between newly and previously diagnosed patients by type of service per patient per year using t tests. Detailed annual medication costs and prescription fills were compared between newly and previously diagnosed patients by therapeutic class per patient per year using 

 tests. Patient demographics and the number of 

 diagnosis claims per user per year were tested across subgroups, also using

Regression analyses were used to determine what effect individual covariates had on schizophrenia-specific costs. A 1-part model was performed using ordinary least squares (OLS) regression. A log-transformed OLS and a generalized linear model with a gamma distribution and log-link function were also performed, but showed similar results and so are not presented here.

 Variables in the regression analysis included age, sex, status of diagnosis, and type of medical and prescription utilization.

We also provided an estimate of total costs for the estimated prevalence and incident US populations of privately insured individuals with schizophrenia based on our per-patient costs. We assumed a 1% population prevalence of schizophrenia and that 40% of this population was privately insured for this calculation.

The average age of patients was 36.6 years (SD 12.2 years) (

). Females, who represented 50.6% of the study sample, were significantly older than males (mean age 40.5 years [SD 12 years] and 32.5 years [SD 11.1 years], respectively; P <.05). The most common 

 codes on all patient claims were for schizoaffective disorder (295.7x) at 45.4%, paranoid type  schizophrenia (295.3x) at 33.6%, and unspecified schizophrenia (295.9x) at 24.2%, indicating the presence of a relatively severe  type of disorder in this population. These 3 diagnoses along with residual type schizophrenia (

 295.6) also were associated with the highest utilization of healthcare resources (25.5, 21.1, 20.3, and 19.4 claims per user per year, respectively). Comparison of patient characteristics between newly and previously diagnosed patients showed no significant differences.

For all patients combined (n = 414), total direct costs averaged $12,885 per patient per year (SD $17,082), withschizophrenia-specific costs averaging $6220 (SD $9987) or 48.3% of the total (Figure 1). This large difference between total direct costs and schizophrenia-related costs may be partly attributable to costs that were associated with schizophrenia, but were not coded as such. Inpatient services ($2762; SD $8451) were highest for schizophrenia-specific costs, followed closely by prescription drugs ($2479; SD $2767). This pattern continued for total direct costs, with inpatient and prescription costs representing the highest costs, followed closely by outpatient services (

Total direct costs for patients newly diagnosed with schizophrenia were numerically higher than those for patients  previously diagnosed with schizophrenia ($15,282 [SD $16,126] vs $12,029 [SD =$17,357] per patient per year; P = .09) (

). Total costs were significantly different between these groups in the preindex period, when newly diagnosed patients lacked schizophrenia-related services ($8139 [SD $11,727] vs $10,868 [SD $11,468] for newly and previously diagnosed patients, respectively; P = .03.)

Previously diagnosed ($3999; SD $3428) patients also had significantly higher prescription costs in the preindex period (presumably due to schizophrenia drug treatment) than the newly diagnosed ($2185; SD $2776) group. During the treatment period, these differences continued between groups, with costs of prescription drugs remaining significantly higher for previously  diagnosed patients, but all other cost categories and total costs remaining significantly higher for the newly diagnosed patients (Table 2). Inpatient services in the newly diagnosed group contributed the most to both total direct costs ($7745; SD $12,791) and schizophrenia-specific costs ($4222; SD $9010). In contrast, total inpatient costs were the second most costly category ($4400; SD $14,522) for the previously diagnosed group, representing about half the cost of that for the newly diagnosed group. For schizophrenia-specific costs, only prescription costs remained significantly higher in previously diagnosed patients than in newly diagnosed patients.

Previously diagnosed patients had similar costs in the preindex and treatment periods. But for the newly diagnosed, the total treatment period costs were significantly higher ($8319) than the total preindex period costs (P <.05). After a diagnosis of schizophrenia, inpatient, outpatient, and prescription costs were significantly higher than they were before diagnosis. Subtracting the preindex costs from the treatment period costs is one method of determining schizophrenia-related costs, and it was found that these cost differences were similar to our schizophrenia-related costs determined by 

Overall, 81.1% of the patients in the study sample used a schizophrenia medication, and 18.9% of patients received no schizophrenia medications throughout their treatment period (

). Many patients utilized more then 1 type of therapeutic drug class, but the majority of patients utilized atypical antipsychotics (73.9%), followed by mood stabilizers (31.9%) and typical antipsychotics (16.7%).

Atypical antipsychotics had the highest costs per user per year ($2215; SD $2547), while the remaining therapeutic classes contributed minimally to overall schizophrenia-related medication costs (Table 3). When looking at individual medications, clozapine costs per user per year were highest ($2396; SD $2269), followed by olanzapine ($2279; SD $2031), despite the former being the least utilized atypical antipsychotic and the latter having the highest utilization. Clozapine was most commonly prescribed as 2 separate prescriptions with 2 separate doses in the small subset of the population receiving it (n = 33). The average days of supply was 12.4, while all the remaining medications assessed had an average of 26.1 days of supply, explaining some of the high fill rates and costs for clozapine. These characteristics of clozapine use are likely due to the restrictive prescribing patterns and high risk of adverse effects associated with the drug, as it is also typically reserved for patients with treatment-resistant schizophrenia. All patients had relatively high rates of prescription fills (11.7 fills [SD 10.7] per user per year).

There was a significant difference in the costs of all schizophrenia drug categories between newly and previously diagnosed patients, except for mood stabilizers (

). Among patients in the previously diagnosed group, 82.6% utilized a schizophrenia medication while only 77.1% of newly diagnosed patients utilized a schizophrenia drug. This resulted in significantly lower costs

in newly diagnosed patients compared with previously diagnosed patients  ($2411 [SD $2195] vs $3269 [SD $2901]; P <.005;  

). Specifically, more previously diagnosed patients utilized an atypical antipsychotic (74.4% vs 72.5%) or a typical  antipsychotic (18% vs 12.8%) compared with newly diagnosed patients (

). More newly diagnosed than previously diagnosed patients used a mood stabilizer (37.6% vs 29.8%), although this difference was not significantly different. It appears that newly diagnosed patients may not have been as compliant as previously diagnosed patients because newly diagnosed patients had only an average of 11.7 prescription claims per user per year while previously diagnosed patients had 16.5 prescription claims per user per year (P = .0003). This trend was further reflected by the significant cost differences (P = .005)  between the 2 groups (Table 4A). 

We used OLS regression, log OLS, and generalized linear models to determine predictors for the differences in costs between newly and previously diagnosed patients. There were no differences in the results between these 3 approaches; therefore, only the results from the OLS regression are reported here, as this method seems to perform better, is more conservative, and is easier to interpret as mean costs (

). Overall, the OLS model explained about 42% (r 2 = 0.42) of the variation in average schizophrenia-specific costs per patient per year. The use of inpatient and outpatient services added significantly to costs, with inpatient services adding almost $12,220 and outpatient services adding $3399. Drug use also added significantly to costs, with atypical antipsychotics adding about $3419, conventional antipsychotics $2475, and mood stabilizers $2121.

Use of inpatient and outpatient services, use of atypical antipsychotics, use of conventional antipsychotics, and use of mood stabilizers were significantly associated with increased costs in all patients as well as in the subset of previously diagnosed patients. In newly diagnosed patients, use of inpatient and outpatient services and antipsychotic use were also significantly associated with an increase in cost, but use of mood stabilizers was not significantly associated with costs. This finding indicates that previously diagnosed patients were likely receiving more complex drug therapy than those who were newly diagnosed.

There was no significant difference between newly and previously diagnosed patients for each parameter in the regression. Only age was found to significantly affect the cost difference between the 2 groups (P = .037).

This study found that total direct costs for privately insured schizophrenia patients fell between the $6220 for schizophrenia-related costs and the $12,885 for total costs per patient per year. Schizophrenia-specific cost results in this study were lower than the  average yearly direct cost per patient of about $8000 found in 2 other studies of privately insured patients.

 However, the study used 2002 dollars, excluded patients older than 65 years, and estimated schizophrenia-related costs by subtracting costs of a matched cohort without schizophrenia.

 Because their cost was between our schizophrenia-related costs and total costs, it may be that their schizophrenia costs still included some comorbidity hospitalization costs (patients were matched on age, sex, and geographic region and not on comorbidities) and did not strictly exclude items based on diagnoses of schizophrenia as we did in this analysis. In addition, we used data from 2001-2004, when the use of atypical antipsychotics and mood stabilizers was more widespread, while Wu et al used 2002 data.

The use of atypical antipsychotics has been demonstrated to be associated with lower hospitalization costs; unlike the Wu et al study, our results reflect this.

 Crown et al also reported a cost of $8000 per patient per year using data from 1991-1993, when there was little atypical antipsychotic use.

 Although the cost of hospitalization was not specifically reported, it was demonstrated that for hospitalized patients the cost was $20,800 and for patients not hospitalized the cost was only $3200. Another study by McDonald et al from 2001-2002 found direct costs of treatment for schizophrenia, for both privately and publicly  insured patients, to be about $4000 per patient per year, which are lower than the costs reported here.

 McDonald et al also reported total treatment costs for schizophrenia patients that are similar to our total costs ($12,300 vs $12,885 per patient per year, respectively).

The variations in costs across all cost studies are likely related to several factors, such as the method of determining schizophrenia-specific costs, the particular characteristics of the population reflected in their insurance coverage, and whether the cost analyses were done before or after the introduction and gradual widespread use of atypical antipsychotics and mood stabilizers.

 The differences in and difficulty around defining which costs are schizophrenia specific and which are not is a major concern. For our study, we feel that our total and schizophrenia-specific costs are best viewed as a maximum and minimum range of the true costs of schizophrenia. For this diagnosis in particular, it may be common for a hospital admission or a physician visit to be attributed to a nonschizophrenia 

 diagnosis, despite actually being closely related to schizophrenia. Therefore, our  schizophrenia-specific cost estimates are very strict and may be underestimated, while our total costs include the costs of all comorbidities, some likely unrelated to schizophrenia, and thus may be overestimated. Therefore, schizophrenia costs are best viewed as falling within a range of schizophrenia-specific and total costs, as we present here.

A comparison of the costs between newly and previously diagnosed patients demonstrated an increase in treatment complexity for patients over time. For example, previously diagnosed patients utilized more types of medications. Also, 5.5% more newly diagnosed patients did not use any drug for schizophrenia compared with those who were previously diagnosed. Furthermore, newly diagnosed patients had 4.8 fewer prescription fills  per user per year than previously diagnosed patients. The lower medication use and costs in the newly diagnosed patients were likely partly due to a lack of patient compliance with schizophrenia medications, which indicates the importance of focusing on compliance in newly diagnosed patients.

The previously diagnosed patients had lower total costs and schizophrenia-related costs than the newly diagnosed patients despite the higher prescription drug costs. This higher drug cost in previously diagnosed patients is associated with much lower inpatient costs (by almost half) for both total and schizophrenia-related costs. These lower hospital costs may be due to a more complex medication regimen in previously diagnosed patients, with greater use of atypical antipsychotics including clozapine as well as the use of typical antipsychotics, a drug class reserved as secondary treatment after failure with initial medication therapies.

Schizophrenia treatment algorithms since 1991 indicate that monotherapy with atypical antipsychotics is the first line of therapy. By 2001 second-generation atypicals accounted for 92% of all expenditures on antipsychotics.

 Our data demonstrate that in this insured population the majority of patients are prescribed atypical antipsychotics, with very little use of typical antipsychotics. We also showed more costly medication use in the previously diagnosed patients, indicating that newly diagnosed patients were not compliant or were not being prescribed antipsychotic treatment as consistently as in previously diagnosed patients. The significantly higher medication cost for previously diagnosed patients also reflects treatment with combination therapy rather than monotherapy and is associated with lower hospital costs.

This study has all the limitations inherent in the use of retrospective claims data analyses. Paid costs were used as the proxy for economic costs to society but may not represent true opportunity costs, and we did not control for comorbidities. This study reported only direct costs of schizophrenia, which are only 48% of the total burden of schizophrenia, although this proportion may be different in a privately insured population.

 It should also be noted that the costs reported here only apply to those who are privately insured. Costs for publicly insured patients are likely significantly higher since this population includes the jobless, homeless, and others who are more severely ill. Our sample size is relatively small because of our strict inclusion criteria. Earlier analysis with more generous inclusion criteria gave similar results, but we were unsure whether we had true schizophrenia patients and those who were truly newly diagnosed. Therefore, we chose to select strict inclusion criteria, resulting in a smaller sample size but one that was better defined for this analysis. Finally our sample was from 2001 to 2004, and so did not include some of the newer long-acting atypical antipsychotics or intramuscularinjectables, which just recently were approved.

 However, that may not affect the treatment of the privately insured as much as a less adherent Medicaid population for which

We estimate that the direct costs of schizophrenia fall between our schizophrenia-specific costs of $6220 and our total costs of $12,885, supporting the conclusion of previous studies that total schizophrenia costs in this population continue to decrease over time with continued use of atypical antipsychotics despite their high costs, due to a corresponding decrease in hospital costs. The extrapolation of the schizophrenia-specific (minimum) and total (maximum) costs to the privately insured US population, assuming a 1% schizophrenia prevalence rate and assuming that 40% are privately insured, brings the annual direct treatment  costs for the privately insured schizophrenia population to anywhere between $8.3 and $17.2 billion.

 Assuming the incidence of schizophrenia is 100,000 in the United States, the total annual direct costs for the 40% of newly diagnosed patients who are privately insured would be $248.8 million to $515.4 million.

The cost difference we found for newly versus previously diagnosed patients indicates that more of an emphasis should be placed on stabilizing treatment and ensuring compliance in newly diagnosed patients. The higher cost of drug treatment and the lower cost of inpatient services in previously diagnosed patients signify that this group of patients may be appropriately receiving combination treatment regimens, are more in control of their condition, and are more stable.

A detailed examination of medication costs indicates that most newly diagnosed patients are appropriately being prescribed atypical antipsychotics, with more frequent use of last-line therapies in previously diagnosed patients. This study continues to demonstrate the high cost burden of treating schizophrenia even in a privately insured population. The costs of incident (or newly diagnosed) cases were found to be lower than the costs of prevalent (or previously diagnosed) cases, which can be used by managed care to better predict total direct schizophrenia costs.

Articles

Previously diagnosed schizophrenia patients have lower inpatient and higher drug costs compared with newly diagnosed patients, possibly indicating better control of their disease.