SEARCH

SEARCH BY CITATION

Keywords:

  • antipsychotics;
  • costs;
  • olanzapine;
  • optimal matching;
  • propensity score;
  • quetiapine;
  • resource utilization;
  • schizophrenia

ABSTRACT

  1. Top of page
  2. ABSTRACT
  3. Introduction
  4. Patients and Methods
  5. Results
  6. Discussion
  7. Conclusions
  8. References

Objective:  Compare annual health-care costs and resource utilization associated with olanzapine versus quetiapine for treating schizophrenia in a Medicaid population.

Methods:  Adult schizophrenia patients were selected from deidentified Pennsylvania Medicaid claims database (1999–2003). Included patients were continuously enrolled and initiated with olanzapine or quetiapine monotherapy after a 90-day washout period. Treatment costs were calculated for 1-year post-therapy initiation and inflation adjusted to year 2003. To control for selection bias, olanzapine and quetiapine patients were 1:1 matched using an optimal matching algorithm on propensity score, which was generated using logistic regression controlling for demographics, prior drug therapy, utilization, and costs. Treatment costs for the matched cohorts were compared directly, as well as using a difference-in-difference analysis.

Results:  A total of 6929 patients treated with olanzapine and 2321 with quetiapine met inclusion criteria. Quetiapine patients appeared more severe at baseline. After propensity score matching, 2321 patient pairs had similar baseline characteristics, including total costs. Compared with matched quetiapine patients, for the 1-year postindex period, olanzapine patients had similar drug costs ($6131 vs. $6014, P = 0.326), lower medical costs ($9897 vs. $11,218, P = 0.0128), and lower total health-care costs ($16,028 vs. $17,232, P = 0.0279). Lower psychiatric hospitalization costs account for most of the total cost difference. Difference-in-difference regression analysis confirmed olanzapine's economic advantage. Further adjusting for baseline variations, the total cost advantage of olanzapine patients was $962 (P = 0.032), and was mostly because of reduced psychiatric hospitalization costs of $992 (P = 0.004).

Conclusion:  Schizophrenia patients treated with olanzapine had lower total costs than quetiapine patients, mostly attributable to reductions in psychiatric hospitalization costs.


Introduction

  1. Top of page
  2. ABSTRACT
  3. Introduction
  4. Patients and Methods
  5. Results
  6. Discussion
  7. Conclusions
  8. References

Schizophrenia is a costly condition with an economic burden estimated at $62.7 billion in 2002 in the United States [1]. The major goals of current pharmacotherapy for schizophrenia are to achieve continuous relief from psychotic symptoms, to maximize patient functioning and quality of life, and to maintain recovery and prevent relapses. Atypical (or second generation) antipsychotics are recommended as the preferred therapy for schizophrenia by clinical guidelines in light of their better safety profile compared with typical (or first generation) antipsychotics (e.g., lower incidence of extrapyramidal symptoms) [2]. In 2003, atypical antipsychotics were prescribed to 71% of patients treated with antipsychotics [3]; risperidone, olanzapine, and quetiapine were the most frequently prescribed atypical agents [3].

The Clinical Antipsychotic Trials in Intervention Effectiveness (CATIE) study showed that olanzapine, compared with other atypical antipsychotics, had longer time to all-cause discontinuation [4,5]. Nevertheless, systematic reviews of randomized clinical trials and more recent published trials of schizophrenia patients did not show consistent benefits for olanzapine compared with other atypicals, especially quetiapine [6–13]. Overall, the evaluation of the efficacy and safety of atypicals are multidimensional in terms of adherence, control of positive and negative symptoms, extrapyramidal adverse events, and other safety profile, such as weight gain, and lipid profiles. Based on the mixed findings regarding this wide array of clinical and safety end points, there is no single atypical that is superior to others across all endpoints.

Clinical trials do not commonly compare cost outcomes among different atypicals. In contrast, observational studies in a real-world setting have been conducted to assess the economic consequences of the use of different antipsychotics. A number of published observational studies compared the total treatment costs between olanzapine and risperidone. Based on comprehensive literature reviews of Liu et al. (2004) and Hargreaves and Gibson (2005), there is no conclusive evidence to distinguish the total costs associated with these two therapies [14,15]. Some studies have analyzed the resource utilization associated with different antipsychotic medications and yielded inconclusive findings regarding olanzapine's overall economic outcomes compared with other atypical antipsychotics [16–18]. Nevertheless, the economic comparison between olanzapine and quetiapine is relatively rare. To the best of our knowledge, no published study compared the total costs of care between olanzapine and quetiapine directly for schizophrenia patients in the United States. Three articles presented relevant information on economic outcomes with both olanzapine and quetiapine. Gianfrancesco et al. (2006a, 2006b) compared the hospitalization rates across major atypical antipsychotics and found no significant differences between olanzapine and quetiapine using two different claims databases [19,20]. Both studies have relatively small sample sizes. A recently published study using California Medicaid database compared olanzapine, risperidone and quetiapine versus conventional antipsychotics, and found that olanzapine and risperidone, not quetiapine, were associated with reductions in total costs relative to typical antipsychotics [21]. Numerically, olanzapine had more favorable costs outcomes, including psychiatric hospitalization cost reduction, which is not consistent with findings from the Gianfrancesco et al. (2006a, 2006b) studies. In addition, a cost-effectiveness analysis based on the CATIE study found that the total treatment costs were lower with perphenazine than with atypical antipsychotics, and among atypical antipsychotics, olanzapine appeared to be associated with the lowest overall costs despite higher drug costs [22].

In light of the paucity in the literature with regard to observational studies comparing the economic outcomes of olanzapine and quetiapine, the objective of this research is to compare the annual health-care costs and resource utilization associated with olanzapine versus quetiapine for treating schizophrenia in a Medicaid population.

We used a Medicaid database for this research because we consider the research question especially important to Medicaid. State Medicaid programs are a major payer for all atypical antipsychotics prescriptions, which is a highly costly therapeutic class and consumes a sizeable portion of the total pharmacy budget for state Medicaid. Given the impact of this drug class on Medicaid budget and differential safety and efficacy profiles across different atypical antipsychotics, it is essential for the policymakers to comprehensively evaluate the economic outcomes associated with different drugs, particularly the medical cost offset, such as a reduction in hospitalizations. Combined with a sound methodology, conducting a comparison of economic outcomes using a Medicaid database provides additional insights beyond clinical trials, and is likely to provide useful information for policymakers, as well as state government health management organizations, to make informed policy decisions.

Patients and Methods

  1. Top of page
  2. ABSTRACT
  3. Introduction
  4. Patients and Methods
  5. Results
  6. Discussion
  7. Conclusions
  8. References

Data

The data for this study consist of the Pennsylvania Medicaid database (1999–2003), which includes deidentified information on patients' enrollment history, outpatient pharmacy claims, behavioral health (BH) and general medical claims, including diagnosis and procedures, physician visits, and hospitalizations. This database includes approximately 2 million beneficiaries per year. These data cover all Medicaid beneficiaries: low-income families/children, people with disabilities, long-term care individuals, and supplemental coverage for low-income Medicare beneficiaries. Data are available for both fee for service (FFS) and managed care organization (MCO) sources of claims. We conducted extensive data quality checks (e.g., to remove duplicate claims) to insure that the data elements and observations are complete and analytically meaningful.

Study Sample and Characteristics

All study sample patients were diagnosed with schizophrenia (ICD-9-CM: 295.xx). To ensure specific selection of schizophrenia patients, we used the behavior health (BH) file only to identify those patients who had either an inpatient hospitalization claim or two independent noninpatient medical claims with schizophrenia diagnosis incurred on different dates. In addition, patients were included if they had at least one qualified index prescription that meets all the following criteria: 1) patients received either olanzapine or quetiapine prescription (the index prescription) during June 2000 to June 2002; 2) before the index prescription of olanzapine or quetiapine, patients had a washout period of at least 90 days, during which patients did not receive either study medication, although prescriptions of other antipsychotics were allowed; 3) the index prescription was a monotherapy, though overlap with other antipsychotics was allowed if the other drug was filled before this prescription date; 4) a minimum of one monotherapy prescription of index drug and the monotherapy period of 30 days or longer were required. Included patients were 18 to 64 years of age as of the index prescription date and were continuously enrolled for at least 1 year before and 1 year after the index prescription date. Because outpatient pharmacy claims are not complete for the MCO source, patients who had MCO claims on or after the index prescription date were excluded. We did not exclude schizophrenia patients with bipolar diagnosis because we wanted to include the entire schizophrenia population in the analysis. Based on the first atypical antipsychotic received (i.e., olanzapine or quetiapine), patients were classified as either an olanzapine user or a quetiapine user regardless of their subsequent switching pattern. Therefore, this study analyzed the treatment outcomes associated with the initial atypical psychotics prescribed. Patients who were on both olanzapine and quetiapine at initiation were excluded.

Comorbidities were defined using claims with International Classification of Diseases, 9th Revision, clinical Modification (ICD-9-CM) diagnosis codes, including 16 mental health comorbidities (defined as the first three digits of the ICD-9 diagnosis codes from 290 to 316), and 8 major physical comorbidities, which were selected using the first three digits of ICD-9 codes 250 to 259, 270 to 279, 401 to 405, 410 to 414, and 428, and which had greater than 1% prevalence in our study sample. In addition, the number of mental comorbidities and the Deyo-Charlson Comorbidities Index for each study patient were also calculated [23].

Study Measures

The analysis compared resource utilization and costs between the two treatment groups over the 12-month post-index period. Medical resource utilization included the proportion of patients with an emergency room visit and any inpatient hospitalization, distinguishing between services that were psychiatric and nonpsychiatric related. Among medical resource users, the incidence and number of inpatient admissions, inpatient days, and number of emergency visits were also analyzed by psychiatric-related and nonpsychiatric-related status. A medical claim is considered to be psychiatric related if it appears in the BH file, indicates psychiatry inpatient hospital as a place of service, or has a primary diagnosis code for a mental health condition.

Pharmacy resource utilization included the use of any antipsychotics, antidepressants (including tricyclics, monamine oxidase inhibitors (MAOIs) and second generation antidepressants), mood stabilizers, benzodiazepines, hypnotics, anxiolytics, antiparkinsonian medication, as well as the total number of days with no antipsychotics usage. The medication possession ratio (MPR) was calculated for patients who used antipsychotics. The MPR was calculated as the ratio of the total number of nonoverlapped days covered by antipsychotics during the year over 365. The number of unique antipsychotic drugs used and the days of simultaneous antipsychotic drugs use were also calculated.

Costs per patient were calculated as the amount reimbursed from the payer's perspective (i.e., Pennsylvania Medicaid) and were inflated to year 2003 USD using the medical care component of the Consumer Price Index. Medical service costs were broken down by type of service (psychiatric or nonpsychiatric) and by place of service (inpatient, office/outpatient, emergency department). Prescription drug costs were broken down by cost of psychotropic drugs and cost of index drug (i.e., olanzapine and quetiapine). Total health-care costs were calculated as the sum of medical service and drug costs.

Statistical Analysis Methods

Outcomes were first compared descriptively across the unmatched cohorts using Wilcoxon tests for continuous variables and chi-square tests for categorical variables. Although a regression model approach is often used to control for differences among study samples, it assumes a given specification form for the model, which may bias the study findings. In observational studies, selection bias can be a potential limitation in evaluating the outcomes of alternative interventions, because patients are not randomly assigned to different therapies. Therefore, to control observable selection bias and render two treatment cohorts more comparable at baseline, a propensity score matching method was used. Propensity scores (i.e., the probability of receiving a given treatment conditional on covariates) are used to reduce the dimensionality of covariates into a single index. Assuming that there are no unmeasured confounders, treatment effect can be estimated by conditioning on the estimated propensity score [24]. Propensity scores can be applied using four methods, which have been developed over the past two decades in biostatistics and econometrics: matching, stratification (e.g., binning), inverse weighting, and covariate adjustment [25].

Two major classes of matching algorithms are available for propensity score: greedy matching and optimal matching. Greedy matching finds the nearest neighbor without replacement, and results were shown to be sensitive to the order of matching [26]. Moreover, greedy matching often yields incomplete matches (i.e., certain patients in one group cannot find a match within an acceptable range in the other group), and, as a result, both groups will have unmatched subjects after the matching process. The incomplete matching limits the study findings' external validity because the findings and conclusions could not be applied to the unmatched sample, which are often very different in characteristics from the matched population. In contrast, optimal matching involves minimizing the global distance (or the average absolute propensity score distance) between matched pairs, and hence the order of matching is irrelevant, and at the same time, the balancing of differences in baseline characteristics of matched pairs is also optimized. In addition, because there is no clear cutoff threshold as in greedy matching, it can retain the maximal number of matched pairs, thereby minimizing the loss of sample size and improving the external validity of study findings. Though the computer algorithm can be complex, statistical routines are available [27,28].

The following process was applied to implement the propensity score method with optimal matching. First, patients' baseline characteristics were profiled during a 12-month preindex period. These characteristics (including demographics, mental and physical comorbidities, resource utilization and costs) were compared between the two study cohorts. Second, propensity scores (i.e., the probability of being in the quetiapine treatment group conditional on observed baseline covariates) were generated using logistic regressions and given patient baseline covariates, which were selected using backward selection. Third, optimal matching was applied to match olanzapine and quetiapine patients based on propensity score using a SAS macro algorithm (SAS Institute, Cary, NC) developed by the Mayo Clinic [27]. Finally, baseline characteristics were compared for the full and the matched cohorts over the 12-month preindex period. Once the propensity score matching process was completed, outcomes were compared using paired t tests for continuous variables, and McNemar's tests for categorical variables for matched pairs. Each cost component was studied separately.

Sensitivity Analysis

Two sensitivity analyses were conducted to test the robustness of the results. First, the monotherapy period was extended to include patients who had at least two index drug prescriptions and at least 60 days of monotherapy, and conducted similar analyses as the ones described above. Because longer monotherapy indicates a better response and tolerability to the index therapy, this sensitivity analysis allows investigating the extent to which the economic difference between the two drugs remains for patients who responded to initial therapy.

The second sensitivity analysis used a difference-in-difference regression analysis, where the changes in 1-year treatment costs were compared between the two cohorts. In combination with propensity score matching, the difference-in-difference regression analysis further controlled for baseline covariates to remove unbalanced differences after matching, if any, between the two treatment cohorts. Controlled baseline covariates include patient demographics (age, sex, race), MCO, index year, comorbidities and Charlson Comorbidity Index, prior medical resource utilization, and prior medications use patterns. The combination of propensity score method and covariate adjustment has been shown to increase the validity of the estimate [29]. Paired t tests were used to evaluate whether the difference-in-difference was significant.

Results

  1. Top of page
  2. ABSTRACT
  3. Introduction
  4. Patients and Methods
  5. Results
  6. Discussion
  7. Conclusions
  8. References

Baseline Characteristics

A total of 6929 olanzapine patients and 2321 quetiapine patients met the sample selection criteria. Table 1 shows the patient selection criteria and corresponding sample sizes. As shown in the baseline characteristics reported in Table 2 (column C), olanzapine patients were older (age 42.8 vs. 41.3, P < 0.0001), more likely to be male (52.8% vs. 39.9%, P < 0.0001), and less likely to be Caucasian than quetiapine patients. Olanzapine patients also appeared less severe in general, as evidenced by fewer mental health comorbidities (1.34 vs. 1.59, P < 0.0001), a lower Deyo-Charlson Comorbidity Index (0.49 vs. 0.55, P < 0.0001), and a lower likelihood to use any medical and pharmacy resources.

Table 1.  Sample selection
Sample selection criteriaNumber of unique patients
Patients with at least one inpatient claim with schizophrenia diagnosis (295.xx) in the behavioral health (BH) file, or at least two independent claims with schizophrenia in BH45,855
Patients who have at least one pharmacy claim for either quetiapine (Q) or olanzapine (O)29,265
Patients with a drug claim (Q or O) within the index window (June 2000–June 2002)22,167
Patients with at least one prescription of Q or O and 12 months of continuous eligibility before and after the prescription17,930
Patients who have no prescriptions of the potential index drug for 90 days (washout period) before a potential index drug prescription14,733
Patients who were of age 18 to 64 as of the potential index date12,768
Patients with potential index dates on which Q and O prescriptions were not filled on the same day12,660
Patients who have at least one monotherapy prescription(s) of the potential index drug and 30 days of monotherapy period10,442
 OlanzapineQuetiapine
Patients with no managed care organization claims on or after the index date, by index drug6,9292,321
Table 2.  Baseline characteristics for schizophrenia cohorts over the 12-month preindex period
 Quetiapine (A)Olanzapine full cohort (B)(C) = (A) vs. (B)Olanzapine matched cohort (D)(E) = (A) vs. (D)
N N P-value (before matching)N P-value (after matching)
  1. Antidepressants include tricyclics, MAOI, and second-generation antidepressants. MPR is calculated as the ratio of total number of days covered by antipsychotics during the year before index date divided by 365. If a day is covered by two prescriptions, it is not double counted. Continuous variables were compared using Wilcoxon test for unmatched samples and paired t-test for matched sample; categorical variables were compared using chi-square test for unmatched samples and NcNemar's test for matched sample.

  2. MAOIs, monoamine oxidase inhibitors; MPR, medication possession ratio.

 2,321 6,929  2,321  
I. Demographics        
  Age (mean, SD)41.310.642.810.5<0.000141.610.80.2854
  Male92639.9%3,65652.8%<0.000193240.2%0.8574
  Race    <0.0001  0.8864
  White1,52065.5%3,82755.2% 1,49364.3% 
  Black65828.3%2,54536.7% 67629.1% 
  Hispanic462.0%2213.2% 451.9% 
  Other974.2%3364.9% 1074.6% 
II. Comorbidities        
 Number of mental comorbidities (mean, SD)1.591.721.341.66<0.00011.601.760.7925
  Charlson Comorbidity Index0.551.180.491.27<0.00010.591.350.6656
III. Medical resource utilization        
 Any psychiatric hospitalization76132.8%1,82726.4%<0.000179834.4%0.2502
 Any nonpsychiatric hospitalization73831.8%1,79926.0%<0.000176733.0%0.3632
 Any inpatient admission within the past 30 days40417.4%1,01214.6%0.001242018.1%0.5388
 Any emergency visit1,18050.8%2,96742.8%<0.00011,18050.8%1.0000
IV. Pharmacy utilization        
 Any use of antipsychotics1,02043.9%1,58222.8%<0.00011,05945.6%0.2497
 Any use of mood stabilizers68029.3%1,17917.0%<0.000166328.6%0.5821
 Any use of benzodiazepines/hypnotics/anxiolytics69930.1%1,17517.0%<0.000168929.7%0.7485
 Any use of antiparkinsonian medication40917.6%74110.7%<0.000141017.7%0.9693
 Total days with no antipsychotics (mean, SD)29011833280.5<0.00012921110.6469
 Among users        
 MPR of antipsychotics (mean, SD)0.470.340.390.31<0.00010.440.310.0556
 Days of simultaneous use of antipsychotics (mean, SD)21.8856.8915.045.90.000518.3351.160.1570
V. Costs per patient        
 Medical service cost (mean, SD)$11,619$20,082$9,679$18,024<0.0001$11,384$19,2300.3325
 Drug cost (mean, SD)$2,070$3,598$964$2,566<0.0001$1,886$3,2770.3448
 Total costs (mean, SD)$13,689$20,455$10,643$18,442<0.0001$13,270$19,7060.1182

Study Outcomes

When generating propensity score using logistic regression with backward selection, 19 variables were selected in the final model. The following variables increased the probability of receiving quetiapine: being Caucasian, starting treatment in 2001, having the following comorbidities (episodic mood disorders, special symptoms or syndromes not elsewhere classified, diabetes mellitus, obesity and other hyperalimentation), having hospital inpatient admissions (both psychiatric and nonpsychiatric related), using antipsychotics, clozapine, and benzodiazepines, and having higher medical service costs.

The propensity score matching process selected 2321 olanzapine patients who were the closest in propensity score values to their counterparts in the quetiapine cohort. Results indicated that after matching, none of the differences between the two cohorts remained significant, and the quetiapine and matched olanzapine cohorts had similar demographics, baseline comorbidities, prior medications, resource utilization, and costs (Table 2, column E). Therefore, the two study groups were comparable at baseline.

The comparison of the two matched cohorts in the 12-month postindex period is reported in Table 3. During the 1-year follow-up period, the matched olanzapine patients had a lower rate of psychiatric hospitalization (28.8% vs. 34.0%, P = 0.001) and emergency visits (47.0% vs. 52.0%, P = 0.0007), and lower pharmacy utilization across all drug categories analyzed (all P < 0.05). They also used fewer antipsychotics on average compared with quetiapine patients (1.64 vs. 1.81, P < 0.0001), had fewer days with simultaneous antipsychotic use (47.8 vs. 66.9 days, P < 0.0001), and had on average 9.6 more days without any antipsychotics compared with quetiapine patients (114.1 vs. 104.5 days, P = 0.0009). Nevertheless, olanzapine patients had a lower MPR of all antipsychotics (0.69 vs. 0.71, P = 0.0003) compared with quetiapine patients.

Table 3.  Comorbidities and outcomes for matched schizophrenia cohorts defined over the 12-month postindex period
 OlanzapineQuetiapineP-value
N N 
  1. Antidepressants include tricyclics, MAOI, and second-generation antidepressants. McNemar's tests are used to compare categorical variables; paired t-tests are used to compare continuous variables, except for values calculated among users, which are compared using Wilcoxon tests.

  2. MAOIs, monoamine oxidase inhibitors; MPR, medication possession ratio.

 2321 2321  
I. Comorbidities     
  Mean number of mental comorbidities (mean, SD)1.451.691.511.660.1830
  Charlson Comorbidity Index0.611.380.601.310.6348
II. Medical resource utilization     
 Any psychiatric hospitalization66828.8%78934.0%0.0001
 Any nonpsychiatric hospitalization55223.8%57624.8%0.4148
 Any emergency visit109147.0%120652.0%0.0007
III. Pharmacy utilization     
 Any use of clozapine1064.6%1647.1%0.0003
 Any use of antidepressants150865.0%165671.3%<0.0001
 Any use of mood stabilizers120551.9%134357.9%<0.0001
 Any use of benzodiazepines/hypnotics/anxiolytics110447.6%121052.1%0.0020
 Any use of antiparkinsonian medication60125.9%67028.9%0.0214
 Among users     
  MPR of all antipsychotics (mean, SD)0.690.280.710.270.0003
  Days of simultaneous use of antipsychotics (mean, SD)47.894.066.9107.8<0.0001
  Number of antipsychotic drugs used (mean, SD)1.640.861.810.94<0.0001

In the 12-month postindex period, olanzapine patients had statistically significantly lower medical costs compared with quetiapine patients ($9897 vs. $11,218, P = 0.0128) and similar drug costs ($6131 vs. $6014, P = 0.3257) (Table 4). The differences in medical costs were primarily driven by lower psychiatric costs ($7352 vs. $9037, P = 0.0002), and, in particular, psychiatric hospitalization costs ($3149 vs. $4220, P = 0.0024). When excluding the cost of the index drug, patients treated with olanzapine had lower psychotropic drug costs compared with patients treated with quetiapine ($1828 vs. $2459, P < 0.0001). In total, olanzapine patients had statistically significantly lower annual total costs compared with matched quetiapine patients ($16,028 vs. $17,232, P = 0.0279).

Table 4.  Descriptive analysis of costs for matched schizophrenia cohorts over the 12-month postindex period
CostsOlanzapine (N = 2321)Quetiapine (N = 2321)Difference (olanzapine – quetiapine)P-value
MeanSDMeanSD
  1. P-values were generated by paired t-test.

Medical service costs$9,897$18,377$11,218$18,592($1,321)0.0128
 Psychiatric costs$7,352$14,282$9,037$16,904($1,685)0.0002
 Nonpsychiatric costs$2,546$11,587$2,181$8,344$3650.2209
 Psychiatric hospitalization costs$3,149$10,638$4,220$13,838($1,071)0.0024
 Nonpsychiatric hospitalization costs$814$7,024$530$2,617$2840.0668
Drug costs$6,131$4,236$6,014$4,180$1170.3257
 Psychotropic drug costs$4,788$3,309$4,609$3,295$1790.0591
 Psychotropic drug costs, excluding index drug$1,828$2,131$2,459$2,477($631)<0.0001
Total costs$16,028$19,182$17,232$19,162($1,204)0.0279

Results of the sensitivity analysis showed that these findings are relatively insensitive to design specifications. Medical services and total costs calculated in the responder analysis exhibited the same trend as in the core analysis, but were generally no longer statistically significantly different. The difference-in-difference analysis, controlling for any residual patients' characteristics differences using regression models, further confirmed that olanzapine patients incurred lower costs than matched quetiapine patients, and cost differences became even more statistically significant. After adjusting for baseline characteristics, the change in medical service costs between the study period and baseline was statistically significantly lower for olanzapine patients (decreasing by $2106) compared with quetiapine patients (which decreased by $869). That is, controlling for differences in baseline costs, treatment with olanzapine is associated with an additional reduction in costs of $1237 (P = 0.0046, Table 5). Similarly, during the 12-month study period, total costs increased less for olanzapine ($1473) than for quetiapine patients ($2435), yielding a net difference in total costs of $962 in favor of olanzapine compared with quetiapine (P = 0.032).

Table 5.  Regression-adjusted cost differences comparing matched schizophrenia cohorts
 Olanzapine (postindex – preindex)Quetiapine (postindex – preindex)Difference (olanzapine – quetiapine)P-value
  1. Ordinary least squares regression was used to control for baseline characteristics.

Least squares mean cost differences    
 Medical service cost($2106)($869)($1237)0.0046
  Psychiatric cost($2017)($587)($1430)0.0004
   Psychiatric hospitalization cost($1566)($574)($992)0.0043
 Drug cost$3578$3304$2740.0059
  Psychotropic drug cost$3097$2736$361<0.0001
  Psychotropic drug cost, excluding index drug$664$965($301)<0.0001
 Total costs$1473$2435($962)0.0320

Discussion

  1. Top of page
  2. ABSTRACT
  3. Introduction
  4. Patients and Methods
  5. Results
  6. Discussion
  7. Conclusions
  8. References

This study used a retrospective cohort design to compare the health-care resource utilization and treatment costs for schizophrenia patients initiated on olanzapine versus those who were initiated on quetiapine. To control for observed selection bias, a propensity score matching method was used with an optimal matching algorithm. The comparability of the quetiapine cohort to the matched olanzapine cohort shows the impact of optimal matching: although before matching, the quetiapine and full olanzapine cohorts had substantially different baseline characteristics, the matching process generated two balanced cohorts. Nevertheless, unlike randomized clinical trials, using a propensity score method in an observational study can only control for observed covariates, and there maybe unobserved factors that could affect both the choice of atypical drugs and the study outcomes. Therefore, unobserved selection bias may still remain in the study and confound the study findings.

This study found that based on two comparable cohorts at baseline, schizophrenia patients covered by the Pennsylvania Medicaid program treated with olanzapine had lower medical and pharmaceutical resource utilization and were less costly than patients treated with quetiapine in the 12 months following the index drug prescription.

Specifically, in the 12-month study period, medical service costs were $1321 lower for patients treated with olanzapine than for those treated with quetiapine, of which $1071 were because of lower psychiatric hospitalization costs. Total drug costs were not significantly different for patients in both treatment groups (P = 0.33), but when excluding the index drugs, olanzapine patients had $631 lower drug costs compared with patients treated with quetiapine. Overall, the total health-care costs in the 12-month study period were $1204 lower for olanzapine patients than for quetiapine patients. Therefore, olanzapine's economic advantage in other psychotropic drug costs and psychiatric hospitalization costs more than offset the higher acquisition costs of the drug, yielding lower total treatment costs compared with quetiapine.

Higher rates for both psychiatric hospitalization and emergency room visits were observed among patients in the quetiapine group, as well as higher psychiatric hospitalization costs. Both of these urgent care resource uses indicate that quetiapine patients were probably more likely to experience relapses [17]. In addition, quetiapine patients used more concomitant psychiatric medications from each of the classes we examined, including more antipsychotics compared with olanzapine patients. Nevertheless, quetiapine patients showed a slightly higher MPR compared with olanzapine patients. A detailed treatment pattern outcome comparison between olanzapine and quetiapine is covered in a separate study [18].

The trend in the results observed was confirmed by the sensitivity analyses conducted and was consistent with some past studies [30,31]. Some studies did not find differences between quetiapine and olanzapine because of smaller sample size and noncomparability of the study cohorts [19,20]. In addition, the two studies analyzed patients who had at least 60 days of monotherapy, while the study group investigated in this study was requested to have at least 30 days of monotherapy, which did not force patients to be responders. Results of the sensitivity analysis conducted among patients with 60 days of monotherapy indicate that olanzapine patients had lower total costs than matched quetiapine patients, although to a lesser extent than in the core analysis. This implies that part of the total economic benefit of olanzapine is due to a better response profile of olanzapine, which is consistent with clinical findings in the CATIE study [4]. Nevertheless, an appropriate comparison of overall economic outcomes would consider the cost advantage due to different response and tolerability associated with the initial therapy of selection, as shown in this study. Moreover, quetiapine patients have numerically more mental comorbidities during the follow-up period than matched olanzapine patients (although not statistically significant). This indicates that patients treated with quetiapine are responding less to treatment than patients treated with olanzapine, which is also reflected by a higher utilization of psychotropic drugs for quetiapine patients during the follow-up period.

The results for olanzapine's overall economic advantage are consistent with findings from a cost-effectiveness study reported for the CATIE trial [32]. The CATIE trial also used an intent-to-treat analysis design and reported that the average monthly cost savings associated with olanzapine compared with quetiapine was $224 ($1433 vs. $1657), despite a higher antipsychotic medication costs. Similar to findings reported in our study, the majority of the CATIE-reported savings were because of the reduced cost in inpatient services ($556 vs. $753 monthly costs). The consistency of the results in our study with those of a randomized clinical trial further validated the propensity score optimal matching methodology employed in this study. The study results are also consistent with those of a recent cost analysis using California Medicaid database analyzing different types of episodes, which found that olanzapine was associated with lower total costs and psychiatric hospitalization costs than quetiapine, although this study did not perform statistical tests comparing the two drugs [33]. Nevertheless, Gianfrancesco et al. (2006b) found to the contrary that olanzapine was associated with a higher risk for hospitalization compared with quetiapine, although the difference was not statistically significant [34]. The Gianfrancesco study allowed a treatment gap of up to 90 days, which potentially included patients who used the drugs episodically. In addition, the study converted drug-specific dose to risperidone-equivalent milligrams, which has been recognized as “highly misleading”[34]. This is especially problematic when that variable is controlled in the regression, because as shown in the study, dose is a significant predictor of hospitalization risk. Finally, the study suffers from selection bias where the study cohorts are not comparable, yet no sufficient baseline information were considered and matched. A regression analysis controlling for limited baseline variables was used. Nevertheless, in contrast to propensity score matching, a conventional regression model would superimpose assumptions about the model specification form with noninteractive effects, thereby potentially biasing study findings when two cohorts are not comparable.

The economic comparison between olanzapine and quetiapine is consistent with their relative clinical effectiveness. In addition to the CATIE trial that found olanzapine to have better effectiveness as previously discussed, a meta-analysis summarized clinical trials comparing atypical antipsychotics, and concluded that olanzapine was significantly more efficacious than typical antipsychotics, whereas quetiapine did not have a significantly better efficacy [33].

Our study has several limitations. Costs estimates presented here have the common limitations associated with claims data-based studies, including the absence of detailed patient clinical symptom and severity information and characteristics of prescribing physicians. Although we attempted to control for selection bias by using propensity scores and adjusted for recorded patient characteristics, prior resource utilization and cost information, unobserved confounders (e.g., unobserved severity of illness) may potentially affect results. Regarding generalizability, relying on a single state's Medicaid data may not be representative of the overall US Medicaid population. In addition, the sample selection criterion of continuous eligibility during 2 years may limit our study sample to a subgroup of Medicaid beneficiaries that may differ from other Medicaid beneficiaries. Further, approximately 32% of schizophrenia patients are enrolled in the Medicaid program, and results of this Medicaid data analysis may not be generalizable to all schizophrenia patients [32]. The data we used is from 1999 to 2003, which may also affect the external validity of the study results, because prescribing practice could change over time, which may influence the economic outcomes. By using a propensity score matching method, we excluded a large proportion of olanzapine patients. As a result, the olanzapine cost advantages can only be applicable to olanzapine patients who are clinically similar to the olanzapine matched cohort, and cannot be extended to the unmatched olanzapine patients who were excluded from the study. Nevertheless, the outcomes comparison may not be applicable among these excluded olanzapine patients because they were unlikely to receive quetiapine in the real world because they had substantially different baseline characteristics from those of the quetiapine patients. Although this propensity score matching approach allowed the two study cohorts to be comparable, unlike randomized clinical trials, this method does not balance unobserved factors and confounders. Furthermore, we could not use covariates (such as comorbid conditions) that were used to estimate the likelihood of cohort assignment, and therefore, we could not estimate the effect of those covariates on medical costs [35]. Moreover, because the cost analysis only covered 12 months, the cost savings associated with olanzapine treatment might be more or less important if observed over a longer period of time. Future research is warranted to extend the comparison between olanzapine and quetiapine to a longer period of time. In addition, this study could not estimate the causal effect of the two drug therapies, but rather, it could only estimate the economic outcomes associated with treatment initiation with olanzapine versus quetiapine.

Finally, although both olanzapine and quetiapine were approved for the treatment of bipolar conditions in recent years, this study considered only schizophrenia patients; therefore, estimating the resource utilization and costs of bipolar patients requires further research using more recent data.

Conclusions

  1. Top of page
  2. ABSTRACT
  3. Introduction
  4. Patients and Methods
  5. Results
  6. Discussion
  7. Conclusions
  8. References

This study demonstrated that in the Pennsylvania Medicaid schizophrenia population, after applying optimal matching algorithm on propensity score, initiation of olanzapine was associated with lower medical resource utilization, in particular psychiatric hospitalizations, and lower utilization of other concurrent psychotropic drugs. Although the acquisition costs of olanzapine are higher than those for quetiapine, the beneficial impact of olanzapine on schizophrenia patients' medical and pharmaceutical resource utilization more than offsets the drug's acquisition costs. Therefore, schizophrenia patients treated with olanzapine are less costly than those treated with quetiapine after controlling for differences in patients' characteristics. Most of the economic benefit was attributable to significant reductions in medical resource utilization and costs of psychiatric hospitalizations and other psychotropic drugs.

Source of financial support: This study was presented at the 2007 American Psychiatric Association Annual Meeting, San Diego, CA, May 19–24, 2007, and was supported by funding from Eli Lilly and Company.

References

  1. Top of page
  2. ABSTRACT
  3. Introduction
  4. Patients and Methods
  5. Results
  6. Discussion
  7. Conclusions
  8. References
  • 1
    Wu EQ, Birnbaum HG, Shi L, et al. The economic burden of schizophrenia in the United States in 2002. J Clin Psychiatry 2005;66:11229.
  • 2
    American Psychiatric Association. Practice Guideline for the Treatment of Patients with Schizophrenia (2nd ed.). Arlington, VA: American Psychiatric Association, 2004.
  • 3
    Aparasu RR, Bhatara V. Antipsychotic use and expenditure in the United States. Psychiatr Serv 2006;57:1693.
  • 4
    Stroup TS, Lieberman JA, McEvoy JP, et al. Effectiveness of olanzapine, quetiapine, risperidone, and ziprasidone in patients with chronic schizophrenia following discontinuation of a previous atypical antipsychotic. Am J Psychiatry 2006;163: 61122.
  • 5
    Lieberman JA, Stroup TS, McEvoy JP, et al. Effectiveness of antipsychotic drugs in patients with chronic schizophrenia. N Engl J Med 2005;353:120923.
  • 6
    Jayaram MB, Hosalli P, Stroup S. Risperidone versus olanzapine for schizophrenia. CD005237. Cochrane Database Syst Rev 2006;19:CD005237.
  • 7
    Davis JM, Chen N, Glick ID. A meta-analysis of the efficacy of second-generation antipsychotics. Arch Gen Psychiatry 2003;60:55364.
  • 8
    Furtado VA, Srihari V. Atypical antipsychotics for people with both schizophrenia and depression. CD005377. Cochrane Database Syst Rev 2008;1:CD005377.
  • 9
    Rummel C, Hamann J, Kissling W, Leucht S. New generation antipsychotics for first episode schizophrenia. Article no. CD004410. Cochrane Database Syst Rev 2003;4:CD004410.
  • 10
    Sirota P, Pannet I, Koren A, Tchernichovsky E. Quetiapine versus olanzapine for the treatment of negative symptoms in patients with schizophrenia. Hum Psychopharmacol Clin Exp 2006;21: 22734.
  • 11
    Sacchetti E, Valsecchi P, Parrinello G. A randomized, flexible—dose, quasi-naturalistic comparison of quetiapine, risperidone, and olanzapine in the short-term treatment of schizophrenia: the QUERISOLE trial. Schizophr Res 2008;98:5565.
  • 12
    Riedel M, Muller N, Spellmann I, et al. Efficacy of olanzapine versus quetiapine on cognitive dysfunctions in patients with an acute episode of schizophrenia. Eur Arch Psychiatry Clin Neurosci 2007;257:40212.
  • 13
    Voruganti LP, Awad AG, Parker G, et al. Cognition, functioning and quality of life in schizophrenia treatment: Results of a one-year randomized controlled trial of olanzapine and quetiapine. Schizophr Res 2007;96:14655.
  • 14
    Liu GG, Sun SX, Christensen DB, Luo X. Cost comparisons of olanzapine and risperidone in treating schizophrenia. Ann Pharmacother 2004;38:13441.
  • 15
    Hargreaves WA, Gibson PJ. Effectiveness and cost of risperidone and olanzapine for schizophrenia: a systematic review. CNS Drugs 2005;19:393410.
  • 16
    Liu GG, Sun SX, Christensen DB, Zhao Z. Cost analysis of schizophrenia treatment with second-generation antipsychotic medications in North Carolina's Medicaid program. J Am Pharm Assoc 2007;47:7781.
  • 17
    Tunis SL, Ascher-Svanum H, Stensland M, Kinon BJ. Assessing the value of antipsychotics for treating schizophrenia: the importance of evaluating and interpreting the clinical significance of individual service costs. Pharmacoeconomics 2004;22:18.
  • 18
    Yu AP, Ben-Hamadi R, Birnbaum HG, et al. Comparing the treatment patterns of schizophrenia patients treated with olanzapine and quetiapine in Pennsylvania Medicaid population. Current Medical Research Opinion. In press.
  • 19
    Gianfrancesco F, Wang RH, Pesa J, Rajagopalan K. Hospitalisation risks in the treatment of schizophrenia in a Medicaid population: comparison of antipsychotic medications. Int J Clin Pract 2006;60:141924.
  • 20
    Gianfrancesco F, Rajagopalan K, Wang RH. Hospitalization risks in the treatment of schizophrenia: comparison of antipsychotic medications. J Clin Psychopharmacol 2006;26:4014.
  • 21
    Chen L, McCombs JS, Park J. The impact of atypical antipsychotic medications on the use of health care by patients with schizophrenia. Value Health 2008;11:3443.
  • 22
    Rosenheck RA, Leslie DL, Sindelar J, et al. Cost-effectiveness of second-generation antipsychotics and perphenazine in a randomized trial of treatment for chronic schizophrenia. Am J Psychiatry 2006;163:20809.
  • 23
    Deyo RA, Cherkin DC, Ciol MA. Adapting a clinical comorbidity index for use with ICD-9-CM administrative databases. J Clin Epidemiol 1992;45:6139.
  • 24
    Rosenbaum PR, Rubin DB. The central role of the propensity score in observational studies for causal effects. Biometrika 1983;70:4155.
  • 25
    Austin PC, Mamdani MM. A comparison of propensity score methods: a case-study estimating the effectiveness of post-AMI statin use. Stat Med 2006;25:2084106.
  • 26
    Rosenbaum PR. Observational Studies. (2nd ed.). New York: Springer Verlag, 2002.
  • 27
    Bergstralh EJ, Kosanke JL, Jacobsen SJ. Software for optimal matching in observational studies. Epidemiology 1996;7:3312.
  • 28
    Ho DE, Imai K, King G, et al. Nonparametric preprocessing for parametric causal inference. Available from: http://gking.harvard.edu/matchit/docs/matchit.pdf[Accessed June 25, 2007.
  • 29
    Little RJ, Rubin DB. Causal effects in clinical and epidemiological studies via potential outcomes: concepts and analytical approaches. Annu Rev Public Health 2000;21:12145.
  • 30
    Rosenheck RA, Leslie DL, Sindelar J, et al. Cost-effectiveness of second-generation antipsychotics and perphenazine in a randomized trial of treatment for chronic schizophrenia. Am J Psychiatry 2006;163:20809.
  • 31
    Gibson PJ, Damler R, Jackson EA, et al. The impact of olanzapine, risperidone, or haloperidol on the cost of schizophrenia care in a medicaid population. Value Health 2004;7:2235.
  • 32
    Wu EQ, Shi L, Birnbaum H, et al. Annual prevalence of diagnosed schizophrenia in the USA: a claims data analysis approach. Psychol Med 2006;36:153540.
  • 33
    Davis JM, Chen N, Glick ID. A meta-analysis of the efficacy of second-generation antipsychotics. Arch Gen Psychiatry 2003;60: 55364.
  • 34
    Hay JW. Appropriate econometric methods for pharmacoeconometric studies of retrospective claims data: an introductory guide. J Manag Care Pharm 2005;11:3448.
  • 35
    Winkelmayer WL, Kurth T. Propensity score: help or hype? Nephrol Dial Transplant 2004;19:167173.