PRACTICE REPORTS Polypharmacy
PRACTICE REPORTS
Polypharmacy with antipsychotic drugs in patients with schizophrenia: Trends in multiple health care systems FangFang Sun, Eileen M. Stock, Laurel A. Copeland, John E. Zeber, Brian K. Ahmedani, and Sandra B. Morissette
S
chizophrenia is a complicated mental illness requiring lifelong treatment with antipsychotic drugs.1 Schizophrenia imparts social and cognitive impairment, pervasively affecting memory, attention, motor skills, executive function, and intelligence.2 Continuous use of antipsychotic drugs is recommended (for schizophrenia uniquely among the major mental illnesses). Although antipsychotic monotherapy is the standard approach for managing symptoms of schizophrenia,3-6 multiple antipsychotic medications may be prescribed. Several factors contribute to pharmacologic treatment patterns in caring for patients with schizophrenia, including patientlevel issues, provider prescribing
Purpose. Patterns of pharmacologic treatment in U.S. outpatients with schizophrenia across multiple health care settings were investigated. Methods. Antipsychotic drug utilization by patients with schizophrenia and related disorders was analyzed using data on 119,662 patients served by the Veterans Affairs (VA) health care system in fiscal years 2005–09, data on 5,440 enrollees in two health maintenance organizations (HMOs) in 2002–09, and National Ambulatory Medical Care Survey (NAMCS) data reflecting the experience of 17.6 million U.S. residents seeking care outside federal systems during the same eight-year period. Polypharmacy was defined as the use of more than one antipsychotic agent during one year (in the VA sample) or one week (in the HMO and NAMCS samples). The association of polypharmacy with hospital
FangFang Sun, M.S., is Health Services Researcher, Center for Applied Health Research, Temple, TX. Eileen M. Stock, Ph.D., is Research Scientist, Center for Applied Health Research, and Assistant Professor, College of Medicine, Texas A&M Health Sciences Center, Bryan. Laurel A. Copeland, Ph.D., is Interim Associate Chief of Staff of Research, Central Texas Veterans Health Care System, Temple, and Associate Director, Center for Applied Health Research, and Associate Professor, College of Medicine, Texas A&M Health Sciences Center. John E. Zeber, Ph.D., is Co-Director, Health Outcomes Core (jointly sponsored by Central Texas Veterans Health Care System and Scott & White Healthcare), Temple, and Associate Professor, College of Medicine, Texas A&M Health Sciences Center. Brian K. Ahmedani, Ph.D., LMSW, is Research Scientist, Center for Health Policy & Health Services Research, Henry Ford Health System, Detroit, MI. Sandra B. Morissette, Ph.D., is Assessment Core Chief, Veterans Affairs VISN 17 Center of Excellence for Research on Re-
728
Am J Health-Syst Pharm—Vol 71 May 1, 2014
admissions was assessed via multivariable logistic regression. Results. Rates of antipsychotic use in the VA sample ranged from 74% to 78%, with lower and more variable rates in the NAMCS sample (69–84%) and the HMO sample (22–67%). VA patients were found to have lower polypharmacy rates (20–22%) than patients in the HMO and NAMCS samples (19–31%). In all samples evaluated, polypharmacy was associated with an increased likelihood of hospital admission (odds ratio range, 1.4–2.4). Conclusion. A multisystem study revealed that antipsychotic use among patients with schizophrenia varied substantially among health care systems and that nearly one fifth of patients with schizophrenia or other psychotic disorders in most of the health care systems experienced antipsychotic polypharmacy. Am J Health-Syst Pharm. 2014; 71:728-38
turning War Veterans, Waco, TX, and Associate Professor, College of Medicine, Texas A&M Health Sciences Center. Address correspondence to Dr. Copeland (laurelacopeland@ gmail.com). Supported by the Center for Applied Health Research, Temple, TX, a research center jointly sponsored by Central Texas Veterans Health Care System and Scott & White Healthcare; and by grants from the Veterans Health Administration, Health Services Research and Development #IIR-09-335. The views expressed in this article are those of the authors and do not necessarily reflect the position or policy of the Department of Veterans Affairs. The authors have declared no potential conflicts of interest. DOI 10.2146/ajhp130471
PRACTICE REPORTS Polypharmacy
decisions, health-system culture, and organizational structure.7 Given the challenges and costs of treating individuals with schizophrenia, the prescribing trends of antipsychotic monotherapy and polypharmacy may vary across health care systems. A number of international guidelines recommend antipsychotic monotherapy using second-generation antipsychotics as first-line medications for the treatment of schizophrenia.4,8-13 While first-generation antipsychotics are acceptable in terms of efficacy, they may be avoided or considered second-line treatments because of their irreversible adverse effects (e.g., tardive dyskinesia). 14,15 However, antipsychotic polypharmacy, defined as the concurrent use of more than one antipsychotic drug for a single clinical condition, is on the rise.16,17 Indeed, since 1996, studies have shown that the rate of antipsychotic polypharmacy has increased among patients with schizophrenia, with the extent of polypharmacy varying by the populations under study, the treatment setting, and how antipsychotic polypharmacy was defined and measured.18 Specifically, antipsychotic polypharmacy among both outpatients and inpatients with schizophrenia has been reported at very high rates: up to 20% among patients receiving care in the Department of Veterans Affairs (VA) in fiscal year 2000,19 35% among outpatients in the public mental health system during the two-year period 1996–98,20 up to 40% among outpatients captured in 1998–2000 Medicaid claims data,16 and 38–55% among inpatients in Asian countries during the period 2001–09.21,22 Some plausible reasons for these figures include patient or clinician dissatisfaction with treatment effects, bothersome adverse effects, and undetected poor adherence leading to persistent symptoms that are incorrectly attributed to ineffective pharmacotherapy. Nonetheless, the prescribing of multiple antipsychotics may result
in adverse outcomes, such as an increased risk of metabolic syndrome23 and emergency department visits due to adverse drug reactions and psychiatric admissions,24 while also increasing direct treatment costs for the health care system, imposing an indirect societal burden, and resulting in impaired quality of life for the patient.23,25,26 Few studies have assessed the differential longitudinal prevalence of polypharmacy between health care systems for patients with schizophrenia. Most studies examined broad antipsychotic polypharmacy rates across countries, 27,28 though one older study compared rates of prescribing of combinations of psychotropic drugs targeting the same set of symptoms (those of acute paranoid schizophrenia) in California, New York, Pennsylvania, and Texas based on a survey of psychiatrists.29 To our knowledge, only two studies have compared pharmacotherapy for patients with schizophrenia across multiple types of health care systems.30,31 One of the two studies found that patients with schizophrenia treated in the VA health care system were more likely to receive an antipsychotic medication and more likely to be dosed above treatment recommendations than patients in the private sector. However, it did not appear that physical comorbidities were included for case-mix adjustment or that longitudinal pharmacotherapy patterns were measured.30 The second study, using data from 1994–96, found that outpatients receiving care in the VA system in two states were less likely to receive an antipsychotic medication than patients outside of this system.31 Despite these prior studies, no large-scale multisystem studies have investigated this issue in the last decade. With the growing national concern regarding increased polypharmacy among patients with schizophrenia, coupled with concern over the high cost of duplicative or non-evidence-based treatment (or
excess emergency department or inpatient care costs), new data are needed to inform policy and clinical decisions. In the study described here, we compared longitudinal data on antipsychotic prescribing and polypharmacy among patients diagnosed with schizophrenia in multiple health care systems representing three different organizational structures: (1) the nationwide VA system (2) nonfederal outpatient settings nationwide, and (3) two large not-for-profit health care systems in the Southwest and Great Lakes area of the United States. We also modeled one-year hospital admissions as a function of polypharmacy (versus monotherapy), adjusting for patient characteristics and disease burden. As an additional methodological aim, we examined how the availability of different amounts of data, depending on the source of information, affected the observed rates of antipsychotic prescribing and polypharmacy. Methods This study used a retrospective design relying on several sources of data, including a public-use data set and data from multiple health care systems for which approval of use for this research was obtained from the respective institutional review boards (IRBs). The study population consisted of patients who were 18–64 years of age at the time of diagnosis of schizophrenia or an unspecified psychotic disorder, as defined by the diagnosis codes (International Classification of Diseases, 9th Revision, Clinical Modification [ICD-9-CM]) listed in the medical record, specifically ICD-9-CM code series 295, “schizophrenic disorders” (excluding 295.5, “latent schizophrenia,” used for persons with symptoms but no psychotic episode before a schizophrenia diagnosis can be definitively assigned) and 298.9, “unspecified psychosis”; the latter code was included when it was determined that
Am J Health-Syst Pharm—Vol 71 May 1, 2014
729
PRACTICE REPORTS Polypharmacy
one health care system did not typically use the 295-series codes but used code 298.9 instead. The latent schizophrenia code was rare, appearing on the records of 135 of 119, 662 VA patients, all of whom also had other codes for schizophrenia; thus, it was determined that code 295.5 may have been used in error. No cases were noted in the other data sources. Elderly patients were excluded from the study population to reduce the impact of “burnout” in older patients with schizophrenia, as these patients typically have few positive symptoms (e.g., delusions, disordered thought, hallucinations) and thus derive little benefit from antipsychotics, may be at increased risk of death from antipsychotics, and are therefore less likely to be prescribed antipsychotics.32,33 Nationally representative sample. The public-use data set was obtained from the National Ambulatory Medical Care Survey (NAMCS) and represented nationwide nonfederal health care utilization among patients with schizophrenia from fiscal year 2002 (FY02, October 2001–September 2002) through FY09.34 NAMCS data contain information on nonfederal office-based physician visits obtained by sampling a set of clinics and requesting that those clinics report on a systematic random sample during a single week of patient visits. A total of 885 visits over eight years, representing the experience of more than 17 million community-based outpatients diagnosed with schizophrenia or a psychotic disorder, were identified from this national sample. Clinicians responding to this survey provided answers based on the past week at their practice. Pharmacy prescription fill data are not assessed by the NAMCS, but the survey captures clinician-reported prescriptions. National federal sample (VA). The study inclusion criteria targeted patients receiving VA care who had two or more outpatient dates of care for which a diagnosis of schizophrenia 730
or psychotic disorder was recorded with the same ICD-9-CM code. Data were only available for VA patients in the years 2005–09 due to IRBapproval issues. VA data were extracted from VA’s all-electronic medical records system for patients treated during the period FY05–FY09. For the VA cohort, we obtained data on all 119,662 veterans diagnosed with schizophrenia or psychotic disorder systemwide. Similar to the NAMCS data, the VA data reflected rolling enrollment (i.e., new patients entered the cohort each year, with some overlapping cases). The major difference between the VA and NAMCS data sets was that VA data captured year-round health care utilization and pharmacy fills per patient rather than utilization during just one week. It is unclear from NAMCS methods how far back into the patient chart the survey responses reach. Thus, for comparability we also assessed one week (June 21–28, per year analyzed) of service use for VA patients during each fiscal year in order to resemble NAMCS data, identifying 47,100 patients (the VA-Week sample). Private health care systems. For the private, not-for-profit health care systems, we used data from two sites (sites 1 and 2) submitting data to the Health Maintenance Organization Research Network (HMORN) Virtual Data Warehouse. HMORN is a federation of 19 care-and-coverage health care systems that have developed a set of standardized data definitions in order to create uniform health care measures from member data, primarily health care claims, from the year 2000 onward.35-37 We identified 699 patients at site 1 and 4887 patients at site 2 diagnosed with schizophrenia or psychotic disorder during the period FY02–FY09 (these sample sizes were insufficient for meaningful one-week look-back analyses, as were performed for the NAMCS and VA-Week samples). Measures. For each system of care represented by the sampled data
Am J Health-Syst Pharm—Vol 71 May 1, 2014
sets (NAMCS, VA, and HMORN), patient demographic characteristics were obtained: age at baseline, sex, race, and ethnicity. Age was used to create a variable representing a decade effect in the analytical models, and ethnicity was dichotomously defined as Hispanic or not Hispanic. Case-mix adjusters included the Charlson Comorbidity Index score and the Selim comorbidity index score (total and physical and mental component scores), with comorbidity indices calculated at a patient’s first year of entry into the applicable data set. The Charlson Comorbidity Index sums weighted indicators (with weights ranging from 1 to 6) of 19 conditions associated with posthospitalization one-year mortality,38 and the Selim comorbidity index sums indicators of 30 chronic physical illnesses and 6 mental health conditions.39 Both the Charlson and Selim indexes are commonly used to adjust for case mix. There is some overlap in ICD-9-CM codes (e.g., some cancer, diabetes, and cardiovascular diagnosis codes); however, the two case-mix adjusters can be used together because they capture different aspects of disease burden. Shared variance between these summary measures produced a correlation of 0.53 (indicating a 28% shared variance) in a study conducted with a chronically ill cohort of veterans.40 Antipsychotic drugs were classified as either first-generation agents (chlorpromazine or promazine, chlorprothixene, droperidol, fluphenazine, haloperidol, loxapine, mesoridazine, molindone, perphenazine, pimozide, prochlorperazine, thioridazine, thiothixene, and trifluoperazine) or second-generation agents (aripiprazole, asenapine, clozapine, olanzapine, paliperidone, quetiapine, risperidone, and zipras idone). The main outcome of interest was the rate of antipsychotic polypharmacy, which was assessed for each study year and defined as the filling of prescriptions for two
PRACTICE REPORTS Polypharmacy
or more antipsychotic medications concurrently in one year (or one week in the VA-Week sample). In addition to using both year-round and one-week VA utilization data to define polypharmacy, a third method of analysis was intended to capture cases in which clinicians may want patients to switch drugs—for example, from risperidone to aripiprazole—and gradually decrease the dose of a drug being discontinued while gradually increasing the dose of the drug being substituted; these cases were not considered to involve polypharmacy, as they represented the planned transition from one antipsychotic to another. To capture this phenomenon of “tapered polypharmacy” among VA outpatients (via the VA-Tapered sample), the definition of polypharmacy was modified to allow up to 60 days of overlap in refill periods. In the other data sources, we were unable to determine whether concurrent use represented tapering off one medication and onto another or intentional, ongoing use of two or more antipsychotics. A secondary outcome was the proportion of patients across systems prescribed any antipsychotic medication each year. Finally, we examined one-year all-cause hospitalization as an outcome. Statistical analysis. Means and frequencies were calculated to describe patient characteristics for each unique system data source. Analyses were performed in SAS, version 9.2 (SAS Inc., Cary, NC), with survey procedures used for the NAMCS weighted data. 41 To compare yearly rates and trends of antipsychotic polypharmacy, we first determined the proportion of patients with schizophrenia who received any antipsychotic medication and then the proportion of those taking antipsychotics who experienced antipsychotic polypharmacy. Logistic regression was used. For each health care system, we modeled antipsychotic polypharmacy (versus
monotherapy) among patients prescribed antipsychotics as a function of patient characteristics. In these models, the outcome was polypharmacy during the one week of utilization for patients in the one-week data models (VA-Week and NAMCS) or polypharmacy during the year after study entry in the remaining models (VA sample, VA-tapered, and HMORN sites 1 and 2). The relative odds of experiencing antipsychotic polypharmacy were estimated as a function of demographic covariates and the year of first entry of patient information into the applicable data set (FY02–FY07) relative to a reference year (FY08); patients whose year of first entry was FY09 were excluded from the multivariable models because a full year of followup data on polypharmacy could not be measured. Logistic regression also modeled the one-year risk of hospital admission for each health care system separately, adjusting for patient characteristics. A maximum type I error rate of a = 0.05 was applied to all analyses. Results Patient characteristics among the four study samples are displayed in Table 1. How the final number of study patients or visits was determined is summarized in Figure 1. Among the 119,662 VA patients with schizophrenia, only 7% were women, whereas women accounted for 45% of the estimated NAMCS sample, 51% of the HMORN site 1 sample of 699 patients, and 49% of the HMORN site 2 sample of 4,887 patients. On average, VA patients were older, with a mean ± S.D. age of 50.6 ± 8.8 years, compared with 43.3 ± 12.2 years for the NAMCS sample, 40.5 ± 14.9 years for the HMORN site 1 sample, and 41.3 ± 12.7 years for the HMORN site 2 sample. Both the Charlson Comorbidity Index score and Selim comorbidity index score were higher among VA patients than among nonVA patients. Race differed by sample.
African Americans were more numerous in the VA and HMORN site 2 samples (34% and 52%, respectively) versus the NAMCS sample (21%) and the HMORN site 1 sample (14%). Hispanic ethnicity was more common in the VA, NAMCS, and HMORN site 1 samples (7%, 6%, and 3%, respectively) than in the HMORN site 2 sample (1%). For the 47,100 VA patients in the one-week (VA-Week) sample, descriptive statistics were very similar to those for the VA sample as a whole but with a slightly lower proportion listed as Hispanic. Prevalence of antipsychotic use and polypharmacy. Trends over time in the use of antipsychotics and in antipsychotic polypharmacy for patients with schizophrenia prescribed antipsychotic medications are depicted in Figure 2. Overall, the proportion of VA patients with schizophrenia who were prescribed some form of antipsychotic pharmacologic treatment stayed fairly consistent during the period FY05– FY09, at 74–78% each year. The VA one-week sample showed stable and comparatively lower rates of antipsychotic prescribing (62–64%). Across all health care systems, the NAMCS sample had the highest antipsychotic utilization rates (84%, 79%, and 78% in FY05, FY07, and FY08, respectively) while the VA sample had the highest utilization rates in FY06 (77%) and FY09 (74%). The HMORN site 2 sample had the lowest antipsychotic utilization rates (22–30%) from FY05 to FY09. In contrast, non-VA estimates of antipsychotic use revealed much greater variability across the same years, ranging from 69% to 84% for the NAMCS sample and from 49% to 67% for the HMORN site 1 sample; the HMORN site 2 sample had the lowest rates (range, 22–41%). In addition, the use of antipsychotics was more prevalent in the sampled nationwide health care systems (as reflected in the VA and NAMCS samples) than in the
Am J Health-Syst Pharm—Vol 71 May 1, 2014
731
732
Am J Health-Syst Pharm—Vol 71 May 1, 2014 663 (74.1) 177 (21.4) 45 (4.5) 0 (0.0) 58 (6.6) 767 (86.7) 60 (6.8) 196 (22.1) 118 (13.3) 395 (44.6) 95 (10.7) 81 (9.2)
27,768 (59.0) 16,138 (34.2) 1,345 (2.9) 1,849 (3.9) 2,958 (6.3) 42,293 (89.8) 1,849 (3.9) 18,536 (39.4) 3,747 (8.0) 19,324 (41.0) 2,872 (6.1) 2,621 (5.6)
69,290 (57.9) 40,857 (34.1) 3,225 (2.7) 6,290 (5.3)
8,197 (6.8) 105,175 (87.9) 6,290 (5.3)
38,478 (32.2) 8,979 (7.5) 47,774 (39.9) 14,601 (12.2) 9,830 (8.2)
361 (58.1) 17 (2.7) 177 (28.5) 45 (7.2) 21 (3.4)
19 (2.7) 347 (49.6) 333 (47.6)
472 (67.5) 96 (13.7) 11 (1.5) 120 (17.2)
3,308 (78.1) 81 (1.9) 616 (14.5) 182 (4.3) 48 (1.1)
54 (1.1) 1,230 (25.2) 3,603 (73.7)
2,043 (41.8) 2,565 (52.5) 126 (2.6) 153 (3.1)
2,391 (48.9) 2,496 (51.1)
401 (45.2) 484 (54.8)
3,487 (7.4) 43,613 (92.6)
8,331 (7.0) 111,331 (93.0)
362 (51.8) 337 (48.2)
2.7 ± 2.2 1.5 ± 1.0 1.2 ± 1.8 0.6 ± 1.5 2.4 ± 1.8 1.1 ± 0.9 1.3 ± 1.7 0.8 ± 1.8
1.3 ± 0.8 1.2 ± 0.7 0.1 ± 0.4 0.1 ± 0.5
3.8 ± 2.2 1.9 ± 1.1 1.9 ± 1.7 0.8 ± 1.5
51.5 ± 8.5 1.5 ± 1.0 1.2 ± 0.6 0.3 ± 0.8 0.2 ± 0.6
50.6 ± 8.8
41.3 ± 12.7
HMORN Samples Site 1 Site 2 (n = 699) (n = 4,887) FY02–FY09 FY02–FY09 40.5 ± 14.9
NAMCS Sample (n = 885) FY02–FY09 43.3 ± 12.2
VA-Week Sample (n = 47,100) FY05–FY09
a VA = Department of Veterans Affairs, VA-Week = VA one-week sample, NAMCS = National Ambulatory Medical Care Survey, HMORN = Health Maintenance Organization Research Network, FY = fiscal year (October 1– September 30). b The Selim comorbidity index total score ranges from 0 to 36 (physical index range, 0–30; mental index range, 0–6); higher scores indicate more chronic physical and mental illness. The Charlson Comorbidity Index score ranges from 0 to 37, with higher scores indicating a greater risk of one-year postdischarge mortality. c For the VA and HMORN samples, assessment of antipsychotic types was limited to patients with one year of available follow-up data; for the VA-Week and NAMCS samples, antipsychotic types were measured during a oneweek period.
Mean ± S.D. age, yr Mean ± S.D. Selim comorbidity index scoreb Total Mental Physical Mean ± S.D. Charlson Comorbidity Index scoreb Sex, no. (%) Female Male Race, no. (%) White Black Other Unknown Hispanic, no. (%) Yes No Unknown Antipsychotic type, no. (%)c None First-generation only Second-generation only 2 or more agents First- and second-generation
Characteristic
VA Sample (n = 119,662) FY05–FY09
Demographic and Clinical Characteristics of Patients With Schizophrenia in Study Samplesa
Table 1.
PRACTICE REPORTS Polypharmacy
18–64 years of age on date of entry into data set for each study year (n = 4,887) 18–64 years of age on date of entry into data set for each study year (n = 699) 18–64 years of age on date of entry into data set during the period FY05–FY09 (n = 119,662)
Two or more outpatient visits during any one year (n = 167,725)
18–64 years of age on date of entry into data set (n = 47,100)
Visit by patient 18–64 years of age on date of entry into data set for each study year (n = 885 visits; (17,596,617 patients represented)
Diagnosis of schizophrenia or psychosis disorder (n = 7,527) Diagnosis of schizophrenia or psychosis disorder (n = 1,808) Visit with accompanying diagnosis of schizophrenia or psychosis disorder (n = 957) Diagnosis of schizophrenia or psychosis disorder (n = 59,680) Diagnosis of schizophrenia or psychosis disorder (n = 258,752)
Utilized outpatient services during period FY02–FY09 (n = 366,645) Outpatient visit during a designated one-week period of each year from FY02 through FY09 (n = 170,954) Utilized outpatient services during period June 21–29 in any year from FY05 through FY09 (n = 2,762,341) Utilized outpatient services during period FY05–FY09 (n = 7,233,362)
Utilized outpatient services during period FY02–FY09 (n = 972,214)
Site 2 Site 1 NAMCS Sample VA-Week Sample VA Sample
HMORN Samples
private, not-for-profit health care systems (as reflected in the HMORN samples). For patients taking antipsychotics, the VA cohorts demonstrated relatively consistent polypharmacy prevalence for the years FY05–FY09: 13–15% for the VA-Tapered, 29–31% for the full VA, and 20–22% for the VA-Week samples. Conversely, the nonfederal national (i.e., NAMCS) sample demonstrated a fluctuating polypharmacy trend during those years (rates of 19–31%), while the nonnational (i.e., HMORN) samples also demonstrated more variable polypharmacy rates (16–26% for site 1 and 21–26% for site 2). Multivariable models of antipsychotic polypharmacy across systems. The results of the multivariable logistic regression models comparing patient characteristics associated with antipsychotic polypharmacy are reported in Table 2 along with results obtained via the various approaches to defining polypharmacy within the VA sample. Significant associations were not observed in the NAMCS sample. However, younger age (modeled in 10-year increments) was associated with increased odds of antipsychotic polypharmacy in the full-year VA sample (one-year odds ratio [OR], 0.79; 95% confidence interval [CI], 0.78–0.81) and the VAWeek sample (one-week OR, 91; 95% CI, 0.88–0.94). The ORs of 0.69 and 0.82 for age in the HMORN samples corresponded to approximately 31% and 18% lower relative odds of polypharmacy for each decade of older age for sites 1 and 2, respectively. For all VA patients with schizophrenia, increased odds of antipsychotic polypharmacy were observed for female veterans (OR, 1.11; 95% CI, 1.05–1.18). Greater total comorbidity per the Selim index was associated with an increased risk of polypharmacy among all VA patients (OR, 1.09; 95% CI, 1.08–1.09) and patients in both the HMORN site 1 sample (OR, 1.46; 95% CI, 1.15–1.86) and the HMORN site 2 sample (OR,
Figure 1. Flowchart of application of study inclusion criteria. VA = Department of Veterans Affairs, NAMCS = National Ambulatory Medical Care Survey, HMORN = Health Maintenance Organization Research Network, FY = fiscal year (October 1–September 30).
PRACTICE REPORTS Polypharmacy
Am J Health-Syst Pharm—Vol 71 May 1, 2014
733
PRACTICE REPORTS Polypharmacy
1.13; 95% CI, 1.04–1.23), indicating cumulative odds increases of approximately 9%, 46%, and 13% per additional chronic illness, respectively; however, decreased risks of polypharmacy were observed in the VA-Tapered sample (OR, 0.97 for each additional condition assessed by the Selim index; 95% CI, 0.96–0.98) and VA-Week sample (OR, 0.94 per additional Selim-assessed condition; 95% CI, 0.91–0.97). Yet, the Charlson Comorbidity Index score was associated with polypharmacy only in the VA-Tapered sample (OR, 1.04; 95% CI, 1.02–1.06). Multivariable models of oneyear admission to hospital across systems. We also examined multi-
variable logistic regression models of one-year hospital admission in the VA and HMORN samples. The model was not conducted on the NAMCS sample due to its lack of follow-up information. Increased antipsychotic polypharmacy was significantly associated with increased odds of one-year admission in all models, with OR values of 1.87 (95% CI, 1.77–1.97) for the VA-Tapered sample, 2.44 (95% CI, 2.34–2.54) for the VA one-year sample, 1.41 (95% CI, 1.33–1.51) for the VA-Week sample, 2.22 (95% CI, 1.25–2.95) for the HMORN site 1 sample, and 1.57 (95% CI, 1.18–2.08) for the HMORN site 2 sample.
Discussion The study described here compared antipsychotic medication use and polypharmacy among patients diagnosed with schizophrenia (or a nonspecific psychotic disorder) across health care systems, including a national federal system (the VA system), national nonfederal health care settings (as captured by the NAMCS), and two nonfederal health care systems (as reflected in the HMORN samples). As a result of the completeness of VA data, several methods of analyses were compared when assessing antipsychotic polypharmacy among patients with schizophrenia. Our examination of unadjusted longitudinal trends,
Figure 2. Percentage of patients in each study sample treated with at least one antipsychotic drug (panel A) and the percentage receiving more than one antipsychotic (panel B). VA = Department of Veterans Affairs, NAMCS = National Ambulatory Medical Care Survey, HMORN = Health Maintenance Organization Research Network.
A Antipsychotic Use (%)
90 80 70 60 50 40 30 20 2002
2003
2004
2005
2006
2007
2008
2009
Fiscal Year
VA + VA-Tapered VA-Week NAMCS Sample Sample VA-TaperedSample VA-Week
VA Sample
a
B
Sample
Sample
HMORN Site 2
HMORN Sample HMORN Site 2 Site 1 Sample Sample
Antipsychotic Polypharmacy (%)
35 30 25 20 15 10 2002
2003
2004
2005
2006
Fiscal Year Includes VA-Tapered sample in panel A.
a
734
HMORN Site 1 Sample NAMCS Sample
Am J Health-Syst Pharm—Vol 71 May 1, 2014
2007
2008
2009
0.003 1.13 (1.04–1.23) 1.46 (1.15–1.86) 0.002 1.20 (0.93–1.56) 0.169
0.94 (0.91–0.97) 0.001
0.002 0.211 0.509 0.631 0.926 0.784 0.837 0.440 0.82 (0.73–0.93) 1.23 (0.89–1.68) 0.82 (0.45–1.48) 1.17 (0.62–2.23) 0.97 (0.48–1.92) 1.10 (0.56–2.16) 0.93 (0.46–1.89) 1.31 (0.66–2.61) 0.004 0.812 0.977 0.314 0.144 0.177 0.568 0.878
1.09 (1.08–1.09)
PRACTICE REPORTS
Polypharmacy with antipsychotic drugs in patients with schizophrenia: Trends in multiple health care systems FangFang Sun, Eileen M. Stock, Laurel A. Copeland, John E. Zeber, Brian K. Ahmedani, and Sandra B. Morissette
S
chizophrenia is a complicated mental illness requiring lifelong treatment with antipsychotic drugs.1 Schizophrenia imparts social and cognitive impairment, pervasively affecting memory, attention, motor skills, executive function, and intelligence.2 Continuous use of antipsychotic drugs is recommended (for schizophrenia uniquely among the major mental illnesses). Although antipsychotic monotherapy is the standard approach for managing symptoms of schizophrenia,3-6 multiple antipsychotic medications may be prescribed. Several factors contribute to pharmacologic treatment patterns in caring for patients with schizophrenia, including patientlevel issues, provider prescribing
Purpose. Patterns of pharmacologic treatment in U.S. outpatients with schizophrenia across multiple health care settings were investigated. Methods. Antipsychotic drug utilization by patients with schizophrenia and related disorders was analyzed using data on 119,662 patients served by the Veterans Affairs (VA) health care system in fiscal years 2005–09, data on 5,440 enrollees in two health maintenance organizations (HMOs) in 2002–09, and National Ambulatory Medical Care Survey (NAMCS) data reflecting the experience of 17.6 million U.S. residents seeking care outside federal systems during the same eight-year period. Polypharmacy was defined as the use of more than one antipsychotic agent during one year (in the VA sample) or one week (in the HMO and NAMCS samples). The association of polypharmacy with hospital
FangFang Sun, M.S., is Health Services Researcher, Center for Applied Health Research, Temple, TX. Eileen M. Stock, Ph.D., is Research Scientist, Center for Applied Health Research, and Assistant Professor, College of Medicine, Texas A&M Health Sciences Center, Bryan. Laurel A. Copeland, Ph.D., is Interim Associate Chief of Staff of Research, Central Texas Veterans Health Care System, Temple, and Associate Director, Center for Applied Health Research, and Associate Professor, College of Medicine, Texas A&M Health Sciences Center. John E. Zeber, Ph.D., is Co-Director, Health Outcomes Core (jointly sponsored by Central Texas Veterans Health Care System and Scott & White Healthcare), Temple, and Associate Professor, College of Medicine, Texas A&M Health Sciences Center. Brian K. Ahmedani, Ph.D., LMSW, is Research Scientist, Center for Health Policy & Health Services Research, Henry Ford Health System, Detroit, MI. Sandra B. Morissette, Ph.D., is Assessment Core Chief, Veterans Affairs VISN 17 Center of Excellence for Research on Re-
728
Am J Health-Syst Pharm—Vol 71 May 1, 2014
admissions was assessed via multivariable logistic regression. Results. Rates of antipsychotic use in the VA sample ranged from 74% to 78%, with lower and more variable rates in the NAMCS sample (69–84%) and the HMO sample (22–67%). VA patients were found to have lower polypharmacy rates (20–22%) than patients in the HMO and NAMCS samples (19–31%). In all samples evaluated, polypharmacy was associated with an increased likelihood of hospital admission (odds ratio range, 1.4–2.4). Conclusion. A multisystem study revealed that antipsychotic use among patients with schizophrenia varied substantially among health care systems and that nearly one fifth of patients with schizophrenia or other psychotic disorders in most of the health care systems experienced antipsychotic polypharmacy. Am J Health-Syst Pharm. 2014; 71:728-38
turning War Veterans, Waco, TX, and Associate Professor, College of Medicine, Texas A&M Health Sciences Center. Address correspondence to Dr. Copeland (laurelacopeland@ gmail.com). Supported by the Center for Applied Health Research, Temple, TX, a research center jointly sponsored by Central Texas Veterans Health Care System and Scott & White Healthcare; and by grants from the Veterans Health Administration, Health Services Research and Development #IIR-09-335. The views expressed in this article are those of the authors and do not necessarily reflect the position or policy of the Department of Veterans Affairs. The authors have declared no potential conflicts of interest. DOI 10.2146/ajhp130471
PRACTICE REPORTS Polypharmacy
decisions, health-system culture, and organizational structure.7 Given the challenges and costs of treating individuals with schizophrenia, the prescribing trends of antipsychotic monotherapy and polypharmacy may vary across health care systems. A number of international guidelines recommend antipsychotic monotherapy using second-generation antipsychotics as first-line medications for the treatment of schizophrenia.4,8-13 While first-generation antipsychotics are acceptable in terms of efficacy, they may be avoided or considered second-line treatments because of their irreversible adverse effects (e.g., tardive dyskinesia). 14,15 However, antipsychotic polypharmacy, defined as the concurrent use of more than one antipsychotic drug for a single clinical condition, is on the rise.16,17 Indeed, since 1996, studies have shown that the rate of antipsychotic polypharmacy has increased among patients with schizophrenia, with the extent of polypharmacy varying by the populations under study, the treatment setting, and how antipsychotic polypharmacy was defined and measured.18 Specifically, antipsychotic polypharmacy among both outpatients and inpatients with schizophrenia has been reported at very high rates: up to 20% among patients receiving care in the Department of Veterans Affairs (VA) in fiscal year 2000,19 35% among outpatients in the public mental health system during the two-year period 1996–98,20 up to 40% among outpatients captured in 1998–2000 Medicaid claims data,16 and 38–55% among inpatients in Asian countries during the period 2001–09.21,22 Some plausible reasons for these figures include patient or clinician dissatisfaction with treatment effects, bothersome adverse effects, and undetected poor adherence leading to persistent symptoms that are incorrectly attributed to ineffective pharmacotherapy. Nonetheless, the prescribing of multiple antipsychotics may result
in adverse outcomes, such as an increased risk of metabolic syndrome23 and emergency department visits due to adverse drug reactions and psychiatric admissions,24 while also increasing direct treatment costs for the health care system, imposing an indirect societal burden, and resulting in impaired quality of life for the patient.23,25,26 Few studies have assessed the differential longitudinal prevalence of polypharmacy between health care systems for patients with schizophrenia. Most studies examined broad antipsychotic polypharmacy rates across countries, 27,28 though one older study compared rates of prescribing of combinations of psychotropic drugs targeting the same set of symptoms (those of acute paranoid schizophrenia) in California, New York, Pennsylvania, and Texas based on a survey of psychiatrists.29 To our knowledge, only two studies have compared pharmacotherapy for patients with schizophrenia across multiple types of health care systems.30,31 One of the two studies found that patients with schizophrenia treated in the VA health care system were more likely to receive an antipsychotic medication and more likely to be dosed above treatment recommendations than patients in the private sector. However, it did not appear that physical comorbidities were included for case-mix adjustment or that longitudinal pharmacotherapy patterns were measured.30 The second study, using data from 1994–96, found that outpatients receiving care in the VA system in two states were less likely to receive an antipsychotic medication than patients outside of this system.31 Despite these prior studies, no large-scale multisystem studies have investigated this issue in the last decade. With the growing national concern regarding increased polypharmacy among patients with schizophrenia, coupled with concern over the high cost of duplicative or non-evidence-based treatment (or
excess emergency department or inpatient care costs), new data are needed to inform policy and clinical decisions. In the study described here, we compared longitudinal data on antipsychotic prescribing and polypharmacy among patients diagnosed with schizophrenia in multiple health care systems representing three different organizational structures: (1) the nationwide VA system (2) nonfederal outpatient settings nationwide, and (3) two large not-for-profit health care systems in the Southwest and Great Lakes area of the United States. We also modeled one-year hospital admissions as a function of polypharmacy (versus monotherapy), adjusting for patient characteristics and disease burden. As an additional methodological aim, we examined how the availability of different amounts of data, depending on the source of information, affected the observed rates of antipsychotic prescribing and polypharmacy. Methods This study used a retrospective design relying on several sources of data, including a public-use data set and data from multiple health care systems for which approval of use for this research was obtained from the respective institutional review boards (IRBs). The study population consisted of patients who were 18–64 years of age at the time of diagnosis of schizophrenia or an unspecified psychotic disorder, as defined by the diagnosis codes (International Classification of Diseases, 9th Revision, Clinical Modification [ICD-9-CM]) listed in the medical record, specifically ICD-9-CM code series 295, “schizophrenic disorders” (excluding 295.5, “latent schizophrenia,” used for persons with symptoms but no psychotic episode before a schizophrenia diagnosis can be definitively assigned) and 298.9, “unspecified psychosis”; the latter code was included when it was determined that
Am J Health-Syst Pharm—Vol 71 May 1, 2014
729
PRACTICE REPORTS Polypharmacy
one health care system did not typically use the 295-series codes but used code 298.9 instead. The latent schizophrenia code was rare, appearing on the records of 135 of 119, 662 VA patients, all of whom also had other codes for schizophrenia; thus, it was determined that code 295.5 may have been used in error. No cases were noted in the other data sources. Elderly patients were excluded from the study population to reduce the impact of “burnout” in older patients with schizophrenia, as these patients typically have few positive symptoms (e.g., delusions, disordered thought, hallucinations) and thus derive little benefit from antipsychotics, may be at increased risk of death from antipsychotics, and are therefore less likely to be prescribed antipsychotics.32,33 Nationally representative sample. The public-use data set was obtained from the National Ambulatory Medical Care Survey (NAMCS) and represented nationwide nonfederal health care utilization among patients with schizophrenia from fiscal year 2002 (FY02, October 2001–September 2002) through FY09.34 NAMCS data contain information on nonfederal office-based physician visits obtained by sampling a set of clinics and requesting that those clinics report on a systematic random sample during a single week of patient visits. A total of 885 visits over eight years, representing the experience of more than 17 million community-based outpatients diagnosed with schizophrenia or a psychotic disorder, were identified from this national sample. Clinicians responding to this survey provided answers based on the past week at their practice. Pharmacy prescription fill data are not assessed by the NAMCS, but the survey captures clinician-reported prescriptions. National federal sample (VA). The study inclusion criteria targeted patients receiving VA care who had two or more outpatient dates of care for which a diagnosis of schizophrenia 730
or psychotic disorder was recorded with the same ICD-9-CM code. Data were only available for VA patients in the years 2005–09 due to IRBapproval issues. VA data were extracted from VA’s all-electronic medical records system for patients treated during the period FY05–FY09. For the VA cohort, we obtained data on all 119,662 veterans diagnosed with schizophrenia or psychotic disorder systemwide. Similar to the NAMCS data, the VA data reflected rolling enrollment (i.e., new patients entered the cohort each year, with some overlapping cases). The major difference between the VA and NAMCS data sets was that VA data captured year-round health care utilization and pharmacy fills per patient rather than utilization during just one week. It is unclear from NAMCS methods how far back into the patient chart the survey responses reach. Thus, for comparability we also assessed one week (June 21–28, per year analyzed) of service use for VA patients during each fiscal year in order to resemble NAMCS data, identifying 47,100 patients (the VA-Week sample). Private health care systems. For the private, not-for-profit health care systems, we used data from two sites (sites 1 and 2) submitting data to the Health Maintenance Organization Research Network (HMORN) Virtual Data Warehouse. HMORN is a federation of 19 care-and-coverage health care systems that have developed a set of standardized data definitions in order to create uniform health care measures from member data, primarily health care claims, from the year 2000 onward.35-37 We identified 699 patients at site 1 and 4887 patients at site 2 diagnosed with schizophrenia or psychotic disorder during the period FY02–FY09 (these sample sizes were insufficient for meaningful one-week look-back analyses, as were performed for the NAMCS and VA-Week samples). Measures. For each system of care represented by the sampled data
Am J Health-Syst Pharm—Vol 71 May 1, 2014
sets (NAMCS, VA, and HMORN), patient demographic characteristics were obtained: age at baseline, sex, race, and ethnicity. Age was used to create a variable representing a decade effect in the analytical models, and ethnicity was dichotomously defined as Hispanic or not Hispanic. Case-mix adjusters included the Charlson Comorbidity Index score and the Selim comorbidity index score (total and physical and mental component scores), with comorbidity indices calculated at a patient’s first year of entry into the applicable data set. The Charlson Comorbidity Index sums weighted indicators (with weights ranging from 1 to 6) of 19 conditions associated with posthospitalization one-year mortality,38 and the Selim comorbidity index sums indicators of 30 chronic physical illnesses and 6 mental health conditions.39 Both the Charlson and Selim indexes are commonly used to adjust for case mix. There is some overlap in ICD-9-CM codes (e.g., some cancer, diabetes, and cardiovascular diagnosis codes); however, the two case-mix adjusters can be used together because they capture different aspects of disease burden. Shared variance between these summary measures produced a correlation of 0.53 (indicating a 28% shared variance) in a study conducted with a chronically ill cohort of veterans.40 Antipsychotic drugs were classified as either first-generation agents (chlorpromazine or promazine, chlorprothixene, droperidol, fluphenazine, haloperidol, loxapine, mesoridazine, molindone, perphenazine, pimozide, prochlorperazine, thioridazine, thiothixene, and trifluoperazine) or second-generation agents (aripiprazole, asenapine, clozapine, olanzapine, paliperidone, quetiapine, risperidone, and zipras idone). The main outcome of interest was the rate of antipsychotic polypharmacy, which was assessed for each study year and defined as the filling of prescriptions for two
PRACTICE REPORTS Polypharmacy
or more antipsychotic medications concurrently in one year (or one week in the VA-Week sample). In addition to using both year-round and one-week VA utilization data to define polypharmacy, a third method of analysis was intended to capture cases in which clinicians may want patients to switch drugs—for example, from risperidone to aripiprazole—and gradually decrease the dose of a drug being discontinued while gradually increasing the dose of the drug being substituted; these cases were not considered to involve polypharmacy, as they represented the planned transition from one antipsychotic to another. To capture this phenomenon of “tapered polypharmacy” among VA outpatients (via the VA-Tapered sample), the definition of polypharmacy was modified to allow up to 60 days of overlap in refill periods. In the other data sources, we were unable to determine whether concurrent use represented tapering off one medication and onto another or intentional, ongoing use of two or more antipsychotics. A secondary outcome was the proportion of patients across systems prescribed any antipsychotic medication each year. Finally, we examined one-year all-cause hospitalization as an outcome. Statistical analysis. Means and frequencies were calculated to describe patient characteristics for each unique system data source. Analyses were performed in SAS, version 9.2 (SAS Inc., Cary, NC), with survey procedures used for the NAMCS weighted data. 41 To compare yearly rates and trends of antipsychotic polypharmacy, we first determined the proportion of patients with schizophrenia who received any antipsychotic medication and then the proportion of those taking antipsychotics who experienced antipsychotic polypharmacy. Logistic regression was used. For each health care system, we modeled antipsychotic polypharmacy (versus
monotherapy) among patients prescribed antipsychotics as a function of patient characteristics. In these models, the outcome was polypharmacy during the one week of utilization for patients in the one-week data models (VA-Week and NAMCS) or polypharmacy during the year after study entry in the remaining models (VA sample, VA-tapered, and HMORN sites 1 and 2). The relative odds of experiencing antipsychotic polypharmacy were estimated as a function of demographic covariates and the year of first entry of patient information into the applicable data set (FY02–FY07) relative to a reference year (FY08); patients whose year of first entry was FY09 were excluded from the multivariable models because a full year of followup data on polypharmacy could not be measured. Logistic regression also modeled the one-year risk of hospital admission for each health care system separately, adjusting for patient characteristics. A maximum type I error rate of a = 0.05 was applied to all analyses. Results Patient characteristics among the four study samples are displayed in Table 1. How the final number of study patients or visits was determined is summarized in Figure 1. Among the 119,662 VA patients with schizophrenia, only 7% were women, whereas women accounted for 45% of the estimated NAMCS sample, 51% of the HMORN site 1 sample of 699 patients, and 49% of the HMORN site 2 sample of 4,887 patients. On average, VA patients were older, with a mean ± S.D. age of 50.6 ± 8.8 years, compared with 43.3 ± 12.2 years for the NAMCS sample, 40.5 ± 14.9 years for the HMORN site 1 sample, and 41.3 ± 12.7 years for the HMORN site 2 sample. Both the Charlson Comorbidity Index score and Selim comorbidity index score were higher among VA patients than among nonVA patients. Race differed by sample.
African Americans were more numerous in the VA and HMORN site 2 samples (34% and 52%, respectively) versus the NAMCS sample (21%) and the HMORN site 1 sample (14%). Hispanic ethnicity was more common in the VA, NAMCS, and HMORN site 1 samples (7%, 6%, and 3%, respectively) than in the HMORN site 2 sample (1%). For the 47,100 VA patients in the one-week (VA-Week) sample, descriptive statistics were very similar to those for the VA sample as a whole but with a slightly lower proportion listed as Hispanic. Prevalence of antipsychotic use and polypharmacy. Trends over time in the use of antipsychotics and in antipsychotic polypharmacy for patients with schizophrenia prescribed antipsychotic medications are depicted in Figure 2. Overall, the proportion of VA patients with schizophrenia who were prescribed some form of antipsychotic pharmacologic treatment stayed fairly consistent during the period FY05– FY09, at 74–78% each year. The VA one-week sample showed stable and comparatively lower rates of antipsychotic prescribing (62–64%). Across all health care systems, the NAMCS sample had the highest antipsychotic utilization rates (84%, 79%, and 78% in FY05, FY07, and FY08, respectively) while the VA sample had the highest utilization rates in FY06 (77%) and FY09 (74%). The HMORN site 2 sample had the lowest antipsychotic utilization rates (22–30%) from FY05 to FY09. In contrast, non-VA estimates of antipsychotic use revealed much greater variability across the same years, ranging from 69% to 84% for the NAMCS sample and from 49% to 67% for the HMORN site 1 sample; the HMORN site 2 sample had the lowest rates (range, 22–41%). In addition, the use of antipsychotics was more prevalent in the sampled nationwide health care systems (as reflected in the VA and NAMCS samples) than in the
Am J Health-Syst Pharm—Vol 71 May 1, 2014
731
732
Am J Health-Syst Pharm—Vol 71 May 1, 2014 663 (74.1) 177 (21.4) 45 (4.5) 0 (0.0) 58 (6.6) 767 (86.7) 60 (6.8) 196 (22.1) 118 (13.3) 395 (44.6) 95 (10.7) 81 (9.2)
27,768 (59.0) 16,138 (34.2) 1,345 (2.9) 1,849 (3.9) 2,958 (6.3) 42,293 (89.8) 1,849 (3.9) 18,536 (39.4) 3,747 (8.0) 19,324 (41.0) 2,872 (6.1) 2,621 (5.6)
69,290 (57.9) 40,857 (34.1) 3,225 (2.7) 6,290 (5.3)
8,197 (6.8) 105,175 (87.9) 6,290 (5.3)
38,478 (32.2) 8,979 (7.5) 47,774 (39.9) 14,601 (12.2) 9,830 (8.2)
361 (58.1) 17 (2.7) 177 (28.5) 45 (7.2) 21 (3.4)
19 (2.7) 347 (49.6) 333 (47.6)
472 (67.5) 96 (13.7) 11 (1.5) 120 (17.2)
3,308 (78.1) 81 (1.9) 616 (14.5) 182 (4.3) 48 (1.1)
54 (1.1) 1,230 (25.2) 3,603 (73.7)
2,043 (41.8) 2,565 (52.5) 126 (2.6) 153 (3.1)
2,391 (48.9) 2,496 (51.1)
401 (45.2) 484 (54.8)
3,487 (7.4) 43,613 (92.6)
8,331 (7.0) 111,331 (93.0)
362 (51.8) 337 (48.2)
2.7 ± 2.2 1.5 ± 1.0 1.2 ± 1.8 0.6 ± 1.5 2.4 ± 1.8 1.1 ± 0.9 1.3 ± 1.7 0.8 ± 1.8
1.3 ± 0.8 1.2 ± 0.7 0.1 ± 0.4 0.1 ± 0.5
3.8 ± 2.2 1.9 ± 1.1 1.9 ± 1.7 0.8 ± 1.5
51.5 ± 8.5 1.5 ± 1.0 1.2 ± 0.6 0.3 ± 0.8 0.2 ± 0.6
50.6 ± 8.8
41.3 ± 12.7
HMORN Samples Site 1 Site 2 (n = 699) (n = 4,887) FY02–FY09 FY02–FY09 40.5 ± 14.9
NAMCS Sample (n = 885) FY02–FY09 43.3 ± 12.2
VA-Week Sample (n = 47,100) FY05–FY09
a VA = Department of Veterans Affairs, VA-Week = VA one-week sample, NAMCS = National Ambulatory Medical Care Survey, HMORN = Health Maintenance Organization Research Network, FY = fiscal year (October 1– September 30). b The Selim comorbidity index total score ranges from 0 to 36 (physical index range, 0–30; mental index range, 0–6); higher scores indicate more chronic physical and mental illness. The Charlson Comorbidity Index score ranges from 0 to 37, with higher scores indicating a greater risk of one-year postdischarge mortality. c For the VA and HMORN samples, assessment of antipsychotic types was limited to patients with one year of available follow-up data; for the VA-Week and NAMCS samples, antipsychotic types were measured during a oneweek period.
Mean ± S.D. age, yr Mean ± S.D. Selim comorbidity index scoreb Total Mental Physical Mean ± S.D. Charlson Comorbidity Index scoreb Sex, no. (%) Female Male Race, no. (%) White Black Other Unknown Hispanic, no. (%) Yes No Unknown Antipsychotic type, no. (%)c None First-generation only Second-generation only 2 or more agents First- and second-generation
Characteristic
VA Sample (n = 119,662) FY05–FY09
Demographic and Clinical Characteristics of Patients With Schizophrenia in Study Samplesa
Table 1.
PRACTICE REPORTS Polypharmacy
18–64 years of age on date of entry into data set for each study year (n = 4,887) 18–64 years of age on date of entry into data set for each study year (n = 699) 18–64 years of age on date of entry into data set during the period FY05–FY09 (n = 119,662)
Two or more outpatient visits during any one year (n = 167,725)
18–64 years of age on date of entry into data set (n = 47,100)
Visit by patient 18–64 years of age on date of entry into data set for each study year (n = 885 visits; (17,596,617 patients represented)
Diagnosis of schizophrenia or psychosis disorder (n = 7,527) Diagnosis of schizophrenia or psychosis disorder (n = 1,808) Visit with accompanying diagnosis of schizophrenia or psychosis disorder (n = 957) Diagnosis of schizophrenia or psychosis disorder (n = 59,680) Diagnosis of schizophrenia or psychosis disorder (n = 258,752)
Utilized outpatient services during period FY02–FY09 (n = 366,645) Outpatient visit during a designated one-week period of each year from FY02 through FY09 (n = 170,954) Utilized outpatient services during period June 21–29 in any year from FY05 through FY09 (n = 2,762,341) Utilized outpatient services during period FY05–FY09 (n = 7,233,362)
Utilized outpatient services during period FY02–FY09 (n = 972,214)
Site 2 Site 1 NAMCS Sample VA-Week Sample VA Sample
HMORN Samples
private, not-for-profit health care systems (as reflected in the HMORN samples). For patients taking antipsychotics, the VA cohorts demonstrated relatively consistent polypharmacy prevalence for the years FY05–FY09: 13–15% for the VA-Tapered, 29–31% for the full VA, and 20–22% for the VA-Week samples. Conversely, the nonfederal national (i.e., NAMCS) sample demonstrated a fluctuating polypharmacy trend during those years (rates of 19–31%), while the nonnational (i.e., HMORN) samples also demonstrated more variable polypharmacy rates (16–26% for site 1 and 21–26% for site 2). Multivariable models of antipsychotic polypharmacy across systems. The results of the multivariable logistic regression models comparing patient characteristics associated with antipsychotic polypharmacy are reported in Table 2 along with results obtained via the various approaches to defining polypharmacy within the VA sample. Significant associations were not observed in the NAMCS sample. However, younger age (modeled in 10-year increments) was associated with increased odds of antipsychotic polypharmacy in the full-year VA sample (one-year odds ratio [OR], 0.79; 95% confidence interval [CI], 0.78–0.81) and the VAWeek sample (one-week OR, 91; 95% CI, 0.88–0.94). The ORs of 0.69 and 0.82 for age in the HMORN samples corresponded to approximately 31% and 18% lower relative odds of polypharmacy for each decade of older age for sites 1 and 2, respectively. For all VA patients with schizophrenia, increased odds of antipsychotic polypharmacy were observed for female veterans (OR, 1.11; 95% CI, 1.05–1.18). Greater total comorbidity per the Selim index was associated with an increased risk of polypharmacy among all VA patients (OR, 1.09; 95% CI, 1.08–1.09) and patients in both the HMORN site 1 sample (OR, 1.46; 95% CI, 1.15–1.86) and the HMORN site 2 sample (OR,
Figure 1. Flowchart of application of study inclusion criteria. VA = Department of Veterans Affairs, NAMCS = National Ambulatory Medical Care Survey, HMORN = Health Maintenance Organization Research Network, FY = fiscal year (October 1–September 30).
PRACTICE REPORTS Polypharmacy
Am J Health-Syst Pharm—Vol 71 May 1, 2014
733
PRACTICE REPORTS Polypharmacy
1.13; 95% CI, 1.04–1.23), indicating cumulative odds increases of approximately 9%, 46%, and 13% per additional chronic illness, respectively; however, decreased risks of polypharmacy were observed in the VA-Tapered sample (OR, 0.97 for each additional condition assessed by the Selim index; 95% CI, 0.96–0.98) and VA-Week sample (OR, 0.94 per additional Selim-assessed condition; 95% CI, 0.91–0.97). Yet, the Charlson Comorbidity Index score was associated with polypharmacy only in the VA-Tapered sample (OR, 1.04; 95% CI, 1.02–1.06). Multivariable models of oneyear admission to hospital across systems. We also examined multi-
variable logistic regression models of one-year hospital admission in the VA and HMORN samples. The model was not conducted on the NAMCS sample due to its lack of follow-up information. Increased antipsychotic polypharmacy was significantly associated with increased odds of one-year admission in all models, with OR values of 1.87 (95% CI, 1.77–1.97) for the VA-Tapered sample, 2.44 (95% CI, 2.34–2.54) for the VA one-year sample, 1.41 (95% CI, 1.33–1.51) for the VA-Week sample, 2.22 (95% CI, 1.25–2.95) for the HMORN site 1 sample, and 1.57 (95% CI, 1.18–2.08) for the HMORN site 2 sample.
Discussion The study described here compared antipsychotic medication use and polypharmacy among patients diagnosed with schizophrenia (or a nonspecific psychotic disorder) across health care systems, including a national federal system (the VA system), national nonfederal health care settings (as captured by the NAMCS), and two nonfederal health care systems (as reflected in the HMORN samples). As a result of the completeness of VA data, several methods of analyses were compared when assessing antipsychotic polypharmacy among patients with schizophrenia. Our examination of unadjusted longitudinal trends,
Figure 2. Percentage of patients in each study sample treated with at least one antipsychotic drug (panel A) and the percentage receiving more than one antipsychotic (panel B). VA = Department of Veterans Affairs, NAMCS = National Ambulatory Medical Care Survey, HMORN = Health Maintenance Organization Research Network.
A Antipsychotic Use (%)
90 80 70 60 50 40 30 20 2002
2003
2004
2005
2006
2007
2008
2009
Fiscal Year
VA + VA-Tapered VA-Week NAMCS Sample Sample VA-TaperedSample VA-Week
VA Sample
a
B
Sample
Sample
HMORN Site 2
HMORN Sample HMORN Site 2 Site 1 Sample Sample
Antipsychotic Polypharmacy (%)
35 30 25 20 15 10 2002
2003
2004
2005
2006
Fiscal Year Includes VA-Tapered sample in panel A.
a
734
HMORN Site 1 Sample NAMCS Sample
Am J Health-Syst Pharm—Vol 71 May 1, 2014
2007
2008
2009
0.003 1.13 (1.04–1.23) 1.46 (1.15–1.86) 0.002 1.20 (0.93–1.56) 0.169
0.94 (0.91–0.97) 0.001
0.002 0.211 0.509 0.631 0.926 0.784 0.837 0.440 0.82 (0.73–0.93) 1.23 (0.89–1.68) 0.82 (0.45–1.48) 1.17 (0.62–2.23) 0.97 (0.48–1.92) 1.10 (0.56–2.16) 0.93 (0.46–1.89) 1.31 (0.66–2.61) 0.004 0.812 0.977 0.314 0.144 0.177 0.568 0.878
1.09 (1.08–1.09)
