Impact of Medicaid prescription copayments on use of antipsychotics and other medications in patients with schizophrenia

Abstract

Objective: To assess the impact of Medicaid prescription copayment policies on anti-psychotic and other medication use among patients with schizophrenia.

Method: The study sample included fee-for-service adult Medicaid patients with schizophrenia. Medicaid claims records from 2003–2005 from 42 states and D.C. were linked with county-level data from the Area Resource File and findings from a state Medicaid policy survey. Patient-level fixed-effects regression models examined the impact of increases in generic copayments and generic/brand copayment differentials on monthly use of anti-psychotic (overall and by generic/brand status) and other non-antipsychotic medications. Medications for hypertension, hyperlipidemia, and diabetes in sub-groups of patients with these comorbidities were also examined.

Results: Prescription copayment changes had a statistically significant but small impact on anti-psychotic use. For instance, for every $1 increase in the minimum or generic copayment per prescription, there was a reduction of 1.4 anti-psychotic drug fills per 100 patient months (relative reduction = 1.9%). The generic/brand copayment differential increases also had a minimal impact in changing utilization of first-generation (generic) and second-generation (brand) anti-psychotics. Effects of copayment changes on non-anti-psychotic medication use were substantially higher; for each $1 generic copayment increase, there was a reduction of 23 non-anti-psychotic drug fills per 100 patient months (relative reduction = 10.1%). Similarly, for each $1 increase in the generic/brand copayment differential, there was a reduction of 15 non-anti-psychotic drug fills (relative reduction = 5.6%). Reductions in the number of prescriptions filled for antidiabetics, antihypertensives, and lipid-lowering agents were 4–11-fold higher than corresponding reductions for anti-psychotics.

Limitations: Because federal law requires pharmacists to fill medications for Medicaid patients regardless of the ability to pay, these results may under-estimate the true impact of copayment increases.

Conclusions: Medicaid prescription copayment increases resulted in only a minimal decline in anti-psychotic medication use, but much larger reductions in use of other medications, particularly cardiometabolic medications.

Keywords:

• Medicaid
• copayments
• anti-psychotics
• schizophrenia
• out-of-pocket costs

Introduction

Faced with serious budgetary constraints and rapid growth in prescription drug expenditures, many state Medicaid programs have adopted prescription cost-containment strategies previously used only in the private insurance sector. These strategies have included increasing patient cost sharing and generic/brand copayment differentials (i.e. tiered copayments). The theoretical rationale behind these policies is that higher out-of-pocket payments will reduce utilization of drugs considered to be of low value, and that tiered copayments will encourage use of less expensive generic drugs, leading to decreased prescription expenditures. Because the individuals covered under Medicaid are low income, federal law has limited the extent of cost sharing that states can require¹. However, even small copayment amounts may be sufficient to impede use of needed medications among low-income patients. In particular, higher copayments may place cumulative burden on individuals receiving medications for multiple conditions.

Medicaid prescription copayment changes have particular significance for individuals with schizophrenia, many of whom require pharmacological treatment for both psychiatric and general medical conditions. Anti-psychotic medications are the mainstay therapy for the management of schizophrenia², and medication adherence is critical to improve disease symptoms, reduce the risk of acute psychotic episodes, and decrease the frequency of psychiatric hospitalizations^3–7. Even relatively small changes in adherence have been associated with changes in the risk of adverse clinical outcomes⁸. Several studies have investigated the impact of copayments on use of anti-psychotics, and they have generally found that higher copayments are associated with decreased anti-psychotic use (especially in the non-Medicaid setting)^9–13. For example, a 25% reduction in psychiatric medication refills was observed among veterans with schizophrenia after their prescription copayments increased from $2 to $7⁹. To date, only two studies using Medicaid data have examined the impact of copayments of up to $3 on anti-psychotic use and treatment continuity in schizophrenia^10,13. However, both of these studies faced methodological limitations; one used a cross-sectional design¹³, and the other lacked a contemporaneous control group¹⁰, thus potentially questioning the validity of the findings of non-linear reductions in anti-psychotic medication continuity across increasing copayment categories in the former¹³ and the significant reduction in anti-psychotic medication use following an increase in copayments in the latter¹⁰. Furthermore, none of these studies have examined how increases in tiered copayment differentials impact utilization of generic vs brand anti-psychotic medications among Medicaid patients with schizophrenia. Because Medicaid is the largest insurer of individuals with schizophrenia in the US¹⁴, a closer examination of the effects of medication copayments on anti-psychotic medication use in adults with schizophrenia in the Medicaid program is warranted.

In addition to the importance of studying the impact in this population on anti-psychotic medications, it is also imperative to examine the effect of copayment increases on other non-anti-psychotic medications for multiple reasons. A complex series of factors (including anti-psychotic medication side-effects and higher rates of risk behaviors such as smoking) contribute to higher rates of metabolic conditions and cardiovascular disease among individuals with schizophrenia, who have up to double the risk of dying from cardiovascular disease compared to the general population^15–18. Evidence indicates that individuals with schizophrenia often receive sub-standard care for hypertension, dyslipidemias, diabetes, and related medical conditions, and thus any additional barriers to appropriate management of these conditions would be cause for concern^15,19–21. In light of evidence that individuals may be more likely to skip medications for asymptomatic as opposed to symptomatic conditions²², cumulative out-of-pocket copayment costs may place individuals at particular risk of skipping medication refills for general medical conditions that are frequently life-shortening. Little is known about whether overall copayment burden could lead Medicaid patients with schizophrenia and co-occurring general medical conditions to skip some medications in favor of others.

In this study, we sought to examine these issues in adults with schizophrenia who were facing stable vs increased prescription copayments, taking advantage of natural longitudinal variation in individual state Medicaid copayment policies. We first examine the impact of increases in Medicaid prescription copayments on use of anti-psychotic medications. By examining data from a time period when first-generation anti-psychotic medications (FGAs) were available as generic medications and second-generation anti-psychotic medications (SGAs) were available only as branded drugs (2003–2005), we were also able to examine the impact of increases in tiered (generic vs brand) copayment differentials on FGA vs SGA use. Next, we evaluate the relative impact of these copayment changes on use of anti-psychotics and other medications in adult patients with schizophrenia. Finally, we assess the relative impact of these copayment changes on use of anti-psychotics and medications for hypertension, hyperlipidemia, and diabetes in sub-groups of patients with schizophrenia with these conditions.

Method

Study design and data sources

This claims-based study utilized a quasi-experimental design, capitalizing on the natural variation in prescription copayment policies that existed across state Medicaid programs from 2003–2005. During this time period, some state Medicaid programs implemented copayment changes, whereas others did not. Individuals in states that changed their copayment policies during the study period served as the study group, whereas individuals in states that did not change their copayment policies served as a contemporaneous control group²³. Data from 44 states plus the District of Columbia (DC) were used in the analyses. The study was reviewed and approved by the University of Pennsylvania Institutional Review Board.

The study combined data from three sources: Centers for Medicare & Medicaid Services Medicaid Analytic eXtract (MAX) data files, the Area Resource File (ARF), and a survey of state Medicaid pharmacy directors conducted by the authors. The 2003–2005 MAX files were used to obtain patient-level information on Medicaid eligibility, medical diagnoses, and prescription drug use. The MAX prescription claims data include elements such as National Drug Code (NDC) numbers, prescription fill dates, quantities of medications dispensed, and number of days supplied. The ARF, which is a county-specific information system on health resources and socioeconomic and environmental characteristics, provides area-level characteristics including statistics on income, education, and availability of psychiatrists in each Medicaid patient’s county of residence. The survey of Medicaid pharmacy directors collected information on each state Medicaid program’s specific policies related to prescription copayments, prescription drug limits, preferred drug lists, prior authorization, and “fail first” (i.e. step therapy) policies, in general and for anti-psychotics specifically. The survey asked each state’s Medicaid program personnel to verify state-specific policy data collected by the authors from published sources and websites against their archived documents, and also to provide data where information was missing. In addition, the dates of implementation for changes in prescription copayments and other policies, if any, were obtained for each state. Medicaid officials were also asked to identify prescription policies that applied to anti-psychotics for each year captured in our data set. The individual-level state-identified data from the MAX files were then merged with state-level policy information from 2003–2005.

Study sample

For analyses of anti-psychotic use, the primary sample included all fee-for-service Medicaid enrollees with full pharmacy benefits and a diagnosis of schizophrenia (ICD-9-CM 295.xx), identified via ≥1 inpatient or ≥2 outpatient claims with that diagnosis during a continuous 12-month identification window during 2003–2005²⁴. Patients were excluded if they were: (1) eligible for both Medicaid and Medicare (i.e. dual eligible), as Medicare is the primary payer for this population and, thus, MAX data are incomplete²⁵; (2) in risk-based Medicaid managed care organizations, since pharmacy benefits are often managed by the private managed care organization rather than the state Medicaid program; (3) less than 21 years of age at the beginning of the identification window; or (4) pregnant at any point from 2003–2005, since federal law exempts these patients from Medicaid copayments.

For the analysis of copayment changes on use of anti-psychotics and medications for hypertension, hyperlipidemia, and diabetes, we selected patients from the primary sample who also had a comorbid diagnosis for hypertension (ICD-9-CM 401–405), hyperlipidemia (ICD-9-CM 272.0–272.4), or diabetes (ICD-9-CM 250, 357.2, 362.01, 362.02, or 366.41), identified via ≥1 inpatient or ≥2 outpatient claims with those diagnoses during a continuous 12-month identification window across 2003–2005.

A primary analytic file with longitudinal, monthly observations was constructed for each patient to capture measures of monthly use of anti-psychotics and other medications and to detect whether changes in medication use occurred before or after a copayment change in the patient’s state of residence. Observations began with the first month of the 12-month identification window, through December 2005. After the 12-month identification window, individual months were excluded if the patients in that month had a change in status that corresponded to any exclusion criterion (i.e. managed care enrollment, dual status, or no pharmacy benefits). This produced 6,115,454 person-month observations from 202,709 unique patients over the study period. For analysis of anti-hypertensive, lipid-lowering, and anti-diabetic medications (among patients who had ≥1 inpatient or ≥2 outpatient claims for the related condition during the 12-month identification window), the analytic files were sub-sets of the primary analytic file, with monthly observations beginning from the month of the first claim with a comorbid diagnosis of hypertension, hyperlipidemia, or diabetes in the 12-month identification window until December 2005.

Anti-psychotic and other medication use measures

Use of anti-psychotics, overall and by FGA/SGA status, was captured through medication adherence and number of 30-day supply-equivalent prescription fills²⁶. First, monthly measures of adherence to anti-psychotics for each patient were calculated using a measure of the proportion of days covered (PDC)²⁶. The PDC is calculated as the number of days with any anti-psychotic drug supply available in the month divided by the number of days in the month²⁷. The monthly PDC was also captured separately for FGA and SGA medications.

Second, the number of anti-psychotic prescription fills per month was calculated for each patient over the study period. The number of prescription fills was converted into 30-day supply equivalents to permit standardized comparisons before and after policy changes, as well as between the study group and comparison group. As with PDC, the number of fills for anti-psychotic medications was also calculated separately for FGA and SGA medications. The number of prescription fills per month for other prescriptions (non-anti-psychotic medications [i.e. all other medications], non-anti-psychotic mental health medications, and non-mental health medications) was calculated in a similar manner for each patient in each month of the study period. Similar measures were calculated for cardiometabolic medications including lipid-lowering drugs, anti-hypertensives, or anti-diabetic agents, respectively, among the sub-groups of patients with these comorbidities.

Prescription copayment measures

Our primary independent variables of interest included state-level prescription copayment requirements in each month. These were captured using two variables in the model specifications. The first key independent variable was the minimum copayment required to obtain a prescription in each state in each month. For states with tiered copayments in a given month, the lowest or generic copayment was assigned as the minimum copayment level. For anti-psychotics, this represented the minimum copayment required to obtain any anti-psychotic prescription (i.e. FGA or SGA).

Our second key independent variable captured the difference between the maximum and minimum copayment (or the copayments for brands and generics) in each state in each month. States that had no copayments or flat copayments in a given month had a $0 differential assigned in that month. For anti-psychotics, this variable represented the copayment differential faced by the patient for SGA medications (available only as brand medications during the study period) vs FGA medications (available as generics) in each month.

Analyses

Patient-level fixed-effects models with robust estimation were used to examine how copayment changes affected monthly anti-psychotic medication use, other mental health medication use, and all other medication use, using observations from patients in states that did not implement policy change as controls²⁸. This fixed-effects estimation strategy controls for relevant time-invariant patient-level and state-level variables²⁹.

The analyses included time-varying patient-level, county-level, and state-level factors as control variables. The prescription drug hierarchical condition category (RxHCC) risk score for each patient was included to control for prescription need and comorbidity burden^30–33. “Adjustments were also made for county-level variables—metropolitan status (urban/rural), mean per capita income, unemployment rate, and education level—that are known to influence medication adherence”³⁴. Measures of the supply of mental healthcare resources at the county level of each beneficiary’s residence included psychiatric hospital beds per 10,000 persons, and the number of psychiatrists per 10,000 persons. Time varying state-level Medicaid policies were also included as control variables. They included limits on the number of prescriptions per month/year, fail first policies, prior authorization programs, and preferred drug lists.

To test the robustness of study findings, we conducted sensitivity analyses using different sample inclusion criteria. We relaxed the diagnostic sampling criteria related to the identification of patients with schizophrenia (i.e. requiring only one or more inpatient or outpatient claim in the 12-month identification window), the requirement for length of continuous eligibility over the study period (i.e. continuously eligible between 2003–2005), and length of the identification window for the schizophrenia diagnosis (i.e. > 1 inpatient or outpatient claim any time between 2003–2005). All statistical analyses were carried out using Stata version 11.1 (Stata Corp, College Station, TX).

Results

Table 1 displays the distribution of copayment requirements across the 44 states and DC at 6-month intervals between January 1, 2003 and December 31, 2005. A total of 12 state Medicaid programs had prescription copayment policy changes during the study period, which included 11 changes in generic or minimum copayments and nine changes in brand or maximum copayments. Although a majority of the states had no copayment requirement change during the study period, there was significant variation in magnitude of the changes when they did occur. Five states increased their minimum copayments by $2 or more, while six states implemented a copayment that increased out-of-pocket cost differentials between brand and generic drugs by $2 or more. Three states experienced copayment decreases for generic medications because of changes from a flat copayment model to either tiered or scaled copayment models.

Table 1. State Medicaid Prescription Copayments, 2003 to 2005

	January 2003	July 2003	January 2004	July 2004	January 2005	July 2005	December 2005
Total number of states	45	45	45	45	45	45	45
Number of states changing prescription copayments since previous period	n/a	3	3	2	2	4	1
Generic/minimum copayments
$0	18	19	19	19	18	18	18
$0.50–<$1	9	10	10	8	9	8	8
$1–<$2	11	9	10	11	12	11	11
$2–$3	7	7	6	6	6	8	8
Number of states changing generic copayment since previous period	n/a	2	1	2	2	3	1
Brand/maximum copayments
$0	17	17	17	17	16	16	16
$1–<$2	6	4	4	3	3	2	2
$2–<$2.50	8	8	6	7	6	6	6
$2.50–$3	14	16	18	18	20	21	21
Number of states changing brand copayments since previous period	n/a	2	2	1	2	1	1
Brand/generic copayment differential
$0	28	26	25	26	24	25	25
$0.50–$1	4	4	4	4	4	4	4
$1.50–$2	5	5	5	5	6	6	6
$2.50–$3	8	10	11	10	11	10	10
Number of states changing brand/generic copayment differential since previous period	n/a	3	2	2	2	3	1

Note: Fourteen states (Delaware, Idaho, Illinois, Indiana, Iowa, Kentucky, Massachusetts, Mississippi, New Hampshire, New Jersey, North Carolina, Tennessee, West Virginia, and Wyoming) had copayment changes between 2003–2005. There were 1,415,780 person months representing 47,035 patients in these states with copayment changes.

Patient characteristics are presented for the entire sample (N = 202,709) and separately for patients in states that did (n = 47,035) and did not (n = 155,674) change Medicaid prescription copayments (Table 2). More than half (55%) of the overall sample was <45 years of age, 51% were males, and 43% were White. Patients in states that implemented copayment changes were more likely to be Black and to live in rural areas and counties with lower incomes than those in states that did not implement copayment changes. On the other hand, patients in both the study and comparison groups were quite similar in their comorbidity burden. In general, baseline medication use measures were also similar across both groups.

Table 2. Sample baseline characteristics.

Characteristics	All patients	Patients in states with copayment change	Patients in states without copayment change	p-value
n	202,709	47,035	155,674
Age, years				<.001
21–34	27.1%	28.1%	26.8%
35–44	28.5%	28.3%	28.5%
45–54	28.3%	27.5%	28.5%
≥55	16.1%	16.1%	16.2%
Gender				<.001
Female	48.8%	49.7%	48.5%
Male	51.2%	50.3%	51.5%
Race				<.001
White	42.7%	49.6%	40.6%
Black	34.1%	42.2%	31.6%
Hispanic	7.2%	2.7%	8.6%
Other race	16.1%	5.6%	19.2%
Urban	82.1%	71.7%	85.3%	<.001
Residence in a county with low education levels^a	16.7%	13.4%	17.7%	<.001
Mean county-level income	$36,603	$29,359	$38,791	<.001
County-level unemployment rate	6.7%	6.7%	6.7%	.054
Number of psychiatrists per 10,000 persons in county	0.3	0.1	0.3	<.001
Number of psychiatric hospital beds per 10,000 persons in county	0.2	0.2	0.2	<.001
RxHCC risk score	0.63	0.64	0.63	.002
Hypertension^b	24.7%	26.6%	24.2%	<.001
Hyperlipidemia^b	12.7%	11.6%	13.1%	<.001
Diabetes^b	16.2%	16.7%	16.1%	<.001
Baseline medication use measures^c
Mean monthly anti-psychotic adherence (PDC)
All anti-psychotic medications	0.52	0.51	0.53	<.001
First generation anti-psychotics	0.11	0.11	0.11	.003
Second generation anti-psychotics	0.46	0.45	0.46	<.001
Mean monthly number of 30-days supply prescription fills
All anti-psychotic medications	0.74	0.72	0.74	<.001
First generation anti-psychotics	0.13	0.12	0.13	<.001
Second generation anti-psychotics	0.61	0.60	0.61	<.001
All non-anti-psychotic medications	2.49	2.59	2.46	<.001
Non-anti-psychotic mental health medications	0.92	0.91	0.92	.510
Non-mental health medications	1.57	1.68	1.54	<.001
Anti-hypertensive medications among those with hypertension	1.00	1.01	1.00	.003
Lipid lowering medications among those with hyperlipidemia	0.54	0.59	0.53	<.001
Anti-diabetic medications among those with diabetes	0.78	0.77	0.78	.480

RxHCC, Prescription Drug Hierarchical Condition Category.

^a Residing in a county with ≥25% adults aged 25–64 years without high school diploma in calendar year 2000.

^b Sub-group of patients who had ≥1 inpatient or ≥2 outpatient claims for the related condition during the 12-month identification window.

^c Calculated based on use in the first month of the 12-month identification window for all medication use measures except the measures on anti-hypertensives, lipid lowering drugs, and anti-diabetic drugs which were measured in the month with the first diagnosis of hypertension, hyperlipidemia, and diabetes during the 12-month identification window.

Table 3 shows the results of fixed-effects regression analyses, whereby within-person changes in medication use were modeled as a function of within-person changes in copayments. In intervention group states, some of those months were subject to changes in copayment amount. In comparison group states, the copayments remained constant over time. The resulting beta coefficients represent the average change in medication use associated with a $1 increase in the medication copayment variable. Results show that prescription copayment changes had a statistically significant but small effect on anti-psychotic use. For every $1 increase in the generic or minimum copayment, there was a reduction of 0.014 anti-psychotic fills per patient per month. The copayment changes also had a statistically significant but small impact on adherence and prescription refills of both FGA and SGA medications (Table 3). In particular, the increases in generic/brand differentials had a minimal impact on reducing SGA (i.e. brand) medication use (–0.007) and almost no impact in terms of magnitude in increasing FGA (i.e. generic) medication use (–0.001). On the other hand, the impact of these copayment changes on non-anti-psychotic medication use was substantially higher. For every $1 increase in the generic or minimum copayments, there was a reduction of 0.23 other medication fills per patient per month. In other words, facing a $1 generic copayment increase, a patient with schizophrenia would reduce one anti-psychotic fill every 71.4 months (1/0.014) and cut one non-antipsychotic fill every 4.3 months (1/0.23). Similarly, for every $1 increase in the generic/brand copayment differential, there was a reduction of 0.15 non-anti-psychotic fills per patient per month, translating into one non-anti-psychotic fill being missed every 6.6 months (1/0.15).

Table 3. Impact of copayment changes on monthly adherence and number of 30-day supply equivalent prescription fills.

		All anti-psychotics	Second generation anti-psychotics	First generation anti-psychotics	Non-anti-psychotic medications	Non-anti-psychotic mental health medications	Non-mental health medications
Monthly adherence (PDC)
Generic copayment
Estimate	–0.005**		–0.005**	–0.001	NA	NA	NA
95% CI	(–0.007 to –0.003)		(–0.007 to –0.003)	(–0.002 to 0.001)
Generic/brand differential
Estimate	–0.007**		–0.007**	–0.001	NA	NA	NA
95% CI	(–0.009 to –0.006)		(–0.008 to –0.006)	(–0.002 to 0.000)
Number of 30-day supply equivalent prescription fills
Generic copayment
Estimate	–0.014**		–0.013**	–0.001	–0.229**	–0.050**	–0.169**
95% CI	(–0.018 to –0.009)		(–0.018 to –0.009)	(–0.003 to 0.001)	(–0.245 to –0.213)	(–0.056 to –0.043)	(–0.181 to –0.156)
Generic/brand differential
Estimate	–0.019**		–0.018**	–0.001**	–0.148**	–0.050**	–0.094**
95% CI	(–0.022 to –0.016)		(–0.021 to –0.015)	(–0.002 to –0.000)	(–0.160 to –0.137)	(–0.054 to –0.045)	(–0.104 to –0.085)

PDC, proportion of days covered.

** p < 0.05

Note: Non-anti-psychotic medications include non-anti-psychotic mental health medications and non-mental health medications.

As compared to anti-psychotic fills, the percentage reduction in non-anti-psychotic fills was over 4-times higher for every $1 increase in generic copayments (8.8% vs 1.9%) and over 2-times higher for every $1 increase in the generic/brand copayment differential (5.7% vs 2.6%; Figure 1). Further examination within the type of non-anti-psychotic drug fills indicated that the relative impact of every $1 increase in generic copayments was almost twice as large for non-mental health drugs as compared to non-anti-psychotic mental health drugs (10.1% vs 5.4%, Figure 1).

Figure 1. Percentage reduction in number of 30-days supply equivalent prescription fills per $1 increase in copayments by anti-psychotic vs non-anti-psychotic drug classes.

Figure 2 shows the relative comparisons of the percentage reductions in anti-psychotics associated with copayment increases as compared with reductions in medications for diabetes, hypertension, and hyperlipidemia in the sub-groups of patients with those conditions. The relative reduction in use due to generic copayment increases was 3.6–11.3-times for all three cardiovascular and metabolic drug classes as compared to anti-psychotics. The relative reduction in use due to generic/brand copayment differential increases was also higher for lipid-lowering drugs and anti-hypertensives as compared to anti-psychotics. Sensitivity analyses using alternative sample inclusion criteria showed similar findings (data not shown).

Figure 2. Percentage reduction in number of 30-day supply equivalent prescription fills per $1 increase in copayments by anti-psychotics vs cardiovascular and metabolic drugs.

Discussion

This study is the first to evaluate the longitudinal impact of Medicaid prescription copayment changes on medication use for mental health and cardiometabolic conditions among patients with schizophrenia, using Medicaid data from 42 states and DC. We found that small prescription copayment increases and generic/brand copayment differentials resulted in only a minimal decline in anti-psychotic medication use and adherence (as measured by PDC), but more substantial reductions in use of other medications.

For each $1 increase in the minimum copayment amount, we observed a reduction of 1.4 anti-psychotic drug fills per 100 patient months, corresponding to a relative reduction of ∼1.9% in the mean number of fills per month. From a clinical perspective, it is reassuring that these small increases in prescription copayments had only a limited impact on anti-psychotic medication use, since these drugs are central to the management of schizophrenia, and anti-psychotic non-adherence is a leading source of preventable morbidity in the community management of schizophrenia³⁵. At the same time, there is an established relationship between anti-psychotic adherence and risk of hospitalization in this population, so each missed prescription fill has the potential to influence clinical outcomes.

From a Medicaid program perspective, generic/brand copayment differentials, which might be expected to lead to decreases in use of second-generation anti-psychotics (that were available only as brand medications during our study period), in fact had minimal impact on reducing SGA use and increasing FGA use. This is particularly interesting in light of evidence from large pragmatic clinical trials, suggesting there are few clinical differences between FGAs and SGAs^36–39. Our findings provide important data for policy-makers who may be considering broader implementation of differential (or tiered) copayment policies.

Interestingly, increases in prescription copayments and generic/brand copayment differentials had a larger negative effect on non-anti-psychotic medication use. For each $1 increase in minimum copayment, there was a reduction of 23 non-anti-psychotic drug fills per 100 patient months, or a relative reduction of ∼10.1%. Similarly, for each $1 increase in generic/brand copayment differential, there was a reduction of 15 non-anti-psychotic drug fills, or a relative reduction of ∼5.6%. The fact that individuals with schizophrenia appear to be less responsive to increases in copayments for their anti-psychotic medications compared with other medications suggests that the price elasticity of demand for anti-psychotic medications was substantially lower than for non-anti-psychotics. These findings also moderate the long-held perception that the demand for mental health services is particularly sensitive to price⁴⁰. These findings also corroborate other research suggesting that patients with certain chronic diseases are not indiscriminate with regard to the choice between disease-specific medications and other medications⁴¹. Patients were also more likely to forego their non-mental health drugs (e.g. statins) than other mental health drugs (e.g. anti-anxiety agents, anti-depressants). For example, among patients with depression, price elasticity estimates for anti-depressants were reported to be –0.08, whereas it was reported to be –0.25 for non-anti-depressant drugs⁴¹. It is possible that individuals prioritized medications that were treating troubling symptoms, as opposed to more asymptomatic cardiovascular conditions. It is unknown whether these reductions had a clinically meaningful impact on the medical outcomes, quality-of-life, and healthcare costs of Medicaid beneficiaries diagnosed with schizophrenia. Nevertheless, it is of concern that individuals were foregoing prescription fills for medications that are used to treat major cardiovascular and metabolic diseases^42,43, which are substantial contributors to premature mortality and reduced life expectancy in individuals with schizophrenia¹⁵. Thus, it is important that clinicians and pharmacists counsel such patients and help them make decisions regarding treatment choices in the face of increased cost sharing.

Several study limitations deserve mention. First, Medicaid copayments were not mandatory during our study period. By federal law, pharmacists cannot deny filling prescription drugs for Medicaid patients who cannot make their copayments. However, work by Fahlman et al.⁴⁴ indicates that 90% of pharmacists always collect copayments. In addition, even if copayments were waived for some patients in states that increased copayments, such waivers would bias our results towards the null; thus, our results may under-estimate the true relationship between copayments and medication use. Second, our study identified patients with schizophrenia in the 3-year study period, but did not include data on duration of illness or prior treatment history. As a result, we are not able to determine whether copayment policies had a differential impact on newly-diagnosed individuals who were initiating anti-psychotic treatment for the first time, as compared to individuals who were further along in their treatment course. We do not expect that duration of illness or treatment history would have differed across our intervention and comparison states, however. Finally, whereas our quasi-experimental study design took advantage of natural variation in state Medicaid copayment policies, our categorization of states into study vs comparison groups may not have been exact if there were other concurrent changes in Medicaid policies (e.g. prior authorization, step therapy requirements, physician visit copays) during the 3-year period. This concern is minimized by the fact that our survey of state Medicaid personnel confirmed that overlap of copayment changes with most other types of policy changes was rare during our study period.

Despite these limitations, our findings have important implications for pharmacological treatment of comorbid conditions in Medicaid patients with schizophrenia and underscore the need for clinicians and pharmacists to be attentive to the risk of interruptions in treatment when patients experience a change in out-of-pocket costs for medication, even if cost sharing increases are small. The findings also have direct implications for the future landscape of Medicaid policies at the state level. Recent legislation has permitted states to increase the maximum nominal copayment limit (otherwise historically fixed at $3 per prescription) in accordance with the percentage increase in the medical care component of the Consumer Price Index (CPI) each year. While the medical component of the CPI rises at about twice the rate of the overall CPI, the incomes of Medicaid enrollees living below the poverty line tend to rise more slowly than the CPI⁴⁵. Few states now require cost-sharing of up to $4 for preferred drugs and $8 for non-preferred drugs for all Medicaid-covered individuals, including individuals with incomes at or below 150% of the federal poverty level¹. This may place many impoverished, Medicaid patients with schizophrenia at risk of facing higher out-of-pocket costs that would consume a larger share of their limited incomes over time. Given the ongoing attempts to repeal and replace the Affordable Care Act, the situation is likely to get worse with the enormous increases in state budgetary pressures if the proposals to revamp Medicaid financing in recently proposed Republican healthcare plans come to fruition⁴⁶. Thus, ongoing research will be needed to monitor whether Medicaid prescription copayment changes reduce essential medication use and health outcomes in this vulnerable population.

Transparency

Declaration of funding

This study was supported by grant R01-MH080976 from the National Institute of Mental Health and 1-R01-HS018389-01 from the Agency for Healthcare Research and Quality.

Declaration of financial/other relationships

JAD reported serving as a consultant for Allergan, Ironwood Pharmaceuticals, Sanofi, Shire, and Vertex; and had received research funding from Abbvie, Biogen, Humana, Janssen, PhRMA, Pfizer, Regeneron, Sanofi, and the National Pharmaceutical Council. Her spouse holds stock in Merck and Pfizer. SCM reported serving as a consultant for Allergan, Janssen, and Eli Lilly. PL and SD have no relevant financial relationships to disclose. Peer reviewers on this manuscript have received an honorarium from JME for their review work, but have no other relevant financial relationships to disclose.

Acknowledgments

No assistance in the preparation of this article is to be declared.

Appendix: R² for the models in Table 3

Table

	All anti-psychotic	Atypical anti-psychotics	Conventional anti-psychotics	Non-anti-psychotic medication	Other mental health medication (excluding anti-psychotics)	Non-mental health medication
Monthly adherence (PDC)	0.0044	0.0046	0.0007	NA	NA	NA
Nmber of 30-day supply equivalent prescription fills	0.004	0.0041	0.0005	0.0168	0.0049	0.0165