Medication Complexity and Affordability in Use of Time Release Antidepressants

Introduction

Time release (TR) drug formulations reportedly offer advantages in improving medication adherence for pharmacological treatment of depression as compared with their immediate‐release (IR) counterparts.1, 2, 3, 4, 5, 6 TR formulations require fewer drug administrations by slowly and consistently releasing a drug into the blood stream over an extended period of time.7 TR formulations are further claimed to improve drug tolerability by lowering peak plasma drug concentrations and reducing fluctuations between peak and trough plasma drug concentrations.5 While greater tolerability increases the likelihood that patients continue to take the medication in theory,8, 9, 10 empirical evidence is inconclusive as to whether TR formulations in fact reduce the rate of medication discontinuance compared to IR formulations.11

Most TR formulations cost more than IR counterparts with monthly drug therapy costs ranging from $123 to more than $330.12 Many drug plans are thus reluctant to designate TR formulations as a preferred tier for insurance coverage unless they are generically available. As the result, their out‐of‐pocket costs tend to be higher than IR formulations. Further, TR formulations are not likely to be generically available immediately following the expiration of market exclusivity. Current regulations of the Food and Drug Administration (FDA) impose a 30‐month stay in generic drug approval when the incumbent pioneer challenges the generic drug maker's patent. Under the regulations, the TR technology patent is more often challenged because of its inherent complexity and sophistication. Finally, there is a perception that generic TR formulations are not equivalent to the pioneer TR formulations.13 In fact, generic TR formulations have been recalled for drug safety concerns.14, 15 Generic drug makers thus face more entry barriers when developing TR formulations than they do when developing IR formulations. As a result, TR formulations would not be as affordable as IR formulations despite the availability of generic versions.

The benefit of drug regimen simplification associated with TR formulations thus comes at a cost. While those who need to simplify drug therapies are willing to pay additional amount for TR formulations, those patients who face financial constraints may not afford the increased cost of TR formulations. This study aimed to examine how the trade‐off between medication affordability and complexity affects the likelihood of using TR formulations among adults with diagnosed depression. Study hypotheses are (1) that the likelihood of TR use is lower for patients who face economic constraints in obtaining medications, and (2) that the likelihood of TR use is higher as medication complexity increases.

Methods

Study design and data source

This study was based on a retrospective analysis of prescription drug utilization data collected three times a year for patients with a diagnosed depression. The prescription utilization data came from the Medical Expenditure Panel Survey (MEPS) for the year 2010. MEPS is a nationally representative longitudinal household survey of health care use, expenditures, sources of payment and health insurance coverage for noninstitutionalized U.S. civilians.16 The prescription drug utilization data are constructed for each event reported first by survey respondents and then supplemented with pharmacy records by contacting pharmacies that survey respondents reported to have visited. Each event has a record of drug name, number of pills dispensed, payment, days of supply, sources of payment, and national drug code (NDC).

Study subjects

Study subjects consisted of adults older than or equal to 24 years of age with a reported diagnosis of depression among noninstitutionalized US civilians. The age was based on the end of the year 2010, in which study subjects were sampled for the MEPS. International Classification of Diseases, Ninth Revision, Clinical Modification (ICD‐9‐CM) diagnosis code (311) was used to identify depression.17 Further, to be included in this study, study subjects must have filled at least one prescription for antidepression medication. The antidepression medication was determined based on Multum therapeutic class codes in the MEPS: 76, 208, 209, 249, 306, 307, and 308. The medication also had to be available generically not only for IR but also for TR formulations. When isomers exist, they were, classified as the same active substance, that is, desvenlafaxine was considered equivalent to venlafaxine. This restriction narrowed the set of antidepression medications to bupropion, paroxetine, and venlafaxine.

Measurements

TR formulation

TR formulation was dummy‐coded based on whether the formulation of a prescribed medicine dispensed was TR or not. TR formulation was determined based on FDA's drug data. The FDA documents drug nomenclature information including formulation type (IR vs. TR) for all drugs approved.18

Medication affordability

Medication affordability was represented by two variables: family income and insurance coverage. Family income was measured as a percentage of the federal poverty level (FPL) and was assigned to five categories: poor (FPL ≤ 100%), near poor (100% < FPL < 125%), low income (125% < FPL < 200%), middle income (200% < FPL < 400%), and high income (> 400% FPL). Insurance coverage was defined as overall insurance coverage. It had categories of uninsured, Medicaid, Medicare, private insurance and others.

Drug therapy complexity

Drug therapy complexity was represented by three variables: the number of medications, drug therapy duration, and comorbidity. The number of medications was measured by the number of different drugs the patient had taken during the year. Drugs were considered different when drug ingredients were different. Different drug ingredients were identified from the FDA's drug nomenclature database. Drug therapy duration was identified based on the year in which a person first started taking a medicine. These “first taken” questions were only asked the first time a prescription event was mentioned by the survey respondent but carried forward from prior rounds over refills of the prescription in subsequent rounds.19 Comorbidity was measured using Charlson comorbidity index. A Charlson comorbidity index was constructed by summing weights assigned to ICD9 codes reported for a study subject.20, 21, 22

Other predictors that can affect the likelihood of TR use were identified using the Anderson's Behavioral model of health services utilization. They were categorized into predisposing, enabling, and need‐related constructs.23 The predisposing construct consisted of sociodemographic factors such as age, gender, race, ethnicity, and geographical region. Race/ethnicity was classified as non‐Hispanic white (NHW), non‐Hispanic black (NHB), Hispanics, and others. The enabling construct consisted of education in addition to the variables related to medication affordability. The need construct basically consisted of the same variables related to drug therapy complexity factors mentioned above.

Statistical analysis

Descriptive statistics were used to summarize the utilization of TR formulations among study subjects. A multiple logistic regression model was used to identify predictors for the likelihood of TR use. In the logistic regression model, the dependent variable was the ratio of the number of TR formulations to the number of all formulations filled by each patient. SAS survey logistic procedure was used to control for complex survey sampling. The procedure incorporated person‐level weights and variance adjustment weights (strata and primary sampling unit) provided by MEPS to produce nationally representative estimates.19, 24 The procedure thus controls for clustering of prescription events within each study subject. The level of statistical significance was p ≤ 0.05, and all analyses were carried out using the statistical package SAS 9.3 (SAS Institute, Cary, NC, USA).

Results

A total of 625 people representing about 8.5 million noninstitutionalized adults in the US met the inclusion/exclusion criteria (Table 1). The majority was female (69.86%), NHW (86.73%), residents of MSA (81.44%), and had college level education (33.49%). About 46% had ages between 46 and 65, and were located predominantly in the South (33.7%). Almost 41.2% of them were from high‐income families (>400% of FPL), and had private insurance (73.02%). About 5% were uninsured. More than half of the study subjects (57%) reported no comorbid conditions (Charlson comorbidity index = 0). However, up to 30.4% had taken more than 10 different medications per year. As for the duration of drug therapy for depression, about one‐third (28.35%) had taken less than 1 year while one‐tenth had taken more than 10 years.

Table 1.

Description of study sample with a choice of TR antidepressants

Variables	Values	N	Weighted N	%
Sex
	Male	182	2,558,145	30.14
	Female	443	5,930,498	69.86
Age
	24–45	28	3,933,391	46.34
	45–65	13	1,773,267	20.89
	65 or older	205	2,781,985	32.77
Race
	Hispanic	63	464,051	5.47
	Non‐Hispanic black	58	344,404	4.06
	Non‐Hispanic white	474	7,361,987	86.73
	Other	30	318,201	3.75
Metropolitan statistical area (MSA)
	No	118	1,575,404	18.56
	Yes	507	6,913,239	81.44
Census region
	Northeast	112	1,693,918	19.96
	Midwest	170	2,226,430	26.23
	South	208	2,858,214	33.67
	West	135	1,710,081	20.15
Education
	Some high school or under	103	1,001,634	11.82
	High school graduate	179	2,086,856	24.63
	Some college	167	2,547,783	30.06
	College graduate	173	2,838,044	33.49
Family income
	Poor/negative	118	1,111,137	13.09
	Near poor	30	277,812	3.27
	Low income	77	945,197	11.13
	Middle income	193	2,659,431	31.33
	High income	207	3,495,067	41.17
Insurance coverage
	Private	398	6,198,002	73.02
	Public	187	1,874,194	22.08
	Uninsured	40	416,447	4.91
Number of medications taken
	0–3 medications	91	1,189,961	14.02
	3–6	160	2,209,243	26.03
	6–10	179	2,508,994	29.56
	10 or more	195	2,580,445	30.40
Charlson comorbidity index
	0	343	4,881,771	57.51
	1	75	1,011,095	11.91
	2	105	1,283,575	15.12
	3 or higher	102	1,312,202	15.46
Depression therapy duration
	≤1 year	158	2,154,868	28.35
	1–2	106	1,348,832	17.74
	2–5	110	1,485,621	19.54
	5–10	131	1,833,104	24.11
	>10	55	779,594	10.26
Total		625	8,488,643	100.00

The study subjects all together filled 3,745 antidepressants that offered a choice between TR and IR formulations (Table 2). Overall, TR formulations were used 59.48% of the time (95% CI: 55.19%–64.10%) with substantial variations across different patient groups. The use of TR formulations was highest among NHW (61.68%); in comparison, Hispanics had used TR formulations 44.03% of the time and NHB used them 45.22% of the time (Figure 1 A). Females were also more likely to use TR formulations than males (63.52% vs. 50.01%). As for variables related to medication affordability (Figure 1 B), those who had private insurance were most likely to use TR formulations (67.03%) while those uninsured having the lowest rate of TR use (32.30%). The rate of TR use tended to fluctuate with increases in family income levels. The rate of TR use also varied with different levels of medication therapy complexity. The higher the number of medications taken per year, the higher the rate of TR use became (Figure 1 C). However, the rate of TR use declined with increases in Charlson comorbidity index.

Table 2.

Bivariate analysis of the likelihood of using TR formulations of antidepressants

Variables	Rxs	TR Use	LCI	UCI
Gender
Male	1,005	50.01%	39.38%	60.64%
Female	2,740	63.52%	58.73%	68.30%
Race/ethnicity
HSP	359	44.03%	29.81%	58.25%
NHB	323	45.22%	27.59%	62.86%
NHW	2,846	61.68%	56.67%	66.69%
OTH	217	53.30%	31.63%	74.98%
Poverty level
Poor	789	59.52%	49.19%	69.85%
Near poor	157	50.09%	31.10%	69.09%
Low income	519	61.91%	50.96%	72.85%
Middle income	1,163	53.26%	44.16%	62.35%
High income	1,117	65.33%	57.65%	73.02%
Insurance status
Private	2,213	67.03%	61.37%	72.69%
Public	1,318	44.36%	36.23%	52.49%
Uninsured	214	32.20%	14.02%	50.39%
Drug therapy complexity (number of medications)
0–3	397	51.38%	35.31%	67.45%
3–6	934	52.02%	41.56%	62.48%
6–10	1,061	62.72%	54.28%	71.15%
10+	1,353	65.67%	58.38%	72.95%
Charlson comorbidity index
0	1,964	64.35%	59.19%	69.51%
1	456	62.85%	45.56%	80.15%
2	669	45.47%	34.21%	56.74%
3+	656	54.35%	43.71%	64.99%
Drug therapy duration (years)
≤1 year	537	62.59%	53.61%	71.58%
1–2	681	63.97%	51.99%	75.96%
2–5	711	66.60%	57.82%	75.38%
5–10	1,027	58.26%	48.50%	68.01%
>10	372	48.37%	29.74%	67.00%
Education
1	669	44.96%	31.53%	58.39%
4	1,139	52.29%	41.54%	63.04%
5	935	66.73%	58.36%	75.10%
6	982	65.33%	56.49%	74.16%
MSA: 0	800	64.81%	55.31%	74.32%
1	2,945	58.28%	53.25%	63.31%
Region
NE	734	65.65%	56.66%	74.64%
MW	1,005	60.13%	51.16%	69.11%
South	1,186	61.31%	53.35%	69.26%
West	820	50.31%	39.13%	61.49%
Total	3,745	59.48%	55.19%	64.10%

Figure 1.

Variations in use of TR antidepressant by (A) gender and race/ethnicity, (B) affordability factors, and (C) medication complexity factors. Error bars are 95% confidence intervals. Meds = medications; TR = time release; Yrs = years.

A multivariable logistic regression model depicts whether likelihoods of TR use are associated with factors related to medication affordability and complexity controlling for all other variables related to Anderson's behavioral model (Table 3). As expected, variables related to medication affordability had significant associations with the likelihoods of TR use. Patients from middle family income had significantly lower odds of using TR formulations (OR = 0.57, p < 0.01) than those from the highest family income. However, rather surprisingly, patients from the poorest family income (≤100% FPL) had a significantly higher likelihood of TR use compared to patients from the highest family income (OR = 1.97, p = 0.01). It is important to note this significant finding, unlike the one from the bivariate analysis, had resulted from the multivariate logistic regression that controls for all other variables. The multivariate logistic results also produced a significantly lower likelihood of TR use among patients without insurance compared to patients with private insurance (OR = 0.52, p = 0.02). However, there was no significant difference between public and private insurance.

Table 3.

Logistic regression model for time release prescription specific events

Variables	Levels	Parameter estimate	p Value	OR	95% CI
Intercept		–0.66	0.01a	0.52	(0.31–0.86)
Sex	Male
	Female	0.28	0.02a	1.33	(1.04–1.69)
Age	24–45
	45–65	0.05	0.73	1.05	(0.79–1.41)
	65 or older	–0.28	0.16	0.76	(0.51–1.12)
Race	NHW
	Hispanics	–0.17	0.55	0.85	(0.49–1.47)
	NHB	–0.27	0.52	0.76	(0.33–1.75)
	Other	0.35	0.40	1.42	(0.63–3.24)
MSA	No
	Yes	0.03	0.81	1.03	(0.79–1.35)
Region	West
	NE	0.22	0.28	1.24	(0.84–1.84)
	MW	0.14	0.45	1.15	(0.80–1.67)
	South	0.11	0.60	1.11	(0.75–1.64)
Education	College graduate (reference)
	<High school	–0.71	0.00b	0.49	(0.30–0.80)
	High school	–0.03	0.88	0.97	(0.66–1.43)
	Some college	0.33	0.06	1.40	(0.98–1.99)
Family income	High income (reference)
	Poor	0.68	0.01b	1.97	(1.18–3.27)
	Near poor	0.10	0.78	1.10	(0.56–2.18)
	Low income	0.04	0.88	1.04	(0.60–1.80)
	Middle income	–0.56	0.00b	0.57	(0.40–0.82)
Insurance	Private
	Public	–0.18	0.35	0.84	(0.58–1.22)
	Uninsured	–0.66	0.02a	0.52	(0.30–0.89)
Number of medications	0–3 Meds
	3–6 Meds	–0.66	0.00b	0.52	(0.35–0.77)
	6–10 Meds	0.25	0.17	1.28	(0.90–1.82)
	10 or more Meds	1.15	<0.0001b	3.15	(1.97–5.03)
Drug therapy duration	0–1 yr
	1–2 yrs	0.07	0.73	1.08	(0.71–1.63)
	3–5 yrs	0.38	0.11	1.47	(0.92–2.35)
	6–10 yrs	–0.10	0.63	0.91	(0.61–1.35)
	More than 10 yrs	–0.68	0.05a	0.51	(0.26–1.00)
Charlson comorbidity	0 (no comorbidity)
	1 (moderate)	0.44	0.11	1.55	(0.90–2.67)
	2 (severe)	–0.79	0.00b	0.46	(0.30–0.70)
	3 or higher (extreme)	–0.20	0.37	0.82	(0.52–1.28)

CI = confidence interval; Meds = medications; MSA = metropolitan statistical area; NHB = non‐Hispanic black; NHW = non‐Hispanic white; OR = odds ratio; yrs = years.

^a p < 0.05,

^b p < 0.01.

As for the medication complexity, patients with the highest complexity (more than 10 different medications per year) had a much higher odds of TR use (OR = 3.15, p < 0.001) than those referent patients with the lowest number of medications taken in the year (less than or equal to 3). However, patients with a moderate complexity (3–6 medications per year) had a lower odds of TR use (OR = 0.52, p < 0.01) than the reference patients when all other variables were held constant. Likewise, drug therapy duration did not have a consistent relationship with likelihoods of TR use. Instead, it had an up and down relationship with the likelihood of TR use. The likelihood of TR use had a tendency to increase with increases in the number of years that patients had been on antidepressants until 5 years, but then to decrease thereafter. However, the only statistically significant association was among patients who had had more 10 years of antidepressant drug therapy compared to the referent who had taken less than 1 year (OR = 0.51, p = 0.05). Similarly, Charlson comorbidity also showed the cyclical relationship with the likelihood of TR use. As Charlson comorbidity increased from 0 to 1, the likelihood of TR use increased (OR = 1.55, p = 0.11). However, when the index increased to 2, the odds of TR use was significantly against compared to that for the index of 0 (OR = 0.46, p < 0.01).

The logistic regression model also shows what other variables related to the Anderson's behavioral model of health service utilization are significantly associated with likelihoods of TR use. As for predisposing variables such as gender and race/ethnicity, females were still significantly more likely to use TR formulations as compared to males (OR = 1.33, p = 0.02) even when all other variables were controlled. However, the distinctive difference among race/ethnicity in the bivariate analysis no longer existed when all other variables were controlled. Further, there were no significant associations either for different ages or for different geographical regions. However, education was found significantly associated with the likelihood of TR use. Compared to patients with a college degree, those without a high school diploma had a lower likelihood of TR use (OR = 0.49, p < 0.01).

Discussion

TR formulations provide the benefit of reduced medication complexity while being less affordable than their IR counterparts. This study thus hypothesized there will be a trade‐off between medication affordability and complexity in the likelihood of using TR formulations. Specifically, we hypothesized that the likelihood of TR use is lower for patients who face problems of medication affordability such as the uninsured and the poor, and that the likelihood of TR use is higher among patients who need to reduce medication complexity.

As hypothesized, insurance status and family income were significantly associated with the likelihood of using TR formulations. Patients who were uninsured were significantly less likely to use TR formulations compared to those who had private insurance, which is consistent with the literature.25 Thus, it is reasonable to assume that patients without insurance must have experienced financial barriers in getting TR formulations. The study results also showed that patients with middle family income were less likely to use TR formulations compared to patients with the higher family income (>400% FPL). However, one striking finding was that patients with a poor family income (≤100% FPL) were significantly more likely to use TR antidepressants (OR = 1.97, p = 0.01) than those with the highest family income (>400% FPL) controlling for all other variables; in contrast, the poor had a lower likelihood of TR use than the rich in the bivariate analysis. While striking at the first glance, the finding could simply suggest that poor patients may have obtained TR formulations from alternative sources such as samples and prescription drug assistance programs. Typically, all major pharmaceutical manufacturers make their drugs such as TR formulations available for free through prescription drug assistance programs. Further, the poor patients could have received low income subsidies that lower their financial burdens in obtaining prescription drugs. For example, currently, Medicare beneficiaries receive financial support to help pay their premiums, deductibles, and copayments.26 This finding could fuel the discussion whether low income subsidies have to expand or to shrink when in fact TR formulations are beneficial to medication treatment.

As hypothesized, factors related to medication complexity were also significantly associated with the likelihood of TR use. The likelihood of TR use was the highest among patients taking more than 10 different medications per year (OR = 3.15, p < 0.001). This finding indicates there is a greater need for reduced medication complexity among those patients who face medication complexity. However, patients who were on 3–6 medications per year were less likely to use TR formulations (OR = 0.52, p < 0.01) compared with the referent patients who were on 0–3 medications per year. The inconsistent finding across different numbers of medications could indicate that the number of medications per year may not accurately represent the concept of medication complexity.

There were two other variables (drug therapy duration and Charlson comorbidity index) this study used to represent medication complexity. Their impacts on the likelihood of TR use were also inconclusive. They both exhibited a cyclical, up‐and‐down relationship with the likelihood of TR use. For example, patients who had been on an antidepressant for 3–5 years were more likely to use TR formulations, although not statistically significant (OR = 1.47, p < 0.11), than those referent patients who had stayed on less than a year. However, patients who had stayed on the drug therapy for more than 10 years were much less likely to use TR formulations than those referent patients (OR = 0.51, p < 0.05). This reversal in the likelihood of TR use suggests that the longer drug therapy duration could impose a heavier financial constraint.

A similar trend was also found for the Charlson comorbidity. While patients with a Charlson comorbidity index of 1 had an insignificant but higher likelihood of TR use than those referent patients with “0” comorbidity index, patients with an index of “2” were much less likely to use TR formulations than those referent patients (OR = 0.46, p = 0.05). Evidently, TR formulations could have been beneficial for patients with moderate comorbidity but not for those with severe comorbidity. It is also possible that obtaining affordable medications is more important than reducing medication complexity when patients face severe comorbidity.

As for racial and ethnic differences, bivariate analysis showed distinctively different results. However, there were no statistical significant differences across race/ethnicities in the multivariate logistic regression analysis. In terms of a gender difference, however, female patients were more likely to use TR formulations (OR = 1.33, p = 0.02) than their male counterparts. It is well known that females are the majority users of antidepressants.10, 27 Given the more frequent use of antidepressants, females would have been aware of the benefits of TR formulations, and might have opted for TR drug therapy. Finally, education was significantly associated with the likelihood of TR use. Education empowers patients to assess cost effectiveness of TR formulations. It also gives more opportunities for high income. Thus, patients without a high school diploma were naturally less likely to use TR formulations than those with a college degree (OR = 0.49, p < 0.01).

Limitations

This study has some limitations worthy of detailing. First, identification of diagnosed depression and drug therapy regimen was based on patient self‐reports. It is possible that there might be some underreporting of the condition or recall bias for some of the medications, which was not captured by MEPS. One measure used for this study to represent medication complexity was the number of different medications taken within the year. A better measure for medication complexity could be the one that controls for synchronicity and multiplicity of daily pill takings for all the medications that the patient takes concurrently. Next, there will be other factors that this study did not include but affect the likelihood of using TR formulations. Practice guidelines, practice styles, and pharmacy benefit designs all affect the likelihood of TR use. Although their impacts on TR use could be random, it is worthwhile to examine how those factors affect the utilization of TR formulations in the future.

Conclusion

This study found that there is a trade‐off between medication affordability and complexity in the likelihood of using TR antidepressant formulations. While those who face medication affordability were less likely to use TR formulations, those who face medication complexity were more likely to use TR formulations. Identifying the trade‐off associated with the use of TR formulations would contribute to improving access as well as medication adherence to antidepressant drug therapy among adults with depression.

Conflict of Interest

All authors have no conflict of interest to declare.