Sociodemographic Determinants of Out-of-Pocket Expenditures for Patients Using Prescription Drugs for Rheumatoid Arthritis

Abstract

Background

Rheumatoid arthritis (RA) is a chronic inflammatory disease that has a substantial economic impact on patients. Patients with RA are at an increased risk for disability and for loss of income. The inclusion of biologic drugs in RA therapy has increased the cost of treatment. Little is known about the relationship between sociodemographic characteristics and the out-of-pocket (OOP) expenditures for prescription drugs for patients with RA, including biologics, disease-modifying antirheumatic drugs (DMARDs), nonsteroidal anti-inflammatory drugs (NSAIDs), corticosteroids, and analgesics.

Objectives

To explore the relationship between sociodemographic characteristics, personal characteristics, and OOP expenditures associated with RA prescription medications. A secondary objective was to measure the average OOP expenditures for different therapeutic classes of RA medications, including biologics, DMARDs, NSAIDs, corticosteroids, and analgesics.

Methods

In this retrospective analysis of Medical Expenditure Panel Survey (MEPS) data from 2009 to 2012, we identified a patient sample of 1090 adults with RA, which represented approximately 9.71 million patients in the MEPS database. The total OOP expenditure was calculated based on the OOP expenditure for each prescription drug corresponding to an individual. Patient variables included age, race, sex, insurance status, number of comorbid conditions, region, area of living, annual family income, and marital status. Logistic regression and generalized linear models were used for analysis. The mean OOP expenditure for therapeutic classes was estimated using nonparametric percentiles from 1000 cluster bootstrap estimates.

Results

Overall, the mean annual OOP expenditure was $273.99 (95% confidence interval [CI], $197.07–$364.75). The OOP expenditures were lower for privately insured (0.31; 95% CI, 0.21–0.45) patients and publicly insured (0.18; 95% CI, 0.12–0.27) patients versus uninsured patients, and for poor (0.60; 95% CI, 0.43–0.84) and low-income (0.69; 95% CI, 0.49–0.97) patients versus high-income patients. The mean annual OOP expenditure decreased with age (0.98; 95% CI, 0.97–0.99), was lower (0.73; 95% CI, 0.58–0.92) for male patients than for female patients, and increased with the presence of comorbidities (1.16; 95% CI, 1.07–1.25). The average annual OOP expenditure was highest for biologics ($2556.73), followed by DMARDs ($89.37). The average annual OOP expenditures were $27.97, $52.36, and $72.51 for corticosteroids, NSAIDs, and narcotic analgesics, respectively.

Conclusions

Age, sex, race, income level, insurance status, and comorbidity status significantly affected patient OOP expenditure. Higher OOP expenditures among the uninsured, female patients, patients with low income levels, and patients with several comorbidities could adversely affect RA therapy. The use of expensive biologics needs to be monitored to reduce prescription-related cost-sharing among patients with RA.

KEY POINTS

• ▸Little is known about how sociodemographic characteristics of patients with RA affect out-of-pocket (OOP) expenditure on prescription drugs.
• ▸This retrospective study analyzed data from the 2009–2012 Medical Expenditure Panel Survey for patients with RA.
• ▸OOP spending was lower for patients with private or public insurance versus uninsured patients, and for poor and low-income patients versus high-income patients.
• ▸The mean annual OOP expenditure decreased with age, was lower for male patients than for female patients, and increased with the presence of comorbidities.
• ▸The average annual OOP expenditure was highest for biologics ($2556.73), followed by DMARDs ($89.37), narcotic analgesics ($72.51), NSAIDs ($52.36), and corticosteroids ($27.97).
• ▸Age, sex, race, income level, insurance status, and comorbidities significantly affected OOP expenditures.
• ▸Increasing OOP expenditures among the uninsured, female patients, patients with low income, and patients with comorbidities can negatively affect medication adherence.
• ▸Adherence to biologic drugs for RA should be monitored to reduce prescription-related patient cost-sharing.

Rheumatoid arthritis (RA) is a chronic inflammatory disease that is typically accompanied by swelling, tenderness of the joints, and destruction of the synovial joints.1 The progression of RA can cause disability and can have serious physical, mental, and economic consequences for patients.1,2 Based on data from 2001 to 2005, the prevalence of RA and other inflammatory polyarthropathies was 682 per 100,000 population, or 1.481 million adults, in the United States.3 The Rochester Epidemiology Project reported an annual RA incidence rate of approximately 41 per 100,000 population in the United States between 1995 and 2007.4 The disease also has substantial clinical and economic impacts on the healthcare system. Compared with the general population, patients with RA have an increased incidence of cardiovascular diseases and higher mortality rates.5 Also, because of an increased risk for work disability, patients with RA are at risk for being without paid employment, and therefore a loss of income.5 Overall, the excess healthcare cost for patients with RA was estimated to be $19.3 billion in 2005 (the most recent data related to excess healthcare cost specific to RA).6

The American College of Rheumatology (ACR) recommends the use of anti-inflammatory pharmacologic interventions, such as nonsteroidal anti-inflammatory drugs (NSAIDs), glucocorticoids, analgesics, and disease-modifying antirheumatic drugs (DMARDs), for the treatment of RA.7 The introduction of biologics, in the form of anti–tumor necrosis factor (TNF) and non-TNF agents, has represented a critical advance in the treatment of patients with RA.8 Biologic agents are recommended by the ACR as monotherapy or are often used in combination with DMARDs in the treatment of RA.8 Because biologics are much more costly than DMARDs, their inclusion in RA therapy has significantly affected the overall cost of RA drug expenditures.9,10 As a result, a significant portion of the direct cost of RA treatment is associated with the costs of medication acquisition, monitoring, and adverse effects.11

Medications were reported to be the second largest contributor to RA-related costs, accounting for more than 25% of the total cost in a 2001 review.12 A previous study published in 1999 showed that drugs such as DMARDs and NSAIDs accounted for almost 26% of the medical care cost of RA, and patients who were uninsured or those with high drug copayments had greater financial hardship in paying for these medications.13 With the advent of biologics, the contribution of RA medication costs to the direct annual medical cost of RA has increased to almost 66%.9 This direct medical cost analysis of 7527 patients with RA in 2003 showed that biologics had the highest annual cost ($3307), followed by DMARDs ($643) and NSAIDs ($591).9

A study that used data from the 2008 Medical Expenditure Panel Survey (MEPS) demonstrated that pharmacy-related costs had become the primary driver of the incremental total expenditure associated with RA treatment, largely as a result of early aggressive treatment with DMARDs and expensive biologics.14 In that study, the average pharmacy-related expenditure for patients with RA was estimated to be $5825, which was significantly higher than $1264 for patients without RA.14 The average incremental difference in annual pharmacy-related expenditures was $1380, which accounted for almost 66% of the total incremental annual expenditure for patients with RA compared with people without RA.14

The increasing RA-related drug costs and subsequent drug expenditures have also significantly affected patient out-of-pocket (OOP) expenditures in the treatment of RA.15–18 Even patients with health insurance can face difficulties paying for their medications, because of health plan restrictions on coverage limits and therapy exclusions. The restriction of health benefit coverage had an effect on the likelihood of initiating and continuing biologic drugs for patients with RA.15 Households that incurred a high OOP expenditure showed a lower initiation of biologic drugs.15 Similarly, an analysis of 1360 drug formularies offered by private insurers under Medicare Part D indicated that adalimumab, a biologic anti-TNF agent indicated for the treatment of RA, was not included in 20% of the drug formularies.19

Another study, based on claims data from 55 health plans offered by 15 large employers, concluded that increased cost-sharing for specialty drugs, such as biologics, did not reduce their use; instead, the financial burden was transferred to the patients.16 A study of 14,929 Medicare Replacement Drug Demonstration enrollees with RA also found that Medicare Part D plans requiring coinsurance instead of copayments for biologic DMARDs were shifting the financial burden to patients.17 A recent nationwide study of the Medicare Part D plans' formularies in 2013 also found that most plans required a high level of cost-sharing from patients, which result in significant OOP expenditures.18

Increasingly, it has been recognized that a growing patient cost-sharing is associated with poorer health outcomes.20 A similar trend is being reported in the population of patients with RA based on a number of studies that have explored the relationship between high drug costs and the subsequent impact on patients' financial hardships and healthcare resource utilization.21–23 A study of 1055 Medicare beneficiaries diagnosed with RA reported a significant increase in mean annual outpatient prescription costs for RA drugs between the years 2000 ($396) and 2006 ($984).10 As a result of the increasing cost, patients with RA reported nonadherence to their medications and made difficult trade-offs in their lives.21–23

A study published in 2008 showed that adherence to RA therapies significantly decreased as the weekly OOP expenditures for therapies increased.21 Patients with weekly OOP expenditures >$50 had a greater likelihood of discontinuing therapy compared with patients with lower expenses.21 The presence of RA increases the likelihood of cost-related medication nonadherence and may force people to spend less on basic needs, such as food or heat.22 Female, disabled, uninsured, and low-income patients had a greater likelihood of experiencing cost-related medication nonadherence than other groups and also faced more problems paying for basic needs.22 A 2005–2007 study of 8545 patients with RA found that 43.6% of the patients experienced problems paying their OOP costs for medical and drug expenses after insurance payments.23 The financial burden was greater for nonelderly patients than for the elderly population. Sex, ethnicity, annual income, education, and marital status were associated with difficulty paying the OOP costs. In addition, people with an annual income of <$35,000 faced severe problems with OOP costs.23 The relationship of OOP expenditures with patient outcomes is not limited to the US healthcare system. Even in Canada, which has a publicly funded healthcare system, the policy of cost-sharing for medications has been shown to increase overall health utilization among elderly patients with RA.24,25

Previous studies have measured how OOP expenditures for patients with RA have affected adherence to treatments, but these studies have not explored the relationship between OOP expenditures related to prescription drugs that are used in the treatment of RA and the sociodemographic and personal characteristics of patients in the general US population.

The primary objective of this study was to explore the relationship between sociodemographic and personal characteristics and OOP expenditures for RA medications in a nationally representative sample of the US civilian noninstitutional population. A secondary objective was to measure the average OOP expenditures for different therapeutic classes of RA medications, including biologics, conventional DMARDs, NSAIDs, corticosteroids, and analgesics.

Methods

This retrospective analysis is based on the MEPS data set between 2009 and 2012. A 2-part regression model was used to answer the objectives of this study.

In this study we used 4 consecutive MEPS data sets from 2009 to 2012. The MEPS has 3 components, including household size, insurance/employer, and medical provider. For this study, the household component was analyzed, and files within the household component, such as the medical conditions files, the full-year consolidated data file, the prescribed medicines files, and the appendix to the MEPS event files, were extracted. The medical conditions files were used to identify individuals who were diagnosed with RA. The prescribed medicines files provided the medication name, the medication therapeutic class based on Cerner Multum Lexicon data source, total expenditures, and prescription-related OOP expenditures. The appendix to the MEPS event files was used to link records of the medical conditions file with the prescribed medicines file.

The full-year consolidated files provided information about the individual's sociodemographic and personal characteristics. We did not include a medical provider component, because the information collected by that component is used only for the purpose of editing and imputation of household component event files. We also did not include the insurance/employer component data, because the data are not available to the public because of confidentiality issues.

Sample Selection

Patients with RA were identified from the medical conditions file using International Classification of Diseases, Ninth Revision (ICD-9), Clinical Modification code 714.0.14 Their records were linked to the prescribed medicines file and were subsequently linked to the full-year consolidated data files to create a final data file that contained information about sociodemographic characteristics, prescribed medications, and cost. Because MEPS uses a panel survey design in which the same individual could be interviewed for 5 rounds over 2 calendar years, it is possible that the same individual might appear in the sample more than once. To identify repeated observations, data from 4 years were combined.

No significant differences in sociodemographic and personal characteristics were found across 4 years of data. A total of 372, 325, 375, and 424 individuals with RA were selected using the ICD-9 diagnosis code 714.0 in the sample from the 2009, 2010, 2011, and 2012 data sets, respectively, which represented 1098 unique individuals. After excluding 7 patients who were aged <18 years and 1 patient with missing data, the final data set included 1090 adults, representing a weighted frequency of approximately 9.71 million adult patients in the United States.

Statistical Analysis

The total OOP expenditure for an individual with RA was calculated by totaling the OOP expenditure for each prescription corresponding to that particular patient. Because the distribution of OOP expenditures was right-skewed, and some individuals had no OOP expenditures, a 2-part modeling method was conducted. In the first part of the model, we conducted a logistic regression analysis with OOP expenditures (measured as binomial, zero, or positive) as the dependent variable. In the second part, we created a generalized linear model (GLM) to model the relationship between the sociodemographic and personal characteristics associated with positive OOP expenditures in patients with RA.

In addition, we used log-link and gamma distribution in the GENMOD procedure (a statistical procedure), which fitted a GLM to the expenditure data by maximum likelihood estimation of the parameter vector. The predictors for both models included age; race, which was categorized as white, black (reference category), Hispanic, or others (ie, Asians, American Indian, Alaska native, native Hawaiian, and Pacific Islander); sex (female as reference); insurance status, categorized as public-only insurance, any private insurance, or uninsured (reference); marital status, categorized as married or unmarried (reference); family size; number of comorbid conditions; region, including Northeast, Midwest, South, or West (reference); area of living, categorized as non–metropolitan statistical area (non-MSA) or MSA (reference); and annual family income as a percentage of the annual federal poverty line, which was included as a categorical variable with 5 categories, including poor, near poor, low income, middle income, and high income (reference).

The annual family income as a percentage of the poverty line was below 100% for the group defined as poor, 100% to <125% for the near-poor group, 125% to <200% for the low-income group, 200% to <400% for the middle-income group, and ≥400% for the high-income group. Because OOP expenditures had right-skewed distribution, the means and confidence intervals (CIs) were estimated using nonparametric percentiles from 1000 cluster bootstrap estimates. Statistical analysis was conducted using SAS version 9.4 (SAS Institute, Inc; Cary, NC). All statistical analyses were performed at a 0.05 level of significance.

Results

The sample included 1090 unique individuals with RA, which represented approximately 9.71 million US adults (Table 1). The sample mean age was 60.7 ± 14.33 years (95% CI, 59.8–61.6 years). The majority of the patients were female (68.8%), white (65.3%), residing in an MSA (76.2%), and married (51.9%). A majority (56.6%) of the individuals were in the middle-income or high-income category, and 18.2% were classified as poor. Overall, 52.1% of the patients had any type of private insurance, 40.6% had public insurance, and 7.3% were uninsured. A total of 35.2% of the patients had ≥3 comorbid conditions. The mean annual family income was $8724.57 (± $5840) in the poor group; $17,003.75 (± $6652) in the near-poor group; $26,879.01 (± $12,260.08) in the low-income group; $47,840.67 (± $19,113) in the middle-income group; and $110,662.86 (± $54,973) in the high-income group. The mean family size was 2.51 (± 1.48) people (95% CI, 2.42–2.60). The sociodemographic and personal characteristics of the patients are listed in Table 1.

Table 1.

Patient Sociodemographic and Personal Characteristics

Variable	Patients, in millionsa (weighted percentage)
Sex
Male	3.03 (31.2)
Female	6.68 (68.8)
Race/ethnicity
White	6.35 (65.3)
Black	1.50 (15.5)
Hispanic	1.25 (12.9)
Others	0.61 (6.3)
Family income
Poor	1.77 (18.2)
Near poor	0.80 (8.3)
Low income	1.64 (16.9)
Middle income	2.73 (28.1)
High income	2.77 (28.5)
Insurance coverage
Any private insurance	5.06 (52.1)
Public-only insurance	3.95 (40.6)
Uninsured	0.70 (7.3)
Marital status
Married	5.04 (51.9)
Unmarriedb	4.67 (48.1)
Geographic region
Northeast	1.53 (15.8)
Midwest	1.92 (19.7)
South	4.41 (45.4)
West	1.85 (19.1)
Reside in metropolitan statistical area
Yes	7.40 (76.2)
No	2.31 (23.8)
Comorbid conditions
None	1.45 (14.9)
1 or 2	4.84 (49.9)
≥3	3.42 (35.2)

^aThe sample patient population was 1090, representing approximately 9.71 million US adults.

^bUnmarried includes divorced, separated, widowed, or never married.

The mean annual OOP expenditure for RA prescription drugs was $273.99 (95% CI, $197.07–$364.75). A total of 987 (90.5%) of the patients had some positive OOP expenditures, whereas 103 (9.5%) had no OOP expenditures for their prescription drugs. The mean annual OOP expenditures for the poor, near-poor, low-income, middle-income, and high-income groups were $148.94, $365.49, $143.90, $390.36, and $342.20, respectively.

A logistic regression analysis indicated that the odds of having no OOP expenditures was significantly greater in patients whose race was classified as “other” (odds ratio [OR], 2.68; 95% CI, 1.16–6.18); the odds of having no OOP expenditures was not significantly different for white (OR, 0.80; 95% CI, 0.41–1.57) and Hispanic (OR, 0.72; 95% CI, 0.36–1.44) patients compared with black patients. The odds of having no OOP expenditures were greater for poor patients (OR, 3.02; 95% CI, 1.09–8.35) than for those in the high-income group. Other factors, such as age, sex, marital status, insurance status, comorbidity, family size, and area and region of living, had no significant effect on whether a person had any OOP expenditures (Table 2).

Table 2.

Relationship Between Sociodemographic and Personal Characteristics, and Out-of-Pocket Expenditures

Variable	Odds ratio (95% CI)a	P valuea
Age	1.00 (0.99–1.02)	.669
White	0.80 (0.41–1.57)	.514
Hispanic	0.72 (0.36–1.44)	.349
Other	2.68 (1.16–6.18)	.021
Black (reference)
Poor	3.02 (1.09–8.35)	.033
Near poor	1.83 (0.64–5.20)	.260
Low income	2.64 (1.00–6.98)	.051
Middle income	2.12 (0.76–5.95)	.152
High income (reference)
Any private insurance	0.75 (0.25–2.24)	.610
Public-only insurance	1.35 (0.48–3.78)	.564
Uninsured (reference)
Male	1.37 (0.75–2.51)	.312
Female (reference)
Married	0.88 (0.46–1.67)	.688
Unmarriedb (reference)
Comorbidity	0.90 (0.75–1.08)	.261
Non-MSA	1.28 (0.65–2.51)	.471
MSA (reference)
Northeast	1.15 (0.52–2.56)	.729
Midwest	0.93 (0.40–2.19)	.870
South	1.32 (0.67–2.58)	.421
West (reference)
Family size	1.16 (0.98–1.36)	.076

^aOdds ratios with 95% CI and corresponding P values are obtained from the logistic regression, which modeled the odds of having no out-of-pocket expenditures as a dependent variable.

^aUnmarried includes divorced, separated, widowed, or never married.

CI indicates confidence interval; MSA, metropolitan statistical area.

The GLM showed that among patients with a positive OOP expenditure, age, sex, race, income category, insurance status, comorbidity status, marital status, and family size had a statistically significant effect on the OOP expenditures (Table 3). As age increased by 1 year, the mean annual expenditure decreased by at least 1% (0.98; 95% CI, 0.97–0.99). The mean annual OOP expenditure was significantly lower (0.73, 95% CI, 0.58–0.92) for men than for women. The mean annual OOP expenditure was the highest for white patients among all of the races, and was at least 1.66 times higher (2.15; 95% CI, 1.66–2.76) for white patients than for black patients. Compared with the high-income group, the OOP expenditures for the poor and low-income groups were at least 16% (0.60; 95% CI, 0.43–0.84) and 3% (0.69; 95% CI, 0.49–0.97) lower, respectively. Compared with uninsured individuals, the OOP expenditures were significantly lower by at least 55% (0.31; 95% CI, 0.21–0.45) for privately insured patients and by at least 73% (0.18; 95% CI, 0.12–0.27) for those with public insurance. The OOP expenditures also increased significantly for married people compared with unmarried people (1.80; 95% CI, 1.43–2.27). As the number of comorbid conditions increased, the OOP expenditures also increased (1.16; 95% CI, 1.07–1.25). The OOP expenditures decreased with an increase in family size by at least 11% (0.83; 95% CI, 0.77–0.89).

Table 3.

Impact of Sociodemographic and Personal Characteristics on Positive Out-of-Pocket Expenditures

Variable	Estimate (95% CI)a	P valuea
Age	0.98 (0.97–0.99)	<.0001
Male	0.73 (0.58–0.92)	.006
Female (reference)
White	2.15 (1.66–2.76)	<.001
Hispanic	1.00 (0.75–1.33)	.976
Other	0.99 (0.65–1.56)	.960
Black (reference)
Poor	0.60 (0.43–0.84)	.003
Near poor	0.94 (0.63–1.42)	.756
Low income	0.69 (0.49–0.97)	.034
Middle income	1.11 (0.82–1.49)	.492
High income (reference)
Any private insurance	0.31 (0.21–0.45)	<.0001
Public-only insurance	0.18 (0.12–0.27)	<.0001
Uninsured (reference)
Married	1.80 (1.43–2.27)	<.0001
Unmarriedb (reference)
Comorbidity	1.16 (1.07–1.25)	.0004
Family size	0.83 (0.77–0.89)	<.0001
Northeast	0.73 (0.52–1.02)	.063
Midwest	1.27 (0.91–1.76)	.158
South	0.99 (0.76–1.30)	.967
West (reference)
Non-MSA	1.07 (0.82–1.42)	.606
MSA (reference)

^aEstimates with 95% CI and corresponding P values are obtained from the generalized linear model with log link, which modeled the positive out-of-pocket expenditure as a dependent variable. Estimates and 95% CIs were obtained by the exponentiation of regression coefficients.

^bIncludes divorced, separated, widowed, or never married.

CI indicates confidence interval; MSA, metropolitan statistical area.

We conducted further analyses to analyze the OOP expenditures associated with the different classes of RA medications (Table 4). The average annual OOP expenditure for biologics was $2556.73 (95% CI, $1145.68–$4359.96), which was almost 12.5% of the average total annual expenditure of $20,534.37 (95% CI, $16,497.30–$24,838.31) for biologic drugs. The OOP expenditure was the highest for biologics followed by conventional DMARDs. The average annual OOP expenditure for conventional DMARDs was $89.37 (95% CI, $74.26–$104.79), which was 23.5% of the average total expenditure of $380.12 (95% CI, $284.69–$493.25) for conventional DMARDs. The average annual OOP expenditure for NSAIDs was $52.36 (95% CI, $41.56–$64.83), which was 15.8% of the average total expense of $330.77 (95% CI, $257.37–$423.83) for NSAIDs, whereas the mean annual OOP expenditure for narcotic analgesics was $72.51 (95% CI, $56.82–$92.43), which amounted to 17.1% of the average total expenditure of $423.04 (95% CI, $271.12–$617.89) on narcotic analgesics. The mean annual OOP expenditure was the lowest for corticosteroids at $27.97 (95% CI, $21.50–$35.65); however, this was 60% of the average total expense of $46.58 (95% CI, $37.83–$56.96) for corticosteroids.

Table 4.

Average Out-of-Pocket Costs for Drug Classes Used in the Treatment of Rheumatoid Arthritis

Type of medication	Mean OOP expenditure, $ (95% CI)a	Mean total expenditure, $ (95% CI)a
Biologic	2556.73 (1145.68–4359.96)	20,534.37 (16,497.30–24,838.31)
Conventional DMARD	89.37 (74.26–104.79)	380.12 (284.69–493.25)
NSAID	52.36 (41.56–64.83)	330.77 (257.37–423.83)
Narcotic analgesic	72.51 (56.82–92.43)	423.04 (271.12–617.89)
Corticosteroid	27.97 (21.50–35.65)	46.58 (37.83–56.96)

^aAll 95% CIs for the mean were estimated based on nonparametric percentiles from 1000 cluster bootstrap estimates.

CI indicates confidence interval; DMARD, disease-modifying antirheumatic drug; NSAID, nonsteroidal anti-inflammatory drug; OOP, out-of-pocket.

Discussion

This study explored the relationship between patients' sociodemographic and personal characteristics and OOP expenditures related to RA prescription drugs in a nationally representative noninstitutionalized US population. The results showed that race and income were significant predictors of OOP expenditure. Among patients with a positive OOP expenditure, age, income, insurance status, sex, comorbidity level, marital status, and family size were significantly associated with having an OOP expenditure. As age increased, the OOP expenditure decreased significantly. This result was consistent with the findings of the study by Wolfe and Michaud that nonelderly patients with RA had more problems with OOP spending for medical bills than patients aged ≥65 years.23 For the elderly population (ie, aged ≥65 years), the availability of Medicare coverage may reduce OOP spending on medications.23

Overall OOP expenditures were significantly higher for the high-income group than for the poor and low-income groups. One possible reason for this is that individuals with a high income were prescribed more costly medications (eg, biologics), which increased their OOP spending. Of note, we did not find a significant difference in OOP expenditures related to biologics among the high-income, poor, and low-income groups. This may be because biologics were offered in those cases when insurance could absorb the high cost. The majority of patients who had at least 1 prescription for a biologic had some form of public or private insurance, which might have provided coverage for the biologic. By contrast, the average OOP expenditure for conventional DMARDs was significantly higher among the high-income group than in the poor group. Because conventional DMARDs are widely used in the treatment of patients with RA, it is possible that higher OOP expenditures for these treatments may cause economic hardships for certain groups of patients with RA.

The OOP expenditures for uninsured individuals were significantly greater than for patients with private or public insurance. This finding was similar to the findings of the study by Wolfe and Michaud, who reported that the absence of health insurance causes a severe financial burden on patients with RA.23 Many patients with RA have functional limitations and disabilities, which can limit their earning potential. Higher OOP expenditures, coupled with limited earning or unemployment and a lack of insurance, impose a tremendous financial burden on these patients.

It is well-established that RA is more common in women, and that sex is significantly associated with financial problems related to paying for drugs and medical bills among patients with RA.23 These observations were confirmed in our study as well, and OOP expenditures were found to be significantly greater for female patients than for male patients. As the number of comorbid conditions increased, the OOP expenditure for prescription drugs increased significantly. More than 85% of patients with RA had ≥1 comorbid conditions, whereas a little more than 33% had ≥3 comorbid conditions. Most of the patients had high blood pressure, high cholesterol, and other cardiac ailments. A previous study of older adults with RA reported that cost-related medication nonadherence was more pronounced in patients with comorbidities than in those without comorbidities.22 The presence of multiple comorbid chronic conditions posed a serious financial challenge to patients with RA in paying for their prescription medications.

Pharmacy-related expenditures have become the primary cost driver of RA treatments in recent years after the introduction of expensive biologics.14 Higher OOP costs have significantly affected adherence to biologic therapies among patients with RA.21 In this context, our study focused on OOP expenditures for drugs among patients with RA and explored how these individuals in different socioeconomic groups were affected by OOP pharmacy-related expenditures. Higher OOP expenditures among the uninsured, female patients, patients in certain income categories, and individuals with various comorbidities could adversely affect adherence to RA therapy, which in turn would cause worse health outcomes and higher healthcare spending.

Limitations

This study has several limitations. First, the MEPS data are based on self-reporting; thus, the correct identification of the disease and proper recall of information greatly influence the validity of these self-reports.

Second, the OOP expenditure information for prescription drugs was collected from 2 sources. For respondents who reported that their pharmacy providers automatically submitted claims for the prescription and for uninsured respondents, the OOP expenditure information was collected from the pharmacy. For respondents who reported that they sent in their prescription drug claims themselves, the payment information was collected directly from them. In the case of self-reporting, it is possible that the expenditure data may not be accurate.

Third, the OOP expenditure model was estimated using a GLM with log-link and gamma distribution. No attempt was made to use other link functions or to compare the results across different link functions. Misspecification of the mean-variance relationship as gamma distribution may cause a potential efficiency loss. Because the information about prescribed medications was obtained through self-reporting by respondents in the MEPS, it could lead to the underreporting of medications. The underreporting of the number of prescribed medications by respondents in MEPS may influence the findings in this present study. A future study may specifically evaluate the accuracy of the number of RA medications as reported by respondents in MEPS by comparing it with other national-level data sources, such as Medicare Part D claims.

Another limitation of the present study is that it included information on OOP expenditures for the time period before the implementation of the Affordable Care Act (ACA). Although the ACA has improved the affordability of healthcare for the US population, the existence of a high number of uninsured people, a high coinsurance rate for biologics, and the availability of high-deductible health plans in the marketplace render the findings of the present study as relevant even after the implementation of the ACA.18,26 However, it will be interesting to conduct a similar study using prescription drug expenditure data from plans available in the commercial marketplace and compare the findings with those of the present study.

Conclusion

This study supports that patients with RA who have certain sociodemographic characteristics are affected by higher OOP expenditures for prescription drugs. With the rising use of biologic drugs in the treatment of RA, the OOP expenditures for RA drugs may increase in the future. The use of these expensive biologics needs to be monitored carefully to reduce prescription drug–related cost-sharing among patients. Clinicians have an important role in identifying patients who are experiencing financial hardship and in assisting them with information on sources of low-cost drugs, such as medication assistance programs. Managed care plans can also provide generous drug plan coverage, because the evidence supports the association of cost offsets of a generous drug plan with a reduction in future healthcare resource utilization. Given the presence of high-deductible health plans in the commercial market and the high coinsurance rate for biologics, it is imperative to monitor adherence to biologics in patients with RA: high OOP expenditures could negatively affect patient adherence. However, it is evident that the strategy of increasing OOP expenditures for prescription drugs to control spending needs some critical thinking.