SEARCH

SEARCH BY CITATION

Abstract

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. AUTHOR CONTRIBUTIONS
  8. REFERENCES

Objective

To evaluate the impact of Medicare Part D on medication utilization, drug expenditures, and medical expenditures in patients with arthritis.

Methods

This was a retrospective study using a national sample of 2,484 Medicare-eligible beneficiaries with arthritis from the pooled Medical Expenditure Panel Survey 2005–2008 data. Quantile regression was used to estimate the following outcomes: 1) number of prescription fills, 2) total drug expenditures, 3) out-of-pocket (OOP) drug expenditures, 4) Medicare-paid drug expenditures, 5) total medical expenditures (including all payments for inpatient/outpatient care, prescription drugs, and other medical services), 6) OOP medical expenditures, and 7) Medicare-paid medical expenditures. For each outcome variable, the 50th, 75th, and 90th percentiles were estimated, adjusting for demographics and comorbidity. All expenditures were inflation adjusted to 2008 dollars.

Results

From 2005 to 2008, the adjusted median annual number of prescription fills increased by 4.2 (14.6% change), from 28.4 to 32.6. The adjusted median OOP drug expenditures and OOP medical expenditures decreased by $151 (25.2% change) and $197 (17.3% change), respectively. The adjusted median Medicare-paid drug and medical expenditures increased by $366 and $896 (39.5% change), respectively. The adjusted total prescription expenditures increased by $845 (25.3% change) at the 75th percentile and by $1,194 (22.0% change) at the 90th percentile. The adjusted total medical expenditures did not change significantly.

Conclusion

Medicare Part D resulted in increased medication utilization and significant reductions in OOP drug and OOP medical expenditures among beneficiaries with arthritis 3 years after its implementation. Part D was not associated with significant differences in total medical spending.


INTRODUCTION

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. AUTHOR CONTRIBUTIONS
  8. REFERENCES

The Medicare Part D prescription drug benefit, introduced in January 2006 as part of the Medicare Prescription Drug, Improvement, and Modernization Act of 2003, was designed to alleviate the burden of out-of-pocket (OOP) expenditures on medications for the elderly and to increase access to prescription medications for beneficiaries who previously had limited or no drug coverage (1). Prior to Part D, high drug costs and lack of prescription coverage were the most common reasons why Medicare beneficiaries discontinued or failed to initiate drug therapy, especially among those with multiple chronic conditions (2). As of 2010, more than 70% of the 47.5 million Medicare beneficiaries had enrolled in Part D. The total cost of Part D in 2010 was $62 billion ($1,789 per capita), or 12% of all Medicare expenditures. Medicare expenditures on Part D were projected to grow continuously by 9.7% through 2020, outpacing the estimated growth rate of the US economy by 4.5% over the same period (3).

A number of studies have evaluated the impact of Medicare Part D since its implementation and found the program to be associated with increased drug utilization and reduced OOP prescription expenditures (4, 5). However, Part D has been found to result in differential effects in subsets of Medicare beneficiaries, such as those with chronic conditions, those without previous prescription coverage, and those who were enrolled in both Medicare and Medicaid (i.e., “dual-eligibles”) (6–12). For example, Part D seems to have brought about little to no improvement in access to medications for patients with mental disorders (8, 9). Total and OOP drug expenditures have not changed significantly for dual-eligibles (7), but reductions in OOP drug expenditures and increases in utilization of statins and proton-pump inhibitors have been observed among those without previous drug coverage (6, 10). In patients with diabetes mellitus with no or minimal previous drug benefit, the implementation of Part D was associated with improved adherence to medications, increased prescription drug expenditures, and decreased spending in other medical services (11, 12).

Despite the increasing number of empirical evaluations on the effects of Part D, little is known regarding the policy's impact on patients with arthritis and other rheumatic conditions, who often face high OOP drug expenditures due to the chronic nature of the disease and the need for expensive biologic agents to prevent disease progression. From 1997 to 2005, the population with arthritis in the US increased from 37 million to 45 million, and the mean prescription expenditures nearly doubled during the same time period, from $970 to $1,811 per person (13). The burden of OOP copayments is significant in these patients. More than 4 of 10 patients with rheumatoid arthritis reported having difficulty paying medical bills (14). One study predicted that after Part D, the median total and OOP drug expenditures for patients with arthritis would decrease by 17% and 32%, respectively; however, these estimates were extrapolated based on simulations using data through 2004 (15). Given the lack of evidence on the effects of Part D on patients with arthritis in current research, a comprehensive evaluation of drug utilization and expenditures using data published after 2006 is needed.

The objective of this study was to evaluate the impact of Medicare Part D on prescription utilization, drug expenditures, and medical expenditures for patients with arthritis using a nationally representative sample of community-dwelling elderly adults in the US. We assessed overall medical expenditures in addition to drug utilization and expenditures because empirical evidence has shown that reduced copayments for prescription drugs may increase drug spending but decrease overall medical expenditures due to improved adherence (16). For prescription and medical spending, we examined the changes in total expenditures, OOP expenditures, and expenditures paid by Medicare.

Significance & Innovations

  • Using nationally representative data through 2008, we found evidence of increased medication utilization in patients with arthritis after Medicare Part D.

  • Out-of-pocket drug and out-of-pocket medical expenditures decreased significantly, suggesting an alleviated burden of copayments for patients.

  • Total medical spending did not change significantly after Part D was implemented.

MATERIALS AND METHODS

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. AUTHOR CONTRIBUTIONS
  8. REFERENCES

Data source and study population.

Using the Medical Expenditure Panel Survey (MEPS), a longitudinal study was conducted assessing utilization by individuals ages ≥65 years with arthritis. MEPS is a survey distributed to a nationally representative sample of the US civilian noninstitutionalized population. Administered by the Agency for Healthcare Research and Quality, the survey collects comprehensive data on health services utilization by the respondents. MEPS includes 3 components: the Household Component (HC), the Medical Provider Component, and the Insurance Component. The MEPS HC uses an overlapping panel design that collects data using 5 rounds of in-person interviews over 2.5 years (17). Survey data from panel 9 to panel 12 (most recent data available at the time of the study) of the MEPS HC were used, which contain detailed personal-level information on demographic characteristics, medical conditions, health status, use of health services, charges and payments, insurance coverage, and income from 2004 to 2008.

We included patients who had at least 1 diagnosis for arthritis or another rheumatic condition and completed all 5 rounds of the survey. Diagnoses were ascertained in the first year of each panel using MEPS truncated 3-digit International Classification of Diseases, Ninth Revision (ICD-9) diagnosis codes, including 710–716, 719–721, and 725–728 (15). To assess the comorbidity of patients, other chronic conditions were identified using ICD-9 codes to compute the Charlson Comorbidity Index (CCI) (18). Demographic characteristics were determined using data from the first year of each panel, including age, race/ethnicity, sex, income, education, marital status, and metropolitan statistical area (MSA).

Outcomes of interest.

Medication utilization, prescription expenditures, and medical expenditures were determined using data in the second year of each panel. Prescription utilization was defined as the number of prescription fills per patient during a calendar year. Total prescription expenditures were the sum of all payments for prescription drugs in 1 year, inclusive of payments from OOP, Medicare, Medicaid, private insurance, and other third party payors. OOP drug expenditures were the sum of payments made by patients, whereas Medicare-paid drug expenditures were the sum of payments made by Medicare. Total medical expenditures included payments for ambulatory care visits, physician office visits, hospital admissions, emergency department visits, prescription medications, and other medical services; the total included payments from OOP, Medicare, Medicaid, private insurance, and other third party payors. OOP medical expenditures were defined as the sum of payments made by patients, and Medicare-paid medical expenditures were defined as the sum of payments paid by Medicare. All expenditures were adjusted for inflation to 2008 US dollars using the medical care Consumer Price Index (19).

Statistical analysis.

First, we described the demographic characteristics of patients in 2005 and 2008. We then used quantile regression to assess the changes in prescription fills, prescription expenditures, and medical expenditures from 2005 to 2008 at the 50th, 75th, and 90th percentiles. These percentiles were chosen because we were interested in examining the trends in utilization and expenditures for average users (i.e., median), above average users (i.e., 75th percentile), and high users (i.e., 90th percentile). Analogous to ordinary least squares regression, quantile regression estimates the median or other quantiles of an outcome variable associated with a set of predictors and covariates without the assumptions of normality and homoscedasticity of the underlying distribution (20). Quantile regression is robust to outliers because it allows for studying the full distribution of the outcome variable and is suitable for modeling outcomes that are highly skewed or not normally distributed (21). In the quantile regression models, patient-specific weights were applied to derive weighted expenditures and number of prescription fills, while controlling for age, age squared, race/ethnicity, sex, income, education, marital status, MSA, and the CCI. To account for the complex survey design of MEPS, bootstrapping was used to provide 95% confidence intervals for the differences in regression coefficients (22). The numbers of prescription fills and expenditures were estimated in reference to white, single women age 75 years with up to 12 years of education, with an income of $27,073 in 2008 dollars, residing in an MSA, and with a CCI score of 1. All statistical analyses were conducted using SAS, version 9.2. The a priori level of statistical significance was P values less than 0.05.

RESULTS

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. AUTHOR CONTRIBUTIONS
  8. REFERENCES

Sample characteristics.

The sample included 2,484 patients with arthritis ages ≥65 years, consisting of 594 patients in 2005, 599 patients in 2006, 621 patients in 2007, and 670 patients in 2008. After multiplying with person-specific weights, these individuals represented between 5.8 million and 7.0 million patients per year nationally. The mean ± SD age was 75.4 ± 6.3 years. The demographic characteristics remained stable between 2005 and 2008, with the majority of the patients being white women with high school education or less. Table 1 shows the demographic characteristics of patients with arthritis in 2005 and 2008.

Table 1. Characteristics of Medicare-eligible patients with arthritis in 2005 and 2008
 2005 (n = 594)2008 (n = 670)
  • *

    Based on ref.18.

Weighted N5,767,6706,972,286
Age, mean ± SD years75.5 ± 6.275.7 ± 6.4
Female sex, %66.465.3
White, %85.589.3
Married, %53.449.6
High school education or less, %77.172.4
Residing in a metropolitan statistical area, %79.480.5
Income, mean ± SD 2008 dollars26,102 ± 23,81927,073 ± 25,052
Charlson Comorbidity Index score, mean ± SD*1.02 ± 1.331.44 ± 1.50
Charlson Comorbidity Index score, %  
 146.734.2
 226.627.2
 314.817.8
 47.211.1
 ≥51.75.0

Changes in medication utilization.

Figure 1 displays the trends in prescription fills from 2005 to 2008. After adjusting for the covariates, the median number of prescription fills increased from 28.4 to 32.6 (4.2 [14.6%]; P = 0.014) and the 75th percentile increased from 54.1 to 60.8 (6.7 [12.4%]; P = 0.033) between 2005 and 2008. An increase was also seen at the 90th percentile, from 74.7 to 82.2, but the difference was not statistically significant (7.5 [10.1%]; P = 0.098).

thumbnail image

Figure 1. Adjusted median, 75th percentile, and 90th percentile of the number of prescription fills from 2005 to 2008, estimated in reference to white, single women age 75 years with up to 12 years of education, with an income of $27,073 in 2008 dollars, residing in a metropolitan statistical area, and with a Charlson Comorbidity Index score of 1.

Download figure to PowerPoint

Changes in prescription expenditures.

Between 2005 and 2008, total prescription expenditures (in 2008 dollars) increased from $3,337 to $4,182 at the 75th percentile ($845 [25.3%]; P < 0.001]) and from $5,430 to $6,624 ($1,194 [22.0%]; P = 0.006) at the 90th percentile, but the increase at the median was not statistically significant ($214 [10.3%]; P = 0.079). Quantile regressions showed reductions in OOP drug expenditures across the 50th (−$151 [−25.2%]; P < 0.001), 75th (−$413 [−27.8%]; P < 0.001), and 90th (−$829 [−31.1%]; P < 0.001) percentiles, as well as increases in drug expenditures paid by Medicare across the 50th ($366; P < 0.001), 75th ($1,549 [622.0%]; P < 0.001), and 90th ($2,583 [281.3%]; P < 0.001) percentiles, after adjusting for the covariates. Table 2 shows the adjusted total drug expenditures, OOP drug expenditures, and Medicare-paid drug expenditures between 2005 and 2008.

Table 2. Total drug expenditures, out-of-pocket drug expenditures, and Medicare-paid drug expenditures in Medicare-eligible patients with arthritis between 2005 and 2008*
 20052008Difference95% CI% changet valueP
  • *

    Expenditures were estimated in reference to white, single women age 75 years with up to 12 years of education, with an income of $27,073 in 2008 dollars, residing in a metropolitan statistical area, and with a Charlson Comorbidity Index score of 1. 95% CI = 95% confidence interval; N/A = not applicable.

Total drug expenditures       
 Median$2,079$2,293$214−$25, $45310.31.760.079
 75th percentile$3,337$4,182$845$437, $1,25225.34.06< 0.001
 90th percentile$5,430$6,624$1,194$346, $2,04122.02.760.006
Out-of-pocket drug expenditures       
 Median$601$450−$151−$226, −$77−25.2−3.97< 0.001
 75th percentile$1,485$1,071−$414−$593, $233−27.8−4.50< 0.001
 90th percentile$2,662$1,833−$829−$1,191, −$467−31.1−4.49< 0.001
Drug expenditures paid by Medicare       
 Median$0$366$366$318, $414N/A14.91< 0.001
 75th percentile$249$1,798$1,549$1,394, $1,704622.019.61< 0.001
 90th percentile$918$3,501$2,583$2,185, $2,980281.312.73< 0.001

Changes in medical expenditures.

No statistically significant differences were found between 2005 and 2008 in overall total medical expenditures (50th percentile: −$181 [−2.8%]; P = 0.714, 75th percentile: −$1,575 [−10.4%]; P = 0.300, and 90th percentile: −$3,212 [−10.3%]; P = 0.281). However, reductions in OOP medical expenditures were observed across the 50th (−$197 [−17.3%]; P = 0.029), 75th (−$568 [−23.2%]; P = 0.001), and 90th (−$736 [−18.5%]; P = 0.023) percentiles, adjusting for the covariates. Medical expenditures paid by Medicare increased at the median ($896 [39.5%]; P < 0.001), but not at the 75th ($13 [0.2%]; P = 0.989) and 90th percentiles (−$2,929 [−13.86%]; P = 0.220). Table 3 shows the adjusted total medical expenditures, OOP medical expenditures, and Medicare-paid medical expenditures between 2005 and 2008.

Table 3. Total medical expenditures, out-of-pocket medical expenditures, and Medicare-paid medical expenditures in Medicare-eligible patients with arthritis between 2005 and 2008*
 20052008Difference95% CI% changet valueP
  • *

    Expenditures were estimated in reference to white, single women age 75 years with up to 12 years of education, with an income of $27,073 in 2008 dollars, residing in a metropolitan statistical area, and with a Charlson Comorbidity Index score of 1. Expenditures included payments for ambulatory care visits, physician office visits, hospital admissions, emergency department visits, prescription medications, and other medical services. 95% CI = 95% confidence interval.

Total medial expenditures       
 Median$6,529$6,348−$181−$1,148, $787−2.8−0.370.714
 75th percentile$15,167$13,591−$1,576−$4,555, $1,403−10.4−1.040.300
 90th percentile$31,289$28,077−$3,212−$9,056, $2,632−10.3−1.080.281
Out-of-pocket medical expenditures       
 Median$1,142$944−$198−$375, −$20−17.3−2.180.029
 75th percentile$2,447$1,879−$568−$891, −$245−23.2−3.450.001
 90th percentile$3,981$3,245−$736−$1,368, −$104−18.5−2.280.023
Medical expenditures paid by Medicare       
 Median$2,267$3,163$896$398, $1,39539.53.530.000
 75th percentile$7,684$7,697$13−$1,845, $1,8710.20.010.989
 90th percentile$21,131$18,203−$2,928−$7,612, $1,754−13.9−1.230.220

DISCUSSION

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. AUTHOR CONTRIBUTIONS
  8. REFERENCES

Using national estimates, the present study set out to determine the impact of Medicare Part D on medication utilization, drug expenditures, and medical spending among noninstitutionalized Medicare beneficiaries with arthritis ages ≥65 years. The results of this investigation showed substantial increases in payments for prescription drugs by Medicare and reductions in OOP drug and OOP medical expenditures. Between 2005 and 2008, the number of prescription fills, total drug expenditures, and Medicare-paid medical expenditures increased at various quantiles, but total medical expenditures (including payments from OOP, Medicare, and all other payors) did not change significantly.

Previous studies have found an increase in medication utilization after Part D using proxies such as days' supply and number of patients filling prescriptions (23–25). In the current study, we used the number of prescription fills as a surrogate to assess the change in medication utilization over time. Our findings that the number of prescription fills has increased at the median and at the 75th percentile support the general trend of increased medication utilization in Medicare beneficiaries after Part D. However, it should be noted that without the ascertainment of days' supply (unavailable in MEPS), it is unclear whether the growth in the number of prescription fills can be translated directly into increased volume of drugs dispensed. In another study using MEPS data, the mean number of prescription fills did not change significantly (22.0 in 2005 and 23.9 in 2006) in the general Medicare population (26). Several possibilities may contribute to this discrepancy. First, it is likely that the effect of Part D on medication utilization did not manifest until the transition year had passed. Second, Part D may have differential effects on specific populations. Compared with the general Medicare population, patients with arthritis seem to have greater medication utilization. In the current study, the unadjusted mean ± SD yearly numbers of prescription fills were 36.1 ± 31.4 and 39.1 ± 33.0 in 2005 and 2008, respectively, which are more than 50% higher than the general Medicare population. Third, the use of quantile regression may be more robust in detecting differences in variables that are not normally distributed. We conducted a separate analysis to examine the difference in mean numbers of prescription fills between 2005 and 2008 using analysis of variance, but did not find a statistically significant difference (F = 1.61, P = 0.185).

Earlier studies have found an increase or no change in total prescription drug expenditures due to Part D (10, 26). In our study, we did not find a statistically significant growth in median total prescription drug expenditures (10.3%); however, the increases were significant at the 75th (25%) and 90th percentiles (22%), suggesting that drug spending grew more markedly in patients with higher drug expenditures after Part D. The magnitude of reduction in OOP prescription expenditures (25%) was similar to the mean reductions in 2 other studies using nationally representative data (21% to 32%) (6, 26), but higher than the estimates from studies that used claims data (13% to 18%) (24, 25). The variation may be a result of the different study populations and methods. It is likely that studies using prescription fill data from specific pharmacy chains had captured patients that were not representative of the general Medicare population, thus underestimating the benefit of Part D. Nevertheless, even in studies using nationally representative data, the effect of Part D may have been underestimated because the analyses were limited to the transition year (2006) and many beneficiaries did not enroll in Part D until several months after the program's implementation. One recent study examined the impact of Part D using national estimates and found that prescription drug utilization did not increase significantly until 2007 for patients who had poorer health status (27). Across the 50th, 75th, and 90th percentiles, the reductions in OOP drug expenditures were smaller than the sizeable increases in prescription expenditures paid by Medicare. Similar findings were seen in earlier studies (26, 28), implying that Part D has shifted the costs for prescription drugs from patients and other third party payors to Medicare. Total drug expenditures, as a proportion of overall medical spending, seem to have increased from 2005 to 2008: from 31.8% to 36.1% at the median, from 22.0% to 30.8% at the 75th percentile, and from 17.4% to 23.6% at the 90th percentile. However, these results should be interpreted with caution because we did not test whether the growths were statistically significant.

In the current study, reductions in total medical expenditures were observed between 2005 and 2008; however, none of the differences were statistically significant. Because the goal of Part D is to improve the health status of Medicare beneficiaries through a decreased burden of OOP payments, further investigations are necessary to examine whether the policy has resulted in decreased total medical spending. The reductions in OOP medical expenditures, largely due to the decreases in OOP drug expenditures, confirm the projections from a modeling study on patients with arthritis (15). Medical expenditures paid by Medicare increased at the median but not at the 75th or 90th percentile. The lack of statistically significant differences at the 75th and 90th percentiles suggests that prior to Part D, Medicare may have already been paying for a large percentage of medical expenditures for patients with high medical spending (i.e., the upper quartile and the top decile).

Several limitations to the current study need to be acknowledged. First, although the study was based on a nationally representative sample, the sample size was relatively small (n = 2,484). Due to the limited sample size, we were not able to examine the impact of “donut hole” on these patients. Under Part D, patients (excluding dual-eligibles) are required to pay 100% of the OOP cost once they reach the coverage gap (i.e., donut hole), before the expense reaches the catastrophic coverage threshold. This issue is particularly important for patients with arthritis because their condition may necessitate the use of expensive biologic agents, and the lack of financial assistance during the coverage gap could lead to discontinuation of drug therapy. In our study, 41% of the beneficiaries reached the initial coverage limit ($2,510) in 2008, higher than the percentage (26%) seen in patients with diabetes mellitus (29). The Part D coverage gap has been found to be associated with worse adherence in patients with chronic disease such as diabetes mellitus (30). Therefore, further research is needed to examine the impact of donut hole for Medicare beneficiaries with arthritis and other rheumatic conditions.

The small sample size also precluded us from evaluating the effects of Part D in subsets of Medicare beneficiaries, such as those without prior drug coverage or those enrolled in both Medicare and Medicaid. While Part D has been found to be associated with increased medication use, decreased nondrug medical spending, and improved adherence for Medicare beneficiaries with limited prior drug coverage (10, 12, 31), the policy has had little impact on health care utilization and expenditures in dual-eligibles (7). Since we did not conduct subgroup analyses based on patients' insurance coverage, our findings may have reflected an averaged effect of Part D among patients with arthritis.

In addition, we only included patients who were age ≥65 years. Since individuals ages <65 years could be enrolled in Medicare due to disabilities or end-stage renal disease, our results may not be generalizable to the overall Medicare population. It is possible that individuals ages <65 years have different patterns of health care utilization and expenditures, and Part D may have had different effects on these patients. Also, our study was limited to the evaluation of patients with arthritis and other rheumatic conditions. Therefore, the study findings may not be transferrable to patients with other disease states.

Finally, given the inherent limitations of the MEPS data, we could not evaluate changes in adherence or long-term health outcomes attributable to Part D. The observed increases in prescription fills and decreased OOP expenditures in the current study imply that access to prescription medications has improved for patients with arthritis after Part D; however, these improvements may not be positively associated with increased adherence to drug therapy. In addition to economic benefits such as decreased OOP payments, the assessment of clinical benefits using long-term health outcomes is necessary to examine whether and how Part D has enhanced the health for the elderly.

In summary, Medicare Part D was associated with significant reductions in OOP prescription expenditures and OOP medical expenditures, as well as increased medication utilization, among a nationally representative sample of patients with arthritis and other rheumatic conditions 3 years after its implementation. While total prescription expenditures increased for those who had higher drug spending, total medical expenditures had not increased after Part D. Further research might explore the differential effects of Part D on subsets of patients with arthritis, including both clinical and economic benefits. As the structure of Part D continues to evolve, it will be necessary to assess its impact on access to medications and long-term health outcomes.

AUTHOR CONTRIBUTIONS

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. AUTHOR CONTRIBUTIONS
  8. REFERENCES

All authors were involved in drafting the article or revising it critically for important intellectual content, and all authors approved the final version to be published. Mr. Cheng had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study conception and design. Cheng, Rascati.

Acquisition of data. Cheng.

Analysis and interpretation of data. Cheng, Rascati.

REFERENCES

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. AUTHOR CONTRIBUTIONS
  8. REFERENCES