|
Read related papers by:
Bruce Stuart and colleagues,
Cindy Thomas and colleagues,
and Robert F. Atlas
D A T A W A T C H
D R U G C O V E R A G E
19 April 2005
Prescription Drug Coverage And Seniors:
Findings From A 2003 National Survey
Where do things stand on the eve of implementing
the new Medicare Part D benefit?
By Dana Gelb Safran, Patricia Neuman, Cathy Schoen,
Michelle S. Kitchman, Ira B. Wilson, Barbara Cooper,
Angela Li, Hong Chang, and William H. Rogers
ABSTRACT:
Beginning in 2006 the Medicare Prescription Drug, Improvement, and Modernization Act (MMA) will offer pharmacy benefits to forty-two million Medicare beneficiaries nationwide. In a 2003 national survey of Medicare beneficiaries age sixty-five and older, more than one-quarter reported no prescription coverage, and nearly half of low-income seniors in some states lacked coverage. Wide coverage differences among states highlight implementation challenges and the need for tailored enrollment strategies. Evidence of Medicaid’s highly effective coverage delineates the importance of assuring this group’s continued protection under Part D plans. Reports of complex drug regimens, multiple prescribing physicians and pharmacies, nonadherence, and reimportation demonstrate the challenges of integrating seniors’ prescription care. We discuss MMA’s potential to improve quality and the need to monitor performance.
The Medicare Prescription Drug, Improvement, and Modernization Act of 2003 (MMA) represents the most substantial expansion of Medicare since the program was enacted. Beginning in 2006, beneficiaries will have access to a prescription drug benefit (Part D) subsidized by Medicare and available through private plans. Additional subsidies are available for low-income beneficiaries and for those with extremely high drug costs. MMA responds to the well-documented need for improved drug coverage, creating an important opportunity to mitigate the burden of rising drug costs and improve health care quality for an aging population. However, enormous challenges lie ahead, given the heterogeneity of the Medicare population; the diversity of current coverage; and the tasks of establishing a nationwide system of private prescription plans, determining millions of low-income seniors’ eligibility for subsidies, and providing forty-two million beneficiaries with information to make good coverage decisions.
This paper presents findings from a 2003 national survey of noninstitutionalized Medicare beneficiaries age sixty-five and older, with a detailed view of seniors’ prescription drug coverage, use, and out-of-pocket spending. Our intention is to inform the MMA implementation process, including the development of effective outreach, education, and enrollment strategies. The data represent the most current information available on seniors’ prescription coverage and use and are the only source of such information with state-level analysis.1 We oversampled low-income seniors with and without Medicaid to profile their experiences in this period leading up to MMA implementation. We also oversampled seniors residing in each of twelve geographically diverse states, encompassing 55 percent of elderly Americans, to assess state-level variations in coverage and prescription care. Among the twelve states are six with state-sponsored pharmacy assistance programs, state discount card programs, or both (New York, Pennsylvania, Illinois, Michigan, Florida, and California) and six states without such programs (Ohio, Tennessee, Louisiana, Texas, Colorado, and Washington).
We document variations in prescription use and out-of-pocket spending by drug coverage, poverty, and disease burden, and we show how sources of coverage vary among states. We examine the extent to which seniors report failing to take prescribed medications because of cost and for other reasons. Finally, we assess the extent to which seniors in different states and with different sources of coverage report purchasing prescriptions from Canada and Mexico. We conclude by discussing the implications for implementing Part D coverage, for helping seniors with rising prescription costs, and for improving the quality of seniors’ care related to medication use.
Study Data And Methods
Study design and sample. For sampling, the Centers for Medicare and Medicaid Services (CMS) provided a 1 percent probability sample of noninstitutionalized Medicare beneficiaries age sixty-five or older from each state and the District of Columbia. CMS data included a current buy-in code indicating whether the state was buying full or partial Medicaid coverage for each beneficiary. We oversampled Medicaid enrollees and seniors residing in low-income neighborhoods, which were identified by linking the CMS file to 2000 U.S. census data through geocoding and using census block group–level data to designate beneficiaries residing in high-poverty neighborhoods (13 percent of residents age sixty-five and older with incomes below 100 percent of the federal poverty level).2 Three strata were defined for sampling: (1) beneficiaries with full Medicaid benefits, (2) beneficiaries without Medicaid living in high-poverty neighborhoods, and (3) beneficiaries without Medicaid living outside of high-poverty neighborhoods. A random starting sample of 36,901 Medicare beneficiaries was drawn, including approximately 2,200 from each of the twelve states targeted for oversampling, and the remaining sample was drawn proportionally from the other states and the District of Columbia.3 Each state’s sample was drawn randomly, with fixed allocations across the three strata described above (25 percent, 50 percent, and 25 percent, respectively).
Questionnaire. The study questionnaire focused on current drug coverage, use, and spending; and it included questions on health status, income, and other sociodemographic characteristics, drawing from instruments that have been extensively tested and validated.4 The survey was administered in English and Spanish between 15 July and 7 October 2003 using a standard five-stage survey protocol involving mail and telephone.5 After we accounted for beneficiaries excluded because of death, institutionalization, relocation, non-English/Spanish language, or severe cognitive or physical impairment, the response rate was 51 percent (N = 17,685).6
Analytic methods and variable definitions. We used a combination of bivariate and multivariate methods to examine national and state-level information on seniors’ prescription drug use, out-of-pocket spending, coverage, nonadherence, and purchase from Canada and Mexico. The analytic sample included all respondents for whom required data elements were present (n = 17,569). Probability sampling weights, defined as the inverse of the sampling probability, were applied to all analyses to correct for unequal sampling probabilities across states and strata. The statistical software used (STATA 7.0) takes these weights into account when computing standard errors. We report bivariate results in the exhibits and, where applicable, summarize multivariate results in the text.
Prescription coverage. For beneficiaries reporting more than one source of prescription coverage, a primary coverage source was assigned based on the following hierarchy: Medicaid, employer-sponsored, health maintenance organization (HMO), Medigap (Medicare supplemental), state prescription program, Veterans Affairs/Department of Defense, and other. In this hierarchy, the leading sources of prescription coverage supersede more minor sources, and sources offering more comprehensive coverage supersede those offering less. For people with multiple drug coverage sources, this approach attributes their experiences and out-of-pocket costs to the source that has the largest influence. Beneficiaries indicated by the CMS as having full Medicaid coverage were classified as having Medicaid prescription coverage (n = 2,657), even if the individual did not self-report Medicaid (n = 564).
Poverty. Using the 2003 federal poverty thresholds ($8,988 annual income for a single person; $12,120 for a married couple), together with self-reported income and marital status, we classified seniors as poor (below 100 percent of poverty), near-poor (101–200 percent), or nonpoor (more than 200 percent). For approximately 10 percent of respondents with missing income data, income was imputed based on Buck’s Method.7
Nonadherence. The survey included a series of previously validated questions concerning nonadherence to prescription regimens. All adherence questions referenced experiences over the past twelve months. Factor analysis confirmed our conceptual model of three types of nonadherence: (1) cost-related nonadherence; (2) nonadherence due to medication experiences (for example, side effects); and (3) nonadherence due to self-assessed need for particular medications. A summary indicator of “any nonadherence” denoted nonadherence in one or more of these areas.
Cost-related nonadherence was evaluated with questions about the following three actions: (1) not filling a prescription because of cost, (2) skipping doses to make a prescription last longer, and (3) taking smaller doses than prescribed to make a prescription last longer. Respondents who reported using prescription medicines for one or more named chronic conditions also indicated whether they had failed to fill any of these prescriptions because of cost.8
Experience-related nonadherence was assessed by asking whether respondents had skipped doses or stopped taking a medicine because (1) it was making them feel worse, or (2) because they didn’t think the medicine was helping them. Nonadherence due to self-assessed need was assessed by asking whether respondents had failed to fill a prescription because (1) they felt they were taking too many medicines, or (2) they didn’t think they needed the medicine.
Study Findings
Prescription medication use and out-of-pocket costs. Nationwide, nine-tenths of seniors reported taking prescription drugs, and, among those using at least one prescription drug, nearly half reported using five or more different drugs (Exhibit 1). Pills were the predominant form of medication, although a large percentage also reported using prescribed inhalers, creams, and eyedrops. More than half of seniors reported having more than one prescribing physician, and about one-third used more than one pharmacy.
Out-of-pocket costs for prescriptions were high, with nearly one-third of seniors spending $100 or more per month (Exhibit 1). Five percent of seniors said that they bought some of their prescription medicines from Canada or Mexico, with significantly higher importation rates among those lacking coverage (10.5 percent).
Medication use and spending patterns highlight the particular vulnerability of three subgroups: those without prescription coverage, low-income seniors, and the complex chronically ill (three or more chronic conditions). Although seniors without coverage used significantly fewer medicines than those with coverage (mean 4.0 versus 5.0, p < .001), they were twice as likely to have monthly out-of-pocket costs of $100 or more (49.9 percent versus 25.8 percent, p < .001). Among low-income seniors, half relied on at least five different prescription medicines, and one-third spent $100 or more per month. Reliance on prescriptions was highest among the complex chronically ill, where 73 percent reported five or more different medications and 42 percent spent $100 or more per month.
Forgoing medicines. In the face of these complex and costly prescription regimens, seniors’ rates of medication nonadherence were high, with both cost and other factors figuring importantly (Exhibits 2 and 3).
Nonadherence because of cost. One-quarter of seniors nationwide reported forgoing prescription medications in the past year because of cost (Exhibit 2). Among all three vulnerable subgroups, more than one-third reported some form of cost-related nonadherence, including one-quarter or more not filling at least one prescription because of cost.9 Approximately one-fifth of seniors in each vulnerable subgroup said that they spent less on basic needs to be able to afford their prescriptions. In multivariate models controlling for beneficiaries’ sociodemographic and health characteristics, all three attributes (no coverage, low income, high disease burden) were significantly associated with higher cost-related nonadherence (p < .001).
The important role of coverage with respect to cost-related nonadherence is further delineated in Exhibit 3, which compares nonadherence rates among four chronically ill subgroups with and without prescription coverage. In every case, those without coverage reported significantly more cost-related nonadherence than those with coverage (p < .001). Even with coverage, more than one-quarter of seniors in each chronic illness group reported cost-related nonadherence during the year before the survey. Of serious clinical concern, more than one-fifth of seniors with congestive heart failure (CHF), diabetes, or multiple chronic conditions who lacked coverage said that they did not fill at least one of their chronic disease medications in the past year because of cost. In multivariate models accounting for sociodemographic and coverage characteristics, nonadherence rates did not differ by disease (p > .20).
Nonadherence for reasons other than cost. The survey also revealed substantial nonadherence for reasons other than cost: medication experiences (feeling worse, medicine not helping) and self-assessed need (medicine unnecessary, taking too many medicines) (Exhibit 2). One-quarter of seniors reported some form of experience-related nonadherence: Nearly one-fifth reported having skipped doses or stopped taking a medicine because of side effects, and nearly as many having stopped medications because they felt they weren’t helping. These actions were more common among those lacking coverage (27 percent), low-income seniors (28 percent), and the complex chronically ill (33.8 percent). Forgoing medicines based on self-assessed need was less common but not rare (14.5 percent). Rates were highest among those with complex chronic illness (18.8 percent).
Overall rates of medication nonadherence. Overall, 40 percent of seniors reported one or more forms of medication nonadherence.10 Among seniors without coverage, those with low incomes, and the complex chronically ill, approximately half reported nonadherence. In multivariate models controlling for beneficiaries’ sociodemographic and health characteristics, all three attributes of vulnerability (no coverage, low income, high disease burden) remained significantly associated with higher nonadherence (p < .001). Cost-related nonadherence rates were considerably higher than other forms of nonadherence for seniors lacking coverage and those with low incomes—highlighting the predominance of cost as a barrier for these groups. The complex chronically ill were at highest risk for all three types of nonadherence, which reflects their reliance on multimedication, costly regimens (Exhibits 2 and 3). And the “double jeopardy” posed by complex chronic illness together with no prescription coverage led to nonadherence rates of 66 percent in this group (Exhibit 3).
Variations by source of prescription drug coverage. Nationwide, about one-quarter of all seniors and one-third of poor and near-poor seniors lacked prescription drug coverage (Exhibit 4). Medicaid was the largest source of coverage for poor seniors but played only a minor role for near-poor seniors. Coverage for near-poor seniors was equally likely to come from employers or to be privately purchased. Among higher-income seniors (above 200 percent of poverty), employer-sponsored coverage was the predominant source.
State differences in prescription coverage. These national averages mask substantial state-level differences in the rates and sources of prescription coverage. Across our twelve-state sample, the percentage of seniors lacking prescription coverage was lowest in New York (16 percent) and highest in Louisiana and Washington (35 percent and 36 percent, respectively). Among poor and near-poor seniors, state differences were even larger (Exhibit 4).11 New York and Pennsylvania had significantly fewer poor and near-poor seniors lacking coverage than the national average (p < .05). Illinois also had significantly fewer poor seniors lacking coverage (p < .05), and California had significantly fewer near-poor seniors lacking coverage (p < .05). In these four states, no single factor, such as Medicaid or state pharmacy assistance programs, accounted for the high rates of coverage for low-income seniors. Rather, higher rates were achieved through a combination of factors including these and other sources. At the other end of the spectrum, five states had 40 percent or more of their poor and near-poor seniors lacking coverage (Ohio, Tennessee, Washington, Louisiana, and Texas). These states tended to have low or average Medicaid rates, and none had state-sponsored pharmacy programs.
Employer-sponsored prescription coverage averaged 29 percent nationally. Several states clustered around this level, but rates were significantly higher (p < .05) in Michigan (47 percent), Ohio (38 percent), and New York (37 percent). In all twelve study states, poor seniors were the least likely to have employer-sponsored coverage, and higher-income seniors, the most likely.
Medicaid prescription coverage among poor seniors also varied widely by state (from a low of 19 percent in Pennsylvania to a high of 49 percent in California). Among near-poor seniors, Medicaid rates were more tightly clustered—8 percent or less in all but three states. California stood out with significantly higher Medicaid rates among both poor (49 percent) and near-poor (32 percent) seniors (p < .05).
Prescription use, spending, and reimportation of medicines, by state and coverage source. To further examine variability in seniors’ experiences by state, we compared state and national rates of prescription medication use, out-of-pocket spending, and two indicators of cost-coping strategies (cost-related nonadherence and purchasing prescriptions from Canada and Mexico) (Exhibit 5). To compare the relative influence of state and coverage effects on each parameter and test for interactions, analysis-of-variance (ANOVA) models were applied, and the standard deviation (SD) of state and coverage effects was computed. Both state and coverage were significantly associated with each parameter (p < .001). However, for all parameters except reimportation, coverage effects were two to three times larger than state effects (SDCoverage versus SDState). There were no significant coverage-state interactions, signaling that coverage effects generalize across states.
The results document that there is no oasis for those without prescription coverage. Irrespective of state, those lacking prescription coverage reported higher rates of medication spending and cost-related nonadherence than any other group—despite using fewer medications. Those lacking coverage were also significantly more likely to purchase medications from Canada and Mexico. The highest rates of reimportation were observed among those lacking prescription coverage in Colorado (18.7 percent), Florida (17.3 percent), Washington (17.0 percent), California (12.8 percent), Illinois (12.6 percent), Pennsylvania (12.5 percent), and Texas (12.4 percent).
Nationally and in all twelve states, seniors with employer-sponsored coverage and low-income seniors with Medicaid reported the lowest rates of financial strain related to prescriptions compared with other groups. Notably, reimportation was rare among seniors in both groups (less than 2 percent). Among low-income seniors, the protection afforded by Medicaid coverage was clear when contrasted with low-income seniors covered by other sources and those lacking coverage. Medicaid beneficiaries used more prescription medications than these other groups, owing to their higher disease burden, but they were far less likely to spend in excess of $100 per month out of pocket for medicines. However, despite quite comprehensive coverage, one-quarter of low-income Medicaid enrollees reported cost-related nonadherence.
Finally, regardless of where they lived, those with privately purchased coverage (Medigap, Medicare+Choice, other private plans) reported high spending and cost-related nonadherence and greater use of reimported prescriptions than did seniors with Medicaid or employer-sponsored coverage. Their experiences contrasted sharply with those of seniors reporting employer-sponsored coverage, although they used a similar number of prescription medications, on average. These results affirm our previous findings that prescription benefits available to seniors through private sources fall short of employer-sponsored coverage and Medicaid.12 The findings highlight again that the coverage-versus-no-coverage distinction misses important and consequential differences in coverage quality.
Conclusions And Policy Implications
Five key issues. This 2003 survey highlights at least five issues with direct implications for policymakers as plans for implementing the new Medicare Part D prescription benefit evolve. First, there is no question about the critical role that prescription medicines play in the health care of America’s elderly, nor of the importance of prescription coverage in enabling seniors to sustain complex and costly medication regimens. Second, the very large percentage of low-income seniors who lack prescription coverage—40 percent or more in several states—highlights the immense potential for MMA’s low-income provisions to bring relief, but also the enormity of the task involved in realizing that potential. Third, the positive role played by Medicaid prescription coverage, evidenced by the study’s findings, delineates the high stakes involved in assuring the continued protection of Medicaid enrollees as they are moved into Medicare Part D plans. Medicaid prescription coverage expires 31 December 2005.
Fourth, the finding that not all sources of coverage are equally protective and that the mix of prescription coverage varies markedly by state has important implications for the types of outreach, education, and enrollment strategies needed in different parts of the country. Finally, the high rates of nonadherence to prescription medications due to both costs and other factors indicate that although the new Part D benefit may help mitigate some aspects of nonadherence, others will remain to be addressed through doctor-patient interaction and the larger health care delivery system.
Need for system solutions. These data support other recent evidence showing that most seniors rely on multiple medications to manage their health problems.13 However, the results go beyond previous findings by highlighting not only the complexity of seniors’ prescription regimens, but also the complexity of the process by which seniors obtain these medications—relying on multiple physicians and multiple pharmacies and, in some cases, obtaining medications from abroad. These findings underscore the immense challenge of integrating the care of America’s elderly. For the most part, these challenges require system solutions that are only modestly addressed by the new legislation.
The findings leave no doubt about the difference that prescription coverage makes. And with one-quarter of seniors now lacking such coverage, Part D clearly has the potential to bring both financial relief and improved health care quality to this group. This is especially true for the millions of low-income seniors without prescription coverage for whom MMA affords considerable subsidies. Nationwide, one-third of low-income seniors lack such coverage, and in several states more than 40 percent do. Reaching this group of seniors—to both establish their eligibility for subsidized coverage and enroll them in a Part D plan—represents an enormous administrative task, and plans for doing so are not yet well specified. Previous experiences and data from this study highlight the difficulties and variable success that federal and state programs have had in achieving high enrollment rates among low-income populations.14
Our findings also demonstrate the considerable protection and advantages that Medicaid provides, which underscores the high stakes involved in moving 6.3 million dually eligible beneficiaries from Medicaid to Part D plans and ensuring the continued adequacy of their protection over the long term.
Source of coverage matters. The study’s findings also underscore that a drug benefit alone does not guard against high prescription costs and that certain sources of coverage offer much more protection against high out-of-pocket costs than others. In particular, employer-sponsored coverage and Medicaid provided considerably more financial protection than privately purchased plans (such as Medigap and Medicare managed care). In some states, these higher-quality coverage sources were much more prevalent than others, which further underscores that MMA implementation strategies will need to account for local circumstances. For example, beneficiaries in states with a large presence of employers offering retiree drug benefits, such as Michigan, Ohio, and New York, will require educating and counseling retirees about their particular coverage options in coordination with employers in those states. By contrast, states such as Pennsylvania, Colorado, California, and Florida, with relatively high enrollment in Medigap and Medicare+Choice plans, will face a very different task in educating seniors about the choices available in 2006.
The observed differences in coverage generosity and the observed consequences of these differences for seniors illustrate that the new Medicare drug benefit’s potential for mitigating high drug spending will depend on the generosity of Part D coverage in 2006 and over the long term. One noteworthy consequence, when coverage was absent or limited, was increased use of reimported medications. The findings suggest that where gaps in coverage persist—for example, among those who reach coverage limits characterized as the “doughnut hole”—seniors will use a variety of coping strategies, including potentially seeking medications from abroad.
Nonadherence. A far more prevalent coping strategy, however, was nonadherence with prescription regimens. Nonadherence rates were extraordinarily high—and costs were a central factor, particularly among seniors who lacked coverage, those with low incomes, and those with multiple chronic illnesses. However, seniors’ experiences with medications and their doubts about the value of various medications also contributed to high nonadherence rates. Available evidence suggests that nonadherence is going undiscussed and possibly unnoticed in the doctor-patient interaction.15 The magnitude and extent of nonadherence reported by seniors nationally—and most notably those reported by seniors with chronic medical conditions—demand that doctors and patients begin to discuss and address these issues. While the clinical encounter itself is outside the scope of the MMA provisions, the law offers a new opportunity for doctors and patients to engage in the long-taboo topic of adherence and the “double-taboo” topic of cost-related nonadherence. Beneficiaries will need assistance finding coverage that helps them best afford their prescriptions, and they will likely turn to their doctors for guidance.16
Study limitations. There are several relevant study limitations. First, the study excluded institutionalized Medicare beneficiaries (2.5 million) and those younger than age sixty-five (5.7 million). Well-established differences in prescription coverage sources and use among these groups, relative to the noninstitutionalized elderly, suggest that our findings would not generalize to them.17
Second, our use of coverage hierarchies ascribes beneficiaries’ prescription experiences to one coverage source. For the minority of beneficiaries with more than one source, the benefits of secondary coverage sources are ascribed to the primary. Despite this limitation, coverage hierarchies are widely used in Medicare research because the alternative of assigning beneficiaries to multiple sources of coverage blurs distinctions between sources and gives secondary sources the appearance of comprehensiveness that owes to the primary source.
Finally, the study achieved only a modest response rate. Customary of survey research, nonrespondents were older and disproportionately of a minority race and lower socioeconomic status. If nonrespondents from these vulnerable subgroups were more disenfranchised than their counterparts who responded, the study could overestimate coverage rates and underestimate nonadherence. However, allowing for well-established market changes since 1999, the observed coverage rates and mix are comparable enough with 1999 MCBS data to suggest that nonresponse did little to distort coverage data.18 In addition, since nonresponse characteristics did not differ by state, state comparisons presented here can be presumed accurate.
The new Medicare Part D benefit represents an important opportunity to help reduce seniors’ prescription cost burden and, in particular, to provide additional assistance to low-income seniors and those with catastrophic drug spending. Findings from this survey call attention to the many challenges that lie ahead in assuring the smooth implementation and transition to the new program. The recent experiences of implementing MMA’s discount card program illustrate the complexity of putting hundreds of new Medicare-endorsed products in place and of informing forty-two million beneficiaries about their choices. Finally, lessons from this study underscore that coverage is necessary but not sufficient, and that continued tracking and monitoring of coverage experiences including prescription use, spending, and adherence will be critical to measuring the impact of the new Medicare prescription drug program.
This research was supported by grants from the Commonwealth Fund and the Henry J. Kaiser Family Foundation. The authors gratefully acknowledge Spike Duzor, Maribel Franey, and Dural Suite (Centers for Medicare and Medicaid Services) for their assistance with the authors’ data request; and Theresa Sommers and Barbara Seltzer (the Health Institute, New England Medical Center) for their dedicated assistance in preparing this manuscript. The contents of this paper are the sole responsibility of the authors and do not necessarily represent the views of the Commonwealth Fund, the Henry J. Kaiser Family Foundation, or Tufts–New England Medical Center.
NOTES
1. The last published national data on seniors’ prescription coverage rates are from the 1999 Medicare Current Beneficiary Survey (MCBS). See, for example, M.A. Laschober et al., “Trends in Medicare Supplemental Insurance and Prescription Coverage, 1996–1999,” Health Affairs, 27 February 2002, content.healthaffairs.org/cgi/content/abstract/hlthaff.w2.127 (14 March 2005). MCBS coverage data for 2001 published to date present combined rates for elderly and disabled beneficiaries. See, for example, Medicare Payment Advisory Commission, A Data Book: Healthcare Spending and the Medicare Program, June 2004, www.medpac.gov/publications/congressional_reports/Jun04DataBook_Entire_report_links.pdf (14 March 2005). We previously reported 2001 prescription coverage rates among seniors in eight states. See D.G. Safran et al., “Prescription Drug Coverage and Seniors: How Well Are States Closing the Gap?” Health Affairs, 31 July 2002, content.healthaffairs.org/cgi/content/abstract/hlthaff.w2.253 (14 March 2005).
2. Geocoding assigns latitude and longitude to a known address. The coordinates are mapped into census block groups (CBGs), allowing census neighborhood variables to be attached to the beneficiary. CBGs are smaller, more homogeneous units than ZIP code areas, counties, or census tracts and thus provide a useful proxy for individual socioeconomic characteristics.
3. Samples in New York and California were somewhat larger to enable further analyses in these states.
4. The questionnaire is available from the authors; send e-mail to dsafran{at}tufts-nemc.org.
5. D.A. Dillman, Mail and Telephone Surveys: The Total Design Method (New York: John Wiley, 1978).
6. As is customary with survey research, nonrespondents were older and disproportionately of minority race and lower socioeconomic status than respondents (p < .001). However, national and state poverty rates and race/ethnicity accord with those reported by the Current Population Survey, which suggests that the study achieved a similarly representative sample. See U.S. Census Bureau, Current Population Survey, 2000–2002 Annual Social and Economic Supplements. For number of people over age sixty-five, see Administration on Aging, “Census Bureau Population Estimates as of July 1, 2003,” www.aoa.gov/prof/Statistics/2003Pop/Stterr2003_files/sheet003.asp (4 April 2005); for 2003 race estimates, see ibid., sheet008.
7. S.F. Buck, “A Method of Estimation of Missing Values in Multivariate Data Suitable for Use with an Electronic Computer,” Journal of the Royal Statistical Society B22 (1960): 302–306. Ordinal logistic regression was used to estimate the conditional distribution (predicted probabilities) of the missing data. Missing values were replaced with values having the highest probability among the categories.
8. Health conditions included hypertension, high cholesterol, heart problems, asthma, emphysema or chronic obstructive pulmonary disease (COPD), diabetes, rheumatoid arthritis, osteoarthritis, degenerative joint disease, and depression.
9. Cost-related nonadherence rates observed here are much higher than rates in the MCBS data from 1996–1999. See B.M. Craig, D.H. Kreling, and D.A. Mott, “Do Seniors Get the Medicines Prescribed for Them? Evidence from the 1996–1999 Medicare Current Beneficiary Survey,” Health Affairs 22, no. 3 (2003): 175–182. At least three factors likely contribute. First, social-desirability response bias is known to be higher in more personal modes of survey administration. In the paper by Benjamin Craig and colleagues, two contemporaneous in-person surveys (the MCBS and Asset and Health Dynamics) found similar nonadherence rates (below 5 percent), while studies using less personal modes found much higher rates (9–14 percent). A second likely contributing factor is different question wording. Open-ended questions about events and behavior have been shown to yield less complete recall than more specific probes. See G. Menon and D.A. Yorkston, “The Use of Memory and Contextual Cues in the Formation of Behavioral Frequency Judgements,” in The Science of Self-Report: Implications for Research and Practice, ed. A.A. Stone et al. (Mahwah, N.J.: Lawrence Erlbaum Associates, 2000). A third likely factor accounting for the differences is time. Recent trending analyses reveal that cost-related nonadherence greatly increased among seniors after 1998. See I.B. Wilson et al., “Cost-related Skipping of Medications and Other Treatments among Medicare Beneficiaries between 1998 and 2000: Results of a National Study,” Journal of General Internal Medicine (forthcoming).
10. Adherence research has long documented nonadherence rates to be on this order of magnitude. See, for example, D.L. Sackett and J.C. Snow, “The Magnitude of Compliance and Noncompliance,” in Compliance in Health Care, ed. R.B. Haynes et al. (Baltimore: Johns Hopkins University Press, 1979); and C.A. Jackevicius, M. Mamdani, and J.V. Tu, “Adherence with Statin Therapy in Elderly Patients With and Without Acute Coronary Syndromes,” Journal of the American Medical Association 288, no. 4 (2002): 462–467.
11. For an expanded exhibit showing the data for all twelve states, see
content.healthaffairs.org/cgi/content/full/hlthaff.w5.152/DC2.
12. Safran et al., “Prescription Drug Coverage and Seniors.”
13. J.F. Moeller, G.E. Miller, and J.S. Banthin, “Looking Inside the Nation’s Medicine Cabinet: Trends in Outpatient Drug Spending by Medicare Beneficiaries, 1997 and 2001,” Health Affairs 23, no. 5 (2004): 217–225; C.P. Thomas et al., “Impact of Health Plan Design and Management on Retirees’ Prescription Drug Use and Spending, 2001,” Health Affairs, 4 December 2002, content.healthaffairs.org/cgi/content/abstract/hlthaff.w2.408 (14 March 2005); and C.W. Burt, “National Trends in Use of Medications in Office-based Practice, 1985–1999,” Health Affairs 21, no. 4 (2002): 206–214.
14. D.K. Remler, J. Rachlin, and S.A. Glied, “What Can the Uptake of Other Programs Teach Us about Take-up of Health Insurance Programs?” NBER Working Paper no. w8185 (Cambridge, Mass.: National Bureau of Economic Research, March 2001).
15. G.C. Alexander, L.P. Casalino, and D.O. Meltzer, “Patient-Physician Communication about Out-of-Pocket Costs,” Journal of the American Medical Association 290, no. 7 (2003): 953–958; J.D. Piette, M. Heisler, and T.H. Wagner, “Cost-related Medication Underuse among Chronically Ill Adults: The Treatments People Forgo, How Often, and Who Is at Risk,” American Journal of Public Health 94, no. 10 (2004): 1782–1787; and J.D. Piette, M. Heisler, and T.H. Wagner, “Cost-related Medication Underuse: Do Patients with Chronic Illness Tell Their Doctors?” Archives of Internal Medicine 164, no. 16 (2004): 1749–1755.
16. Henry J. Kaiser Family Foundation, “New Survey Assesses Senior’s Views of Medicare Drug Law,” Health Poll Report, January 2005, www.kaiserfamilyfoundation.org/kaiserpolls/pomr012705pkg.cfm (14 March 2005).
17. Laschober et al., “Trends in Medicare Supplemental Insurance”; J.A. Poisal and G.S. Chulis, “Medicare Beneficiaries and Drug Coverage,” Health Affairs 19, no. 2 (2000): 248–256; and B. Breisacher et al., “Medicare’s Disabled Beneficiaries: The Forgotten Population in the Debate over Drug Benefits,” September 2002, www.kff.org/medicare/upload/14181_1.pdf (14 March 2005).
18. Laschober et al., “Trends in Medicare Supplemental Insurance.”
Dana Safran (dsafran{at}tufts-nemc.org) directs the Health Insurance at Tufts-New England Medical Center in Boston. Patricia Neuman is vice president and director of the Medicare Policy Project, Henry J. Kaiser Family Foundation (KFF), in Washington, D.C. Cathy Schoen is vice president for health policy, research, and evaluation at the Commonwealth Fund in New York City. Michelle Kitchman is a senior policy analyst at the KFF. Barbara Cooper is senior program director; Medicare's Future, at Commonwealth. Ira Wilson is an associate professor of medicine at Tufts-New England Medical Center and the Tufts University School of Medicine. Angela Li is a research associate; Hong Chang, a statistician; and William Rogers, a senior scientist, at the Health Institute.
Read related papers by: Bruce Stuart and colleagues, Cindy Thomas and colleagues, and Robert F. Atlas
DOI:
10.1377/hlthaff.w5.152
©2005 Project HOPE–The People-to-People Health
Foundation, Inc.
|
