Income-, education- and gender-related inequalities in out-of-pocket health-care payments for 65+ patients - a systematic review

As shown in Figure 1, an online PubMed search was conducted to identify studies. Search terms included combinations of the following keywords: "cost sharing"[All Fields] OR "copay"[All Fields] OR "copayments"[All Fields] OR "out of pocket"[All Fields] OR "direct payments"[All Fields] OR "incentive based"[All Fields] OR "patient charge"[All Fields] OR "prescription charge"[All Fields] OR "coinsurance"[All Fields] OR "deductible"[All Fields] OR "extra billing"[All Fields] AND "aged"[MeSH Terms]. The search process ended on November 9^th, 2009. Studies' abstracts were examined in detail, and, if required, extended by a full text revision.

Unless separate data analyses were conducted for the 65+ subsamples, studies not limited to the senior population (defined as 65 years of age or older) were excluded. Only studies in English and German language based on evaluation of primary data were considered. Finally, articles had to contain relevant information concerning education, sex and/or income in relation to out-of-pocket payments for people 65+ (as illustrated in Figure 1).

To facilitate comparisons, all cost estimates were inflated to 2008 US dollars using the Chain Type Price Index for Gross Domestic Product (US Census Bureau 2010). The following information was systematically extracted (see Table 1): authors, year of publication, sample size indicating the validity of the article (varying from 113 to 30,791,751 persons), study design, analyzed outcome, key findings stratified according to income, education and gender, and confounders controlled for.

This review is structured as follows: every socio-demographic variable has its own chapter describing its association with OOPP. As income contains the most information, the respective chapter is further structured by overall expenditures, and their biggest share, prescription drugs. Furthermore, details on impacts like insurance types or cost-reducing strategies are reported. After describing the association of education and sex with OOPP, a conclusion at the end of this review summarizes all findings, confronts them with theories to put them in scientific perspective, and gives implications.

995 papers were found via PubMed, and 17 papers were found via bibliographic search in reference lists of eligible articles, resulting in a total of 1012 studies. After exclusion of studies not focusing on out-of-pocket payments for people 65 years of age or older, 217 studies remained. Out of these, 29 articles remained containing relevant information concerning education, sex and/or income in relation to out-of-pocket payments for people 65+ (as illustrated in Figure 1). Included are 11 longitudinal and 18 cross-sectional surveys, of which 9 articles use the MCBS (Medicare Current Beneficiary Survey), making it the most utilized source of data. Five studies are purely descriptive [3‐7], while the remaining 24 use inferential statistics to test differences in out-of-pocket payments between income, education and gender groups. Of these, 14 controlled for confounders, meaning independent control variables were evaluated (for respective confounders in articles see columns in Tables 2, 3 and 4). The articles used in this review originate from 1993-2009, from which 21 were published before the implementation of the Medicare Prescription Drug, Improvement, and Modernization Act of 2003, an important break marking the inclusion of a voluntary drug benefit (Part D), taking effect in January 2006 [8]. Studies varied in analysis of absolute expenditures and its relation to income [9‐11], or just focusing on the burden [7, 12, 13].

Several articles refer to education as a significant cause for differences among older persons and their burden of OOPP for medical services. The main finding is that the lower the level of education, the higher the burden becomes [10‐12]. Lower education may lead to lower income and lack of employer based health programs [15, 20]. Concerning the burden of overall out-of-pocket costs in 1995, Crystal et al. [12] found that beneficiaries with college degrees spent 12.8% of their income on health-care, while those without a high school diploma spent 21.4%. Focusing on burden of prescription drugs costs only, Rogowski et al. [11] found that beneficiaries with less than 12 years of education paid 4.5% of their incomes on prescription drugs in 1990, while those with more than 12 years of education only 1.6%. As suggested before, this was mainly caused by lower retirement incomes and less prevalence of private insurance. Sambamoorthi et al. [10] found that 12.1% of those without a high school diploma had to pay more than 10% of their income on prescription drugs in 1997, but only 3.9% of seniors with college education had to. Conversely, a burden of 0-5% was determined for 56.9% of those without a high school diploma, while 77.8% of college graduates belonged to this group. Increments of education confirm this trend proportionally. Regarding insurance coverage, Pourat et al. [15] found in 2000 that the likelihood of an employment-based or MediGap coverage is raised by better education, while a poor education increases the probability of basic Medicare fee-for-service insurance only, or no coverage at all, leading to higher OOPP in case of morbidity. Saver et al. [20] confirmed this observation in 2004: only 39% of those without coverage faced monthly expenditures below $50, while 79% of patients with private prescription coverage did. Equivalently, 8% of not-covered beneficiaries faced more than $100 per month, while nobody of the second group had to pay this amount. The poorest are secured by Medicaid, but the ones shortly above the income limit for eligibility are confronted with catastrophic costs that eat up all their savings before Medicaid steps in [9]. Another point is the Medicare Part D asset test, denoting specific income and asset thresholds set to qualify for low-income subsidies [6]. Those low-income seniors failing the test mostly have better education than those who qualify for subsidies (college degree: 9.7% vs. 4.4%). But because approximately 70% of those only have incomes less than 135% of the federal poverty line, and around 50% of them only have assets less than $35.000 above the allowing thresholds. These assets would not pay a single year of nursing home care. Compared to beneficiaries without any assets, they have to deal with higher OOPP without governmental support, and in relation to those with higher incomes and larger assets, they have to pay similar OOPP with less means [6]. Furthermore, a low educational standard is associated with lower utilisation of preventive measures. In various studies [24, 25] clarify that a college degree is highly associated with preventive behaviour like mammography. Not using preventive measures finally leads to higher OOPP due to worse health status.

There are certain gender related differences concerning OOPP. Selden et al. [7] found that in women, prevalence of high burden for medical services is significantly higher than in men for all age groups. For example, 12.2% of females aged 75 years and older faced burdens larger than at least 40% of after-tax disposable income in 1996, but only 9.2% of men. Concerning prescription drugs, in 1995 women had 18% more annual OOPP than men ($551 vs. $454) [26]. In a longitudinal study from 1992-2000, Wei et al. [27] found the similar inequality ($526 vs. $432) in expenditures. Regarding the burden of OOPP for prescription drugs, Wei et al. describe a gender difference of 28% (4.4% vs. 3.2%) [27]. While finding equal absolute OOPP in 1996, Rogowski et al. [11] approve the finding that women have a higher relative burden (3.3% vs. 2.8%). Sambamoorthi et al. [10] found that 9.4% of women faced a burden of at least 10% of their income for OOPP on prescription drugs, compared to only 5.7% in men. The more distinct the gender differences in health status and income are, the higher the difference in burden appears. One reason for the differences is that women are less likely to be employed, and thereby generate less wealth. Even if women pursue a career, on average they have significantly lower income than men. Wei et al. stated that in 1999 the mean income for women aged at least 65 years was $19.097, compared to $35.676 for men [27]. 26% of female Medicare beneficiaries live below the federal poverty line, whereas only 11% of men. Being less often employed also means fewer possibilities to be covered by an employer-financed supplemental insurance (including drug coverage) to protect women from high burdens for out-of-pocket medical expenses [26]. Aggravating this is the fact that women consume more medication than men [27], for example for not incurred complementary and alternative medicine [28]. Considering that women in general have to pay more out-of-pocket for medication and special equipment, as Lapsley et al. illustrate [29], they are more likely to rely on the aforementioned strategies to reduce costs [8]. Furthermore, gender-specific morbidities like breast cancer cause higher OOPP, since appropriate health-care plans have less generous benefits [27]. An additional aspect is women's longer life expectancy, thereby producing the greater likelihood for chronic diseases in general [4]. For financial issues, it is also relevant that women, due to their longer lifespan, have a higher probability become a widow. Those in the lowest income quartile (below $12.000) pay approximately 70% of their income in their final two years for health-care, whereas the average is around 30%, leading to higher poverty rates (17% vs. 5%). Among other reasons, OOPP in the last year of the spouses' life and the loss of the spouses' income raise the likelihood of old age poverty for widows by 56% [5]. Rice et al. deliver proof that if widowed women benefit from an inheritance, they are very likely to be excluded from needed low-income subsidies, because their assets exceed the limit Medicare Part D foresees, so that the assets first have to be spent out-of-pocket before the federal government steps in: 46% of those failing the asset test are widowed, and almost all are female (~93%) [6].

Most of the reviewed studies describe inequalities in OOPP for medical services, in relation to diverging income proportions. The largest amount of OOPP can be assigned to prescription medications, leading to cost-induced strategies like non-adherence, implicating further health problems. This effect is further enhanced by the widespread lack of preventive measures among low-income beneficiaries, partly caused by lack of supplemental insurance. This describes women and low-educated layers of society. Considered as percentage share, women, lower social and low-income classes are far heavier affected. While these variables explain a great part of income inequalities, which finally lead to inequalities in OOPP, they are also predictors of inequalities in their own right. For example, being female and low-educated leads to being employed less or working in worse paid jobs than higher-educated people. A cycle of low income, no supplemental insurance, less paid services, worse health and thereby financially unfeasible need is set in motion. Besides inequalities associated with lower income, several studies indicate higher out-of-pocket burdens on a gender- and education-specific level. Concerning sex, gender-specific illnesses (like breast cancer or other chronic diseases that are more likely to occur due to a longer lifespan), and higher costs for medications and special equipment influence the health-care burden without being associated with income. On the education side, problems like lower awareness of a healthy lifestyle, reflecting less use of preventive measures, may constitute income-independent disadvantages. Furthermore, another effect can be observed: Neuman et al. [30] show that the implementation of Medicare Part D in 2006, resulting in a rising number of private insurers, led to a larger variety of available plans to choose from. This could have left low-educated beneficiaries in a confusing situation, making it hard to find adequate insurance options trying to face all possibilities and background information. As public health policy is unable to influence education's role in the short-term by providing information and transparency, possible approaches to a solution should be based on the strong correlation of income and education. Thus, appropriate measures ought to be largely coherent with those for income and burden, for example income thresholds according to beneficiaries' burden. Besides education, an enhanced preference of low-risk beneficiaries in popular plans could have been initiated by the insurers, leaving disadvantaged seniors aside. This could explain why despite all reform efforts since 2006, the number of about four million beneficiaries without drug coverage remains constant up to today. As several studies reviewed show the importance of access to prescription drug coverage for the amount of OOPP prior to the implementation of Medicare Part D in 2006 [10‐12, 14, 17‐20, 22, 23, 26, 27, 31], its initiation should have settled this problem. But as studies analyzing data after 2006 reveal [8, 30, 32], details like the "Donut Hole" coverage gap still remain responsible for inequalities. Based on the studies reviewed, an explanation for inequalities in OOPP by these factors remains ambiguous. Also, methodically, the diversity of controlled confounders could have caused difficulties in comparing results.

To give a starting point for subsequent studies examining mechanisms causing these inequalities, hereafter, the findings will shortly be put in conjunction with relevant theories. The aim is to show what other variables could cause effects on OOPP inequalities. Mojtabai et al. [23] demonstrated that severe health problems lead to more medication, and thereby higher OOPP, resulting in a higher probability of non-adherence, poorer health and more hospitalizations, especially for those without supplemental insurance and low income paired with high OOPP. This summary is quintessential for most of the findings in this review. The coherence between high OOPP and non-adherence to medication is observed by several authors [8, 21, 22], as well as the association of cost-induced strategies and worse health outcomes [19]. The intensification of this problem for low-income beneficiaries in a disadvantageous insurance situation is also evident [22]. As this finding refers to income only, Sambamoorthi et al. [10] put other variables in perspective by arguing that generation-comprehensive predispositions may impact actual developments. Parents' low income leads to poor education. A corresponding workplace, which results in low income (and a lack of sufficient assets), leads to a lifestyle detrimental to good health and preventive measures, which itself leads to high OOPP that one cannot afford. Reinforcing this development, inequalities intensify with age. The connection of low education, poorly paid jobs and corresponding income and insurance, is supported by several studies reviewed [8, 25]. Also, coherence is found between less income and a lack of awareness of health, reflecting less adherence to medication and preventive measures [19]. The resulting effect of higher OOPP has also been demonstrated [15, 22]. To put all questioned variables in perspective, several articles confirm that being female is an additional aspect worsening the situation by, for example, lower income and a longer lifespan [5, 6, 17, 26]. So generally, Sambamoorthi et al.'s thesis can be confirmed. But it also becomes clear that the variables income, education and gender cannot explain the whole context independently, as other aspects throw different light on the argumentation. Ahrens [33] also identifies overall working-, living- and environmental circumstances as main influences on health status. He states that predisposing factors (age, sex, congenital diseases), medical infrastructure (quantity and quality, financial means), social status (lifestyle, income, education) and comprehensive variables (education system, environmental quality, labour conditions) as individual and systematic circumstances finally determine the means to deal with high OOPP. Therefore, it can be assumed that a concluding bottom line of coherence cannot be drawn solely on the basis of income, education and gender, as numerous other variables seem to bear an unexplored amount of influence, whose extent has to be analyzed in future studies.

Finally, a short windup addressing possible approaches to overcome inequalities in burden will be provided. Although beneficiaries can influence their situation by, for example, granting preventive measures a high priority, their information deficit puts them in a dependent position. Their state is defined by the physician's diagnosis, who profits from the divide in information, competence and a lack of transparency concerning the iatrogenic costs [33]. This may be a point to exercise pressure, by developing a system not paying the physician by the number of measures prescribed, but by the quality of care, reflecting his patients' health status. McCormick et al. [34] found that only 9.1% of US-physicians probed support the status quo, while reform proposals like adding tax credits or tax penalties (49.2%) or single payer NHI programs (41.6%) are favoured. When raising OOPP is necessary due to fiscal reasons, it should occur in relation to the patient's income. Chandra et al. further suggest that one regards health status as a second factor to determine the income-related limit of cost-sharing in order to protect chronically ill patients from catastrophic expenditures [31]. The focus should lie on equal burdens, not equal expenditures. By doing so, inequalities could be alleviated. Another problem to solve is "cream skimming", the insurer's selection of low-risk beneficiaries. As this issue is intensified by an increasing number of plans to choose from, the variety of plans should be reduced to a reasonable and transparent extent. Also, the government could limit the influence of private insurers on the market, so that high-income beneficiaries cannot exclude themselves from the solidarity system, leaving society socially unbalanced [30]. Additionally, the inequality problems in connection with the Donut Hole [8, 18] and the Medicare Part D asset test [6] should be identified as a financial threat. If OOPP have to be imposed to stabilize the budget situation for future tasks, their implementation should be carefully thought over, with great importance attached to socially balanced arrangements.