To explore implications of using systematic drug class reviews to develop U.S. Medicaid drug formularies. We assess racial/ethnic, gender, and socioeconomic status (SES) concordance between Medicaid populations and studies synthesized in Drug Effectiveness Review Project (DERP) systematic reviews.
Study Design and Setting
Review of 32 DERP systematic reviews for subpopulation reporting/analysis and concordance with Medicaid populations.
Results
Among Medicaid recipients in DERP member states and nationally, minorities are overrepresented (21% to 57%) compared with their presence in the population (10% to 30%). Fifty-nine percent of DERP reviews reported insufficient evidence to evaluate drug effects by race/ethnicity or gender. Three percent of reviews found evidence of differential effects by race and 13% by gender. Twenty-four percent found evidence of no difference by race and 9% found no difference by gender. Most of this evidence was described as weak, limited, or of poor quality. Eighty percent of Medicaid recipients are poor or near-poor. DERP does not report on SES.
Conclusion
DERP reviews reveal deficiencies of the evidence when applied to Medicaid populations. To increase health equity and provide evidence for policies that serve socially disadvantaged populations, drug trials, and other studies should include more members of these populations. Systematic reviews should include low-SES as a prespecified subgroup.
Introduction
What is new?
Key findings
•
Systematic drug class reviews conducted by the Drug Effectiveness Review Project (DERP) are used by some U.S. states to develop Medicaid drug formularies. Our examination of DERP reviews reveals an evidence base that is not representative of poor and minority populations that are overrepresented on Medicaid rolls.
•
Most DERP reviews reported insufficient evidence for any analysis of differential effects in subgroups defined by gender or race/ethnicity. Much of the evidence that was found in the remaining reviews was described as “weak” or “limited.” DERP reviews did not include socioeconomic status (SES) as a prespecified subgroup.
What this adds to what was known.
•
It is not possible to determine from currently available evidence if poor, minority, or female patients respond differently to many pharmaceutical therapies, or if social factors affect treatment response.
•
Most of the evidence for differential drug response in women and minorities is of poor-to-fair quality.
What is the implication, what should change now?
•
Evidence should be collected and analyzed in ways that allow policymakers to develop policies that support health equity in disadvantaged populations. Low-SES individuals should be an included subgroup in systematic reviews. Systematic reviews can help to identify the deficiencies of evidence in pharmaceutical research when applied to specific subgroups and guide the direction of future research and health policy.
The World Health Organization (WHO) recommends that national health care policies include evidence-based lists of essential medicines to allow equitable availability of drugs that have public health relevance, are safe, efficacious, and cost-effective, and are appropriate for the populations served by national health care programs [1], [2]. Promoting health equity requires that policies and programs improve health outcomes for all sectors of the population by eliminating avoidable inequalities, particularly those that result from injustice or social exclusion. These policies should be based on inclusive research, and they should be evaluated to ensure that they are not doing harm or having unintended negative effects [3], [4]. Health disparities are likely to be increased, not decreased, by the widespread implementation of policies that seem to improve overall health outcomes for the population as a whole, but which benefit the advantaged, while providing little benefit for the disadvantaged [5], [6].
Although the United States has no national health plan, Medicaid programs provide medical care and drug coverage to some low-income adults and children. Only individuals and families that fit into eligible groups defined by federal and state regulations qualify for and receive Medicaid. Although the federal government provides funding, Medicaid programs are administered by individual states. Therefore, eligibility requirements and benefits vary greatly from state to state, and services may be provided on a fee-for-service basis or through privatized managed care. Income is only one of many criteria for determining Medicaid eligibility, and not all low-income people, not even the very poor, qualify for the program. Some of the other factors that affect eligibility include age, assets, citizenship or immigration status, pregnancy, and disability. Some Medicaid recipients also receive benefits from welfare programs or disability insurance, and some disabled people receive medical care through programs other than Medicaid. Although most Medicaid recipients are poor or near-poor, some recipients are “medically needy.” These individuals have certain types of excessive medical expenses, but their income or assets are above the usual Medicaid eligibility limits. [7].
At the time this research was conducted, 131 of the 50 U.S. states participated in the Drug Effectiveness Review Project (DERP), which conducts systematic drug class reviews. DERP describes itself as “a self-governing collaboration of public and private organizations…which have joined together to provide systematic, evidence-based reviews of the comparative effectiveness and safety of drugs in many widely used drug classes and to apply the findings to inform public policy and related activities in local settings” [8]. States may use DERP reviews and other sources of evidence to develop formularies, also known in Medicaid as preferred drug lists (PDLs), which are lists of drugs routinely covered by Medicaid programs.
The use of drug formularies is not unique to Medicaid programs; private and public health insurers and providers in the United States and internationally are increasingly using evidence-based formularies in attempts to contain costs while providing prescription drug benefits to their beneficiaries [2], [9], [10], [11]. This article explores the health equity implications of using scientific evidence, as summarized in systematic drug class reviews, in the development of PDLs for the fee-for-service component of Medicaid and related programs. We examine racial/ethnic, gender, and socioeconomic status (SES) concordance between study populations included in DERP systematic reviews and in Medicaid populations in the United States and in DERP member states. We then examine the availability and relevance of evidence for the development of appropriate and equitable evidence-based health policy for Medicaid populations.
A systematic review is a study that uses systematic and explicit methods and clearly formulated questions to identify and critically appraise all studies on a particular topic. Relevant studies are included in the review, and reviewers assemble and analyze data from the included studies [12]. Statistical methods (i.e., meta-analysis) may or may not be used to analyze and summarize the results of the included studies. Systematic drug class reviews compare drugs within a class, for example, antihistamines or statins, to look for evidence of difference in efficacy and safety among the various drugs in the class. Although some critics raise questions about the methodology used to produce systematic reviews [13], [14], they are widely recognized as a useful and reliable source of evidence [15], [16], [17], [18], and are used in the development of drug formularies and in other health care decision making [19]. Participating states and other DERP member agencies determine which drug classes will be reviewed, and what key questions will be addressed in the reviews [8].
Policymakers say that to provide the best and most effective health care, they need access to accurate and unbiased research. However, the large number of published studies, often with conflicting conclusions, makes it difficult to use the available scientific evidence in policy decisions [20]. Although systematic drug class reviews like those produced by DERP and the Cochrane Collaboration do not include drug costs in their studies, public and private decision makers incorporate cost information into their calculations when deciding which drugs to cover [14], [17]. Medicaid decision makers worry that soaring drug costs may threaten states' ability to maintain coverage and services while still providing prescription drug benefits [10]. Because federal law does not permit states to implement closed formularies as Medicaid PDLs, patients are allowed access to nonformulary drugs if prescribed by their medical providers. Barriers to obtaining nonformulary drugs vary by state, and some states' previous authorization and appeal processes are more restrictive than others. The use of PDLs is one of several strategies used by states to limit the sharply rising costs associated with Medicaid prescription drug benefits. Other cost-containment measures include restricting the number of prescriptions beneficiaries can receive, and capping payments to drug providers [21].
Well-conducted systematic drug class reviews are useful to policymakers and consumers because they consolidate information, and can identify harmful effects and benefits of treatments. Although both individual studies and review articles that are funded by pharmaceutical companies have been shown to produce systematic bias in favor of the drug that is produced by the company sponsoring the research [22], [23], [24], systematic reviews produced by authors with no financial interest in promoting the drugs they are reviewing are less likely than industry-funded studies to be skewed toward the most profitable treatments. Systematic reviews are also less biased than narrative reviews or consensus-based decisions [15], [16], [18], [25]. Nonetheless, critics and supporters of systematic reviews have concerns about the methods used to produce them and the applicability of the findings they generate. Even systematic review authors with no financial conflicts of interest rely at least partly on the industry-funded studies that make up much of the evidence base, and this may in turn effect the accuracy of the results. Arguments have also been made that systematic reviews are based primarily on randomized clinical trials, which have been criticized for being poorly generalizable to high-risk populations and for not examining how drugs work outside of comparatively well-controlled study conditions. Furthermore, critics argue that patients' responses to drugs can vary greatly, especially in diverse or high-risk populations, and that it is unclear whether or not the results of trials conducted in predominantly white and middle-class populations may be generalizable to other patients and other settings [6], [13], [26], [27].
Concerns about the generalizability of clinical research and questions about the potential harms and benefits of using evidence-based PDLs are also raised by those outside the pharmaceutical industry, including patient advocacy groups, insurers, policymakers and other decision makers, and researchers [9], [13], [26], [28]. Although some patient advocacy groups are financed by the pharmaceutical industry, others are not. For example, consumers united for evidence-based Health care, which is affiliated with the Cochrane Collaboration, http://apps1.jhsph.edu/cochrane/uscccc.htm, is made up of patient and consumer groups that do not receive any substantial industry funding. However, it is sometimes difficult to accurately determine the source of funding for patient and consumer advocacy groups.
Most Medicaid enrollees are poor or near-poor. Nationally, whites are the largest single racial/ethnic group among Medicaid enrollees, but racial/ethnic minorities make up more than half of the Medicaid population [29]. Although guidelines issued by National Institutes of Health (NIH) and the U.S. Food and Drug Administration require that women and racial/ethnic minorities be included in clinical trials, reporting for these groups and also for low-SES populations is often inadequate and these groups may be inadequately enrolled and analyzed [30], [31], [32], [33], [34]. Pharmaceutical industry supporters sometimes argue against the use of drug formularies on the grounds that there is insufficient evidence about differential drug effects in diverse populations [3], [13]. However, trials funded by the industry are less likely to analyze or report enrollment of women and minorities than clinical trials that are funded by NIH grants or other sources [33], [35]. In addition, drug trials are more likely than other intervention trials to explicitly exclude participants based on SES, gender (female), age, and comorbidity and multiple medications [36].
Although some racial differences in responses to pharmaceutical therapies have been found [37], [38], [39], race, by many accounts, is a poor proxy for genetic or biological difference [40], [41], [42], [43], [44]. There is substantial controversy associated with racial profiling for heath care [41], [45], [46], [47], [48], [49], [50], [51], with competing ideas of social and biological causes of differences in disease rates and response to treatment, and theories of multiple, interconnected health determinants [34], [42], [52], [53], [54]. While racial categories may not represent meaningful genetic differences, health outcomes vary by race and ethnicity, and self-identified race can capture living experience in ways that genetics cannot [53], [55], [56]. Poverty, income inequality, and SES are well-known factors in health inequities, and poor and socially disadvantaged populations, such as those served by Medicaid programs, have higher rates of morbidity, chronic disease, and excess death, [57], [58]. The mechanisms through which these factors affect health are complex and multidimensional, and may include suboptimal nutrition, socially mediated health behaviors, poor housing and bad neighborhoods, transportation issues, chronic stress, discriminatory prescribing and treatment patterns, and other discrimination [59], [60], [61], [62], [63], [64], [65], [66]. There is also evidence for a cumulative and negative effect of race and SES on health [43], [67], [68]. Moreover, emerging evidence from the field of epigenetics suggests that aside from functioning as barriers to good health, factors in the social and physical environment may also have direct effects on gene expression, affecting such diverse phenomena as maternal behavior, mental illness, and drug response [69], [70], [71]. Although it is widely acknowledged that social factors affect health outcomes, it is unknown whether or not they also affect patients' clinical responses to pharmaceutical therapies.
Section snippets
Methods
We examined racial/ethnic and gender differences among Medicaid enrollees in DERP member states and nationally. We evaluated DERP drug class reviews to assess the inclusion of African Americans, Hispanics, women, and low-SES populations in the pharmaceutical research that forms the basis of the reviews. We examined the concordance between Medicaid enrollee characteristics and populations included in DERP reviews. We evaluated DERP reviews and review protocols to determine whether DERPs
Results
The 13 DERP member states have a combined Medicaid/SCHIP enrollment of over 11 million, more than 25% of the U.S. total Medicaid population. Table 1 shows population and Medicaid enrollment by race and ethnicity in the United States and in the 13 DERP member states. The proportion of whites in the overall and Medicaid populations is generally somewhat greater in DERP states compared with in the United States. Nationally and in all DERP states, more whites receive Medicaid than any other group,
Discussion
Jonathan Mann, former director of the WHO global AIDS program and a life-long advocate for health equity and human rights, wrote that public health as a discipline has an ethical obligation to explore the impact of underlying social conditions on health and equitable access to health care [80]. We suggest that this may also be said of the impact of social conditions on who is included in and who benefits from research. Nationally, whites comprise the largest group of Medicaid recipients, but
Implications
The wide acceptance that social factors affect health gives rise to a question: Are there social determinants of response to treatment, including pharmaceutical therapies? Our findings suggest that there is not enough evidence available from drug trials to determine if social or biological factors are associated with differential responses to drugs in low-income and nonwhite populations. In its 2008 final report, the WHO Committee on Social Determinants of Health calls for research that both
Acknowledgments
The authors would like to thank Mark Gibson for comments and clarifications, Lisa Hirsch for editing and proofreading, and our colleagues at the Philip R. Lee Institute for Health Policy Studies writing seminar for review and suggestions. Funding for this research was provided by the Flight Attendant Medical Research Institute.
Other actions steps may be relevant for both models. For example, ensuring that drug formularies (Odierna & Bero, 2009) and other approval processes are reviewed to ensure that both innovative drugs and procedures that may disproportionately benefit men (e.g., treatment for testicular cancer) and those that may disproportionately benefit women (e.g., treatment for ovarian cancer) are available. Further, attention to an organizational culture that acknowledges and addresses gender dynamics in administrative and provider interactions with patients could help address existing gaps in patient experience.
Medicare beneficiaries annually select fee-for-service Medicare or a private Medicare insurance (managed care) plan; information about plan performance on quality measures can inform their decisions. Although there is drill-down information available regarding quality variation by race and ethnicity, there remains a dearth of evidence regarding the extent to which care varies by other key beneficiary characteristics, such as gender. We measured gender differences for six patient experience measures and how gender gaps differ across Medicare plans.
We used data from 300,979 respondents to the 2015–2016 Medicare Advantage Consumer Assessment of Healthcare Providers and Systems surveys. We fit case mix–adjusted linear mixed-effects models to estimate gender differences and evaluate heterogeneity in differences across health plans.
Nationally, women's experiences were better than men's (p < .05) by 1 percentage point on measures involving interactions with administrative staff (+1.6 percentage point for customer service) and timely access to care (+1.1 percentage point for getting care quickly), but worse on a measure that may involve negotiation with physicians (getting needed care). Gender gaps varied across plans, particularly for getting care quickly and getting needed care, where plan-level differences of up to 5 to 6 percentage points were observed.
Although the average national differences in patient experience by gender were generally small, gender gaps were larger in some health plans and for specific measures. This finding indicates opportunities for health plans with larger gender gaps to implement quality improvement efforts.
Sepsis, defined as a life-threatening organ dysfunction caused by a dysregulated host response to inflammation [6], is a heterogeneous syndrome affecting both females and males, yet the impact of sex on outcomes remains unclear [7,8]. Social and contextual interactions as well as biological factors may affect health outcomes for sepsis [9,10]. Until around the mid-1980s, women were broadly excluded as participants in biomedical research [11,12].
The objective of the study was to assess female representation in primary studies underpinning recommendations from clinical guidelines and systematic reviews for sepsis treatment in adults.
We conducted a bibliometric study. We removed studies pertaining to sex-specific diseases and included quasirandomized, randomized clinical trials (RCTs), and observational studies. We analyzed the female participation-to-prevalence ratio (PPR).
We included 277 studies published between 1973 and 2017. For the 246 studies for which sex data were available, the share of female participation was 40%. Females overall were under-represented relative to their share of the sepsis population (PPR 0.78). Disaggregated results were reported by sex in 57 studies. In univariate analyses, non–intensive care unit setting and consideration of other social health determinants were significantly associated with greater female participation (P < 0.001 and P = 0.023, respectively). In regression models, studies published in 1996 or later were likely to report sex, while RCTs were unlikely to do so (P = 0.019 and P < 0.001, respectively).
Our study points to female underenrollment in sepsis studies. Primary studies underpinning recommendations for sepsis have poorly reported their findings by sex.
In these cases, reviewers may consider including other study designs, knowing that these studies have an inherently higher risk of bias than RCTs [1–3]. Published reports of RCTs often do not contain detailed data on secondary outcomes or particular subgroups, despite the clinical relevance of such results [4,5]. Reasons given for the omission of this important data include limitations on space and length imposed by journals, that secondary outcomes or subgroup analyses are not planned a priori, and the lack of statistical power for subgroups or secondary outcomes in individual RCTs, resulting in insignificant findings that are not considered to be interesting or important [6].
Pooled-studies publications (PSPs) present statistical analyses of multiple randomized controlled trials without a systematic literature search or critical appraisal. We explored the characteristics of PSPs and their potential impact on a systematic review (SR).
We systematically evaluated PSPs excluded from an SR of second-generation antidepressants. We analyzed their basic characteristics, risk of bias, and the effect of new data on review conclusions.
We identified 57 PSPs containing a median of five trials (range, 2–11) and 1,233 patients (range, 117–2,919). Ninety-six percent of PSPs were industry funded, and 49% of PSPs contained unpublished data. The median number of citations for PSPs was 29 (range, 0–549). Only 7% planned pooling a priori, and 19% combined trials with identical protocols. Fifty-nine percent of PSPs eligible for general efficacy provided no new data. For some subgroups and accompanying symptoms (e.g., anxiety, insomnia, melancholia, fatigue, sex, and race), more than 30% of PSPs presented entirely new data or data that could alter the strength of the evidence available in the SR.
In this case study, PSPs provided new information on subgroups and secondary outcomes; however, guidance for reviewers and development of a system to assess their susceptibility to bias are required.