Incorporating health equity into value assessment: frameworks, promising alternatives, and future directions

Exploring methods that value diverse perspectives is critical to better understand the value of health equity in value assessment frameworks. In this paper, we examined emerging value assessment frameworks in the United States and present examples where evidence on outcomes and preferences for value do not take into consideration diverse perspectives. We identify possible solutions to improve existing value assessment methods and illustrate—using a hypothetical shared decision-making case study—an alternative to current value-assessment frameworks, “equitable multicriteria decision analysis,” which could be implemented in the context of the value-based assessment of prevention choices for women at high risk of developing breast cancer. These proposed alternatives and solutions can be used by researchers and decision makers to incorporate health equity into value assessment. In response to growing concerns about spiraling health care costs, the US health care system is undergoing a paradigm shift in the way it delivers and pays for care. 1 It is in this context that value assessment frameworks (VAFs) have emerged to assist different stakeholders, including clinicians and payers, in realigning health care decisions based on the most robust clinical and economic evidence. However, the use of these frameworks has sparked growing

concerns, of which health equity achievement has become central, triggering a debate over the meaning of value and its measurement. 2 Thus, there is a critical need to identify and assess avenues to incorporate health equity measures into VAFs.
In this paper, we examined current VAFs and document examples in which available evidence on outcomes and preferences traditionally accounted for in value assessments do not consider diverse perspectives. We document processes that may inadequately address health disparities and underserved populations. Next, we propose solutions to improve existing value assessment methods. Finally, we illustrate, using a hypothetical shared decision-making case study, how an alternative to current value-assessment frameworks, equitable multicriteria decision analysis, could be implemented as part of the value-based assessment of prevention choices for women at high risk of developing breast cancer.

Overview of Value Assessment Frameworks: A Comparative Analysis
The primary goal of VAFs is to support evidence-based health care decision making using robust clinical and health economic evidence. 1,3 Since the 1990s, the United States has fallen behind many high-income countries in relation to the benefit-cost of health care expenditures. 4 Lately, the United States has begun shifting from a volume-based system to one based on value. 3 Figure 1 shows the evolution of government and private investments in infrastructure supporting evaluation of costs and effectiveness of medicines and health care technologies in the United States.
In this context, several professional organizations such as the American Society of Clinical Oncology (ASCO), 5 American Heart Association/the American College of Cardiology, 6 National Comprehensive Cancer Network, 7 Memorial Sloan Kettering Cancer Center (MSKCC [DrugAbacus]) 8 and the Institute for Clinical and Economic Review (ICER) 9 have developed their VAFs. In this section, we examine these US VAFs according to their target audience, rationale, elements of value considered, methodological approach, and other considerations including equity attainment. Table 1 presents a summary of this comparative analysis. An initial examination of Table 1 reveals that VAFs have different rationales and perspectives. Elements of value considered in these frameworks differ based on perspectives, target audience (clinical/shared decision making versus coverage and policy), and expected outcomes, either direct or indirect.
Clinical evidence used in these frameworks is generated from population-level sources including randomized controlled trials (RCTs). As an example, the ASCO framework solely depends on RCTs to generate the net health benefit (NHB) scores and does not reflect the real-world clinical practice. Clinical trials represent the cornerstone of evidence generation when it comes to establishing treatment efficacy. Yet, it has been reported that only 5% and 1% of trial participants are Black/African American and Hispanic/Latino, despite making up 13.4% 10 and 18.1% 11 of the US population, respectively. Knowing that participation of racial/ethnic minority groups in clinical trials is low, such evidence is not tailored to meet the needs of minority patients/communities, or underrepresented subgroups of the population including the young and the elderly. 12 Except for the ICER framework, VAFs completely ignore patient heterogeneity (eg, age, sex, race/ethnicity, socioeconomic status), complex demographic factors (eg, geography), and health care system-level factors (eg, access to care) that are known contributors to health care disparity. This potentially leads to inefficient allocation of health care resources, coverage, and drug pricing decision making, 13 with a huge burden on minority patients as they may not be provided with the optimal clinical services they deserve.
Similarly, equity measures (eg, unmet needs, health disparity, burden of illness) are not explicitly and quantitatively incorporated in value assessment as evidenced in Table 1. Only the MSKCC considers unmet needs and burden of illness quantitatively, whereas ICER uses an ad hoc process to include health disparity contextually, if judged feasible. Finally, the VAFs discussed in this section have previously been criticized for their limited transparency in their data aggregation processes, their ability to be reproduced, 14 and the lack of decision criteria that matter the most to patients. 1

Solutions to Improve Existing Value Assessment Frameworks
In recent years, the need to curtail the spiraling of health care costs coupled with the imperative to deliver value-based care has legitimated the emergence of VAFs in the United States. With the progressive adoption of these VAFs, major limitations to VAFs have become apparent. In this section, we propose a range of solutions to address the limitations of existing VAFs, starting from the generation of appropriate elements of value, equity data specifically, to the adoption of robust data aggregation methods.
A call for action to make recruitment and inclusion of underrepresented minority groups a priority in clinical trial implementation is critical as it will ensure that RCTs contribute to the development of interventions that are responsive to the needs of a diverse population. 15,16 Evidence obtained from these trials can then feed VAFs. Pharmaceutical companies could be incentivized to develop drugs for rare diseases and diseases that affect underrepresented minorities the most such as sickle cell disease. Including the patient voice in the clinical drug development to identify endpoints and achieve outcomes that are important to them will expand elements of value used in VAFs.
Criticized for their lack of patient centeredness while being agnostic of health equity considerations, emerging VAFs need to build on existing methodological approaches that allow for the explicit incorporation of the health inequality impacts of the health care interventions evaluated. 17 We briefly present 3 approaches, 2 of which build on traditional cost-effectiveness analyses. These approaches are (1) 2-part health technology appraisal, (2) distributional cost-effectiveness analysis (DCEA), and (3) equitable multi-criteria decision analysis.

TWO-PART HEALTH TECHNOLOGY APPRAISAL
The goal of the 2-part health technology appraisal is to ensure patient access to health care interventions that generate extensive value to society, which is not captured by the incremental cost-effectiveness ratios traditionally used in economic evaluations. 18 Specifically, the incremental cost-effectiveness ratio is used jointly with a comprehensive benefits and value (CBV) score for decision making about the value of health care interventions. The CBV score is a composite and qualitative score obtained from the aggregation performance of the interventions evaluated against expanded elements of value including innovativeness, disease severity, and unmet need. 18 This approach was initially proposed as an alternative process to the National Institute for Health and Care Excellence health technology assessment.

DISTRIBUTIONAL COST-EFFECTIVENESS ANALYSIS
The term "DCEA" serves as an umbrella for economic evaluations that model distributions of health (health gains/disease burden) associated with health care interventions at both population (societal) and subgroup (eg, sex, race/ethnicity) levels. [19][20][21] In other words, DCEA provides breakdowns of health gains and losses per equity-relevant sociodemographic variables and per disease categories. In making decisions about the value of health care technologies, decision makers make tradeoffs between improving total population health and reducing unfair health inequality. [19][20][21]

EQUITABLE MULTICRITERIA DECISION ANALYSIS
Value assessments of health care interventions are complex in nature as they involve multiple criteria of varying importance to decision makers and data from diverse sources. This conundrum calls for the development of more structured frameworks that allow for the transparent aggregation of value elements into a decision composite metric. One of the most comprehensive quantitative approaches to choose, rank, and select treatment options is multicriteria decision analysis (MCDA). 22,23 The explicit consideration of criteria importance and the transparency inherent in the use of this approach ensure that the underlying process of arriving at any decision based on treatment profiles and decision criteria is consistent and clear. In this context, equitable multicriteria decision analysis can help support decision makers faced with evaluating treatment options by considering multiple criteria in an explicit manner, among which is the treatment impact on health equality. 24,25 In the next section, we develop a hypothetical case study illustrating the application of equitable multicriteria decision analysis to the value-based assessment of prevention choices for women at high risk of developing breast cancer.

CASE STUDY PRESENTATION
Imagine that Jane Doe, a Hispanic woman at high risk of developing breast cancer, has a clinical encounter with her care provider. She is offered a choice among 4 chemoprevention alternatives: low-dose tamoxifen, tamoxifen, exemestane, and raloxifene. Treatment recommendation in this shared decision-making process is based on the performance of the treatment options (ie, how well each treatment option meets prespecified decision criteria) against 4 criteria that reflect the following dimensions of value: cost (out-of-pocket costs), safety (drug safety profile), clinical benefit (quantification of cancer risk reduction), and equity (treatment impact on health equity). These criteria are defined as follows: • Out-of-pocket costs: Monthly co-pay expressed in dollars. This criterion is measured on a quantitative scale and needs to be minimized (ie, the lower the out-of-pocket costs, the better).
• Safety profile: This criterion is measured on a quantitative scale from 1 to 5, 5 being the best. This criterion needs to be maximized (ie, the higher the score, the better).
• Quantification of cancer risk reduction: This criterion is measured on a quantitative scale from 0 to 100 (percent), 100 being the best. This criterion needs to be maximized (ie, the higher the score, the better).
• Treatment impact on health equity: In this hypothetical case, the treatment impact on health equity is captured through its ability to reduce the differences in terms of cancer risk reduction among racial and ethnic groups. ) and is expected to be maximized (ie, VL is the least preferred, whereas VH is the most preferred). Table 2 summarizes the performance of each treatment option against the 4 decision criteria.
It is worth noting that the data contained in this table are used for illustrative purposes. In a real-world setting, this data would be obtained from clinical trials, comparative effectiveness studies, relevant cost data sources, and experts as needed.
As part of the shared decision-making process, the provider has quantified Ms. Doe's preferences using an approach called ELICIT, 27 which yielded the following weights: 0.28, 0.22 0.35, and 0.15 for out-of-pocket costs, safety, quantification of cancer risk reduction, and treatment impact on health equity, respectively. These weights indicate the contribution of each criterion to the decision about the best chemoprevention agent. For example, the criterion safety makes up 22% of the decision while OPC makes up 28%. The weights sum to 100%.

METHODS: VALUE ASSESSMENT USING EQUITABLE MULTICRITERIA DECISION ANALYSIS
We use partial or noncompensatory MCDA methods for transparent aggregation of elements of value, notably outranking models. These models establish dominance relations among alternatives such that even if an alternative A (treatment option in our case) is as good as an alternative B on every decision criterion, a decision-maker (provider/patient) can conclude that A outranks B if a majority of the decision criteria support this assertion as long as there is no criterion on which A is worse than B. The majority rule is fulfilled when the sum of the preference weights associated to these criteria are superior or equal to 55%. This is the basic principle of the Elimination and Choice Expressing Reality (ELECTRE) methods. [28][29][30] In ELECTRE methods, alternatives are compared pairwise to establish outranking relationships by considering the weights of the criteria in favor of the outranking relation and also the possibility that an opposing criterion vetoes ("veto threshold") that relation. Outranking relations are analyzed using an ELECTRE method to select the best alternative, rank the alternatives, or sort them into predefined categories. The issue of ELECTRE methods not yielding a single winner or best choice is circumvented by using the Copeland's pairwise aggregation method 31 where alternatives, treatment options in this case, are ranked by the number of pairwise "victories" (ie, number of times an alternative is dominant over others) minus the number of pairwise "defeats" (ie, number of times an alternative is dominated by others). The higher the Copeland's score, the better the rank. Deterministic and probabilistic sensitivity analyses can be conducted to address the impact of uncertainty of the analysis results using "equitable multicriteria decision analysis." Figure 2 shows the relationship among the treatment options, which translate into the ranking of these treatments by order of preference (Table 3). Each arrow represents a dominance relationship between 2 treatment options (Ax versus Ay). An arrow departing from an alternative/treatment (Ax) would suggest that Ax is dominant over Ay (Victory for Ax) while an arrow pointing towards Ax would suggest that Ax is dominated by Ay (Defeat for Ax). For example, looking at the pairwise comparison between A1-low dose tamoxifen and A2-tamoxifen in Table 1, A1 has a performance that is at least as good as A2 on all the criteria. The majority rule is fulfilled given that the sum of the weights of the criteria for which A1 is at least as good as A2 is superior to 55% (Majority weight = 0.28 + 0.22 + 0.35 + 0.15). Thus, A1 has 2 victories and 0 defeat. These dominance relationships are established for all pairwise treatment comparisons. Using equitable multicriteria decision analysis, which accounted for both patient preferences and included an equity criterion in the decision process, low-dose tamoxifen has the highest Copeland score and therefore ranked first. As a result, low-dose tamoxifen is the treatment returning the most value.

Conclusions
Recent societal and cultural movements have helped place a spotlight on inequities that for far too long have not been addressed in our health care system. As a case in point, these inequities can be found where underrepresented communities are underserved, including VAFs. Thus, the solution is to focus resources on these communities using decision toolkits that are already available but can be adapted to ensure decision makers appropriately measure equity-efficiency tradeoffs. Innovations for improved health outcomes for marginalized groups should be appropriately incentivized to ensure the attainment of health equity. While concerns remain about VAFs, the solutions to their expansion are attainable and include fostering inclusion and diversity in clinical trials and translational research, use of appropriate value aggregation methods that can capture equity, and input from the patients on what endpoints/criteria matters most.   c Difference between "victories" and "defeats."