New York State of Health Pilot Yields Increased Race and Ethnicity Question Response Rates
Colin Planalp, State Health Access Data Assistance Center at the University of Minnesota
Race response rate grew 20 percentage points, ethnicity grew 8 percentage points
- New York set out to improve race and ethnicity response rates by piloting changes to the question on the Marketplace application.
- The state enhanced its explanation on the importance of the question for applicants and assistors, and it provided new training for assistors and navigators.
- Applicants did not have to share their race or ethnicity, but they could not leave the question blank; instead they could respond with “don’t know” or “choose not to answer”.
- Among participants in the test, race response rates increased 20 percentage points and ethnicity response rates increased 8 percentage points, while response rates for a comparison group saw minimal change.
- Based on the pilot findings, New York is expanding changes to the race and ethnicity questions system-wide for the next open enrollment period, and the state is considering additional revisions in hopes of further enhancing the quality and completeness of its data.
Even during the COVID-19 pandemic, states are striving to enhance health equity. In addition to racial justice movements that arose during 2020, the disproportionate impact of the pandemic itself on people of color highlighted existing health inequities in the United States. These factors have influenced many states to redouble their focus on closing health gaps. However, in order to identify priorities and evaluate improvement efforts, states need high quality and more complete data—a challenge when state health agencies’ data on race and ethnicity commonly contains large gaps. For instance, a recent report by the Centers for Medicare and Medicaid Services (CMS) found that 19 states’ race and ethnicity data was more than 20 percent incomplete, with some more than 50 percent incomplete.
Many states are looking to fill those gaps in race and ethnicity data for Medicaid and related agencies. This expert perspective highlights an effort by New York’s official state-based marketplace, NY State of Health, to improve the completeness of race and ethnicity data that applicants share when applying for Medicaid; Child Health Plus, the state’s Children’s Health Insurance Program (CHIP); the Essential Plan, New York’s Basic Health Program (BHP); or Qualified Health Plan (QHP) coverage through its Marketplace.
Setting Out to Improve Race and Ethnicity Response Rates
In the fall of 2020, staff in NY State of Health began a project to improve response rates in the collection of race and ethnicity data that people are asked to share during the application process. Historically, the agency had seen substantial gaps in those data, with roughly 40 percent of respondents skipping the question on race and 15 percent skipping the question on ethnicity. Because the NY State of Health serves as a single application point for all of the health insurance programs administered by the state, any issues with missing data affect all of those programs.
The state started with a pilot project to test strategies for improving race and ethnicity question response rates. By employing a smaller-scale pilot, the approach would allow them to test a “proof of concept” to determine whether their changes resulted in the intended improvements in question response rates before embarking on a larger effort that would apply to all new and existing program applicants. Demonstrating a proof of concept would also provide the state with results to bolster stakeholder engagement, which was particularly important in the New York context because the state has many health plans and assister organizations, which provide assistance to 80 percent of NY State of Health enrollees.
Test Strategies to Improve Question Response Rates
Working with the State Health and Value Strategies (SHVS) technical assistance team, New York tested multiple strategies aimed at encouraging applicants to answer the optional race and ethnicity questions.
Making the case to applicants
Under the Affordable Care Act (ACA), states are prohibited from requiring applicants to share information that is not necessary for determining eligibility for coverage, including race and ethnicity. However, best practices demonstrate that people may be more willing to voluntarily share that information if told how the data are intended to be used (e.g., responses to these questions will be used to inform health equity efforts by identifying health care gaps and enhancing outreach efforts), and how they will not be used (e.g., responses to these questions will not affect program eligibility, plan choices, or access to programs). Based on that understanding, New York deployed a revised script for the assistors and navigators participating in the pilot:
“Here at [agency name] we are testing a new approach for the next two questions which are on race and ethnicity. Obtaining this information can help us reach and possibly bridge healthcare gaps in traditionally underserved communities.
Please answer the following questions on race and ethnicity. We use this data to improve services to the community and to enhance outreach efforts. You do not have to answer these questions and giving us this information will not affect your eligibility, plan choices, or access to programs.”
While the message remained similar, and crucially still notified applicants that providing race and ethnicity information was optional, the revised introductory language made two key changes. First, it provided additional detail on how the information may be used, “to improve services to the community and to enhance outreach efforts,” rather than the original version that more generically states that “answering them can help us serve your community better.”
Second, the revised language flipped the order of the paragraph, starting with the request to share the information and a brief explanation of how the information may be used, followed by an acknowledgement that providing the information is optional. Conversely, the original introductory paragraph began by stating that “you do not have to answer any questions about race or ethnicity,” which could have discouraged applicants before they even learned why they might consider it.
Requiring a response
Although states are prohibited from requiring applicants from sharing race and ethnicity information, New York wanted to discourage people from rolling past the question with a quick skip. To achieve that, the state asked the organizations participating in the pilot to treat the question as if it required a response. However, applicants were given the option to respond with choices of “Don’t Know” and “Choose Not to Answer” to ensure they had the ability to opt out of sharing the information (Figure 1). That way, applicants could still decline to share the requested information, but the structure of requiring a response made answering the question just as easy as declining to answer it.
Figure 1. New York’s race and ethnicity pilot question
Training for assistors and navigators
In addition to piloting changes to its race and ethnicity question, New York also emphasized the importance of the data to the health plan and assistor organization that participated in the test pilot. The state provided training for those entities and their navigators and assistors. That training—including a computer-based presentation and written materials—described in detail how assistors and navigators should ask the race and ethnicity questions, including presenting a standardized script, and it explained the importance and purpose of those data: allowing the state to better understand who they are reaching with coverage and who is still being missed.
The state felt that approach was important because most people who obtain health insurance through NY State of Health do so with the help of an assistor (80 percent in 2020), rather than filling out the application entirely on their own. For that reason, navigators and assisters can serve as champions to help collect those data—or alternatively, they could discourage applicants from sharing that data if they do not understand the potential value it carries.
Evaluating the pilot
In addition to testing changes to how the state collects race and ethnicity data, New York also wanted to evaluate whether the pilot appeared to improve response rates. While it is difficult to definitively attribute any changes to the pilot—especially because it coincided with complicating factors related to the pandemic—the state found encouraging results.
In the health plan and assister organization that participated in the pilot, the response rates for race increased 20 percentage points, to 64.0 percent in the test period during 2021, compared to 44.0 percent during a similar period in 2020 (see Figure 2). The comparison group of other assistor organizations in the state, meanwhile, saw an increase of only 0.9 percentage points. The same health plan and assister group also saw the response rate for ethnicity increase 8.0 percentage points (from 80.3 percent to 88.3 percent), compared to a decline of 1.4 percentage points in the comparison group.
Figure 2. Race and ethnicity response rates, pilot test and comparison groups
Based on the successful results of its test, New York plans to implement the piloted changes on its application used by all people to enroll in and renew their coverage. Additionally, the state has continued to consider other revisions to the way in which it collects race and ethnicity data from applicants. The state has not yet implemented changes related to the following strategies, but it is considering making them in time for the next Marketplace open enrollment period.
Single race and ethnicity question
Research from the U.S. Census Bureau has found that a single, combined race and ethnicity question yields a greater response rate and improved accuracy as compared to two separate questions for race (e.g., American Indian/Alaska Native, Asian, White, Black/African American, etc.) and ethnicity (i.e., Hispanic/Latino) (see Figure 3). Historically, New York has combined the two questions into a single page of its enrollment system, but it is considering combining them into a single question that lists all response options (race and ethnicity) together with instructions to select all that apply.
Figure 3. U.S. Census Bureau-tested combined race and ethnicity question
Tailoring race and ethnicity response options
New York also is considering making revisions to the race and ethnicity options it lists as part of the state’s question, with two potential benefits. First, offering options that reflect how individuals identify may make them more likely to respond. Second, offering more granular and intuitive response options could allow the state to analyze the data with more specificity than the typical response categories allow. For instance, the state has a relatively large Middle Eastern and North African population, which typically is considered part of the White response category. However, that may not be intuitive to applicants and limits the state’s ability to analyze the data with concern to health inequities for that population. To facilitate its consideration of that option, the state worked with SHVS technical assistance providers to pull Census Bureau data on the size of various racial and ethnic populations within the state.
With a growing focus on health equity, high quality and complete data on beneficiaries’ race and ethnicity can help Medicaid and other state health programs to identify and mitigate gaps. This example from New York shows how states can test and evaluate strategies to improve the collection of race and ethnicity data to determine and build on approaches that work. And the state’s consideration of new strategies illustrates how improving these data is likely to require persistent and iterative efforts over time rather than any single, easy fix.
 Baker, D.W., et al. (2005). Patients’ attitudes toward health care providers collecting information about their race and ethnicity. J Gen Intern Med, 20(10), 895-900.