Survey Shows Racial, Ethnic Disparities in Social Health 1 Year Into COVID-19 Pandemic

In this interview, Dr Nancy Gordon discusses findings from a recent survey conducted in an integrated delivery network, which aimed to shed light on demographic differences in the prevalence of social needs and risks 1 year into the pandemic. Dr Gordon also advises how health plans can improve their approach to social health screening.

What inspired your research into social needs in US adults during the pandemic?

I have been involved with the collection and analysis of information about the social health of Kaiser Permanente members since 2014. This work includes identifying questionnaire items captured in Kaiser Permanente’s electronic health record (EHR) system that can be used to estimate prevalence of different social risks and social needs.

For this specific survey, my aim was to collect baseline data on the prevalence of various social risks and social needs that Kaiser Permanente Northern California’s middle-aged and older adult health plan members identified as areas in which they required assistance. I also wanted to study whether the narrow focus of the current Epic social determinants of health (SDoH) risk screening questions—many of which are also used in the Accountable Health Communities screener developed by the US Centers for Medicare & Medicaid Services (CMS) Center for Medicare and Medicaid Innovation—could potentially under-identify adults experiencing difficulties. 

We focused on the segment of members not covered by Medicaid because social health screening of Medicaid-covered members has already been conducted routinely for several years. Additionally, because of their low income, Medicaid-covered adults with social needs were eligible for many existing community-based social and financial services like the Supplemental Nutrition Assistance Program (aka SNAP) and housing and transportation vouchers. These groups are studied quite often. 

Much less is known about social health concerns of the adults who are not in these known vulnerable groups and who comprise the largest segment of the adult population served by health plans. This includes how receptive they would be to being asked by their health care providers about social risk/social needs such as financial strain and food, housing, and transportation insecurity. We also wanted to learn whether there were substantive differences between demographic subgroups (ie middle-aged vs older adults, racial/ethnic groups, different genders within these groups) in the prevalence of social risks, the desire for help with social needs, and the receptivity to being screened. We believed this information was important to understand the potential impact and accuracy of social risk screening programs that will be implemented in sociodemographically diverse populations like ours, as well as in targeted populations, like adults covered by Medicare.  

How was your survey designed, and who responded to it?

We wanted to estimate social risks based on questions that went beyond Epic’s current screener. For this reason, we drew survey items from both Epic and several other social risk/social needs screening tools, as well as from my previous member health surveys, our first Kaiser Permanente social risk screener, other research studies, and government-sponsored national surveys. We decided to ask questions about social risks using a 3-month lookback window rather than the 12-month lookback used in the Epic SDoH screener because of the ongoing and continually changing impact of COVID-19. We also captured information about social connection, social support, and loneliness, which we intend to address in a separate paper. The survey was designed to be completed in under 10 minutes by most people.

We selected a stratified (by age group, sex, and race/ethnicity) random sample of 10,000 members aged 35 to 90 years who had been with Kaiser Permanente for at least 2 years. We limited the survey to people who had English listed as their preferred spoken and written language in their EHR, which is a limitation of our study, but necessary due to budget constraints. We mailed a print questionnaire and emailed a link to an online version of the questionnaire up to 2 times. We also provided contact information for resources that provide assistance with social, financial, or emotional difficulties. As an incentive, we explained that 100 survey respondents would be randomly selected to receive a $100 Target or Amazon.com gift card. Approximately 29% of members we reached out to responded. In our analyses, we assigned a survey weighting factor to each respondent so our results would be based on a study population that better reflected the age (35 to 64 years and 65 to 90 years), sex (male and female), and racial/ethnic (White, Black, Latinx, Asian and Pacific Islander) composition of Kaiser Permanente Northern California’s adult membership.

What are some key findings from the survey? Did any of the results stand out or surprise you?

The response to the survey was much lower than we expected based on response to a general member health survey I conducted the previous year. I interpreted this as an indication that future large-scale social health screening efforts will likely yield similarly low response rates because there will be people who consider the questions too sensitive or not relevant to them. In many general surveys, questions about income have a much higher percentages of missing data than other sociodemographic and health items. The same could be true for our survey. This tells us we need to be aware that while it may seem proactive and helpful to ask people about their social risks, there may be many people who do not want to answer these questions, especially if the information will be a permanent part of their EHR. 

We were not surprised to find that a higher percentage of Black and Latinx adults had recently experienced strains related to food, housing, and overall financial security, compared to White adults. This has been seen in other surveys. However, our survey results suggested pandemic-related financial strain impacted larger percentages of middle-aged adults vs older adults, both overall and within vulnerable racial and ethnic groups. This was likely due to the sudden reduced employment income, increased cost of living expenses, and mounting credit debt faced by working age adults during the onset of the COVID-19 pandemic. We also found that, compared to older adults, a higher percentage of middle-aged adults often felt lonely or socially isolated and that they were not getting enough social and emotional support.

It also was interesting to see that the needs people most frequently indicated wanting help with were not food, housing, or transportation—the domains most universally screened for—but dental care; vision care; health-related expenses; debt, loan, or credit card repayment; and utilities expenses. We only saw modest overlap between identified social risks (eg food and housing insecurity) and interest in getting help with needs related to those risks. 

Our study confirmed our suspicion that the current narrowly focused questions used to screen for food insecurity, housing instability, and financial strain under-identified—and thus also underestimated—the prevalence of these social risks in the population. For example, including a question about use of free community food resources, like food pantries, along with the widely used 2-item Hunger Vital Sign (HVS) screener doubled the percentage of adults identified as food insecure, compared to using the HVS questions alone. Similarly, when questions that addressed concerns about paying for or losing housing were added to questions about recent problems paying for housing, we significantly increased our estimates of adults who were housing-insecure, especially among middle-aged Black and Latinx adults. And when we included questions about decreased or foregone use of medical care, prescription medicines, and dental care, we saw significantly increased percentages of people who were experiencing financial strain compared to using single questions that only addressed general levels of financial strain.

Finally, with regard to acceptability of social health screening and screening modalities, we found that while 70% of adults said they would be okay with being screened for social risks and needs in a clinic setting, nearly 20% would not want to be screened, with variations across demographic subgroups. Among those who were okay with being screened, most preferred that screening be conducted using a questionnaire that they completed themselves, rather than being asked these questions by health care staff.

How can these data inform future efforts to address SDoH within integrated delivery networks?

Our study suggests that solely screening for social risks without asking a separate question about assistance wanted for parallel social needs increases the likelihood that large percentages of adults who have social needs may not be identified. Kaiser Permanente’s Brief Social Health Screener, which is currently being rolled out across Kaiser Permanente regions for universal screening (all ages and settings), addresses this by including a question to ascertain social needs for which people want assistance in addition to the food, housing, transportation, and overall financial strain domain questions in the Epic SDoH screener. 

We also found that more adults wanted help with health care-related needs like dental, vision, and paying for medical expenses than help with food, housing, and transportation. This suggests health care-related domains should be included when patients are asked about their concerns for which they want assistance.

Importantly, our lower-than-expected survey response, along with our finding that 70% of adults were okay and 20% were definitely not okay with being screened for social health concerns, suggest that social health screening will not necessarily be embraced by all patients, including those who might benefit from screening. Health plans that want to implement large-scale outreach screening efforts, especially using patient portal questionnaires, will need to do extensive pretesting with diverse segments of their patient populations to develop communications that encourage patients to complete screening questionnaires. It will be important to communicate to patients that even if they do not want help with social needs, information about social risks can inform clinician recommendations, treatment protocols, and modalities of care delivery.

Is there anything else you would like to add?

Currently, social risk questionnaires ask people to reflect on their circumstances during the past 12 months. Responses to questions that use such a long lookback timeframe may not accurately reflect a member or patient’s current life situation or near future needs.

First, people who temporarily experienced a social risk, as occurred with many people during the early months of the pandemic, but do not still have that risk, may not want that risk to be noted in their EHR as it no longer applies.

Second, in the future, health plans may be judged by CMS, the National Committee for Quality Assurance, and other health care regulators and accreditation bodies not only on the percentage of patients who are screened for social risks, but also on the percentage of those patients who screened positive for a risk and received adequate follow-up for their risk. If a risk identified using a 12-month lookback is noted in the EHR, health care organizations with limited social services staff may still be held accountable for showing that they provided follow-up assessment and referral to patients who did not have that risk at the time of the screening or did not want help with that risk even if it was still present. It would be more efficient to require EHR systems to include ways for patients or health care staff to indicate whether a patient has a current risk that requires follow-up, and whether the patient wants help. Having a way to easily document that a past risk is not current is essential to show a patient’s situation has changed, potentially in response to follow-up actions taken by the health plan staff.  

Third, past problems might not reflect a patient’s current or near-future situation. For example, a patient who drives may not have had transportation problems in the past but may now need transportation assistance due to a leg injury or use of medication that should not be used while driving. An older patient with a diet-related health condition may not screen positive for food insecurity but may need assistance obtaining healthy meals from meal-delivery organizations like Meals on Wheels or may need to be shown how to use online grocery store delivery services or meal delivery services like Doordash. 

It is not enough to say we screen people for social risks. We must implement social risk screeners that are flexible, relevant, and accessible for patients and staff to document both past and current patient risks as well as help offered, received, or refused.

About Dr Gordon

Nancy Gordon, ScD, is a research scientist at Kaiser Permanente Northern California’s Division of Research. She has also worked as a consultant to the Kaiser Permanente National Program Office, assisting with the development and deployment of social risk and social needs screening questionnaires and surveys.