Understanding the impact of unmet social needs on consumer health and healthcare

Articles

Income, employment, education, food security, housing, transportation, safety, and social support are all factors that affect health and well-being.

Healthcare industry stakeholders increasingly recognize that in order to deliver improved health outcomes, it is necessary to address the social factors that influence consumer health. Findings from McKinsey’s 2019 Consumer Health Insights and Consumer Social Determinants of Health surveys demonstrate that social needs are a salient issue for all types of consumers, and that they are also linked with consumer preferences and perceptions of the care they receive. This paper sheds light on differences in preferences, perceptions, health, and utilization across consumers with varying degrees of unmet social need. By better understanding consumers’ perspectives regarding their own health and social needs, healthcare stakeholders may improve programs that optimize utilization, improve outcomes, and enhance consumer experience.

Insight 1:
Unmet social needs exist across all types of consumers

Income, employment, education, food security, housing, transportation, safety, and social support are all factors that affect health and well-being. Additionally, trauma or adverse childhood experiences may have long-term effects on health and well-being.

We surveyed consumers with employer-sponsored group insurance, individual market coverage, Medicare coverage, Medicaid coverage, and those who were uninsured. We focused on understanding six social needs: food security, adequate housing, reliable transportation, social support, community safety, and personal safety.1 Although a larger proportion of Medicaid respondents reported having unmet social needs than respondents with other types of insurance coverage (likely due to the linkage with socioeconomic status), more than 35 percent of respondents in each line of business reported experiencing at least one unmet social need, and nearly a quarter of all respondents experienced two or more (Exhibit 1). This finding suggests that there are subsets of consumers in each line of business who could benefit from support to address their unmet social needs.

Furthermore, healthcare stakeholders could consider social needs not only by line of business but also by health condition. For instance, within Medicaid, respondents who reported poor mental health were more likely to report multiple unmet social needs as compared to those who reported good mental health.2 Understanding where unmet social needs may exist across types of insurance coverage and health conditions may help to focus efforts to address unmet social needs for different types of consumers.

Sidebar 1

Definitions of unmet social need and social risk score

This paper refers to an individual’s social context in two ways:

  • Unmet social need: An unmet social need is defined as the lack of a basic resource such as food, safe housing, or transportation. An unmet social need may or may not have an adverse effect on a person’s health. For example, an individual may not have reliable transportation, but that lack of transportation may or may not impact their existing health conditions or ability to receive needed healthcare. In this case, we refer to the lack of transportation as an unmet social need, as the linkage to health is not known for this individual.
  • Social risk score: Analytics can help to inform the relationship between a person’s social context and their health. Using a predictive model, we analyzed the relative importance of six unmet social needs in predicting a survey respondent’s health status. We then created a composite metric that weighted each unmet social need by its relative importance in predicting health status, thereby quantifying the linkage between unmet social needs and health. We defined this metric as an individual’s social risk score, where higher scores indicate not only the presence of a greater number of unmet social needs, but also the degree to which those unmet social needs may impact health within the surveyed population.1 See technical appendix for additional detail on methodology.

1We defined this metric as a social risk score as it aligns with the definition of social risk factors as adverse social conditions associated with poor health, such as food insecurity and housing instability. Alderwick H and Gottlieb LM, “Meanings and misunderstandings: a social determinants of health lexicon for health care systems,” Milbank Quarterly, June 2019, 97(2):407-19; Health Affairs Blog, “When talking about social determinants, precision matters,” blog entry by Green K and Zook M, October 29, 2019, healthaffairs.org.

Insight 2:
Consumers with higher utilization are more likely to report unmet social needs

The survey results identify a link between healthcare services utilization and social needs. For example, respondents with higher inpatient or emergency room (ER) utilization are more likely to report unmet social needs (Exhibit 2).3

Importantly, the relationship between social needs and healthcare utilization varies by population. For example, an unmet transportation need may have a different impact on the utilization of a healthy Medicaid consumer versus a Medicare consumer with poor health. Advanced analytics can help to demonstrate the relative impact of each unmet social need in different populations. For instance, we built analytical models to determine which social needs were most predictive of health status4 and utilization in different populations, indicating that they may play a larger role in driving overall health and healthcare utilization (Exhibit 3, see technical appendix for additional detail on methodology):

  • Among Medicare respondents, housing and transportation were relatively more important than other social needs in predicting health status.
  • Among Medicare respondents with poor health status, community safety was relatively more important in predicting inpatient or ER utilization.
  • Among Medicaid respondents and those who are likely to cycle in and out of the Medicaid program (referred to here as the “Medicaid churn population”),5 food and personal safety were relatively more important in predicting health status.
  • Among Medicaid and Medicaid churn respondents with poor health status, community safety was relatively more important in predicting inpatient utilization but not ER utilization.

These findings provide empirical support for what may seem intuitive—unmet social needs and factors associated with them may impact health and utilization differently across populations. Thus, in developing interventions, clearly defining objectives and tailoring an approach to specific populations of interest are critical to achieving the desired outcomes. Put simply, a program that targets Medicare consumers with poor health status may need to address different social factors than one that addresses the broader Medicare population. Similarly, an initiative that aims to optimize ER utilization may need to address different social factors for Medicare consumers than for Medicaid consumers. Considering the degree to which different social needs may impact different types of utilization can help stakeholders prioritize which unmet social needs to address in each target population and ensure that local social service providers have enough capacity to address these needs.

Insight 3:
Consumers with significant social risk often have decreased access to care and lower satisfaction with care when received

We analyzed access and satisfaction among respondents with Medicare coverage, Medicaid coverage, the Medicaid churn population, and those with dual Medicare and Medicaid coverage. We created a social risk score for each respondent in this group by weighting the relative importance of each social need in predicting health status, and segmented the population into those with limited, moderate, and significant social risk.6

Thirty-seven percent of respondents with significant social risk reported not getting all the healthcare they needed. Among respondents who reported at least one healthcare encounter in the previous twelve months, only 52 percent of respondents with significant social risk were satisfied with their overall healthcare experience, compared to 81 percent of respondents with limited social risk (Exhibit 4). Though respondents with significant social risk were more likely to report not getting all needed healthcare, eight of the 10 most common reasons for not getting all needed care were not related to specific social factors. Instead, they were related to affordability, healthcare access, health literacy, and low expectations of the healthcare system. This suggests that addressing unmet social needs may be necessary but not sufficient to remove barriers to receiving healthcare for consumers with significant social risk. Instead, healthcare stakeholders could consider addressing specific unmet social needs, such as transportation, in conjunction with other levers to expand access, improve satisfaction, and encourage appropriate use of healthcare services.

This is especially relevant for Medicaid and Medicaid churn populations, for whom the gap between those with limited social risk and those with significant social risk is smaller. Twenty-four percent of respondents with limited social risk also reported they did not get all the care they need, and only 64 percent of those who reported at least one healthcare encounter were satisfied with their overall healthcare experience. While social risk is important to assess, the similarity between those with limited and significant social risk in this population indicates that there are likely additional barriers not directly related to food, transportation, housing, social support, or safety for this population (such as economic status, healthcare access, and health literacy).

Insight 4:
Consumers with significant social risk may prefer alternative sites of care and support

Differences in care preferences reveal opportunities to improve care delivery for populations with significant social risk. Focusing again on the Medicare, Medicaid, and Medicaid churn populations, respondents with significant social risk are more likely to prefer the ER and alternative sites of care for routine and ongoing care (e.g., urgent care, telemedicine, pharmacy) (Exhibit 5). Eighty-one percent of respondents with significant social risk want support in managing their health and may seek support from a variety of sources—for example, 58 percent want support from a primary care provider, 27 percent from family or friends, and 21 percent from a therapist. In comparison, similar percentages of those with limited social risk said they would want the support of a primary care provider (PCP) or family or friends, though only 2 percent said they would want the support of a therapist.

There is also an opportunity to engage respondents with significant social risk digitally—86 percent have access to a smartphone and 83 percent were open to digital health solutions. However, even when interested respondents had access to programs such as extended hours at drop-in care clinics and digital health tools, 30 percent did not use these services. Respondents with significant social risk were less likely to find digital tools such as medication reminders useful after trying them, despite their original interest.

However, even when interested respondents had access to programs such as extended hours at drop-in care clinics and digital health tools, 30 percent did not use these services. Respondents with significant social risk were less likely to find digital tools such as medication reminders useful after trying them, despite their original interest.

Together, these findings suggest a need to engage individuals, communities, and caregivers in creating solutions that reflect individual needs and local preferences. Equipping alternative sites of care to address social needs as well as physical health could help to improve overall outcomes. The disconnect between program interest, uptake, and continued use suggests misalignment between what users desire and what each solution ultimately delivers. Ensuring that programs and digital solutions are designed and tested with individuals with unmet social needs may help to meet these consumers’ needs and expectations, and ultimately create opportunities to further scale support for these populations. Importantly, because local context often influences social needs, successful solutions may also require flexibility to respond differently to communities across geographies.

* * *

Recognizing that individuals with unmet social needs exist across all types of health insurance coverage, a nuanced understanding of consumers’ social needs and preferences can enable organizations to move away from broad social interventions to targeted, effective solutions.

Healthcare organizations could consider investing in quantitative and qualitative methods to understand the impact of unmet social needs on the health and healthcare preferences of their consumers and gain insight into which unmet social needs most influence their healthcare utilization and outcomes. For example, providers could tailor a patient’s care plan and follow-up to the patient’s social context, partnering with local community organizations to provide social services where needed. Payers could incorporate data about social needs into existing analytics and systems to better equip care managers and customer care teams in supporting members, and to make informed investments in expanding the capacity of local social service providers. Creating targeted solutions and interventions with input from consumers with varied unmet social needs may yield higher engagement, improved effectiveness, and better consumer outcomes.

Sidebar 2

Technical appendix

Survey methodology

This paper draws upon findings from two surveys, McKinsey’s 2019 Consumer Social Determinants of Health survey and McKinsey’s 2019 Consumer Health Insights survey.

  • 2019 Consumer Social Determinants of Health survey: Survey was fielded in December 2018. Respondents included US residents between the ages of 18–84 and were uninsured or had health insurance coverage through Medicare, Medicaid, or an individual market plan. Respondents who had individual market plans or were uninsured were limited to those with household incomes at or below 250 percent of the federal poverty level. Individuals with employer-sponsored health insurance were excluded. Results are shown for a nationally representative weighted subset of 2,010 respondents.
  • 2019 Consumer Health Insights survey: Survey was fielded in May 2019. Respondents included 4,958 US residents between the ages of 18–84 who were involved in health insurance decisions for their household. Respondents were uninsured or had health insurance coverage through their employer, Medicare, Medicaid, or an individual market plan. Results were weighted to be nationally representative.

In both surveys, participants were asked questions regarding health status, utilization, preferences, and experiences, as well as questions across the following categories of social needs: income, family size, employment, education, food security, community safety, personal safety, housing, transportation, and social support.

Definition of social needs

Impact of food security, community safety, personal safety, housing security, social support, and transportation access were defined per the methodology below. Respondents reporting an unmet social need were defined as those who were considered to be adversely impacted by the social factor.

  • Food environment: Considered to be adversely impacted by food security if they have “very low food security” or “low food security,” using questions and methodology from the United States Department of Agriculture’s (USDA) screening tool to assess household food security. Methodology uses US Household Food Security Survey Questions, USDA.
  • Community safety: Considered to be adversely impacted by community safety if they reported low-scoring responses to questions about general feelings about their community (e.g., whether people in the community take care of each other), concerns about community crime (e.g., worries about theft or assaults), and perceptions of community problems (e.g., presence of litter, abandoned buildings, noise violations).
  • Personal safety: Considered to be adversely impacted by personal safety if they reported needing medical care due to physical violence or indicated being afraid of a partner or co-resident.
  • Housing: Considered to be adversely impacted by housing if they indicated that they do not have housing or have housing but are worried about losing it.
  • Social support: Considered to be adversely impacted by social support if they reported low-scoring responses to questions about emotional and social loneliness, and about availability of emotional and physical social support (e.g., sources of emotional strength and help with practical activities).
  • Transportation: Considered to be adversely impacted by transportation if they indicated that transportation has kept them from work, meetings, or getting things needed for daily living, or that it takes 60 minutes or longer to get to a doctor’s appointment.

Impact of socioeconomic status, employment, and education were not analyzed in this research:

  • Socioeconomic status: Analyses were performed across and within different types of healthcare coverage, which can control for income to some degree. Additionally, the 2019 Consumer Social Determinants of Health survey was fielded to respondents with household incomes below 250 percent of the federal poverty level for certain types of healthcare coverage in order to control for income levels. Additional factors related to socioeconomic status were not analyzed separately.
  • Employment: Analyses were performed across and within different types of healthcare coverage, which can control for employment to some degree. Additional factors related to employment status or type of employment were not analyzed separately.
  • Education: The majority of respondents attained a high school education. Due to small sample sizes of respondents who did not attain a high school education, analyses related to educational attainment were not conducted.

Health status

A proprietary algorithm was used to calculate a composite measure of health status for each respondent, incorporating answers to questions about number of chronic conditions and number of ER visits, inpatient stays, and visits with a PCP or general practitioner in the past year.

Relative importance of social needs

Weighted Random Forest predictive models were built to understand the relative importance of each of the six defined social needs for predicting health status of respondents in the 2019 McKinsey Consumer Social Determinants of Health Survey. These models were used instead of univariate correlations as they were able to capture the interactions between variables and with the outcome, whereas a univariate analysis would have only considered the relationship between one variable and the outcome. The models were weighted to compensate for the imbalance in the outcome variable classes. Additionally, the weighted Random Forest method was selected over other machine learning methods because it provided a measure of feature importance with regard to the outcome. Three unique models were created (one for all survey respondents, one for the Medicare population, and one for the Medicaid and Medicaid churn population).

Additional weighted Random Forest predictive models were built to understand the relative importance of each unmet social need in predicting ER and inpatient utilization among respondents with poor health status. Three unique models were created to predict ER utilization (one for all survey respondents, one for the Medicare population, and one for the Medicaid and Medicaid churn population) and three unique models were created to predict inpatient utilization (one for all survey respondents, one for the Medicare population, and one for the Medicaid and Medicaid churn population).

Social risk score

A weighted Random Forest predictive model was built to understand the relative importance of each of the six defined social needs for predicting the health status of all survey respondents in the 2019 McKinsey Consumer Social Determinants of Health Survey. The relative importance of each social need was used to create weighted values for each social need, which were then summed to create a composite social risk score that ranges from 0 to 1. Using the social risk score, survey respondents were categorized into three groups: limited social risk, moderate social risk, and significant social risk.

By accounting for the relative importance of each social need in predicting a respondent’s health status, the social risk score places more emphasis on the social needs that may impact health. In this way, the social risk score provides a more nuanced measure of the social needs that may impact downstream healthcare utilization.

The authors would like to thank Glen Graves, Eric Bochtler, Abigail Charles, Violet Dang, Shriram Bhutada, and Himani Kohli for their contributions to this paper.

  1. Because analyses were conducted across and within different types of healthcare coverage which themselves control for income and employment to some degree, analyses related to income and employment were excluded from this research. The majority of respondents attained at least a high school education; therefore, analyses related to education were excluded from this research. This research focused on understanding current social context; past unmet social needs, e.g., during childhood, were not assessed.
  2. Erica Coe and Jenny Cordina, “Insights on mental health from a 2019 McKinsey Consumer survey,” McKinsey & Company, February 2020, McKinsey.com.
  3. Analysis uses correlations and trends to identify possible relationships between unmet social needs and health; therefore, results cannot be interpreted to determine causality between unmet social needs and health.
  4. Calculated as a composite score for each respondent, incorporating the number of ongoing chronic conditions a respondent has and their past healthcare utilization in inpatient, emergency room, and primary care settings. See technical appendix for additional detail on methodology.
  5. We defined the Medicaid churn population as individuals with household incomes below 250 percent of the federal poverty level who were uninsured or who received health insurance through the individual market.
  6. See technical appendix for additional detail on methodology.