Provision of Social Services and Health Care Quality in US Community Health Centers, 2017

Abstract

Objectives. To describe the types of social services provided at community health centers (CHCs), characteristics of CHCs providing these services, and the association between on-site provision and health care quality.

Methods. We surveyed CHCs in 12 US states and the District of Columbia during summer 2017 (n = 208) to identify referral to and provision of services to address 8 social needs. Regression models estimated factors associated with the provision of social services by CHCs and the association between providing services and health care quality (an 8-item composite).

Results. CHCs most often offered on-site assistance for needs related to food or nutrition (43%), interpersonal violence (32%), and housing (30%). Participation in projects with community-based organizations was associated with providing services on-site (odds ratio = 2.48; P = .018). On-site provision was associated with better performance on measures of health care quality (e.g., each additional social service was associated with a 4.3 percentage point increase in colorectal cancer screenings).

Conclusions. Some CHCs provide social services on-site, and this was associated with better performance on measures of health care quality.

Public Health Implications. Health care providers are increasingly seeking to identify and address patients’ unmet social needs, and on-site provision of services is 1 strategy to consider.

Mounting evidence suggests that social determinants of health (SDOH), inclusive of social, behavioral, and economic factors and the physical and built environments, strongly influence an individual’s health.¹ Recent studies report that lack of stable housing, limited incomes and education, and food insecurity are associated with poor mental and physical health,² worse outcomes for patients with diabetes,³ and overall poor health outcomes.⁴ Some have estimated that SDOH explain more than 80% of the differences in population health outcomes.⁵ There is increasing attention to addressing SDOH in primary health care settings. Professional associations are encouraging screening for social needs,^6,7 and state Medicaid programs are offering flexibility in paying for nonmedical services.⁸ There is also growing federal, state, and communitywide engagement to test strategies to better align health and social services—for example, the Centers for Medicare and Medicaid Services’ Accountable Health Communities initiative.⁹

Many studies have examined the effects of addressing social needs in medical settings. A 2017 systematic review of interventions to address patients’ social needs in medical settings found 67 studies, of which all interventions included screening for SDOH followed by varying approaches for linking patients with on-site or external resources (e.g., use of resource sheets, navigators, or community health workers).¹⁰ Only a minority of these studies, however, examined the effects of SDOH interventions on health (30% of studies) or health care cost or utilization (27% of studies), with interventions leading to mixed but mostly positive results for these outcomes. Despite these mostly positive findings, a recent nationally representative survey of medical practices reported that only 15.6% of practices reported screening for 5 common social needs: food insecurity, housing instability, utility needs, transportation needs, and interpersonal violence.¹¹ Moreover, evidence suggests that patients who are referred for social services from health care providers do not always receive them.¹⁰

Community health centers (CHCs) are nonprofit, community-focused primary care providers that treat all patients regardless of ability to pay and are well positioned to address their patients’ social needs. CHCs provide primary care to more than 27 million patients annually, many of whom have complex medical conditions and high need for social services.¹² CHC patients often have difficulty meeting their basic needs; more than half rely on public insurance, 23% are uninsured, 92% have incomes at or below 200% of the federal poverty level, and about 1.4 million patients are homeless.¹³ CHCs were early adopters of screening for SDOH and are more than twice as likely to screen for SDOH as the average medical practice.¹¹ However, rigorous evidence regarding the effects of addressing social needs in CHCs is limited. The recent systematic review included only 4 studies focused on CHCs,¹⁰ and none of these studies were designed to address social needs on-site.^14–17

To address patients’ unmet social needs, some CHCs have sought to provide some of these services on-site. For example, a CHC in Connecticut is enrolling eligible patients in the Supplemental Nutrition Assistance Program (SNAP) and offering snacks where behavioral health services are offered, a CHC in Illinois offers exercise and cooking classes on-site, and a CHC in California offers job training on-site.¹⁸ On-site provision of social services can help address patients’ SDOH immediately upon screening, which may help to ensure successful completion of referrals as patients do not have to schedule additional visits across multiple days at different locations. Although on-site provision of social services may be a promising strategy for some CHCs, it requires both physical space, operational support, and additional funding. Given the resources needed to support these efforts, it is important to understand the characteristics of CHCs engaging in this strategy and the types of services that are being offered. Additionally, although prior studies have primarily focused on evaluating individual interventions,¹⁰ less is known about the population-level impact of these types of interventions across a wide range of social needs. Given CHCs’ emphasis on tracking and improving population health and health care providers’ increasing interest in screening for and addressing social needs, it is important to know if offering conveniently located on-site social services is associated with population health improvements.

To fill this gap in the literature, we combined a survey of CHCs in 12 states and the District of Columbia with information on CHC characteristics and measures of health care quality from the Health Resources and Services Administration (HRSA). We used these data to describe the types of social services that CHCs either refer their patients to or provide on-site, assess how common these activities are for each service, and examine the factors associated with on-site provision of social services. Additionally, we explore the association between on-site provision of social services and performance on an aggregate measure of population health and health care quality.

METHODS

Our conceptual model for this study was informed by Maslow’s hierarchy of needs, which illustrates how basic human needs must be met for individuals to reach their full potential.¹⁹ We extended this model to illustrate approaches to screening and addressing social determinants of health in primary care settings.²⁰ Building on these models, we hypothesized that CHCs’ provision of social services on-site is likely to reduce time, transportation, and logistical barriers to receiving social services by patients and therefore lead to fewer unmet social needs, which should give patients the time and energy to focus on obtaining preventive health care services and making other healthy choices.

Data Sources

We administered a Web-based survey during summer 2017 to medical directors at 407 HRSA-funded CHCs in 12 states (CA, CO, IL, LA, ME, MN, NJ, OR, UT, VT, WA, WI) and the District of Columbia. We purposely sampled these states to include those with active Medicaid accountable care organization programs at the time of the survey (7 states: CO, ME, MN, NJ, OR, UT, and VT), as we hypothesized that these programs might encourage CHCs to address SDOH using diverse strategies. We supplemented this sample with 5 additional states (CA, IL, LA, WA, WI) and the District of Columbia to improve geographic diversity and to include states that expanded Medicaid under the Affordable Care Act and those that did not. HRSA’s Uniform Data System database provided contact information for medical directors, CHC characteristics, and measures of health care quality.

Survey

The survey’s goal was to gather information about how CHCs collaborate with hospitals, specialists, and community-based organizations (CBOs),²¹ with the latter being the focus of this article. We informed survey items by reviewing the literature and interviewing subject matter experts and representatives of CHCs and state primary care associations, which provide training and technical assistance to safety net providers. We used items from existing surveys whenever possible.^22–24 We sought comments on the draft survey from primary care associations in all sample states; associations in 6 states provided comments. We conducted 3 cognitive interviews with CHC medical directors outside of our sample to ensure that survey items were eliciting the intended information. The survey was designed to be completed in 30 minutes. We offered CHCs an incentive of a $50 gift card for survey completion, which we later increased to $100 to improve the response rate after 4 weeks in the field.

Variables

Survey items asked CHC respondents about their efforts to address the following 8 social service issues:

• 1.food or nutrition assistance, including through the Special Supplemental Nutrition Program for Women, Infants, and Children (WIC);
• 2.interpersonal violence (including intimate-partner violence, elder abuse, and child maltreatment);
• 3.housing;
• 4.nonmedical transportation (defined as transportation to work, the grocery store, or other locations for nonmedical activities);
• 5.utility bills;
• 6.immigration;
• 7.employment; and
• 8.child care and early education.

We purposefully avoided detailed descriptions of these needs to allow CHCs to use their own interpretation of patient needs and their efforts to address them. For each of the 8 patient needs, CHCs were asked to check all applicable response boxes regarding provision of services and coordination with CBOs: (1) “Yes, our staff or another organization provides services on-site”; (2) “Yes, we refer patients to 1 or more organizations”; (3) “No, we do not provide or refer”; (4) “No organization is available in our community” (for this survey item, see Appendix A, available as a supplement to the online version of this article at http://www.ajph.org). For each CHC, we summed responses to items 1 and 2 to generate counts of social services provided on-site or through referral, respectively.

Survey items also asked about potential facilitators and barriers to CHCs’ provision of any social services. One item asked how often in the past 2 years CHCs participated in joint projects and meetings with CBOs. Another item asked how often CHCs send or receive electronic health information with hospitals, which was intended to measure CHCs’ use of health information technology. Another asked how often a chaotic environment in the health center affected a CHC’s ability to care for patients. Three items asked how often the following factors affected the CHCs’ ability to coordinate with CBOs: few CHC staff available to coordinate care, few staff available at CBOs to coordinate care, and insufficient number of CBOs that meet cultural or language needs of patients. We measured responses on a 5-point Likert-type scale (never, rarely, sometimes, often, and always) and dichotomized them by their distribution (often or always vs never, rarely, or sometimes).

Using multiple performance measures from HRSA’s 2017 Uniform Data System, we constructed a composite index of performance on measures of health care quality. The composite measure included 2 measures of health outcomes (controlled blood pressure [< 140/90 mm Hg] among patients with hypertension and hemoglobin A1c < 8% for patients with diabetes) and 6 measures of health care processes (screening for colorectal and cervical cancers; appropriate treatment of patients with asthma, coronary heart disease, and ischemic vascular disease; and children’s receipt of vaccines). For each CHC, we estimated a composite measure of health care quality using a 2-parameter logistic item response model, a common methodology for aggregating performance measures.²⁵ We estimated these composite measures using maximum likelihood, and we specified the likelihood as a series of logistic regressions predicting each performance measure from the composite. In this sense, the composite can be thought of as the “best” predictor of the individual measures, and it is found by exploiting the correlation between the measures. This approach also models the fact that each measure may be more or less difficult to achieve and may be better or worse at discriminating CHC quality. Appendix B (available as a supplement to the online version of this article at http://www.ajph.org) includes a full description of the construction of the composite measure. We also obtained the following characteristics of patients and CHCs from the Uniform Data System: percentage of patients that were homeless, uninsured, enrolled in Medicaid, or best served in a language other than English; total patients; primary care providers per 10 000 patients; and rural versus urban location. Of note, we report all Uniform Data System variables at the grantee level, and the survey was administered at the grantee level; each grantee may have multiple practice sites.²⁶

Analyses

For each of the 8 social service needs, we report the percentage of CHCs that provide referrals or offer the service on-site. We describe the characteristics of CHCs, stratified by whether or not they provide any social services on-site. We estimated a logistic regression model to determine factors associated with CHCs’ provision of social services. Finally, we estimated ordinary least squares regression models to examine the association between a continuous measure of total social services provided on-site and CHC performance on the composite measure of health care quality. We adjusted these regression models for the aforementioned characteristics of CHCs and CHC patients; we performed them with Stata 15.1 (StataCorp, College Station, TX). To contextualize the association between social service provision and the composite measure of health care quality, we used additional ordinary least squares regressions to estimate the association between each component measure and the composite measure of health care quality (e.g., quantifying the average increase in colorectal cancer screening rate for each unit increase in the composite).

RESULTS

Of 407 CHCs invited to participate in the survey, 208 responded (response rate = 51.1%). CHCs responding to the survey were similar to nonresponding CHCs on nearly all characteristics, except that nonresponding CHCs served a higher percentage of homeless patients (full results available in Appendix C, available as a supplement to the online version of this article at http://www.ajph.org). Responding CHCs served an average of about 24 000 patients annually, primarily in urban communities (Table 1). Among the sample, 20.4% of patients were uninsured and 4.2% were homeless.

TABLE 1—

Characteristics of Community Health Centers (CHCs), Stratified by On-Site Provision of Social Services: United States, 2017

	All CHCs (n = 208)	No. of Social Services Provided On-Site
		None (n = 73)	≥ 1 (n = 135)
No. of social services provided on-site, mean (range)	1.8 (0–8)	0	2.8 (1–8)
Rural, %	37.0	43.9	33.3
No. of primary care providers per 10 000 patients, mean	9.6	8.3	10.3
No. of patients, mean	23 824	17 311	27 347
Homeless patients, %	4.2	4.0	4.3
Uninsured patients, %	20.4	19	21.2
Medicaid-insured patients, %	53.1	52.8	53.2
Patients best served in a language other than English, %	22.0	21.5	22.3
Chaotic environment, %	18.1	17.2	18.7
Few CHC staff available to coordinate care with community-based organizations (CBOs), %	42.3	46.6	40.0
Few CBO staff available to coordinate care with CHCs, %	38.5	38.4	38.5
Insufficient no. of CBOs that meet cultural or language needs of patients, %	24.5	26.0	23.7
Participate in projects or joint leadership meetings with CBOs, %	30.3	19.2	36.3
CHC participates in health information exchange with hospitals, %	25.0	27.4	23.7

Note. Sample includes CHCs in CA, CO, DC, IL, LA, ME, MN, NJ, OR, UT, VT, WA, and WI.

Compared with CHCs providing no social services on-site, CHCs providing at least 1 social service on-site served larger patient populations, employed more primary care providers, and were less likely to be located in rural areas (Table 1); additionally, they were more likely to participate in joint projects or leadership meetings with CBOs (36.3% vs 19.2%) and less likely to report an insufficient number of CBOs in their communities that meet the cultural or language needs of patients (23.7% vs 26.0%). After adjusting for covariates listed in Table 2, we found that CHC participation in joint projects or meetings with CBOs (odds ratio [OR] = 2.48; P = .018), higher primary care provider staffing levels (OR = 1.08; P = .010), and having more patients (OR = 1.19; P = .031) were significantly associated with CHCs’ provision of any social services on-site.

TABLE 2—

Factors Predicting Provision by Community Health Centers (CHCs) of On-Site Social Services and Quality of Care: United States, 2017

	Factors Predicting CHCs’ Provision of On-Site Social Services,a OR (SE)	Factors Predicting Quality of Care,b Coefficient (SE)
No. of social services provided on-site (range: 0–8)		0.03** (0.01)
Rural (Ref = not rural)	0.80 (0.35)	−0.04 (0.05)
Total primary care providers per 10 000 patients	1.08* (0.03)	0.005 (0.003)
Total patients (per 10 000)	1.19* (0.10)	0.01 (0.01)
Homeless patients, %	1.91 (4.73)	0.08 (0.28)
Uninsured patients, %	1.03 (0.02)	−0.007*** (0.002)
Medicaid insured patients, %	1.00 (0.01)	−0.005*** (0.002)
Patients best served in a language other than English, %	0.99 (0.01)	0.015*** (0.001)
Reporting a chaotic environment in CHC (Ref = not reporting a chaotic environment)c	1.40 (0.62)	−0.05 (0.05)
Reporting few CHC staff available to coordinate care with community based organizations (CBOs; Ref = not reporting few CHC staff available to coordinate care with CBOs)c	0.72 (0.29)	−0.01 (0.05)
Reporting few CBO staff available to coordinate care with CHCs (Ref = not reporting there are few CBO staff available to coordinate care with CHCs)c	1.26 (0.52)	−0.07 (0.05)
Reporting an insufficient no. of CBOs that meet cultural or language needs of patients (Ref = not reporting an insufficient no. of CBOs that meet cultural or language needs of patients)c	0.73 (0.29)
Report participating in projects or joint leadership meetings with CBOs (Ref = report participating in neither projects nor joint leadership meetings with CBOs)c	2.48* (0.94)
CHC participates in health information exchange with hospitals (Ref = CHC does not participate in health information exchange with hospitals)c	0.71 (0.27)

Note. OR = odds ratio. Sample size was n = 208. Sample includes CHCs in CA, CO, DC, IL, LA, ME, MN, NJ, OR, UT, VT, WA, and WI.

^aThis column presents results from a logistic regression model examining factors predicting CHC’s provision of 1 or more social services on-site.

^bThis column presents results from an ordinary least squares regression model examining the association between number of social services provided on-site and a composite measure of health care quality.

^cAs noted in the Methods section, responses to these items were measured on a 5-point Likert-type scale (never, rarely, sometimes, often, and always) and dichotomized on the basis of the distribution of responses (often or always vs never, rarely, or sometimes [Ref]).

^*P < .05; **P < .01; ***P < .001.

Figure 1 illustrates the types of social services for which CHCs referred patients to another organization or that they provided on-site. CHCs were more likely to refer patients for most services than provide services on-site. For example, more than 70% of CHCs reported referring patients to child care and early education services (77%) and assistance for victims of interpersonal violence (74%). CHCs most commonly reported providing food or nutrition assistance, including through WIC (43%), and assistance for victims of interpersonal violence (32%) on-site. Fewer than 20% of CHCs reported providing the following services on-site: assistance for utility bill payments, immigration, employment, and child care or early education services. Nearly all CHCs reported providing referrals for 1 or more services, whereas nearly 35% of CHCs did not provide any social services on-site (Figure 2). Approximately 1 in 4 CHCs (23%) reported providing only 1 social service and 42% reported providing 2 or more social services on-site.

FIGURE 1—

Percentage of Community Health Centers (CHCs) Offering Each of 8 Social Services by Referral or On-Site: United States, 2017

Note. WIC = Special Supplemental Nutrition Program for Women, Infants, and Children. Respondents were permitted to select both response options (i.e., refer patients to another organization and provide services on-site). Thus, the denominator for each bar is the total number of survey respondents (n = 208). CHCs were more likely to report referring for each service than to report providing each service on-site (for all 8 χ² tests, P < 0.01). Sample includes CHCs in CA, CO, DC, IL, LA, ME, MN, NJ, OR, UT, VT, WA, and WI.

FIGURE 2—

Number of Social Services Provided On-Site or by Referral by Community Health Centers (CHCs): United States, 2017

Note. Results displayed were constructed from a survey question asking respondents to indicate, for each of 8 social services, if services were offered in any of the following ways: (1) “Yes, our staff or another organization provides services on-site”; (2) “Yes, we refer patients to one or more organizations”; (3) “No, we do not provide or refer”; and (4) “No organization is available in our community.” Respondents were instructed to check all that applied. Sample includes CHCs in CA, CO, DC, IL, LA, ME, MN, NJ, OR, UT, VT, WA, and WI.

In an adjusted regression model (Table 2), we found that on-site provision of social services was associated with better performance on the composite measure of health care quality. Each additional social service provided on-site was associated with a 0.03-unit increase in a CHC’s performance on the measure of health care quality (P = .004). Because the composite quality measure does not have a direct interpretation, we estimated how each 0.03-unit increase of the composite was associated with each component measure in the composite. Our models suggest that a 0.09-unit increase on the composite (corresponding to provision of 3 social services on-site compared with no social services) was associated with a 4.3 and 3.8 percentage point increase in rates of colorectal and cervical cancer screening from means of 40.0% and 51.3%, respectively. Associations with other measures were smaller (ranging from 0.4 to 2.5 percentage point increases). Further details, including detailed accounting of the relationship between the composite and the component measures, are available in Appendix B.

DISCUSSION

This study, conducted in 12 states and the District of Columbia, found that CHCs most commonly reported providing assistance to patients for nutrition, interpersonal violence, or housing, and that nearly two thirds of CHCs provided at least 1 social service on-site. CHCs may be motivated to provide these services on-site because they address more immediate needs related to food, shelter, and safety. Additionally, CHCs may be more likely to provide certain services in response to the funding or policy environment, such as federal funding from the Public Housing Primary Care program to provide housing services²⁷ and HRSA-facilitated training on interpersonal violence and trauma-informed care.²⁸ Fewer CHCs reported providing assistance on-site for selected services, including immigration, employment, and child care needs, although CHCs commonly provided referrals for these services. These lower rates may reflect challenges to offering these types of services or suggest that CBOs in the community already address these needs. For example, addressing immigration issues commonly requires legal assistance, and CHCs that have larger budgets and are located in urban areas might be better able to arrange and subsidize on-site legal aid to immigrants.²⁹ Additionally, employment assistance (e.g., job placement services and career counseling) may be offered by a local government agency or CBO in close proximity to the CHC, making referral a better option.

We found that CHCs with greater organizational capacity, as measured by size and relationships with CBOs, were more likely to offer services on-site. This might be due to the resources required to offer these services, such as physical space and other administrative costs. This likely explains why 35% of CHCs did not provide any social services on-site and those that did had more patients, providers, and engagement with CBOs. Although CHC financial resources are limited, our findings offer insights into strategies that may facilitate on-site provision of social services. We found that CHCs reporting participation in projects or joint meetings with CBOs were more likely to report that they provided social services on-site. These interactions may include meetings to discuss shared goals or pursuing collaborative quality improvement projects. For example, CHCs in San Diego, California, meet monthly with CBOs, along with other health care and government partners, to discuss experiences and best practices for using the Community Information Exchange, an online platform connecting social service and health care providers.³⁰

Because on-site provision of social services may not be feasible or the best solution for all CHCs, other strategies are being used to address unmet social needs. Online portals, for use by both social service and health care providers, are increasingly being used to connect patients to resources and to close the loop on referrals. Both Kaiser Permanente and Veterans Administration medical centers use these types of portals, which include resource directories, allow tracking of social service referrals, and enable bidirectional communication between social service and health care providers.^31,32 These virtual portals are also being deployed communitywide; examples include the long-standing Community Information Exchange in San Diego County³⁰ and the newer NCCARE360 platform available across North Carolina.³³ These communitywide solutions can leverage communitywide health outcomes data—such as those from the 500 Cities Project³⁴—for monitoring and evaluating the effects of these strategies on population health.

Limitations

Studies report that unmet social service needs are associated with poor health outcomes,^2,3,35 and here we provide evidence that CHCs’ on-site provision of social services is associated with better performance on measures of health care quality. Our composite measure of performance primarily included process measures, suggesting that availability of on-site assistance for social services may facilitate patients’ receipt of recommended screenings and treatments. This finding is consistent with our conceptual model hypothesizing that CHC provision of social services is likely to reduce barriers to social services, leading to fewer unmet social needs and thus giving patients the time and energy to focus on obtaining health care services. In this cross-sectional study, however, we cannot disentangle whether CHCs that have better performance on measures of health care quality are more likely to offer social services or whether it is the provision of social services that leads to better quality. Given this limitation and our finding that CHCs with greater organizational capacity were more likely to offer on-site provision of social services, we encourage additional research designed to evaluate the causal relationship between on-site provision of social services and health care quality.

Additional limitations of our study include those common to analyses of survey data. Despite cognitive testing of the survey instrument with stakeholders, respondents might have interpreted the survey questions in different ways. Additionally, survey responses were self-reported and may suffer from social desirability bias if medical directors and staff responded to survey items in a way they thought would be viewed favorably instead of accurately. Lastly, CHCs responding to our survey may not be representative of all CHCs nor fully capture the variety of social services that CHCs provide. However, our sample was geographically diverse, and respondents reported variability about services provided and referred.

Public Health Implications

CHC patients often have difficulty meeting their basic needs and may require help finding food, shelter, or employment. Within the context of widespread support for screening for social needs^6,7 and greater flexibility offered by insurers in paying for nonmedical services,^8,9 we found that many CHCs reported providing on-site assistance to patients for nutrition, interpersonal violence, and housing, and that most CHCs provided at least 1 social service on-site. Although on-site provision of social services may not be feasible or even the best solution for all CHCs, addressing SDOH is essential for promoting health care access and improving health outcomes. Given the increasing interest in addressing social needs in medical settings, future research should focus on identifying the best practices used by high-performing CHCs to address patients’ SDOH to ensure that all patients have their social and health care needs met.