Prevalence of social risk factors and social needs in a Medicaid Accountable Care Organization (ACO)

Health-related social needs (HRSN) are associated with high chronic disease prevalence, poor disease control [1‐4], and high health care utilization in both adults and children [5‐7]. Increasingly, health systems are screening patients for HRSN during routine care and integrating responses into the electronic medical record (EMR), with the goal to refer or provide resources to address identified needs [8, 9]. Screening for HRSN has also become a priority for public payors in the Accountable Care Organization (ACO) model as a strategy to prevent and treat chronic disease [10‐12], and several states allow use of Medicaid funds to directly address HRSN like food and housing [13, 14].

Screening for HRSN may be completed using many instruments, which can include questions on social risk factors as well as social needs [15‐17]. The relationship between social risk factor and social need screening is unclear. Consistent with prior literature, we use the term health-related social needs to mean a group of individual-level adverse social determinants of health, such as those assessed in a screening instrument [10]. We use social risk factors or social risks as the specific adverse social and economic conditions associated with poor health as measured on the individual level, for example food insecurity. We use the term social needs when individuals express their own preferences and priorities to address these conditions, such as requesting assistance with food [18, 19].

Previous studies have shown that the identification of social risk factors is not always consistent between different screening instruments assessing the same domain, such as housing risk [20]. There has also been substantial variability in the extent to which individuals identified as having social risk factors on screening instruments report social needs by requesting additional assistance [21‐24]. In smaller studies of patient questionnaires in a research context, 8.6 to 26% of participants who screened negative or declined to answer social risk screening questions still indicated they were interested in receiving resources for social needs [23, 25, 26]. It is therefore not clear how social risk factor and social need screening overlap in identifying HRSN in a population.

In our study, we sought to understand the prevalence of social risk factors and social needs in a large population screened for HRSN as part of routine clinical care. We included both patient and practice level characteristics in our model, drawing from the Drivers of Health framework, which includes indirect factors (such as public policy, gender, and racial identity) that affect direct factors (such as environment, access to and quality of healthcare, and social circumstances) that affect health outcomes [27].

The goals of this study were: (1) to compare the differential prevalence of social risk factors and social needs in a Medicaid ACO population, specifically describing the characteristics of patients who would be missed by screening for social risk factors only, and (2) to identify the patient and practice characteristics associated with reporting social needs in a different screening domain from social risk factors.

In this study, we demonstrate that screening with social risk factors as compared to social needs identifies different patient populations across a large primary care population in varied practice settings in a Medicaid ACO. Patients who identified as a race/ethnicity other than non-Hispanic White were more likely to report social needs, and more often reported social needs without reporting social risk factors. Among those with social risk factors, patients were more likely to report social needs in a domain discordant to social risks if they identified as non-Hispanic Black, preferred a language other than English, had higher social risk overall, or received care in a practice with higher proportions of patients with LEP and/or Medicaid enrollment. These patients would have been missed if they were screened with social risk factor questions alone (Fig. 1). These individuals are also more likely to experience HRSN due to structural racism and systemic poor access to health services [29, 30], emphasizing the importance of including both social risk factor and social need questions in integrated screening tools to improve the equity and accuracy of clinical screening programs.

It is difficult to precisely compare the prevalence of social risk factors and social needs to other studies due to differences in the populations and screening tools examined. In our study, the 44.6% social risk and 19.1% social need falls in the mid-range of previously published data, with reports of multi-domain HRSN screening in primary care showing a prevalence of 15 to 90% for social risk factors [21, 31, 32], and studies specifically assessing request for assistance finding 15 to 37% with social needs [9, 23, 32]. Our prevalence of food and housing insecurity specifically were also comparable to those reported in other studies [21‐23, 31‐34].

Our study expands upon prior research identifying a discrepancy between social risk factor and social need screening [23, 25, 35]. This observational study of routine-care screening in a large population across varied practice settings adds to the understanding of HRSN prevalence in clinical practice, and expands upon existing literature by identifying specific patient and practice characteristics associated with domain discordant screening. Our findings are supported by insights from prior research, including a study of an emergency department population in the same health system finding that non-Hispanic Black and Spanish speaking patients more often reported social need rather than social risks [35].

There are multiple reasons why patients may report social needs but not social risk factors on a screening tool. Patients may experience stigma regarding their social circumstances or have privacy concerns about who will see the information [22, 36]. Others may perceive questions on social needs to be more relevant or actionable compared to social risk screening. The finding that patients report social needs in the absence of social risks underscores the limitation of using social risk factor screening alone, and lends further support to implementing patient-centered strategies that engage individuals in determining their own needs and priorities [37].

This study has several potential limitations. First, patients in the sample were only those without substantial churn in Medicaid eligibility (at least 11/13 member-months) and who engaged in routine primary care, limiting the portion of the ACO examined. These patients are likely to be different from the portion of the Medicaid ACO population who experiences more disruptions in eligibility or is unable to participate in scheduled office-based care. Second, the study is a secondary analysis of data that was collected during routine clinical care rather than to answer a specific research question, leading to potential misclassification and missing data. The race, ethnicity, and language data from the EMR were not complete for all included patients, though unavailable data was limited to 8% of race/ethnicity and 3% of language preference. Third, while the institutional screening tool used questions from validated screeners, the entire instrument was not formally validated prior to clinical deployment, leading to potential bias in the patient responses collected. Additionally, we used the request for more information item from this screener to define social need because the answers expressed patient prioritization of a domain and preference for additional engagement. We recognize that patients were not specifically asked if they would like help addressing the health-related social need and this may have led to misclassification of patient responses. The question is an imperfect proxy, although provides an opportunity to understand patient prioritization of their own needs in a real-world clinical screener. Finally, our analysis was limited to patients with Medicaid in a single large, integrated health care system. Because the Medicaid population is more likely to have high social risk and needs, the results may not be generalizable to other patient populations. The practice settings included were varied in size, location, practice ownership, and resources for patients with LEP ranging from on-site interpreters to third-party phone services. However, the results may not be generalizable to patients who are uninsured or who receive medical care in different health system settings.

The findings from this study have important implications for health policy and practice. Health systems, payors, and policy makers who wish to screen for HRSN should carefully consider how to conduct population-based screening as asking about social risk factors alone are not sufficient to identify all patients with HRSN in a Medicaid population. Populations with systematically higher HRSN may be more likely to report social needs rather than social risk factors. Health systems and Medicaid programs should consider screening tools that include questions which assess both social risk factors and patient-identified social needs. Identifying both populations of patients would increase the opportunity for intervention to reduce the burden of HRSN and associated adverse health outcomes.