Health-Related Social Needs, Methods, and Concerns for a Polysocial Risk Score: An Expert Panel

ABSTRACT

Objectives: A polysocial risk score, which summarizes multiple health-related social needs (HRSNs) into a single likelihood of risk, could support more effective population health management. Nevertheless, a polysocial risk score faces uncertainties and challenges due to the HRSNs’ differing etiologies and interventions, cooccurrence, and variation in information availability.

Study Design: A national expert panel provided guidance on the development and potential application of a polysocial risk score in a 3-round Delphi process.

Methods: Expert panel members from across the US included physicians (n = 8), social service professionals and staff (n = 9), and patients (n = 6). Round 1 obtained an initial sense of the importance of HRSNs for general health and well-being and total health care cost. Panelists also suggested additional HRSNs. Responses served as discussion points for round 2, during which 5 focus groups explored how HRSNs should be ranked, additional HRSNs to include, timing of measurements, management of nonresponse and missing data, and concerns about bias and equity. We analyzed the transcripts using a consensus coding approach. Panelists then completed a follow-up survey (round 3).

Results: Panelists identified 17 HRSNs relevant to health and well-being for inclusion in a polysocial risk score. Methodology concerns included the sources and quality of data, nonrandom missing information, data timeliness, and the need for different risk scores by population. Panelists also raised concerns about potential bias and misapplication of a polysocial risk score.

Conclusions: A polysocial risk score is a potentially useful addition to the growing methodologies to better understand and address HRSNs. Nevertheless, development is potentially complicated and fraught with challenges.

Am J Manag Care. 2025;31(2):In Press

Takeaway Points

• Translating information regarding patients’ health-related social needs (HRSNs) into action remains a challenge.
• A polysocial risk score, which summarizes multiple HRSNs into a single likelihood of risk, could support more effective population health management.
• A national expert panel recommended a wide and diverse set of HRSNs for inclusion in a polysocial risk score.
• A polysocial risk score is a potentially useful addition to the growing methodologies to better understand and address HRSNs. Nevertheless, development is potentially complicated and fraught with challenges.

Health-related social needs (HRSNs) encompass patients’ nonclinical, economic needs.1 HRSNs are distinct from social determinants of health (SDOH), although the terms are often used synonymously. HRSNs are individual-level characteristics, whereas SDOH are communities’ broader socioeconomic and policy conditions that shape individuals’ lives and opportunities.2 HRSNs negatively affect health and well-being,3,4 are associated with increased health care utilization,5-7 and result in greater health care costs.8 Moreover, HRSN information can improve risk prediction models,9 identify patients in need of services,10 and illuminate disparities.11 In light of CMS’ new screening quality measures,12 HRSN screening will likely become more common and important.

However, translating HRSNs into action remains a challenge. HRSNs such as housing instability, financial strain, and transportation barriers frequently co-occur,13 but organizations often collect information on only select HRSNs.14 Likewise, HRSN information is often stored in different locations within electronic health records (EHRs), thereby inhibiting comprehensive views of the patient’s circumstances.15 Despite increased screening, most patients’ HRSNs go unmeasured and unresolved.16

A polysocial risk score, which summarizes the occurrence of different HRSNs into a single likelihood of risk for each individual patient, could support more effective use of HRSN information.17 A single risk measure has analogues in genetics’ polygenic risk scores, clinical care’s comorbidity indices, and public health’s community vulnerability indices (which measure SDOH). Additionally, limited applications of an overall social risk score suggest potential usefulness: Counts of HRSNs have been associated with health outcomes in national surveys.18-20

Nevertheless, developing a polysocial risk score faces uncertainties. HRSNs have differing etiologies and require different interventions.21 Also, risk scores that simply count a patient’s risks fail to account for HRSNs’ interrelated nature and the need to prioritize some factors over others during care.22 Importantly, relevant HRSN data may go uncollected or patients may not supply all relevant information.23

We sought to lay the foundations for the development of a polysocial risk score appropriate for the general US adult patient population. We convened a nationwide expert panel to identify relevant HRSNs, assess importance, and provide methodological guidance. Expert panels can identify which factors should be included in risk scores, guide the relative weighting of factors in a score,24 and identify approaches to account for missing information.25 Our work sets the stage for future development and quantitative assessment.

MATERIALS AND METHODS

We used a 3-round Delphi technique approved by the Indiana University Institutional Review Board (#17119).

Expert Panel

We recruited 23 participants: physicians (n = 8), social service professionals and staff (n = 9), and patients (n = 6) from 11 states by emailing authors of research studies or commentaries on HRSNs in medical education and practice. To identify nonphysicians who are in practice, we requested that each recruited physician recommend a relevant staff member (eg, nurse, social worker, case manager, patient advocate, community health worker). We also relied on our professional networks. Patients were recruited through health system contacts. Panelists self-reported gender, age, and race/ethnicity, which we used to ensure diversity of life experiences and perspectives (Table 1).

Round 1: Preliminary Ranking and Identification

Following guidelines on risk score development,26 panelists rated the importance of an initial set of HRSNs (adapted from the topics included in a major EHR vendor’s screening tool) on a Likert-type scale ranging from “absolutely critical” to “not important at all” (eAppendices 1 and 2 [eAppendices available at ajmc.com]). Panelists then ranked the same HRSNs from most important to least important. Panelists also suggested additional HRSNs. First-round responses were used to stimulate conversation during focus groups.

Round 2: Focus Groups

We conducted, recorded, and transcribed 5 focus groups online (mean of 93 minutes): 2 focus groups of physicians (participant quotes indicated with “MD”), 2 of social service professionals and staff (“SO”), and 1 of patients (“PT”). Prior to each focus group, we reviewed the objective of the study and shared survey results. A semistructured interview guide covered the following: how HRSNs should be ranked, additional HRSNs to be included, measurement timing, management of nonresponse and missing data, and concerns about bias and equity.

Analysis followed a consensus coding approach.27 Three team members read 1 transcript independently to create codes using open coding. Through a joint reading session, coding was compared and refined to develop an initial code book. Pairs of coders coded the remaining transcripts independently and resolved differences through joint readings. The third member of the team served as the adjudicator. We summarized codes into overall themes and identified representative quotes. Across all focus groups, we identified 17 different HRSNs (eAppendix 3).

Round 3: Final Importance Rating

All panelists completed a follow-up survey using Likert-type items to assess the 17 HRSNs’ risk to health from very low to very high (eAppendix 4). We also asked panelists to rate the potential effectiveness of the polysocial risk score in each of the National Academy of Medicine’s 5 social care activities.28 A single question assessed the appropriate level for using a polysocial risk score: individual-level care, population-level activities, or both. We compared median responses by participant type29 using the Kruskal-Wallis test.

RESULTS

The panelists provided guidance on why particular HRSNs should be included in a polysocial risk score, methodological considerations, and the positives and potential negatives (Table 2).

Reasons for Inclusion

The HRSNs discussed were recognized as important. A physician (MD 5) summarized: “They’re all critically important to health…and everything is interconnected.” Nevertheless, several HRSNs were identified as more important due to their relevance to health. A director of community services (SO 2) described financial strain, food insecurity, and housing instability as having “a little bit more urgency,” a social worker (SO 1) called them “root causes,” and a patient (PT 1) equated those 3 HRSNs with Maslow’s hierarchy of needs. A physician (MD 8) noted, “Cost of daily living, buying food, paying your rent, keeping the utilities active in your home—many people, especially those who have children, are going to prioritize the family’s stability over their own personal health.” At the same time, panelists called out other HRSNs such as exposure to violence, legal problems, health literacy, discrimination, education, and adverse childhood experiences as still important.

Relatedly, several factors were noted as more relevant due to personal experience. Providers and staff recalled care delivery events; patients described salient events or contexts that warranted including some HRSNs. For example, a patient (PT 5) introduced the concept of immigration status: “I go to church every Sunday and I go to a Spanish branch, and most of them, if not all of them, have legal documents problems, and so I just see them struggling more.” In similar fashion, patients from rural areas (PT 3 and PT 6) recounted how transportation barriers posed a significant barrier to their health.

Available measurements and available interventions were the logic for inclusion and greater importance of some HRSNs. One physician (MD 1) noted, “We may tend to rank these things with things [ie, HRSNs] that we can do something about, right? So food insecurity is one where people find it a little bit easier to refer people to the 8 million programs that are out there vs solving a senior citizen’s social isolation, which is way harder.” Another physician (MD 7) illustrated the concept through a contrast: “Certainly they’re all associated with health outcomes. But food and security...in terms of operations, [are] something that a lot of people are thinking about clinically and…there are state programs and local programs. Social isolation—I think a lot of people shy away from this because it maybe seems less tangible to address it in clinical settings.”

Panelists’ general view was to exclude behavioral factors (eg, nutrition, physical activity) and behavioral health from a polysocial risk score. Relevant quotes included that “the interventions are different” (MD 5), the “solution for those 2 areas is different” (SO 4), and these require “totally different resources” (SO 1).

The final rankings (round 3) also reflected the general perception that many different HRSNs were important to overall health and well-being (Table 3); specifically, on a 5-point scale, every HRSN had at least a median rating of 3. The highest median score was for access to care. The next highest-rated factors mirrored much of the discussion and included adverse childhood events, discrimination, exposure to violence, financial strain, food insecurity, housing instability, language barrier, and stress. The only social factor for which rating varied significantly by respondent type (P = .0107) was immigration status; patients had a higher median rating of this factor.

Risk Score Methodology

Panelists’ recommendations were broad regarding the data sources that could be used for HRSNs, as any one source could be insufficient. For example, one social worker (SO 9) noted, “I think a survey alone doesn’t really provide the response or the feedback that [something] a nurse wrote as a note could give. We need to be pulling information from all places.” Surveys were viewed as “cleaner” (SO 5), and survey information was viewed favorably by social service panelists because it was “patient reported” (SO 1, SO 2, and SO 4). At the same time, panelists noted limitations. Regarding Z codes in the International Statistical Classification of Diseases, Tenth Revision, one physician (MD 7) observed, “You never know when it was documented. I think relying on them would be certainly underestimating a lot.” A nurse (SO 8) echoed, “I don’t know any consistent provider who actually uses those [Z] codes.… It’s not [going to] be super robust.” A social worker (SO 1) noted, “I do think there are other elements that absolutely can be pulled from the record; you just need to be thoughtful about what exactly.”

Although surveys had advantages, a population health director (SO 5) noted challenges with missing information: “The full completion rate is about 15%, which means they’re leaving a lot blank.” Multiple physicians reported that missing information was not at random (MD 1, MD 3, and MD 4) and that responses of “prefer not to answer” and skipped questions were different (MD 6). A patient confirmed this view: “Somebody may miss a question because they don’t understand it. They prefer to leave it blank than give inaccurate information” (PT 2). Reflecting this possibility, a social worker reported that their organization still offered resources to individuals who declined screening (SO 3). Options for handling missing information included statistical imputation (MD 1, MD 3, MD 8, and SO 1) and rescreening (SO 1).

A physician (MD 3) captured the tensions in the discussions of timeliness: “I mean, ideally, that time of visit, that’s when you’re making decisions. But that’s not practical.” A social worker (SO 7) explained, “I want to know where my patient is today.” Likewise, panelists noted that social program benefits can change and issues such as homelessness require immediate attention (SO 3). Similarly, a patient (PT 6) endorsed the value of more recent information: “It depends on what day it is as to which one of these things are most important.” Nevertheless, panelists noted that the burden of “survey fatigue” (MD 1) and that some HRSNs can remain “unchanged”(MD 4) for months. Longer time frames were still seen as useful. A physician (MD 7) noted, “Even having that sense of the past in there can give you some sort of data that [are] relevant to health outcomes.” Likewise, longer time frames were “easier to define, like a year of observation, for so many reasons, including insurance and eligibility” (MD 2). Timing was attributed to “kind of a pragmatic decision” (MD 5), trying to strike “a good clinical balance as well as a good logistic balance” (MD 1), or to what was defined by the screening tool (SO 3 and SO 9). A social worker (SO 6) summarized, “As a clinician, a history is important for me regarding treatment outcomes. In terms of an accurate current risk score, a more recent time frame is more helpful to me.”

Panelists noted ways that a polysocial risk score may lack generalizability or need stratification. Immediate needs and resources vary between the emergency department and primary care (MD 8), patients may experience very different “built economic and social environments” (SO 4), and older patients face different financial risks (MD 4). Additionally, patient panelists were attentive to rural and urban differences (PT 3, PT 5, and PT 6).

Application

The overall survey responses suggested that a polysocial risk score would be “very effective” for all 5 social care activities (ie, awareness, adjustment, assistance, alignment, and advocacy) (Table 4). However, physicians reported significantly lower perceptions of effectiveness for adjustment (P = .0144), alignment (P = .0202), and advocacy activities (P = .0312). A physician (MD 6) noted, “I think you actually need individual items largely for adjustment. If you’re [going to] do telehealth follow-up, you need very detailed information about digital access or transportation….” Another (MD 4) stated, “I worry about the creation of a single unique score because it kind of disaggregates our ability to act on the information. If you tell me the risk score is 13, I can’t do anything. Rather, telling me that they screen positive for housing and food insecurity and something else...I might have an intervention to deliver, or at least resources to provide.”

Physicians largely agreed (62.5%) with more population-level usage of the risk score (Figure). In discussions, physicians’ population-level applications of the polysocial risk score were for risk stratification. A physician (MD 2) noted, “Population-level stratification, I still think that it is a starting point for the care teams to decide which subpopulation of patient I should start with questioning or which one requires more frequent or in-depth assessment of their social needs. And then based on those specific needs, I take action.” Phrases such as “distinguish which track of intervention patients should go in” (MD 5), “triage” (MD 7), or “trying to find people” (MD 1) indicated applying a polysocial risk score to a larger patient group in order to identify individuals with the highest needs. Also referring to the population level, social service panelists reported potential applications that were more for organizational decision-making and policy making around the groups of patients for which the organization was responsible. For example, “I think it is helpful…aggregating data so that you can start to see patterns or not see patterns. That’s really valuable for allocating resources” (SO 1). Likewise, a director of community services (SO 2) mentioned that such a score might support “a better population-level understanding of what’s happening in our institutions.”

Panelists noted the potential for bias and stigma in a polysocial risk score and also for HRSN data generally. A patient (PT 5) reported, “I don’t want to get myself in trouble.… I think there can be bias or answers that are not accurate.” Another patient (PT 6) agreed, “Everything’s not fair game.… Something from 10 years ago, it’s not really relevant anymore.” Another (PT 2) said, “If they don’t need to know, they don’t need to know.” Biases in the data could also arise from system problems. A physician (MD 2) noted, “Some people don’t have access to care.… The implicit bias in the health care system and the patterns of cost of care [are] very different.” Also, a director of community services (SO 2) asked, “How much of these scores are really just going to reflect more of that and how poorly resourced our system and our structured system is, and we’re just continuing that to reinforce that…?” Even with these concerns, some panelists were in favor of a polysocial risk score. For example, a physician (MD 3) noted parallels to genetic risk scores: “The concern for discrimination is accurate.… Unfortunately, it’s [going to] happen anyway.… I think you have to be very careful on how you present this.… I don’t think that’s necessarily a hard stop either.” Another physician (MD 1) stated, “I’m a person who believes in doing a score, mostly to trigger interventions to do things.… I absolutely worry about bias...increased child protection reports…differential interactions…differential treatments.”

DISCUSSION

Experts identified a set of HRSNs relevant to health and well-being for inclusion in a polysocial risk score. Views and recommendations provided substantial details beyond the existing calls suggesting a polysocial risk score in terms of relevant HRSNs, methodological concerns, and application. Establishing clarity is all the more pressing as HRSN information becomes more widespread to US health care organizations and researchers.

The panelists recommended a wide and diverse set of HRSNs. In fact, recommendations were much broader than what many organizations currently collect14 or those that are included in widely used screening tools,30 indicating the importance of HRSNs on health and well-being. However, although the panel was not unanimous, the general preference was to exclude behavioral factors (ie, social was different than sociobehavioral). Justifications exist for such separation: Social and behavioral risks have different origins and require different interventions by different professionals.31 The panelists’ distinctions may help the scientific community describe and classify the growing number of polysocial risk scores appearing in the literature. For example, several disease-specific studies combining HRSNs also included behavioral factors19,31-33 and clinical conditions or relied on a small set of HRSNs.20,34 In contrast, a pair of studies creating a polysocial risk score for older adults considered a wide range of HRSNs and omitted behavioral factors.35,36 As pointed out by our findings, these studies do not use comparable definitions of a polysocial risk score.

Polysocial risk scores in the literature include counts19,20,31,32,34,37 and more advanced analytic models.33,35,36 Our study using experts was not able to effectively discriminate HRSNs by importance, suggesting that weights may be best determined by empirical methods. However, the high rating of all factors suggests that simply counting unweighted factors could also be appropriate. Additional methodological concerns included the sources and quality of data, missing information, timeliness, and the need for different risk scores by population. These challenges are generally the same ones faced by any risk score,25 but the nature of HRSN data may exacerbate these challenges. For example, HRSNs can change quickly over time (appearing or resolving),38 which is different from polygenic risk scores or the unidirectional slow increase of a comorbidity index.39 Additionally, no agreed-upon measurement frequency exists. Further, although clinical information systems like EHRs include data on many HRSNs, these systems/processes of care delivery were not designed to systematically collect HRSN data. Therefore, any effort to measure all the HRSNs needed for a risk score will require multiple data collection modalities.

Increasing awareness (ie, identifying patients with HRSNs) is a potential use case for a polysocial risk score.28 A polysocial risk score would give clinicians and other providers a comprehensive summary of patient risk, instead of the current situation of disjointed and sometimes difficult-to-find individual HRSN information. Nevertheless, some panelists noted potential difficulties in using a polysocial risk for individual care. The aggregated, summative nature of a risk score, combined with challenges in measurement timing, could conflict with providers’ need to know individual-level HRSNs to adjust care or to assist directly or through referrals. However, measures for individual patients may be aggregated into overall measures for entire panels or attributed patients to create population-level awareness. Such a usage would be conceptually closer to risk stratification, where organizations would be able to identify larger patient groups for additional screening or for follow-up activities.40 Stratification based on polysocial risk score could be a means of more effectively using organizational resources.16 Additionally, a polysocial risk score could be useful as an adjustment factor in quality reporting metrics.41

Summary risk scores have been demonstrated to be biased,42,43 and bias or misuse was a concern of panelists. Selective reporting and collecting of HRSNs create a distorted picture of risk.44 Moreover, HRSN data are particularly susceptible due to differential access to care and explicit and implicit biases.45 Underserved and underrepresented populations are disproportionally burdened by HRSNs,46 heightening the need to ensure that a polysocial risk score does not have negative consequences on care.

Limitations

Although we had a diverse set of experts from across the US and we observed saturation, different experts may have generated different results. We used anonymous ratings, stratified focus groups by expert type, and actively moderated discussions, but it is still possible that more dominant personalities influenced the opinions obtained through our Delphi process. We were focused on a general US population, so our findings may not generalize to specific conditions or other populations. Prospective data collection on all suggested HRSNs for a diverse patient population would be necessary to test the validity and formulation of a polysocial risk score.

CONCLUSIONS

A polysocial risk score is a potentially useful addition to the growing methodologies to better understand and address HRSNs. Nevertheless, development is potentially complicated and fraught with challenges.

Acknowledgments

The authors thank the panelists (Carlos F. Albizu, PhD; Johona Bailey, LBSW; Jason J. Bischof; Samantha Boch; Wendy C. Coates, MD; Kyaien O. Conner, PhD, MPH, LSW; Shelbi Cummings, CCHW; Brooke Goodwin, LICSW, LCSW-C; Elham Hatef, MD, MPH, FACPM; Ivanna Cáceres Hogan; Bri Morris; Matthew S. Pantell; Wendy Parisi; E. Stanley Richardson; Alexandra Rucker, MD; Margaret Samuels-Kalow, MD, MPhil, MSHP; Kristin Topel; Dorica Watson; and Kiley Wanecke, MHA) for their participation and Ms Brittany Miller for assistance in coding.

Author Affiliations: Indiana University Richard M. Fairbanks School of Public Health (JRV, CM), Indianapolis, IN; Center for Biomedical Informatics, Regenstrief Institute (JRV), Indianapolis, IN; Department of Emergency Medicine, Indiana University School of Medicine (PIM), Indianapolis, IN.

Source of Funding: This work was supported by the Health Disparity and Equity Research Program at the Regenstrief Institute.

Prior Publication: A prior version of this work appears as a preprint at doi:10.1101/2023.10.17.23297142.

Author Disclosures: Dr Vest is the founder of and holds equity interest in Uppstroms. The remaining authors report no relationship or financial interest with any entity that would pose a conflict of interest with the subject matter of this article.

Authorship Information: Concept and design (JRV, PIM); acquisition of data (JRV, CM, PIM); analysis and interpretation of data (JRV, PIM); drafting of the manuscript (JRV, PIM); critical revision of the manuscript for important intellectual content (PIM); statistical analysis (JRV); provision of patients or study materials (JRV, CM); obtaining funding (JRV); and administrative, technical, or logistic support (JRV, CM).

Address Correspondence to: Joshua R. Vest, PhD, MPH, Indiana University, 1050 Wishard Blvd, Indianapolis, IN 46032. Email: joshvest@iu.edu.

REFERENCES

1. National Academies of Sciences, Engineering, and Medicine. Investing in Interventions That Address Non-Medical, Health-Related Social Needs: Proceedings of a Workshop. The National Academies Press; 2019. doi:10.17226/25544

2. Green K, Zook M. When talking about social determinants, precision matters. Health Affairs Forefront. October 29, 2019. Accessed December 3, 2019. https://www.healthaffairs.org/do/10.1377/hblog20191025.776011/full/

3. McCoy RG, Lipska KJ, Herrin J, Jeffery MM, Krumholz HM, Shah ND. Hospital readmissions among commercially insured and Medicare Advantage beneficiaries with diabetes and the impact of severe hypoglycemic and hyperglycemic events. J Gen Intern Med. 2017;32(10):1097-1105. doi:10.1007/s11606-017-4095-x

4. Engelberg Anderson JK, Jain P, Wade AJ, Morris AM, Slaboda JC, Norman GJ. Indicators of potential health-related social needs and the association with perceived health and well-being outcomes among community-dwelling Medicare beneficiaries. Qual Life Res. 2020;29(6):1685-1696. doi:10.1007/s11136-019-02410-7

5. Mayes KD, Cash RE, Schiavoni KH, et al. Social risk, social need, and use of the emergency department. JAMA Netw Open. 2024;7(1):e2352365. doi:10.1001/jamanetworkopen.2023.52365

6. Canterberry M, Figueroa JF, Long CL, et al. Association between self-reported health-related social needs and acute care utilization among older adults enrolled in Medicare Advantage. JAMA Health Forum. 2022;3(7):e221874. doi:10.1001/jamahealthforum.2022.1874

7. Holcomb J, Highfield L, Ferguson GM, Morgan RO. Association of social needs and healthcare utilization among Medicare and Medicaid beneficiaries in the Accountable Health Communities Model. J Gen Intern Med. 2022;37(14):3692-3699. doi:10.1007/s11606-022-07403-w

8. Jones KG, Roth SE, Vartanian KB. Health and health care use strongly associated with cumulative burden of social determinants of health. Popul Health Manag. 2022;25(2):218-226. doi:10.1089/pop.2021.0255

9. Vest JR, Ben-Assuli O. Prediction of emergency department revisits using area-level social determinants of health measures and health information exchange information. Int J Med Inform. 2019;129:205-210. doi:10.1016/j.ijmedinf.2019.06.013

10. Gottlieb LM, Hessler D, Long D, et al. Effects of social needs screening and in-person service navigation on child health: a randomized clinical trial. JAMA Pediatr. 2016;170(11):e162521. doi:10.1001/jamapediatrics.2016.2521

11. Bazemore AW, Cottrell EK, Gold R, et al. “Community vital signs”: incorporating geocoded social determinants into electronic records to promote patient and population health. J Am Med Inform Assoc. 2016;23(2):407-412. doi:10.1093/jamia/ocv088

12. Screening for social drivers of health. CMS Measures Inventory Tool. Accessed March 25, 2024. https://cmit.cms.gov/cmit/#/MeasureView?variantId=10405&sectionNumber=1

13. Wan W, Li V, Chin MH, et al. Development of PRAPARE social determinants of health clusters and correlation with diabetes and hypertension outcomes. J Am Board Fam Med. 2022;35(4):668-679. doi:10.3122/jabfm.2022.04.200462

14. Cartier Y, Gottlieb L. The prevalence of social care in US health care settings depends on how and whom you ask. BMC Health Serv Res. 2020;20(1):481. doi:10.1186/s12913-020-05338-8

15. Pantell MS, Adler-Milstein J, Wang MD, Prather AA, Adler NE, Gottlieb LM. A call for social informatics. J Am Med Inform Assoc. 2020;27(11):1798-1801. doi:10.1093/jamia/ocaa175

16. RTI International. Accountable Health Communities (AHC) Model Evaluation: Second Evaluation Report. CMS; May 2023. Accessed November 2, 2023. https://www.cms.gov/priorities/innovation/data-and-reports/2023/ahc-second-eval-rpt

17. Figueroa JF, Frakt AB, Jha AK. Addressing social determinants of health: time for a polysocial risk score. JAMA. 2020;323(16):1553-1554. doi:10.1001/jama.2020.2436

18. Shah SJ, Oreper S, Jeon SY, Boscardin WJ, Fang MC, Covinsky KE. Social Frailty Index: development and validation of an index of social attributes predictive of mortality in older adults. Proc Natl Acad Sci U S A. 2023;120(7):e2209414120. doi:10.1073/pnas.2209414120

19. Echouffo-Tcheugui JB, Caleyachetty R, Muennig PA, Narayan KM, Golden SH. Cumulative social risk and type 2 diabetes in US adults: the National Health and Nutrition Examination Survey (NHANES) 1999-2006. Eur J Prev Cardiol. 2016;23(12):1282-1288. doi:10.1177/2047487315627036

20. Caleyachetty R, Tehranifar P, Genkinger JM, Echouffo-Tcheugui JB, Muennig P. Cumulative social risk exposure and risk of cancer mortality in adulthood. BMC Cancer. 2015;15:945. doi:10.1186/s12885-015-1997-z

21. Kinner SA, Borschmann R. Polysocial risk scores for assessing social determinants of health. JAMA. 2020;324(16):1680-1681. doi:10.1001/jama.2020.15699

22. Fuchs VR. Social determinants of health: caveats and nuances. JAMA. 2017;317(1):25-26. doi:10.1001/jama.2016.17335

23. Wallace AS, Luther B, Guo JW, Wang CY, Sisler S, Wong B. Implementing a social determinants screening and referral infrastructure during routine emergency department visits, Utah, 2017-2018. Prev Chron Dis. 2020;17:E45. doi:10.5888/pcd17.190339

24. Clack L, Passerini S, Manser T, Sax H. Likelihood of infectious outcomes following infectious risk moments during patient care—an international expert consensus study and quantitative risk index. Infect Control Hosp Epidemiol. 2018;39(3):280-289. doi:10.1017/ice.2017.327

25. Iezzoni LI. Risk Adjustment for Measuring Health Care Outcomes. 4th ed. Health Administration Press; 2012. Accessed July 13, 2020. https://www.ache.org/learning-center/publications/books/2205

26. Harris PA, Taylor R, Minor BL, et al; REDCap Consortium. The REDCap Consortium: building an international community of software platform partners. J Biomed Inform. 2019;95:103208. doi:10.1016/j.jbi.2019.103208

27. Hill CE, Knox S, Thompson BJ, Williams EN, Hess SA, Ladany N. Consensual qualitative research: an update. J Couns Psychol. 2005;52(2):196-205. doi:10.1037/0022-0167.52.2.196

28. National Academies of Sciences, Engineering, and Medicine. Integrating Social Care Into the Delivery of Health Care: Moving Upstream to Improve the Nation’s Health. The National Academies Press; 2019. doi:10.17226/25467

29. Lundorff P, Brölmann H, Koninckx PR, et al; Anti-Adhesions in Gynaecology Expert Panel (‘ANGEL’). Predicting formation of adhesions after gynaecological surgery: development of a risk score. Arch Gynecol Obstet. 2015;292(4):931-938. doi:10.1007/s00404-015-3804-0

30. Moen M, Storr C, German D, Friedmann E, Johantgen M. A review of tools to screen for social determinants of health in the United States: a practice brief. Popul Health Manag. 2020;23(6):422-429. doi:10.1089/pop.2019.0158

31. Khurana D, Leung G, Sasaninia B, Tran D, Khan M, Firek A. The Whole PERSON Health Score: a patient-focused tool to measure nonmedical determinants of health. NEJM Catalyst. 2022;3(8). doi:10.1056/CAT.22.0096

32. Chen P, Yang Z, Fan Z, et al. Associations of polysocial risk score with incident rosacea: a prospective cohort study of government employees in China. Front Public Health. 2023;11:1096687. doi:10.3389/fpubh.2023.1096687

33. Bu F, Deng XH, Zhan NN, et al. Development and validation of a risk prediction model for frailty in patients with diabetes. BMC Geriatr. 2023;23(1):172. doi:10.1186/s12877-023-03823-3

34. Lee HH, Kang AW, Lee H, Cha Y, Operario D. Cumulative social risk and cardiovascular disease among adults in South Korea: a cross-sectional analysis of a nationally representative sample. Prev Chronic Dis. 2020;17:E39. doi:10.5888/pcd17.190382

35. Ping Y, Oddén MC, Stawski RS, Abdel Magid HS, Wu C. Creation and validation of a polysocial score for mortality among community-dwelling older adults in the USA: the Health and Retirement Study. Age Ageing. 2021;50(6):2214-2221. doi:10.1093/ageing/afab174

36. Tang J, Chen Y, Liu H, Wu C. Examining racial and ethnic differences in disability among older adults: a polysocial score approach. Maturitas. 2023;172:1-8. doi:10.1016/j.maturitas.2023.03.010

37. Javed Z, Valero-Elizondo J, Dudum R, et al. Development and validation of a polysocial risk score for atherosclerotic cardiovascular disease. Am J Prev Cardiol. 2021;8:100251. doi:10.1016/j.ajpc.2021.100251

38. Xu S, Goodrich GK, Moore KR, et al. Identifying relative changes in social risk factors: an analytic approach. Med Care. 2021;59(2):e9-e15. doi:10.1097/MLR.0000000000001441

39. Fraccaro P, Kontopantelis E, Sperrin M, et al. Predicting mortality from change-over-time in the Charlson Comorbidity Index. Medicine (Baltimore). 2016;95(43):e4973. doi:10.1097/MD.0000000000004973

40. Parikh RB, Jain SH, Navathe AS. The sociobehavioral phenotype: applying a precision medicine framework to social determinants of health. Am J Manag Care. 2019;25(9):421-423.

41. Nerenz DR, Austin JM, Deutscher D, et al. Adjusting quality measures for social risk factors can promote equity in health care. Health Aff (Millwood). 2021;40(4):637-644. doi:10.1377/hlthaff.2020.01764

42. Obermeyer Z, Powers B, Vogeli C, Mullainathan S. Dissecting racial bias in an algorithm used to manage the health of populations. Science. 2019;366(6464):447-453. doi:10.1126/science.aax2342

43. Vyas DA, Eisenstein LG, Jones DS. Hidden in plain sight — reconsidering the use of race correction in clinical algorithms. N Engl J Med. 2020;383(9):874-882. doi:10.1056/NEJMms2004740

44. Cook LA, Sachs J, Weiskopf NG. The quality of social determinants data in the electronic health record: a systematic review. J Am Med Inform Assoc. 2021;29(1):187-196. doi:10.1093/jamia/ocab199

45. Olteanu A, Castillo C, Diaz F, Kıcıman E. Social data: biases, methodological pitfalls, and ethical boundaries. Front Big Data. 2019;2:13. doi:10.3389/fdata.2019.00013

46. Hill L, Artiga S, Haldar S. Key facts on health and health care by race and ethnicity. KFF. January 26, 2022. Accessed August 19, 2022. https://web.archive.org/web/20220625120424/https:/www.kff.org/report-section/key-facts-on-health-and-health-care-by-race-and-ethnicity-social-determinants-of-health/