Performance of 2-Stage Health-Related Social Needs Screening Using Area-Level Measures

ABSTRACT

Objectives: Screening for health-related social needs (HRSNs) has increased in importance, but screening large patient populations comes with a cost and potential burden for care delivery organizations. This study evaluated the performance of 2-stage HRSN screening that used residence in a high-poverty area to determine which patients were administered screening questions.

Study Design: Screening evaluation.

Methods: Adult primary care patients in Indiana and Florida completed HRSN screening questions included in an electronic health record (EHR) system and a set of additional questionnaires that served as the gold standard for assessing HRSN presence. Responses were linked to patients’ residential zip code (n = 1351). The first stage of screening applied residence in a high-poverty zip code, and the second stage was the EHR-based HRSN screening questions. Using the response to the gold-standard questions, we calculated sensitivity, specificity, positive and negative predictive values, and area under the curve (AUC) for each HRSN.

Results: The highest AUC value was for food insecurity (80%), which was largely driven by the strong performance of the EHR-based HRSN screening questions. The remaining HRSNs had lower AUC values, which were driven by the overall low sensitivities of the screening questions and the overall low performance of the first-stage area-level screen. Positive predictive values were high.

Conclusions: Two-stage HRSN screening based on geography is suboptimal. Although a 2-stage approach based on area-level socioeconomic measures can reduce the number of patients requiring individual-level HRSN screening, large percentages of patients in need would go unidentified.

Am J Manag Care. 2025;31(12):In Press

Takeaway Points

For organizations responsible for large populations, the sheer number of patients and repeated screening for health-related social needs (HRSNs) may translate into a significant source of costs. Two-stage screening, in which a lower-cost and easy-to-implement screening method is first applied to the patient population followed by more intensive screening methods when necessary, is a potential means for more efficient HRSN screening. In the case of HRSNs, area-level socioeconomic measures may serve as an appropriate first-stage screening method. This study evaluated the performance of 2-stage HRSN screening that used area poverty levels as a first-stage screening tool combined with screening questions as a second stage and found the following:

• The 2-stage screening approach using area-level socioeconomic measures performed worse than screening using patient-level questions alone.
• A 2-stage approach would reduce organizations’ screening burdens, but it comes at the expense of lower overall accuracy.

Patients’ health-related social needs (HRSNs) include a wide range of nonclinical, economic, and contextual characteristics.^1-3 HRSNs such as food insecurity, financial strain, and housing instability are common among US patient populations. Growing evidence indicates that HRSNs are important drivers of morbidity, mortality, disparities, and costs.^4,5 In the US, HRSN screening is an increasingly common practice among health care delivery organizations⁶ because CMS has made HRSN screening a quality reporting metric.⁷ Moreover, HRSN screening is endorsed by multiple professional associations.⁸ Screening for HRSNs has multiple applications to patient management and care delivery: Screening may be used to match patients to appropriate social services,^9,10 to identify patients in need of a referral to community partners for services,^11,12 or to increase clinician awareness of relevant patient issues.¹³

Regardless of the potential benefits, screening for HRSNs is a resource- and time-intensive activity. Screening incurs costs and workflow burdens for care delivery organizations.¹⁴ For example, HRSN screening asks additional time of patients and providers for data collection^15-17 and can impede typical clinical workflows.¹⁸ Also, HRSN screening activities require dedicated resources.¹⁴ Estimates suggest that the direct cost of HRSN screening may be as high as $5 to $8 per patient if screening tools are administered by staff, with additional costs depending upon responses to positive screens.^19,20 Additionally, HRSN screening is potentially associated with recurring costs and challenges. The optimal frequency for HRSN screening is an ongoing and open question, but using the CMS Accountable Care Communities Model requirements as a guide, HRSN screening could occur at least annually.²¹ Yet it may need to be more frequent to meet patients’ changing HRSNs.²² Therefore, for organizations responsible for large populations, the sheer number of patients and repeated HRSN screenings may translate into a significant source of costs.

Thus, health care organizations could benefit from effective and appropriate HRSN screening approaches that reduce burden and costs. One possible solution to these burdens and costs is the use of 2-stage screening, in which a lower-cost and easier-to-implement screening method is applied to the patient population. The purpose of the first stage is to identify, or prioritize, those individual patients in need of additional screening in a second stage with a more intensive and informative screening method.²³ In the case of HRSNs, area-level socioeconomic measures may serve as an appropriate first-stage screening method. Areas of known high need would identify which patients should receive HRSN screening questionnaires.^24-26 For health care organizations, the advantage is clear: A 2-stage screening approach could dramatically reduce the number of patients screened by more labor- and time-intensive questionnaire-based screening methods.²⁶

The conceptual logic and justification for a 2-stage screening approach are relatively strong. For example, prior studies have incorporated area-level socioeconomic measures into patient records^25,27 and utilized those measures to decide who should be referred to case management²⁸ or to identify areas for intervention.²⁹ Also, an Institute of Medicine report on HRSNs considered area-level socioeconomic measures as potentially useful data,³⁰ and a National Academies of Sciences, Engineering, and Medicine report suggested that using area-level socioeconomic data could reduce HRSN screening burdens.³¹ Furthermore, area-level socioeconomic measures are easily available and help identify disparities in populations,³² and their use is consistent with the established public health and epidemiology strategy of hotspotting (ie, identifying high-risk individuals or areas). Importantly, some health care organizations may already be using area-level socioeconomic measures as a type of first-stage “prescreener” to determine which patients should be prioritized for HRSN screening questionnaires.³³

However, using area-level socioeconomic measures as part of an HRSN screening approach is not without limitations. The correlation between area-level socioeconomics and individual-level HRSNs may be weak.^24,34,35 Relatedly, area-level hotspotting may not effectively identify high-risk patients.³⁶ Also, using area-level measures may raise concerns about equity because, as a screening approach, these measures rely on presumed socioeconomic status as opposed to actual HRSNs.³⁷ Lastly and importantly, the performance of 2-stage HRSN screening using area-level socioeconomic measures is unknown. This is not surprising given the high methodological hurdles in evaluating screening methods. Establishing the performance of a 2-stage screening approach is similar to any screening evaluation: All screening results have to be compared against measures of true condition status.^23,38

Given the potential advantages and disadvantages of 2-stage screening, we sought to evaluate the performance of a 2-stage HRSN screening approach in which the first stage prioritized patients with residence in a high-poverty area and the second stage was direct screening via administration of an HRSN questionnaire. Although prior studies have described the correlation between patient-level HRSN questionnaire responses and area-level measures²⁴ or have used area-level measures to predict responses on HRSN questionnaires,^34,39 this study goes further by evaluating the performance of 2-stage screening against a gold-standard measure of patient-level HRSN status. In a secondary analysis of data collected for another study,⁴⁰ we analyzed a sample of adult primary care patients who completed an HRSN screening questionnaire from an electronic health record (EHR) as well as a set of questions that independently measured HRSNs. The latter set of questions served as gold-standard measures. Using these unique data, we were able to estimate the performance of a 2-stage screening approach, with residence in a high-poverty geographic area as the first stage and responses to the EHR-based screening questions as the second stage (Figure).

METHODS

We evaluated the performance of a 2-stage HRSN screening approach using area-level socioeconomic measures and screening questionnaires for 5 HRSNs: food security, housing stability, financial needs, transportation needs, and legal issues.

Setting and Sample

We conducted a secondary analysis of adult (aged ≥ 18 years) primary care patients who completed a set of HRSN questions in Indianapolis, Indiana, and Gainesville, Florida, between January 2022 and February 2023.⁴⁰ Patients in Indiana were invited by trained research assistants to complete the questionnaires during visits at 1 of 3 primary care sites operated by 2 health systems in urban areas. Patients in Florida completed the same questionnaires in person during visits at 2 primary care sites or via phone or email after their visits at 6 additional clinic sites. Patients who were unable to respond to a questionnaire in English or Spanish were excluded. Questionnaires were collected on tablets, paper, or online for those contacted by phone and email.^41,42 Responses reflected 52% of the patients with office visits and 5% of the patients contacted by email or phone. Patients were offered a gift card as an incentive. A total of 1355 patients were included in the study. Patients self-reported their race, ethnicity, and highest level of education completed. We assessed the performance of 2-stage screening on this sample retrospectively (ie, all patients completing the questionnaires were included in the analyses).

First-Stage HRSN Screening

We linked patients’ responses to each health system’s respective EHR system using identifiers (eg, record numbers, names, and date of birth). For each patient, we abstracted age, sex, and residential zip code. We selected zip code as the geography for the area-level measurement because it best reflects a real-world use case: Zip codes are routinely collected within EHR data, and their use by health care organizations would not require any additional geocoding of addresses. We obtained the percentage of population in poverty in the past 12 months from the 5-year estimates in the American Community Survey at the zip code tabulation area level.⁴³ We chose poverty as a first-stage measure for both practical reasons (ie, the data are publicly available and have a straightforward interpretation for organizational leaders) and theoretical reasons (ie, poverty is related to multiple dimensions of material resources⁴⁴ and deprivation, and the financial strain, housing instability, and food insecurity screening questions have explicit financial components). Multiple studies have used the highest quartile as the definition of in-need or at-risk areas.^24,45,46 We were able to assign poverty rates for 1351 of the 1355 (99.7%) patients. The first stage of screening was residence in a high-poverty area, defined as the zip codes with the highest-quartile percentage of poverty within the respective states of Indiana and Florida.

Second-Stage HRSN Screening

Participating patients completed questionnaires that included the HRSN screening items from Epic, a widely implemented EHR system. As is the case with other HRSN screening tools, these EHR-based screening questions were derived from existing publications and are summarized in eAppendix 1 (eAppendices available at ajmc.com).^47-50 For each of the 5 HRSNs, we created binary indicators of a positive response following each of the questionnaire developers’ instructions. The EHR-based screening questionnaire was the second-stage screening. In this study, regardless of zip code poverty levels, all patients were administered the EHR-based HRSN screening questions.

Gold-Standard HRSN Measure

At the time of recruitment, patients also completed an additional set of questionnaires that independently measured the same 5 HRSNs. We selected these questionnaires on the basis of their published psychometric properties: the US Department of Agriculture’s Six-Item Short Form of the Food Security Survey Module,⁵¹ the Housing Instability Index,⁵² the Consumer Financial Protection Bureau’s Financial Well-Being Scale,⁵³ items from the legal status section of the Addiction Severity Index,⁵⁴ and a previously validated measure of transportation barriers.⁵⁵ These additional questions served as the study’s gold standard or external criterion for all analyses and are also reported in eAppendix 1. Following each questionnaire developers’ instructions, we created binary indicators for the presence of each of the 5 HRSNs.

Through this data collection approach, each of the 5 HRSNs were measured by both the EHR-based screening questions and our gold-standard questions and for all patients regardless of zip code characteristics. Completion rates for each HRSN (in both the gold-standard and EHR-based questions) ranged from 96.8% to 99.0%.

Analyses

We described the study sample using frequencies, percentages, and means. We calculated the overall sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and area under the curve (AUC) value for a 2-stage HRSN screening approach^56-58 using the responses to the additional HRSN questionnaires as the gold standard. The resulting measures illustrate the performance of a 2-stage screening approach in which only patients in high-poverty zip codes would have been administered the EHR-based screening questionnaires. We conducted the analysis separately for each HRSN. Additionally, we provide the same performance measures for each screening stage independently calculated.

Supplemental Analyses

We conducted 2 additional sets of analyses. First, we describe the characteristics of first-stage false-negatives for each HRSN. This represents patients who would be incorrectly “missed” in the first stage of the screening process. We defined false-negatives as negative on the first-stage screening but positive per the gold-standard screeners. We compared these patients with those who were true-negatives at the first stage using t tests and χ² tests. Second, as a check on the robustness of our findings, we repeated the analyses using the Area Deprivation Index at the census block group level as our first-stage screening measure. The Area Deprivation Index is a composite measure of overall neighborhood disadvantage.⁴⁵

The study was approved by the Indiana University Institutional Review Board (IRB) and the University of Florida IRB.

RESULTS

The study sample included 1351 patients for whom zip codes could be assigned. Consistent with an urban primary care patient care population, the sample was predominately female (65%) with a mean age of 49 years (Table 1). Likewise, the sample was racially and ethnically diverse, with 41% identifying as Black non-Hispanic and 42% as White non-Hispanic. A majority of the sample resided in higher-poverty-rate zip codes (60%), and the presence of HRSNs was relatively common. Patients in Indiana accounted for the majority of the sample (n = 1054; 78%) and were slightly more diverse and younger than those from Florida. Based on the gold-standard questionnaires, respondents from Indiana more frequently reported food insecurity (43% vs 31%) and more housing instability (43% vs 34%) than those from Florida. In both Indiana and Florida, most patients resided in high-poverty-rate zip codes (62% and 52%, respectively).

Specificity

For all 5 HRSNs (Table 2), the 2-stage screening approach resulted in high specificity (each was greater than 93%). The high specificity was the product of the characteristics of the second-stage screening questions. Each of the screening questions demonstrated high specificity, ranging from 93% to 99%. The first-stage area-based screening was much less specific, with specificity ranging from 41% to 45%.

Sensitivity

The 2-stage screening approach was less sensitive than specific (Table 2). Sensitivities ranged from a low of 18% for financial strain to 65% for food insecurity. With the exception of food insecurity, sensitivities for both the first-stage area-level socioeconomic measure and the second-stage EHR-based HRSN screening questions were low. Of all the EHR-based HRSN screening questions, food insecurity had the highest sensitivity (95%).

Positive and Negative Predictive Values

Patients who screened positive through the 2-stage approach most likely had the HRSN (Table 2). For example, 9 of 10 patients who screened positive for food insecurity or transportation barriers had the HRSN according to responses to the gold-standard questionnaires. Likewise, more than 8 of 10 positive screens for housing and financial strain were reflective of true presence of the HRSN. Negative screens were more often correct (ie, reflective of the absence of the HRSN according to the gold-standard questionnaires), but not to the same extent as PPVs. Again, the observed PPVs and NPVs were driven by the stronger performance of the second-stage screening questions. The first-stage area-level screening had generally poor PPVs.

AUC

The overall discriminatory performance of the 2-stage screening approach was generally low. The highest AUC value was for food insecurity (80%), which was largely driven by the strong performance of the EHR-based screening questions. The remaining HRSNs had lower AUC values that were driven by the overall low sensitivities of the EHR-based screening questions and the overall low performance of the first-stage area-level measures.

Supplemental Analyses

Patients who had false-negatives per the 2-stage screening tended to be younger on average and have lower levels of educational attainment than those who had true-negatives. This was the case for food insecurity, housing instability, financial strain, and transportation barriers (eAppendix 2). In addition, false-negatives for food insecurity and housing instability had a higher percentage of patients who were not White non-Hispanic. For our robustness check, the use of the composite Area Deprivation Index at the block group for the first stage was consistent with the primary analysis findings (eAppendix 3). Overall performance of the 2-stage screening approach remained low for all HRSNs.

DISCUSSION

For all 5 HRSNs considered, the 2-stage screening approach using area-level socioeconomic measures performed worse than screening using patient-level EHR-based HRSN screening questions alone. In fact, only for food insecurity did the 2-stage approach result in performance that might be considered useful in clinical practice.⁵⁹ Two-stage screening approaches maximize specificity^56-58; however, in this study, the EHR-based HRSN screening questions were already very specific,⁴⁰ so actual improvements in this measure were minimal. Critically, the high specificity came at the cost of substantially reduced sensitivity and overall worse performance in discerning patients with HRSNs compared with using EHR-based screening questions alone. The very low sensitivity means organizations focused on population health would likely find that 2-stage screening based on geography fails to achieve widespread improvement in patients’ social risks and needs. In particular, the low sensitivity means that large percentages of patients with HRSNs would go unidentified and, therefore, never be referred to services. Missing patients with HRSNs could exacerbate existing social inequities and frustrate organizations’ efforts to reduce the unnecessary and costly utilization driven by HRSNs. Additionally, as reimbursement moves toward adjusting for, or documenting, HRSNs,⁶⁰ these organizations may be financially harmed by any screening approaches that fail to identify large portions of their population with HRSNs.

Even though area-level measures have been used²⁸ or suggested^30,31 as part of HRSN screening, this study illustrates the ongoing concerns with the application of area-level measures in clinical decision-making.^24,26,34 Measures at the zip code level are attractive, as they are readily available for nearly all patients, definitionally clear, and easy to calculate. Moreover, zip code is easily identified by patients and providers, and it offers a structured value that can be automated in screening protocols and procedures. In addition, area-level measures may avoid some of the overtly biased and subjective practices used to identify patients for HRSN screening that have been previously reported in the literature.^37,61,62 For all those benefits, residence in a high-poverty zip code is a generally poor-performing screening tool given that the AUC values we observed were barely above a coin flip for all the HRSNs considered in this study. It is possible that 2-stage screening with smaller, different geographical units might be useful, but that is yet to be determined. Smaller geographies may be more homogenous units in terms of socioeconomics and demographics but require additional geocoding addresses and, therefore, are less readily accessible to health care organizations.⁶³

Nevertheless, some health care organizations simply do not possess the capacity to screen or address HRSNs for their entire population^11,64 and need some approach to reduce burdens and costs. In such organizations, a 2-stage approach does reduce the number of patients requiring question-based screening. A reduction in screening burden was observed in this urban-based primary care sample because more than half of the patients would have been screened out at the first stage. Excluding that large proportion could represent important practical differences in terms of data collection workflows and personnel capacity. Although area-level measures may be an ineffective first stage, as an alternative, a first-stage serial screening study could use more advanced analytic approaches with higher sensitivities, such as machine learning risk prediction models, which have shown effectiveness in identifying patients at higher risk for HRSNs.^9,65 Although such an approach could be more effective and potentially mitigate the inequities in false-negative rates, it would be both more costly to implement and require more technical sophistication than a basic area-level measure. In the absence of more sophisticated risk stratification, one option could be to leverage technology that facilitates universal self-screening at the patient level, such as through personal health record portals and kiosks or tablet computers in clinics.⁶⁶ Such self-reported approaches may reduce the direct cost of screening dramatically.¹⁹ Although less costly on a per-patient basis, self-screening would not reduce patients’ overall screening burdens and may pose a challenge for those with lower literacy or with technology access barriers.⁶⁷

Limitations

This study has the advantage of using additional questions on HRSNs to act as a gold standard (more accurately described as external criterion measures). This assessment of HRSNs outside the EHR-based screening questions enabled us to evaluate the 2-stage screening strategy. Nevertheless, the findings are limited in generalizability to the patient population and the measurement instruments utilized. Although we have multiple states and health systems represented, the exact performance may not be seen in other locations serving different patient populations, particularly if the prevalence of HRSNs differs. Likewise, those responding via email communications after office encounters may not be generalizable to nonrespondents given the low response rates. Alternative geographies or area-level measures of socioeconomic position exist and may have yielded different results and deserve further investigation. However, another investigation found that index-based area-level measures were poor predictors of positive EHR-based HRSN screening results.³⁴ Also, there is some potential misclassification in terms of patients’ geography because the US Census Bureau’s zip code tabulation areas are not exact geographical matches for self-reported zip codes. In addition, other social screening tools may perform differently from the EHR-based screening questions used in this study. However, we selected HRSN questions from the largest EHR vendor and a readily available area-level measure. Lastly, the actual constructs measured by the EHR-based HRSN screening questions and referent instruments may be different, so different screeners or referent instruments may have yielded different results.

CONCLUSIONS

Using area-level socioeconomic status as part of 2-stage HRSN screening is an approach to reduce screening burdens for organizations. However, it comes at the expense of lower overall accuracy than in-person screening alone. Health care organizations may need to avoid 2-stage HRSN screening unless utilizing more sensitive and overall higher-performing patient-level instruments or applying an alternative first-stage screening test.

Acknowledgments

The authors thank Ms Cassidy McNamee, Ms Cara McDonnell, Ms Abby McCleary, Indiana University Health, Eskenazi Health, Regenstrief Data Services, and University of Florida Integrated Data Repository for their assistance.

Author Affiliations: Indiana University Richard M. Fairbanks School of Public Health (JRV, CAH, JB), Indianapolis, IN; Center for Biomedical Informatics, Regenstrief Institute (JRV, CAH), Indianapolis, IN.

Source of Funding: This work was supported by the Agency for Healthcare Research and Quality (1R01HS028636-01; principal investigator: Vest). Additionally, data collection was supported by the National Center for Advancing Translational Sciences of the National Institutes of Health under the University of Florida Clinical and Translational Science Award (UL1TR001427).

Author Disclosures: Dr Vest has a grant under review at Patient-Centered Outcomes Research Institute and is a founder and equity holder in Uppstroms, LLC, a technology company. Dr Harle has received and has pending research grants to his institution from the following state and federal agencies: National Institutes of Health, Agency for Healthcare Research and Quality, and Indiana Department of Correction. Dr Blackburn reports 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, CAH, JB); analysis and interpretation of data (JRV, JB); drafting of the manuscript (JRV, CAH, JB); critical revision of the manuscript for important intellectual content (JRV, CAH, JB); statistical analysis (JRV, JB); obtaining funding (JRV); and supervision (JRV).

Address Correspondence to: Joshua R. Vest, PhD, MPH, Indiana University Richard M. Fairbanks School of Public Health, 1050 Wishard Blvd, Indianapolis, IN 46202. Email: joshvest@iu.edu.

REFERENCES

1. 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/

2. Alderwick H, Gottlieb LM. Meanings and misunderstandings: a social determinants of health lexicon for health care systems. Milbank Q. 2019;97(2):407-419. doi:10.1111/1468-0009.12390

3. Institute of Medicine and National Research Council. U.S. Health in International Perspective: Shorter Lives, Poorer Health. The National Academies Press; 2013. doi:10.17226/13497

4. Commission on Social Determinants of Health. Closing the Gap in a Generation: Health Equity Through Action on the Social Determinants of Health. World Health Organization; 2008. Accessed November 14, 2025. https://iris.who.int/server/api/core/bitstreams/cb08095c-55c8-484e-bff6-0e9c78fd38dd/content

5. Pruitt Z, Emechebe N, Quast T, Taylor P, Bryant K. Expenditure reductions associated with a social service referral program. Popul Health Manag. 2018;21(6):469-476. doi:10.1089/pop.2017.0199

6. De Marchis EH, Brown E, Aceves B, et al. State of the Science on Social Screening in Healthcare Settings. Social Interventions Research & Evaluation Network; 2022. Accessed February 1, 2023. https://sirenetwork.ucsf.edu/sites/default/files/2022-06/final%20SCREEN%20State-of-Science-Report%5B55%5D.pdf

7. Quality ID #487: screening for social drivers of health. CMS. November 2022. Accessed March 25, 2024. https://qpp.cms.gov/docs/QPP_quality_measure_specifications/CQM-Measures/2023_Measure_487_MIPSCQM.pdf

8. Gurewich D, Garg A, Kressin NR. Addressing social determinants of health within healthcare delivery systems: a framework to ground and inform health outcomes. J Gen Intern Med. 2020;35(5):1571-1575. doi:10.1007/s11606-020-05720-6

9. Vest JR, Menachemi N, Grannis SJ, et al. Impact of risk stratification on referrals and uptake of wraparound services that address social determinants: a stepped wedged trial. Am J Prev Med. 2019;56(4):e125-e133. doi:10.1016/j.amepre.2018.11.009

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. Gold R, Bunce A, Cowburn S, et al. Adoption of social determinants of health EHR tools by community health centers. Ann Fam Med. 2018;16(5):399-407. doi:10.1370/afm.2275

12. Gottlieb L, Sandel M, Adler NE. Collecting and applying data on social determinants of health in health care settings. JAMA Intern Med. 2013;173(11):1017-1020. doi:10.1001/jamainternmed.2013.560

13. Page-Reeves J, Kaufman W, Bleecker M, et al. Addressing social determinants of health in a clinic setting: the WellRx pilot in Albuquerque, New Mexico. J Am Board Fam Med. 2016;29(3):414-418. doi:10.3122/jabfm.2016.03.150272

14. Fraze TK, Brewster AL, Lewis VA, Beidler LB, Murray GF, Colla CH. Prevalence of screening for food insecurity, housing instability, utility needs, transportation needs, and interpersonal violence by US physician practices and hospitals. JAMA Netw Open. 2019;2(9):e1911514. doi:10.1001/jamanetworkopen.2019.11514

15. Kusnoor SV, Koonce TY, Hurley ST, et al. Collection of social determinants of health in the community clinic setting: a cross-sectional study. BMC Public Health. 2018;18(1):550. doi:10.1186/s12889-018-5453-2

16. Kent J. Top 3 data challenges to addressing the social determinants of health. Healthtech Analytics. February 17, 2020. Accessed February 18, 2020. https://healthitanalytics.com/news/top-3-data-challenges-to-addressing-the-social-determinants-of-health

17. O’Gurek DT, Henke C. A practical approach to screening for social determinants of health. Fam Pract Manag. 2018;25(3):7-12.

18. Solberg LI. Theory vs practice: should primary care practice take on social determinants of health now? no. Ann Fam Med. 2016;14(2):102-103. doi:10.1370/afm.1918

19. Drake C, Reiter K, Weinberger M, et al. The direct clinic-level cost of the implementation and use of a protocol to assess and address social needs in diverse community health center primary care clinical settings. J Health Care Poor Underserved. 2021;32(4):1872-1888. doi:10.1353/hpu.2021.0171

20. Basu S, Berkowitz SA, Davis C, Drake C, Phillips RL, Landon BE. Estimated costs of intervening in health-related social needs detected in primary care. JAMA Intern Med. 2023;183(8):762-774. doi:10.1001/jamainternmed.2023.1964

21. Alley DE, Asomugha CN, Conway PH, Sanghavi DM. Accountable Health Communities—addressing social needs through Medicare and Medicaid. N Engl J Med. 2016;374(1):8-11. doi:10.1056/nejmp1512532

22. 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

23. Zhou XH, Obuchowski NA, McClish DK. Statistical Methods in Diagnostic Medicine. 2nd ed. John Wiley & Sons Inc; 2011. Accessed January 27, 2023. https://www.wiley.com/en-us/Statistical+Methods+in+Diagnostic+Medicine%2C+2nd+Edition-p-9780470183144

24. Cottrell EK, Hendricks M, Dambrun K, et al. Comparison of community-level and patient-level social risk data in a network of community health centers. JAMA Netw Open. 2020;3(10):e2016852. doi:10.1001/jamanetworkopen.2020.16852

25. 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

26. Gottlieb LM, Francis DE, Beck AF. Uses and misuses of patient- and neighborhood-level social determinants of health data. Perm J. 2018;22:18-078. doi:10.7812/tpp/18-078

27. Cantor MN, Chandras R, Pulgarin C. FACETS: using open data to measure community social determinants of health. J Am Med Inform Assoc. 2018;25(4):419-422. doi:10.1093/jamia/ocx117

28. Knighton AJ, Savitz L, Belnap T, Stephenson B, VanDerslice J. Introduction of an area deprivation index measuring patient socioeconomic status in an integrated health system: implications for population health. EGEMS (Wash DC). 2016;4(3):1238. doi:10.13063/2327-9214.1238

29. Lichkus J, Liaw WR, Phillips RL. Utilizing PHATE: a population health–mapping tool to identify areas of food insecurity. Ann Fam Med. 2019;17(4):372. doi:10.1370/afm.2387

30. Institute of Medicine. Capturing Social and Behavioral Domains in Electronic Health Records: Phase 2. The National Academies Press; 2014. doi:10.17226/18951

31. 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

32. Krieger N, Chen JT, Waterman PD, Rehkopf DH, Subramanian SV. Painting a truer picture of US socioeconomic and racial/ethnic health inequalities: the Public Health Disparities Geocoding Project. Am J Public Health. 2005;95(2):312-323. doi:10.2105/AJPH.2003.032482

33. Cantor MN, Thorpe L. Integrating data on social determinants of health into electronic health records. Health Aff (Millwood). 2018;37(4):585-590. doi:10.1377/hlthaff.2017.1252

34. Bensken WP, McGrath BM, Gold R, Cottrell EK. Area-level social determinants of health and individual-level social risks: assessing predictive ability and biases in social risk screening. J Clin Transl Sci. 2023;7(1):e257. doi:10.1017/cts.2023.680

35. Brignone E, LeJeune K, Mihalko AE, Shannon AL, Sinoway LI. Self-reported social determinants of health and area-level social vulnerability. JAMA Netw Open. 2024;7(5):e2412109. doi:10.1001/jamanetworkopen.2024.12109

36. Lee NS, Whitman N, Vakharia N, Taksler GB, Rothberg MB. High-cost patients: hot-spotters don’t explain the half of it. J Gen Intern Med. 2017;32(1):28-34. doi:10.1007/s11606-016-3790-3

37. Garg A, Boynton-Jarrett R, Dworkin PH. Avoiding the unintended consequences of screening for social determinants of health. JAMA. 2016;316(8):813-814. doi:10.1001/jama.2016.9282

38. Gilbert R, Logan S, Moyer VA, Elliott EJ. Assessing diagnostic and screening tests: part 1. concepts. West J Med. 2001;174(6):405-409. doi:10.1136/ewjm.174.6.405

39. Holcomb J, Oliveira LC, Highfield L, Hwang KO, Giancardo L, Bernstam EV. Predicting health-related social needs in Medicaid and Medicare populations using machine learning. Sci Rep. 2022;12(1):4554. doi:10.1038/s41598-022-08344-4

40. Harle CA, Wu W, Vest JR. Accuracy of electronic health record food insecurity, housing instability, and financial strain screening in adult primary care. JAMA. 2023;329(5):423-424. doi:10.1001/jama.2022.23631

41. 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

42. Harris PA, Taylor R, Thielke R, Payne J, Gonzalez N, Conde JG. Research electronic data capture (REDCap)—a metadata-driven methodology and workflow process for providing translational research informatics support. J Biomed Inform. 2009;42(2):377-381. doi:10.1016/j.jbi.2008.08.010

43. Manson S, Schroeder J, Van Riper D, Kugler T, Ruggles S. IPUMS NHGIS: version 17.0. IPUMS Center for Data Integration. 2022. doi:10.18128/D050.V17.0

44. Bossert W, Chakravarty SR, D’Ambrosio C. Multidimensional poverty and material deprivation with discrete data. Rev Income Wealth. 2013;59(1):29-43. doi:10.1111/j.1475-4991.2012.00519.x

45. Johnson AE, Zhu J, Garrard W, et al. Area deprivation index and cardiac readmissions: evaluating risk-prediction in an electronic health record. J Am Heart Assoc. 2021;10(13):e020466. doi:10.1161/JAHA.120.020466

46. Phelos HM, Deeb AP, Brown JB. Can social vulnerability indices predict county trauma fatality rates? J Trauma Acute Care Surg. 2021;91(2):399-405. doi:10.1097/TA.0000000000003228

47. Rottapel R, Sheward R. The Hunger Vital Sign: Best Practices for Screening and Intervening to Alleviate Food Insecurity. Children’s HealthWatch; August 2016. Accessed November 16, 2020. https://childrenshealthwatch.org/wp-content/uploads/CHW_HVS_whitepaper_FINAL.pdf

48. Hager ER, Quigg AM, Black MM, et al. Development and validity of a 2-item screen to identify families at risk for food insecurity. Pediatrics. 2010;126(1):e26-32. doi:10.1542/peds.2009-3146

49. What is PRAPARE? National Association of Community Health Centers. Accessed November 14, 2025. https://prapare.org/what-is-prapare/

50. Puterman E, Adler N, Matthews KA, Epel E. Financial strain and impaired fasting glucose: the moderating role of physical activity in the Coronary Artery Risk Development in Young Adults study. Psychosom Med. 2012;74(2):187-192. doi:10.1097/PSY.0b013e3182448d74

51. U.S. Household Food Security Survey Module: Six-Item Short Form. US Department of Agriculture Economic Research Service. September 2012. Accessed March 10, 2020. https://web.archive.org/web/20200208091131/https://www.ers.usda.gov/media/8282/short2012.pdf

52. Rollins C, Glass NE, Perrin NA, et al. Housing instability is as strong a predictor of poor health outcomes as level of danger in an abusive relationship: findings from the SHARE study. J Interpers Violence. 2012;27(4):623-643. doi:10.1177/0886260511423241

53. CFPB Financial Well-Being Scale: scale development technical report. Consumer Financial Protection Bureau. May 26, 2017. Accessed December 28, 2021. https://www.consumerfinance.gov/data-research/research-reports/financial-well-being-technical-report/

54. McLellan AT, Kushner H, Metzger D, et al. The fifth edition of the Addiction Severity Index. J Subst Abuse Treat. 1992;9(3):199-213. doi:10.1016/0740-5472(92)90062-s

55. Locatelli SM, Sharp LK, Syed ST, Bhansari S, Gerber BS. Measuring health-related transportation barriers in urban settings. J Appl Meas. 2017;18(2):178-193.

56. Weinstein S, Obuchowski NA, Lieber ML. Clinical evaluation of diagnostic tests. AJR Am J Roentgenol. 2005;184(1):14-19. doi:10.2214/ajr.184.1.01840014

57. Muschelli J. ROC and AUC with a binary predictor: a potentially misleading metric. J Classif. 2020;37(3):696-708. doi:10.1007/s00357-019-09345-1

58. Tenny S, Hoffman MR. Prevalence. In: StatPearls. StatPearls Publishing; 2023. Accessed April 10, 2023. http://www.ncbi.nlm.nih.gov/books/NBK430867/

59. de Hond AAH, Steyerberg EW, van Calster B. Interpreting area under the receiver operating characteristic curve. Lancet Digit Health. 2022;4(12):e853-e855. doi:10.1016/S2589-7500(22)00188-1

60. Innovative Approaches to Addressing Social Determinants of Health for Medicare Advantage Beneficiaries. Better Medicare Alliance Center for Innovation in Medicare Advantage; August 2021. Accessed April 27, 2023. https://bettermedicarealliance.org/wp-content/uploads/2021/08/Innovative-Approaches-to-Addressing-SDOH-for-MA-Beneficiaries-FINAL.pdf

61. Mazurenko O, Hirsh AT, Harle CA, McNamee C, Vest JR. Health-related social needs information in the emergency department: clinician and patient perspectives on availability and use. BMC Emerg Med. 2024;24(1):45. doi:10.1186/s12873-024-00959-2

62. 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 Chronic Dis. 2020;17:E45. doi:10.5888/pcd17.190339

63. Krieger N, Waterman P, Chen JT, Soobader MJ, Subramanian SV, Carson R. Zip code caveat: bias due to spatiotemporal mismatches between zip codes and US Census–defined geographic areas—the Public Health Disparities Geocoding Project. Am J Public Health. 2002;92(7):1100-1102. doi:10.2105/AJPH.92.7.1100

64. Greenwood-Ericksen M, DeJonckheere M, Syed F, Choudhury N, Cohen AJ, Tipirneni R. Implementation of health-related social needs screening at Michigan health centers: a qualitative study. Ann Fam Med. 2021;19(4):310-317. doi:10.1370/afm.2690

65. Kasthurirathne SN, Vest JR, Menachemi N, Halverson PK, Grannis SJ. Assessing the capacity of social determinants of health data to augment predictive models identifying patients in need of wraparound social services. J Am Med Inform Assoc. 2018;25(1):47-53. doi:10.1093/jamia/ocx130

66. Palacio A, Suarez M, Tamariz L, Seo D. A road map to integrate social determinants of health into electronic health records. Popul Health Manag. 2017;20(6):424-426. doi:10.1089/pop.2017.0019

67. Drewry MB, Yanguela J, Khanna A, et al. A systematic review of electronic community resource referral systems. Am J Prev Med. 2023;65(6):1142-1152. doi:10.1016/j.amepre.2023.06.001