Food Insecurity Identification Modeling for Medicare Enrollees Using Administrative Data

ABSTRACT

Objectives: Food insecurity is a critical health-related social need (HRSN) disproportionately impacting vulnerable populations. Although health care systems are expected to address social needs, many lack the infrastructure for universal screening. Predictive modeling offers a scalable alternative for targeting individuals. This study aimed to develop and evaluate a parsimonious, mixed-effects model to predict the likelihood of food insecurity among Elevance Health Medicare Advantage beneficiaries using administrative and social risk data.

Study Design: Retrospective cohort study using hierarchical generalized linear mixed modeling with cross-validation.

Methods: We analyzed data from 462,251 unique Medicare Advantage members with completed HRSN assessments between January 2021 and June 2024. Food insecurity, the dependent variable, was defined using self-reported health risk assessments. Predictors included demographic characteristics, prior social needs coded according to Logical Observation Identifiers Names and Codes, dual Medicare-Medicaid enrollment, Social Vulnerability Index (SVI) tertiles, and disability status. Models incorporated random intercepts by Medicare market state. Model performance was evaluated using 10-fold cross-validation.

Results: The final model demonstrated strong predictive performance (area under the curve = 0.82; SD = 0.002). The most influential predictors were documentation of multiple prior social needs (β = 3.52; 95% CI, 3.47-3.59) and dual enrollment (β = 2.96; 95% CI, 2.88-3.06). Chronic conditions were not significantly associated with food insecurity. SVI and disability status also contributed meaningfully.

Conclusions: This mixed-effects model offers a scalable strategy for identifying food insecurity risk using existing data sources. It may enable managed care organizations to better target interventions and improve equity in addressing unmet social needs.

Am J Manag Care. 2026;32(5):e141-e146. https://doi.org/10.37765/ajmc.2026.89939

Takeaway Points

Identification of health-related social needs is becoming an increasingly rich opportunity among health care organizations for reducing individual and systemic health care burdens.

• We offer a methodologically informed prediction model to identify adult Medicare enrollees’ likelihood of food insecurity for intervention use with social risk assessments and administrative data.
• By targeting food insecurity with probabilistic methods, we aim to prioritize member outreach scaled to match intervention program needs.

Unmet social needs have been associated with missed medical appointments, more frequent emergency department use, hospital readmission,^1,2 and increased health care expenditures.³ There is growing evidence that social interventions can improve access to preventive health care, enhance chronic disease management,⁴ reduce hospital admissions,^5,6 and lower health care costs.^7-9 Research suggests that when people face adverse social conditions that impact their ability to maintain good health, or health-related social needs (HRSNs), they make difficult trade-off decisions with more flexible expenses, primarily food.¹⁰ These unmet needs, such as food insecurity, have been associated with depression, diabetes, and other chronic health conditions.^11-15 Identifying and addressing social needs, particularly in vulnerable populations, is essential to improving health outcomes.^16-18 Older adults are especially vulnerable, given their increased risk for acute and chronic health conditions. Additionally, a large percentage of older adults live on a fixed income and are often forced to make spending trade-offs.¹⁹

Current approaches to social needs intervention in US health care settings often rely on universal screening of patients, which poses logistic and operational problems for managed care organizations (MCOs).²⁰ An alternative is to implement risk prediction strategies that proactively identify and prioritize patients at highest risk for HRSNs.

Our approach integrates diverse data sources, including health risk assessments (HRAs), to better detect and flag patients who are likely experiencing food insecurity. An opportunity exists to better understand the potential for integrating risk prediction to proactively identify patients in need of further social need assessment or intervention. Applying predictive models to anticipate HRSNs is a relatively novel strategy within health care analytics. This method is particularly impactful given that effective screening is a prerequisite to meaningful intervention.²¹

HRSNs remain underdocumented in medical claims data. Additionally, the US Preventive Services Task Force (USPSTF) recently concluded that there is insufficient evidence to assess the balance of benefits and harms of screening for food insecurity on health outcomes in primary care.²² Historically, USPSTF grades have been influential in guiding provider behavior. Practices not strongly recommended by USPSTF are often less likely to be widely adopted. Considering these limitations, we propose a predictive model that incorporates administrative claims and publicly available data to identify members at highest risk of acute food insecurity. This study aims to develop and validate a predictive model to prioritize Medicare Advantage (MA) members at risk of food insecurity for further screening and targeted outreach.

METHODS

Data Description

We used data from Elevance Health (EH; formerly Anthem Inc) MA-enrolled members with HRSN assessments between January 2021 and June 2024, consisting of 462,251 unique adult members.

The dependent variable was the presence of food insecurity identified through HRSN screenings, which we refer to as F-HRSN. A variety of assessments have been used to identify HRSNs within Medicare, each potentially with different scripts, wording, and responses. Our social needs data warehouse stores both industry-standard questions mapped to the corresponding Logical Observation Identifier Names and Codes (LOINC) codes and nonstandard questions mapped to an internally defined member attribute taxonomy. To leverage all the available data, we used wildcard searches for food insecurity–related questions on the HRAs. Some of the searches included food would run out, did not last, didn’t last, food you bought, and food we bought. Other criteria included difficulty with food access, food affordability, or being sufficiently fed to align with variations from the Hunger Vital Sign.^23,24

Other independent variables were included in testing based on stepwise improvement of Akaike information criteria (AIC) and significant likelihood ratio tests. Social needs are defined as having a documented social LOINC code prior to, or within 1 calendar year of, being screened by an HRA. LOINC codes have been established as a reliable metric for medical clinical documentation.^25,26 A total count of social determinant–related LOINC codes was used as a measure of total HRSN load, based on select supplemental questions in the following LOINC categories: social determinant of health (SDH) education, SDH health care, SDH community, SDH economics, SDH environment, SDH food, and smoking-related terms.

LOINC identifiers used to compile the total HRSN load were pulled separately from data used to assess F-HRSN. International Statistical Classification of Diseases, Tenth Revision codes for HRSNs, also called Z-codes, were not used because each Z-code was expected to contain a LOINC referent.²⁷

To assess the cumulative burden of social needs among members, we constructed a categorical variable representing the number of distinct social need domains identified via LOINC. This variable was derived based on the count of unique social need types documented through LOINC-coded responses within 1 calendar year before the HRA. Members were assigned to 1 of 4 mutually exclusive categories: 0 social needs identified, 1 social need identified, 2 social needs identified, or 3 or more social needs identified. This classification reflects the extent of social risk burden and enables stratified analysis of how multiple co-occurring social needs relate to food insecurity.

Chronic clinical conditions were assigned to each member based on their impact on overall costs. Three of these clinical conditions were included in the model: diabetes, cardiovascular, and none. Although only 1 condition is assigned to each member, they may have more than 1. This approach focuses the analysis on how a member’s primary chronic condition may influence social risk factors, particularly food insecurity.

Two indicators from CMS²⁸ were also included: a low-income subsidy indicator and an indicator of disability. Medicare-Medicaid plan was used as an indicator of dual Medicare-Medicaid enrollment. There are several types of care management (CM) programs offered to EH members, including transition of care, complex care, and care coordination. If the status of the most recent CM case was flagged as open, active, new, or specialty, the member was considered active with a CM program.

Control variables considered Social Vulnerability Index (SVI) tertiles,²⁹ which were mapped to individual members’ census tract locations based on enrollment addresses and coded as low (0-0.30), medium (0.30-0.60), or high (> 0.60), with low as the reference group. Metropolitan area based on US Department of Agriculture (USDA) Economic Research Service designation of rural-urban status was also included. Each respective indicator was dummy coded, where the absent or lowest rank condition served as the reference category. Demographic variables included member age, sex (female coded as reference category), and minority race/ethnicity status. EH employs a prioritization method for determining racial and ethnic identity. Racial identity is determined by using standard values derived from hierarchical logic, prioritizing the highest quality sources from enrollment to clinically acquired data to RAND imputation, aligned to Office of Management and Budget standards from the federal government. Approximately 10% of member race and/or ethnicity is imputed using a Bayesian improved surname geocoding method. Minority status refers to categories other than White. The USDA reported in 2024 that there were meaningful differences in food insecurity across and within racial and ethnic groups.³⁰ Medicare market states, covering 26 US states referred to as market states, were used as the random intercept.

Model Design

Hierarchical linear modeling (HLM) is increasingly recognized as a valuable approach for analyzing health determinants, particularly when accounting for complex, multilevel influences, such as social determinants of health (SDOH). This modeling framework enables the partitioning of variability across nested data structures—such as individuals within geographic regions—allowing for the simultaneous estimation of both fixed effects (population-level influences) and random effects (group-level variation). Given that SDOH often operate at multiple contextual levels, single-layer analyses may oversimplify these dynamics. HLM addresses this by capturing both overall associations with the outcome variable and variability across subgroups, thereby reducing the risk of assuming homogeneous error distributions across inherently diverse populations.^31,32

Data were split into training and testing sets based on 60% and 40% splits. Analyses were performed using R 4.4.1 (R Foundation for Statistical Computing). Generalized linear mixed models fit by maximum likelihood were performed using package lme4.³³

Model Evaluation

Model performance was assessed using likelihood ratio tests comparing the null model on training data to a series of nested models increasing in complexity, as follows:

Model 1: Fixed effects only (baseline model)

Model 2: Random intercept for Medicare market state

Model 3: Random intercept for member age nested within Medicare market state

Model 4: Random slope for coded social needs and random intercept for member age nested within Medicare market state

Additional model validation was performed with a 10-fold cross-validation. The data set was randomly partitioned into 10 equal subsets; in each iteration, 9 subsets were used for model training and 1 for testing. This procedure was repeated 10 times to ensure robust assessment of model generalizability.

Predictions were classified using a fixed probability threshold of 0.51 across all iterations. To evaluate model performance, we computed a comprehensive set of metrics, including mean absolute error, root mean square error (RMSE), and area under the receiver operating characteristic curve (AUC). Additional classification performance measures included Cohen κ, sensitivity, specificity, positive predictive value (PPV), and negative predictive value, aggregated across all folds.

RESULTS

Descriptive statistics for model predictors are presented in Table 1. Likelihood ratio tests demonstrated that model 2, which included a random intercept for Medicare market state, significantly outperformed the fixed-effects–only model (model 1) (χ²₁₈ = 2049.00; P < .0001; AIC = 264,705.2). Furthermore, model 2 also outperformed model 3, which included a random intercept for age nested within market state (χ²₁₈ = 160.58; P < .0001; AIC = 264,865.8). Model 4 did not converge due to singularity, likely reflecting overparameterization. Model 2 revealed between-state variation in baseline food insecurity risk, with an SD of 0.352 (variance = 0.124) for market-level intercepts. This indicates that substantial heterogeneity persists across states after adjusting for individual-level fixed effects. Conditional and marginal R² values indicated moderate explanatory power (0.417 and 0.394, respectively), and the RMSE was 0.397 on the test set.

Fixed Effects

Among all predictors, prior social needs identification was the strongest predictor of food insecurity (Table 2). Having 3 or more social needs was associated with a large positive effect (β = 3.52; 95% CI, 3.47-3.59), followed by 2 needs (β = 3.18; 95% CI, 3.15-3.22) and 1 need (β = 2.56; 95% CI, 2.53-2.60). Dual Medicare-Medicaid enrollment also showed a strong positive association (β = 2.96; 95% CI, 2.88-3.06). Chronic conditions, including diabetes (β = 0.005; 95% CI, –0.04 to 0.05) and cardiovascular disease (β = –0.021; 95% CI, –0.06 to 0.02), were not statistically significant, whereas the absence of a chronic condition was positively associated (β = 0.125; 95% CI, 0.09-0.16). SVI scores in the high (β = 0.107; 95% CI, 0.07-0.14) and medium (β = 0.051; 95% CI, 0.01-0.09) ranges were both significantly associated with food insecurity. The CMS low-income subsidy flag showed a negative association (β = –0.316; 95% CI, –0.35 to –0.28), whereas disability status was positively associated (β = 0.194; 95% CI, 0.16-0.22). Minority race/ethnicity (β = 0.077; 95% CI, 0.05-0.10) was statistically significant, whereas metropolitan area status was not (β = 0.01; 95% CI, –0.02 to 0.03). Both age (β = –0.008; 95% CI, –0.01 to –0.01) and male sex (β = –0.024; 95% CI, –0.05 to 0.00) were significantly negatively associated.

Cross-Validation Results

Results from the 10-fold cross-validation are summarized in Table 3. The model demonstrated consistent performance across iterations, with a mean (SD) AUC of 0.82 (0.002) and a mean (SD) PPV of 0.61 (0.004). No evidence of multicollinearity was detected, as assessed by variance inflation factors and tolerance values.

DISCUSSION

Our model identified multiple social needs and dual Medicare-Medicaid enrollment as the strongest predictors of F-HRSN identification, with meaningful variability observed across Medicare market states. These findings underscore the role of nonclinical risk factors in identifying members vulnerable to food insecurity. Notably, chronic conditions such as diabetes and cardiovascular disease were not significantly associated with food insecurity risk in our model, highlighting the relative importance of social and environmental context. Early indicators, including documented HRSN responses in HRAs and relevant LOINC codes, may provide MCOs with valuable signals for initiating proactive interventions.²³

Recent guidance from CMS and the National Committee for Quality Assurance has emphasized the importance of addressing food insecurity as a health equity priority.^34-37 Although many Medicare MCOs have expanded their screening infrastructure—often integrating social needs questions into electronic health record workflows—implementation remains inconsistent across plans. Barriers such as resource constraints, insufficient staff training, and variation in prioritizing SDOH continue to limit the reach and effectiveness of these efforts. Moreover, even when needs are identified, challenges persist in linking members to services such as the Supplemental Nutrition Assistance Program or meal delivery due to fragmented referral networks and the lack of closed-loop referral systems.^22,37

Metrics for evaluating the impact of food insecurity interventions on health outcomes and cost reduction are still evolving, limiting MCOs’ ability to justify and scale such programs. As a result, although there has been progress in identifying food insecurity, sustained investment, regulatory alignment, and infrastructure development are needed to ensure timely, equitable, and effective responses across Medicare populations.³⁸

Limitations

Although we were able to leverage the strengths of HLM to account for the complexity of SDOH within our analysis, it also faced several limitations. The analytic sample used for modeling consisted of 462,251 MA members who had completed an HRSN assessment, representing approximately 29% of the total eligible population of 1.6 million members. Although this is a large and valuable subsample, it may not fully capture the diversity and variability of the broader MA population. Consequently, predictive modeling based on this subset introduces potential biases, including the risk of overfitting. However, results from 10-fold cross-validation suggest that the model demonstrates strong internal validity, with consistent predictive performance across folds. Specifically, the AUC of 0.82 indicates strong discriminatory power in identifying individuals with F-HRSN.

Including the SVI as a random variable did not substantively improve model fit. Although composite indices such as the SVI are increasingly used to predict health care needs, their aggregation may obscure variation in constituent components.³⁹ As Kaalund et al noted, vulnerability indices are most effective when disaggregated into their core dimensions.⁴⁰ In our analysis, the fixed effects of SVI tertiles were statistically significant, but treating SVI as a grouping variable in the random-effects structure failed to improve the model’s log likelihood. This may reflect the ecological fallacy: attributing individual-level conclusions based on group-level data. Researchers are advised to apply area-based indices with caution and consider multilevel designs or disaggregation where appropriate.

Finally, although standardized tools exist for assessing food insecurity, we observed considerable variability in HRA implementations across the data set. In particular, the lookback period of food insecurity questions varied from the past 30 days to 12 months. As a result, we defined food insecurity using the wildcard search of self-reported HRA responses as the dependent variable while clinical indicators of additional documented social needs (coded via medical encounter) were included as predictors. Although these variables were derived from distinct data sources, they demonstrated a moderate correlation (r_462,247 = 0.461; P < .001), suggesting some overlap between identified food insecurity and other documented social needs, reflecting a shared underlying construct. In a recent study, Chatterjee et al found that Z-code use patterns likely underrepresent social risk among clinically complex patients.⁴¹ Future studies may benefit from harmonizing assessment tools and exploring the temporal alignment of HRSN measures across clinical and administrative data streams.

CONCLUSIONS

As health care in the US continues to prioritize HRSN, the demand for scalable, data-driven screening strategies will increase. Many current programs lack the infrastructure to conduct universal screening. To address this gap, we propose a theoretically grounded, technically efficient mixed-effects model to identify members at elevated risk of food insecurity. Our findings indicate that the strongest predictors of food insecurity identification include prior social needs–related documentation and dual Medicare-Medicaid enrollment, with additional contributions from medium to high SVI scores and disability status. By enabling targeted member identification, this approach can support health care administrators in improving the efficiency and reach of intervention programs, expanding outreach capacity, and better addressing the needs of vulnerable populations.

Author Affiliations: Elevance Health Inc (JW, DH), Indianapolis, IN.

Source of Funding: None.

Author Disclosures: Dr Wrathall and Ms Hersh contributed to this article as part of their terms of employment with Elevance Health.

Authorship Information: Concept and design (JW, DH); acquisition of data (JW); analysis and interpretation of data (JW, DH); drafting of the manuscript (JW, DH); critical revision of the manuscript for important intellectual content (JW, DH); statistical analysis (JW, DH); and administrative, technical, or logistic support (JW, DH).

Address Correspondence to: Debra Hersh, MPH, Elevance Health, 551 Siena Ln, Glen Allen, VA 23059. Email: debra.hersh@elevancehealth.com.

REFERENCES

1. Berkowitz SA, Seligman HK, Meigs JB, Basu S. Food insecurity, healthcare utilization, and high cost: a longitudinal cohort study. Am J Manag Care. 2018;24(9):399-404.

2. McQueen A, Li L, Herrick CJ, et al. Social needs, chronic conditions, and health care utilization among Medicaid beneficiaries. Popul Health Manag. 2021;24(6):681-690. doi:10.1089/pop.2021.0065

3. Palakshappa D, Garg A, Peltz A, Wong CA, Cholera R, Berkowitz SA. Food insecurity was associated with greater family health care expenditures in the US, 2016-17. Health Aff (Millwood). 2023;42(1):44-52. doi:10.1377/hlthaff.2022.00414

4. Hill-Briggs F, Adler NE, Berkowitz SA, et al. Social determinants of health and diabetes: a scientific review. Diabetes Care. 2020;44(1):258-279. doi:10.2337/dci20-0053

5. Cassarino M, Robinson K, Quinn R, et al. Impact of early assessment and intervention by teams involving health and social care professionals in the emergency department: a systematic review. PLoS One. 2019;14(7):e0220709. doi:10.1371/journal.pone.0220709

6. Hatef E, Ma X, Rouhizadeh M, Singh G, Weiner JP, Kharrazi H. Assessing the impact of social needs and social determinants of health on health care utilization: using patient- and community-level data. Popul Health Manag. 2021;24(2):222-223. doi:10.1089/pop.2020.0043

7. Knighton AJ, Stephenson B, Savitz LA. Measuring the effect of social determinants on patient outcomes: a systematic literature review. J Health Care Poor Underserved. 2018;29(1):81-106. doi:10.1353/hpu.2018.0009

8. Berkowitz SA, Baggett TP, Edwards ST. Addressing health-related social needs: value-based care or values-based care? J Gen Intern Med. 2019;34(9):1916-1918. doi:10.1007/s11606-019-05087-3

9. 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

10. Leroux J, Morrison K, Rosenberg M. Prevalence and predictors of food insecurity among older people in Canada. Int J Environ Res Public Health. 2018;15(11):2511. doi:10.3390/ijerph15112511

11. Silverman J, Krieger J, Kiefer M, Hebert P, Robinson J, Nelson K. The relationship between food insecurity and depression, diabetes distress and medication adherence among low-income patients with poorly controlled diabetes. J Gen Intern Med. 2015;30(10):1476-1480. doi:10.1007/s11606-015-3351-1

12. Chambers EC, McAuliff KE, Heller CG, Fiori K, Hollingsworth N. Toward understanding social needs among primary care patients with uncontrolled diabetes. J Prim Care Community Health. 2021;12:2150132720985044. doi:10.1177/2150132720985044

13. Nagata JM, Palar K, Gooding HC, Garber AK, Bibbins-Domingo K, Weiser SD. Food insecurity and chronic disease in US young adults: findings from the national longitudinal study of adolescent to adult health. J Gen Intern Med. 2019;34(12):2756-2762. doi:10.1007/s11606-019-05317-8

14. Venci BJ, Lee SY. Functional limitation and chronic diseases are associated with food insecurity among U.S. adults. Ann Epidemiol. 2018;28(3):182-188. doi:10.1016/j.annepidem.2018.01.005

15. Jih J, Stijacic-Cenzer I, Seligman HK, Boscardin WJ, Nguyen TT, Ritchie CS. Chronic disease burden predicts food insecurity among older adults. Public Health Nutr. 2018;21(9):1737-1742. doi:10.1017/S1368980017004062

16. Cole MB, Nguyen KH. Unmet social needs among low-income adults in the United States: associations with health care access and quality. Health Serv Res. 2020;55(suppl 2):873-882. doi:10.1111/1475-6773.13555

17. 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.

18. Byhoff E, Freund KM, Garg A. Accelerating the implementation of social determinants of health interventions in internal medicine. J Gen Intern Med. 2018;33(2):223-225. doi:10.1007/s11606-017-4230-8

19. Mavegam Tango Assoumou BO, Coughenour C, Godbole A, McDonough I. Senior food insecurity in the USA: a systematic literature review. Public Health Nutr. 2023;26(1):229-245. doi:10.1017/S1368980022002415

20. Chisolm DJ, Brook DL, Applegate MS, Kelleher KJ. Social determinants of health priorities of state Medicaid programs. BMC Health Serv Res. 2019;19(1):167. doi:10.1186/s12913-019-3977-5

21. 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):4454. doi:10.1038/s41598-022-08344-4

22. Nicholson WK, Silverstein M, Wong JB, et al; US Preventive Services Task Force. Screening for food insecurity: US Preventive Services Task Force recommendation statement. JAMA. 2025;333(15):1333-1339. doi:10.1001/jama.2025.0879

23. Gunderson C, Engelhard EE, Crumbaugh AS, Seligman HK. Brief assessment of food insecurity accurately identifies high-risk US adults. Public Health Nutr. 2017;20(8):1367-1371. doi:10.1017/S1368980017000180

24. 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-e32. doi:10.1542/peds.2009-3146

25. Frazier P, Rossi-Mori A, Dolin RH, Alschuler L, Huff SM. The creation of an ontology of clinical document names. Stud Health Technol Inform. 2001;84(pt 1):94-98.

26. Bhargava A, Kim T, Quine DB, Hauser RG. A 20-year evaluation of LOINC in the United States’ largest integrated health system. Arch Pathol Lab Med. 2020;144(4):478-484. doi:10.5858/arpa.2019-0055-OA

27. Gravity Project. Accessed June 9, 2025. https://thegravityproject.net/

28. Data dictionaries. Chronic Conditions Data Warehouse. Accessed April 8, 2026. https://www2.ccwdata.org/web/guest/data-dictionaries

29. Social Vulnerability Index. CDC. July 22, 2024. Accessed September 9, 2024. https://www.atsdr.cdc.gov/place-health/php/svi/index.html

30. Hales LJ, Coleman-Jensen A. Household Food Insecurity Across Race and Ethnicity in the United States, 2016-21. US Department of Agriculture Economic Research Service report EIB-269. April 2024. Accessed September 9, 2024. https://ers.usda.gov/sites/default/files/_laserfiche/publications/108905/EIB-269.pdf?v=59861

31. Bryk AS, Raudenbush SW. Hierarchical Linear Models: Applications and Data Analysis Methods. Sage Publications Inc; 1992.

32. Luke DA, Krauss M. Where there’s smoke there’s money: tobacco industry campaign contributions and U.S. Congressional voting. Am J Prev Med. 2004;27(5):363-372. doi:10.1016/j.amepre.2004.08.014

33. Bates D, Mächler M, Bolker B, Walker S. Fitting linear mixed-effects models using lme4. J Stat Softw. 2015;67(1):1-48. doi:10.18637/jss.v067.i01

34. Proposed changes to existing measure for HEDIS MY 2026: social need screening and intervention (SNS-E). National Committee for Quality Assurance. 2025. Accessed June 9, 2025. https://wpcdn.ncqa.org/www-prod/04.-SNS-E.pdf

35. NCQA responds to CMS list of screening instruments for housing stability, food security, and transportation questions on health risk assessments. National Committee for Quality Assurance. December 22, 2022. Accessed June 9, 2025. https://www.ncqa.org/comment-letter/ncqa-responds-to-cms-list-of-screening-instruments-for-housing-stability-food-security-and-transportation-questions-on-health-risk-assessments/

36. Addressing food and nutritional insecurity a use case in leveraging the value-based insurance design (VBID) model. CMS. Accessed June 9, 2025. https://www.cms.gov/priorities/innovation/media/document/vbid-cy2023-food-insecurity-use-case

37. Jordanova KE, Suresh A, Canavan CR, et al. Addressing food insecurity in rural primary care: a mixed-methods evaluation of barriers and facilitators. BMC Prim Care. 2024;25(1):163. doi:10.1186/s12875-024-02409-1

38. McCurley JL, Fung V, Levy DE, et al. Assessment of the Massachusetts flexible services program to address food and housing insecurity in a Medicaid accountable care organization. JAMA Health Forum. 2023;4(6):e231191. doi:10.1001/jamahealthforum.2023.1191

39. Mah JC, Penwarden JL, Pott H, Theou O, Andrew MK. Social vulnerability indices: a scoping review. BMC Public Health. 2023;23(1):1253. doi:10.1186/s12889-023-16097-6

40. Kaalund K, Thoumi A, Bhavsar NA, Labrador A, Cholera R. Assessment of population-level disadvantage indices to inform equitable health policy. Milbank Q. 2022;100(4):1028-1075. doi:10.1111/1468-0009.12588

41. Chatterjee P, Macneal E, Roberts E. Measurement bias in documentation of social risk among Medicare beneficiaries. JAMA Health Forum. 2025;6(7):e251923. doi:10.1001/jamahealthforum.2025.1923