Sage Journals: Discover world-class research

Abstract

Objective: To evaluate how and from where social risk data are extracted from EHRs for research purposes, and how observed differences may impact study generalizability. Methods: Systematic scoping review of peer-reviewed literature that used patient-level EHR data to assess 1 ± 6 social risk domains: housing, transportation, food, utilities, safety, social support/isolation. Results: 111/9022 identified articles met inclusion criteria. By domain, social support/isolation was most often included (N = 68/111), predominantly defined by marital/partner status (N = 48/68) and extracted from structured sociodemographic data (N = 45/48). Housing risk was defined primarily by homelessness (N = 39/49). Structured housing data was extracted most from billing codes and screening tools (N = 15/30, 13/30, respectively). Across domains, data were predominantly sourced from structured fields (N = 89/111) versus unstructured free text (N = 32/111). Conclusion: We identified wide variability in how social domains are defined and extracted from EHRs for research. More consistency, particularly in how domains are operationalized, would enable greater insights across studies.

Keywords

data mining electronic health records machine learning social determinants of health social domains social risk factors structured data unstructured data

Introduction

Over the past decade, healthcare initiatives to collect information about patients’ social risks, also known as adverse social determinants of health (SDH; e.g., food insecurity and housing instability) have expanded.^1,2 In theory, these efforts can inform work to intervene on social risks to improve health outcomes.^3–9 Despite growing efforts to collect and document social risk data in electronic health records (EHRs), there is significant variability in data collection and documentation.^10,11 This raises the possibility that research using EHR-sourced social risk data may operationalize the same social risk domain using different concepts, or operationalize the same domain using the same concept but vary on what data are used to support that concept. These both may impact researchers’ ability to conduct population comparisons and intervention comparative effectiveness studies across settings and populations.

Existing measures for social risk domains can track distinct concepts related to an individual domain.¹² For instance, some housing stability tools assess patients’ current housing status; others ask about lifetime experiences of homelessness; still others ask about overcrowding or housing quality concerns.¹³ Differences in how social risk domains are defined can present challenges to interpreting and comparing social risk data across settings.

Social risk domains can also be defined using the same concept but be based on different data with variable validity. Some healthcare settings ask patients about social risks using screening tools with standardized survey questions. Even among these screening tools, however, questions related to any one social risk domain are not standardized. As one example, food insecurity measures used in Kaiser Permanente’s Your Current Life Situation (YCLS) and the Centers for Medicare & Medicaid Innovation’s (CMS) Accountable Health Communities (AHC) Health-Related Social Needs (HRSN) Screening tools differ in the look-back timeframe (3-months vs 12-months, respectively) and in assessment of severity of risk (YCLS includes response options for “very often” and “often” worrying about food running out, whereas the AHC tool does not distinguish beyond “often”).¹⁴ Though a brief, validated (compared to gold standard measure) food insecurity screening measure exists,^12,15 many other social risk domains lack validated measures and measures are often adapted without additional validity testing.¹² Beyond screening tools, some researchers may depend on social risk data from unstructured free-text data collected more informally in the course of clinician verbal history-taking. If one study classifies patients as food insecure using a proxy of documented use of a food bank in a clinician note whereas another uses self-report of food insecurity in a patient screening tool, the reported data will be different: one tells you about whether a patient is using a food resource whereas the other the prevalence of patient food insecurity disclosure. These differences can influence conclusions evaluating how food insecurity is associated with patient health outcomes (e.g., patients who use a food resource may have a higher level of food insecurity than patients who report experiencing food insecurity, or vice versa if the patient using the resource is having their need addressed while the other patient is not).

In parallel, the way social risk data are documented also impacts how readily they can be extracted and evaluated. These differences in social risk content and documentation have implications for the usability and representativeness of social risk data, especially in cases where researchers aim to compare rates of social risk or the efficacy of an intervention across settings and populations. Data documented in structured fields (e.g., socio-demographic fields, screening tools with select answer options that are directly embedded in the EHR or integrated into the EHR after completion on a paper form) or in other standardized data (e.g., billing codes, International Classification of Disease (ICD) diagnosis codes) can be more readily extracted from EHRs. Identifying data from unstructured and non-standardized data is more challenging and has historically required manual review of free-text notes.^16–18 The emergence of advanced data analytic techniques such as natural language processing (NLP) offers alternatives to manual chart reviews, though their sensitivity, specificity, and scalability across settings has not yet been well-established.^19–21 Research studies that rely on structured versus unstructured data—or a combination of the two—are likely to calculate different levels of social risk based on methodological differences rather than actual differences between populations.

This review builds on two prior systematic reviews that explore SDH data captured in EHRs.^21,22 These reviews (Chen et al. and Patra et al.) evaluated the use of EHR-based SDH data in risk prediction modeling (including area/neighborhood-level data (i.e., social risks at the level of the community)) and NLP techniques used for social risks data extraction, respectively. This scoping review explores both the kinds of patient-level social risk data (i.e., experiences of social risk at the level of the individual patient) that are being extracted from EHRs for research purposes and the EHR locations where the data are found. We focused on six social risk domains that have been most commonly included in national healthcare sector discussions about social risk data collection and interventions: housing, food, transportation, utilities, personal safety, and social support/isolation. Due to the expansive and evolving nature of social risk screening and data extraction and lack of existing literature on the review topic, a systematic scoping review was appropriate for broadly synthesizing the literature and identifying knowledge gaps.²³ The goal was to better understand the comparability and generalizability of social risk data pulled from EHRs. A better understanding of the state of social risk data being used in research should inform future data collection efforts as well as subsequent design of social risk-related interventions. To our knowledge, this review is the first to examine a range of ways that researchers have extracted patient-level social risk data from EHRs and explored how these data are used to define specific social risk domains.

Methods

In collaboration with an experienced medical librarian, two study team members (E.H.D., N.V.S.) developed a search strategy to identify studies that included EHR-based social risk data. We further refined the search to maximize yield based on previously identified articles deemed relevant by expert referral.^24,25 This resulted in a three-concept search that included terms for 1) EHRs, 2) social risks, and 3) data extraction.

The social risks concept included both broad terms for the social determinants of health (SDH), as well as terms related to six specific social risk domains. Five of these domains—housing, food, transportation, utilities, and personal safety—are the core domains included in CMS’s AHC HRSN screening tool.²⁶ The AHC tool has been adopted by healthcare settings even outside the CMS Innovation Center’s demonstration as well as included in new quality measure designs by both the National Commission for Quality Assurance (NCQA) and CMS.^27,28 We added a sixth domain by also searching for terms related to social support/isolation given the recognition that social isolation is a significant driver of health and mortality.^29–31 Data extraction terms included the different techniques used to extract both structured (e.g., ICD codes or templated social risk fields) and unstructured data (e.g., social risk information included in free-text notes). See Appendix 1 for more information about the search strategy and a complete list of search terms.

We limited our review to the peer-reviewed literature. Databases searched included: PubMed, EMBASE, CINAHL, and Web of Science. We adapted the search strategy for each included database. To be included in our review, research studies had to analyze patient-level social risk data, defined as information on patients’ individual experiences of social risks, for at least one of six social risk domains (housing, food, transportation, utilities, personal safety, and social support/isolation). We did not include articles evaluating proxies for patient-level social risk data, such as area/neighborhood-level data pulled from publicly available sources. Studies that relied on social risk data from sources outside the EHR, such as the U.S. Census Bureau data, were excluded from our study. Our study focus was on evaluating data stored in the EHR and prior research has shown that individual-level social risk factors do not correlate highly with area-level measures.³² Articles had to be available in English and published in the peer-reviewed literature from 1/1/2009 to 7/20/2020. Our search was limited to articles published after 2009, based on the reasoning that the 2009 Health Information Technology for Economic and Clinical Health (HITECH) Act spurred widespread EHR adoption.^33,34 We did not distinguish between type of evidence included in our review; studies could include both quantitative and/or qualitative information.

Search results were uploaded to, and duplicates removed from, a group library in Zotero 5.0.96.2 reference manager. Title and abstract screening, followed by full-text review, were completed independently by three study team members (G.H.L., S.P., H.K.). Following full-text review, every study recommended by one of the reviewers was reviewed by an additional study team reviewer, such that the full-text of each article was reviewed by at least two study authors. Differences of opinion between reviewers at both the title/abstract and full-text level were resolved through group discussion and with additional input from a fourth study team member (E.H.D.).

Data extracted from each article included study design, setting, patient population, social risk factor(s) documented in the EHR, documentation location in the EHR, domain definition, method of data extraction, and outcomes evaluated (e.g., prevalence of social risks, patient health, healthcare utilization). The review was registered with the International Prospective Register of Systematic Reviews (CRD42020180539) and followed the Preferred Reporting Items for Systematic reviews and Meta-Analyses extension for Scoping Reviews.³⁵

Results

General characteristics

Our initial search yielded 9022 unique articles. Following an initial title/abstract screen, 133 articles underwent full-text review. (Figure 1). After full text review, 111 articles met inclusion criteria (Appendix Table 1). Most studies (N = 88/111, 79%) were from the United States (see Appendix Figure 1 for information on country of origin). Thirty percent of the studies (N = 33/111) extracted information for multiple social risk domains; the sum of the studies within each domain is therefore greater than the total number of included studies. The most frequently included social risk domain was social support/isolation (N = 68/111, 61%), followed by housing-related social risks (N = 49/111, 44%), personal safety concerns (N = 26/111, 23%), food insecurity (N = 10/111, 9%), transportation barriers (N = 9/111, 8%), and utilities insecurity (N = 5/111, 5%). (Appendix Table 2).

Figure 1.

PRISMA flow sheet.

The purpose for using patient-level social risk data differed across studies. The majority (N = 71/111, 64%) extracted social risk data to evaluate whether social risk factors were predictors for health or healthcare utilization outcomes. A smaller number of articles characterized social risks in the patient population (N = 16/111, 14%) or evaluated the performance of machine learning for extracting social risk factors from EHRs (N = 13/111, 12%). (Appendix Table 3).

Social risk domain definitions

Social risks related to housing, personal safety, social support/isolation, and transportation were not consistently defined across studies. For example, housing-related social risks were represented alternately as current or lifetime status of homelessness, housing instability, and/or housing quality. Food and utilities insecurity were consistently classified as their respective insecurities (i.e., lacked further sub-domains). (Table 1; Appendix Table 4).

Table 1.

Social risk domain breakdown across included studies (N = 111).

Social domain	Subdomains	N	%
Social support/isolation	Marital/partner status	48	70.6
	Living situation^a	11	16.2
	Involvement with community	9	13.2
	Self-reported loneliness or isolation	4	5.9
	Lack of social support not specified^b	9	13.2
Housing-related social risk	Homelessness	39	79.6
	Housing instability	25	51.0
	Housing quality	8	16.3
Personal safety concern^c	Physical abuse	22	80.8
	Sexual abuse	16	57.7
	Emotional abuse	14	53.8
Food insecurity	Food insecurity	10	100
Transportation barrier	Access to non-emergency medical transportation	7	77.8
Transportation barrier	Lack of transportation not specified^b	2	18.2
Utilities insecurity	Ability to pay for utilities	5	100

^aLiving situation refers to information about whether patients live alone versus with partner and/or roommates.

^bNot specified: articles did not provide adequate information to identify how the social risk domain was operationalized.

^cPersonal safety concern assessed both intimate partner violence (N = 18/26, 69.2%) and interpersonal violence (N = 3/26, 11.5%). Some articles did not include details about the perpetrator of violence (N = 5/26, 19.2%).

Data extraction methods

Across domains (e.g., housing) and subdomains (e.g., housing quality), different types of data were used to capture risks. Eighty percent of studies (N = 89/111) analyzed structured data whereas 29% (N = 32/111) analyzed unstructured data. These categories were not mutually exclusive; 17% studies (N = 19/111) analyzed both structured and unstructured data (Appendix Table 5). See Figure 2 and Appendix Table 6 for breakdown of data extraction method by social risk domain.

Figure 2.

Study use of structured versus unstructured data, by social risk domain^*.

Structured data

Of the 89 studies that relied on structured data for analysis, 53% (N = 47) used sociodemographic fields, 32% (N = 28) used billing codes, and 30% (N = 27) used social risk screening tools integrated into the EHR (which included screenings that were done on a paper form and manually entered into the EHR at a later time) (Appendix Table 5). By domain, 50% of the articles on housing-related social risks and personal safety concerns used information from billing codes (N = 15/30; N = 10/20; respectively). Screening tools were used in at least 50% of articles on food insecurity (N = 7/9, 78%), utilities insecurity (N = 3/4, 75%), personal safety concerns (N = 11/20, 55%), and transportation barriers (N = 4/8, 50%). Eighty-four percent (N = 47/57) of articles examining structured data on social support/isolation used sociodemographic data, 96% (N = 45/47) of which were used to determine marital status. (Figure 3(a); Appendix Table 7).

Figure 3.

Study data extraction methods, by social risk domain. (a) Sources of structured data, by social risk domain. (b) Sources of unstructured data, by social risk domain.

Unstructured data

Of the studies that used unstructured EHR data, 44% (N = 14/32) extracted the data from free-text notes using manual chart review and 66% (N = 21/32) used machine learning techniques (Appendix Table 5). By domain, machine learning techniques were used in 88% of the articles on social support/isolation (N = 14/16), 58% of the articles on housing-related social risks (N = 11/19), and 38% of the articles on personal safety (N = 3/8). Only one article pulled unstructured data on food insecurity and one on transportation barriers; both utilized machine learning. (Figure 3(b); Appendix Table 8).

Both structured and unstructured data

Of the studies that examined both structured and unstructured data, 37% (N = 7/19) compared the performance of structured and unstructured data for identifying social risk information.^{19,20,36–40} The remainder used both structured and unstructured data to increase the yield of social risk detection (N = 12/19, 63%).^41–52 Articles in this category extracted data pertaining to social support/isolation (N = 11/19, 58%),^{19,20,36,37,39,40,47–49,51,52} housing-related social risks (N = 10/19, 53%)^{19,20,38,41,44–46,49,51,52} and/or personal safety concerns (N = 5/19, 26%).^{42,43,48,50,52}

Discussion

This is a novel review that examines the techniques by which researchers are capturing patient-level social risk information from the EHR, along with the ways that data are defined into specific social risk domains. Our study is relevant and timely as many healthcare systems stand to benefit from a new crop of federal, state, and private payer incentives designed to increase screening and documentation of patients’ social risks.^27,28,53,54 These measures will likely increase the prevalence of standardized social risk screening and subsequent data availability to researchers.

We found only 27% of articles identified in this scoping review relied on data from integrated patient social risk screening tools. Although many common social risk screening tools include questions about transportation, food, and utilities, these domains were less represented than social support/isolation, housing-related risks, and personal safety in the patient-level data extracted from EHRs.^26,55 While screening tools are increasingly embedded into EHRs,^56,57 studies of screening programs frequently report using paper collection tools, which may not be consistently transferred to the EHR.^23,34–38 It is possible that data on food, transportation, and utilities securities are less frequently documented in the EHR than other domains, which is consistent with a previous study reporting low rates of food and utilities securities screening by healthcare providers,⁶⁴ the reasons for which are unclear. As multi-domain screening for social risks increases, we would expect an increase in documentation for food insecurity, transportation barriers, and utilities insecurity, provided integration of these tools within EHRs also increases.

Our findings also highlight how a lack of standardization in social risks definitions can influence social risk research outcomes and comparability of findings across studies. Many studies had differences in how they operationalized the same social risk domain or used very different data to support/refute the presence of a social risk. Others lacked detailed methodology explaining what social risk data were used to represent domains. For instance, several studies did not specify how or what data they used to define the term “social support” (e.g., it was unclear if they relied solely on martial/partner status from demographic screening, or if they used more robust information on interpersonal relationships and involvement with the community). Without this information, it is not possible to compare study results. The inconsistency of social risk domain definitions across studies limits the interpretation and synthesis of research findings, given that different studies may utilize very different representations of individual social risks. Further standardizing definitions of social risk domains or reporting requirements for social risk data would improve the ability to pool data and/or compare results across settings and populations. Organizations like the Gravity Project, which is trying to develop standards for social risk data for healthcare settings, can serve as a resource for researchers interested in capturing social risk data.⁷²

The method of extracting social risk data from the EHR also impacts the generalizability and comparability of research studies. Social support/isolation and housing-related risks were commonly extracted from both structured and unstructured fields, whereas food security and personal safety were primarily extracted from structured fields. While a growing number of studies are utilizing machine learning tools to extract free-text data from the EHR, these technologies are still in the research stage and limited by the quality of the written text.^21,73 Heterogeneity in where social risk data is documented and can be extracted from limits the accessibility of that data to both clinicians and researchers, even if shared definitions were to be used. The inadequate standardization in social risk data documentation and extraction limits the comparability of data across clinical settings and institutions and provides technological barriers to efficiently access social risk data. Until there are standards on how/where to document social risks within EHRs, researchers will likely need to mine EHRs for multiple types of data to achieve a more complete picture of patients’ social risks. We believe it is therefore important for researchers to be transparent and explicit about their extraction methods, as well as domain definitions, so that study results can be contextualized and compared only when appropriate.

Health systems face substantial barriers to implementing comprehensive social risk screening tools and standardized documentation of social risk factors. Barriers to screening described in the existing literature include workforce and time constraints.⁷⁴ These barriers likely contribute to both inconsistent documentation and under-documentation of social risk data,⁷⁵ which then impacts the availability and reliability of these data for both individual- and public health-level interventions, and research. It is also important to acknowledge the challenges and potential pitfalls of standardized documentation. Condensing a nuanced conversation about social risks between a patient and provider into a structured field could inappropriately oversimplify patients’ experiences.^73,76 Standardized documentation of social risks, however, may in fact facilitate more involved discussions between patients and providers about how patients’ experiences impact health and wellness. Moving forward, diverse stakeholder involvement (including patients, caregivers, and frontline healthcare team members), training and education of personnel, and necessary resources/infrastructure are vital to overcome the barriers of integrating social risk data into the EHR.^77,78

This scoping review builds on two prior systematic reviews. In contrast to a review from Chen et al, in this study we focus on six social risk domains that have been prioritized in recent government and private payer programs, rather than including any SDH used in risk prediction models. Furthermore, Chen et al. largely drew from studies that relied on neighborhood-level SDH data that were used in risk-prediction models.²² In contrast, in this study we explored the many ways researchers use patient-level social risk data from EHRs, including how individual social risks are both defined and extracted. We believe this study’s findings are especially relevant in the context of new government and private payer incentives (e.g., quality measures) that focus on patient-reported data. The review from Patra et al. synthesized studies that specifically used NLP to extract patient social risk data from unstructured EHR fields.²¹ While Patra et al. report on NLP methods used to identify social risk,¹⁹ our review examined a wider range of ways researchers have extracted social risks from both structured and unstructured EHR fields. In fact, NLP is currently only used to identify select social risk domains; given persistent EHR documentation barriers, we believe that efforts to identify social risks using a combination of structured and unstructured data may provide a more comprehensive picture of patients’ social risks.

Limitations

The review has several limitations that should be considered when interpreting findings. First, our academic database search was not designed to capture initiatives underway outside of research. Social risk data available outside of academia (e.g., healthcare systems or EHR vendors) may be more standardized than those used for research purposes due to data sharing policies and organization research priorities. Second, this is a rapidly evolving field, which means new articles have been published since the search was conducted. Since no new standards for social risk data in the EHR have been developed, however, we believe the findings are still relevant and should be used to prompt more standardization efforts. Third, we had low search specificity. In developing our search terms, we were unable to add or remove terms to reduce our yield without unacceptably reducing search sensitivity. As a scoping review, our priority was to broadly evaluate the literature²³ and therefore prioritize sensitivity. Lastly, we did not critically appraise each manuscript for risk of bias or other measure of quality. As a scoping review with a high volume of articles, our priority was to focus on the comparability of data based on how it was documented and extracted.

Conclusion

In this systematic scoping review, we explored how and what types of social risk data are extracted from EHRs for research. Evaluating how social risks are documented at the patient-level in the research literature is important and timely given the increased impetus and incentives to screen for and document patient social risk factors. Our findings highlight that current data documentation and extraction approaches have meant social risk data are not consistently comparable across populations and settings. Consensus on definitions and more standardized documentation practices cannot be achieved overnight. But as consensus emerges and implementation improves, social risk data will be both more reliable and generalizable. Standardization of efforts can improve both patient- and population-level activities to integrate social and medical care, by providing more consistent, reliable, and ideally complete information. Attention will need to focus on what data needs to be collected and where those data can and should be stored to ensure they meaningfully influence interventions, payment, and research that improve patient and population health.

Supplemental Material

Supplemental Material - Evaluating the comparability of patient-level social risk data extracted from electronic health records: A systematic scoping review

Supplemental Material for Evaluating the comparability of patient-level social risk data extracted from electronic health records: A systematic scoping review by Gaia H Linfield, Shyam Patel, Hee Joo Ko, Benjamin Lacar, Laura M Gottlieb, Julia Adler-Milstein, Nina V Singh, Matthew S Pantell and Emilia H De Marchis in Health Informatics Journal

Supplemental Material

Supplemental Material - Evaluating the comparability of patient-level social risk data extracted from electronic health records: A systematic scoping review

Footnotes

Acknowledgements

We would like to thank our medical librarian, Evans M. Whitaker, MD, MLIS, for his expert guidance on generating a search strategy for our scoping review. We would also like to thank the Robert Wood Johnson Foundation (RWJF) and Kaiser Permanente (KP) for their support of the UCSF Social Interventions Research & Evaluation Network (SIREN). L.M.G. is the founding co-director at SIREN and E.H.D. and M.P. are affiliate faculty.

Declaration of conflicting interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by the G.H.L. was supported by The David Vanderryn Memorial Fund Preceptorship/Project Program during summer 2020. S.P. was supported by the University of California, San Francisco Summer Explore Research Fellowship during summer 2020. B.L.’s work was supported by the Innovate for Health program, including the UC Berkeley Institute for Data Science, the UCSF Bakar Computational Health Sciences Institute, and Johnson & Johnson. M.P.’s work was funded by the NIH (Loan Repayment Program award 9L40-HD-106442-02A1), the Agency for Healthcare Research and Quality (award K12-HS-026383), and the National Center for Advancing Translational Sciences (award KL2-TR-001870).

ORCID iDs

Gaia H Linfield

Shyam Patel

Benjamin Lacar

Emilia H De Marchis

Supplemental material

Supplemental material for this article is available online.

References

Green

Zook

. When talking about social determinants, precision matters. Health Aff Forefr. 2019. doi:10.1377/forefront.20191025.776011

Center for Medicare & Medicaid Services . CMS issues new roadmap for states to address the social determinants of health to improve outcomes, lower costs, support state value-based care strategies. CMS. https://www.cms.gov/newsroom/press-releases/cms-issues-new-roadmap-states-address-social-determinants-health-improve-outcomes-lower-costs (accessed 23 March 2022).

Kangovi

Mitra

Grande

, et al. Evidence-based community health worker program addresses unmet social needs and generates positive return on investment. Health Aff (Millwood) 2020; 39: 207–213.

Berkowitz

Delahanty

Terranova

, et al. Medically tailored meal delivery for diabetes patients with food insecurity: a randomized cross-over trial. J Gen Intern Med 2019; 34: 396–404.

Rhodes

Simon

Hume

, et al. Safety-net accountable health model partnership drives inpatient connection to outpatient social services, reducing readmissions in a population experiencing homelessness. Prof Case Manag 2021; 26: 150–155.

DeLia

Nova

Chakravarty

, et al. Effects of permanent supportive housing on health care utilization and spending among New Jersey medicaid enrollees experiencing homelessness. Med Care 2021; 59: S199–S205.

Jones

VanWassenhove-Paetzold

Thomas

, et al. Impact of a fruit and vegetable prescription program on health outcomes and behaviors in Young Navajo Children. Curr Dev Nutr 2020; 4: nzaa109.

Pantell

Hessler

Long

, et al. Effects of in-person navigation to address family social needs on child health care utilization. JAMA Netw Open 2020; 3: e206445.

Tyris

Keller

Parikh

. Social risk interventions and health care utilization for pediatric asthma: a systematic review and meta-analysis. JAMA Pediatr 2022; 176: e215103.

10.

Bettencourt-Silva

Mulligan

Cullen

, et al. Bridging clinical and social determinants of health using unstructured data. Stud Health Technol Inform 2018; 255: 70–74.

11.

DeVoe

Bazemore

Cottrell

, et al. Perspectives in primary care: a conceptual framework and path for integrating social determinants of health into primary care practice. Ann Fam Med 2016; 14: 104–108.

12.

Henrikson

Blasi

Dorsey

, et al. Psychometric and pragmatic properties of social risk screening tools: a systematic review. Am J Prev Med 2019; 57: S13–S24.

13.

De Marchis

Ettinger de Cuba

Chang

, et al. Screening discordance and characteristics of patients with housing-related social risks. Am J Prev Med 2021; 61: e1–e12.

14.

Lewis

Wellman

Jones

SMW

, et al. Comparing the performance of two social risk screening tools in a vulnerable subpopulation. J Fam Med Prim Care 2020; 9: 5026–5034.

15.

Hager

Quigg

Black

, et al. Development and validity of a 2-item screen to identify families at risk for food insecurity. Pediatrics 2010; 126: e26–e32.

16.

Polnaszek

Gilmore-Bykovskyi

Hovanes

, et al. Overcoming the challenges of unstructured data in multi-site, electronic medical record-based abstraction. Med Care 2016; 54: e65–e72.

17.

Kreimeyer

Foster

Pandey

, et al. Natural language processing systems for capturing and standardizing unstructured clinical information: a systematic review. J Biomed Inform 2017; 73: 14–29.

18.

Hatef

Weiner

Kharrazi

. A public health perspective on using electronic health records to address social determinants of health: the potential for a national system of local community health records in the United States. Int J Med Inf 2019; 124: 86–89.

19.

Hatef

Rouhizadeh

Tia

, et al. Assessing the availability of data on social and behavioral determinants in structured and unstructured electronic health records: a retrospective analysis of a multilevel health care system. JMIR Med Inform 2019; 7: e13802.

20.

Navathe

Zhong

Lei

, et al. Hospital readmission and social risk factors identified from physician notes. Health Serv Res 2018; 53: 1110–1136.

21.

Patra

Sharma

Vekaria

, et al. Extracting social determinants of health from electronic health records using natural language processing: a systematic review. J Am Med Inf Assoc 2021; 28: 2716–2727.

22.

Chen

Tan

Padman

. Social determinants of health in electronic health records and their impact on analysis and risk prediction: a systematic review. J Am Med Inf Assoc 2020; 27: 1764–1773.

23.

Munn

Peters

MDJ

Stern

, et al. Systematic review or scoping review? Guidance for authors when choosing between a systematic or scoping review approach. BMC Med Res Methodol 2018; 18: 143.

24.

Levac

Colquhoun

O’Brien

. Scoping studies: advancing the methodology. Implement Sci 2010; 5: 1–9.

25.

Arksey

O’Malley

. Scoping studies: towards a methodological framework. Int J Soc Res Methodol 2005; 8: 19–32.

26.

Center for Medicare & Medicaid Services . The accountable health communities health-related social needs screening tool. https://innovation.cms.gov/files/worksheets/ahcm-screeningtool.pdf (accessed 6 July 2022).

27.

Center for Medicare & Medicaid Services . FY 2023 hospital inpatient prospective payment system (IPPS) and long-term care hospital prospective payment system (LTCH PPS) final rule — CMS-1771-F. https://www.cms.gov/newsroom/fact-sheets/fy-2023-hospital-inpatient-prospective-payment-system-ipps-and-long-term-care-hospital-prospective (2022, accessed 12 August 2022).

28.

National Committee for Quality Assurance . Proposed new measure for HEDIS measurement year (MY) 2023: social need screening and intervention (SNS-E). https://www.ncqa.org/wp-content/uploads/2022/02/04.-SNS-E.pdf (accessed 12 August 2022).

29.

Pantell

Rehkopf

Jutte

, et al. Social isolation: a predictor of mortality comparable to traditional clinical risk factors. Am J Public Health 2013; 103: 2056–2062.

30.

Kotwal

Holt‐Lunstad

Newmark

, et al. Social isolation and loneliness among san francisco bay area older adults during the COVID‐19 shelter‐in‐place orders. J Am Geriatr Soc 2021; 69: 20–29.

31.

Sepúlveda-Loyola

Rodríguez-Sánchez

Pérez-Rodríguez

, et al. Impact of social isolation due to COVID-19 on health in older people: mental and physical effects and recommendations. J Nutr Health Aging 2020: 1–10.

32.

Cottrell

Hendricks

Dambrun

, et al. Comparison of community-level and patient-level social risk data in a network of community health centers. JAMA Netw Open 2020; 3: e2016852.

33.

Adler-Milstein

Jha

. HITECH act drove large gains in hospital electronic health record adoption. Health Aff 2017; 36: 1416–1422.

34.

US Department of Health & Human Services . HITECH Act enforcement interim final rule. https://www.hhs.gov/hipaa/for-professionals/special-topics/hitech-act-enforcement-interim-final-rule/index.html (2017, accessed 6 July 2022).

35.

Tricco

Lillie

Zarin

, et al. PRISMA extension for scoping reviews (PRISMA-ScR): checklist and explanation. Ann Intern Med 2018; 169: 467–473.

36.

Anzaldi

Davison

Boyd

, et al. Comparing clinician descriptions of frailty and geriatric syndromes using electronic health records: a retrospective cohort study. BMC Geriatr 2017; 17: 248.

37.

Bucher

Shi

Pettit

, et al. Determination of marital status of patients from structured and unstructured electronic healthcare data. AMIA Annu Symp Proc 2019; 2019: 267–274.

38.

Feller

Bear Don’t Walk Iv

Zucker

, et al. Detecting social and behavioral determinants of health with structured and free-text clinical data. Appl Clin Inform 2020; 11: 172–181.

39.

Kan

Kharrazi

Leff

, et al. Defining and assessing geriatric risk factors and associated health care utilization among older adults using claims and electronic health records. Med Care 2018; 56: 233–239.

40.

Kharrazi

Anzaldi

Hernandez

, et al. The value of unstructured electronic health record data in geriatric syndrome case identification. J Am Geriatr Soc 2018; 66: 1499–1507.

41.

Amato

Nobay

Amato

, et al. Sick and unsheltered: homelessness as a major risk factor for emergency care utilization. Am J Emerg Med 2019; 37: 415–420.

42.

Bell

Gilyan

Moore

, et al. Long-term evaluation of a hospital-based violence intervention program using a regional health information exchange. J Trauma Acute Care Surg 2018; 84: 175–182.

43.

Feder

Davies

Baird

, et al. Identification and referral to improve safety (IRIS) of women experiencing domestic violence with a primary care training and support programme: a cluster randomised controlled trial. Lancet 2011; 378: 1788–1795.

44.

Feller

Zucker

Don’t Walk

, et al. Towards the inference of social and behavioral determinants of sexual health: development of a gold-standard corpus with semi-supervised learning. AMIA Annu Symp Proc 2018; 2018: 422–429.

45.

Hatef

Rouhizadeh

, et al. Assessing the impact of social needs and social determinants of health on health care utilization: using patient- and community-level data. Popul Health Manag. Epub ahead of print 2020. DOI: 10.1089/pop.2020.0043.

46.

Kanak

Stewart

Vinci

, et al. Trends in homeless children and young adults seeking shelter in a Boston pediatric emergency department following state housing policy changes, 2011–2016. Am J Public Health 2018; 108: 1076–1078.

47.

Kang

McHugh

Chittams

, et al. Utilizing home healthcare electronic health records for telehomecare patients with heart failure: a decision tree approach to detect associations with rehospitalizations. Comput Inform Nurs 2016; 34: 175–182.

48.

Klopfstein

Kamber

Zimmermann

. On the way to light the dark: a retrospective inquiry into the registered cases of domestic violence towards women over a six year period with a semi-quantitative analysis of the corresponding forensic documentation. Swiss Med Wkly 2010; 140: w13047.

49.

LaWall

Fan

, et al. Living alone and homelessness as predictors of 30-day potentially preventable hospital readmission. Prev Chronic Dis 2019; 16: E16.

50.

Ranney

Mello

. A comparison of female and male adolescent victims of violence seen in the emergency department. J Emerg Med 2011; 41: 701–706.

51.

Tulloch

Khondoker

Fearon

, et al. Associations of homelessness and residential mobility with length of stay after acute psychiatric admission. BMC Psychiatr 2012; 12: 121. Epub ahead of print 2012. DOI: 10.1186/1471-244X-12-121.

52.

Williams

Bossert

Chen

, et al. Influence of social determinants of health and substance use characteristics on persons who use drugs pursuit of care for hepatitis C virus infection. J Subst Abuse Treat 2019; 102: 33–39.

53.

Department of Health Care Services . CalAIM. https://www.dhcs.ca.gov/calaim (accessed 6 July 2022).

54.

Kaiser Family Foundation . A first look at North Carolina’s section 1115 Medicaid Waiver’s healthy opportunities pilots. https://www.kff.org/report-section/a-first-look-at-north-carolinas-section-1115-medicaid-waivers-healthy-opportunities-pilots-issue-brief/ (2019, accessed 6 July 2022).

55.

National Association of Community Health Center . PRAPARE: protocol for responding to and assessing patients’ assets, risks, and experiences.

56.

Nitkin

. A new way to document social determinants of health. Published June 16, 2017. https://www.hopkinsmedicine.org/office-of-johns-hopkins-physicians/best-practice-news/a-new-way-to-document-social-determinants-of-health (accessed 6 July 2022).

57.

National Association of Community Health Center . PRAPARE: protocol for responding to and assessing patients’ assets, risks, and experiences FAQ content. https://prapare.org/wp-content/uploads/2021/09/FAQ_Inventory-6.30.pdf (accessed 12 August 2022).

58.

Buitron de la Vega

Losi

Sprague Martinez

, et al. Implementing an EHR-based Screening and Referral System to Address Social Determinants of Health in Primary Care. Med Care 2019; 57: S133–S139.

59.

Gold

Bunce

Cowburn

, et al. Adoption of social determinants of health EHR tools by community health centers. Ann Fam Med 2018; 16: 399–407.

60.

Greenwood-Ericksen

DeJonckheere

Syed

, et al. Implementation of health-related social needs screening at michigan health centers: a qualitative study. Ann Fam Med 2021; 19: 310–317.

61.

Knowles

Khan

Palakshappa

, et al. Successes, challenges, and considerations for integrating referral into food insecurity screening in pediatric settings. J Health Care Poor Underserved 2018; 29: 181–191.

62.

Jones

Nolte

O’Brien

, et al. Factors related to providers screening children for behavioral health risks in primary care settings. J Pediatr Nurs 2021; 59: 37–44.

63.

Morgenlander

Tyrrell

Garfunkel

, et al. Screening for social determinants of health in pediatric resident continuity clinic. Acad Pediatr 2019; 19: 868–874.

64.

Fraze

Brewster

Lewis

, et al. 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: e1911514.

65.

Blosnich

Montgomery

Dichter

, et al. Social determinants and military veterans’ suicide ideation and attempt: a cross-sectional analysis of electronic health record data. J Gen Intern Med 2020; 35: 1759–1767.

66.

Chen

Dredze

Weiner

, et al. Extraction of geriatric syndromes from electronic health record clinical notes: assessment of statistical natural language processing methods. JMIR Med Inform 2019; 7: e13039. Epub ahead of print March 2019. DOI: 10.2196/13039.

67.

Chen

Dredze

Weiner

, et al. Identifying vulnerable older adult populations by contextualizing geriatric syndrome information in clinical notes of electronic health records. J Am Med Inf Assoc 2019; 26: 787–795.

68.

Logue

Smucker

Regan

. Admission data predict high hospital readmission risk. J Am Board Fam Med 2016; 29: 50–59.

69.

Primeau

Freund

Ramachandran

, et al. Social service barriers delay care among women with abnormal cancer screening. J Gen Intern Med 2014; 29: 169–175.

70.

Rabovsky

Rothberg

Rose

, et al. Content and outcomes of social work consultation for patients with diabetes in primary care. J Am Board Fam Med 2017; 30: 35–43.

71.

Trinacty

LaWall

Ashton

, et al. Adding social determinants in the electronic health record in clinical care in Hawai’i: supporting community-clinical linkages in patient care. Hawai‘i J Med Public Health 2019; 78: 46–51.

72.

Gravity Project . Our Mission. https://thegravityproject.net/overview/ (accessed 6 July 2022).

73.

Vale

Perkins

. Discuss and remember: clinician strategies for integrating social determinants of health in patient records and care. Soc Sci Med 1982 2022; 315: 115548.

74.

De Marchis

Brown

Aceves

. State of the science on social screening in healthcare settings. https://sirenetwork.ucsf.edu/sites/default/files/2022-06/final_SCREEN_State-of-Science-Report%5B55%5D.pdf (2022, accessed 12 August 2022).

75.

Torres

Lawlor

Colvin

, et al. ICD social codes: an underutilized resource for tracking social needs. Med Care 2017; 55: 810–816.

76.

Cruz

Paine

. Capturing patients, missing inequities: data standardization on sexual orientation and gender identity across unequal clinical contexts. Soc Sci Med 1982 2021; 285: 114295.

77.

Wark

Woodbury

LaBrie

, et al. Engaging stakeholders in social determinants of health quality improvement efforts. Perm J 2022; 0: 1–11.

78.

Iott

Pantell

Adler-Milstein

, et al. Physician awareness of social determinants of health documentation capability in the electronic health record. J Am Med Inf Assoc 2022; 29: 2110–2116. ocac154.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.21 MB

0.07 MB