Evaluation of Structured Knowledge Parts for Social Threat Elements
Aims: Computable social threat issue phenotypes derived from routinely collected structured digital well being document (EHR) or well being data change (HIE) knowledge could signify a possible and strong strategy to measuring social components. This research convened an professional panel to determine and assess the standard of particular person EHR and HIE structured knowledge parts that may very well be used as parts in future computable social threat issue phenotypes.
Examine Design: Technical professional panel.
Strategies: A 2-round Delphi method included 17 consultants with an in-depth data of obtainable EHR and/or HIE knowledge. The primary-round identification periods adopted a nominal group strategy to generate candidate knowledge parts which will relate to socioeconomics, cultural context, social relationships, and neighborhood context. Within the second-round survey, panelists rated every knowledge aspect in keeping with total knowledge high quality and probability of systematic variations in high quality throughout populations (ie, bias).
Outcomes: Panelists recognized a complete of 89 structured knowledge parts. About half of the information parts (n = 45) had been associated to socioeconomic traits. The panelists recognized a various set of knowledge parts. Parts utilized in reimbursement-related processes had been usually rated as increased high quality. Panelists famous that a number of knowledge parts could also be topic to implicit bias or mirror biased programs of care, which can restrict their utility in measuring social components.
Conclusions: Routinely collected structured knowledge inside EHR and HIE programs could mirror affected person social threat components. Figuring out and assessing accessible knowledge parts serves as a foundational step towards growing future computable social issue phenotypes.
Am J Manag Care. 2022;28(1):e14-e23. https://doi.org/10.37765/ajmc.2022.88816
Computable phenotypes are measurements of affected person circumstances or traits that may be obtained from current knowledge by combining an outlined set of variables and logical expressions. Routinely collected structured knowledge inside digital well being information and well being data change programs are reflective of traits of social and financial well-being and thus could also be amenable to make use of in social threat issue phenotype growth. An professional panel recognized and assessed structured knowledge parts to help future growth of social threat issue computable phenotyping.
- Computable phenotypes signify a further methodology of measuring affected person social components that leverages current knowledge sources and workflows.
- Structured knowledge parts utilized in reimbursement-related processes could also be of the best high quality to be used in phenotyping.
- At the moment collected structured knowledge parts reminiscent of Worldwide Classification of Ailments, Tenth Revision Z codes and Logical Remark Identifiers Names and Codes are doubtlessly prone to bias.
- Computable phenotyping would require reworking or combining knowledge parts into novel and doubtlessly extra informative measures.
Social threat components embrace sufferers’ nonclinical, financial, and contextual traits which will adversely have an effect on well being.1-3 As essential drivers of morbidity, mortality, utilization, and well being care prices, social threat components are essential for well being threat evaluation and each particular person and inhabitants well being administration.4,5 Particularly, social threat issue data could enhance threat prediction fashions6,7 and determine sufferers in want of social providers.8 Due to the potential worth of social threat issue data, federal companies, scientific organizations, and well being system consultants advocate for higher assortment and use of affected person social threat issue data.9
Regardless of the potential worth of social threat issue data to particular person affected person care and inhabitants well being administration actions, well being care organizations’ present strategies to measure this data are fraught with challenges. Affected person-facing social threat questionnaires haven’t been constantly validated,10 diagnostic codes reminiscent of Worldwide Classification of Ailments, Tenth Revision (ICD-10) Z codes are underutilized,11-13 area-level measures (eg, zip code–stage demographics) can masks heterogeneity throughout people and are susceptible to the ecological fallacy,14,15 and extracting free-text scientific documentation from digital well being information (EHRs) stays troublesome for a lot of organizations.16,17 Consequently, any considered one of these strategies will not be adequate for well being care organizations to gather the data essential to make inferences about sufferers’ and populations’ social threat components.
Computable social threat issue phenotypes derived from routinely collected structured EHR or well being data change (HIE) knowledge could also be an alternate strategy to measuring social components.18 Computable phenotypes are representations of affected person circumstances or traits that may be obtained from EHR knowledge by combining an outlined set of variables and logical expressions.19-21 Knowledge reminiscent of demographics, insurance coverage data, billing histories, appointment standing, emergency contacts, and language preferences exist in most EHRs. Though such knowledge could not contribute to understanding sufferers’ scientific standing, they’re reflective of traits of social and financial well-being and thus could also be amenable to make use of in social threat issue phenotypes. Moreover, utilizing structured knowledge parts already routinely collected as a part of scientific and enterprise operation workflows could mitigate the challenges of underutilization of screening surveys and analysis coding, extra knowledge assortment burden, and the technical implementation hurdles of pure language processing (NLP) for textual knowledge. Furthermore, when utilized to HIE knowledge, which mix affected person knowledge throughout organizations over time, strong computable social threat issue phenotypes could also be constructed that cut back lacking knowledge challenges22,23 and improve explanatory energy.19 Nonetheless, biomedical informatics and well being providers analysis have devoted little consideration to the potential worth of growing phenotypes from current structured knowledge for social threat issue measurement in favor of questionnaires, area-level knowledge linkage, and NLP.18,24
From current work on computable phenotypes, we all know that poor knowledge high quality25 and knowledge which can be inconsistently collected throughout affected person populations could end in inaccurate and in any other case biased phenotypes.26 Subsequently, as a foundational step, we convened an professional panel to determine and assess the standard of particular person EHR and HIE knowledge parts that may be helpful, correct, and unbiased parts to incorporate in future computable social threat issue phenotype growth. Our work units the stage for the longer term quantitative growth of computable social threat phenotypes by offering professional perception to information choice of candidate knowledge parts.
MATERIALS AND METHODS
We used a 2-round Delphi method to determine and preliminarily consider structured knowledge parts as candidates to be used in future computable phenotype growth.27,28
Knowledgeable Panel Formation
We recruited 17 people (of 18 invites) with in-depth data of EHR and/or HIE knowledge primarily based on publications or observe expertise in 1 or extra of the next 3 areas: EHR or HIE expertise administration in a well being care group; scientific or operational observe that concerned knowledge assortment; or EHR or HIE analysis. The vast majority of respondents had been affiliated with analysis or educational medical establishments (n = 14) and the rest had been people in management roles at well being data expertise organizations (n = 3). The panel represented organizations situated on the East Coast and West Coast and within the South and Midwest. 5 of the panel members had been physicians. Knowledgeable panel members obtained a monetary incentive of $250 for collaborating within the focus group and follow-up survey. Panelists had been cut up throughout 2 identification periods that every adopted a typical protocol.
Prematurely of the identification periods, we supplied every panelist with a abstract of the analysis targets of figuring out and assessing potential structured knowledge parts, an outline of the expectations, and a shared definition of social threat components (ie, any patient-level nonclinical financial, contextual, and psychosocial traits and components). As a result of computable phenotypes are helpful if generalizable,21 we instructed panelists to deal with structured knowledge parts that they might count on to be generally accessible in EHR or HIE knowledge. We requested panelists to exclude unstructured knowledge parts (eg, scientific be aware or different textual content knowledge), knowledge parts requiring linkage to sources exterior of typical EHR or HIE programs (eg, tax information), and patient-facing social threat issue questionnaires as a result of these will not be extensively adopted. These restrictions had been supposed to prioritize potential structured knowledge parts that may be extensively accessible. Panelists’ concepts weren’t restricted to particular age teams. We supplied this data to every panel member throughout a brief preparatory telephone name prematurely of the identification periods.
Spherical 1: Identification
The research crew (consisting of authors J.R.V., H.E.Ok., C.M., and C.A.H.) carried out two 90-minute group identification periods (n = 8 and n = 9) by way of videoconference. In every of those periods, we adopted a nominal group strategy by which every panelist, in flip, was requested to recommend a knowledge aspect, till all concepts had been exhausted. A analysis crew member documented the concepts generated in actual time and displayed them on display screen throughout the session. To assist manage the concept technology, when suggesting knowledge parts, panelists had been requested to categorize them into 1 or extra broad classes of social threat components1,29,30: socioeconomic standing (eg, employment, monetary, meals insecurity/starvation, housing instability); cultural context (eg, language, well being literacy); social relationships (eg, social help, incarceration); and neighborhood context (eg, housing high quality, transportation, security/violence).
The two identification periods had been handled as impartial (ie, findings from the primary weren’t shared with the second). The analysis crew deduplicated instructed knowledge parts from throughout the two panels.
Spherical 2: Evaluation
Panelists had been requested to price knowledge parts recognized in spherical 1 primarily based on 2 traits (outlined under): high quality and probability of systematic variations throughout demographic teams. The aim of this score train was to start evaluating the feasibility and appropriateness of real-world EHR or HIE knowledge for future computable phenotypes. The score survey was carried out utilizing REDCap.31,32 Earlier than administering, 4 core analysis crew members and a couple of panelists pilot examined the survey.
Is the information aspect top quality? We outlined high-quality knowledge parts as people who had been concurrently full, correct, and updated in an EHR or HIE system. These 3 dimensions are frequent parts of knowledge high quality frameworks.33,34 Panel members rated every aspect on a 5-point scale from poor high quality to wonderful high quality (eAppendix A [eAppendices available at ajmc.com]).
What’s the probability of systematic variations in knowledge high quality? Differential knowledge high quality throughout affected person demographic teams (eg, race/ethnicity, gender, age, sexual orientation) can result in inaccurate and biased social threat issue measurement, threat prediction, and inhabitants well being administration actions. Systematic variations in high quality may very well be attributable to an absence of affected person range, differential work processes, structural limitations to care, or broader societal circumstances.35 Panel members rated every aspect on a 5-point scale from extraordinarily unlikely totally different knowledge high quality to extraordinarily possible totally different knowledge high quality.
Lastly, to raised perceive every panel member’s body of reference when finishing the survey, we included a single merchandise to gauge if responses had been rooted in experiences with knowledge from hospital settings, doctor/group observe settings, and/or HIE programs.
Analyses had been divided by the identification and evaluation phases of knowledge assortment. First, we decided counts of recognized potential knowledge parts by social threat issue class. We additionally decided knowledge parts instructed in every social threat class throughout the identification periods. These periods additionally generated group dialogue on potential dangers and limitations of every knowledge aspect, which we summarized. Subsequent, we computed frequencies and percentages to explain panelists’ rankings of every knowledge aspect throughout the evaluation portion of the panel. We created 2-way scatterplots for instance the plurality of panelists who responded on the 2 extremes of the respective scales (eg, high 2 field strategy).36 The plots assist determine these components that they typically perceived to be of upper knowledge high quality and likewise as much less prone to have systematic variations throughout teams. We plotted the information parts for every social issue class individually. To facilitate visualization, we labeled reported knowledge parts as billing and cost, diagnoses and scientific knowledge, encounters and appointments, identifiers and speak to data, language, referrals and orders, social determinants of well being codes, and different. The complete distribution of responses for each knowledge aspect is introduced in eAppendix B. To look at consistency in rankings throughout panel members, we additionally grouped common knowledge aspect rankings by social issue class and stratified by panel member sort (doctor or nonphysician) and first body of reference when answering questions (EHR or HIE) (eAppendix C).
Identification: Potential Knowledge Parts (identification session)
Throughout the two identification periods, panelists generated a complete of 89 structured knowledge parts (Desk). Nonetheless, because of the cooccurring nature of social threat components, a number of instructed knowledge parts pertained to a number of classes. About half of the recognized parts (n = 45 of 89) had been related to the socioeconomic standing class. Inside the socioeconomic standing class, most knowledge parts had been instructed as related to monetary standing adopted by employment, meals insecurity/starvation, and housing instability. The socioeconomic class additionally included a number of knowledge parts that we thought-about to be common indicators of socioeconomic-related wants. Candidate knowledge parts mirrored billing, identifiers, orders, and utilization knowledge.
Identification: Observations on Issues, Concerns, and Limitations Concerning Knowledge Parts (identification session)
Throughout identification, panelists acknowledged a number of limitations associated to structured knowledge parts. These issues included the potential for inherently biased knowledge parts, inconsistent knowledge assortment processes, potential variation throughout affected person populations, and the constraints of area-level measures. For instance, a panelist famous that discrimination happens within the care supply processes and that underserved populations face limitations in accessing providers, which might result in biased knowledge. Equally, one other panelist famous that credit score scores could also be very predictive of sufferers’ monetary dangers and desires, however this knowledge aspect is understood to be biased by race. Equally, one other famous that inconsistent knowledge assortment additionally restricted the usefulness of some knowledge parts. As one panelist acknowledged about documenting homelessness, “There are some ICD codes that no person makes use of.” One other panelist agreed with the restricted adoption, however he famous that “…the ICD code goes to be very particular when used.” Panelists famous that computable phenotypes could have to be developed for various affected person populations. For instance, some knowledge parts may very well be related for adults however wouldn’t be related for pediatric populations. Alternatively, a phenotype might show helpful for sufferers with excessive well being care utilization solely. Lastly, one panelist famous the “poor overlap” between area-level measures and sufferers’ self-reported social threat components.
Evaluation: Perceptions of Knowledge High quality and Chance of Systematic Variations in High quality
When responding to the evaluation surveys, most (n = 12 of 17) of the panel members reported primarily desirous about knowledge that come from HIE programs. Clinicians and nonclinicians didn’t range considerably of their assessments of knowledge high quality and probability of systematic variations in knowledge high quality throughout populations. Panelists’ evaluation of the probability of systematic variations in knowledge high quality throughout populations did range primarily based on whether or not they reported primarily pondering of HIE programs vs EHR programs. Those that reported pondering primarily about HIE programs most regularly reported that high quality was prone to be totally different throughout populations for the socioeconomic, social relationship, and neighborhood context classes (eAppendix C).
Socioeconomic Standing Knowledge Component Evaluation
Knowledge parts which can be each increased high quality and unlikely to have systematic variations throughout populations are preferable. For these knowledge parts instructed by the panelists as related to socioeconomic standing, solely identifier and speak to data and billing and payment-related knowledge parts had been regularly rated as “superb” or “wonderful” high quality and on the identical time rated as additionally “unlikely” or “extraordinarily unlikely” to have differential high quality throughout affected person teams (Determine 1). These parts included date of beginning, final title, deal with, insurance coverage sort, payments in assortment, cost methodology, days in accounts receivable, and excellent payments. Many extra knowledge parts had been usually seen as low high quality (ie, “honest” or “poor”). Notably, panelists rated the ICD-10 Z and Logical Remark Identifiers Names and Codes (LOINC) codes that signify varied social threat components, a number of knowledge parts associated to referrals to particular social providers and suppliers, and incapacity to do telehealth visits as decrease high quality (ie, “honest” or “poor”) and concurrently “possible” or “extraordinarily possible” to have differential high quality throughout affected person teams.
Cultural Context Knowledge Component Evaluation
Panelists rated solely deal with and EHR portal account presence and utilization as top quality (ie, “superb” or “wonderful”) (Determine 2). Panelists additionally thought-about these parts to usually be “unlikely” or “extraordinarily unlikely” to have differential high quality throughout affected person teams. Conversely, greater than half of panelists rated presence of advance directives, language of discharge directions, main language, and the necessity and use of interpreters as low high quality (ie, “honest” or “poor”) and as “possible” or “extraordinarily possible” to have differential high quality throughout affected person teams. Once more, ICD-10 Z codes and related LOINC codes associated to schooling and literacy had been rated of low high quality and prone to have totally different knowledge high quality throughout populations.
Social Relationships Knowledge Component Evaluation
Panelists rated few social relationship knowledge parts as top quality total (ie, “superb” or “wonderful”) (Determine 3). The vast majority of panelists rated social relationship–related ICD-10 Z and LOINC codes as low high quality and “possible” or “extraordinarily possible” to have differential high quality throughout affected person teams.
Group Context Knowledge Component Evaluation
Locally context area (Determine 4), some knowledge parts related to identification and speak to data, diagnoses and scientific knowledge, and encounters and appointments, reminiscent of deal with, arrival by ambulance, and emergency division visits related to trauma or damage, had been seen by extra panelists as increased high quality and “unlikely” or “extraordinarily possible” to have differential knowledge high quality throughout populations. Different diagnoses and scientific knowledge had been additionally thought-about “unlikely” or “extraordinarily possible” to have differential knowledge high quality throughout populations however had been nonetheless seen as having “honest” or “poor” knowledge high quality. Once more, panelists rated neighborhood context–related ICD-10 Z and LOINC codes as “possible” or “extraordinarily possible” to have differential knowledge high quality throughout populations and to be of poorer knowledge high quality.
Our panel of 17 EHR and HIE knowledge consultants recognized and commented on routinely collected structured knowledge parts for potential use sooner or later growth of computable social threat issue phenotypes. Panelists highlighted a number of particular issues about total knowledge high quality and the potential for systematic high quality variations throughout populations which will result in bias and different knowledge inaccuracies. This novel and foundational work can be utilized to assist develop future computable phenotypes for social components.
Knowledge high quality (outlined on this research as full, correct, and updated) is a long-standing concern in biomedical informatics, notably when knowledge are used for functions apart from these for which they had been initially collected.37 Panelists usually perceived knowledge parts of the best high quality to be these utilized in reimbursement-related processes (eg, identifiers and speak to data, billing and payment-related knowledge, diagnoses), which is in line with prior research.38,39 As well as, panelists reported that these knowledge parts had been amongst these much less prone to be systematically totally different in high quality throughout populations. Given these advantageous qualities, reimbursement-related knowledge parts could also be viable candidates to be used in computable phenotypes growth.
Basically, essentially the most constantly recognized high quality issues associated to structured knowledge parts which have been designed to doc social threat components: ICD-10 Z and LOINC codes. Proof signifies that these codes are considerably underutilized in observe.11-13 Not solely did the professional panel outcomes query the standard of those knowledge, however additionally they indicated that these had been among the many almost definitely to have totally different knowledge high quality throughout populations. These perceptions align with a current quantitative evaluation indicating that ICD-10 Z codes are a particular indicator of social threat, however one that’s collected in a biased style.40 Rising adoption, specific reimbursement for documenting social wants, or the mapping of screening questionnaires to those requirements41 could ultimately improve the utility of those knowledge parts. Nonetheless, at the moment their utility to computable phenotyping or different measurement actions seems restricted.
Poor-quality knowledge can undermine any utility to care supply. Nonetheless, when knowledge high quality is systematically totally different throughout populations, the dangers improve that any measurement technique, together with computed phenotypes, might perpetuate societal biases and inequitable practices in well being care.42,43 Associated purposes of well being knowledge have demonstrated the danger of drawing the improper inferences about sufferers. For instance, a extensively utilized threat stratification device systematically really useful more healthy White sufferers over sicker Black sufferers for care administration packages, as a result of it didn’t account for differential ranges of entry.44 Equally, illness threat fashions developed in homogenous majority populations don’t carry out effectively for minority teams.45 The longer term growth of computable social threat issue phenotypes would require consideration to the dangers of biased and differential high quality knowledge as a result of the processes for gathering well being care knowledge are extremely variable.46,47 A number of frameworks and methodologies for figuring out and mitigating bias exist, which may very well be utilized to those knowledge.48,49 Moreover, like every superior analytics interventions, computable social issue phenotypes would have to be regularly evaluated and monitored for effectiveness and lack of bias.50
Successfully figuring out sufferers’ social dangers is critical for well being care organizations to provoke acceptable referrals to providers.51 Distributors, collaboratives, and well being care organizations have efficiently built-in screening questions into EHR programs and workflows to help knowledge assortment.41,52,53 Nonetheless, utilization of screening questionnaires in observe is very variable54 and, when used, they’ve elevated workers’s knowledge assortment burden.55 As a measurement technique reliant on current knowledge and one that may be doubtlessly automated, computable social issue phenotypes might help the screening use case whereas avoiding the challenges of administering questionnaires. Nonetheless, screening for social threat components in well being care generally is a delicate concern for sufferers56,57 and automatic computable phenotypes are admittedly not as clear a screening technique as patient-completed questionnaires. If computable social issue phenotypes may very well be efficiently developed, future work ought to embrace assessments of affected person acceptability.
Creating computable phenotypes was past the scope of this Delphi panel. Nonetheless, the findings on this paper present a candidate listing of knowledge parts that may very well be additional evaluated for setting up computable phenotypes. As a part of the identification course of, panelists defined or justified their instructed knowledge parts. Usually, these explanations took the type of strategies for reworking or combining knowledge parts into novel and doubtlessly extra informative measures. For instance, panelists emphasised the potential to achieve data by taking a look at modifications in knowledge parts over time. Essentially the most salient examples had been modifications in addresses to determine housing instability, in insurance coverage standing for monetary standing, or in emergency contact data for social relationships, that are recurrently up to date for billing and reimbursement causes. Others have instructed related makes use of of change in deal with knowledge over time.58 Likewise, panel members famous the data to be gained by explainable missingness (ie, cases by which social circumstances would end in knowledge not being recorded or deliberately recorded as lacking, as within the case of lacking zip codes, telephone numbers, or addresses for homeless people). Different essential suggestions associated to combining knowledge throughout sufferers. For instance, novel knowledge factors may very well be created by figuring out the variety of people sharing an deal with to point monetary pressure or by noting reciprocal emergency contacts as indicators of social help. Panelists additionally surfaced the chance that social want may very well be recognized utilizing proof of discordant utilization. An instance of this kind included ordering dietary dietary supplements or particular meals with out an accompanying analysis that instructed a scientific want. Lastly, concepts from panel members additionally included utilizing the data from the parts of the affected person document. This included the presence, absence, size, or accessing of social employee notes and even the frequency with which the affected person portal was used. Subsequently, along with uncooked, untransformed knowledge parts, future work to develop computable phenotypes ought to think about these knowledge transformations or combos—together with modifications over time, explainable missingness, combining knowledge throughout sufferers, discordant utilization, and the parts of the affected person document.
Given the variation within the utilization of social threat components, dangers, and desires in observe and the literature, it’s doable that panel members had totally different conceptualizations of the social threat components mentioned. For identification, such variation possible had little impact, however the precise motion to phenotype development would require clear assemble definitions. Equally, our survey didn’t mirror the multidimensional nature of knowledge high quality (eg, conformance, completeness, plausibility) however relied on a single query to scale back respondent burden. Further work could be mandatory to know the totally different knowledge high quality rankings of every knowledge aspect. For instance, we can’t inform from this research if parts had been rated as low high quality attributable to perceived inaccuracy or that the information parts couldn’t be relied upon as a result of they had been used too sometimes. Further research could be required to match panelists’ perceptions with precise knowledge high quality metrics. Additionally, the panelists had been instructed to exclude nonstructured and different sources of knowledge. In fact, these are essential sources of social threat issue data, however their potential utilization in computable phenotypes represents a unique set of challenges than those explored on this panel. Nonetheless, the recognized structured knowledge parts may very well be mixed with structured knowledge from survey questionnaires and even unstructured knowledge extracted from NLP, the place accessible. Such combos may very well be extra informative. Though professional panel members acknowledged that social threat components change over time, the difficulty of acceptable intervals for measuring social threat components was not included on this Delphi panel. Lastly, our professional panel mirrored people with data about EHR and HIE knowledge sources and processes that generated knowledge for scientific, analysis, and enterprise functions. A special set of panel members, with totally different backgrounds, could have recognized different knowledge parts.
EHRs and HIE programs comprise structured knowledge parts that mirror affected person social circumstances, and these knowledge could also be helpful in growing computable phenotypes. Efforts to develop phenotypes ought to think about knowledge high quality and dangers for systematic variations throughout populations. Future computable phenotyping analysis ought to validate methods for incorporating ideas reminiscent of modifications over time, explainable missingness, combining knowledge throughout sufferers, discordant utilization, and the parts of the affected person document.
The authors thank Lindsey Sanner, MPH, for her help with visualizations.
Creator Affiliations: Indiana College Richard M. Fairbanks Faculty of Public Well being (JRV, HEK, WMT), Indianapolis, IN; Regenstrief Institute (JRV), Indianapolis, IN; Division of Drugs (JA-M) and Division of Household and Group Drugs (LMG), College of California, San Francisco, San Francisco, CA; Division of Well being Outcomes and Biomedical Informatics, School of Drugs, College of Florida (JB, CM, EAS, CAH), Gainesville, FL; Inhabitants Well being Sciences, Weill Cornell Medical School (TRC), New York, NY; Mathematica (GRC), Washington, DC; New York eHealth Collaborative (ND), New York, NY; Owl Well being Works LLC (JH), Indianapolis, IN; Departments of Household and Group Drugs and Biomedical Informatics, School of Drugs, The Ohio State College (TRH), Columbus, OH; Indiana Well being Info Change (JPK), Indianapolis, IN; Johns Hopkins Faculty of Public Well being (HK), Baltimore, MD; Division of Inhabitants Well being, Dell Medical Faculty, The College of Texas at Austin (AK), Austin, TX; Anthem, Inc (JMO), Indianapolis, IN; Division of Pediatrics, Middle for Well being and Group, College of California, San Francisco (MSP), San Francisco, CA; College of Rochester Medical Middle (WP), Rochester, NY; Division of Pediatrics, Faculty of Drugs, Indiana College (SW), Indianapolis, IN.
Supply of Funding: This work was supported, partly, by the Indiana Medical and Translational Sciences Institute Fund and partly by award No. UL1TR002529 from the Nationwide Institutes of Well being, Nationwide Middle for Advancing Translational Sciences, Medical and Translational Sciences Award. The content material is solely the duty of the authors and doesn’t essentially signify the official views of the Nationwide Institutes of Well being.
Creator Disclosures: Dr Vest supplied consulting to New York eHealth Collaborative and Pima County; is a founder and fairness holder in Uppstroms, LLC, a expertise firm; and has patents pending with Uppstroms. Dr Wiehe reviews receiving an incentive for collaborating within the panel. Dr Harle’s establishment has obtained analysis grants associated to social threat components and their measurement. The remaining authors report no relationship or monetary curiosity with any entity that may pose a battle of curiosity with the subject material of this text.
Authorship Info: Idea and design (JRV, CAH); acquisition of knowledge (JRV, JA-M, LMG, JB, TRC, GC, ND, JH, TRH, JPK, HK, AJ, HEK, CM, CM, JMO, MSP, WP, EAS, WMT, SW, CAH); evaluation and interpretation of knowledge (JRV, JA-M, LMG, JB, TRC, GC, ND, JH, TRH, JPK, HK, AJ, HEK, CM, CM, JMO, MSP, WP, EAS, WMT, SW, CAH); drafting of the manuscript (JRV, JA-M, LMG, CM, HEK, CAH); important revision of the manuscript for essential mental content material (JRV, JA-M, LMG, JB, TRC, GC, ND, JH, TRH, JPK, HK, AJ, HEK, CM, CM, JMO, MSP, WP, EAS, WMT, SW, CAH); statistical evaluation (JRV, HEK, CAH); provision of sufferers or research supplies (JRV, CM, HEK, CAH); acquiring funding (JRV); administrative, technical, or logistic help (JRV, HEK, CM, CAH); supervision (JRV, CAH).
Deal with Correspondence to: Joshua R. Vest, PhD, MPH, Indiana College Richard M. Fairbanks Faculty of Public Well being, 1050 Wishard Blvd, Indianapolis, IN 46202. E mail: [email protected].
1. Inexperienced Ok, Zook M. When speaking about social determinants, precision issues. Well being Affairs. 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 well being lexicon for well being care programs. Milbank Q. 2019;97(2):407-419. doi:10.1111/1468-0009.12390
3. Woolf S, Aron L, eds. U.S. Well being in Worldwide Perspective: Shorter Lives, Poorer Well being. The Nationwide Academies Press; 2013.
4. Fee on Social Determinants of Well being. Closing the Hole in a Era: Well being Fairness By Motion on the Social Determinants of Well being. World Well being Group; 2008.
5. Pruitt Z, Emechebe N, Quast T, Taylor P, Bryant Ok. Expenditure reductions related to a social service referral program. Popul Well being Manag. 2018;21(6):469-476. doi:10.1089/pop.2017.0199
6. Bardsley M, Billings J, Dixon J, Georghiou T, Lewis GH, Steventon A. Predicting who will use intensive social care: case discovering instruments primarily based on linked well being and social care knowledge. Age Ageing. 2011;40(2):265-270. doi:10.1093/ageing/afq181
7. Tan M, Hatef E, Taghipour D, et al. Together with social and behavioral determinants in predictive fashions: tendencies, challenges, and alternatives. JMIR Med Inform. 2020;8(9):e18084. doi:10.2196/18084
8. Kasthurirathne SN, Vest J, Menachemi N, Halverson PK, Grannis SJ. Assessing the capability of social determinants of well being knowledge to enhance predictive fashions figuring out sufferers in want of wraparound social providers. J Am Med Inform Assoc. 2018;25(1):47-53. doi:10.1093/jamia/ocx130
9. Institute of Drugs. Capturing Social and Behavioral Domains in Digital Well being Data: Section 2. The Nationwide Academies Press; 2014.
10. Henrikson NB, Blasi PR, Dorsey CN, et al. Psychometric and pragmatic properties of social threat screening instruments: a scientific evaluation. Am J Prev Med. 2019;57(6 suppl 1):S13-S24. doi:10.1016/j.amepre.2019.07.012
11. Matthew J, Hodge C, Khau M. Z codes utilization amongst Medicare fee-for-service (FFS) beneficiaries in 2017. CMS. January 2020. Accessed January 19, 2021. https://www.cms.gov/information/doc/cms-omh-january2020-zcode-data-highlightpdf.pdf
12. Truong HP, Luke AA, Hammond G, Wadhera RK, Reidhead M, Joynt Maddox KE. Utilization of social determinants of well being ICD-10 Z-codes amongst hospitalized sufferers in the US, 2016-2017. Med Care. 2020;58(12):1037-1043. doi:10.1097/MLR.0000000000001418
13. Guo Y, Chen Z, Xu Ok, et al. Worldwide Classification of Ailments, Tenth Revision, Medical Modification social determinants of well being codes are poorly utilized in digital well being information. Drugs (Baltimore). 2020;99(52):e23818. doi:10.1097/MD.0000000000023818
14. Gottlieb LM, Francis DE, Beck AF. Makes use of and misuses of patient- and neighborhood-level social determinants of well being knowledge. Perm J. 2018;22:18-078. doi:10.7812/tpp/18-078
15. Buajitti E, Chiodo S, Rosella LC. Settlement between area- and individual-level earnings measures in a population-based cohort: implications for inhabitants well being analysis. SSM Popul Well being. 2020;10:100553. doi:10.1016/j.ssmph.2020.100553
16. Chapman WW, Nadkarni PM, Hirschman L, D’Avolio LW, Savova GK, Uzuner O. Overcoming limitations to NLP for scientific textual content: the function of shared duties and the necessity for extra inventive options. J Am Med Inform Assoc. 2011;18(5):540-543. doi:10.1136/amiajnl-2011-000465
17. Lasser EC, Kim JM, Hatef E, Kharrazi H, Marsteller JA, DeCamp LR. Social and behavioral variables within the digital well being document: a path ahead to extend knowledge high quality and utility. Acad Med. 2021;96(7):1050-1056. doi:10.1097/ACM.0000000000004071
18. Parikh RB, Jain SH, Navathe AS. The sociobehavioral phenotype: making use of a precision drugs framework to social determinants of well being. Am J Manag Care. 2019;25(9):421-423.
19. Frey LJ, Lenert L, Lopez-Campos G. EHR large knowledge deep phenotyping: contribution of the IMIA Genomic Drugs Working Group. Yearb Med Inform. 2014;9(1):206-211. doi:10.15265/iy-2014-0006
20. Verchinina L, Ferguson L, Flynn A, Wichorek M, Markel D. Computable phenotypes: standardized methods to categorise folks utilizing digital well being document knowledge. Perspect Well being Inf Manag. 2018;(Fall):1-8.
21. Richesson RL, Hammond WE, Nahm M, et al. Digital well being information primarily based phenotyping in next-generation scientific trials: a perspective from the NIH Well being Care Techniques Collaboratory. J Am Med Inform Assoc. 2013;20(e2):e226-e231. doi:10.1136/amiajnl-2013-001926
22. Hripcsak G, Albers DJ. Subsequent-generation phenotyping of digital well being information. J Am Med Inform Assoc. 2013;20(1):117-121. doi:10.1136/amiajnl-2012-001145
23. Basile AO, Ritchie MD. Informatics and machine studying to outline the phenotype. Knowledgeable Rev Mol Diagn. 2018;18(3):219-226. doi:10.1080/14737159.2018.1439380
24. Feller DJ, Bear Don’t Stroll OJ IV, Zucker J, Yin MT, Gordon P, Elhadad N. Detecting social and behavioral determinants of well being with structured and free-text scientific knowledge. Appl Clin Inform. 2020;11(1):172-181. doi:10.1055/s-0040-1702214
25. Ahmad FS, Ricket IM, Hammill BG, et al. Computable phenotype implementation for a nationwide, multicenter pragmatic scientific trial: classes realized from ADAPTABLE. Circ Cardiovasc Qual Outcomes. 2020;13(6):e006292. doi:10.1161/CIRCOUTCOMES.119.006292
26. Richesson R, Smerek M. Digital well being records-based phenotyping. rethinking scientific trials. June 27, 2014. Accessed April 26, 2021. https://websites.duke.edu/rethinkingclinicaltrials/ehr-phenotyping/
27. McPherson S, Reese C, Wendler MC. Methodology replace: Delphi research. Nurs Res. 2018;67(5):404-410. doi:10.1097/nnr.0000000000000297
28. Hasson F, Keeney S, McKenna H. Analysis tips for the Delphi survey method. J Adv Nurs. 2000;32(4):1008-1015. doi:10.1046/j.1365-2648.2000.t01-1-01567.x
29. Social determinants of well being. HealthyPeople.gov. Accessed March 5, 2020. https://www.healthypeople.gov/2020/topics-objectives/subject/social-determinants-of-health
30. Nationwide Academies of Sciences, Engineering, and Drugs. Accounting for Social Threat Elements in Medicare Cost: Figuring out Social Threat Elements. The Nationwide Academies Press; 2016.
31. Harris PA, Taylor R, Thielke R, Payne J, Gonzalez N, Conde JG. Analysis digital knowledge seize (REDCap)—a metadata-driven methodology and workflow course of for offering translational analysis informatics help. J Biomed Inform. 2009;42(2):377-381. doi:10.1016/j.jbi.2008.08.010
32. Harris PA, Taylor R, Minor BL, et al; REDCap Consortium. The REDCap consortium: constructing a world neighborhood of software program platform companions. J Biomed Inform. 2019;95:103208. doi:10.1016/j.jbi.2019.103208
33. Kahn MG, Callahan TJ, Barnard J, et al. A harmonized knowledge high quality evaluation terminology and framework for the secondary use of digital well being document knowledge. EGEMS (Wash DC). 2016;4(1):1244-1244. doi:10.13063/2327-9214.1244
34. Lee YW, Sturdy DM, Kahn BK, Wang RY. AIMQ: a technique for data high quality evaluation. Inf Handle. 2002;40(2):133-146. doi:10.1016/s0378-7206(02)00043-5
35. Ferryman Ok, Pitcan M. Equity in precision drugs. Knowledge & Society. February 26, 2018. Accessed March 12, 2020. https://datasociety.internet/library/fairness-in-precision-medicine/
36. Russell GJ. Itemized score scales (Likert, semantic differential, and Stapel). In: Kamakura W, ed. Advertising and marketing Analysis. Wiley & Sons; 2010. Sheth J, Malhotra NK, eds. Wiley Worldwide Encyclopedia of Advertising and marketing; vol 2. Accessed March 17, 2021. https://onlinelibrary.wiley.com/doi/abs/10.1002/9781444316568.wiem02011
37. Weiskopf NG, Weng C. Strategies and dimensions of digital well being document knowledge high quality evaluation: enabling reuse for scientific analysis. J Am Med Inform Assoc. 2013;20(1):144-151. doi:10.1136/amiajnl-2011-000681
38. Callahan A, Shah NH, Chen JH. Analysis and reporting concerns for observational research utilizing digital well being document knowledge. Ann Intern Med. 2020;172(suppl 11):S79-S84. doi:10.7326/M19-0873
39. Horsky J, Drucker EA, Ramelson HZ. Accuracy and completeness of scientific coding utilizing ICD-10 for ambulatory visits. AMIA Annu Symp Proc. 2018;2017:912-920.
40. Weeks WB, Cao SY, Lester CM, Weinstein JN, Morden NE. Use of Z-codes to document social determinants of well being amongst fee-for-service Medicare beneficiaries in 2017. J Gen Intern Med. 2020;35(3):952-955. doi:10.1007/s11606-019-05199-w
41. Weir RC, Proser M, Jester M, Li V, Hood-Ronick CM, Gurewich D. Amassing social determinants of well being knowledge within the scientific setting: findings from nationwide PRAPARE implementation. J Well being Care Poor Underserved. 2020;31(2):1018-1035. doi:10.1353/hpu.2020.0075
42. Rutjes AWS, Reitsma JB, Di Nisio M, Smidt N, van Rijn JC, Bossuyt PMM. Proof of bias and variation in diagnostic accuracy research. CMAJ. 2006;174(4):469-476. doi:10.1503/cmaj.050090
43. Rajkomar A, Hardt M, Howell MD, Corrado G, Chin MH. Guaranteeing equity in machine studying to advance well being fairness. Ann Intern Med. 2018;169(12):866-872. doi:10.7326/M18-1990
44. Obermeyer Z, Powers B, Vogeli C, Mullainathan S. Dissecting racial bias in an algorithm used to handle the well being of populations. Science. 2019;366(6464):447-453. doi:10.1126/science.aax2342
45. Adamson AS, Smith A. Machine studying and well being care disparities in dermatology. JAMA Dermatol. 2018;154(11):1247-1248. doi:10.1001/jamadermatol.2018.2348
46. Cohen GR, Friedman CP, Ryan AM, Richardson CR, Adler-Milstein J. Variation in physicians’ digital well being document documentation and potential affected person hurt from that variation. J Gen Intern Med. 2019;34(11):2355-2367. doi:10.1007/s11606-019-05025-3
47. Overhage JM, McCallie D. Doctor time spent utilizing the digital well being document throughout outpatient encounters. Ann Intern Med. 2020;173(7):594-595. doi:10.7326/M18-3684
48. Bellamy RKE, Dey Ok, Hind M, et al. AI Equity 360: an extensible toolkit for detecting, understanding, and mitigating undesirable algorithmic bias. ArXiv. Preprint posted on-line October 3, 2018. Accessed Could 29, 2020. http://arxiv.org/abs/1810.01943
49. Haneuse S, Daniels M. A common framework for contemplating choice bias in EHR-based research: what knowledge are noticed and why? EGEMS (Wash DC). 2016;4(1):1203. doi:10.13063/2327-9214.1203
50. Embi PJ. Algorithmovigilance—advancing strategies to investigate and monitor synthetic intelligence–pushed well being look after effectiveness and fairness. JAMA Netw Open. 2021;4(4):e214622. doi:10.1001/jamanetworkopen.2021.4622
51. Garg A, Boynton-Jarrett R, Dworkin PH. Avoiding the unintended penalties of screening for social determinants of well being. JAMA. 2016;316(8):813-814. doi:10.1001/jama.2016.9282
52. Buitron de la Vega P, Losi S, Sprague Martinez L, et al. Implementing an EHR-based screening and referral system to deal with social determinants of well being in main care. Med Care. 2019;57(6 suppl 2):S133-S139. doi:10.1097/mlr.0000000000001029
53. Gold R, Bunce A, Cowburn S, et al. Adoption of social determinants of well being EHR instruments by neighborhood well being facilities. Ann Fam Med. 2018;16(5):399-407. doi:10.1370/afm.2275
54. Cottrell EK, Dambrun Ok, Cowburn S, et al. Variation in digital well being document documentation of social determinants of well being throughout a nationwide community of neighborhood well being facilities. Am J Prev Med. 2019;57(6 suppl 1):S65-S73. doi:10.1016/j.amepre.2019.07.014
55. Greenwood-Ericksen M, DeJonckheere M, Syed F, Choudhury N, Cohen AJ, Tipirneni R. Implementation of health-related social wants screening at Michigan well being facilities: a qualitative research. Ann Fam Med. 2021;19(4):310-317. doi:10.1370/afm.2690
56. Kusnoor SV, Koonce TY, Hurley ST, et al. Assortment of social determinants of well being in the neighborhood clinic setting: a cross-sectional research. BMC Public Well being. 2018;18(1):550. doi:10.1186/s12889-018-5453-2
57. Pinto AD, Glattstein-Younger G, Mohamed A, Bloch G, Leung FH, Glazier RH. Constructing a basis to scale back well being inequities: routine assortment of sociodemographic knowledge in main care. J Am Board Fam Med. 2016;29(3):348-355. doi:10.3122/jabfm.2016.03.150280
58. Gold R, Cottrell E, Bunce A, et al. Creating digital well being document (EHR) methods associated to well being heart sufferers’ social determinants of well being. J Am Board Fam Med. 2017;30(4):428-447. doi:10.3122/jabfm.2017.04.170046