Capturing Social and Behavioral Domains and Measures in Electronic Health Records: Phase 2 (2014)

Summary

BACKGROUND

Substantial empirical evidence of the contribution of social and behavioral factors to functional status and the onset and progression of disease has accumulated over the past few decades. Traditionally, research and interventions on social and behavioral determinants of health have largely been the purview of public health, which has focused on disease prevention, protection, and promotion of the public’s health. Health care systems, in contrast, have focused primarily on the treatment of disease in individual patients, and, until recently, social determinants of health have not been linked to clinical practice or been of concern to health care delivery systems. A variety of pressures are increasing the need for health care systems and providers to attend to the social and behavioral determinants of health. These include the relatively poor health status of the U.S. population despite high—and possibly unsustainable—investments in health care and new incentive structures through the Patient Protection and Affordable Care Act such as accountable care organizations which reward health systems for achieving better health with less use of costly medical services.

Electronic health records (EHRs) provide valuable information about the determinants of health and the effectiveness of treatment. This information can enable more effective responses to the pressures noted above when used by health systems, including public health officials, researchers, and providers treating individual patients. Inclusion of information on social and behavioral characteristics will provide vital knowledge to inform and improve all three uses.

The Health Information Technology for Economic and Clinical Health Act of 2009 (HITECH) and the Patient Protection and Affordable Care Act place new importance on the widespread adoption and meaningful use of EHRs. “Meaningful use” in a health information technology context refers to the use of EHRs and related technology within a health care organization to achieve specified objectives. Achieving meaningful use also helps determine whether an organization can receive payments from the Medicare and/or Medicaid EHR incentive programs. The Centers for Medicare & Medicaid Services (CMS) is working with the Office of the National Coordinator for Health Information Technology (ONC) and other parts of the U.S. Department of Health and Human Services (HHS) to establish regulations for the third stage of the meaningful use incentive program. Meaningful Use Stage 3 is in development, and implementation for this stage is expected to start in 2017.

Meaningful Use regulations can incentivize the inclusion of social and behavioral data in EHRs. Expansion beyond the traditional medical information collected in EHRs to include social and behavioral health determinants requires the identification and application of criteria for determining what domains should be included in all EHRs and for specific populations. The rapid adoption of EHRs and the exigent Meaningful Use Stage 3 criteria formulation by the ONC and CMS add urgency to this effort.

THE FORMATION AND CHARGE TO THE COMMITTEE

The Office of Behavioral and Social Sciences Research and the National Institutes of Health, together with the Association of State and Territorial Health Officials, the Blue Shield of California Foundation, the California HealthCare Foundation, the Centers for Disease Control and Prevention, CMS, the Department of Veterans Affairs, The Lisa and John Pritzker Family Foundation, the Robert Wood Johnson Foundation, and the Substance Abuse and Mental Health Services Administration, requested that the Institute of Medicine (IOM) conduct a two-phase study to identify social and behavioral domains and their measures for inclusion in electronic health records. The charge to the committee for the project is presented in Box S-1.

In response to that request, the IOM convened a multidisciplinary committee of 13 members with a wide variety of expertise, including leaders from the fields of health information technology, clinical care and health systems, social and behavioral determinants of health, and measurement.

THE COMMITTEE’S APPROACH IN PHASE 1

To meet its charge, the committee first established the rationale for adding social and behavioral determinants of health into EHRs and con-

sidered how EHRs may assist providers in their decision making, resulting in improved health outcomes for their patients regardless of Meaningful Use adoption and implementation. The committee held four information-gathering meetings to hear from other experts in the field, stakeholders, and

the public. In addition, the committee met in closed sessions to allow for discussion and deliberation.

Before the first meeting and throughout the study process, the committee reviewed relevant literature. Its formal review of the literature focused on identifying peer-reviewed, published literature and reports; evidence-based reviews from governmental and other agencies; and previous IOM reports that were germane to the statement of task. For this study, the committee uses the term candidate to refer to the core domains (the third item of the Statement of Task) because the specific task during Phase 1 was to identify domains that should be considered by ONC for Stage 3 Meaningful Use. In this context the core domains are those that the committee proposed as candidates for being selected for Meaningful Use. The committee erred on the side of inclusion at this stage while also trying to limit the number of candidate domains. Throughout this study, the term domain is used to refer to determinants of health, which could include health conditions that, in turn, influence other health outcomes. The committee also embraced the use of the World Health Organization’s definition of health being “a state of complete physical, mental and social well-being and not merely the absence of disease or infirmity” (WHO, 1946).

Domains and Criteria

Several existing conceptual frameworks identify categories of health determinants and the ways in which they link to mortality as well as to disease onset and progression. These models generally distinguish individual-level characteristics (such as biological factors, emotional and cognitive traits, and health-related behaviors) from features of the context in which they emerge and operate (such as the physical and social environment). The committee reviewed a number of existing frameworks and selected three that it used in developing an outline of domains for the committee to review (see Chapter 2) as an initial step in identifying domains to be considered for inclusion in all EHRs.

The committee then decided on the following criteria for domains to be given a high priority for inclusion in EHRs:

1. Strength of the evidence of the association of the domain with health.
2. Usefulness of the domain, as measured for
   1. a. The individual patient for decision making between the clinician and patient for management and treatment;
   2. b. The population to describe and monitor population health and making health care–related policy decisions that affect the

2. 1.      population cared for by the particular health system or as a whole; and
   2. c. Research to conduct clinical and population health research to learn about the causes of health, the predictors of outcomes of care, and the impact of interventions at multiple levels.
3. Availability and standard representation of a reliable and valid measure(s) of the domain.
4. Feasibility, that is, whether a burden is placed on the patient and the clinician and the administrative time and cost of interfaces and storage.
5. Sensitivity, that is, if patient discomfort regarding revealing personal information is high and there are increased legal or privacy risks.
6. Accessibility of data from another source (i.e., information from external sources may be accessible to meet the needs of patient care, population health, and research; if so, the domains would have less priority for inclusion in the EHR).

The committee worked to narrow the number of domains in the outline using the first two criteria through a consensus process. The result reduced the number of domains constituting the candidate set to 17 for which the committee found sufficient evidence of relevance and usefulness to consider for inclusion in all EHRs. Given the limited time that the committee had to complete its Phase 1 tasks in order for its recommendations to be useful for Meaningful Use Stage 3 deliberations, the committee only used criteria 1 and 2 to select the 17 candidate domains. Chapter 3 of the report and the criteria presented above are intended to serve as resources to support their drafts and final decisions.

Committee’s Key Conclusions in Phase 1

In addition to developing criteria for selecting domains and measures to recommend for inclusion in all electronic health records, the major focus of Phase 1 was identifying a candidate set of domains relevant for all individuals. The committee’s conclusions are listed in Table S-1. Of note, the committee opted to include domains even if they are already routinely captured in EHRs to ensure that they will continue to be prioritized and to encourage the use of standard measures for these domains. The domains are not listed in order of priority. Rather, they are organized by the committee’s initial outline, which ordered domains in terms of types of levels they represented.

TABLE S-1 Summary of Selected and Non-Selected Domains

Domains Not Selected

Narrowing the initial set of domains covering the whole range of social and behavioral determinants was a difficult task. The committee’s decisions were guided by awareness of the need to identify the domains for which there was adequate evidence of the association of the domains with health outcomes and of the usefulness in having the information in EHRs. Most of the domains excluded from the final candidate set simply lacked an adequate evidence base to support routine capture of these data. (See Appendix A for more information on domains not selected.) Because the Phase 1 study serves as a foundation for the committee’s deliberations regarding their remaining task, the Phase 1 study report is woven into this report largely unchanged as Chapters 1–3.

PHASE 2: PRIORITIZING DOMAINS AND MEASURES

As the committee entered its Phase 2 work, two information-gathering sessions were held (see Appendix C for the agendas). The committee began to compile measurement tools for the candidate domains. These measures are essential ingredients in EHRs; they must be consistently defined and used in order to achieve interoperable standards, a priority goal for ONC. The committee also saw opportunities for increasing standardization despite logistical challenges to achieving routine, harmonized measurement tools. Toward meeting this objective, the committee applied criteria 3 through 6 to the 17 candidate domains, along with their subdomains, which had been selected as the best candidates for inclusion in EHRs, while continuing to take account of criteria 1 and 2. In several instances, multiple measures of a domain were carefully considered. In other instances, a single accepted standard measure, which had been tested for its reliability, validity, and scoring stood out. The committee weighed the usefulness and feasibility of collecting data using each measure, and potential concerns about the sensitive nature of the information or violations of privacy in collecting, storing, or using the data were also considered. Finally, the committee examined the accessibility of the data from other sources.

Initially, the committee considered each domain and measure individually. However, as work progressed, the committee stepped back and considered overlap between domains and their measures and interactions among them. To recommend core measures of social and behavioral domains for inclusion in all EHRs, the committee saw greater value in considering the measures as a whole and identifying a parsimonious panel of measures that would be complete, interoperable, and efficient.

The committee also considered the stability of the measure and the implications for how often they need to be assessed. Sociodemographic

characteristics of the person, which help determine their resources and adverse exposures, are relatively unlikely to change, especially once one reaches adulthood. Other risk factors are more fluid. Although some health behaviors are habitual, they may fluctuate independently or in response to an intervention or treatment. Social relationships and affective states are likely to vary over time and with changing circumstances.

The stability of a domain affects the frequency with which it needs to be assessed for use in direct patient care (for screening, intervention, or monitoring), by the health system or public health, or for purposes of research. A stable domain can be assessed once at entry; others require periodic screening with detailed assessment and follow up on a positive screen. Others, such as the patient’s address, require verification at every visit. Frequency of assessment is a factor in evaluating the feasibility of including a measure of the domain in EHRs; a detailed assessment may be feasible if needed only at entry, but not if needed frequently. Similarly, domains that can be assessed with a brief screen with targeted follow up are more feasible than in-depth assessments that need to be repeated.

Constructing a Parsimonious Panel of Measures:
Usefulness, Readiness, and Committee Judgment

In Phase 2, evaluations of measures collapsed the four criteria into two dimensions. One dimension represents the readiness of a measure for use in the EHR. Readiness was quantified across the following criteria: availability of a standard, freely available measure; feasibility; and lack of sensitive information. The second dimension represents the usefulness of having the information generated by the measure in the patient record for clinical, population management, and research purposes. Table S-2 presents a summary of the committee’s process from its identification of conceptual frameworks to its final steps in constructing a parsimonious panel of measures.

Figure 5-1 in Chapter 5 displays all the measures and where they fall on these two dimensions. The committee also assigned an overall score to each measure. In general, the strongest endorsement was for measures that scored highest on both readiness and usefulness, but some measures that were high on usefulness but somewhat lower on readiness were also included. Informed by these ratings, the committee reached consensus on four domains that are currently being collected in many clinical settings and eight domains that are not yet routinely collected. Table S-3 summarizes the 11 domains and 12 measures that compose the selected panel and the number of questions in each measure.

TABLE S-2 The Committee’s Analytic Process in Narrowing Domains and Measures to a Parsimonious Measurement Panel

NOTE: EHR = electronic health record; SBD = social and behavioral determinants.

IMPLEMENTATION CHALLENGES

The need to adopt new standards and incorporate more social and behavioral information in EHRs is driven by the contrasts in the performance of the U.S. health system, which has achieved technological advances but is fragmented, uncoordinated, costly, and yielding poor population health outcomes. Implementing changes to EHRs involves not just modifications to technologies but also an expanded view of the determinants of health and adaptation in the way clinical teams work and how patients engage in their own care.

It is beyond this committee’s charge to address the general challenges of EHR use. The committee was acutely aware that adding additional data to the EHR could increase the burden on health systems, clinicians, patients, vendors, as well as implementers of meaningful use regulations. Accordingly, the committee used a systematic approach to weighing the trade-offs and aimed for the most parsimonious set of measures.

TABLE S-3 Core Domains and Measures

NOTE: Q = question(s).

Self-Reported Data

Most of the recommended measures rely on self-reported data. Obtaining such data does not need to add to clinicians’ time as it does not necessarily need to be collected through an interview with a member of the clinical team. It can also be collected directly from the patient on paper or via a computer. Self-report can be subject to error and bias, and it is important for health care systems to help patients understand the purpose and the value of the information being collected. Future technological advances may allow collection of more objective indicators and information on experiences that individuals may not be able to remember and report reliably. For example, sensors which record data for review and upload to the EHRs if appropriate—while not without their own limitations—may eliminate or reduce the need for having to ask individuals about behaviors such as exercise or sleep.

Select measures for some determinants of health may be found in other sources related to the patient, including EHRs from other institutions; personal health records; health risk appraisals gathered by insurers, employers, or clinical data registries; community agency datasets; national surveys; and datasets gathered by third-party data integrators such as retail. Presently there are few straightforward ways to transfer data from external data sources to EHRs or vice versa.

The concept of a robust data infrastructure in a recent report, Robust Health Data Infrastructure (AHRQ, 2014), developed by the JASON/MITRE Corporation, offers potential in ensuring that data flows needed to make social and behavioral determinants of health accessible to the patient,

to the clinical care team, to the health system, and to society are realized. Data could be stored at the point of acquisition and integrated at the point of need. With such an open architecture, the committee’s recommended measures could be acquired from a wide variety of sources.

Privacy Protections

Risks to the patient in some sensitive areas, such as substance use or violence, represent considerable challenges to collecting data. However, basic safety measures are widely used. When possible, data can be de-identified to provide anonymity. For example, in syndromic surveillance the public health entity only needs to know how many cases there are—and perhaps associated information such as age, sex, and neighborhood—but the specific names of individuals are not needed. Privacy concerns are more likely in cases where there is a need to individually link EHRs to a public health registry and the data cannot be de-identified. However, the transmitted data can be encrypted.

Institutions should inform patients about the specifics of data sharing. For example, if data are being shared with public health officials, patients should be informed that this is occurring and informed of the benefits that may incur through sharing that information. Routine collection of these types of potentially sensitive data may have the additional benefit of normalizing or destigmatizing their discussion in clinical practice.

Linking Data

Linking data from EHRs to local public health departments and community agencies provides several advantages to patients, providers, and the broader community. Information can flow in both directions. For example, data in EHRs can enable public health practitioners to identify groups of persons affected by environmental pollutants and identify areas that may need environmental mitigations. Clinicians can use geocoded environmental data to coach individual patients on risk mitigation or to tailor treatment.

Public health departments or community agencies are often in the best position to address certain problems, such as food insecurity, lack of housing, and social isolation. The manner in which social and behavioral domains may be addressed falls far outside the typical interventions found in health care. For example, food insecurity may be alleviated by access to government-funded food assistance programs, but patients may need help in navigating the enrollment process, or individuals may benefit from health interventions such as group home visits, but some may also need community-level interventions.

Resource Considerations

The business model for capturing social and behavioral domains and measures in the EHR has yet to be fully realized, and few examples exist. The committee believes that cost savings will accrue from addressing the social and behavioral determinants of health. However, those bearing the costs of identifying and addressing these determinants do not necessarily benefit from the resulting savings. The benefits accrue to society, health care payers, and health systems who are reimbursed for population health management. While some of these benefits are near term, many accrue over years. The costs of adding social and behavioral domains to EHRs, such as programming, modifying workflows, and intervening on positive screens often fall on the individual health practice or hospital. The movement toward population health management and accountable care organizations may address this malalignment over time. In the meantime, costs remain a barrier.

The ultimate value of incorporating the social and behavioral domains of health in the EHR lies in engaging the patient and aligning health service and care. Such redesign is a long-term answer to facing and addressing the implementation challenges summarized here. The barriers and suggested interventions highlighted are intended to act as a reference to guide stakeholders along this journey.

LOOKING FORWARD

The inclusion of the committee’s recommended measures in all EHRs (as well as those which are appropriate for specific populations) will enable:

• More effective treatment of individual patients in health care settings,
• More effective population management for health care systems and for public health agencies, and
• Discovery of the pathways that link social and behavioral factors to functioning, disease processes, and mortality that may inform new treatments and interventions.

The committee’s judgments and recommendations necessarily reflect not only the current status of knowledge about the social and behavioral determinants of health and of the measures of the identified domains of health determinants, but also a tactical decision of the committee to put forward at this time a parsimonious initial set of social and behavioral domains and measures for inclusion in EHRs. A number of domains and measures narrowly missed inclusion in this set, and are thus readily avail-

able to be added to EHRs when the opportunity next arises. In addition, over the coming years, new research may point to the importance and usefulness of domains and measures that were not selected based on current knowledge. A number of measures are very promising and potentially important, but the committee found that they currently fell short on aspects of readiness for inclusion in all EHRs. These domains and measures that were not included in the recommended panel merit greater attention as valuable targets of research. What follows are the committee’s findings and recommendations.

FINDINGS AND RECOMMENDATIONS

Finding 5-1: Four social and behavioral domains of health are already frequently collected in clinical settings. The value of this information would be increased if standard measures were used in capturing these data.

Finding 5-2: The addition of selected social and behavioral domains, together with the four domains that are already routinely collected, constitute a coherent panel that will provide valuable information on which to base problem identification, clinical diagnoses, treatment, outcomes assessment, and population health measurement.

Finding 7-1: Standardized data collection and measurement are critical to facilitate use and exchange of information on social and behavioral determinants of health. Most of these data elements are experienced by an individual and are thus collected by self-report. Currently, EHR vendors

and product developers lack harmonized standards to capture such domains and measures.

Finding 7-2: The addition of social and behavioral data to EHRs will enable novel research. The impact of this research is likely to be greater if guided by federal prioritization activities.

Finding 7-3: Advances in research in the coming years will likely provide new evidence of the usefulness and feasibility of collecting social and behavioral data beyond that which is now collected or which is recommended for addition by this committee. In addition, discoveries of interventions and treatments that address the social and behavioral determinants and their impact on health may point to the need for adding new domains and measures. There is no current process for making such judgments.

With the passage of the Patient Protection and Affordable Care Act of 2010 (ACA) the United States has begun to expand health coverage to millions of uninsured Americans, and the nation is poised to reduce existing

health disparities. Currently, the absence of social and behavioral determinants of health in EHRs limits the capacity of health systems to address key contributors to the onset and progression of disease. The addition and standardization of a parsimonious panel of social and behavioral measures into EHRs can help spur policy, system design, interoperability, and innovation to improve health outcomes and reduce health care costs.

REFERENCES

AHRQ (Agency for Healthcare Research and Quality). 2014. A robust health data infrastructure. AHRQ Publication No. 14-0041-EF. Washington, DC: Agency for Healthcare Research and Quality.

Elo, A.-L., A. Leppänen, and A. Jahkola. 2003. Validity of a single-item measure of stress symptoms. Scandinavian Journal of Work, Environment & Health 29(6):444–451.

WHO (World Health Organization). 1946. Preamble to the constitution of the World Health Organization as adopted by the International Health Conference, New York, 19–22 June 1946. http://www.who.int/about/definition/en/print.html (accessed October 22, 2014).

This page intentionally left blank.

A version of the following Chapters 1–3 first appeared in Capturing Social and Behavioral Domains in Electronic Health Records: Phase 1, originally released on April 8, 2014. Editorial and content changes have been made to these three chapters for inclusion in this report, Capturing Social and Behavioral Domains and Measures in Electronic Health Records: Phase 2. Substantive content changes have been footnoted.

This page intentionally left blank.