3

Applying Selection Criteria to Social Risk Factors and Health Literacy

In its first report, the committee presented a conceptual framework that illustrates the primary hypothesized conceptual relationships by which five social risk factors—socioeconomic position (SEP); race, ethnicity, and cultural context; gender; social relationships; and residential and community context—as well as health literacy may directly or indirectly affect measures of health care use, health care outcomes, and resource use outcomes among Medicare beneficiaries (NASEM, 2016a). The conceptual framework applies to all Medicare beneficiaries, including beneficiaries with disabilities and those with end-stage renal disease (ESRD). Although the committee acknowledges heterogeneity among Medicare beneficiaries (including among beneficiaries age 65 and older), the committee expects the effect of social risk factors to be similar for all Medicare subpopulations. As described in its first report, the committee considers variations in the effect of social risk factors among beneficiaries under age 65 with disabilities, beneficiaries age 65 and older, and beneficiaries with ESRD to fall within a continuous spectrum of effects. Notably, Medicare beneficiaries with disabilities differ systematically from persons with disabilities more generally, because eligibility for federal disability benefits requires that a person is unable to work, has a low income, and meets certain medical criteria (SSA, n.d.). Therefore, Medicare beneficiaries with disabilities are by definition a socially at-risk group. Additionally, the Centers for Medicare & Medicaid Services (CMS) uses the same measures to assess quality, outcomes, and resource use for Medicare beneficiaries regardless of the origin for entitlement (i.e., whether an individual qualified because of age, disability, or ESRD). The committee still holds these assumptions to be true.

In its first report, the committee also identified specific indicators that correspond to the five social risk factors. These indicators represent ways to measure the latent constructs of the social risk factors and are distinct from specific measures. For example, education is an indicator of socioeconomic position that can be measured in multiple ways (e.g., highest degree attained, years of education). Figure 3-1 presents a modified version of the committee’s conceptual framework, expanded to include indicators of each social risk factor. The framework also groups the domains that the committee embraced in its expanded definition of “health outcomes” in its first report (health care use, health care outcomes, and resource use) under the umbrella of performance indicators for value-based payment (VBP).

In this chapter, the committee applies the criteria identified in Chapter 2 of this report to these social risk factors and health literacy (and their respective indicators), and also identifies the rationale and limitations of each factor and indicator relative to those criteria. To review, the committee identified three broad categories of criteria for selecting social risk factors that could be accounted for in Medicare VBP programs:

1. The social risk factor is related to the outcome;
2. The social risk factor precedes care delivery and is not a consequence of the quality of care; and
3. The social risk factor is not something the provider can manipulate.

The committee also identified practical considerations. These are empirical questions that may be best assessed using specific measures and data. Additionally, data collection and measurement is the subject of the committee’s fourth and next report. Because the committee can recommend new data sources and new methods of data collection in this next report (and is therefore not limited to existing measures and data sources), the criteria related to practical considerations such as issues of measurement feasibility are not discussed exhaustively in this report.

SOCIOECONOMIC POSITION

Socioeconomic position describes an individual’s absolute and relative position in a society’s stratification system. SEP captures a combination of access to material and social resources as well as relative status—prestige- or rank-related characteristics (Krieger et al., 1997). As described in the committee’s first report (NASEM, 2016a), the committee prefers socioeconomic position to the more common phrase socioeconomic status, because socioeconomic position is a broader term encompassing resources as well as status (Krieger et al., 1997). SEP is commonly measured using indicators including income and wealth, education, and occupation and employment. In the medical field, insurance status (whether someone has insurance and the type of insurance—i.e., public or private) is also used as a proxy for SEP.

Income

Individual income can affect health and health care outcomes through multiple pathways (Braveman et al., 2005). It can affect health directly as a means of purchasing health care and indirectly as a means of acquiring health promoting resources, such as better education, housing, and nutrition (Adler and Newman, 2002). This effect is stronger at lower levels of income (i.e., poverty and deprivation). However, the association between income and health is graded such that increases in income are associated with increases in health status above a threshold of material deprivation (i.e., poverty level) (Adler et al., 1994; Braveman et al., 2010; NASEM, 2016a). Thus, literature supports a conceptual relationship between income and health-relevant measures of interest to Medicare quality and payment programs. As identified in the committee’s first report, literature indicates that income may influence health care utilization, clinical processes of care, health care costs, health outcomes, and patient experience (NASEM, 2016a). Therefore, literature also supports an empirical association as well. Income is generally not a consequence of health care. Income is potentially susceptible to rapid changes as a consequence of a health event across individual trajectories. However, income’s average association with health care outcomes is not likely to change rapidly. In other words, income precedes care delivery and is not a consequence of the quality of care.

Income is the most commonly used measure of economic resources (Braveman et al., 2005), largely because there are available measures, but also because income has strong face validity. When self-reported, measuring income can be sensitive to collect, but reliable methods exist to accurately, reliably, and feasibly collect income data (Moore and Welniak, 2000). These measures are likely to be resistant to gaming or manipulation. Although there may be less variation in income among older populations, especially the very old (age 80 and older), because income includes both earned and unearned income, there is likely to be sufficient variation in income among the Medicare population (albeit a narrower range than among the general

population) to capture the full variation in SEP (HHS, 2015c). In sum, income is related to health care outcomes of interest, precedes care delivery and is not a consequence of the quality of care, is not something the provider can manipulate, and meets practical considerations.

Wealth

Wealth is an alternate measure of economic resources that represents total accumulated economic resources (assets). Wealth is likely to be as important for health and health care outcomes as is income as a means of acquiring health care and health-promoting resources (Braveman et al., 2005; Deaton, 2002; NASEM, 2016a). Moreover, whereas income may capture less variation in economic attainment among Medicare beneficiaries, especially the very old, wealth may capture more variation. Therefore, wealth may be a more sensitive indicator of SEP for the very old (Allin et al., 2009). Wealth can also buffer the effects of changes in income (such as those due to unemployment or illness) (Cubbin et al., 2011). However, wealth may still be susceptible to changes as a consequence of health events among individuals (Lee and Kim, 2008). For example, onset of a new chronic condition may require out-of-pocket medical expenditures and costs associated with rearranging housing or transportation. These costs can lead an individual to incur both a sudden increase in health-related costs as well as costs that accrue over time, both of which could deplete wealth. This may be particularly relevant for low-income persons who also share a disproportionate of disease burden. However, as with income, the association between wealth and health at a population level is unlikely to be a consequence of health care. An additional challenge of using wealth as an indicator of SEP is that there are substantial differences by subgroups, especially racial and ethnic subgroups and by gender. For example, blacks have significantly less wealth compared to whites even at the same income levels (Kochhar and Fry, 2014; Shapiro et al., 2013). Moreover, relatively few studies have examined the relationship between wealth and health care outcomes (Braveman et al., 2005; NASEM, 2016a). Hence, there is little evidence documenting an empirical association. This may be due in large part to the difficulty of measuring net worth. Like income, it can be sensitive to assess. Unlike income, although some good measures exist, missing data at the upper and lower ends of the wealth distribution can be problematic (Cubbin et al., 2011; Eckerstorfer et al., 2015; Eggleston and Klee, 2015). Collecting self-reported net worth is challenging because many individuals do not know the value of their net worth or what assets they have (Braveman et al., 2005; Eggleston and Klee, 2015). Some studies have used simplified or proxy measures such as home or car ownership, but there remains little empirical evidence on the association between wealth and health care outcomes (Braveman et al., 2005). Literature supports a conceptual relationship between wealth and health care outcomes of interest, but a lack of available measures and thus evidence of an empirical association present limitations for using wealth as an indicator of SEP. Practical considerations present challenges for collecting accurate wealth data.

Insurance Status: Dual Eligibility

In health research, numerous studies assess the effects of insurance coverage on health status (see, for example, IOM, 2009a), but its use here is restricted to its use as a proxy for resources to support health and health care and thus as an indicator of SEP. For the Medicare population, Medicaid eligibility—also referred to as dual (Medicare and Medicaid) eligibility—is an indicator of insurance status that can be used as a proxy that captures elements of both income and wealth. Dual eligibility captures elements of income, because Medicaid eligibility

requires an income below a certain threshold (set at a national minimum level of 133 percent of the federal poverty level [FPL]) (CMS, n.d.-a). However, like any measure of insurance, it is generally an imperfect proxy of income, because it does not capture the full continuum of SEP. This is particularly true for dual eligibility, which is a dichotomous measure representing high or low income. Additionally, individuals with low incomes that exceed Medicaid income thresholds may be eligible for Medicaid coverage under “spend down” rules that allow medically needy individuals to spend down (or, subtract) medical expenses from their income (CMS, n.d.-b). Dual eligibility also captures elements of wealth, because Medicaid eligibility also includes asset limits (CMS, n.d.-a). Similar to income, individuals with few assets that nonetheless exceed the Medicaid asset threshold may “spend down” their assets to become eligible for Medicaid coverage. Because dual eligibility status interacts with the health system in this way, it is a measure that captures both income and wealth in a particular functional form (that of the eligibility criteria) that may or may not be the best predictor of performance indicators used in VBP. Importantly, because states establish Medicaid eligibility, what dual eligibility represents also varies by state. Similarly, Medicaid covers long-term care for those who meet additional eligibility requirements, in which case dual eligibility would capture still another functional form of health-related resources among institutionalized persons.

Dual eligibility may also capture dimensions of health status that are unmeasured by other data sources, because it represents insurance coverage as a concept distinct from SEP. For example, dual eligibles receive more generous health coverage through Medicare and Medicaid than uninsured or underinsured persons who have relatively higher SEP, but who are ineligible for Medicaid coverage because they have income and/or wealth just above the eligibility threshold. Relatedly, dual eligibility may capture clinical characteristics covering those who are under age 65 and eligible for Medicaid coverage based on disability. As noted in the introduction, the committee expects social risk factors to operate similarly among all Medicare beneficiaries including disabled persons. However, the committee notes that in its use here as a proxy measure for SEP as a social risk factor that could be accounted for in Medicare quality measurement and payment (and not as a characteristics of the population to which the social risk factor framework applies), dual eligibility may capture health status–related elements of disability because of eligibility criteria for Medicaid coverage based on disability. Because dual eligibility captures elements of income, wealth, and health status, dual eligibility can be considered a broader measure of health-related resource availability that captures medical need.

Dual eligibility is empirically associated with health and health care outcomes including health care utilization, clinical processes of care, and patient experience (NASEM, 2016a). Dual eligibility also has face validity, particularly among health and medicine researchers, and is a relatively easy to measure and collect. Additionally, dual eligibility is not a consequence of care and likely to be resistant to gaming and manipulation at the population level. Hence, dual eligibility is an available proxy measure of resources available for goods and services to support health and health care.

Occupation

Occupation includes both employment status (whether an individual participates in the paid labor force or not, and if so, to what degree), as well as the type of occupation among the employed (Adler and Newman, 2002; NASEM, 2016a). Additionally, occupation can be collected in its current state or in a past state, as primary lifetime occupation. Among Medicare beneficiaries, fewer of whom participate in the paid work force than the general population—

especially disabled Medicare beneficiaries who by definition cannot work—employment status may be more relevant than job type. Occupation can affect health through exposure to environmental health hazards as well as through psychosocial risks associated with job strain, lack of control, and increased stress (Kasl and Jones, 2000; Theorell, 2000). Additionally, literature suggests that employment and occupation are associated with health outcomes including unhealthy behaviors, morbidity, and mortality (NASEM, 2016a). Literature therefore supports a conceptual relationship between occupation and performance indicators used in VBP. However, there is relatively little empirical evidence on the association between employment or occupation and health care outcomes, especially using U.S. data (NASEM, 2016a). This is likely because of the difficulty of collecting and classifying occupation in the United States. Measures of occupation and employment are likely to be resistant to gaming and manipulation and the United States maintains a Standard Occupational Classification System, but many of the categories are too heterogeneous to be meaningful (Braveman et al., 2005). Additionally, some groups such as retired persons and homemakers may not have an employment related to occupation, making it difficult to identify their SEP. Despite these measurement challenges, occupation and employment are not logical consequences of the quality of care, although like other measures of SEP, employment is potentially susceptible to changes as a consequence of a health effect, such as losing a job because one becomes too ill to work. However, again, at the population level, occupation is unlikely to be a consequence of health care quality. In short, like wealth, occupation is a conceptually powerful indicator of SEP, but practical considerations limit its potential use.

Education

Education is important for health because it shapes future employment and economic resources (Adler and Newman, 2002; IOM, 2014; NASEM, 2016a). Education can therefore affect health indirectly through other indicators of SEP—employment, occupation, and income. At the same time, education can also affect health by enabling individuals to access and understand health information and health care and to make decisions that promote health and reduce health risks, and by contributing to a patient’s ability to advocate for him-or herself in health care (Cutler and Lleras-Muney, 2006; IOM, 2014). Thus, literature supports a conceptual relationship between education and performance indicators used in VBP. Education is strongly associated with health behavior, health status, morbidity, and mortality (IOM, 2014). However, the relationship between health and health care outcomes may vary across age cohorts owing to changes in the distribution of education over time (Lynch, 2003). Nevertheless, as identified in the committee’s first report, literature indicates that education may influence health care utilization, health outcomes, and patient experience, thus providing support for an empirical association (NASEM, 2016a). Education has face validity, precedes care delivery, and is not a logical consequence of care. Education can be measured as continuous or categorical years of schooling completed or as educational attainment measured by credentials of formal schooling (e.g., high school diploma, college degree) (Braveman et al., 2005; IOM, 2014). These measures are feasible to collect and likely to be resistant to gaming. In short, education is related to health care outcomes of interest, precedes care delivery and is not a consequence of the quality of care, is not something the provider can manipulate, and meets practical considerations.

Summary

Income and education are promising indicators. Wealth is likely to be strongly associated with health and health care outcomes, but accurate data is difficult to collect. Dual eligibility meets practical criteria and can be considered a proxy for SEP as a measure of resources available for goods and services to support health and health care. Occupation is likely to be strongly associated with performance indicators used in VBP, but practical considerations limit its potential use.

RACE, ETHNICITY, AND CULTURAL CONTEXT

Race and Ethnicity

Race and ethnicity are social categories that represent dimensions of a society’s stratification system by which resources, risks, and rewards are distributed. Categories of race and ethnicity capture a range of health-relevant dimensions, especially those related to social disadvantage. These include access to social institutions and rewards; behavioral norms and other sociocultural factors; inequitable distribution of power, status, and material resources; and psychosocial exposures like discrimination and bias (Phelan and Link, 2015; Williams, 1997). Race and ethnicity are strongly associated with health and health care outcomes, even after accounting for measures of SEP (Krieger, 2000; LaVeist, 2005; NASEM, 2016a; Williams, 1999; Williams et al., 2010). This effect may be caused by the lack of comparability of a given SEP measure across racial and ethnic groups (for example, as described above, wealth is differentially correlated with income by race), the importance of other unmeasured social factors that are patterned by race and ethnicity (for example, neighborhood environments, discrimination, immigration-related factors, language), and measurement error in SEP (NASEM, 2016a). Together, this literature supports a conceptual relationship between race and ethnicity and health. In its first report, the committee identified literature indicating that race and ethnicity may influence health care utilization, clinical processes of care, health care costs, health outcomes, patient safety, and patient experiences of care (NASEM, 2016a). Thus, literature supports an effect. Race and ethnicity precede care delivery and are not logical consequences of care. However, observed differences by race and ethnicity may represent differences in the quality of care received, including differences related to poor communication, poor cultural competence, discrimination, and bias (IOM, 2003a).

Race and ethnicity are typically identified through self-reported categories, and measures of race and ethnicity are resistant to gaming or manipulation. Refinement of standardized race and ethnicity measures is still needed. In health research, Hispanic ethnicity is frequently combined with racial categories. The most commonly used “racial” categories are: non-Hispanic white, non-Hispanic black, Hispanic, and Asian (see, for example, AHRQ, 2016; CMS, 2016). This categorization is problematic because it conceals substantial heterogeneity within certain categories. In particular, there are substantial differences across Asian groups from different countries. Additionally, Hispanic groups from different (Latin American) countries use racial classifications that differ from U.S. racial classifications (for example, who is considered black), because they reflect different sociopolitical constructs (Wade, 1997). Some existing standards include federal standards from the White House Office of Management and Budget, which the Department of Health and Human Services (HHS) is increasingly adopting (CDC, 2010; IOM,

2009c; OMB, 1995), and those recommended in a 2009 Institute of Medicine (IOM) report (IOM, 2009c). Because race and ethnicity are conceptually distinct, these standards recommend using separate items for collecting race and ethnicity. In sum, race and ethnicity are related to health care outcomes, precede care delivery and are not a consequence of the quality of care, are not things a provider can manipulate, and meet practical considerations. At the same time, the committee acknowledges that causal pathways by which race and ethnicity influence health include mechanisms that can be related to quality of care.

Language

Language typically represents language barriers, such as speaking a primary language that is not English, having limited English proficiency, or otherwise needing interpreter services. This includes deaf American Sign Language users. Language barriers are strongly associated with health and health care outcomes—in particular, poorer access to health care, poorer health status, poorer quality care, including less recommended care, and more adverse health events (NASEM, 2016a). Thus, literature supports both a conceptual relationship and an empirical association between language and health care outcomes. Language is not a consequence of health care. Measures of language are resistant to gaming or manipulation and are also relatively easy to assess. The same 2009 IOM report recommending standards for collecting and measuring race and ethnicity data also included recommended standards for language data (IOM, 2009c). Language is thus related to health care outcomes, precedes care delivery and is not a consequence of the quality of care, is not something the provider can manipulate, and meets practical considerations.

Nativity, Immigration History, and Acculturation

Nativity refers to country of origin. Immigration history includes refugee and documentation status, as well as duration in the United States. Acculturation describes the extent to which an individual adheres to the social norms, values, and practices of his or her own ethnic group or home country or to those of the United States (NASEM, 2016a). Because acculturation is expected to increase with the amount of time spent in the United States, duration in the United States is also used as a proxy for acculturation. Nativity and duration in the United States may influence health and health care outcomes through differences in language, communication, and health care use (IOM, 2014). Nativity and immigration history may also expose individual to different health risks or protective factors prior to arriving in the United States. Risks include environmental exposures, infectious diseases, and poverty, whereas protective factors may arise from cultural differences that shape health behaviors such as smoking, diet, and physical activity (IOM, 2014). These characteristics are likely to have important interactions with race and ethnicity (Jerant et al., 2008; Newhouse et al., 2012). Literature therefore supports several pathways by which nativity, immigration history, and acculturation may affect health. In its first report, the committee identified literature indicating that nativity may influence clinical processes of care and patient experience, supporting an empirical association (NASEM, 2016a). Evidence on the relationship between acculturation and health care outcomes is not well established, in part due to measurement challenges (Abraído-Lanza et al., 2006; IOM, 2014; NASEM, 2016a). These factors are not logical consequences of health care or health events.

Measures of nativity include identifying a specific country of origin or a dichotomous measure comparing foreign-born to U.S.-born individuals. These measures of nativity and

measuring duration in the United States could therefore feasibly be collected during an office visit or in an electronic health record. Measures of nativity and time in the United States are also less sensitive than measures of documentation status or citizenship (IOM, 2014). Because there is a strong interaction between acculturation and race and ethnicity, measures of acculturation frequently assess acculturation among specific subgroups (e.g., Hispanic immigrants) (HHS, 2014). Nativity, duration in the United States, and measure of language can be crude proxies for acculturation. Measures of nativity, immigration history, and acculturation are likely to be resistant to gaming or manipulation. All told, measures of nativity and immigration history are related to health care outcomes, precede care delivery and are not a consequence of the quality of care, are not things a provider can manipulate, and meet practical considerations. Literature supports a conceptual relationship between acculturation and health care outcomes of interest, but existing measures pose challenges to feasibility. Consequently, there is a lack of empirical evidence about the relationship between acculturation and performance indicators used in VBP.

Summary

Race, ethnicity, language (especially limited English proficiency), and nativity are promising indicators, particularly in combination (Goodell and Escarce, 2007). Literature supports a conceptual relationship between acculturation and health care outcomes of interest, but existing measures have limitations and empirical evidence is lacking. Documentation status as a measure of immigration history is likely to be sensitive to collect.

GENDER

The committee uses the term gender broadly to capture the social dimensions of gender as distinguished from biological effects of sex. Gender captures both normative gender identity and gender minorities, including individuals who identify as transgender, intersex, or otherwise non-conforming gender. Normative gender categories (men and women) are strongly associated to health and health care outcomes (NASEM, 2016a). However, deconstructing the effects of gender and sex can be challenging. Frequently, investigators do not specify which construct they are measuring and use the terms interchangeably (for example, incorrectly referring to sex differences as gender differences), and because sex and gender may interact to produce health outcomes (Krieger, 2003). Nevertheless, gender has face validity, is not a consequence of care, and there are good self-reported measures that are resistant to gaming. For accountability purposes in Medicare payment, gender is already included as a risk factor in clinical adjustment.

Gender Identity

Gender minorities may experience differences in health and health care outcomes, but there remains little empirical evidence. Additionally, although gender identity is not a consequence of health care, what evidence does exist suggests that differential health care outcomes may arise from miscommunication, lack of cultural competence, or bias in the patient-provider encounter (IOM, 2011). The lack of evidence is due in part to the lack of a good existing measure, although, based on recommendations from a 2011 IOM report, HHS has been actively working to improve data collection. In recent years, questions on gender identity have been included in national surveys such as the Behavioral Risk Factor Surveillance System,

National Health Interview Survey, National Survey of Family Growth, National Survey on Drug Use and Health, and National Health Service Corps Patient Satisfaction Survey (Copen et al., 2016; HHS, 2015b; Ward et al., 2014). Additionally, the Office of the National Coordinator’s (ONC’s) final rule specifying meaningful use criteria included gender identity measures (HHS, 2015a). Measures of gender identity are likely to be resistant to gaming or manipulation, but because there is a very low prevalence of gender minorities, gender identity is unlikely to have a significant effect in adjustment models and other methods of accounting for social risk factors. Emerging literature supports a relationship between gender identity and health care outcomes of interest, but existing measures pose challenges to feasibility. Hence, the empirical association is poorly established.

Sexual Orientation

Sexual orientation includes individuals who identify as lesbian, gay, bisexual, queer, questioning, or otherwise non-conforming. Sexual orientation is typically defined with respect to three dimensions: attraction, behavior, and identity (IOM, 2011). Like gender minorities, sexual minorities may experience differences in health and health care outcomes although there is currently little empirical evidence (NASEM, 2016a). Moreover, as with gender identity, emerging evidence suggests that differential health care outcomes among sexual minorities may be largely attributable to drivers related to the quality of care provided (e.g., miscommunication, poor cultural competence, discrimination) (Elliott et al., 2015; IOM, 2011). Similar to the practical challenges of establishing better evidence between gender identity and health care outcomes, there are no good existing measures, although the HHS has also included sexual orientation items in the surveys discussed above, and ONC also recommended inclusion of sexual orientation in its meaningful use criteria (CDC, 2010; HHS, 2015a,b). One limitation of existing measures is that they frequently only capture one dimension of sexual orientation, and identifying the dimension or dimensions most relevant to the outcome of interest can be conceptually challenging (IOM, 2011). Specifically, some individuals do not present consistently across the three dimensions of sexual orientation. For example, some men report that they have sex with other men, but do not identify as gay. In cases of such inconsistency across dimensions, identifying the dimension or dimensions most relevant for the outcome of interest will be important to accurately classify individuals. Taken together, like gender identity, emerging literature supports a relationship between sexual orientation and health care outcomes of interest, but poor existing measures have limited available evidence.

Summary

Normative gender categories (men and women) are strong candidates for inclusion in accounting methods, despite the fact that effects of gender are difficult to separate from biological effects of sex in measurement. However, the committee notes that gender is already included in clinical risk adjustment. Promising measures of gender identity and sexual orientation that HHS is currently testing and collecting data on could be revisited for potential inclusion when there is more empirical evidence supporting the relationship between gender identity and sexual orientation and health care outcomes. Certainly, in the short term, there is likely to be a very low prevalence of individuals who have non-normative gender identities. In addition, the relationship of these constructs to health care outcomes is not well established.

Thus, accounting for variations in gender identity is unlikely to have a significant effect in accounting methods.

SOCIAL RELATIONSHIPS

Many dimensions of social relationships are important to health, health care use, and health care outcomes (Berkman and Glass, 2000; Cohen, 2004; Eng et al., 2002; Holt-Lunstad et al., 2010; House et al., 1988; Umberson and Montez, 2010). These include access to social networks that can provide access to resources, including material resources and emotional and instrumental social support. Social relationships may be especially relevant to health care access and outcomes among older adults and persons with limitations in activities of daily living (ADLs) and instrumental activities of daily living (IADLs) (Cornwell and Waite, 2009; Hawton et al., 2011; Houser et al., 2010; Seeman et al., 2001; Tomaka et al., 2006). Hence, literature supports a conceptual relationship between social relationships and health care outcomes of interest. In health research, social relationships are typically assessed using three indicators: marital/partnership status, living alone, and emotional and instrumental social support.

Marital/Partnership Status

Marital or partnership status is a foundational structural element of social relationships that is also often considered an important indicator of social support. Being married or partnered is associated with better health care outcomes, while being single, widowed, or otherwise unpartnered is associated with worse health care outcomes (NASEM, 2016a). Literature suggests that this relationship holds true for both heterosexual partners and same-sex couples (Liu et al., 2013). Additionally, the relationship between marriage and health outcomes interacts with gender. Not only might marriage affect health in different ways by gender, but some evidence also suggests that marriage is also more beneficial to men than women (IOM, 2014). Thus, there is a conceptual relationship between marital/partnership status and health. In its first report, the committee identified literature indicating that marital status may influence health care utilization, clinical processes of care, health care outcomes, patient experiences of care, and health care costs (NASEM, 2016a). Thus, there is evidence of an empirical association. It is important to note demographic shifts in family structure over the past several decades—marriage rates have declined while the number of cohabiting individuals and persons who never married has increased (Wang and Parker, 2014). Some evidence suggests that the relationship between marital status and health is changing along with these demographic shifts (Liu and Umberson, 2008). It will therefore be important to monitor the empirical association between marital/partnership status and health and revisit assumptions about their conceptual relationship over time.

Marital or partnership status is not a logical consequence of care, but is potentially susceptible to rapid changes—both gaining and losing a partner—across individual trajectories. However, at the population level, marital/partnership status is not likely to be susceptible to rapid changes. Marital and partnership status is likely to contribute to unique variation in outcomes of interest, especially among older adults. Additionally, there is likely to be greater variability in the future with the increase in the never-married and cohabiting populations, which are increasingly tied to SEP, race, ethnicity, and community of residence (Aughinbaugh et al., 2013; Tamborini, 2007; Wang and Parker, 2014). Measures of marital or partnership status include dichotomous

measures of whether someone is married or not and whether someone is partnered or lacks a partner. Other measures include more categories, such as individuals who are single, widowed, and divorced. These measures are relatively easy and acceptable to collect via self-report and are likely to be resistant to gaming. Marital or partnership status is therefore related to health care outcomes, precedes care delivery and is not a consequence of the quality of care, is not something the provider can manipulate, and meets practical considerations. However, demographic changes suggest that monitoring the relationship between marital/partnership status and health outcomes over time is needed.

Living Alone

Living alone is a structural element of social relationships. In health research, living alone is typically an indicator of social isolation or loneliness, which have been shown to have important consequences for health (Berkman and Glass, 2000; Brummett et al., 2001; Cohen, 2004; Eng et al., 2002; House et al., 1988; Wilson et al., 2007). Living alone is also likely to at least partly capture elements of social support. Thus, literature supports a conceptual relationship between living alone and health care outcomes. Living alone is strongly associated with health, although literature on the association between living alone and health care outcomes is sparse (NASEM, 2016a). Nevertheless, the committee identified literature indicating that living alone may influence health care utilization, clinical processes of care, and health outcomes in its first report (NASEM, 2016a). Living alone is potentially susceptible to rapid changes, including changes resulting from a health care interaction. For example, an ill parent may temporarily move in with his or her child following a health event or the advice of a doctor. However, living alone is not likely to be susceptible to rapid changes on average. Living alone is unlikely to vary across reporting units substantially, although there may be specific geographic regions with substantially higher prevalence of older adults living alone. Therefore, it may be important to measure living alone with regional interactions. Living alone can be fairly easily and feasibly assessed in the clinical setting using a dichotomous measure (living alone or not) or more finely graded household composition measures (e.g., living alone, with one other person, two other persons, and so on). Thus, living alone is related to performance indicators used in VBP, precedes care delivery and is not a consequence of the quality of care, is not something the provider can manipulate, and meets practical considerations.

Emotional or Instrumental Social Support

Social support is a key function of social relationships and includes emotional elements (such as through caring and concern) as well as instrumental components (such as material and other practical supports). Emotional social support may affect health through psychosocial mechanisms—for example, boosting self-efficacy to practice health-promoting behaviors like quitting smoking or to follow a treatment regimen (Berkman and Glass, 2000). Emotional social support may also buffer negative effects of health risks or facilitate health behaviors (IOM, 2014). At the same time, social support can have a negative effect on health, for example, from distress caused by negative social interactions or because negative influences promote risky health behaviors (Uchino, 2006). Instrumental social support can support access to health-promoting resources (e.g., delivery of nutritious meals) and health care (e.g., providing transportation to a doctor’s appointment) (Berkman and Glass, 2000). Hence, literature supports a conceptual relationship between social support and performance indicators used in VBP. In its

first report, the committee identified literature indicating that social support may influence health care utilization, clinical processes of care, health outcomes, and patient experiences of care (NASEM, 2016a). Generally, higher levels of social support are associated with better health care outcomes while lower levels of social support are associated with poorer health care outcomes (NASEM, 2016a). Thus, there is evidence of an empirical association.

Similar to indicators of SEP, social support is potentially susceptible to rapid changes, including changes that result from health care interactions. For example, a person who previously lacked social support may gain it following a health event, because members of their social network reach out to help the person in their recovery. Or, a physician may provide instrumental support such as organizing meal deliveries or transportation services during a clinical encounter. However, on average, social support is not a consequence of the quality of care. For this reason, researchers often measure an individual’s perceived or potential social support through measures of social connections or social integration, which may represent potential sources of social support (IOM, 2014).

Measures of social support are likely resistant to gaming and manipulation, but they may pose feasibility issues. Some measures have many items and are burdensome to collect or may only assess one element of social support (e.g., instrumental but not emotional support; perceived support versus actual support). Additionally, because social support is multidimensional, identifying the measure that represents the most relevant dimension for a given health care outcome can be challenging. Despite these limitations, measures of social support are still likely to capture elements of social relationships that are relevant for health care outcomes. Taken together, emotional or instrumental social support is related to health care outcomes, precedes care delivery and is not a consequence of the quality of care, is not something the provider can manipulate, and generally meets practical considerations, with some limitations.

Summary

Marital status and living arrangements (living alone) are likely to influence health and health care outcomes, are easy to measure, and may at least partly capture elements of emotional and instrumental social support. Some evidence suggests that the relationship between marital/partnership status and health is changing along with demographic shifts, which point to a need to reassess the empirical associations and revisit assumptions about the conceptual relationship over time. Emotional social support and instrumental social support are likely to influence health care outcomes. However, because social support is multidimensional, identifying the measure that represents the most relevant dimension for a given health care outcome may pose both conceptual and practical challenges for data collection and measurement.

RESIDENTIAL AND COMMUNITY CONTEXT

Residential and community context refers to a broad set of characteristics that could be important to health and health care processes. These include compositional characteristics that represent aggregate characteristics of neighborhood residents, characteristics of physical and social environments (i.e., environmental measures), as well as policies, infrastructural resources, and opportunity structures that influence individuals’ everyday lives (NASEM, 2016a).

Compositional Characteristics

Compositional characteristics of communities include, for example, dimensions of SEP, the proportion of racial and ethnic minority residents, foreign-born residents, single-parent households, and English language proficient residents. Compositional characteristics can be interpreted to represent a combination of environmental effects, group-level effects, and as a proxy for effects of individual characteristics. Compositional characteristics might affect health care outcomes in similar ways to their individual-level correlates. For example, lower education or lower income on average may influence health and health care outcomes through differences in accessing health-promoting and health care resources. Compositional characteristics might also affect health care outcomes through genuine group-level effects. For example, one study found that for foreign-born Latinos, living in neighborhoods with high-proportions of foreign-born Latinos was protective for health, potentially through greater levels of social support or through lower levels acculturation and its related health-damaging effects (e.g., less nutritious diets, less physical activity) (Acevedo-Garcia and Bates, 2008). Thus, literature supports a conceptual relationship. Studies have shown that community composition may affect health behaviors and other risk factors, morbidity, and mortality (Diez Roux and Mair, 2010). In its first report, the committee identified literature indicating that compositional characteristics may influence health care utilization, clinical processes of care, health care outcomes, and patient safety (NASEM, 2016a). Thus, literature also supports an empirical association between compositional characteristics and performance indicators used in VBP. Neighborhood compositional characteristics are not logical consequences of care (NASEM, 2016a).

Compositional characteristics can be assessed and used individually—for example, neighborhood racial and ethnic composition or neighborhood SEP. Compositional characteristics can also be assessed using composites, such as a summary indicator of neighborhood deprivation or neighborhood SEP. For example, Roblin (2013) developed a summary measure to assess the neighborhood SEP of a managed care organization’s enrollees measured at the Census tract level using seven indicators: percent of households with income below the FPL, percent of households receiving public assistance, percent of households with low income, percent of unemployed adult males, percent of adults with low educational attainment, median household income, and median home value. Of note, the level of aggregation (e.g., Census tracts, block groups, zip codes) is important when measuring compositional characteristics, because effects may vary based on the units of aggregation used (e.g., Krieger et al., 2002). Additionally, compositional characteristics can be messy to measure, because they can represent an individual characteristic or a genuine area-level effect. Furthermore, when used as a proxy for individual-level effects, they may also pick up area-level (environmental) effects. Measures are likely resistant to gaming or manipulation and relatively easy to assess (IOM, 2014). In total, despite some measurement issues, compositional characteristics of residential and community context are related to performance indicators used in VBP, precede care delivery and are not a consequence of the quality of care, are not things a provider can manipulate, and generally meet practical considerations.

Environmental Measures

Environmental measures are indicators of residential and community context. They represent dimensions of residential environments including the physical or built environment (e.g., housing, walkability, transportation options, and proximity to services—including health

care services) as well as social environments (e.g., safety and violence, social disorder, the presence of social organizations, and social cohesion). Neighborhood environments can affect health through the distribution of health-relevant resources (e.g., access to recreational spaces, healthy foods, or health care services) or by exposing residents to environmental hazards (e.g., air pollution) (Diez Roux and Mair, 2010; IOM, 2003b). Neighborhood environments can also expose residents to physical and social exposures (e.g., decay, safety and violence, discrimination, segregation) that negatively affect health through stress and other psychosocial processes (Diez Roux and Mair, 2010; IOM, 2003b). Thus, there is a conceptual relationship between environmental measures of residential and community context and health care outcomes of interest. Additionally, environmental characteristics are not logical consequences of care.

Although environmental measures are likely to be associated with health and health care outcomes, evidence is currently limited (NASEM, 2016a). Environmental measures are potentially easy to collect, although measures need to be tested further. These measures are also likely to be resistant to gaming or manipulation. For example, a growing body of literature shows that some neighborhoods have substantially fewer safe recreation spaces, purveyors of healthy foods, and health care resources (Blustein et al., 2010; Diez Roux and Mair, 2010). However, evidence regarding the effect of these factors on health care outcomes is still lacking. This is therefore an emerging area of research that could be reevaluated for potential inclusion as more evidence emerges. In sum, environmental measures are conceptually powerful, but this is an emerging area of research and the empirical association with health care outcomes is poorly established. Two environmental measures for which there is more empirical evidence—urbanicity and housing—are discussed in more detail.

Urbanicity

Urbanicity describes where an individual’s place of residence falls on the spectrum from urban to rural. On one end of the spectrum, rural areas are associated with poorer access to health care owing to both distance and availability. Rural areas are also associated with increased risks from environmental hazards associated with rural industries such as pesticides in farming (IOM, 2003b). On the other end, urban areas may have regions with concentrated disadvantage that expose residents to negative effects of poverty, negative psychosocial exposures, and physical decay. Cities may also expose residents to environmental hazards associated with air pollution and safety hazards of old or densely populated buildings (IOM, 2003b). Thus, urbanicity is conceptually related to health care outcomes of interest. In its first report, the committee identified literature indicating that urbanicity may influence health care utilization, clinical processes of care, health care costs, and patient experiences of care, particularly at the far ends of the spectrum (NASEM, 2016a,b). This supports an empirical association. Urbanicity is not a logical consequence of care.

Urbanicity can be measured dichotomously (i.e., urban or rural), trichotomously (i.e., urban, suburban, rural), or on a graded spectrum (e.g., percent urban). Urbanicity can be measured as a provider or patient characteristic. Urbanicity as a provider characteristic (e.g., urbanicity of a hospital) can only measure between-unit effects, whereas patient urbanicity (e.g., rural versus urban patients) can be used to assess both within- and between-unit effects. However, patient urbanicity may differ in significant ways across provider urbanicity because, for example, rural patients who receive care from urban hospitals are likely to differ significantly from rural patients who receive care at rural hospitals. Measures are resistant to gaming and manipulation, and they are relatively easy to collect. However, assessing urbanicity may pose

some potential measurement issues related to identifying the appropriate size to avoid misclassification (Krieger et al., 2002). For example, at the Census tract level, there can be substantial variation in population and geographic size. Additionally, Census tracts may be too small to capture truly rural or urban areas, misclassifying, for example, areas within a large metropolitan county as “rural” or small towns in rural areas as “urban” (Hart et al., 2005). Taken together, urbanicity is related to health care outcomes, precedes care delivery and is not a consequence of the quality of care, is not modifiable through provider action, and generally meets practical considerations, with some limitations.

Housing

Health-relevant elements of housing include housing stability homelessness, and quality and safety. Homelessness and housing instability (lack of access or threats to reasonable quality housing) (Frederick et al., 2014) are associated with lower access to care and higher physical and mental morbidity, as well as increased mortality (NASEM, 2016a). Additionally, poor housing conditions can expose individuals to harmful exposures such as lead or poor air quality, infectious disease, poor sanitation, and injury (IOM, 2003b; NASEM, 2016a). Thus, literature supports a conceptual relationship. Substantial literature supports associations between poor housing, housing instability, and homelessness with a wide range of health conditions covering physical and mental health (IOM, 2003b; Krieger, 2003). However, the empirical association between housing and health care outcomes is less well established. Literature suggests that homeless persons have high hospital readmission rates (Buck et al., 2012; Doran et al., 2013). In its second report, the committee also identified case studies in which housing conditions—stairs and loose wires—were considered risk factors for poor health care outcomes (e.g., falls) (NASEM, 2016b). In its first report, the committee identified a small number of studies examining the relationship between type of residence (namely, private or institutional postdischarge residence) and readmissions, and these studies found no association with either short-term (30-day) or long-term (1-year) readmissions (NASEM, 2016a). To that end, housing is potentially susceptible to rapid changes as a consequence of health care. For example, after a serious health event, a hospital may discharge a patient to an institutional setting such as a skilled nursing facility, which may have resources and conditions that differ substantially from the patient’s residence in the community. However, at the population level, housing is unlikely to be susceptible to rapid changes.

Measures of housing and homelessness are likely to be resistant to gaming or manipulation, but currently present some practical limitations. Homelessness is typically assessed using counts, which requires large teams to physically count homeless persons residing within a given geographic area (HUD, 2012). Some measures of housing insecurity also exist (e.g., how often an individual was worried about paying rent in the past month) (CDC, 2013), but these measures tend to be proxies for financial stress or SEP rather than assessing housing adequacy. Other measures, such as housing characteristics collected through the Medicare Current Beneficiary Survey (CMS, 2006) and those the Department of Housing and Urban Development uses to assess housing quality under its Section 8 program include many items requiring comprehensive inspections and can therefore be burdensome to collect (HUD, 1998).

Summary

Compositional characteristics and environmental measures of residential and community context are related to health care outcomes, precede care delivery and are not a consequence of the quality of care, are not modifiable through provider action, and generally meet practical considerations, with some limitations. A measure of Census-tract neighborhood deprivation (i.e., a composite measure of neighborhood compositional characteristics) is likely a good proxy for a range of individual and true area-level constructs (compositional and environmental) relevant to performance indicators used in VBP. These measures are also feasible to obtain. Measures of urbanicity and housing are also available. Environmental measures are an emerging area of research and other measures could be revisited for potential inclusion when there is more empirical evidence and better measures.¹

HEALTH LITERACY

Health literacy is “the degree to which individuals have the capacity to obtain, process, and understand basic health information and services needed to make appropriate health decisions” (IOM, 2004, p.32). The committee does not conceive of health literacy as a social risk factor, but rather as the product of an individual’s skills and abilities (e.g., reading and other critical skills), social and cultural factors, education, health system demands, and the health care context (IOM, 2004). However, the committee included health literacy in its conceptual framework and retained it for consideration in this report because health literacy is included in the committee’s charge and because it is specifically mentioned in the Improving Medicare Post-Acute Care Transformation (IMPACT) Act and therefore of interest to Congress. Additionally, social risk factors like education and language influence health literacy. The committee’s conception of health literacy also captures the related concept of numeracy, or the ability to understand information presented in mathematical terms and to use mathematical knowledge and skills in a variety of applications across a variety of settings (IOM, 2014).

By definition, health literacy and numeracy are conceptually related to health care outcomes. Furthermore, in part because health literacy and numeracy are defined as barriers to accessing health care and adhering to treatment regimens, they may have strong face validity, especially among health care professionals. Low health literacy is associated with poorer knowledge of disease management and health-promoting behaviors and poorer health status (IOM, 2004). In its first report, the committee also identified literature indicating that health

___________________

¹ The committee sees no conflict between this report and the 2013 IOM report Variation in Health Care Spending: Target Decision Making, Not Geography, which recommended against using area-level payment adjustments to account for regional practice patterns. That committee’s charge was to evaluate whether area-level differences in per-beneficiary spending were real and if so, to develop explanations for the variation. That report examined whether health care markets (characterized using relatively large geographies such as hospital service areas, hospital referral regions, or metropolitan statistical areas) were characterized by persistent patterns of spending driven by commonalities in medical decision making or other provider behavior and concluded that area spending variability was mainly due to price markups in the commercial insurance market and variation in the use of post-acute care in Medicare. In contrast, this report focuses on differences in performance indicators used in VBP (including variations in health care utilization and resource use, but also quality) driven by differences in social characteristics of a provider or other risk-bearing entity’s patient population. The use of area-level measures is therefore at much smaller geographic units (e.g., Census tracts of patient place of residence) and serves to more accurately characterize providers’ patient populations in Medicare quality measurement and payment programs.

literacy may influence health care utilization, clinical processes of care, health care cost, and patient experiences of care (NASEM, 2016a). This literature supports an empirical association. There is less evidence on effects of numeracy on health and health care outcomes (NASEM, 2016a). Available measures of health literacy and numeracy exist, but some of these instruments are long and may be burdensome to collect in the clinical setting (IOM, 2009b). Others capture limited components of health literacy—for example, reading and writing skills, but not listening and speaking skills, or an individual’s lack of background knowledge or cultural differences that may influence his or her understanding (IOM, 2004). The committee expects these measures to be resistant to gaming and manipulation.

Although the committee acknowledges that the burden of improving health literacy does not fall solely on the health care system, the health care system does carry significant responsibility. Health care providers can mitigate the effects of low health literacy (IOM, 2004; Pleasant et al., 2016). For example, a systematic review identified methods that are effective at improving patient health literacy (Berkman, 2011). Thus, taking a universal precautions approach, which assumes that it may be difficult for all patients to understand health information and access health services, health care providers can tailor care to each patient’s level of health literacy and numeracy to ameliorate the effects that low health literacy and numeracy have on health care outcomes (Kripalani et al., 2014). Similarly, health literate health care organizations can align the demands of the health care system with patients’ skills and abilities to make it easier for patients to access, understand, navigate, and use health information and health care services (Brach et al., 2012; IOM, 2012). Thus, health literacy is something providers can act upon and can be a consequence of the quality of health care provided.

Summary

Health literacy is related to health care outcomes of interest and generally meets practical considerations. However, provider actions can mitigate the effects of low health literacy. Thus, to preserve incentives to provide effective care to patients with low health literacy, it would not be desirable to adjust or otherwise account for differences in health literacy in performance indicators used in VBP. Nevertheless, it may be desirable to reward or incentivize the greater effort or greater costs required to provide health literate care and thereby produce good health care outcomes in other ways.

SYNOPSIS

Table 3-1 summarizes the social risk factors as well as health literacy, along with their rationale for inclusion in methods to account for them and potential limitations. In the table, specific criteria as they apply to indicators of social risk factors are indicated using the criteria numbers from the previous chapter (in parentheses). To review, the criteria are:

1. The social risk factor is related to the outcome.
   1. The social risk factor has a conceptual relationship with the outcome of interest.
   2. The social risk factor has an empirical association with the outcome of interest.

2. The social risk factor precedes care delivery and is not a consequence of the quality of care.
   3. The social risk factor is present at the start of care.
   4. The social risk factor is not modifiable through provider actions.
3. The social risk factor is not something the provider can act upon and manipulate.
   5. The social risk factor is resistant to manipulation or gaming.

TABLE 3-1 Application of Selection Criteria to Indicators of Social Risk Factors and Health Literacy

After applying the selection criteria to indicators of the five social risk factors and health literacy, the committee made the following conclusions:

• Income, education, and dual-eligibility;
• Race, ethnicity, language, and nativity;
• Marital/partnership status and living alone; and
• Neighborhood deprivation, urbanicity, and housing.

• Wealth,
• Acculturation,
• Gender identity and sexual orientation,
• Emotional and instrumental social support, and
• Environmental measures of residential and community context.

IMPLEMENTATION CONSIDERATIONS

The committee applied selection criteria to each social risk factor and relevant indicators of these factors individually. However, as discussed in the previous chapter, the goal is to identify a set of measures that perform well together. To that end, a combination of measures might perform differently than the sum of its parts. Additionally, some social risk factors may have regional interactions. For example, as previously described, living alone may not vary substantially across reporting units except in certain communities with exceptionally high proportions of older adults living alone. Furthermore, as discussed with regard to measures of neighborhood deprivation and indicators of social support, proxy measures may cover multiple indicators. Finally, as described in Chapter 2, the committee expects the relationships between social risk factors and health and health care outcomes to change over time. Thus, it will be important to continuously evaluate the individual risk factors, indicators, and measures as well as the overall set of measures over time. These are empirical issues to test and apply when using real data.

REFERENCES

Acevedo-Garcia, D., and L. M. Bates. 2008. Latino health paradoxes: Empirical evidence, explanations, future research, and implications. In Latinas/os in the United States: Changing the face of America. New York, NY: Springer. Pp. 101-113.

Adler, N. E., and K. Newman. 2002. Socioeconomic disparities in health: Pathways and policies. Health Affairs (Millwood) 21(2):60-76.

Adler, N. E., T. Boyce, M. A. Chesney, S. Cohen, S. Folkman, R. L. Kahn, and S. L. Syme. 1994. Socioeconomic status and health: The challenge of the gradient. American Psychologist 49(1):15-24.

AHRQ (Agency for Healthcare Research and Quality). 2016. 2015 national healthcare quality and disparities report and 5th anniversary update on the national quality strategy. http://www.ahrq.gov/sites/default/files/wysiwyg/research/findings/nhqrdr/nhqdr15/2015nhqdr.pdf (accessed June 8, 2016).

Allin, S., C. Masseria, and E. Mossialos. 2009. Measuring socioeconomic differences in use of health care services by wealth versus by income. American Journal of Public Health 99(10):1849-1855.

Aughinbaugh, A., O. Robles, and H. Sun. 2013. Marriage and divorce: Patterns by gender, race, and educational attainment. Monthly Labor Review 136:1.

Berkman, L., and T. Glass. 2000. Social integration, social networks, social support, and health. In Social epidemiology. New York: Oxford University Press.

Berkman, N. D. 2011. Health literacy interventions and outcomes: An updated systematic review. Vol. 199. Rockville, MD: Agency for Healthcare Research and Quality.

Blustein, J., W. B. Borden, and M. Valentine. 2010. Hospital performance, the local economy, and the local workforce: Findings from a us national longitudinal study. PLoS Medicine 7(6):e1000297.

Brach, C., D. Keller, L. M. Hernandez, C. Baur, B. Dreyer, P. Schyve, A. J. Lemerise, and D. Schillinger. 2012. Ten attributes of health literate health care organizations. Washington, DC: Institute of Medicine.

Braveman, P. A., C. Cubbin, S. Egerter, S. Chideya, K. S. Marchi, M. Metzler, and S. Posner. 2005. Socioeconomic status in health research: One size does not fit all. JAMA 294(22):2879-2888.

Braveman, P. A., C. Cubbin, S. Egerter, D. R. Williams, and E. Pamuk. 2010. Socioeconomic disparities in health in the United States: What the patterns tell us. American Journal of Public Health 100(S1):S186-S196.

Brummett, B. H., J. C. Barefoot, I. C. Siegler, N. E. Clapp-Channing, B. L. Lytle, H. B. Bosworth, R. B. Williams, Jr., and D. B. Mark. 2001. Characteristics of socially isolated patients with coronary artery disease who are at elevated risk for mortality. Psychosomatic Medicine 63(2):267-272.

Buck, D. S., C. A. Brown, K. Mortensen, J. W. Riggs, and L. Franzini. 2012. Comparing homeless and domiciled patients’ utilization of the Harris County, Texas public hospital system. Journal of Health Care for the Poor and Underserved 23(4):1660-1670.

CDC (Centers for Disease Control and Prevention). 2010. National Health Interview Survey annotated bibliography. http://www.cdc.gov/nchs/nhis/rhoi/rhoi_bibliography.htm (accessed April 26, 2016).

CDC. 2013. Behavioral Risk Factor Surveillance System questionnaire. http://www.cdc.gov/brfss/questionnaires/pdf-ques/2013%20BRFSS_English.pdf (accessed May 18, 2016).

CMS (Centers for Medicare & Medicaid Services). 2006. Medicare current beneficiary survey. https://www.cms.gov/Research-Statistics-Data-and-Systems/Research/MCBS/Downloads/OLD/2006CBQah.pdf (accessed June 6, 2016).

CMS. 2016. Racial and ethnic disparities in health care in medicare advantage. https://www.cms.gov/About-CMS/Agency-Information/OMH/Downloads/National-Level-Results.pdf (accessed June 8, 2016).

CMS. n.d.-a. Eligibility. https://www.medicaid.gov/medicaid-chip-program-information/bytopics/eligibility/eligibility.html (accessed May 18, 2016).

CMS. n.d.-b. Get help paying costs: Medicaid. https://www.medicare.gov/your-medicarecosts/help-paying-costs/medicaid/medicaid.html (accessed May 18, 2016).

Cohen, S. 2004. Social relationships and health. American Psychologist 59(8):676-684.

Copen, C., A. Chandra, and I. Febo-Vazquez. 2016. Sexual behavior, sexual attraction, and sexual orientation among adults aged 18-44 in the United States: Data from the 2011-2013 National Survey of Family Growth. National Health Statistics Reports (88):1-14.

Cornwell, E. Y., and L. J. Waite. 2009. Social disconnectedness, perceived isolation, and health among older adults. Journal of Health and Social Behavior 50(1):31-48.

Cubbin, C., C. Pollack, B. Flaherty, M. Hayward, A. Sania, D. Vallone, and P. Braveman. 2011. Assessing alternative measures of wealth in health research. American Journal of Public Health 101(5):939-947.

Cutler, D. M., and A. Lleras-Muney. 2006. Education and health: Evaluating theories and evidence. Cambridge, MA: National Bureau of Economic Research.

Deaton, A. 2002. Policy implications of the gradient of health and wealth. Health Affairs 21(2):13-30.

Diez Roux, A. V., and C. Mair. 2010. Neighborhoods and health. Annals of the New York Academy of Sciences 1186:125-145.

Doran, K. M., K. T. Ragins, A. L. Iacomacci, A. Cunningham, K. J. Jubanyik, and G. Y. Jenq. 2013. The revolving hospital door: Hospital readmissions among patients who are homeless. Medical Care 51(9):767-773.

Eckerstorfer, P., J. Halak, J. Kapeller, B. Schütz, F. Springholz, and R. Wildauer. 2015. Correcting for the missing rich: An application to wealth survey data. Review of Income and Wealth.

Eggleston, J. S., and M. A. Klee. 2015. Reassessing wealth data quality in the survey of income and program participation. http://www.census.gov/library/workingpapers/2016/demo/SEHSD-WP2016-17.html (accessed May 18, 2016).

Elliott, M. N., D. E. Kanouse, Q. Burkhart, G. A. Abel, G. Lyratzopoulos, M. K. Beckett, M. A. Schuster, and M. Roland. 2015. Sexual minorities in England have poorer health and worse health care experiences: A national survey. Journal of General Internal Medicine 30(1):9-16.

Eng, P. M., E. B. Rimm, G. Fitzmaurice, and I. Kawachi. 2002. Social ties and change in social ties in relation to subsequent total and cause-specific mortality and coronary heart disease incidence in men. American Journal of Epidemiology 155(8):700-709.

Frederick, T. J., M. Chwalek, J. Hughes, J. Karabanow, and S. Kidd. 2014. How stable is stable? Defining and measuring housing stability. Journal of Community Psychology 42(8):964-979.

Goodell, S. G., and J. J. Escarce. 2007. Racial and ethnic disparities in access to and quality of health care.

http://www.health.state.mn.us/divs/hpsc/hep/transform/novdocuments/rwjsynthesis.pdf (accessed June 6, 2016).

Hart, L. G., E. H. Larson, and D. M. Lishner. 2005. Rural definitions for health policy and research. American Journal of Public Health 95(7):1149-1155.

Hawton, A., C. Green, A. P. Dickens, S. H. Richards, R. S. Taylor, R. Edwards, C. J. Greaves, and J. L. Campbell. 2011. The impact of social isolation on the health status and health-related quality of life of older people. Quality of Life Research 20(1):57-67.

HHS (Department of Health and Human Services). 2014. Improving cultural competence: A treatment improvement protocol. Rockville, MD: U.S. Substance Abuse and Mental Health Services Administration.

HHS. 2015a. 2015 edition health information technology (health IT) certification criteria, 2015 edition base electronic health record (EHR) definition, and ONC health IT certification program modifications. Federal Register. (80): 62601 -62759.

HHS. 2015b. Developing better information. http://www.hhs.gov/programs/topicsites/lgbt/better-information/index.html (accessed April 21, 2016).

HHS. 2015c. A profile of older Americans: 2015. http://www.aoa.acl.gov/aging_statistics/profile/2015/docs/2015-Profile.pdf (accessed June 6, 2016).

Holt-Lunstad, J., T. B. Smith, and J. B. Layton. 2010. Social relationships and mortality risk: A meta-analytic review. PLoS Medicine 7(7):e1000316.

House, J. S., K. R. Landis, and D. Umberson. 1988. Social relationships and health. Science 241(4865):540-545.

Houser, A., M. J. Gibson, and D. L. Redfoot. 2010. Trends in family caregiving and paid home care for older people with disabilities in the community: Data from the national long-term care survey. AARP Public Policy Institute. http://nasuad.org/sites/nasuad/files/hcbs/files/196/9758/caregiving2010.pdf (accessed May 18, 2016).

HUD (Department of Housing and Urban Development). 1998. Resident assessment of housing quality: Lessons from pilot surveys. https://www.huduser.gov/portal/Publications/pdf/quality.pdf (accessed May 18, 2016).

HUD. 2012. Using data to understand and end homelessness. https://www.huduser.gov/portal/periodicals/em/summer12/highlight2.html (accessed May 10, 2016).

IOM (Institute of Medicine). 2003a. Unequal treatment: Confronting racial and ethnic disparities in health care. Washington, DC: The National Academies Press.

IOM. 2003b. The future of the public’s health in the 21st century. Washington, DC: The National Academies Press.

IOM. 2004. Health literacy: A prescription to end confusion. Washington, DC: The National Academies Press

IOM. 2009a. America’s uninsured crisis: Consequences for health and health care. Washington, DC: The National Academies Press.

IOM. 2009b. Measures of health literacy: Workshop summary. Washington, DC: The National Academies Press.

IOM. 2009c. Race, ethnicity, and language data: Standardization for health care quality improvement. Washington, DC: The National Academies Press.

IOM. 2011. The health of lesbian, gay, bisexual, and transgender people: Building a foundation for better understanding. Washington, DC: The National Academies Press

IOM. 2012. How can health care organizations become more health literate?: Workshop summary. Washington, DC: The National Academies Press.

IOM. 2013. Variation in health care spending: Target decision making, not geography. Washington, DC: The National Academies Press.

IOM. 2014. Capturing social and behavioral domains and measures in electronic health records: Phase 2. Washington, DC: The National Academies Press

Jerant, A., R. Arellanes, and P. Franks. 2008. Health status among us hispanics: Ethnic variation, nativity, and language moderation. Medical Care 46(7):709-717.

Kasl, S. V., and B. A. Jones. 2000. The impact of job loss and retirement on health. Social Epidemiology 118-136.

Kochhar, R., and R. Fry. 2014. Wealth inequality has widened along racial, ethnic lines since end of great recession. Pew Research Center 12.

Krieger, N. 2000. Refiguring “race”: Epidemiology, racialized biology, and biological expressions of race relations. International Journal of Health Services 30(1):211-216.

Krieger, N. 2003. Genders, sexes, and health: What are the connections—and why does it matter? International Journal of Epidemiology 32(4):652-657.

Krieger, N., D. R. Williams, and N. E. Moss. 1997. Measuring social class in U.S. public health research: Concepts, methodologies, and guidelines. Annual Review of Public Health 18:341-378.

Krieger, N., J. T. Chen, P. D. Waterman, M.-J. Soobader, S. Subramanian, and R. Carson. 2002. Geocoding and monitoring of U.S. socioeconomic inequalities in mortality and cancer incidence: Does the choice of area-based measure and geographic level matter? The public health disparities geocoding project. American Journal of Epidemiology 156(5):471-482.

Kripalani, S., K. Wallston, K. Cavanaugh, C. Osborn, S. Mulvaney, A. McDougald Scott, and R. Rothmann. 2014. Measures to assess a health-literate organization. Vanderbilt Center for Effective Health Communication, http://www.nationalacademies.org/hmd/Activities/PublicHealth/HealthLiteracy/~/media/Files/Activity%20Files/PublicHealth/HealthLiteracy/Commissioned-Papers/Measures_to_Assess_HLO.pdf (accessed June 27, 2016).

LaVeist, T. A. 2005. Disentangling race and socioeconomic status: A key to understanding health inequalities. Journal of Urban Health 82(2 Suppl 3):iii26-iii34.

Lee, J., and H. Kim. 2008. A longitudinal analysis of the impact of health shocks on the wealth of elders. Journal of Population Economics 21(1):217-230.

Liu, H., and D. J. Umberson. 2008. The times they are a changin’: Marital status and health differentials from 1972 to 2003. Journal of Health and Social Behavior 49(3):239-253.

Liu, H., C. Reczek, and D. Brown. 2013. Same-sex cohabitors and health the role of race-ethnicity, gender, and socioeconomic status. Journal of Health and Social Behavior 54(1):25-45.

Lynch, S. M. 2003. Cohort and life-course patterns in the relationship between education and health: A hierarchical approach. Demography 40(2):309-331.

Moore, J. C., and E. J. Welniak. 2000. Income measurement error in surveys: A review. Journal of Official Statistics 16(4):331.

NASEM (The National Academies of Sciences, Engineering, and Medicine). 2016a. Accounting for social risk factors in medicare payment: Identifying social risk factors. Washington, DC: The National Academies Press.

NASEM. 2016b. Systems practices for the care of socially at-risk populations. Washington, DC: The National Academies Press.

Newhouse, J. P., M. Price, J. Huang, J. M. McWilliams, and J. Hsu. 2012. Steps to reduce favorable risk selection in medicare advantage largely succeeded, boding well for health insurance exchanges. Health Affairs 31(12):2618-2628.

OMB (Office of Management and Budget). 1995. Standards for the classification of federal data on race and ethnicity. https://www.whitehouse.gov/omb/fedreg_race-ethnicity (accessed April 21, 2016).

Phelan, J. C., and B. G. Link. 2015. Is racism a fundamental cause of inequalities in health? Annual Review of Sociology 41:311-330.

Pleasant, A., R. E. Rudd, C. O’Leary, M. K. Paasche-Orlow, M. P. Allen, W. Alvarado-Little, L. Myers, K. Parson, and S. Rosen. 2016. Considerations for a new definition of health literacy. http://nam.edu/wp-content/uploads/2016/04/Considerations-for-a-New-Definition-of-Health-Literacy.pdf (accessed May 18, 2016).

Roblin, D. W. 2013. Validation of a neighborhood SES index in a managed care organization. Medical Care 51(1):e1-e8.

Seeman, T. E., T. M. Lusignolo, M. Albert, and L. Berkman. 2001. Social relationships, social support, and patterns of cognitive aging in healthy, high-functioning older adults: Macarthur studies of successful aging. Health Psychology 20(4):243-255.

Shapiro, T., T. Meschede, and S. Osoro. 2013. The roots of the widening racial wealth gap: Explaining the black-white economic divide. Institute on Assets and Social Policy. http://iasp.brandeis.edu/pdfs/Author/shapiro-thomas-m/racialwealthgapbrief.pdf (accessed June 24, 2016).

SSA (Social Security Administration). n.d. Disability planner: How we decide if you are disabled. https://www.ssa.gov/planners/disability/dqualify5.html (accessed May 18, 2016).

Tamborini, C. R. 2007. Never-married in old age: Projections and concerns for the near future, the. Social Security Bulletin 67:25.

Theorell, T. 2000. Working conditions and health. In Social epidemiology, edited by L. F. Berkman and I. Kawachi. New York: Oxford University Press.

Tomaka, J., S. Thompson, and R. Palacios. 2006. The relation of social isolation, loneliness, and social support to disease outcomes among the elderly. Journal of Aging and Health 18(3):359-384.

Uchino, B. N. 2006. Social support and health: A review of physiological processes potentially underlying links to disease outcomes. Journal of Behavioral Medicine 29(4):377-387.

Umberson, D., and J. K. Montez. 2010. Social relationships and health a flashpoint for health policy. Journal of Health and Social Behavior 51(1 Suppl):S54-S66.

Wade, P. 1997. Race and ethnicity in latin america. Sterling, VA: Pluto Press.

Wang, W., and K. C. Parker. 2014. Record share of americans have never married: As values, economics and gender patterns change. Washington, D.C.: Pew Research Center, Social & Demographic Trends Project.

Ward, B. W., J. M. Dahlhamer, A. M. Galinsky, and S. S. Joestl. 2014. Sexual orientation and health among U.S. adults: National Health Interview Survey, 2013. National Health Statistics Reports (77):1-10.

Williams, D. R. 1997. Race and health: Basic questions, emerging directions. Annals of Epidemiology 7(5):322-333.

Williams, D. R. 1999. Race, socioeconomic status, and health: The added effects of racism and discrimination. Annals of the New York Academy of Sciences 896:173-188.

Williams, D. R., S. A. Mohammed, J. Leavell, and C. Collins. 2010. Race, socioeconomic status, and health: Complexities, ongoing challenges, and research opportunities. Annals of the New York Academy of Sciences 1186:69-101.

Wilson, R. S., K. R. Krueger, S. E. Arnold, J. A. Schneider, J. F. Kelly, L. L. Barnes, Y. Tang, and D. A. Bennett. 2007. Loneliness and risk of alzheimer disease. Archives of General Psychiatry 64(2):234-240.