a publication of the behavioral science & policy association

volume 4 issue 1 2018

featured topic community health & health disparities

behavioralpolicy.org ` founding co-editors disciplinary editors Craig R. Fox (UCLA) Behavioral Economics Sim B Sitkin (Duke University) Senior Disciplinary Editor Dean S. Karlan (Northwestern University) senior policy editor Associate Disciplinary Editors Oren Bar-Gill (Harvard University) Colin F. Camerer (California Institute ofTechnology) Carol L. Graham (Brookings Institution) M. Keith Chen (UCLA) bspa executive director Julian Jamison (World Bank) Kate B.B. Wessels Russell B. Korobkin (UCLA) advisory board Devin G. Pope (University of Chicago) Jonathan Zinman (Dartmouth College) Paul Brest (Stanford University) Cognitive & Brain Science David Brooks (New York Times) Senior Disciplinary Editor Henry L. Roediger III (Washington University) John Seely Brown (Deloitte) Associate Disciplinary Editors Yadin Dudai (Weizmann Institute & NYU) Robert B. Cialdini (Arizona State University) Roberta L. Klatzky (Carnegie Mellon University) Adam M. Grant (University of Pennsylvania) Hal Pashler (UC San Diego) Daniel Kahneman (Princeton University) Steven E. Petersen (Washington University) James G. March (Stanford University) Jeremy M. Wolfe (Harvard University) Jeffrey Pfeffer (Stanford University) Decision, Marketing, & Management Sciences Denise M. Rousseau (Carnegie Mellon University) Senior Disciplinary Editor Eric J. Johnson (Columbia University) Paul Slovic (University of Oregon) Associate Disciplinary Editors Linda C. Babcock (Carnegie Mellon University) Cass R. Sunstein (Harvard University) Max H. Bazerman (Harvard University) Richard H. Thaler (University of Chicago) Baruch Fischhoff (Carnegie Mellon University) executive committee John G. Lynch (University of Colorado) Morela Hernandez (University of Virginia) Ellen Peters (Ohio State University) Katherine L. Milkman (University of Pennsylvania) John D. Sterman (MIT) Daniel Oppenheimer (Carnegie Mellon University) George Wu (University of Chicago) Todd Rogers (Harvard University) Organizational Science David Schkade (UC San Diego) Senior Disciplinary Editors Carrie R. Leana (University of Pittsburgh) Joe Simmons (University of Pennsylvania) Jone L. Pearce (UC Irvine) bspa team Associate Disciplinary Editors Stephen R. Barley (Stanford University) Rebecca M. Henderson (Harvard University) Kaye N. de Kruif, Managing Editor (Duke University) Thomas A. Kochan (MIT) Carsten Erner, Statistical Consultant (FS Card) Ellen E. Kossek (Purdue University) Lea Lupkin, Media Manager Elizabeth W. Morrison (NYU) A. David Nussbaum, Director of Communications (Chicago) William Ocasio (Northwestern University) Daniel J. Walters, Financial Consultant (UCLA) Sara L. Rynes-Weller (University of Iowa) Ricki Rusting, Editorial Director Andrew H. Van de Ven (University of Minnesota) consulting editors Social Psychology Dan Ariely (Duke University) Senior Disciplinary Editor Nicholas Epley (University of Chicago) Shlomo Benartzi (UCLA) Associate Disciplinary Editors Dolores Albarracín (University of Illinois) Laura L. Carstensen (Stanford University) Susan M. Andersen (NYU) Susan T. Fiske (Princeton University) Thomas N. Bradbury (UCLA) Chip Heath (Stanford University) John F. Dovidio (Yale University) David I. Laibson (Harvard University) David A. Dunning (Cornell University) George Loewenstein (Carnegie Mellon University) E. Tory Higgins (Columbia University) Richard E. Nisbett (University of Michigan) John M. Levine (University of Pittsburgh) M. Scott Poole (University of Illinois) Harry T. Reis (University of Rochester) Eldar Shafir (Princeton University) Tom R. Tyler (Yale University) Sociology policy editors Senior Disciplinary Editors Peter S. Bearman (Columbia University) Henry J. Aaron (Brookings Institution) Karen S. Cook (Stanford University) Matthew D. Adler (Duke University) Associate Disciplinary Editors Paula England (NYU) Peter Cappelli (University of Pennsylvania) Peter Hedstrom (Oxford University) Thomas D’Aunno (NYU) Arne L. Kalleberg (University of North Carolina) J.R. DeShazo (UCLA) James Moody (Duke University) Brian Gill (Mathematica) Robert J. Sampson (Harvard University) Michal Grinstein-Weiss (Washington University) Bruce Western (Harvard University) Ross A. Hammond (Brookings Institution) Ron Haskins (Brookings Institution) Arie Kapteyn (University of Southern California) John R. Kimberly (University of Pennsylvania) Mark Lubell (UC Davis) Annamaria Lusardi (George Washington University) Timothy H. Profeta (Duke University) Donald A. Redelmeier (University of Toronto) Rick K. Wilson (Rice University) Kathryn Zeiler (Boston University)

Copyright © 2017 Behavioral Science & Policy is Authorization to photocopy Copyright Clearance Center, general distribution, or for Behavioral Science & Policy a publication of the Behavioral items for internal or personal 222 Rosewood Drive, creating new collective works, Association Science & Policy Association, use or the internal or personal Danvers, MA 01923. For more or for sale. Specific written Brookings Institution P.O. Box 51336, Durham, NC use of specific clients is granted information, please contact, permission for other copying 27717-1336, and is published by the Brookings Institution CCC, at +1 978-750-8400 and must be obtained from the ISSN 2379-4607 (print) twice yearly with the Brookings for libraries and other users online at www.copyright.com Permissions Department, ISSN 2379-4615 (online) Institution, 1775 Massachusetts registered with the Copyright Brookings Institution Press, ISBN 978-0-8157-3249-5 (pbk) Avenue, NW, Washington Clearance Center Transactional This Authorization does not 1775 Massachusetts Avenue, NW ISBN 978-0-8157-3250-1 (epub) DC 20036, and through the Reporting Service, provided that extend to other kinds of Washington, DC, 20036; e-mail: Brookings Institution Press. the basic fee is paid to the copying, such as copying for [email protected]. Behavioral Science & Policy Volume 4 Issue 1 2018

ii Editors’ note Steven Patierno & Sim B Sitkin

Features 1 17 Review Proposal What is health equity? Applying population health science principles Paula Braveman, Elaine Arkin, Tracy Orleans, to guide behavioral health policy setting Dwayne Proctor, Julia Acker, & Alonzo Plough Catherine Ettman, Salma M. Abdalla, & Sandro Galea

27 39 Essay Report The ubiquity of data & communication: Using pay-for-success financing for supportive A double-edged sword for disparities housing interventions: Promise & challenges Robert M. Califf Paula M. Lantz & Samantha Iovan

51 Essay Improving the match between patients’ needs & end-of-life care by increasing patient choice in Medicare Donald H. Taylor, Jr.

62 Editorial policy

Support for this special issue of BSP was provided by the Robert Wood Johnson Foundation. The views expressed here do not necessarily reflect the views of the Foundation. editors’ note

elcome to this special edition of is needed is a shared, unambiguous definition Behavioral Science & Policy, dedicated of health equity that can withstand social and Wto the complex issues surrounding political forces that seek to bend the definition health equity. This Spotlight Topic Forum was to promote particular policies and practices. co-edited by Steven Patierno (Duke University), They propose a definition of health equity that is Ingrid Gould Ellen (New York University), aimed at ensuring fair and just practices across and Todd T. Rogers (Harvard University) and all stakeholder sectors, is actionable, can be draws on research presented at the daylong operationalized and measured, and accounts workshop Achieving Health Equity: The Impact for social concerns. Achieving health equity of Housing, Employment, and Education on will require reducing health disparities, both by Health Disparities. This event was hosted by the improving the health of socially disadvantaged Behavioral Science & Policy Association, Duke groups and by addressing social determinants University, and Duke Health. of health disparities, including poverty and discrimination. The symposium drew attention to the interconnectedness of social, structural, and In the second article, Catherine Ettman, Salma biological determinants of health and provided M. Abdalla, and Sandro Galea propose a policy-­ elegant examples of the ways that interwoven impacting framework that allows for the assess- socioeconomic and geospatial factors drive ment of a broad range of global, national, health inequity. structural, and environmental health determi- nants and how these affect individual behaviors. As early as 1989, Dr. Samuel Broder, former They identify four principles that can serve as director of the National Cancer Institute, guides in the development of more effective acknowledged that “poverty is a carcinogen.”1 health policies: (1) recognize that population In 2003, the Institute of Medicine published health is not binary (sick versus not sick) but a Unequal Treatment: Confronting Racial and continuum of symptoms from mild to severe; Ethnic Disparities in Healthcare, in which it stated, (2) focus on high-prevalence determinants “A large body of published research reveals that that affect the most people rather than high- racial and ethnic minorities experience a lower risk, low-prevalence behaviors of fewer indi- quality of health services, and are less likely to viduals; (3) consider the trade-offs between receive even routine medical procedures than health interventions that may be easy to carry are white Americans.”2 More than a decade later, out (but can unintentionally exacerbate dispar- policymakers continue to grapple with how to ities) and interventions that are more challeng- achieve health equity, but they now recognize ing to implement but may be more effective that health inequity is driven by factors as diverse at mitigating disparities and have broader as housing, education, and employment and that impacts; and (4) carry out quantitative return- achieving health equity will require multilevel on-­investment analyses. In one example, they interventions that address social and structural provide interesting insights into how the mental inequities in these domains. health of a large population, as a reflection of drug abuse rates, could be improved by reduc- This issue begins with an article in which Paula ing population-wide­ stressors that trigger Braveman, Elaine Arkin, Tracy Orleans, Dwayne depression, such as food and housing instabil- Proctor, Julia Acker, and Alonzo Plough provide ity. In another example, they note that setting a carefully crafted definition of health equity. ­colorectal screening guidelines and encourag- This article, whose authors include scholars ing people to be screened through their doctors at the Robert Wood Johnson Foundation, is a increases screening rates overall but that detailed and thought-provoking follow-up to screening rates can be differentially and nega- Braveman’s 2014 commentary in Public Health tively affected by differences in race, education, Reports, “What Are Health Disparities and Health income, and access to a primary care provider. Equity? We Need to Be Clear.”3 In the article In contrast, when Delaware instituted an inter- here, Braveman and her coauthors explain that vention that made screenings readily accessible because a “lack of shared understanding can to the state’s whole population, the program be a serious obstacle to effective action,” what eliminated the health disparity between Blacks

ii behavioral science & policy | volume 4 issue 1 2018 eliminated the health disparity between Blacks low-income and vulnerable groups (such as the and Whites, reduced colorectal cancer mortal- homeless, addicted, or formerly incarcerated), ity in Blacks by 51%, and in the end proved to be as a method of providing a variety of promising cost-effective to the state. ways to enhance health equity and overall community health. They describe seven pay- The third article responds to the recognition for-success intervention models that have that health policymakers, institutional and the potential for significant impact and thus clinical decisionmakers, and patients today are merit more careful study. In the other article, drowning in data. This is not only because the Donald H. Taylor, Jr., tackles whether the United volume of health-related data is exponentially States could improve health equity by enabling expanding but also because our capacity to patients to choose to forgo low-value health effectively use the plethora of data has lagged. care in favor of high-value options that fit their Here, Robert M. Califf, former commissioner of needs better. Specifically, he makes the case the U.S. Food and Drug Administration and now for testing changes in Medicare that could give vice chancellor for health data science at Duke patients new evidence-based choices about University School of Medicine, considers the ways their treatment that would allow them to receive that the ubiquity of data and communication more impactful and equitable care. He also technology can be a double-edged sword provides provocative and intriguing ideas about for disparities. He notes that the computing the kinds of constructive and balanced choices power of a single smartphone exceeds the patients would make if given information and computing power of entire universities only options suited to optimizing individual and a few years ago and that it is now possible to community health. analyze health by collecting data on individuals and populations at the levels of neighborhoods We hope you find this collection of Spotlight and households based on how they interact articles valuable. As always, we look forward to with their digital devices. These data—biological, your feedback and suggestions for additional medical, social, and environmental—can either Spotlight Topic Forums for future issues of guide interventions to reduce disparities and Behavioral Science & Policy. help achieve health equity or be manipulated to exacerbate disparities. Califf identifies specific ways that addressing this challenge will require the collaborative and purposeful engagement references 1. American Cancer Society. (2011). Special section: of thinkers in medicine, law, technology, and Cancer disparities and premature deaths. ethics as well as of community members and Retrieved from https://www.cancer.org/content/ policymakers—all of whom must work together dam/cancer-org/research/cancer-facts-and- to use health data to achieve health equity. statistics/annual-cancer-facts-and-figures/2011/ special-section-cancer-disparities-and- Our special issue ends with two granular premature-deaths-cancer-facts-and- proposed interventions. First, Paula M. Lantz and figures-2011.pdf Samantha Iovan argue that safe and affordable 2. Institute of Medicine. (2003). Unequal treatment: housing is a critical social precondition for Confronting racial and ethnic disparities in health health and well-being and that a focus on care. https://doi.org/10.17226/10260 housing-related issues can pay health-related 3. Braveman, P. (2014). What are health disparities dividends far beyond the investment. They focus and health equity? We need to be clear. Public on pay for success, a public–private partnership Health Reports, 129(1, Suppl. 2), 5–8. https://doi. model, to finance housing interventions for org/10.1177/00333549141291S203references

Steven Patierno Sim B Sitkin Spotlight Editor Founding Editor

a publication of the behavioral science & policy association iii

review What is health equity? Paula Braveman, Elaine Arkin, Tracy Orleans, Dwayne Proctor, Julia Acker, & Alonzo Plough

abstract * Policymakers and others concerned about public health often speak of the need to achieve health equity. Yet the term can mean different things to different people. For government, other organizations, and communities, lack of shared understanding can be a serious obstacle to effective action. This lack of understanding makes it difficult to agree on concrete goals and criteria for success and can lead to wasted efforts, with policies and practices that work at cross-purposes. This article provides a carefully constructed definition of health equity and discusses the definition’s implications both for action and for assessing progress toward health equity.

Braveman, P., Arkin, E., Orleans, T., Proctor, D., Acker, J., & Plough, A. (2018). What is health equity? Behavioral Science & Policy, 4(1), 1–14.

a publication of the behavioral science & policy association 1 ver the past two decades, the term • be conceptually and technically sound and health equity has been used with consistent with current scientific knowledge; Oincreasing frequency in public health practice and research. But definitions for this • reflect the importance of fair and just prac- term vary widely. Some differ inconsequen- tices across all sectors, not only the health tially. Others, however, reflect deep divides in care sector, because health is a product of values and beliefs and can be used to justify and conditions and actions occurring in virtually promote very different policies and practices. all social domains; Clarity is particularly important when health equity is at stake because pursuing equity often • be actionable and sufficiently unambig- w involves a long uphill struggle against consid- uous to substantively guide decisions about erable resistance; in most cases, this struggle resource allocation priorities (some defini- Core Findings must strategically engage diverse stakeholders tions may be meaningful or even inspiring who have their own agendas. Under those to a segment of the public health commu- What is the issue? circumstances, lack of clarity about the desired nity with experience in thinking about and Different audiences tend goal can put efforts to achieve health equity at pursuing health equity, but not specific or to understand health risk of failure. concrete enough to guide action, especially equity differently. This for a wider audience); can frustrate attempts to achieve desired health In this article, based on a report published by outcomes. Public health the Robert Wood Johnson Foundation,1 we aim • be capable of being operationalized for the stakeholders need a purpose of measurement, which is crucial in common understanding to stimulate discussion and promote greater of health equity in order to consensus about the meaning of health equity assessing whether interventions are working; guide decision-making and and the implications this meaning has for action and resource allocation while and research. In recommending a definition of maintaining respect for social groups of concern. the term, we are not aiming to have everyone use • reflect respect for the social groups of exactly the same words to define health equity. concern. How can you act? Rather, our goal is to identify crucial elements Selected recommendations that can guide action in both public and private include: 1) Simultaneously spheres. (The Robert Wood Johnson Founda- The Definition emphasizing the benefit tion report, written by five of us—Braveman, Application of the criteria led to a two-part defi- of health equity measures Arkin, Orleans, Proctor, and Plough—includes nition. The first part is geared toward a broad, to society at large and not only targeted groups content not in this article, such as examples of nontechnical audience; the second is needed to 2) Constant monitoring health equity efforts and resources for under- guide measurement and monitoring of how well of overall levels of health taking health equity initiatives.) efforts to improve health equity are working: and health determinants within and across given populations Throughout this article, the term health refers to Health equity means that everyone has a health status or outcomes, distinct from health fair and just opportunity to be as healthy as Who should take care, which is only one of many important influ- possible. Achieving this requires removing the lead? ences on health. The term social encompasses obstacles to health—such as poverty and Researchers, policymakers, economic, psychosocial, and other societal discrimination and their consequences, and stakeholders in public health domains, although at times we refer separately which include powerlessness and lack of to social and economic domains for emphasis. access to good jobs with fair pay; quality The Appendix provides definitions of many education, housing, and health care; and terms that are used in this article and often arise safe environments. in discussions of health equity. For the purposes of measurement, health equity means reducing and ultimately Criteria for a Definition eliminating disparities in health and in The following criteria were key to developing the determinants of health that adversely the definition of health equity that we share in affect excluded or marginalized groups.2–5 this article. The definition had to:

2 behavioral science & policy | volume 4 issue 1 2018 Different Definitions for Different Audiences For many audiences or settings, the above definition will be too long or complex. The following are briefer and generally less complex alternatives, to be used with the understanding that they are backed up by the full definition:

An 8-second version for general audiences (defining health equity as a goal or outcome): Health equity means that everyone has a fair and just opportunity to be as healthy as possible.

Another 8-second version for general audiences (defining health equity as a process): Health equity means removing social and economic obstacles to health, such as poverty and discrimination.

A 15-second version for audiences concerned with measurement: Health equity means reducing and ultimately eliminating disparities in health and in the determinants of health that adversely affect excluded or marginalized groups.2–5

A 30-second definition for general audiences (consisting of the first part of the full definition above, minus the second part about measurement): Health equity means that everyone has a fair and just opportunity to be as healthy as possible. Achieving this requires removing obstacles to health such as poverty and discrimination and their consequences, which include powerlessness and lack of access to good jobs with fair pay; quality education, housing, and health care; and safe environments.

A 20-second definition to clarify the relationship between health equity and health disparities: Health equity is the ethical and human rights principle that motivates people to eliminate dispar- ities in health and in the determinants of health that adversely affect excluded or marginalized groups. Progress toward health equity is measured by reductions in health disparities.

Explaining the Definition therefore, can be measured by assessing the Both fairness and justice are invoked in this social determinants of health—such as income, definition to signify that achieving health equity wealth,14 education,15,16 neighborhood charac- in a population (for example, of a city, county, teristics,17,18 or social inclusion19—that people state, nation, or globally) involves not only experience across their lives. This concept meeting widely held standards of fairness, but acknowledges that individual responsibility is also addressing broader ethical concerns and important, while recognizing that too many adhering to human rights laws and principles. people lack access to the opportunities, Before people can achieve health equity, they conditions, and resources needed to make must first be able to fully realize their human healthy choices and live the healthiest possible rights in all domains essential for health, dignity, lives.7,8,11,12 Societal action is needed to address and participation in society. They must be able these obstacles. to freely exercise not only civil and political rights—such as freedom of speech, assembly, Health equity and health disparities are inti- and religion—but also social, economic, and mately related to each other. Health equity cultural rights, including rights to education, is the ethical and human rights principle that decent living conditions, and freedom from motivates people to eliminate health disparities, avoidable obstacles to good health.6 which are presumably avoidable differences in health or in its key determinants (such as good A large and growing literature demonstrates that jobs with fair pay; quality education, housing, opportunities to be healthy depend on living and health care; and safe environments) that and working conditions and other resources adversely affect marginalized or excluded that vary across social groups.7–13 The extent groups. Disparities in health and its key determi- of a population’s opportunities to be healthy, nants are the metrics used to assess the extent

a publication of the behavioral science & policy association 3 of extensive scientific research, it becomes “Lack of political will does not apparent that most disparities in health are tenaciously rooted in profound inequities in the justify considering a health opportunities and resources that are needed to be healthier. The literature reveals that social disparity to be unavoidable” inequities produce health inequities, which cannot be addressed effectively or in a lasting way without addressing their underlying causes. of health equity and how it changes over time for different groups of people. A large body of knowledge indicates that pursuing health equity requires addressing Being as healthy as possible refers to the highest equity not only in health care but also in a level of health that could be within an individual’s range of social determinants of health, partic- reach5,20,21 if society makes adequate efforts to ularly poverty,10–12,14,22,23 discrimination,11,19,24,25 provide opportunities to achieve it. This notion and their consequences, including power- acknowledges and takes into account the exis- lessness and lack of access to a range of tence of some unavoidable variations in genetic resources, services, and conditions needed for endowment that may limit an individual’s health optimal health. Achieving health equity calls for potential. Even if someone has serious unavoid- removing obstacles and improving access to able biological disadvantages, the best health the conditions and resources known to strongly possible for people with those biological disad- influence health, including good jobs with fair vantages could be achieved if societal efforts pay;26 high-quality education,15,16 housing,27 addressed that goal. For example, a person and health care; and health-promoting phys- with a disability that makes her unable to walk ical and social environments,17,28 particularly can achieve better health if has a properly for those who lack access to these conditions designed wheelchair and if access to fixtures and resources and who have worse health.29,30 at home, on buses, and at work enable her to Although this strategy should ultimately improve be more physically active, less isolated, and less health and well-being for everyone,31 the dependent on others. Adequate societal efforts systematic focus of action for equity should be often depend on political will. Lack of polit- on groups that have been excluded or marginal- ical will does not justify considering a health ized.30 The definition explicitly points to poverty disparity to be unavoidable. A health disparity and discrimination as underlying causes of should be considered avoidable if current scien- health inequity. We wrote it this way to make the tific knowledge indicates that it could potentially definition concrete and to reduce the ambiguity and plausibly be reduced or eliminated if polit- of more abstract and less specific definitions, ical will were present. which could be misused, perhaps unwittingly, to justify directing resources away from health This definition implies that advancing health equity. equity requires societal actions to increase opportunities to be as healthy as possible, Discrimination refers to adverse treatment of particularly for the groups that have suffered members of a social group based on prejudi- avoidable ill health and encountered the cial assumptions about the group as a whole. greatest social obstacles to achieving optimal Discrimination may be based on any number of health. Workers in the health sector and much characteristics, such as race, ethnic group, reli- of the public will be motivated to take action gion, national origin, disability status, skin color, for greater health equity by seeing evidence of gender or gender identity, or sexual orientation. significant health disparities—that is, presumably Discrimination or oppression is not neces- avoidable health differences on which excluded sarily conscious or intentional. Evidence has or marginalized groups fare worse than socially revealed that unconscious bias in interpersonal better-off groups. If one looks beneath the interactions is strong, widespread, and deeply surface, however, and examines the results rooted. Whatever the cause of the bias, it can

4 behavioral science & policy | volume 4 issue 1 2018 take a heavy toll on the health of its victims. Systemic discrimination has many other guises This conclusion is partly based on an under- as well. Voter registration requirements in some standing of the physiological mechanisms states, such as the need to show a birth certif- involved in responding to stress, particularly icate, may discriminate against immigrants chronic stress.24 and homeless persons, who are less likely to have the necessary documentation even when Discrimination does not occur only on the they meet federal voter qualifications. People interpersonal level, though. It is often systemic, of limited financial means, meanwhile, face that is, built into institutional structures, poli- discrimination in the judicial system. A nonvi- cies, and practices—consider policing, bail, and olent, first-time criminal offender may qualify sentencing practices that put people of color at for a diversion program, which would allow the a profoundly unfair disadvantage in the justice offender to avoid going to jail and to have the system; bank lending procedures that make it offense expunged from records, but only if the difficult or even impossible to build wealth in offender pays substantial fees. Thus, people Health Disparities low-income, largely minority communities; and with low incomes are far more likely to serve jail Avoidable differences the underfunding of schools in racially segre- time and have criminal records than are more in health or in its gated, poor communities, which denies children affluent people who have committed similar or key determinants from these neighborhoods a good education worse offenses.34 People of color are more likely that adversely affect marginalized or and hence a good, decently paying job. These than White people to be incarcerated for the excluded groups built-in features can have inequitable effects same offenses, and a history of incarceration regardless of whether any individual consciously is a formidable obstacle to future employment, intends to discriminate. This systemic form housing, and participation in society.35 of discrimination is also known as structural or institutional discrimination32 or systemic Powerlessness is both an objective and a oppression. subjective phenomenon. Poverty and discrimi- nation deprive people of economic and political Healthy As Possible Racial segregation in housing in the United power and make them less able to gain control Highest level of health States is an example of systemic discrimination of their lives and to access resources. Power- that could be within based on race or skin color. It is the product lessness becomes internalized when people an individual’s reach of deliberately discriminatory policies enacted perceive their inability to influence outcomes as in the past, including the Jim Crow laws that a personal failure rather than a result of discrim- enforced segregation of dark-skinned people in ination or systemic oppression.36 Repeated or the United States and practices affecting the sale persistent experiences of powerlessness may and rental of housing.33 Even though housing lead to feelings of hopelessness and, subse- discrimination is no longer legal, many people quently, immobilization and an inability to assert of color continue to be relegated into neighbor- one’s rights or needs. Structural/ hoods that pose multiple challenges to health Institutional by exposing residents to a range of physical Excluded or marginalized groups are made up Discrimination Systemic form of hazards (such as air pollution, other toxins, and of people who have often suffered discrimina- discrimination built into unsafe housing conditions) and social hazards tion or been pushed to society’s margins, with institutional structures, (such as concentrated poverty, absence of local little or no access to society’s health-promoting policies, and practices employment, inadequate transportation to resources and key opportunities.7,24 They suffer work and to better job prospects, poor schools, economic or social disadvantages or both,37 crime, an unhealthy food environment, hope- and they lack privilege. Examples of histori- lessness, and powerlessness). These places also cally disadvantaged groups who have been lack the assets required for optimal health, such excluded or marginalized include—but are not as good schools, optimism, clean air, green limited to—people of color;19 people living in spaces, traffic patterns that minimize pedestrian poverty, particularly across generations;22,38,39 danger, a feeling of safety, and the presence of religious minorities; people with physical or many role models who set positive norms for mental disabilities;40,41 LGBTQ persons;25,42 and healthy behaviors.19 women.43

a publication of the behavioral science & policy association 5 “A key feature of the definition of health equity is that it deliberately avoids the need to establish a causal role for any given factor in creating a health inequity”

A key feature of the definition of health equity Implications for Action is that it deliberately avoids the need to estab- The definition presented here deliberately lish a causal role for any given factor in creating restricts what can be called an effort for health a health inequity. According to the definition, equity. Many actions may be worthwhile public differences in health are inequitable if members health endeavors but not health equity efforts. of an excluded or marginalized group experi- For example, it could be important to address ence poor health that could plausibly have been a health problem that primarily affects a high-­ avoided, given political will. It is important not to income community; this, however, would not require proof of causation. The causes of some be a health equity endeavor, which prioritizes important health disparities—for example, racial actions disproportionately benefiting those disparities in premature birth—may be unknown who have been socially disadvantaged. Similarly, or contested, making some people reluctant to an initiative to improve nutrition for the entire call them inequities. These disparities should population of a state or nation might be worth- nevertheless be addressed in a health equity while but would not be a health equity effort agenda because they put people who are part unless it devoted considerably more resources of a socially disadvantaged group at further to improving nutrition among the disadvan- disadvantage with respect to their health, taged. Likewise, an initiative to expand green regardless of the causes. If the disparities are spaces and recreational areas in solidly middle- known to be rooted in social inequities in access class communities could be worthwhile from a to the opportunities and resources needed for public health perspective, but it, too, would not health, they can be referred to as health ineq- be a health equity initiative. Health equity should uities. If the causes are not known, we prefer to be one of the most central considerations emphasize the distinction by using a different driving policies that influence health, but not term: disparities or inequalities (a term generally the only principle; other key principles that must used outside the United States). Both disparity also be considered are effectiveness, efficiency, and inequality imply more than just a neutral overall population impact, and sustainability. difference, though: they suggest that there is something suspect about an observed differ- Policies, systems, and environmental improve- ence and that discrimination may be involved. ments can prevent and reduce health inequities, but, in most cases, only if they explicitly and This definition of health equity treats it as both energetically focus on health equity and are a process44 and an outcome, and it can be well designed and implemented; otherwise, measured as either. The process is removing even well-meaning interventions may inad- obstacles to health, particularly among those vertently widen health inequities. For example, who have been excluded and marginalized. in the early decades of anti-smoking efforts, It also can be thought of as the process of messages about the health dangers of tobacco reducing and ultimately eliminating disparities in use were disseminated across entire popula- health and health’s determinants that adversely tions. At some point, however, it became clear affect excluded or marginalized groups. Health that the messages were primarily reaching White equity also can be viewed as an outcome, people of higher education levels. Smoking was namely, the ultimate goal of achieving fair and declining among all groups, but the decline just opportunities to be healthy for everyone, was far slower among people of color and or the elimination of health and health-­ less educated people. The understanding determinant disparities that adversely affect emerged that different messages and different disadvantaged groups. methods for transmitting them were needed for

6 behavioral science & policy | volume 4 issue 1 2018 “Equity is not the same as equality. Those with the greatest needs and fewest resources require more, not equal, effort and resources to equalize opportunities.”

anti-smoking communications to be effective when conscious intent to discriminate is no among less privileged groups. longer present.

Achieving health equity requires societal action Ideally, a health equity effort would aim to to remove obstacles to health and increase improve the fundamental and structural causes opportunities for everyone to be healthier, while of ill health, notably poverty and discrimination, focusing particularly on those who have worse as opposed to addressing only the conse- health, face more social obstacles to health, quences of those causes. It may not always be and have fewer resources to improve their possible in the foreseeable future to alter the health. In line with basic ethical concerns (such underlying causes, however. In those circum- as for autonomy and respect for persons) and stances, it would be desirable, while alleviating human rights principles (such as participation in suffering by addressing the consequences of society and in making decisions that affect one’s the root problems, to also raise awareness well-being), advancing health equity requires (among the public, policymakers, and those engaging excluded or marginalized groups in most affected) of the need to address the root planning and implementing the actions needed causes, thus paving the way for more effec- to achieve greater health equity. Equity is not the tive action targeting the root causes in the same as equality. Those with the greatest needs future. For example, the problem of obesity is and fewest resources require more, not equal, an important health equity issue, with a dispro- effort and resources to equalize opportunities. portionate burden of obesity among people of lower income and education and among Although those who advocate for health equity people of color. A policymaker will probably will necessarily focus on the health needs of not want to wait until all the upstream determi- excluded or marginalized groups, they will nants of obesity and effective solutions for them garner support if they simultaneously call atten- are identified before putting in motion some tion to the ways that achieving greater health downstream efforts—such as making it easier equity will benefit all of society. For example, and more appealing for low-income people to greater health equity should result in a more engage in physical activity, increasing funding productive workforce and reduced spending for physical education at schools, requiring on medical care for preventable conditions. that the caloric content of all foods be clearly Furthermore, advancing health equity requires noted, or taxing sugary sodas—that could have achieving a more generally equitable society, at least some impact in the short or interme- and it has repeatedly been observed that overall diate term. But if the policymaker is aware of health is better in more equal societies.31 Some the more fundamental factors that are strongly scholars have hypothesized that this pattern suspected to be at the root of the problem— arises because more equal societies enjoy factors related to poverty and discrimination—a greater social cohesion and trust, which bene- more long-term and ultimately more effective fits everyone.31 strategy addressing poverty and discrimination and why they often, but do not always, inter- Achieving health equity requires more than iden- sect can be pursued at the same time, with the tifying and addressing overt discrim­ination. It also understanding that the results may not be seen requires addressing unconscious and implicit for quite a while. bias and the discriminatory effects—intended and unintended—of systemic structures and Many groups of people are socially disadvan- policies created by historical injustices, even taged. To be effective, an organization may

a publication of the behavioral science & policy association 7 choose to focus on one or a select few disad- or marginalization, or privilege versus lack of vantaged groups. The depth and extent of privilege); the information must identify which disadvantage faced by a group (such as multiple groups are most and least advantaged and versus single disadvantages),20,23,38,45 as well as define who should be compared. Because where maximal impact could be achieved, are health equity is concerned with fairness and legitimate considerations in choosing where to justice, gaps should be assessed using both focus.20,29,30 In addition, it should be noted that measures that are absolute (such as differ- some individuals in an excluded or marginal- ences between groups in the percentage of ized group may have escaped from some of the infants who survive until their first birthday) and disadvantages experienced by most members measures that are relative (such as infants in of that group; these exceptions do not negate Group X are twice as likely as infants in Group Y the fact that the group as a whole is disadvan- to die in their first year of life). The gaps between taged in ways that can be measured. the advantaged and disadvantaged are closed by making concerted efforts to improve the health of excluded or marginalized groups, not Implications for Accountability: by worsening the health of those who are better Measuring & Monitoring off.49 For example, the relative gap between Health Equity Black and White infants in the incidence of low As the definition of health equity implies, birth weight narrowed during the period 1990– measurement is not a luxury: it is crucial for 2010 in the United States; however, that trend documenting disparities and inequities and did not represent the achievement of greater for motivating and informing efforts to elim- health equity, because it instead reflected inate them. Without measurement, there is an increase in the incidence of preterm birth no accountability for the effects of policies or among Whites rather than real improvement in programs. that measure among Blacks.50

A commitment to health equity requires Disadvantaged groups should be compared constant monitoring of overall (average) levels of with those who are most advantaged, not health and health determinants in a population, with the whole population (or the popula- as well as routine comparisons of how more and tion average). Comparing the disadvantaged less advantaged groups within that population with the general population is not appropriate are faring on relevant measures of health and unless information on advantaged groups is health determinants. Overall levels of health are unavailable, for a simple reason: when disad- useful to know and are important, but they can vantaged groups represent a sizable portion of hide large disparities among subgroups within the population—as is increasingly occurring in a population. Measuring gaps in health and in the United States—this approach compares the opportunities for optimal health is important not disadvantaged groups largely with themselves, only to document progress, but also to motivate thereby substantially underestimating the size action and identify the kinds of actions needed of the gap between the disadvantaged and the to achieve greater equity. advantaged.

The definition of health equity calls for exam- Social advantage, privilege, inclusion, disad- ining how well socially disadvantaged (excluded vantage, discrimination, exclusion, and or marginalized) groups in a population fare marginalization can be measured in various on health and its determinants compared with ways, including by assessing indicators of advantaged or privileged groups.46–48 Making wealth (such as income or accumulated finan- this assessment requires having information cial assets),14,51,52 influence,7,36 and prestige or on both (a) important measures of health and social acceptance (for example, educational its determinants, including social determinants, attainment and representation in high executive, and (b) the distribution of social advantage political, and professional positions).53 They also and disadvantage (inclusion versus exclusion can be measured by well-documented historical

8 behavioral science & policy | volume 4 issue 1 2018 evidence of oppression or discrimination (such author affiliation as slavery; displacement from ancestral lands; lynching and other hate crimes; denial of voting, Braveman: University of California, San Fran- marriage, and other rights; and discriminatory cisco. Arkin: independent consultant. Orleans practices in housing, bank lending, and justice and Proctor: Robert Wood Johnson Foun- system). dation. Acker: University of California, San Francisco. Plough: Robert Wood Johnson Foundation. Corresponding author’s e-mail: Final Remarks [email protected] Health equity may seem to be a complex and elusive concept. The essence, however, consists of two basic elements: (a) reducing health author note disparities by improving the health of socially disadvantaged groups, and (b) addressing We thank the following individuals who provided the social determinants of health dispari- comments on drafts of the report on which this ties, including poverty and discrimination. It is article is based. The authors are solely respon- important to be clear about what health equity is sible for the contents of this article: and what it is not; for example, it is a core aspect of public health, but it is not the only aspect Deborah Austin, ReachUp Inc. that needs to be considered in public health Stephanie V. Boarden, PolicyLink actions. Clarity is important because efforts to Karen Bouye, CDC Office of Minority Health move toward health equity will inevitably face and Health Equity powerful challenges. If those of us who wish to Renee B. Canady, Michigan Public Health contribute to achieving greater health equity are Institute not clear about where we are headed and why, Ana Diez Roux, Drexel University we can be detoured from promising paths and Tyan P. Dominguez, University of Southern perhaps even lose our way. California Mary Haan, University of California, San Francisco Erin Hagan, Evidence for Action Robert Hahn, Centers for Disease Control and Prevention George Kaplan, University of Michigan (Emeritus) Mildred Thompson, PolicyLink Naima Wong Croal, National Collaborative for Health Equity

a publication of the behavioral science & policy association 9 Appendix. Definitions of require establishing that the disparities/inequal- terms used in the article ities were caused by social disadvantage; it requires only observing worse health in socially discrimination or economically disadvantaged groups. Health This is a broad term that includes but is not disparities/inequalities are ethically concerning limited to racism. (Bold type indicates words even if their causes are not clear, because they defined in this appendix.) Prejudicial treatment, affect groups already at underlying economic social exclusion, and marginalization have or social disadvantage (due to poverty, discrim- been based on a wide range of characteristics, ination, or both) and they indicate that these including not only racial or ethnic group but socially disadvantaged groups are further disad- also poverty, disability, religion, LGBTQ status, vantaged by having ill health on top of social gender, and other characteristics. disadvantage; this double whammy seems especially unfair because good health often is ethnicity or ethnic group needed to escape social disadvantage. These terms refer to belonging to a group of people who share a common culture (which It may seem reasonable to use the term dispar- may consist of beliefs, values, or practices, such ities or inequalities to refer to only descriptive as modes of dress, diet, or language) and usually or mathematical differences without implying a common ancestry in a particular region of any judgment about whether they suggest the world. Some people use the term ethnicity cause for moral or ethical concern. However, or ethnic group to encompass both racial social movements in the United States and and ethnic groups, based on the recognition other countries for nearly 30 years have treated that race is fundamentally a social rather than these terms as indicating differences that are biological construct. (See race or racial group worrisome from ethical and human rights below.) perspectives (although the groups of concern are not always the same). In the United States, health health disparities have often referred to racial or Throughout the article, health refers to health ethnic differences in health, whereas in Europe status, that is, to physical and mental well-being, and other regions, health inequalities have distinguished from health care, which is only generally referred to health differences among one of many important influences on health. people of different socioeconomic means. In theory, one might want to bring the definitions health disparity and health inequality into alignment to simplify discussions of how to These terms are synonyms; disparity is used achieve health equity. But legislation and poli- more often in the United States, whereas other cies have been written based on the existing countries use inequality. Progress toward health understandings of the terms, so redefinitions equity is measured by assessing health dispari- might have unintended consequences that ties/inequalities. The concept of health equity is could unwittingly threaten the achievements the underlying principle that motivates action to and momentum gained over decades. For eliminate health disparities. example, some have proposed using the term disparity only to mean a difference, without any The terms disparity and inequality do not neces- implication regarding whether the difference is sarily imply that social disadvantage is the cause morally suspect, and using the term inequity for of or a contributor to worse health, but they racial or socioeconomic differences in health. suggest that such a causal link should be consid- If that change were made, then the resources ered. For over 25 years in the fields of public now directed to national, state, and local efforts health and medicine, the terms health disparity to reduce health disparities could be used and health inequality have referred to plau- for virtually any health improvement effort, sibly avoidable, systematic health differences including efforts focused on privileged groups. adversely affecting socially or economically Furthermore, indiscriminately calling any racial disadvantaged groups. This definition does not or socioeconomic difference in health unfair

10 behavioral science & policy | volume 4 issue 1 2018 (inequitable) would be unwise from a commu- not to take concerted action to eliminate it; nications perspective, because there are some failure to act is unjust because the situation health differences whose etiology we do not puts an already socially disadvantaged group know; the term health disparity is convenient to at further disadvantage on health, and good use for these differences, signaling reason for health is often needed to escape social disad- concern but not necessarily proof of a health vantage. Where there is reasonable (but not inequity. necessarily definitive) evidence that underlying inequities in opportunities and resources to be Health disparity and health inequality are broad healthier have produced a health disparity, that terms that include health inequity and signify disparity can be called a health inequity; it needs more than just difference or variation: they to be addressed through efforts to eliminate signify a health difference that raises moral or inequities in the opportunities and resources ethical concerns. These terms are very useful required for good health. Inequity is a powerful in identifying problematic areas (that is, an word; its power may be diminished if it is used avoidable health difference that puts a socially carelessly, exposing health equity efforts to disadvantaged group at further disadvantage on potentially harmful challenges. It should be used health) and being measurable, but they do not thoughtfully. necessarily imply definitive knowledge of the causes. opportunity This means access to goods, services, and the health equity benefits of participating in society. Financial This phrase means that everyone has a fair and barriers and geographic distance are not the just opportunity to be as healthy as possible. only obstacles to access; others can include Achieving health equity requires removing past discrimination, fear, mistrust, and lack of obstacles to health such as poverty, discrimi- awareness, as well as transportation difficulties nation, and their consequences, which include and family caregiving responsibilities. Measuring powerlessness and lack of access to good jobs the real (or realized) access to opportunities with fair pay; quality education, housing, and that different social groups have requires not health care; and safe environments. just measuring their potential access54 but also assessing which groups actually have the rele- For the purposes of measurement, health equity vant goods, services, and benefits. Because means reducing and ultimately eliminating of past and ongoing racial discrimination in disparities in health and health determinants housing, lending, and hiring policies and prac- that adversely affect excluded or marginalized tices, there is great variation in the quality of the groups. places where people of different racial or ethnic groups live, work, learn, and play; these differ- Health equity is the ethical and human rights ences in quality often affect the opportunities principle motivating efforts to eliminate health groups have to be as healthy as possible. disparities; health disparities are the metric for assessing progress toward health equity. race or racial group This generally refers to a group of people who health inequity share a common ancestry from a particular A health inequity is a particular kind of health region of the globe. Common ancestry is often disparity, one that is a cause for concern in accompanied by superficial secondary physical that it is potentially a reflection of injustice. characteristics such as skin color, facial features, Views of what constitutes adequate evidence and hair texture. Given the extensive racial of a health inequity can differ. Some will argue mixing that has occurred historically, these that to call a disparity an inequity, one must superficial differences in physical appearance know its causes and demonstrate that they are are highly unlikely to be associated with funda- unjust. Others would maintain that regardless mental, widespread, underlying differences in of the causes of a health disparity, it is unjust biology. This low probability of an association

a publication of the behavioral science & policy association 11 does not rule out the possibility that some highly social specific genetic differences associated with Unless specified otherwise, this term encom- ancestry could affect susceptibility to particular passes (but is not limited to) economic, diseases (for example, sickle cell anemia, other psychosocial, and other societal domains. In hemoglobinopathies, or Tay-Sachs disease) this article, at times economic is specified in or responsiveness to treatments. These highly addition to social, for clarity. specific differences, however, are not funda- mental and do not define biologically distinct social determinants of health racial groups; they generally occur in multiple These are nonmedical factors that influ- racial groups at different frequencies. The ence health, such as employment, income, primary drivers of health inequities are differ- housing, transportation, child care, education, ences in social and economic opportunities to discrimination, and the quality of the places be healthier. Scientists, including geneticists, where people live, work, learn, and play. Social concur that race is primarily a social—not a refers broadly to society—that is, people, their biological—construct.55–57 actions, and relationships. Social determinants are social in the sense that they are shaped by racism social policies. The World Health Organiza- This term refers to prejudicial treatment based tion Commission on the Social Determinants on racial or ethnic group and the societal struc- of Health7 chose to include medical care (the tures or institutions that systemically perpetuate services provided by trained medical or health this unfair treatment. Racism can be expressed personnel, such as doctors, nurses, therapists, on interpersonal, systemic, and internalized pharmacists, and their support staff) among the levels.32 social determinants, presumably because the provision of medical care—including access to Interpersonal racism is race-based unfair it and its quality—is under the control of social treatment of a person or group by individuals. policy. Generally, however, and in this article, Examples include hate crimes; name-calling; the term social determinants refers to factors or the denial of a job, promotion, equal pay, or outside of medical care that influence health. access to renting or buying a home on the basis of race. social exclusion or marginalization This term refers to barring or deterring partic- Structural or institutional racism (also known ular social groups—for example, on the basis as systemic racism) is race-based unfair treat- of skin color, national origin, religion, wealth, ment built into policies, laws, and practices. It disability, sexual orientation, gender identity, often is rooted in intentional discrimination that or gender—from full participation in society occurred historically, but it can exert its effects and from sharing the benefit of participation. even when no individual currently intends to Socially excluded or marginalized groups have discriminate. Racial residential segregation is an less power and prestige and, generally, less excellent example: it has steered people of color wealth. Because they lack those basic resources, into residential areas where opportunities to be the places where they are able to live often are healthier and to escape poverty are limited. characterized by health-damaging conditions or conditions that fail to promote health, such Internalized racism occurs when victims of as pollution, lack of access to jobs and services, racism adopt (perhaps unconsciously) race- and inadequate schools. based prejudicial attitudes toward themselves and their racial or ethnic group, resulting in a structural racism loss of self-esteem and potentially in prejudi- See racism. cial treatment of members of their own racial or ethnic group.

12 behavioral science & policy | volume 4 issue 1 2018 references

1. Braveman, P., Arkin, E., Orleans, T., Vol. 1186. The biology of disadvantage: 25. Meyer, I. H., & Northridge, M. E. (Eds.). Proctor, D., & Plough, A. (2017). What Socioeconomic status and health (pp. (2007). The health of sexual minorities: is health equity? And what difference 1–4). New York, NY: New York Academy Public health perspectives on lesbian, does a definition make? Retrieved from of Sciences. gay, bisexual and transgender Robert Wood Johnson Foundation 13. Norman, D., Kennedy, B., & Kawachi, populations. New York, NY: Springer website: https://www.rwjf.org/content/ I. (1999). Why justice is good for our Nature. dam/farm/reports/issue_briefs/2017/ health: The social determinants of 26. Burgard, S. A., & Lin, K. Y. (2013). rwjf437393 health inequalities. Daedalus, 128(4), Bad jobs, bad health? How work and 2. United States Department of Health 215–251. working conditions contribute to and Human Services, Office of Disease 14. Isaacs, S. L., & Schroeder, S. A. (2004). health disparities. American Behavioral Prevention and Health Promotion. Class: The ignored determinant of Scientist, 57, 1105–1127. (2018). Disparities. Retrieved from http:// the nation’s health. The New England 27. Edmonds, A., Braveman, P., Arkin, E., www.healthypeople.gov/2020/about/ Journal of Medicine, 351, 1137–1142. & Jutte, D. (2015). Making the case for foundation-health-measures/Disparities 15. Cutler, D. M., & Lleras-Muney, A. (2006). linking community development and 3. Braveman, P. A., Kumanyika, S., Education and health: Evaluating health. Princeton, NJ: Robert Wood Fielding, J., LaVeist, T., Borrell, L. N., theories and evidence (NBER Working Johnson Foundation. Manderscheid, R., & Troutman, A. (2011). Paper No. 12352). Cambridge, MA: 28. Gordon-Larsen, P., Nelson, M. C., Page, Health disparities and health equity: National Bureau of Economic Research. P., & Popkin, B. M. (2006). Inequality The issue is justice. American Journal of in the built environment underlies key Public Health, 101(S1), S149–S155. 16. Egerter, S., Braveman, P., Sadegh- Nobari, T., Grossman-Kahn, R., & health disparities in physical activity and 4. United States Department of Health Dekker, M. (2011). Education matters obesity. Pediatrics, 117, 417–424. and Human Services, Office of Minority for health. Princeton, NJ: Robert Wood 29. Daniels, N., Kennedy, B. P., & Kawachi, Health, National Partnership for Action Johnson Foundation. I. (2000, February 1). Justice is good for to End Health Disparities. (2010). Toolkit 17. Roux, A. V. D., & Mair, C. (2010). our health. Boston Review. Retrieved for community action. Retrieved from from http://www.bostonreview.net/ http://www.minorityhealth.hhs.gov/npa Neighborhoods and health. In N. E. Adler & J. Stewert (Eds.), Annals of the forum/norman-daniels-bruce-kennedy- 5. Braveman, P. (2006). Health disparities New York Academy of Sciences: Vol. ichiro-kawachi-justice-good-our-health and health equity: Concepts and 1186. The biology of disadvantage: 30. Rawls, J. (1971). A theory of justice. measurement. Annual Review of Public Socioeconomic status and health (pp. Cambridge, MA: Belknap Press. Health, 27, 167–194. 125–145). New York, NY: New York 31. Pickett, K. E., & Wilkinson, R. G. (2015). 6. Braveman, P., & Gruskin, S. (2003). Academy of Sciences. Income inequality and health: A causal Poverty, equity, human rights and 18. Acevedo-Garcia, D., Osypuk, T. L., review. Social Science & Medicine, 128, health. Bulletin of the World Health McArdle, N., & Williams, D. R. (2008). 316–326. Organization, 81, 539–545. Toward a policy-relevant analysis of 32. Jones, C. P. (2000). Levels of racism: 7. World Health Organization Commission geographic and racial/ethnic disparities Theoretic framework and a gardener’s on the Social Determinants of Health. in child health. Health Affairs, 27, tale. American Journal of Public Health, (2008). Closing the gap in a generation: 321–333. 90, 1212–1215. Health equity through action on the 19. Williams, D. R., & Mohammed, S. A. 33. Rothstein, R. (2017). The color of social determinants of health. Geneva, (2013). Racism and health I: Pathways Switzerland: World Health Organization. law: A forgotten history of how our and scientific evidence. American government segregated America. New 8. Berkman, L. F., Kawachi, I., & Glymour, Behavioral Scientist, 57, 1152–1173. York, NY: Liveright. M. (2014). Social epidemiology (2nd ed.). 20. Whitehead, M. (1991). The concepts and New York, NY: Oxford University Press. 34. Looney, A., & Turner, N. (2018). principles of equity and health. Health Work and opportunity before and 9. Marmot, M. (2015). The health gap: The Promotion International, 6, 217–228. after incarceration. Washington, DC: challenge of an unequal world. The 21. Braveman, P., & Gruskin, S. (2003). Brookings Institution. Lancet, 386, 2442–2444. Defining equity in health. Journal of 35. Wildeman, C., & Wang, E. A. (2017). 10. Marmot, M., Friel, S., Bell, R., Houweling, Epidemiology & Community Health, 57, Mass incarceration, public health, and T. A., Taylor, S., & Commission on 254–258. widening inequality in the USA. The Social Determinants of Health. (2008). 22. Wagmiller, R. L., & Adelman, R. M. Lancet, 389, 1464–1474. Closing the gap in a generation: Health (2009). Childhood and intergenerational equity through action on the social 36. Wallerstein, N. (1992). Powerlessness, poverty: The long-term consequences empowerment, and health: Implications determinants of health. The Lancet, 372, of growing up poor. New York, NY: 1661–1669. for health promotion programs. National Center for Children in Poverty. American Journal of Health Promotion, 11. Braveman, P., Egerter, S., & Williams, D. 23. Evans, G. W. (2004). The environment 6, 197–205. R. (2011). The social determinants of of childhood poverty. American 37. United Nations General Assembly. health: Coming of age. Annual Review Psychologist, 59, 77–92. of Public Health, 32, 381–398. (1966). International covenant on 24. Williams, D. R., & Mohammed, S. economic, social and cultural rights. 12. Adler, N. E., & Stewart, J. (2010). A. (2009). Discrimination and racial Retrieved from http://www.ohchr.org/ Preface to the biology of disadvantage: disparities in health: Evidence and EN/ProfessionalInterest/Pages/CESCR. Socioeconomic status and health. In needed research. Journal of Behavioral aspx N. E. Adler & J. Stewart (Eds.), Annals Medicine, 32, 20–47. of the New York Academy of Sciences:

a publication of the behavioral science & policy association 13 38. Reeves, R., Rodrigue, E., & Kneebone, 48. Mackenbach, J. P., & Kunst, A. E. E. (2016). Five evils: Multidimensional (1997). Measuring the magnitude of poverty and race in America. socio-economic inequalities in health: Washington, DC: Brookings Institution. An overview of available measures 39. Cheng, T. L., Johnson, S. B., & illustrated with two examples from Goodman, E. (2016). Breaking the Europe. Social Science & Medicine, 44, intergenerational cycle of disadvantage: 757–771. The three generation approach. 49. Whitehead, M., & Dahlgren, G. (2006). Pediatrics, 137, 1–14. Levelling up (part 1): A discussion paper 40. United States Department of Health on concepts and principles for tackling and Human Services, Office of Disease social inequities in health. Copenhagen, Prevention and Health Promotion. Denmark: WHO Regional Office for (2018). Disability and health. Retrieved Europe. from http://www.healthypeople. 50. Martin, J. A., Hamilton, B. E., Ventura, gov/2020/topics-objectives/topic/ S. J., Osterman, M. J. K., Wilson, E. C., disability-and-health & Matthews, T. J. (2012). Births: Final 41. World Health Organization. (2018). data for 2010. National Vital Statistics Disabilities. Retrieved from http://www. Reports, 61(1). who.int/topics/disabilities/en/ 51. Pollack, C. E., Chideya, S., Cubbin, C., 42. Ward, B. W., Dahlhamer, J. M., Galinsky, Williams, B., Dekker, M., & Braveman, P. A. M., & Joestl, S. S. (2014). Sexual (2007). Should health studies measure wealth? A systematic review. American orientation and health among U.S. adults: National Health Interview Survey, Journal of Preventive Medicine, 33, 2013 (National Health Statistics Report 250–264. No. 77). Retrieved from Centers for 52. Yeung, W. J., Linver, M. R., & Brooks- Disease Control and Prevention: https:// Gunn, J. (2002). How money matters for www.cdc.gov/nchs/data/nhsr/nhsr077. young children’s development: Parental pdf investment and family processes. Child 43. Moss, N. E. (2002). Gender equity and Development, 73, 1861–1879. socioeconomic inequality: A framework 53. Keyes, C. L. M. (1998). Social well-being. for the patterning of women’s health. Social Psychology Quarterly, 61(2), Social Science & Medicine, 54, 121–140. 649–661. 54. Andersen, R., & Aday, L. A. (1978). 44. Bluvas, E. (2016). Camara Jones inspires Access to medical care in the US: at the 9th Annual Clyburn Lecture with Realized and potential. Medical Care, 16, discussion on health disparities and 533–546. racism. Retrieved from https://sc.edu/ 55. Yudell, M., Roberts, D., DeSalle, R., & study/colleges_schools/public_health/ Tishkoff, S. (2016, February 5). Taking about/news/2016/clyburn2016_recap. race out of human genetics. Science, php 351, 564–565. 45. Ferraro, K. F., & Kelley-Moore, J. A. 56. McCann-Mortimer, P., Augoustinos, M., (2003). Cumulative disadvantage and & LeCouteur, A. (2004). ‘Race’ and the health: Long-term consequences of Human Genome Project: Constructions obesity? American Sociological Review, of scientific legitimacy. Discourse & 68, 707–729. Society, 15, 409–432. 46. Harper, S., Lynch, J., Meersman, S. C., 57. Witherspoon, D. J., Wooding, S., Rogers, Breen, N., Davis, W. W., & Reichman, A. R., Marchani, E. E., Watkins, W. S., M. E. (2008). An overview of methods Batzer, M. A., & Jorde, L. B. (2007). for monitoring social disparities Genetic similarities within and between in cancer with an example using human populations. Genetics, 176, trends in lung cancer incidence by 351–359. area-socioeconomic position and race- ethnicity, 1992–2004. American Journal of Epidemiology, 167, 889–899. 47. Hosseinpoor, A. R., Bergen, N., Koller, T., Prasad, A., Schlotheuber, A., Valentine, N., . . . Vega, J. (2014). Equity-oriented monitoring in the context of universal health coverage. PLoS Medicine, 11, 1–9.

14 behavioral science & policy | volume 4 issue 1 2018

proposal Applying population health science principles to guide behavioral health policy setting Catherine Ettman, Salma M. Abdalla, & Sandro Galea

abstract † Many behaviors, such as smoking and overeating, strongly affect a population’s health. Further, social, physical, and economic contexts—for example, housing, transportation, education, and employment—shape health-related behaviors. To improve a population’s health, policies must include actions that alter elements of these larger contexts. But the elements are vast and complex, and resources are limited. How can policymakers determine the right priorities to focus on? Building on the emerging field of population health science, we suggest four principles to guide priority setting: view population health as a continuum, focus on affecting ubiquitous influences on health, consider the trade-offs between efficiency and equity, and evaluate return on investment. This proposal offers a novel approach to setting policy for improving health behaviors.

Ettman, C., Abdalla, S. M., & Galea, S. (2018). Applying population health science princi- ples to guide behavioral health policy setting. Behavioral Science & Policy, 4(1), 17–24.

a publication of the behavioral science & policy association 17 opulation health science researchers aim There is no easy way to determine what matters to understand the factors that affect the most. A recent book on population health Pdistribution of health-related features, science has, however, proposed a formal set such as cardiovascular disease, in a population of nine principles that can guide scholarship so that policymakers can intervene and improve in population health4 and the setting of policy. health on a societal scale.1 This endeavor Here we focus on four of these principles—the requires population health scholars to assess a ones readily translatable to policy—and their broad range of health determinants, including application. w global and national influences, urban structures and environments, individual behaviors, and the Core Findings mechanisms that explain how each of these Principle 1: View Population factors affects health.2 Health as a Continuum What is the issue? The first principle we explore holds that popu- Policymakers need Consider, for instance, how the principles of lation health is best viewed as a continuum. This to implement formal population health science could help guide notion nudges thinking away from conceptual- principles from population health science into policymakers deciding on the right interven- izing health as a binary (someone is sick or not decisionmaking. These tions for addressing the obesity epidemic. sick) and toward recognizing that a population emphasize a broader Obesity arises from molecular, individual, social includes people with symptoms ranging from understanding of health and equity while measuring network, and national causes. At the molec- mild to severe, with only the people toward the the appropriate return on ular level, genes shape people’s vulnerability to severe end of the range meeting the criteria for their investments. This will obesity to some extent. Individual motivation a diagnosis. If health is framed as a continuum, allow them to increase dictates individual approaches to weight control, behavioral health policies should focus on both the effectiveness and cost savings of public and friends in social networks affect individual improving health in as broad a swath of the health care interventions. decisions. National factors related to food avail- population as possible rather than focusing ability—such as food policy and accessibility primarily on finding and treating people with a How can you act? of safe areas for physical exercise—also deter- specific diagnosis. Selected recommendations include: mine whether people are likely to eat well and 1) Lowering obesity by exercise. Therefore, any intervention to reduce The common approach to cholesterol testing increasing retail access to obesity should rest on an understanding of the in the United States is an example of misplaced healthy food through tax breaks or subsidized loans causes of obesity; their prevalence, complexity, emphasis. If a screening shows a person has 2) Providing preventative and interactions; and how amendable any of high cholesterol, a health care provider is likely screening to populations these causes are to an intervention. to worry about that person being at increased composed of particularly risk of cardiovascular disease. To counteract vulnerable member groups 3) Measuring the savings The challenges posed by a population health the high cholesterol and its possible effects, the from improvements in science approach to health policy are enor- health care provider is thus likely to prescribe health care outcomes mous and require enough insight into all the cholesterol-lowering agents and recommend against the costs of direct and/or area-adjacent factors that affect health to be confident in eating fewer saturated fats and exercising more. policy interventions the chosen interventions. If researchers and This practice, however, ignores the burden policymakers are to understand and inter- of poor health being borne by those whose Who should take vene in factors ranging from national policy to cholesterol is certainly higher than the popula- the lead? individual behaviors, from urban planning to tion’s mean cholesterol but not over the cutoff Researchers, policymakers, and stakeholders the molecular mechanisms that affect health, that might suggest the need for intervention. in health care what should they focus on, and which of these These “borderliners” may get no such medicine factors are most likely to contribute to improved or advice. A population health recommendation health in populations? Ultimately, to answer would rely on policy approaches that encourage these questions, they must ask additional ques- everyone to eat healthy foods, not just those tions: What matters most?3 What are the most who already have high cholesterol, and would important elements to study, and what are the thus also protect the health of people who best policy investments for improving popula- fall below the cutoff for what is considered a tion health? dangerous cholesterol level. Such advice, if

18 behavioral science & policy | volume 4 issue 1 2018 followed, might prevent some from raising their cholesterol in the first place. “This notion nudges thinking

This emphasis on healthy eating rather than away from conceptualizing on cholesterol management would also help improve other aspects of population health that health as a binary” occur on a continuum. Policies to reduce the consumption of unhealthy food on a population scale could reduce the number of people who costs to both the health care system and the have or would otherwise come to have a high labor market. This approach was successful in body mass index (BMI), which is a sign of being both Finland and Japan.7 overweight or obese. Like having high choles- terol, being overweight or obese can increase Another strategy to encourage a popula- the risk for heart disease. It also increases the tion to make better food choices would be to risk of diabetes, which can contribute to heart impose taxes on sugar-sweetened food and attacks and other disorders. drinks, which play a role in increasing a popu- lation’s BMI. Several countries and cities have How might policy achieve the more far-reaching implemented these taxes, which have reduced goal of increasing healthy eating across a consumption of the taxed items. In Mexico, population? What people eat is driven in no taxes on sugar-sweetened beverages reduced small part by what is accessible, and there is a sales by 5% during the first year of their impo- gap in healthy food accessibility in the United sition and by almost 10% further during the States. People living in low-income or minority-­ second year. In Berkeley, California, a 25% majority areas are more likely than those who tax increase on sugar-sweetened beverages live in middle-income areas to have access resulted in a 21% reduction of sales in low-­ to overprocessed food, through inexpensive income neighborhoods merely four months fast food outlets and convenience stores, and after implementation.8 limited access to healthy food, which is usually available in large supermarkets. However, These examples suggest that policymakers among participants in the U.S. food stamp who want to improve health behaviors related program, easy access to supermarkets that to food should shift their focus from trying to provide fresh fruit and vegetables is associated understand how to change people’s specific with increased consumption of both.5 One way dietary choices to thinking about how to to increase access to healthy food would be to ensure that healthy food is available to all and encourage the establishment of retail stores how to reduce the population’s consumtion of and supermarkets that sell healthy foods in low-­ unhealthy food. Although this advice may make income neighborhoods, perhaps via subsidized intuitive sense, it has not typically been followed. loans or tax breaks. To date, enormous effort has been expended on behavior modification efforts that can only In England, opening supermarkets in low-­ plausibly benefit people who are at high risk for income neighborhoods led to a 60% increase in heart disease or other specific conditions rather the consumption of fruit and vegetables among than serving whole populations. those who had poor diets before the interven- tion.6 Opening stores in urban areas, where property is rarely cheap, may seem expensive, Principle 2: Focus on Affecting but here is why it makes economic sense: imple- Ubiquitous Influences on Health menting interventions that shift a population’s Health policymakers and health science cholesterol or blood pressure levels in the right researchers have historically been drawn direction will lead to fewer people experiencing to tackling factors that dramatically affect a heart attacks or strokes in the future, reducing person’s health. They therefore tend to expend

a publication of the behavioral science & policy association 19 substantial energy on mitigating very dangerous Once again, this principle can suggest a behaviors, such as injecting heroin.9 These sea change in priority setting in behavioral efforts are important, and we do not mean to science, from the factors that policymakers and suggest that extraordinarily harmful behaviors researchers may be accustomed to focusing should be ignored. on—high-risk behaviors—to more common behavioral influences that may affect many Yet, because extremely harmful behaviors are more people on a daily basis. not particularly prevalent, behavioral policies aimed at them have a very small effect on overall population health. For example, in 2016, an esti- Principle 3: Consider the Trade- mated 948,000 people in the United States used Offs Between Efficiency & Equity heroin. By comparison, an estimated 3.7 million A danger of thinking in terms of populations is adults—nearly four times as many people—had that it is easy to forget they consist of individ- a major depressive episode that same year. In uals of different races, ethnicities, genders, and 2016, roughly 35 million adults received mental socioeconomic classes and that these differ- health care, 37 times as many people as there ences, as well as a range of other factors, can 3.7M 10 are heroin users. Although major depression is lead to variance in how these individuals behave Health Disparities not as acutely threatening as heroin abuse, it is and respond to different conditions. Helping Americans who had a major depressive an important risk factor for a range of adverse one part of a population by implementing the episode in 2016 consequences, including drug abuse11 and easiest health policy intervention will certainly suicide.12 A population health approach would boost overall measures of health, but it may encourage policymakers to consider interven- fail to assist other parts of the population, tions that could influence the mental health of often those who are disadvantaged. To choose whole populations rather than that of people among potential interventions, policymakers in one small, specific subgroup of the popula- therefore need to consider whether they value tion. For instance, depression is influenced by efficiency over health equity or vice versa. stressors that may be ubiquitous in populations, such as food insecurity and housing instability.13 The United States approach to colorectal cancer $147 - $210 billion Cost of the obesity Society may be better served, then, by insti- screening illustrates this trade-off. To increase epidemic per tuting policies that reduce food insecurity and screening rates, the U.S. Preventive Services Task year in the US housing instability than by concentrating efforts Force developed national guidelines. The guide- solely on high-risk, low-prevalence behaviors lines, which focused on reaching health care that affect the health of only a few. Putting such providers and on educational campaigns, led 28.2% to 62.9% policies in place will also help put a dent in the to an increase in screening rates in the United U.S. opioid epidemic.14 States from 38.2% in 2000 to 62.9% in 2015.17 Increase in US colorectal cancer screening Yet follow-up studies consistently showed a gap between 2000 and 2015 The city of Denver offers evidence for the in screening rates. One nationally representative wisdom of this approach. A supportive housing analysis found that people with a primary health initiative for the chronically homeless there led care provider (that is, someone they thought to improvement in the overall health of partic- of as their doctor) were almost four times as ipants. Specifically, 43% of those served by the likely to receive a screening test as were those initiative showed better mental health outcomes without such a provider. The analysis also found and a 15% reduction in substance use.15 Another that race, educational level, and income all example is the Moving to Opportunity experi- contributed to the probability of undergoing a ment in New York City, which relocated families screening test. Those with at least one primary living in public housing in high-poverty neigh- health care provider tended to be older, female, borhoods to low-poverty neighborhoods. Adult and non-Hispanic White; tended to have higher participants in the experiment showed a 20% income, more education, and health insurance; reduction in depressive symptoms compared and were most likely to receive up-to-date with participants in the control group.16 colorectal cancer screening.18

20 behavioral science & policy | volume 4 issue 1 2018 Informational campaigns that notify people who have stable health care providers about the “screening programs that availability of screenings will encourage those individuals to connect with their provider and focus on narrowing health arrange a screening. Overall screening rates will increase. But this approach is unlikely to do much gaps can indeed reduce for marginalized populations who do not have regular care providers, thereby widening gaps these gaps” between health haves and health have-nots. where screening is more dependent on an indi- By contrast, screening programs that focus on vidual’s ability to access the health care system.21 narrowing health gaps can indeed reduce these Population-based approaches may, in the short gaps. To shrink racial disparities in disease inci- run, be more difficult and costly to implement dence and mortality in Delaware, the state than education campaigns, but these European government created a screening program countries made a priority of improving health in available to the entire population (that is, a popu- disadvantaged groups. lation-based intervention). Further, the program offered treatment at no cost for uninsured indi- The national colorectal cancer screening educa- viduals who screened positive for colorectal tion program in the United States efficiently cancer. In addition to increasing the overall improved screening rates when the population screening rate, the Delaware program reduced is viewed as a whole but at the cost of increasing morality rates from colorectal cancer among inequities within the population. Is this trade-off African Americans by 51%, nearly eliminating the justifiable? This question is not a scientific issue gap between them and Whites.19 Although this but a values question, and it is one that can be program cost the state $1 million per year, as we answered only if policymakers are aware of the note later, it was highly cost effective. values they bring to their work. In some circum- stances, they may consider a trade-off between Massachusetts General Hospital Chelsea efficiency and equity acceptable. For example, HealthCare Center, a community health center, when an infectious disease epidemic is raging, adopted a different approach to colorectal achieving high rates of vaccination quickly is cancer screening, reducing the screening gap important, regardless of the cost or uneven between Latino patients and all patients visiting distribution of services. At other times, making the center. The hospital provided outreach decisions without thought to the trade-offs and workers who matched patients both cultur- how to value them is indefensible. Conscious ally and linguistically to help them navigate the consideration of trade-offs between efficiency health care system and tackle barriers to cancer and equity should be front and center in behav- care. Within four years, the program improved ioral science health policy discussions of both both the overall screening rates and health researchers and policymakers. There are no equity in vulnerable populations, especially rules of thumb about what should be valued, when compared with the performance of other but the very act of raising the notion that values practices in the area.20 dictate how people act can push policymakers to reckon with the trade-offs we are making Similar trends have been seen with both implicitly, to the end of forcing us to be honest cervical and breast cancer screenings. A review about why we choose to act in the way we do. of screening programs in 22 European coun- tries found smaller differences in screening rates between lower socioeconomic and Principle 4: Evaluate higher socioeconomic groups in countries that Return on Investment provided national screening programs for their Prevention is the heart of population health entire population, as compared with countries thinking and public health practice. Most people

a publication of the behavioral science & policy association 21 disease-specific mortality, common metrics to “Supporting public measure return on investment include improve- ments in disability adjusted life years (DALYs) transportation would also or quality adjusted life years (QALYs) gained through an intervention. Both measures assess help address the obesity the effects of interventions on years and quality epidemic” of life, albeit in different ways. Let’s look at transportation investments for a fuller example of return-on-investment consid- would prefer not being sick in the first place to erations. In a city of a million people, a 40% being treated for illness. When policymakers are expansion of public transit systems delivers an setting priorities, they should consider another annual health benefit worth more than $200 compelling argument for favoring programs million.22 This yield comes from spurring people that could prevent disease: such policies can to walk more and reducing pollution, among yield a good return on investment, in terms other benefits. This finding is a compelling of both improved population health and cost argument for investing in transportation as a savings. Policymakers who want to improve health policy. public health should assess programs’ potential return on investment as they consider which Yet that is not the only argument for expanding ones to implement. public transportation. Supporting public trans- portation would also help address the obesity The Denver program supporting housing epidemic, which has real, crippling costs stability mentioned earlier offers a case in point: ranging from $147 to $210 billion per year in the it led to the city achieving a net cost savings United States.23 Such an intervention can be a of $4,745 per participant by preventing unfa- win–win for city planning, health system costs, vorable health outcomes.16 The colorectal and the health of populations alike.24 The benefit screening program in Delaware cost the state of reducing obesity would extend even further, $1 million annually, but it led to $8.5 million in because of obesity’s contribution to the burden annual savings from reductions in costs related of such chronic conditions as diabetes, heart to colorectal cancer.19 disease, and cancer. Health care for people with multiple chronic conditions represented 71% of A return-on-investment approach examines health care expenditures in the United States the yield on a particular policy intervention. in 2010.25 In 2012, the estimated costs of diag- Potential interventions can be evaluated by nosed cases of diabetes were $245 billion.26 A considering the extent to which any partic- 10% reduction in mortality due to heart disease, ular approach is likely to yield returns in health, cancer, and diabetes in the United States would whether that return is worth the financial and generate a return on investment of $10.9 tril- other costs of a particular effort, and, most lion.27 Viewed as a return-on-investment practically, how one intervention compares with argument, investments in public transportation another on those features. Metrics to measure clearly have the potential to deliver enormous return on investment in population health can yields in population health. be described in terms of actual health benefits, cost benefits, or many other parameters. For Returns on early childhood education invest- example, one metric by which one can assess ments provide more support for this principle. the success of a subsidized gym membership One program showed, for example, that early program is the number of sick days taken during childhood education provides a 5:1 return rela- a time period. (Society benefits from having tive to costs, with positive outcomes taking the healthier workers who miss fewer days of work.) form of reductions in crime rates, child maltreat- In addition to occurrences of a specific health ment, and teen pregnancy, as well as gains in event during a time period and all-cause or academic achievement.28 The Perry Preschool

22 behavioral science & policy | volume 4 issue 1 2018 Project, established in the 1960s, is also instruc- health, return-on-investment assessments for tive. The school delivered high-quality education proposed recommendations can help sell those to 3- and 4-year-old African-­American chil- recommendations to leaders in the private dren living in poverty. Children attended daily sector, whose decisions inevitably influence educational sessions and received weekly home how people behave and how healthy they are. visits to involve their mothers in the educational process. Forty years later, 77% of those children had graduated from high school, compared In Conclusion with 60% of the children from the control group. Figuring out how best to enhance population Participants in the Perry Preschool Project were health is a daunting undertaking, considering all 20% more likely than those in the control group the public health, social, and economic levers to earn more than $20,000 a year, and they had that can be pulled. The principles outlined in this lower crime rates.29 The effects of early educa- article should help policymakers organize their tion extended to providing both direct and thinking and establish policies and programs indirect health benefits. Early education predicts that will do the most good, maximally improving higher education attainment, which, in turn, the health of the communities they serve. predicts a better ability to make health-­related decisions as well as higher income levels. All of those factors ultimately play roles in determining author affiliation the health of an individual. Ettman: Boston University and Brown University. Beyond providing clarity to policymakers Abdalla and Galea: Boston University. Corre- directly concerned with improving population sponding author’s e-mail: [email protected].

a publication of the behavioral science & policy association 23 references

1. Keyes, K. M., & Galea, S. (2016). Setting Substance Abuse, 7, 481–497. https:// 21. Palència, L., Espelt, A., Rodríguez- the agenda for a new discipline: doi.org/10.1016/0899-3289(95)90017-9 Sanz, M., Puigpinós, R., Pons-Vigués, Population health science. American 12. Nock, M. K., Hwang, I., Sampson, N. A., M., Pasarin, M. I., . . . Borrell, C. (2010). Journal of Public Health, 106, 633–634. & Kessler, R. C. (2010). Mental disorders, Socio-economic inequalities in breast 2. Kaplan, R. M. (2004). Shared medical comorbidity and suicidal behavior: and cervical cancer screening practices decision making: A new tool for Results from the National Comorbidity in Europe: Influence of the type of preventive medicine. American Journal Survey Replication. Molecular screening program. International of Preventive Medicine, 26, 81–83. Psychiatry, 15, 868–876. https://doi. Journal of Epidemiology, 39, 757–765. https://doi.org/10.1093/ije/dyq003 3. Keyes, K., & Galea, S. (2015). What org/10.1038/mp.2009.29 matters most: Quantifying an 13. Black, M. M., Quigg, A. M., Cook, J., 22. Litman, T. (2010). Evaluating public epidemiology of consequence. Casey, P. H., Cutts, D. B., Chilton, M., . . transportation health benefits. Retrieved Annals of Epidemiology, 25, . Frank, D. A. (2012). WIC participation from https://www.apta.com/resources/ 305–311. https://doi.org/10.1016/j. and attenuation of stress-related child reportsandpublications/Documents/ annepidem.2015.01.016 health risks of household food insecurity APTA_Health_Benefits_Litman.pdf 4. Keyes, K. M., & Galea, S. (2016). and caregiver depressive symptoms. 23. Strategies to Overcome and Prevent Population health science. New York, Archives of Pediatric Adolescent Obesity Alliance & The George NY: Oxford University Press. Medicine, 166, 444–451. https://doi. Washington University School of org/10.1001/archpediatrics.2012.1 Public Health & Health Services. 5. Rose, D., & Richards, R. (2014). Food (2017). Fast facts: The cost of obesity store access and household fruit and 14. Abt Associates. (n.d.). Opioid use disorder (OUD), housing instability, and [Fact sheet]. Retrieved from http:// vegetable use among participants in the stopobesityalliance.org/wp-content/ US food stamp program. Public Health housing options for recovery. Retrieved from https://www.abtassociates.com/ themes/stopobesityalliance/pdfs/ Nutrition, 7, 1081–1088. https://doi. Fast%20Facts%20Cost%20of%20 org/10.1079/PHN2004648 projects/opioid-use-disorder-oud- housing-instability-and-housing- Obesity.pdf 6. Wrigley, N., Warm, D., & Margetts, B. options-for-recovery 24. Cawley, J., & Meyerhoefer, C. (2012). (2003). Deprivation, diet, and food- 15. Perlman, J., & Parvensky, J. (2006). Cost The medical care costs of obesity: retail access: Findings from the Leeds An instrumental variables approach. ‘food deserts’ study. Environment and benefit analysis and program outcomes report. Retrieved from https://shnny. Journal of Health Economics, 31, Planning A: Economy and Space, 35, 219–230. https://doi.org/10.1016/j. 151–188. https://doi.org/10.1068/a35150 org/uploads/Supportive_Housing_in_ Denver.pdf jhealeco.2011.10.003 7. World Health Organization. (2002). 16. Leventhal, T., & Brooks-Gunn, J. (2003). 25. Gerteis, J., Izrael, D., Deitz, D., LeRoy, L., Chapter 6: Strengthening risk prevention Ricciardi, R., Miller, T., & Basu, J. (2014). policies. Retrieved from http://www. Moving to opportunity: An experimental study of neighborhood effects on Multiple chronic conditions chartbook: who.int/whr/2002/en/Chapter6. 2010 Medical Expenditure Panel Survey pdf?ua=1 mental health. American Journal of Public Health, 93, 1576–1582. https:// data. Retrieved from https://www. 8. Sweet success: Will sugar taxes improve doi.org/10.2105/AJPH.93.9.1576 ahrq.gov/sites/default/files/wysiwyg/ health? [Editorial]. (2017). The Lancet professionals/prevention-chronic-care/ Diabetes & Endocrinology, 5, 235. 17. National Cancer Institute. (2017). decision/mcc/mccchartbook.pdf Colorectal cancer screening. Retrieved Retrieved from https://doi.org/10.1016/ 26. American Diabetes Association. (2018). S2213-8587(17)30070-0 from https://progressreport.cancer.gov/ detection/colorectal_cancer The cost of diabetes. Retrieved from 9. Irwin, A., Jozaghi, E., Weir, B. W., Allen, http://www.diabetes.org/advocacy/ S. T., Lindsay, A., & Sherman S. G. (2017). 18. Cardarelli, R., & Thomas, J. E. (2009). news-events/cost-of-diabetes.html Mitigating the heroin crisis in Baltimore, Having a personal health care provider and receipt of colorectal cancer testing. 27. Harvard School of Public Health. (2012). MD, USA: A cost-benefit analysis of Infographic: The dollars and sense of a hypothetical supervised injection Annals of Family Medicine, 7, 5–10. https://doi.org/10.1370/afm.904 chronic disease. Retrieved from https:// facility. Harm Reduction Journal, www.hsph.harvard.edu/news/magazine/ 14, 1–14. https://doi.org/10.1186/ 19. Verma, M., Sarfaty, M., Brooks, D., & f12-infographic-chronic-disease/ s12954-017-0153-2 Wender, R. C. (2015). Population-based programs for increasing colorectal 28. Halfon, N., & Hochstein, M. (2002). 10. Center for Behavioral Health Statistics Life course health development: An and Quality. (2016). Key substance use cancer screening in the United States. CA: A Cancer Journal for Clinicians, integrated framework for developing and mental health indicators in the health, policy, and research. The United States: Results from the 2015 65, 496–510. https://doi.org/10.3322/ caac.21295 Milbank Quarterly, 80, 433–479. https:// National Survey on Drug Use and Health doi.org/10.1111/1468-0009.00019 (HHS Publication No. SMA 16-4984, 20. Percac-Lima, S., López, L., Ashburner, J. NSDUH Series H-51). Retrieved from M., Green, A. R., & Atlas, S. J. (2014). The 29. Schweinhart, L. J., Montie, J., Xiang, Z., https://www.samhsa.gov/data/sites/ longitudinal impact of patient navigation Barnett, W. S., Belfield, C. R., & Nores, default/files/NSDUH-FFR1-2015/ on equity in colorectal cancer screening M. (2005). Lifetime effects: The High/ NSDUH-FFR1-2015/NSDUH-FFR1-2015. in a large primary care network. Scope Perry Preschool study through pdf Cancer, 120, 2025–2031. https://doi. age 40 (Monographs of the High/Scope org/10.1002/cncr.28682 Educational Research Foundation, No. 11. Grant, B. F. (1995). Comorbidity 14). Ypsilanti, MI: High/Scope Press. between DSM–IV drug use disorders and major depression: Results of a national survey of adults. Journal of

24 behavioral science & policy | volume 4 issue 1 2018

essay The ubiquity of data & communication: A double-edged sword for disparities Robert M. Califf

abstract ‡ The fourth industrial revolution, in which most information is stored in digital form, is characterized by connectivity and communication among people and groups via, for instance, cell phones and smart watches. The amount of information now generated about people’s health-related activities is multiple log orders more voluminous and complex than the data currently captured in the electronic health record from patient interactions with clinicians. Despite the data’s complexity, it is now possible for health care administrators, policymakers, and clinical researchers to develop—and then test—data-informed interventions that could reduce health disparities. For example, programs initiated by a county government and a major medical system have, respectively, improved asthma management and reduced lead exposure in their localities. Use of big data can be a double-edged sword, however. The technology that allows for high-end use of data also opens the way to increasing disparities, as could happen, for instance, if geospatial information were used to locate clinics in places that optimize profit rather than meet health needs. Efforts are underway to limit this risk.

Califf, R. M. (2018). The ubiquity of data & communication: A double-edged sword of disparities. Behavioral Science & Policy, 4(1), 27–37.

a publication of the behavioral science & policy association 27 or decades, it has been common knowl- into the algorithms to improve them. Some key edge that vast disparities in health innovations relate to data fluidity—the capacity Foutcomes and access to health care occur for data to flow easily and without undue fric- both in the United States and across the globe. tion among all the users who need it—and data Until recently, clinicians, health care adminis- latency—the speed at which data, once gath- trators, policymakers, and clinical researchers ered, is available for analysis. have lacked timely access to the type and w quantity of information that would enable What if a similar approach were taken to human interventions that might ameliorate disparities health care, including treatment and preven- Core Findings between individuals and populations. With the tion? I argue that the ubiquity of data and advent of the fourth industrial revolution,1 that its rapid communication provide previously What is the issue? situation is quickly changing. unimaginable resources for understanding The advent of the fourth and addressing health disparities. One simple industrial revolution has Whereas the third industrial revolution was possibility is that home services could be better brought a wealth of new technologies and an marked by the introduction of digital tech- coordinated by systematically applying opti- exponential increase in nologies such as the Internet and personal mized transportation routing and scheduling of information. This presents computers, the fourth industrial revolution the kind now available with smartphone apps. an unprecedented opportunity to address is characterized by the merging of biolog- Thus, a patient with uncontrolled diabetes living disparities in health care. ical, physical, and information sciences.1 This in an underresourced neighborhood would New methods of data fourth revolution has enormously expanded receive more frequent home visits from a nurse curation and analysis can connectivity and communication among indi- than a less ill person would. lead to more effective interventions. But without viduals and groups. For instance, today’s cell societal involvement and phones reach almost everyone on the planet, Already, in Durham County, North Carolina, a participation in guiding regardless of income, education, and physical data system connected to the Duke University new technology, health disparities are at risk of location, and they provide instant access to vast Health System has contributed to a substantial becoming more acute. digital communication and information systems. improvement in locating children with elevated Given the almost universal ability to connect to blood lead levels.2 The Duke system created How can you act? distributed cloud computing, the computing a map that estimates household lead expo- Selected recommendations include: power accessible through a typical cell phone sure risk based on county tax assessor data, 1) Developing a single today exceeds the computational resources of blood lead screening results from clinic visits, standard for health entire universities or medical centers from just and census data. Stakeholders, including the care data pertinent to a few years ago. Durham County Health Department and several health outcomes 2) Using new technologies community advocacy groups, have used this to provide useful New technologies for mapping and wayfinding map to reach at-risk families. information in a way illustrate the potential of the connectivity and that is tailored to the specific health needs communication that are hallmarks of the fourth Similarly, in Louisville, Kentucky, the public– of an individual, family, revolution. Many drivers already rely on global private AIR Louisville consortium is helping local or population group positioning systems to help them navigate unfa- residents manage their asthma, a disorder that miliar streets and highways, using information disproportionately affects Black children and Who should take that is accessed, contextualized, and integrated people living below the poverty line.3 Relying the lead? Researchers, policymakers, in real time. Smart wayfinding apps provide a on electronic sensors in inhalers, the program and stakeholders driver with current data about the driver’s loca- provides feedback about triggers, adherence in health care tion, traffic and road conditions, upcoming to treatment, and level of control to asthma businesses and landmarks, and reports from patients, which has resulted in a 78% reduction other drivers. This information helps drivers in rescue inhaler use and a 48% improvement avoid accidents, potholes, and traffic jams. in the number of symptom-free days. After Drivers also receive predictions about the combining crowd-sourced data on inhaler use impact of interventions (such as the effect on with environmental information, a government driving time that taking a different route might and community activist team crafted policy have) to support their decisionmaking. Actual recommendations to lower the incidence of results of driving experiences are then fed back asthma attacks citywide, such as increasing the

28 behavioral science & policy | volume 4 issue 1 2018 “Researchers and clinicians are now able to amass novel kinds of biological data, such as an individual’s genetic code”

tree canopy (to reduce air pollution and urban revolution—consequences that may turn heat), requiring that facilities with vulnerable advances into a double-edged sword. populations (such as children and the elderly) be located at least 500 feet from roadways with high traffic and high emissions, and developing Advances in Data & Computing a community notification system that alerts Recent improvements in society’s ability to asthma sufferers when high-risk conditions are store and retrieve information, communicate about to occur. rapidly via digital networks, and analyze data using increasingly powerful methods have Although this powerful technology can be fundamentally enlarged the country’s capacity harnessed to reduce health disparities, it may to assess and intervene in health disparities. If also exacerbate them.4 For instance, data anal- policymakers, researchers, and clinicians take ysis has revealed that the Russian government is full advantage of this powerful combination of using bots to spread skepticism about the safety factors, they will be able to describe people’s of vaccines on Twitter in an apparent attempt to health in multiple dimensions simultaneously create discord in the United States.5 But many and access information as needed. Combined times, negative consequences may well happen with navigation systems at the personal, neigh- inadvertently. Machine learning is a type of arti- borhood, or community level, new analytic ficial intelligence that uses computer algorithms data capacities could identify, deal with, and to predict, for instance, what products you remeasure health problems in a previously might like on Amazon or what music you might unimaginable time frame. enjoy on Spotify. According to a recent report in JAMA Dermatology, a machine-learning Newly available data sets contain immensely algorithm that distinguishes between images of more information about individuals than is benign and malignant moles has the potential currently found in personal health records and to spot skin cancers missed by dermatologists.6 other transactions captured by the health care Early skin cancer diagnosis could particularly system. Researchers and clinicians are now aid Black patients, who are less likely than White able to amass novel kinds of biological data, patients to develop melanoma but are more such as an individual’s genetic code. With the likely to die from it. However, the machine price of whole-genome sequencing drop- learning algorithms have been trained largely ping dramatically,7 scientists can envision a on examples drawn from White patients and are time when this information, consisting of more only now being designed in a way that would than 3 billion base pairs, will be routinely avail- help control for potential bias. able as part of a person’s health record. The collection of such biomolecular data could In this article, I assess changes in the informa- lead to analyses that provide significant insight tion and data ecosystem that should enable into the impact of innate biology on a person’s policymakers, researchers, and clinicians to health and responses to the physical and social harness this ubiquitous information archi- environments. tecture to identify health disparities, provide a method for evaluating them, and create But purely genomic information is only a small effective interventions. I point out develop- part of the data that can increasingly be used ments that allow this new technology to move to construct a biomolecular profile. A profile forward at a very fast pace. In addition, I strike can also include information drawn from tran- a cautious note about the potential nega- scriptomics (the study of RNA molecules), tive consequences of the fourth industrial metabolomics (the study of molecules involved

a publication of the behavioral science & policy association 29 in metabolism), and analyses of the detailed cell phones, computers, and other personal workings of the immune system,8,9 in addition devices. This information is deeply informative to integrative physiological information, such of the 99% of their lives spent outside of clinics as heart rate and blood pressure, that can be and hospitals. Readily available and increasingly measured with digital sensors. Although defini- inexpensive sensors in cell phones and watches tive evidence indicates that social determinants collect detailed information about individuals outweigh genetic influences on health risk at a continuously over long intervals. Wearable population level, strong evidence also shows sensors can measure activity levels, tremors, that biology has an impact on individual dispar- gait, and flexibility. Analyses of keyboard use ities in disease susceptibility and outcome. and gait provide a deep measure of cognitive Further, as described below, when complex function, mood, and physical function. Use of data become less expensive to collect, taking cell phone apps and associated social media a sample or image, recording the information, can provide detailed insights into social activity. and digitally storing it for later use becomes Given the dominance of wealth, education, race, much more feasible. Thus, key outputs of a and location as mediators of health outcomes, successful data-intensive approach to health the ability to directly measure behavior and disparities will include the delineation of biolog- social interactions will provide insights that ical mechanisms by which disparities lead to could not be gained by asking patients ques- poor health outcomes as well as the develop- tions during visits to a clinic or study site. For ment of interventions able to counteract those example, if your goal is to reduce cardiovas- mechanisms—as was seen in AIR Louisville’s cular disease in a population, it may be more efforts to improve asthma control. Such work important to insert green space and healthy will also enable the planning of interventions food into neighborhoods than to increase the that simultaneously deal with biological and number of medical clinic visits. social determinants of health—for example, an asthma intervention that involves both The geospatial dimension is a particularly using medications and improving the home important factor in health disparities. It generally environment. holds true that the most important predictors of health are zip code and income.11,12 Current A fast-growing area of health measurement is technology allows health outcomes and called digital phenotyping,10 which character- determinants to be measured at a more gran- izes people based on the way they interact with ular level: household, street, neighborhood, county, and state levels. This type of measure- ment feeds into a potential understanding of Key Priorities as Identified by Stakeholders social networks but also provides a substantial in the Durham Health Innovations Project opportunity to make changes in the delivery system for both traditional medicine and social • Increase health care coordination and eliminate barriers to services and resources. services and then feed information about those changes back to residents of affected neigh- • Integrate social, medical, and mental health services. borhoods and to medical clinicians and social • Expand health-related services provided in group settings. service providers. As the speed of information • Leverage information technology. acquisition, access, and analysis continues to • Use social hubs such as places of worship, community centers, increase, it will be possible to craft interventions salons, and barbershops as sites for the distribution of clinical and social services and information. at geographical (for example, neighborhood) or social (for example, workplace, school, and • Increase local access to nurse practitioners, physician assistants, and certified nurse midwives. church) levels and measure outcomes to fine- • Use traditional marketing methods to influence health behavior. tune the interventions (see Figure 1).

Note. From “History,” by Durham Health Innovations, n.d. (https://sites.duke.edu/durham- healthinnovations/history). Copyright 2015 by Duke Division of Community Health. Reprinted The dimension of time is also especially with permission. important in gathering individual data. In the past, clinicians and clinical researchers who

30 behavioral science & policy | volume 4 issue 1 2018 Figure 1. Leveraging diverse sources of data to improve population health Population Health

Actionable Information

Analysis

Note. Information from across the spectrum of individual biology, clinical records, behavior, and social interactions can be combined with information about the fixed environment and the ambient environment (bottom). This information is summarized across individuals to produce population-based health measures that can be shared at the levels of the individual, the household, the neighborhood, the voting precinct, and the state. These lead to analysis and actionable information (middle sections). With appropriate analyses, evidence-based policies and interventions that are implemented (top) and assessed with repeated measures (dotted line on the left) can be devised to document the degree of eectiveness of the interventions. The information on more and less eective interventions can then be communicated to individuals and groups such as schools, churches, and neighborhood groups (dotted line on right). The partnership of these stakeholders is essential to the success of the eort. wanted to evaluate changes over time were interventions at the group level, as evidenced limited mostly to measurements made during by the use of social media to deal with environ- periodic clinic visits. Digital technologies and mental catastrophes. the massive increase in the ability to manage data make it possible to use both passive and The study of sleep offers an example of the active data collection to accurately portray the potential importance of collecting data that impact of time. Passive measurements, such as cover temporal, spatial, and behavioral dimen- the capture of heart rate and physical activity sions. In the past, this kind of research has relied on smart watches and wearable fitness moni- on either patient recall or intensive study in tors, can be obtained almost continuously. Cell sleep units, which are artificial environments phones can also be used to collect frequent with little resemblance to the home environ- passive impressions from the research partici- ment. Now, with passive sensors on wearable pant or patient. As discussed above, this radical technology, the quantity and quality of sleep reduction in data latency is also pertinent for and attributes of the home environment can be

a publication of the behavioral science & policy association 31 measured without disrupting normal patterns presence of green spaces, the availability of and without great expense to the patient. It physicians’ offices, and traffic patterns.2 Third, is likely that many of the medical, social, and multiple projects have shown that these data environmental factors that affect sleep are also can be merged to produce data sets that can be associated with differences in longevity. Enabled easily shared across systems, with the goal of by a system of integrated information such as improving outcomes. the one described above, a clinician could intervene at the individual level by prescribing The National Patient-Centered Clinical weight loss procedures (to reduce sleep apnea) Research Network (PCORnet; http://pcornet. or sleep medication. Medical institutions, health org) illustrates the magnitude of data integra- planners, and legislators could make changes tion that is possible. This project, funded by the at the group level through social networks or at Patient-Centered Outcomes Research Insti- the geographic and environmental levels by, for tute, has developed a systematic approach to instance, reducing noxious sounds by changing curating data across multiple health systems, traffic patterns. including both clinical care data derived from 650 electronic health records and insurance claims Health care systems that data, for over 100 million Americans. PCORnet account for 90% of all Harnessing Big Data is now evolving into the People-Centered hospital discharges The consolidation of clinical care systems, in Research Foundation (http://pcrfoundation. which big health systems absorb smaller ones, org/), a not-for-profit organization dedicated provides opportunities to measure and change to conducting pragmatic randomized trials health disparities. As of several years ago, the (research performed not at an artificial research Agency for Healthcare Research and Quality center but at a real patient point of contact estimated that approximately 650 health systems with an eye to informing decisionmakers of account for over 90% of hospital discharges and the comparative balance of benefits, burdens, are increasingly integrating hospital and outpa- and risks of a biomedical or behavioral health There are approximately tient care, including assisted living, nursing intervention) and observational studies (nonran- 3 billino base pairs in 13 the human genome. homes, and hospice. These systems are devel- domized studies that allow for historical oping sophisticated data “lakes” and warehouses comparisons). to aggregate information that is curated so that it can show how health care can be delivered One interesting component of the People-­ in a financially sustainable manner. (The word Centered Research Foundation is the ADVANCE 60% curated deliberately evokes the way historical Collaborative, a network of federally qualified Health Disparities archives work. Curated data include critical community health centers.14 This organiza- Variation in countywide accompanying information about a data point’s tion includes OCHIN, Health Choice Network, life expectancy due to context—where it was gathered and how it was Fenway Health, Kaiser Permanente Center socioeconomic and race/ethnicity factors gathered, stored, transmitted, and transformed, for Health Research, Legacy Health, Care- and by whom.) Oregon, and Oregon Health and Science University. Intended as a learning laboratory Because these health systems need standard- for policymakers to better understand patients ized information to conduct business, data in who use safety-net services, the system has several dimensions need to become integrated digital records on more than 3 million patients, in a variety of ways. First, the various entities including large numbers of homeless, unin- within each system must use common stan- sured, underinsured, and undocumented dards and definitions for the system to function people, as well as members of other under- efficiently and for patients to benefit from the represented populations. This type of network, sharing of information across practices. Second, if coordinated with more traditional integrated health care delivery system data should be health care systems, could provide a mecha- integrated with information from other dimen- nism not only for evaluating disparities but also sions that are pertinent to health—for example, for designing and testing interventions, such as social and environmental factors such as the drug counseling, at the system level.

32 behavioral science & policy | volume 4 issue 1 2018 Opportunities for Data to Characterize Disparities “it should be possible to Several organizations are already providing analyses of data that depict health disparities in provide useful information in intuitively understandable displays. Perhaps the most far-reaching of these reports is put out by a way that is tailored to the collaborators working on the Global Burden of Disease Study.15 Another recent series of reports specific health needs of an done mainly at Harvard University has clearly demonstrated that variation in longevity and individual or family” disease burden in the United States is a func- tion of geographic location at the county level.16 The striking impact of residence in rural coun- Health Disparities). At the level of a community ties is highlighted by the visual depiction of both or health system, interactions between tradi- current longevity and trends in longevity over tional health systems and social services tend to time. One recent report in this series deter- be inefficient. Improved labeling of government, mined, for instance, that 60% of the variation private, and volunteer services and coordina- in countywide life expectancy is explained by tion of these services with clinics, schools, and socioeconomic and race and ethnicity factors businesses could lead to a much more directly and that rural counties fare worst on such effective intervention system. measures as mortality rates, suicides, drug overdose deaths, rates of teenage pregnancy, In addition, the ubiquity of cell phones and and fetal and maternal mortality.12 Detailed steep reductions in the cost of sensors make analyses of these data have demonstrated that, it possible for clinicians to communicate as expected, wealth, education, race, sex, and directly with individuals and groups at any location are key factors in longevity,12 mediated interval that is desired. Additionally, almost all in common chronic diseases by factors such as people use search engines to seek informa- blood pressure, low physical activity, tobacco tion on a routine basis. Search engine results use, obesity, depression, and diabetes mellitus. are tailored by machine-learning algorithms to an individual’s pattern of communication. Many organizations routinely produce compar- As the curation and organization of informa- ative reports of health status. Within the United tion continues to improve, it should be possible States, significant efforts are aimed at curating to provide useful information in a way that is actionable data at the level of the city, county, tailored to the specific health needs of an indi- or state. Some of the most potent information vidual or family. For instance, these approaches comes from the evaluation of boroughs in New can be used to fine-tune search results, much York 17 and from state-level reporting by the as consumer goods are currently surfaced in Robert Wood Johnson Foundation.18 Research a manner consistent with the preferences of in Durham County, North Carolina, demon- the consumer. For example, when someone strated the power of this information when it searches on the term stage 1 breast cancer, was applied at the level of individual households it is technically possible for the high-ranked and neighborhoods to, among other goals, results to be tailored to the medical literacy of reduce exposure to lead poisoning.2 the individual as well as authoritative, relevant, and trustworthy. For the most common health searches, Google is currently providing “knowl- New Ways to Use Data edge panels” that are vetted by medical experts. to Reduce Disparities The same information infrastructure used for Consider the ongoing epidemic of asthma. measurement could also be used for imple- Asthma is often exacerbated by environ- menting interventions (See Key Actions Needed mental triggers, both within and outside the to Collect and Use Actionable Data to Reduce home. Futher, research has demonstrated that

a publication of the behavioral science & policy association 33 disparities in asthma incidence and access to more powerful than medication in preventing care are functions of wealth, education, physical asthma exacerbations. location, and race.19 Inexpensive sensor tech- nology and ubiquitous data networks would Similarly, obesity and diabetes contribute to an enable clinicians to monitor environmental enormous amount of death and disability, and quality at the household and neighborhood the geographic and social profiles of relevant levels, which would make it possible for them health disparities are clear.20 Although special to deploy precise interventions to reduce medical clinics, surgical and medical interven- stimuli that exacerbate asthma. The previously tions, and wide dissemination of accurate and mentioned report from the study in Louisville3 useful information that reaches the people who points out that this sort of intervention, which need it are all possible solutions, there is ample focuses on cleaning up the home and neighbor- reason to believe that constant exposure to hood environments, would potentially be much advertising for food, long distances to grocery stores that sell healthful food, and cultural and environmental influences on physical activity limit the success of medical interventions for people with lower incomes or other socio- Key Actions Needed to Collect & Use economic disadvantages. A wealthy, highly Actionable Data to Reduce Health Disparities educated person who can afford a personal Engage people and communities as partners trainer or gym membership and is not caring • Requires face-to-face time and use of social media for family members is more likely to be able to • Transparency is critical at all steps engage in a healthy lifestyle. Health systems, • Issues of privacy and confidentiality require considerable work advocacy groups, community leaders, and individuals at all levels of government can Collect diverse sources of data engage with people more productively within • Biological, clinical, behavioral, social, and environmental data are their personal digital environments, helping needed them to use geospatial information to locate • Full data use will require solutions to engagement, partnership, and healthful food resources and enabling and privacy/confidentiality issues encouraging them to integrate physical activity Curate and organize data into their routines in a more economically • Curation and organization are currently the most underinvested feasible manner that intrudes less into other area of data science, requiring significant investment aspects of life. • Requires conscious investment at the institutional level by health systems and government entities Analyze Importance of Community • Methods involving geospatial orientation and hierarchical analysis Engagement & the Development from the level of the individual to population will be informative of Shared Approaches • These data are big One area that researchers need to study more • Speed of access to data and fluidity of data are critical factors in thoroughly is how to best transmit new infor- making data actionable mation directly to those who are affected by it • Identifiable data will be most actionable but also riskiest from a as well as to those who can implement inter- privacy/confidentiality perspective ventions and policies to improve outcomes. Use outputs of analyses to formulate policies Although research that engages communities21 • Requires collaboration across health systems, neighborhoods, and continues to advance, and many communities policymakers at local, state, and national levels are involved in direct interventions, policymakers • Participants in the effort need education on quantitative and still lack clarity on which methods are likely community engagement methods to be most effective at linking personal health Implement policies data with social and environmental information in ways that yield measurable improvement in Measure again and adjust on the basis of outcomes outcomes. A promising approach has been developed by the Abdul Latif Jameel Poverty

34 behavioral science & policy | volume 4 issue 1 2018 Action Lab, which is using observational, exper- imental, and quasi-experimental methods “it is critical that data holders to understand which social policies lead to improved outcomes.22 One such study showed discuss their intentions with that charging fees for preventive medicine tools in low-income countries drastically reduces the data providers” their usage.23 in the introduction. It pulled in diverse sources Over a decade ago, Durham County, North of data to provide a holistic understanding of Carolina, and Duke University initiated an health needs, with geographic information inte- ambitious program that showed how to create grated to guide intervention. But because health successful communication between govern- data do not enjoy the kind of fluidity and latency ment, health institutions, and the community. advantages leveraged by wayfinding apps, DHI Called Durham Health Innovations (DHI),24 the was limited in what it could accomplish. Some program existed in the context of a history of the critical roadblocks to implementing that included decades of both outstanding fourth industrial revolution–style solutions can collaboration between the university and be overcome once project directors can access the community and well-documented divi- the right data, in the right way, at the right time. sions and disparities.25 For the project, teams of volunteers were organized so that they had equal representation from the community Steps to Limit the Risks and the university (including its health system of Data Sharing and academic medical center). Each team was Sharing data between patients, physicians, focused on addressing a particular health issue and institutions requires a degree of trust. And of significant concern to the community. Teams clearly, for all the good data sharing can do, a were then supported with the data assets of the markedly enhanced system of measurement, academic health system, the Durham County assessment, and intervention could be used public health department, and other entities and for negative as well as positive purposes. For asked to devise an approach to health care and instance, Facebook’s brokerage of personal community intervention that would improve data during the 2016 U.S. presidential election health outcomes relevant to the issue they had showed that social media data could be put to chosen. The teams winnowed their issues down nefarious uses.26 Theoretically, health care infor- to 10 major problems affecting the commu- mation, perhaps hacked from a medical center’s nity: adolescent health; asthma and chronic patient portal, could be leveraged to, say, target obstructive pulmonary disease; cancer preven- underinsured cancer patients with an ineffective tion and early detection; cardiovascular disease; but expensive “cure.” False news is particularly diabetes; HIV, sexually transmitted diseases, dangerous. Empirical evidence is accruing that and hepatitis; maternal health; obesity; pain false statements, many of which are intended management; and healthy aging in place.24 to stoke differences among people, reach more people faster and persist longer than truthful Remarkably, after a series of meetings and statements do because their novelty gives discussions aided by intensive data analysis, them “legs”.27 These issues are not new, but the proposed approaches to these seemingly they have emerged as critical considerations different problems all converged on a common in deciding how to use information to improve set of interventions that could improve the the outcomes of those who suffer from health health and health care delivery in Durham disparities. regardless of disease or therapeutic area (see Key Priorities as Identified by Stakeholders in To increase transparency, engagement, and the Durham Health Innovations Project). The trust by patients, clinicians, and institutions, it project was ahead of its time and in many ways is critical that data holders discuss their inten- anticipated the wayfinding approach mentioned tions with the data providers. In a body of work

a publication of the behavioral science & policy association 35 on the moral obligations of health systems that A Possible Future are continuously evaluating accruing data to The amount of information now available about improve health care and health outcomes—that individuals and their health—constantly gener- is, learning health systems—ethicists Nancy E. ated and recorded by new, interconnected Kass and Ruth R. Faden explore a crucial funda- devices—is multiple log orders more volumi- mental concept that bears on these challenges: nous and complex than the data from patient the reciprocal obligation of those who use interactions currently available to clinicians, data to those who provide the data.28 Although health care administrators, policymakers, and these concepts are reasonable, they have not clinical researchers, and the cost of managing yet been fully implemented. It is interesting that the data is rapidly declining. It is technologi- Kass and Faden’s scheme considers reducing cally possible to observe, describe, and analyze disparities to be an essential element of learning health disparities at the levels of individuals, health systems. households, streets, neighborhoods, cities, and counties. Information-­based profiles at Given the complexity of interactions needed each of these levels could be further segre- to successfully implement interventions that gated by biological, medical, behavioral, social, reduce health disparities, it will be important for or environmental characteristics. The relevant multiple societal entities to involve themselves information that could guide and measure in developing the cultural and legal expectations interventions to improve health and reduce that will enable big data to be used effectively health disparities could be displayed in any for desirable purposes. For example, the current time interval desired. Interventions could be HIPAA laws deal with health care data in a planned at any of these levels and the results manner that many consider to be overly restric- measured in ways that would reveal cause and tive, whereas the data from “the other 99%” of effect and suggest useful interventions. These life, which have a much larger impact on health same methods, however, if applied for selfish outcomes, are governed by much less restric- or ignoble purposes, could be leveraged to tive rules. Perhaps it would be better to have a increase health disparities. Nevertheless, here at single data standard that is pertinent to health the leading edge of the fourth industrial revolu- outcomes. For this expectation-setting effort tion, new methods of data curation and analysis to succeed, societal entities that are not moti- can provide the foundation for a dramatically vated by profit and are capable of convening improved approach to health disparities for both diverse interests should help to devise stan- individuals and populations. dards. Universities are in a special position to engage with communities and offer the bene- fits of faculty knowledge and skills in medicine, author affiliation law, technology, and ethics as they work with community groups and individuals to devise Califf: Duke University School of Medicine, policies that will achieve the desired results. Stanford University School of Medicine, and Verily Life Sciences. Corresponding author’s e-mail: [email protected].

36 behavioral science & policy | volume 4 issue 1 2018 references

1. Schwab, K. (2016). The fourth industrial 2001–2014. JAMA, 315, 1750–1766. diabetes-and-its-burden-in-the-united- revolution: What it means, how to https://doi.org/10.1001/jama.2016.4226 states.pdf respond. Retrieved from https:// 12. Dwyer-Lindgren, L., Bertozzi-Villa, 21. Michener, L., Cook, J., Ahmed, S. M., www.weforum.org/agenda/2016/01/ A., Stubbs, R. W., Morozoff, C., Yonas, M. A., Coyne-Beasley, T., & the-fourth-industrial-revolution-what- Mackenbach, J. P., van Lenthe, F. J., . . Aguilar-Gaxiola, S. (2012). Aligning it-means-and-how-to-respond/ . Murray, C. J. L. (2017). Inequalities in the goals of community-engaged 2. Miranda, M. L., Ferranti, J., Strauss, life expectancy among US counties, research: Why and how academic B., Neelon, B., & Califf, R. M. (2013). 1980 to 2014: Temporal trends and key health centers can successfully engage Geographic health information systems: drivers. JAMA Internal Medicine, 177, with communities to improve health. A platform to support the “triple aim.” 1003–1011. https://doi.org/10.1001/ Academic Medicine, 87, 285–291. Health Affairs, 32, 1608–1615. https:// jamainternmed.2017.0918 22. Abdul Latif Jameel Poverty Action Lab. doi.org/10.1377/hlthaff.2012.1199 13. Agency for Healthcare Research (n.d.). About us. Retrieved from https:// 3. Barrett, M., Combs, V., Su, J. G., and Quality. (2016). Compendium of www.povertyactionlab.org/about-j-pal Henderson, K., Tuffli, M., & the AIR U.S. health systems, 2016. Retrieved 23. Abdul Latif Jameel Poverty Action Louisville Collaborative. (2018). AIR from https://www.ahrq.gov/chsp/ Lab. (2018). The impact of price on Louisville: Addressing asthma with compendium/index.html#data take-up and use of preventive health technology, crowdsourcing, cross- 14. Patient-Centered Outcomes Research products. Retrieved from https://www. sector collaboration, and policy. Institute. (2017). Accelerating Data povertyactionlab.org/policy-insight/ Health Affairs, 37, 525–534. https://doi. Value Across a National Community impact-price-take-and-use-preventive- org/10.1377/hlthaff.2017.1315 Health Center Network (ADVANCE). health-products 4. World Bank. (2016). World development Retrieved from http://pcornet.org/ 24. Durham Health Innovations. (n.d.). report 2016: Digital dividends. https:// clinical-data-research-networks/ History. Retrieved from https://sites. doi.org/10.1596/978-1-4648-0671-1 cdrn10-oregon-community-health- duke.edu/durhamhealthinnovations/ 5. BBC News. (2018, August 24). information-network-ochin/ history Russia trolls “spreading vaccination 15. GBD 2016 Causes of Death 25. Baker, J. P. (2017, September 19). A misinformation” to spread discord. Collaborators. (2017). Global, regional, tale of two cities? The history of Duke Retrieved from http://www.bbc.com/ and national age-sex specific mortality Health and Durham’s health [Video file]. news/world-us-canada-45294192 for 264 causes of death, 1980–2016: Retrieved from https://dukemed-source. 6. Adamson, A. S., & Smith, A. (2018). A systematic analysis for the Global mediasite.com/Mediasite/Play/6b89b77 Machine learning and health care Burden of Disease Study 2016. Lancet, ef0fa465daa848b6e139653ce1d 390, 1151–1210. disparities in dermatology. JAMA 26. Solon, O. (2018, April 4). Facebook Dermatology. Advance online 16. Murray, C. J. L., Kulkarni, S. C., Michaud, says Cambridge Analytica may have publication. https://doi.org/10.1001/ C., Tomijima, N., Bulzacchelli, M. T., gained 37M more users’ data. The jamadermatol.2018.2348 Iandiorio, T. J., & Ezzati, M. (2006). Guardian. Retrieved from https://www. 7. National Human Genome Research Eight Americas: Investigating mortality theguardian.com/technology/2018/ Institute. (2016). The cost of sequencing disparities across races, counties, and apr/04/facebook-cambridge-analytica- a human genome. Retrieved from race-counties in the United States. PLoS user-data-latest-more-than-thought https://www.genome.gov/27565109/ Medicine, 3, 1513–1524. https://doi. org/10.1371/journal.pmed.0030260 27. Vosoughi, S., Roy, D., & Aral, S. (2018, the-cost-of-sequencing-a-human- March 9). The spread of true and false genome/ 17. New York City Department of Health news online. Science, 359, 1146–1151. and Mental Hygiene. (n.d.). New 8. Institute of Medicine. (2014). 28. Faden, R. R., Kass, N. E., Goodman, Capturing social and behavioral York City community health profiles. Retrieved from https://www1.nyc.gov/ S. N., Pronovost, P., Tunis, S., & domains in electronic health records: Beauchamp, T. L. (2013). An ethics Phase 1. Retrieved from http:// site/doh/data/data-publications/profiles. page framework for a learning health care www.nationalacademies.org/hmd/ system: A departure from traditional Reports/2014/Capturing-Social-and- 18. Robert Wood Johnson Foundation. research ethics and clinical ethics. Behavioral-Domains-in-Electronic- (n.d.). County health rankings & The Hastings Center Report, 43(Suppl. Health-Records-Phase-1.aspx roadmaps. Retrieved from https:// 1), S16–S27. https://doi.org/10.1002/ 9. Institute of Medicine. (2014). www.rwjf.org/en/how-we-work/ hast.134 Capturing social and behavioral grants-explorer/featured-programs/ domains in electronic health records: county-health-ranking-roadmap.html Phase 2. Retrieved from http:// 19. Volerman, A., Chin, M. H., & Press, V. G. www.nationalacademies.org/hmd/ (2017). Solutions for asthma disparities. Reports/2014/EHRdomains2.aspx Pediatrics, 139, 1–9. https://doi. 10. Insel, T. R. (2017). Digital phenotyping: org/10.1542/peds.2016-2546 Technology for a new science of 20. Centers for Disease Control and behavior. JAMA, 318, 1215–1216. Prevention. (2014). National diabetes 11. Chetty, R., Stepner, M., Abraham, S., Lin, statistics report: Estimates of diabetes S., Scuderi, B., Turner, N., . . . Cutler, D. and its burden in the United States, (2016). The association between income 2014. Retrieved April 2, 2015, from and life expectancy in the United States, https://www.cdc.gov/diabetes/pdfs/ data/2014-report-estimates-of-

a publication of the behavioral science & policy association 37

report Using pay-for-success financing for supportive housing interventions: Promise & challenges Paula M. Lantz & Samantha Iovan

abstract § Pay for success (PFS) is an emerging public–private partnership strategy for providing housing to chronically homeless individuals, people with mental or behavioral disorders, and adults recently released from prison. Socially minded private investors from both for-profit and nonprofit organizations provide the up-front funding for the projects. If an independent evaluation demonstrates that the intervention achieved predetermined metrics of success—such as decreasing the number of days children spend in foster care or increasing the number of people with stable housing—the public sector then “pays for success” by repaying the private investors, sometimes with interest. In this article, we describe seven ongoing PFS housing projects in the United States. Most are “housing first” interventions that provide permanent supportive housing to a chronically homeless population without setting any preconditions, such as sobriety. As projects are completed, analyses of the results should provide further insights into the complexities of designing behavioral- based PFS housing programs.

Lantz, P. M., & Iovan, S. (2018). Using pay-for-success financing for supportive housing interventions: Promise & challenges. Behavioral Science & Policy, 4(1), 39–49.

a publication of the behavioral science & policy association 39 afe and affordable housing is important for funds are spent only if specific predetermined, the health and well-being of individuals contracted outcomes are achieved. PFS can Sand communities. Unaffordable housing be used to finance interventions and services puts significant economic pressure on individ- that provide value to the public sector. PFS can uals and families, forcing them to make stressful also be used to conduct a proof-of-concept trade-offs between, for example, paying rent demonstration of a potentially cost-effective or w and buying food, paying utilities, or making cost-saving intervention, which in turn might investments in their children. Unaffordable convince government leaders to directly fund Core Findings housing also leads to eviction and home- such a program in the future.6 lessness. A strong body of social science and What is the issue? epidemiological research has demonstrated the Second, the PFS financing model addresses Pay for success (PFS), positive health effects of housing interventions the difficulty of investing in preventive inter- or social impact bonds, targeting low-income and vulnerable groups, are an innovative way to ventions with long-term impact when acute improve social outcomes including the chronically homeless, individuals public needs urgently require funding. By using in high-risk communities. with mental and behavioral health disorders, private sector capital for up-front financing, Private sector financing and adults recently released from prison.1 Some is used to implement PFS allows governments to plan for potential proven interventions and of these interventions have also been shown to future payouts that are based on the terms of services, which is repaid reduce the amount of money the public sector a performance-based contract.7 This kind of by the public sector only spends on high-need populations (those who when contractual targets funding is politically attractive to taxpayers— typically have multiple complex medical and for desired outcomes especially when the interventions are aimed at have been met. social needs and a higher likelihood of chronic socially marginalized or perceived “undeserving” The early results from homelessness), primarily from reductions in ­populations, such as the homeless—because seven PFS projects that expensive medical, emergency, and criminal provide supportive public funds are not used unless the projects 2 housing for chronically justice services. succeed. homeless people are promising illustrations of An emerging funding strategy for social welfare the PFS financing model. PFS projects are challenging to establish and interventions is called pay for success (PFS), launch. In addition, most PFS projects world- How can you act? in which governmental and socially minded wide are still in progress. As such, it is not yet Selected recommendations private entities (for profit or nonprofit) partner possible to draw conclusions about the impact include: to finance and implement such interventions. 1) Building interest and of PFS projects on social welfare. However, a Metrics for success are laid out by contract in capacity across local comprehensive review of the 82 PFS projects and state government advance. The private investors initially pay for launched globally through 2017 revealed that all agencies to coordinate PFS the program. Then, if a third-party evaluation project buy-in, oversight, of them addressed at least one social determi- demonstrates that an intervention has met and measurement nant of health, with the majority implementing 2) Incentivizing private the contractual criteria for success, the public educational, behavioral, and psychosocial partners to reinvest sector “pays for success” by repaying the private interventions, including 21 aimed at housing.8 success payments back investors, sometimes with interest.3 The first into the PFS project PFS project was implemented in 2010 in the Through 2017, the PFS financing model garnered more than $390,000,000 of private Who should take United Kingdom to reduce criminal recidivism sector capital for the delivery and evaluation of the lead? through social and behavioral case manage- social welfare interventions, primarily in under- Advocates, policymakers, ment services.4 Since then, more than 100 PFS served populations.5 government officials, projects (also known as social impact bonds) private investors and stakeholders in have been launched or are being planned housing and health worldwide.5 In this article, we describe several PFS programs in the United States that focus on an interven- The PFS financing model is designed to address tion known as permanent supportive housing. two well-known challenges in public adminis- We also examine the strengths and challenges tration. The first is government waste, real and of the PFS approach to supportive housing in an perceived. The results-oriented PFS model can effort to glean insights into improving those and reduce inefficiencies and waste because public other PFS programs.

40 behavioral science & policy | volume 4 issue 1 2018 Permanent Supportive Housing Interventions “permanent supportive Used in PFS Programs Permanent supportive housing is a broad term housing can be a cost-saving used to describe certain housing interventions aimed at high-risk, high-need populations. intervention in high-need These interventions provide long-term housing linked to support services, which are delivered populations” on site or in the community and are meant to improve health and housing stability. Such supports typically include mental and behavioral independent housing, vulnerable individuals health care, family interventions, social welfare should be better able to sustain housing in the 12 services, and legal aid. For instance, a woman long term. with a drug dependency problem could receive counseling and support on site and assistance Another way housing first interventions incor- in connecting with other medical and social porate supportive services is through assertive services in the community. Extensive research community treatment (ACT). This model of has shown that permanent supportive housing intensive case management includes 24-hour, can be a cost-saving intervention in high-need seven-day-a-week access to individualized care populations such as the chronically homeless and services. ACT provides intensive support or adults recently released from prison.9 These services that are normally available only in interventions are largely based on theory and inpatient treatment settings. ACT has a strong research from the fields of health behavior evidence base behind its ability to provide change and social psychology. intensive case management, crisis interven- tion, substance use counseling, mental health Housing first is a specific type of permanent treatment, and primary care referrals. Although supportive housing program that connects indi- originally developed to serve individuals with viduals to long-term housing without any sort of severe mental illness, ACT has been adapted for precondition, such as sobriety or participation and evaluated in a variety of populations.13 in treatment or services.10 In other words, the approach ensures that individuals have safe and reliable housing before they attempt to address The Analysis their social or behavioral challenges. Research We designed and implemented a PFS surveil- shows that supportive services are more effec- lance system in 2016, through which we tive when individuals choose to participate—as continuously collect and analyze informa- is more likely when no preconditions are set tion on PFS projects that have been launched for the receipt of housing—rather than being around the world. This information includes required to do so.11 details about the design features, interventions (including the evidence base and relevance to Some housing first interventions use a critical population health), investors, governments time intervention (CTI) approach, in which indi- involved, metrics of success, payout terms and viduals receive case management services to other contractual elements, evaluation features, assist with the major adjustment that occurs outcomes, and challenges. during a move into community housing after being homeless or incarcerated. A social We collect information only on projects that worker or other social services professional have officially launched (with a signed contract, pulls together and manages a tailored set of secured funding, and actual service delivery) services and resources to meet the individu- and those in which the back-end payer is a al’s needs over time. By receiving support and government entity. Although a number of other continuity of care throughout the transition to websites describe PFS activity, we go further

a publication of the behavioral science & policy association 41 by using descriptive project data to follow the PFS Supportive Housing Projects research, administrative, policy, and popula- in the United States tion impacts of PFS initiatives.14 In this article, As of May 1, 2018, 21 housing-related PFS proj- we use our PFS surveillance data to describe ects have been launched globally, of which key elements of the housing projects underway seven (33%) are in the United States.8 At least in the United States and to identify some of 11 additional PFS housing projects are in devel- the strengths and challenges of using the PFS opment in the United States, including projects financing model for supportive housing inter- funded through the U.S. Department of Housing ventions in low-income populations. and Urban Development’s Pay for Success Permanent Supportive Housing Demonstration This second aspect of our article includes initiative.17 Next, we summarize the seven estab- an assessment of whether PFS housing proj- lished PFS housing projects in the United States, ects generally meet established criteria for providing a comparison of the major compo- using PFS programs to improve social welfare, nents of each contract in Table 1. such as having a strong evidence base behind the chosen interventions.15 Many resources, Partnering for Family Success, Cuyahoga including the Urban Institute’s Project Assess- County, Ohio. The Partnering for Family ment Tool, provide guidance for developing Success project aims to reduce the number of successful PFS projects.16 Projects that are most days in foster care for children whose caregivers likely to succeed should meet the following are homeless.18 In 2014, the year this program criteria: launched, Cuyahoga County budgeted more than $50 million for foster care.19 By providing • The intervention must address a problem of homeless parents with stable housing, the interest to the public sector. county hopes to improve the well-being of homeless families while also saving money. In • The intervention should have published addition to a housing first intervention, clients evidence of effectiveness in a clearly identi- receive CTI, trauma-adapted family connec- fied population. tions, and child–parent psychotherapy, three psychosocial interventions aimed at improving • The intervention should provide economic relationships within families, taking into account value to the public sector by being either the traumatic context of their current or past cost-effective or cost saving. situations.20,21

• Outcomes must be clearly defined and This project will serve 135 homeless families measurable. over a five-year period, with the aim of reducing foster care placement days by 25%. If that is • Outcomes must be achievable in a reason- achieved, investors will receive full repayment of able time period. their investment. If the target is exceeded, inves- tors will be repaid with interest. Investors have • Outcomes must be achievable without stated that they plan to reinvest any success significant administrative, political, or stake- payments back into the program, which will holder challenges, such as objections from provide long-term funding and sustainability. local leadership, project partners, or the community.15 Chronic Homelessness PFS Initiative, Massa- chusetts. To address the issue of homelessness and the costly use of public services, Massachu- Our description and analysis of PFS housing setts launched a PFS project in 2015 to deliver interventions should be useful for government the Home & Healthy for Good (HHG) program leaders and socially minded investors who are to chronically homeless individuals.22 Using exploring potential PFS initiatives in and beyond a housing first approach to address the high supportive housing. usage of emergency services by chronically

42 behavioral science & policy | volume 4 issue 1 2018 Table 1. Key components of initiatives related to pay-for-success supportive housing in the United States, May 2018

Project Chronic Project Partnering for Homelessness Welcome Housing to Variable Family Success PFS Initiative Home Health Homes Not Jail REACH Just in Reach Government Cuyahoga Commonwealth Santa Clara Denver, Salt Lake Salt Lake Los Angeles County, Ohio of County, Colorado County, Utah County, Utah County, Massachusetts California California Duration 5 years 5 years 6 years 5 years 6 years 6 years 4 years Total investment $4M $3.5M $6.9M $8.6M $5.3M $5.4M $10M Investors George Gund Santander Bank, Reinvestment Northern Trust, Northern Trust, Northern Trust, United Foundation, United Way of Fund, Walton Family Ally Bank, QBE Ally Bank, QBE Healthcare, Cleveland Massachusetts Corporation Foundation, Insurance, Insurance, Conrad N. Hilton Foundation, Bay and for Supportive Piton Reinvestment Reinvestment Foundation Sisters of Charity Merrimack Housing, Foundation, Fund, Sorenson Fund, Sorenson Foundation of Valley, Sobrato Family Laura and Impact Impact Cleveland, Laura Corporation Foundation, John Arnold Foundation Foundation and John Arnold for Supportive California Foundation, Foundation, Housing Endowment, Living Cities, Reinvestment Health Trust, Nonprofit Fund, Nonprofit James Irvine Finance Fund Finance Fund Foundation, Google.org Service delivery FrontLine Massachusetts Adobe Services Colorado The Road Home First Step House L.A. County organization(s) Service Housing and Coalition for Department of Shelter Alliance the Homeless, Health Services, Mental Health Brilliant Corners Center of Denver, Colorado Access Other housing Enterprise N/A N/A Enterprise N/A N/A Corporation organizations Community Community for Supportive Partners Partners, Housing Corporation for Supportive Housing

Intervention Housing First, Home & Healthy Housing First, Housing First, Rapid Risk–Needs– Housing First Critical Time for Good, Assertive Assertive Re-Housing Responsivity supportive Intervention, Housing First Community Community (Housing First Model housing Trauma supportive Treatment Treatment supportive Adapted Family housing housing), Connections, trauma- child-parent informed care, psychotherapy motivational interviewing

Target Homeless Chronically Chronically Chronically Chronically Formerly Chronically population caregivers with homeless homeless homeless homeless with incarcerated homeless children in foster substance use adults care disorders Success Reduction in 12 months of 3 months of 365 total Improvement in Reduction 6 months and metric(s) foster care days continuous continuous adjusted days the number of in days 12 months in housing stable tenancy in housing, months without incarcerated, stable housing, reduction in jail being in jail reduction reduction in bed days or the shelter, in statewide arrests graduation to arrests, increase permanent in employment, housing, program substance engagement abuse treatment enrollment, mental health treatment enrollment

a publication of the behavioral science & policy association 43 homeless individuals with complex needs, HHG justice system resources. This project combines was created in 2006 to provide health, social, housing first and the ACT model of intensive case and behavioral support after individuals are management to address a wide range of social placed into housing. Since its inception, HHG and behavioral needs. Project Welcome Home has assisted 973 formerly homeless individuals will ultimately serve 150–200 chronically home- with permanent supportive housing in Massa- less individuals over the course of the six-year chusetts, resulting in an average annual savings project. Success payments will initiate when a of $12,428 per tenant housed, according to a participant reaches a minimum of three months state Medicaid analysis.23 of continuous stable tenancy. The goal of Project Welcome Home is for 80% of participants to The Chronic Homelessness PFS Initiative achieve 12 months of continuous tenancy. represents a scaling of the HHG services already delivered by the Massachusetts state govern- Housing to Health, Denver, Colorado. The ment. Over the five-year PFS project period, Denver Housing to Health initiative was supportive housing will be provided to approx- launched in 2016 to address the high use of imately 800 chronically homeless individuals. expensive city and county safety-net services Repayment to private investors depends on by chronically homeless individuals.25 Housing participants achieving housing stability for 12 to Health is using a housing first approach to months. If 80% of individuals meet this mile- provide 250 residential units to chronically stone, investors will receive full repayment of homeless individuals over the five-year project their principal investment from the state. If period. As in Project Welcome Home, service more than 80% of project participants achieve providers are using ACT intensive case manage- 12 months of continuous housing, investors ment to provide supportive services to enrollees. will be repaid with interest. First-year outcomes revealed a housing retention rate of 92%, The Housing to Health initiative is being eval- resulting in an interim repayment to investors uated with respect to two outcomes: housing (see Table 2 for additional reported results). stability and jail days. Housing stability payments will be calculated on the basis of total adjusted Project Welcome Home, Santa Clara County, days in housing for each individual who reaches California. Project Welcome Home was a threshold of at least 365 days housed in the launched in 2015 to provide supportive housing community. Jail reduction payments are based to chronically homeless individuals living in on the reduction of jail days in the interven- Santa Clara County.24 The project targets adults tion participants, with a minimum threshold identified as high-cost users of county services of a greater than 20% reduction in jail days like emergency care, inpatient care, and criminal compared with a control group.

Table 2. Interim payout & results data from the Chronic Homelessness PFS Initiative in Massachusetts Project Feature Characteristic or result Project launch date June 2015 Total investment $2,500,000 Investors (amount invested) Santander Bank ($1,000,000) United Way of Massachusetts Bay and Merrimack Valley ($1,000,000) Corporation for Supportive Housing ($500,000) First success payments to investors announced February 2018 Participants housed (through 2/2018) 656 Participants meeting success metric (through 2/2108) 92% of participants remained permanently housed 1 year after placement Success payments to investors (through 2/2018) Santander Bank ($102,200) United Way of Massachusetts Bay and Merrimack Valley ($102,200) Corporation for Supportive Housing ($51,000)

44 behavioral science & policy | volume 4 issue 1 2018 Homes Not Jail, Salt Lake County, Utah. receive individualized services such as short- The Homes Not Jail project was launched in term housing, case management, substance 2017 in Salt Lake County to serve persistently abuse treatment, mental health services, and homeless adults with substance use disorders. employment support. Homes Not Jail uses a housing first inter- vention called rapid rehousing that provides REACH will eventually serve approximately individuals with fast-paced move-in support, 225 formerly incarcerated individuals over the rental assistance, peer support, and financial six-year project. Success payments are deter- and case management services.26 Homes Not mined on the basis of four outcomes among Jail explicitly uses a harm-reduction approach, participants: reduction in the number of days allowing participants who are currently strug- incarcerated, reduction in the number of state- gling with substance abuse to obtain housing wide arrests, improvement in the number without any social or behavioral preconditions. of quarters of employment, and successful Motivational interviewing and trauma-informed program engagement. Any significant improve- care are also used to help participants make ment in the first three outcomes compared with positive behavioral and psychosocial changes. a control group will result in success payments As with the other PFS housing interventions, to investors. service delivery partners provide comprehen- $390m sive wraparound services to assist with lingering Just in Reach, Los Angeles County, California. PFS capital raised through social issues, such as food insecurity and Los Angeles County launched the Just in Reach the private sector in 2017 unemployment. PFS initiative in 2017 to reduce recidivism and end the cycle of homelessness among individ- Over the six-year project period, Homes Not uals with repeat county jail stays.29 This housing Jail will serve 315 persistently homeless indi- first program links chronically homeless indi- 25.6 viduals in Salt Lake County. Four outcomes will viduals to permanent supportive housing. Percentage of PFS serve as measures of success: months without Once participants enter stable housing, they projects worldwide staying in a shelter or jail, mental health service are provided with social, behavioral, and health aimed at housing participation, substance abuse service enroll- services, including mental health therapy, through 2017 ment, and graduation to permanent supportive substance abuse treatment, employment housing. Any significant improvement in the first services, connections to public benefits, and three measures relative to a control group will mentors. A 2008 demonstration project showed result in a payment. Payment for graduating to a significant decrease in the recidivism rate 33% permanent supportive housing is made for each for program participants compared with the Health Disparities participant who is living in permanent housing general jail population.30 US share of global PFS when discharged from the program. The project projects launched in 2018 goals are a 30% improvement for participants in The Just in Reach PFS initiative aims to serve the number of months without a stay in jail or 300 homeless individuals who are currently in a shelter and 80% of participants graduating to the county jail; have had prior jail stays; and permanent housing. have complex social or behavioral problems such as mental illness, substance use disorder, REACH, Salt Lake County, Utah. REACH or posttraumatic stress disorder. The four- (Recovery, Engagement, Assessment, Career, year project will make payments on the basis and Housing), launched in 2017, is a broad- of housing retention and jail avoidance rates. based intervention tailored specifically to the Housing retention payments will be made needs of formerly incarcerated adult men who for each participant who reaches six months are currently under the supervision of Utah Adult and then 12 months in stable housing. The Probation & Parole.26,27 The REACH program jail avoidance rate is based on the number of uses the risk–need–responsivity model, which ­re­arrests during the two years following entry takes into account the risk a person will reoffend into supportive housing, with success payments and his other specific social, behavioral, psycho- based on participants with two or fewer returns social, and structural needs.28 Participants to jail in a two-year follow-up period.

a publication of the behavioral science & policy association 45 All the PFS housing projects to date in the “Both nonprofit and for-profit United States address a problem of interest to the public sector by implementing cost-effec- investors have provided capital” tive and perhaps even cost-saving interventions that have a strong research evidence base in the Summary. Although the seven PFS housing target populations. What is more, the project projects have differences, they all include the outcomes are clear, measurable, and achievable delivery of evidence-based interventions to in a reasonable time period (four to six years) marginalized or vulnerable groups with complex and do not appear to have serious stakeholder needs. With the exception of REACH in Salt challenges. However, the administrative costs Lake County, all the projects use a housing of these projects are currently not well under- first approach combined with some variant of stood. In addition, the final outcomes from permanent or long-term supportive housing. these projects (including investor payouts) are They deliver a range of supportive services to not yet known. Only two projects thus far have address the complex psychosocial, behavioral, resulted in interim payouts to investors. Never- and medical needs of the target population. theless, these interventions, if implemented with Both nonprofit and for-profit investors have fidelity to the intervention research literature, provided capital, and key agencies and orga- should be able to achieve their objectives. nizations in the field of housing, including the Corporation for Supportive Housing, the It is important to note that in all seven PFS Reinvestment Fund, and Enterprise Commu- housing projects in the United States to date, nity Partners, have been involved in many of the payouts are contractually based on the the projects. Success payments to the private achievement of behavioral outcomes, such investors are contingent on some measure of as stable housing, treatment enrollment, and sustained housing in all but two projects (Part- lack of recidivism—not on evidence of public nering for Family Success in Cuyahoga County savings. This is a major strength of this approach and REACH in Salt Lake County). for making social progress.

The Big Picture Challenges. Even though the research literature The PFS financing model is being used by suggests that supportive housing interventions governments and private entities to support the save money that is often spent on high-risk dissemination of evidence-based interventions. populations, there are significant administra- The projects described here provide models tive and legal challenges to explicitly capturing for how this public–private financing approach public savings. These include the “wrong has been implemented in the important area of pockets” problem, in which the savings from housing. Private nonprofit and for-profit inves- a PFS initiative accrue across multiple govern- tors have demonstrated interest in investing in ment agencies and their budgets, which makes evidence-based supportive housing interven- it difficult to identify and capture savings for tions, bringing new sources of private revenue the purpose of repaying private investors. to address housing in high-risk, complex-need Furthermore, legal barriers can prohibit some populations. Although PFS is in the early stages government programs (such as the federal arm of development, evidence presented here and of Medicaid) from making success payments to elsewhere suggests it holds promise as a way to private investors.14 finance housing, a critical component of health and social equity.14 Additional challenges can complicate growing or scaling up the PFS financing model for Strengths. A clear strength of the seven PFS supportive housing. One is the need for the housing projects we have described is that they government to increase its interest and capacity meet the minimum criteria for interventions for engaging in PFS activity, which is a unique appropriate for PFS, as described earlier.15 Not all type of results-driven contracting. Local and launched PFS projects have met these criteria. state governments need capacity in a number

46 behavioral science & policy | volume 4 issue 1 2018 of key areas—including leadership buy-in, procurement policies, contract management, “A clear strength of the seven and the data systems for measuring outcomes and cost-effectiveness.6,31 PFS housing projects . . . is

Another challenge is that even though success that they meet the minimum payments do not depend on the government saving money, to be economically attractive criteria for interventions for PFS financing, housing interventions must target individuals at the highest risk of needing appropriate for PFS” expensive public services, such as chronically homeless populations with mental health, substance abuse, and other disabling problems. housing-related projects implemented to date Although such individuals are in obvious need provide excellent examples of how this financing of supportive housing, this focus on those at model can enable the spread of evidence-based highest risk is open to the criticism that such permanent supportive housing that improves programs neglect individuals and families housing stability and other outcomes. The long- who are also in need of stable, affordable, and term social, behavioral, and health impacts of supportive housing but who are not high users the PFS housing projects that are underway are of costly public services. As a public–private not yet known, but it does appear that PFS has partnership financing model, however, PFS is opened the door for evidence-based program best suited for interventions that provide signif- delivery to populations who may not otherwise icant economic efficiencies or savings to the be served via traditional funding mechanisms. public sector and thus are bound to target the Although the early results are promising, the outlying, highest need populations. final evaluations of these pioneering proj- ects will more fully reveal the potential of PFS Third, given the challenges that local and state financing for behavioral-based supportive governments face in funding expensive inter- housing and other social welfare interventions ventions like permanent supportive housing, in high-need populations. the long-term sustainability of these interven- tions depends on maintaining the enthusiasm of private and public sector participants. To sustain author affiliations an intervention, either the investors must be willing to reinvest success payments back into Lantz & Iovan: University of Michigan, Ford the project or the public sector must itself take School of Public Policy. Corresponding author’s over the financing and oversight of the inter- e-mail: [email protected]. vention. Although such interest is increasing among health care systems, Medicaid managed care organizations, public health agencies, and author note researchers, using Medicaid mechanisms to finance housing and other social and nonmed- This work was supported by an award from ical interventions related to health can be the Robert Wood Johnson Foundation (award problematic because of significant administra- 73217) for the University of Michigan Policies for tive obstacles.32 Action Research Hub.

Looking Ahead In summary, despite the challenges, PFS remains an important way to finance housing interventions in populations that are high users of government programs and services. PFS

a publication of the behavioral science & policy association 47 references

1. Gibson, M., Petticrew, M., Bambra, 10. HUD Exchange. (2014). Housing us/pdf_council/en-US/2014- C., Sowden, A. J., Wright, K. E., & First in permanent supportive 2015Budget/2014-2015%20 Whitehead, M. (2011). Housing and housing. Retrieved from https:// Recommended%20Budget_2.pdf health inequalities: A synthesis of www.hudexchange.info/resources/ 20. National Child Traumatic Stress systematic reviews of interventions documents/Housing-First-Permanent- Network. (2012). TA-FC: Trauma aimed at different pathways linking Supportive-Housing-Brief.pdf adapted family connections [Fact sheet]. housing and health. Health & Place, 17, 11. National Alliance to End Homelessness. Retrieved from https://www.nctsn.org/ 175–184. (2016). Housing First. Retrieved from sites/default/files/interventions/tafc_ 2. Ly, A., & Latimer, E. (2015). Housing First https://endhomelessness.org/resource/ fact_sheet.pdf impact on costs and associated cost housing-first/ 21. Bassuk, E. L., DeCandia, C. J., offsets: A review of the literature. The 12. Social Programs That Work. (2017). Tsertsvadze, A., & Richard, M. K. Canadian Journal of Psychiatry, 60, Evidence summary for the critical (2014). The effectiveness of housing 475–487. time intervention. Retrieved from interventions and housing and service 3. Galloway, I. (2014). Using pay-for- https://evidencebasedprograms.org/ interventions on ending family success to increase investment in the document/critical-time-intervention- homelessness: A systematic review. nonmedical determinants of health. evidence-summary/ American Journal of Orthopsychiatry, Health Affairs, 33, 1897–1904. 13. Phillips, S. D., Burns, B. J., Edgar, E. R., 84, 457–474. 4. Ministry of Justice. (2017). Peterborough Mueser, K. T., Linkins, K. W., Rosenheck, 22. Pulster, R. (2015). Massachusetts social impact bond: Background. R. A., . . . McDonel Herr, E. C. (2001). launches pay for success initiative Retrieved from https://assets.publishing. Moving assertive community treatment to address chronic homelessness. service.gov.uk/government/uploads/ into standard practice. Psychiatric Retrieved from https://www.usich.gov/ system/uploads/attachment_data/ Services, 52, 771–779. news/massachusetts-launches-pay-for- file/633271/peterborough-social- 14. Lantz, P. M., Rosenbaum, S., Ku, L., & success-initiative-to-address-chronic- impact-bond-background-information. Iovan, S. (2016). Pay for success and homele pdf population health: Early results from 23. Massachusetts Housing and Shelter 5. Social Finance. (n.d.). Impact bond eleven projects reveal challenges and Alliance. (2018). Permanent supportive global database. Retrieved from https:// promise. Health Affairs, 35, 2053–2061. housing: A solution-driven model. sibdatabase.socialfinance.org.uk/ 15. Lantz, P. M., & Iovan, S. (2017, December March 2018 Home & Healthy for Good 6. Whistler, C. (2017). The next phase 12). When does pay-for-success make progress report. Retrieved from https:// of pay for success: Driving public sense? Stanford Social Innovation www.mhsa.net/sites/default/files/ sector outcomes. Retrieved from Review. Retrieved from https://ssir.org/ March%202018%20HHG%20Report.pdf https://www.thirdsectorcap.org/blog/ articles/entry/when_does_pay_for_ 24. County of Santa Clara. (2017). County the-next-phase-of-pay-for-success- success_make_sense of Santa Clara launches California’s driving-public-sector-outcomes/ 16. Milner, J., Eldridge, M., Walsh, K., & first “pay for success” project [Press 7. Carrillo, O. (2017). Pay for success: Roman, J. K. (2016). Pay for success release]. Retrieved from https:// Opportunities and challenges in project assessment tool. Retrieved www.sccgov.org/sites/opa/nr/Pages/ housing and economic development. from https://www.urban.org/sites/ ProjectWelcomeHome.aspx Retrieved from http://www.jchs.harvard. default/files/publication/85391/pay-for- 25. Corporation for Supportive Housing. edu/sites/default/files/carrillo_pay_for_ success-project-assessment-tool_1.pdf (2016). Fact sheet: Denver social impact success_gramlich_2017.pdf 17. U.S. Department of Housing and bond program to address homelessness. 8. Iovan, S., Lantz, P. M., & Shapiro, S. Urban Development. (2015). Pay for Retrieved from https://www.csh.org/ (2018). “Pay for success” projects: Success Permanent Supportive Housing wp-content/uploads/2011/12/Denver- Financing interventions that address Demonstration [Fact sheet]. Retrieved SIB-FactSheet.pdf social determinants of health in 20 from https://www.hudexchange. 26. Salt Lake County Council. (2017). Pay countries. American Journal of Public info/resources/documents/ for success reports: December 2017 Health, 108, 1473–1477. PFS-Demonstration-Fact-Sheet.pdf update. Retrieved from https://slco.org/ 9. Office of the Assistant Secretary for 18. Enterprise Community Partners & council/pay-for-success-reports/ Planning and Evaluation. (2014). 3.1. Third Sector Capital Partners. (2016). 27. Salt Lake County. (2016). Fact sheet: Salt Permanent supportive housing and Developing the Cuyahoga Partnering Lake County pay for success initiatives. services for people experiencing for Family Success Program. Retrieved Retrieved from https://slco.org/ chronic homelessness. In A primer on from https://www.thirdsectorcap. uploadedFiles/depot/fMayor/mayor16/ using Medicaid for people experiencing org/wp-content/uploads/2016/02/ PFS_launch_Factsheet.pdf chronic homelessness and tenants Final-Cuyahoga-Partnering-for-Family- 28. Bonta, J., & Andrews, D. A. (2007). in permanent supportive housing. Success-Program-Lessons-Learned- Risk-need-responsivity model for Retrieved from https://aspe.hhs.gov/ Report.pdf offender assessment and rehabilitation. report/primer-using-medicaid-people- 19. FitzGerald, E. (2013). Cuyahoga Rehabilitation, 6, 1–22. experiencing-chronic-homelessness- County, Ohio 2014–2015 29. Corporation for Supportive and-tenants-permanent-supportive- executive’s recommended housing Housing. (2017). “Just in Reach” biennial budget. Retrieved from supportive housing. Retrieved from http://council.cuyahogacounty. https://www.csh.org/2017/10/ just-in-reach-supportive-housing/

48 behavioral science & policy | volume 4 issue 1 2018 30. Los Angeles County Sheriff’s Department, Community Transition Unit. (YEAR). Just in Reach (JIR) Services for Homeless Inmates. Retrieved from http://www.naco.org/sites/default/ files/documents/Session%204%20 -%20Los%20Angeles%20County%20 Sheriff’s%20Department.pdf 31. Harvard Kennedy School Government Performance Lab. (2016). Results-driven contracting: An overview. Retrieved from http://govlab.hks.harvard.edu/files/ siblab/files/results-driven_contracting_ an_overview_0.pdf 32. McGinnis, T., Bonney, J., & Chang, D. I. (2018). Executive summary: Innovative Medicaid payment strategies for upstream prevention and population health. Retrieved from http://www. chcs.org/media/Innovative-Medicaid- Payment-Strategies-for-Upstream- Prevention-and-Population-Health.pdf

a publication of the behavioral science & policy association 49

essay Improving the match between patients’ needs & end-of-life care by increasing patient choice in Medicare Donald H. Taylor, Jr.

abstract ¶ One way to achieve health equity—ensuring everyone has fair and just opportunities to be as healthy as possible—in the United States would be to reallocate Medicare spending from low-value medical care (expensive treatments that do little good) toward high-value medical and social care (respectively, medical interventions that have been shown to work well but are not covered by Medicare and nonmedical interventions, such as help with activities of daily living, that patients find more helpful than low-value care). In the current policy milieu, the most practical, direct step in that direction may be for Medicare—an already established, universal health care program for the elderly—to provide patients with more choices and autonomy.

Taylor, D. H., Jr. (2018). Improving the match between patients’ needs & end-of-life care by increasing patient choice. Behavioral Science & Policy, 4(1), 51–61.

a publication of the behavioral science & policy association 51 ell-documented inequalities in health conducted by Elizabeth H. Bradley and Lauren insurance coverage, access to care, Taylor of Yale University3 shows that lack of Wand population health clearly show investment in social services that affect health— that the United States has far to go to reach such as education, income support, housing, anything approaching health equity—a goal the nutrition, and child care—explains a substan- Robert Wood Johnson Foundation has defined tial portion of the nation’s poor health in spite as everyone having “a fair and just opportunity of its high health care spending. What Taylor to be as healthy as possible.”1 What are the best and Bradley have called “the national invest- w approaches for moving toward health equity? ment in health”—the combined money devoted to health care and social services—is merely Core Findings In this article, I propose that the most direct average compared with that of the other nations and far-reaching action that might be achiev- in the Organisation for Economic Co-operation What is the issue? able in the current political climate would be and Development. A middling national invest- U.S. health outcomes lag for Medicare, which offers medical coverage ment in health yields middling health outcomes. behind those of other to everyone 65 years of age and older, to shift developed countries 3 despite high levels of health away from primarily covering and promoting Bradley and Taylor concluded that policy- care investment. Public medical care services near the end of life that makers concerned with health equity should spending investments often turn out to be low value, and instead move broaden their focus beyond simply expanding through Medicare tend to cover low-value end-of- toward enabling patients to receive high-value access to health insurance and should work life medical care and not medical care and social services paid for by to expand social interventions. The important broader high-value care Medicare. By low-value medical care services, effects of social factors on health are well docu- alternatives that include I mean expensive medical interventions that mented and may explain over half the variation social service options. An evidence-based shift do little good, such as delivering last-ditch in observed health outcomes between nations toward these options can chemotherapy to a cancer patient who has little and between groups within nations.4–6 serve both efficiency and chance of responding and who is more likely to equity in health outcomes. be harmed by side effects than helped. By high- In spite of ample evidence that health outcomes How can you act? value medical care and social services, I mean are influenced by many factors, for the last Selected recommendations medical care that has been shown to work well decade, the health policy focus in Washington, include: but that is not directly financed by Medicare, DC, has primarily centered on passing (or 1) Identifying specific medical conditions that such as comfort-focused palliative care given opposing) and implementing (or sabotaging) frequently result in low- before a patient elects to receive hospice care the Affordable Care Act (ACA), which prioritized value care provision and forgo curative therapy, and nonmedical expanding insurance coverage to nonelderly 2) Testing the efficacy services, such as meal preparation or help with individuals who lacked access to employer-­ of covering lump sums as an end-of-life health transportation to doctors’ offices, that tend to sponsored health insurance, a relatively small care alternative, to be less expensive than medical care and are slice of the overall population. The controversy encourage use of social more predictably beneficial to elderly persons generated by this fairly narrow reform, which services like home nursing and transportation across many health circumstances. was nevertheless the most comprehensive since for doctor visits the creation of Medicare and Medicaid in 1965, The need for changes in health-related spending demonstrates how difficult large-scale efforts to Who should take is undeniable. The United States invests a great disrupt the status quo can be. the lead? deal in health care: the nation’s expenditure on Researchers, policymakers, and stakeholders health care, which represents around half of the One could imagine an alternative policy initiative in health care country’s total spending, is approximately equiv- that invests the same magnitude of resources alent to the combined governmental and private into social services for children, for example. The spending in most high-income nations. Yet the ACA was financed by a mix of cuts in reimburse- United States has only middling population-­level ment to the Medicare program and increased health outcomes.2 This pattern has often been taxes. If the same money were instead invested viewed as evidence that inequality in access to in social services, funding better education and and use of care leads to poor outcomes, but housing for low-income children, the allocation that is not the whole explanation. Research would move the nation’s investment in health

52 behavioral science & policy | volume 4 issue 1 2018 in a direction that the social science and public The kinds of changes I am recommending health literatures suggest is conducive to better could apply, for instance, to an elderly person societal-level health outcomes,6–8 such as more suffering from advanced heart failure, which has children going on to earn good incomes and no clear medical therapy to cure the disease, or living in healthier conditions. to a patient with lung cancer who has already tried the existing chemotherapy and radiation A shift from spending on the elderly toward treatments. There is almost always something spending on children might be expected to have else to try medically, but I am proposing to allow the biggest bang for the buck in moving the U.S. patients to decide when they have had enough population toward health equity. Such a shift medical care that is not working and to instead would, of course, be politically impossible in use their Medicare coverage to pay for other the United States today, where the elderly advo- types of care or social services that would be cate powerfully for the health care complex that more likely to improve their quality of life. provides them with care and where the govern- ment is currently inclined to cut social spending. A change in Medicare policy that reallocated money within the program to make changes Yet shifting expenditures within the Medicare driven by patient choices might be more palat- program from the kind of health care that is able to policymakers than other proposals for often delivered near the end of a person’s life improving health outcomes in the United States to other medical care and social services would because it would not require added funding or probably be more politically feasible and would creating a new program. The approach would better meet the needs of many, as I argue in benefit many elderly patients—a growing this article. I also describe ways to determine segment of the population—and potentially which services people prefer and to discover reduce health inequity between disadvantaged whether shifting Medicare coverage in this and advantaged senior citizens. For instance, direction would, in fact, improve outcomes shifting resources from low-value medical care while increasing the autonomy and participation to social services in a program that already of the elderly in determining the best ways to covers everyone after they reach the age of address their illnesses and disabilities. 65 years could help to compensate for long- standing sources of inequity, including race, income, education, and rural residence, in that Why Shifting Away From age group. No similar universal insurance struc- Low-Value Medical Care ture exists for younger persons. In addition, Near the End of Life given that much spending by Medicare near the Makes Policy Sense end of life is of questionable value, the approach One reason to focus on care delivered near the has the potential to reallocate some program end of life is that the United States overspends spending without the change being detrimental on low-value care at that time, as abundant to one group while benefiting another. evidence indicates. Since 1970, one in four Medicare dollars has been spent during the last The proposal has another benefit as well: if year of a Medicare beneficiary’s life.9 Yet the evidence-based reallocation of low-value expensive care that is provided in a person’s medical spending to high-value social spending last days, weeks, or months often does not could be achieved in Medicare by enabling extend life or improve other health outcomes patients to play a larger role in determining their and may even harm patients.10,11 Many families own care, that accomplishment could catalyze experience regret over care choices made for considerations of similar reallocations in other loved ones just before death, and studies have programs that could improve health equity. documented posttraumatic stress disorder in survivors who witness a loved one die in an There is a problem with directing a policy intensive care unit.12–14 toward the end of life: the “end of life” concept is

a publication of the behavioral science & policy association 53 Why End-of-Life Care Has “most people in a given Been Hard to Change One part of the two-sided learning problem clinical situation will not standing in the way of better end-of-life care is summarized by the truism “your mother benefit from last-ditch only dies once.” That is, after a loved one dies, family members and other caregivers who medical treatment” learned how to navigate health care decisions for the patient often do nothing with their hard-won wisdom. There are no clear feed- inherently retrospective. In other words, you do back mechanisms through which they can not know when the last year of life started until share knowledge with those who are beginning it ends. Predicting death involves a great deal of the same journey, and so a wealth of practical uncertainty, even for very sick elderly patients, knowledge is lost. and so it it is impossible to design policies that specifically address the last year or months of The second part of the problem is the converse patients’ lives prospectively, which is the only of the first: the health care system copes way to change observed spending patterns. repeatedly with people near the end of their Indeed, physicians often do not know how long life (after all, everyone dies!), and providers can a person will survive or whether a given inter- see after the fact that much of a patient’s last vention is futile. As Lisa Rosenbaum of Brigham year of treatment was useless or harmful. But and Women’s Hospital noted in a recent essay15 the retrospective knowledge that low-value that pushed back against what she termed the care is common at life’s end does not typically “less-is-more crusade” in treatment, “sometimes get translated into an effective, evidence-based less is more, sometimes more is more, and often strategy for changing treatment and spending we just don’t know.” patterns near the end of life, for a variety of reasons. For instance, a multifaceted inertia A recent analysis of Medicare claims data favors the systematic, aggressive provision of supports quantitatively Rosenbaum’s caution care, much of which is understood in retrospect about the difficulty of predicting who will die, to have provided little benefit. even among seriously ill elderly persons.16 Standing in the way of reduced low-value The provider’s dilemma—how to decide what health spending are existing systemwide to do in the face of uncertainty about, on the financial incentives that favor delivering more one hand, any given individual’s prognosis and, treatment—incentives that align well with the on the other hand, reasonable evidence that professional ethos in American medicine that most people in a given clinical situation will not more is better. (In Rosenbaum’s essay,15 she benefit from last-ditch medical treatment—can suggested that professional norms and a desire be addressed in part by providing better infor- for certainty—which can prompt excessive mation and additional care options to patients testing and multiple follow-up procedures— who are afforded the autonomy to make their may actually be more influential than financial own decisions with the best information avail- gain in driving the delivery of much care that is able. This approach is also the most plausible later recognized to have been of low value.) The way to address a common two-sided learning United States’ complex incentive structure did problem that contributes to the perpetuation not form in a vacuum, and it is not surprising of Medicare-funded low-value-care delivery. that health care providers in a culture that uses In the balance of this article, I outline a process military metaphors for health problems (“We for addressing such problems, one that keeps will wage a war on cancer”; “She lost her fight”) research evidence and patient preferences at assume that patients and their families want all the fore of attempts to reform the system. illnesses treated aggressively.

54 behavioral science & policy | volume 4 issue 1 2018 The behavioral economics and social of daily living. More radical options could also psychology literatures have detailed factors that be imagined, such as giving cash to patients interfere with individuals’ ability to make more who forgo care that is understood to be of low cost-effective end-of-life health care decisions, value; the money can then be used for whatever particularly well-known behavioral biases that purpose they choose. Right now, hospice care is limit people’s ability to make rational decisions. limited to cases in which physicians certify that First, when there is a possibility, however slight, a patient is likely to die within six months; such of a miracle recovery, hope springs eternal. limitations could be relaxed, allowing patients to According to prospect theory,17,18 people tend choose to receive palliative care earlier in their to overweight low-probability events (which disease course, without first having to cease explains why they pay a premium for both lottery curative care. tickets and expensive insurance coverage), and they do so especially in emotionally charged Before instituting specific plans along these situations, such as when they are judging the lines, Medicare will need to perform careful pilot potential for recovery from an illness that has tests, and monitoring will be essential to ensure been deemed terminal.19 that patients and family caregivers understand the options offered and the choices they make. Second, people tend to give undue weight to But a study called CHAT (Choosing Health Plans 25% outcomes in the very near term, such as the All Together) that I conducted at Duke Univer- Medicare dollars spent during the last year of possibility of keeping a loved one alive just a sity with several colleagues already supports the a beneficiary’s life little bit longer, and to drastically discount future notion that patients would appreciate adjust- outcomes. They tend, for instance, to under- ments in what Medicare will cover and that value the years of financial misery that may seriously ill patients are able to engage in diffi- result from this decision or the regret that they cult trade-offs, especially when they are able to may feel about the poor quality of life a loved talk about them with other patients. We found 9% 22 one experienced during their weeks, months, evidence that Medicare beneficiaries with Health Disparities or years of extended life.20 Third, most people advanced cancer and their family members or Six-month survival rate find even thinking about sacrificing life out of other caregivers would be willing to forgo last- in patients diagnosed with platinum-resistant financial concern terribly unpleasant—people ditch cancer treatments that are often judged ovarian cancer tend to avoid even contemplating making trade- retrospectively to be of low value in return for offs between sacred values, such as human life, having the flexibility to receive “high-touch, and secular values, such as money, when the low-tech” care designed to improve quality decision involves a particular individual who is of life. In the cancer setting, last-ditch care “infinitely important.”21 typically means experimental chemotherapy, 38% whereas high-touch, low-tech care could take Point drop in adult the form of hospice-like services or social care smokers since 1950 such as a nurse’s aide who can help an elderly A Strategy for Moving Away person with activities of daily living instead of a From Low-Value Care long-shot bid for a miracle cure. As I noted earlier, allowing Medicare patients who are well-informed about their care options The CHAT study provided theoretical choices to refuse last-ditch medical care in return for to patient participants,22 who were essentially reimbursement of medical and social services given a budget and asked to select multiple not currently covered by the Medicare program’s care options from a list of 15 benefit catego- benefit package could improve the value that ries, including three options that Medicare did patients and their families receive from Medi- not cover: visits by a nurse’s aide for a few hours care spending. The new services might include, each day to help with basic tasks like using the for instance, flexible home-based social care toilet, dressing, or cooking (perhaps to allow an that helps patients deal with limitations in adult child to have a break); concurrent palliative dressing, bathing, eating, and other activities care, which involves hospice-like services that

a publication of the behavioral science & policy association 55 “Medicare’s home hospice coverage provides a nursing visit only every two to three days, even though the patient’s care needs are often much greater”

a patient can receive before deciding to cease and ethical perspectives and to identify their curative care (a decision currently required for impact (if any) on the cost of care that patients hospice care to begin); and cash that could be receive. Applying the principles that follow used for anything, including such nonmedical should help to ensure that the outcomes of purposes as paying for rent or food. More than these proposed studies are translated into policy 40% of participants chose to allocate some of changes that better meet the needs of patients their budget to one or more of the services that and reduce disparities in the care given to disad- the Medicare benefit package does not now vantaged groups. cover, which reduced the amount of traditional medical care they could receive. Principle 1: In each demonstration study, select a condition that frequently results in provision Although the patients knew that the study was of low-value care at the end of life and offer hypothetical and their answers did not affect options that are more flexible than those Medi- the care they were allowed to receive later, care now provides. One condition that could the results indicate that patients and families be considered for such a study is platinum-­ would not only be willing to exert more choice resistant ovarian cancer in patients who have and take more responsibility when allocating been hospitalized. Such patients have a 9% their Medicare benefits, but they would also chance of surviving for six months, with none do so in ways that could improve satisfaction surviving 12 months.23 These patients are usually with end-of-life care and potentially reduce the offered a choice between third- or fourth-line cost of the care they choose to receive. The chemotherapy and hospice care. Many patients tendency of participants to allocate Medicare and families who opt for home-based (instead resources away from last-ditch, low-value care of institutional) hospice care are surprised and toward other care suggests, as well, that to discover that Medicare’s home hospice more freedom of choice could improve health coverage provides a nursing visit only every two equity by allowing individuals who have different to three days, even though the patient’s care preferences because of disparities (such as diffi- needs are often much greater. culty affording transportation to doctors’ offices or not having a family member who can afford A pilot study could offer patients in this situation to miss work to help them out) to improve the a choice between last-ditch medical treatment value of their medical spending by choosing the or a lump sum to be used as desired, such as services most important to them. by paying for home-based care to help with tasks such as bathing, dressing, and cooking, Three Guiding Principles for which is not currently covered by the Medicare Experimentation in Medicare benefit package. The traditional hospice benefit The CHAT study22 provided important evidence would remain, and the new benefit might be that patients might choose different care paths thought of as “hospice plus.” If pilot studies if they had the option, but Medicare (via the provide evidence that this approach can work, Centers for Medicare and Medicaid Innovation then similar studies could be developed for very or a similar governmental office) needs to test common conditions, such as congestive heart the merits of different options and examine failure,24 in which the length of survival is less whether patients will stick with expressed pref- clear than in the ovarian cancer example and erences when making actual care decisions. patients are likely to make longer use of the It also needs to determine if such coverage high-touch, low-tech option if it is selected. changes are acceptable from policy, financial, Medicare could design studies so that they

56 behavioral science & policy | volume 4 issue 1 2018 evaluate the degree to which health equity is is the only way to solve the two-sided learning addressed by the decisions patients make. problem—ensuring that that the insights gained by families and by providers get captured and I am talking here about the types of pilot tests used instead of going nowhere. that should be undertaken to improve the match between covered services and patient needs, Principle 3: Adopt an ethic of harm reduction. but policymakers are sure to also consider the The goal of reducing low-value care should be results from a financial perspective. If the goal viewed through a lens of harm reduction, or is improving health equity, shifting funding from the acceptance that some negative outcomes low- to high-value care would be enough to or behaviors will not be eradicated but can be achieve such a goal, and saving money would reduced. Requiring new Medicare policies to not be a key consideration. If saving money for instantly eliminate all mismatches between Medicare were a key aspect of pilot tests, then patients’ needs and their care would be unre- the structure of the test would likely be different. alistic; small gains and improvements are Either type of test is reasonable, but the goals victories and should be valued for the reduction of a test should be made clear to patients and in suffering they facilitate. families, who will have to be meaningfully involved in the allocation decisions that are an The evolution of smoking policies in the United inherent part of such pilot tests. For example, States offers an example of the value of focusing if a low-income Medicare beneficiary chooses persistently on harm reduction. In 1950, 55% of home-based care in lieu of expensive last-ditch the adult population smoked, and the current chemotherapy, that decision would likely reduce rate of 18% was unimaginable. The transition Medicare’s overall costs for this person’s care. If, took 75 years of multifaceted policy efforts, on the basis of the individual’s low income, the combined with shifting cultural norms that were person was also granted cash to pay bills and influenced by policy changes but also enabled reduce family strain, this provision would reduce the changes to be enacted.25–27 Policymakers the cost savings to Medicare but could improve need to adopt a long time horizon to judge health equity. success. Today, many Americans find it hard to believe that airlines still allowed smoking on Principle 2: Commit to an evidence-based planes in 1994, yet people in 1975 would have process. Rosenbaum15 has noted that the less- found it hard to believe that the practice would is-more crusade is backed more by belief than ever end. by evidence, and I agree that a full commit- ment to evidence is required if an attempt to A reduction in low-value care for one condi- shift from low-value spending to high-value tion, such as platinum-resistant ovarian cancer, spending is to be made. The outcomes of all would have only a small impact on the Medicare participants—patients, families, and providers— program as a whole. However, it could be the need to be measured and recorded, along beginning of a sustained effort that could have with the effects on Medicare’s finances. As a large impact over time as the general idea is the evidence base accumulates over time, the applied to more common conditions. information provided to patients, families, and providers (who will have to communicate these Following Through options to patients) should be updated. New Using these guiding principles, Medicare could treatment options—such as a new drug that is design and test a series of pilot studies in which clearly beneficial for late-stage ovarian cancer— patients and families could decline care that would have to be taken into account, and a pilot evidence suggested was often of low value and test might even have to be stopped in such a select benefits that are not currently covered by case, much as a clinical trial of a new drug is Medicare, such as long-term support for caring often stopped if the early results are convincing. for the elderly at home, hospice-like services Ever-improving information, collected while that focus on symptom relief and maximizing following patients from choices to outcomes, the quality of life before a patient becomes

a publication of the behavioral science & policy association 57 eligible for hospice, and even cash that could • Can the communication of uncertainty to be used for any purpose chosen by the patient. patients and families be improved? Such pilot studies could provide insight into whether and how patients and families are able Of course, it is one thing to offer patients a high- to make use of existing clinical evidence relating value home-care option through Medicare; it to the prognoses associated with the treatments is another thing to get patients to choose this available for the patients’ condition. The findings high-value option. An abundance of behavioral would then be used to help patients and family research suggests that the way in which options members overcome their lack of knowledge due are presented to patients and their families (that to the two-sided learning problem by providing is, the choice architecture) can critically influ- them with information about the experiences ence their decisions.28,29 The optimal choice of other patients. The collected results could architecture must be carefully designed and potentially lead to changes in the benefits that tested, but behavioral research provides some the Medicare program agrees to cover. educated guesses about which approaches might work best. Congress and officials in the executive branch responsible for determining what Medicare First, research suggests that policymakers covers and the public (which both uses and pays should be careful to avoid any language that for Medicare) would need to keep the following suggests a trade-off between the patient’s life questions in mind when considering whether to expectancy and money, focusing instead on adjust coverage rules in response to the findings improving the well-being of the patient. Second, of pilot studies: numerous studies have found that defaults have an outsized impact on choices.30 Thus, • What are the differences in survival and a poor prognosis by a clinician might trigger a quality of life in patients given the most protocol in which Medicare presents the home common treatments? care option as the default choice from which patients must opt out to receive continued • What are the costs of these different options, low-value treatment. This presentation may to Medicare and to patients and their families? convey an implicit endorsement of home care and lead patients to construe home support, • Of the common treatments, are any more palliative care, and additional financial support expensive and less effective than others? as something they would have to give up to Should coverage be eliminated for the obtain low-value treatment,31 thereby making least effective approaches? (Such decisions the home care option more attractive. Third, would be controversial if implemented via a a home care default could be bolstered by an top-down administrative process, but they explanation that the default was set because may be accepted by providers, patients, and of high satisfaction scores among families who families if they are driven by the results of pilot have chosen it, as compared with the satisfac- studies in which patients make the decisions.) tion scores of familes who have chose low-value hospital treatments; research suggests that • How can new evidence on patient and family when people face difficult choices, they can be satisfaction with different kinds of coverage swayed by the preferences of others who faced options tested in pilot studies be used to a similar choice.32 ensure that the menu of benefits made avail- able by Medicare to patients remains up to date with the options patients and families Implications for Health Equity currently desire? The possibility that the pilot study research program I have described could identify • How can the way the health care system low-value spending in a health insurance obtains information about patient and family program open to everyone age 65 years and preferences be improved? older means that resources could be freed for

58 behavioral science & policy | volume 4 issue 1 2018 reallocation to high-value spending, which could, in turn, improve health equity. For “funding could be steered instance, funding could be steered to benefits more useful to disadvantaged groups, such as to benefits more useful to cash; home-based long-term care that is not currently covered by Medicare; or home modi- disadvantaged groups, such fications, such as ramps, walk-in showers, and the like, that would allow people to stay at home as cash” in spite of illness. Of course, Medicare officials and Congress, which approves the Medicare budget, would have to choose to reallocate the government decided to respond to such a spending in a way that would invest resources change by lowering (or at least not increasing) in options that are not currently covered in the amount that younger generations pay in Medicare’s benefit package, instead of using the payroll and income taxes to finance Medi- savings to reduce the size of Medicare’s overall care today for elderly beneficiaries. Easing the budget. financing burden on workers would dispro- portionately help low-income workers, which The CHAT study conducted in North Carolina should increase health equity, given the correla- gives an indication of how evidence-based tion between income and health. revisions to Medicare offerings could improve health equity. Recall that the CHAT protocol In research seminars, when I discuss the general hypothetically offered three types of benefits idea of altering Medicare in ways that would that Medicare does not cover. Nearly one in improve end-of-life care, people often invoke a five participants reallocated at least some of study called SUPPORT33 as an argument against their finite spending money to all three types of it. They say that the approach has been tried and benefits (home-based long-term care, concur- failed—in the sense that, although SUPPORT rent palliative care, and cash that could be used documented problems with aggressive care for any purpose); 40% choose at least one. near the end of life, the information did not The most important predictor was race: Black change patient and family preferences or the participants were nearly twice as likely as Whites care people received. (odds ratio = 1.91, 95% confidence interval [1.14, 3.23]; see note A) to consistently allocate The criticism that the approach has been tried resources to those options. Race was the only unsuccessfully is wrong for two reasons. First, statistically significant predicator of choosing the SUPPORT study33 targeted patient and family all three noncovered benefits, after controlling decisionmaking in the intensive care unit. When for age, gender, income, marital status, health a patient arrives in the intensive care unit, it is status, and out-of-pocket spending. This finding too late for well-reasoned and nuanced deci- suggests that some people who typically face sionmaking; at that point, patients are already health disparities (such as less access to care a part of a system set up to do everything by and worse health outcomes) may be more default. Care decisions need to be made far interested in choosing to receive some of their upstream. Second, SUPPORT is more than two Medicare entitlement through the types of decades old, and the baby boomers who are benefits that they anticipate would be of higher flooding into Medicare differ culturally from value to them when they are facing an end-of- their parents: they are more likely to want to life situation. Although the exercise22 described direct more of their care, an inclination that was theoretical, all the study participants had could be harnessed in the way I have suggested. cancer, so the experimental situation was not implausible. I also respond to doubt by noting that the persistence of health inequity and Medicare’s A reduction in Medicare costs could even financial problems mean that out-of-the-box have an indirect impact on health equity if changes need to be considered and discussed.

a publication of the behavioral science & policy association 59 If patients and families who are given evidence- author affiliation based information decide to take advantage of new high-value care options, this outcome Taylor: Duke University. Corresponding author’s provides some evidence that patients and fami- e-mail: [email protected]. lies may be willing to consider more radical changes to what benefits are provided by Medi- care, so long as patients maintain control over author note their choice of benefits. The author thanks Lisa Rosenbaum and Amitabh Chandra for comments on earlier drafts. Errors A Brighter Future and conclusions are the authors responsibility. The United States needs to engage in a broad discussion about the care its citizens receive as they age and endure illness and disability. Chil- endnote dren, grandchildren, and great-grandchildren A. Editors’ note to nonscientists: An odds ratio foot the bill as their elders join the Medicare conveys how the presence of one factor increases program. Because the only thing that everyone the odds of having a second factor present. In this will inevitably do is die, health researchers case, an odds ratio of 1.91 means that the odds and policymakers urgently need to solve the of reallocating resources were almost twice as likely for Blacks as for Whites. The 95% confidence two-sided learning problem, which keeps interval indicates that there is less than a 5% prob- patients’ and providers’ insights into the flaws of ability that the odds ratio would fall outside the today’s end-of-life treatments from being trans- range of 1.14–3.23. In other words, if you took 20 lated into care that matches patients’ needs. samples from this population, you would expect Solving the problem could provide large bene- that 19 out of 20 times, the odds ratio would be fits to each of us as individuals and to society as higher than 1.14 and lower than 3.23. a whole and help to transform the health care system into one that learns.34 Such a system would provide a more just and equitable distri- bution of spending in the Medicare program and, in so doing, could spur broader reconsid- erations of spending across the life course.

60 behavioral science & policy | volume 4 issue 1 2018 references

1. Braveman, P., Arkin, E., Orleans, T., with symptoms of depression and 24. Downar, J., Goldman, R., Pinto, R., Proctor, D., & Plough, A. (2017). What is posttraumatic stress disorder among Englesakis, M., & Adhikari, N. K. J. health equity? And what difference does family members of patients who die in (2017). The “surprise question” for a definition make? Princeton, NJ: Robert the ICU. Chest, 139, 795–801. predicting death in seriously ill patients: Wood Johnson Foundation. 14. Siegel, M. D., Hayes, E., Vanderwerker, A systematic review and meta-analysis. 2. Woolf, S. H., & Aron, L. (Eds.). (2013). L. C., Loseth, D. B., & Prigerson, H. G. CMAJ, 189, E484–E493. U.S. health in international perspective: (2008). Psychiatric illness in the next of 25. Sloan, F. A., Ostermann, J., Picone, G., Shorter lives, poorer health. Washington, kin of patients who die in the intensive Conover, C., & Taylor, D. H., Jr. (2004). DC: National Academies Press. care unit. Critical Care Medicine, 36, The price of smoking. Cambridge, MA: 3. Bradley, E. H., & Taylor, L. A. (2013). 1722–1728. MIT Press. The American health care paradox: 15. Rosenbaum, L. (2017). The less-is-more 26. Sloan, F. A., Smith, V. K., & Taylor, D. Why spending more is getting us less. crusade—Are we overmedicalizing or H., Jr. (2003). The smoking puzzle: Washington, DC: Public Affairs. oversimplifying? New England Journal Information, risk perception, and choice. 4. Williams, D. R., McClellan, M. B., of Medicine, 377, 2392–2397. Cambridge, MA: Harvard University & Rivlin, A. M. (2010). Beyond the 16. Einav, L., Finkelstein, A., Mullanainathan, Press. Affordable Care Act: Achieving real S., & Obermeyer, Z. (2018, June 29). 27. Sloan, F. A., Smith, V. K., & Taylor, D. H., improvements in Americans’ health. Predictive modeling of U.S. health Jr. (2002). Information, addiction, and Health Affairs, 29, 1481–1488. care spending in late life. Science, 360, “bad choices”: Lessons from a century 5. McGinnis, J. M., Williams-Russo, P., & 1462–1465. of cigarettes. Economic Letters, 77, Knickman, J. R. (2002). The case for 17. Kahneman, D., & Tversky, A. (1979). 147–155. more active policy attention to health Prospect theory: An analysis of decision 28. Thaler, R. H., & Sunstein, C. R. (2008). promotion. Health Affairs, 21, 78–93. under risk. Econometrica, 47, 263–291. Nudge: Improving decisions about 6. Smedley, B. D., & Syme, S. L. (2000). 18. Tversky, A., & Kahneman, D. (1992). health, wealth, and happiness. New Promoting health: Intervention Advances in prospect theory: Haven, CT: Yale University Press. strategies from social and behavioral Cumulative representation of 29. Johnson, E., Shu, S. B., Dellaert, B. G. research. Washington, DC: National uncertainty. Journal of Risk and C., Fox, C. R., Goldstein, D. G., Haubl, Academies Press. Uncertainty, 5, 297–323. G., & Larrick, R. P. (2012). Beyond 7. Cutler, D. M., & Lleras-Muney, A. (2006). 19. Rottenstreich, Y., & Hsee, C. K. (2001). nudges: Tools of a choice architecture. Education and health: Evaluating Money, kisses, and electric shocks: Marketing Letters, 23, 487–504. theories and evidence (NBER Working On the affective psychology of risk. 30. Johnson, E. J., & Goldstein, D. (2003, Paper No. 12352). Cambridge, MA: Psychological Science, 12, 185–190. November 21). Do defaults save lives? National Bureau of Economic Research. 20. Frederick, S., Loewenstein, G., Science, 302, 1338–1339. 8. Schoeni, R. F., House, J. S., Kaplan, G. & O’Donoghue, T. (2002). Time 31. Dinner, I., Johnson, E. J., Goldstein, A., & Pollack, H. (Eds.). (2010). Making discounting and time preference: A D. G., & Liu, K. (2011). Partitioning Americans healthier: Social and critical review. Journal of Economic default effects: Why people choose economic policy as health policy. New Literature, 40, 351–401. not to choose. Journal of Experimental York, NY: Russell Sage Foundation. 21. Tetlock, P. E., Kristel, O. V., Elso, S. B., Psychology: Applied, 17, 332–341. 9. Riley, G. F., & Lubitz, J. D. (2010). Long- Lerner, J. S., & Green, M. C. (2000). The 32. Cialdini, R. B. (2009). Influence: Science term trends in Medicare payments in psychology of the unthinkable: Taboo and practice (5th ed.). Boston, MA: Allyn the last year of life. Health Services trade-offs, forbidden base rates, and & Bacon. Research, 45, 565–576. heretical counterfactuals. Journal of 33. The Writing Group for the SUPPORT 10. Colla, C. H., Morden, N. E., Sequist, T. Personality and Social Psychology, 78, Investigators. (1995). A controlled D., Schpero, W. L., & Rosenthal, M. B. 853–870. trial to improve care for seriously ill (2015). Choosing wisely: Prevalence 22. Taylor, D. H., Jr., Danis, M., Zafar, S. hospitalized patients: The Study to and correlates of low-value health care Y., Howie, L. J., Samsa, G. P., Wolf, S. Understand Prognoses and Preferences services in the United States. Journal of P., & Abernethy, A. P. (2014). There is a for Outcomes and Risks of Treatments General Internal Medicine, 30, 221–228. mismatch between the Medicare benefit (SUPPORT). JAMA, 274, 1591–1598. 11. Rosenberg, A., Agiro, A., Gottlieb, M., package and the preferences of patients 34. Committee on the Learning Health Barron, J., Brady, P., Liu, Y., Li, C., & with cancer and their caregivers. Care System in America & Smith, M., DeVries, A. (2015). Early trends among Journal of Clinical Oncology, 32, Saunders, R., Stuckhardt, L., & McGinnis, seven recommendations from the 3163–3168. M. (Eds.). (2013). Best care at lower Choosing Wisely campaign. JAMA 23. Foote, J., Lopez-Acevedo, M., Samsa, cost: The path to continuously learning Internal Medicine, 175, 1913–1920. G., Lee, P. S., Kamal, A. H., Secord, A. health care in America. Washington, DC: 12. Petrinec, A. B., & Daly, B. J. (2016). A., & Havrilesky, L. J. (2018). Predicting National Academies Press. Post-traumatic stress symptoms in 6- and 12-month mortality in patients post-ICU family members: Review and with platinum-resistant advanced methodological challenges. Western stage ovarian cancer: Prognostic Journal of Nursing Research, 38, 57–78. model to guide palliative care referrals. International Journal of Gynecological 13. Kross, E. K., Engelberg, R. A., Gries, Cancer, 28, 302–307. C. J., Nielsen, E. L., Zatzick, D., & Curtis, J. (2011). ICU care associated

a publication of the behavioral science & policy association 61 editorial policy

Behavioral Science & Policy (BSP) is an international, peer-­ • Findings (≤ 4,000 words) report on results of new studies reviewed publication of the Behavioral Science & Policy Asso- and/or substantially new analysis of previously reported ciation and Brookings Institution Press. BSP features short, data sets (including formal meta-analysis) and the policy accessible articles describing actionable policy applications of implications of the research findings. This category is most behavioral scientific research that serves the public interest. appropriate for presenting new evidence that supports a Articles submitted to BSP undergo a dual-review process: For particular policy recommendation. The additional length each article, leading disciplinary scholars review for scientific of this format is designed to accommodate a summary rigor and experts in relevant policy areas review for practicality of methods, results, and/or analysis of studies (though and feasibility of implementation. Manuscripts that pass this some finer details may be relegated to supplementary dual-­review are edited to ensure their accessibility to policy online materials). makers, scientists, and lay readers. BSP is not limited to a • Reviews (≤ 5,000 words) survey and synthesize the key particular point of view or political ideology. findings and policy implications of research in a specific disciplinary area or on a specific policy topic. This could Manuscripts can be submitted in a number of different formats, take the form of describing a general-purpose behavioral each of which must clearly explain specific implications for tool for policy makers or a set of behaviorally grounded public- and/or private-sector policy and practice. insights for addressing a particular policy challenge. External review of the manuscript entails evaluation by at least • Other Published Materials. BSP will sometimes solicit two outside referees—at least one in the policy arena and at or accept Essays (≤ 5,000 words) that present a unique least one in the disciplinary field. perspective on behavioral policy; Letters (≤ 500 words) that provide a forum for responses from readers and Professional editors trained in BSP’s style work with authors contributors, including policy makers and public figures; to enhance the accessibility and appeal of the material for a and Invitations (≤ 1,000 words with links to online Supple- general audience. mental Material), which are requests from policy makers Each of the sections below provides general information for for contributions from the behavioral science community authors about the manuscript submission process. We recom- on a particular policy issue. For example, if a particular mend that you take the time to read each section and review agency is facing a specific challenge and seeks input from carefully the BSP Editorial Policy before submitting your manu- the behavioral science community, we would welcome script to Behavioral Science & Policy. posting of such solicitations.

Manuscript Categories Review and Selection of Manuscripts Manuscripts can be submitted in a number of different cate- On submission, the manuscript author is asked to indicate the gories, each of which must clearly demonstrate the empirical most relevant disciplinary area and policy area addressed by basis for the article as well as explain specific implications for his/her manuscript. (In the case of some papers, a “general” (public and/or private-sector) policy and practice: policy category designation may be appropriate.) The relevant • Proposals (≤ 2,500 words) specify scientifically grounded Senior Disciplinary Editor and the Senior Policy Editor provide policy proposals and provide supporting evidence an initial screening of the manuscripts. After initial screening, including concise reports of relevant studies. This category an appropriate Associate Policy Editor and Associate Disci- is most appropriate for describing new policy implications plinary Editor serve as the stewards of each manuscript as it of previously published work or a novel policy recommen- moves through the editorial process. The manuscript author dation that is supported by previously published studies. will receive an email within approximately two weeks of • Reports (≤ 3000 words) provide a summary of output and submission, indicating whether the article has been sent to actionable prescriptions that emerge from a workshop, outside referees for further consideration. External review working group, or standing organization in the behavioral of the manuscript entails evaluation by at least two outside policy space. In some cases such papers may consist of referees. In most cases, Authors will receive a response from summaries of a much larger published report that also BSP within approximately 60 days of submission. With rare includes some novel material such as meta-analysis, exception, we will submit manuscripts to no more than two actionable implications, process lessons, reference to rounds of full external review. We generally do not accept related work by others, and/or new results not presented re-submissions of material without an explicit invitation from in the initial report. These papers are not merely summa- an editor. Professional editors trained in the BSP style will ries of a published report, but also should provide substan- collaborate with the author of any manuscript recommended tive illustrations of the research or recommendations and for publication to enhance the accessibility and appeal of the insights about the implications of the report content or material to a general audience (i.e., a broad range of behav- process for others proposing to do similar work. Submitted ioral scientists, public- and private-sector policy makers, and papers will undergo BSP review for rigor and accessibility educated lay public). We anticipate no more than two rounds that is expedited to facilitate timely promulgation. of feedback from the professional editors. Standards for Novelty in the studies presented. (A template for these disclosures is BSP seeks to bring new policy recommendations and/or new included in our checklist for authors, though in some cases evidence to the attention of public and private sector policy may be most appropriate for presentation online as Supple- makers that are supported by rigorous behavioral and/or social mental Material; for more information, see Simmons, Nelson, & science research. Our emphasis is on novelty of the policy ­Simonsohn, 2011, Psychological Science, 22, 1359–1366). application and the strength of the supporting evidence for that Copyright and License recommendation. We encourage submission of work based on new studies, especially field studies (for Findings and Proposals) Copyright to all published articles is held jointly by the Behav- and novel syntheses of previously published work that have a ioral Science & Policy Association and Brookings Institution strong empirical foundation (for Reviews). Press, subject to use outlined in the Behavioral Science & Policy publication agreement (a waiver is considered only in cases BSP will also publish novel treatments of previously published where one’s employer formally and explicitly prohibits work studies that focus on their significant policy implications. For from being copyrighted; inquiries should be directed to the instance, such a paper might involve re-working of the general BSPA office). Following publication, the manuscript author may emphasis, motivation, discussion of implications, and/or a post the accepted version of the article on his/her personal re-analysis of existing data to highlight policy-relevant implica- web site, and may circulate the work to colleagues and tions or prior work that have not been detailed elsewhere. students for educational and research purposes. We also allow In our checklist for authors we ask for a brief statement that posting in cases where funding agencies explicitly request explicitly details how the present work differs from previously access to published manuscripts (e.g., NIH requires posting on published work (or work under review elsewhere). When in PubMed Central). doubt, we ask that authors include with their submission copies of related papers. Note that any text, data, or figures excerpted Open Access or paraphrased from other previously published material must BSP posts each accepted article on our website in an open clearly indicate the original source with quotation and citations access format at least until that article has been bundled into as appropriate. an issue. At that point, access is granted to journal subscribers and members of the Behavioral Science & Policy Association. Authorship Questions regarding institutional constraints on open access Authorship implies substantial participation in research and/ should be directed to the editorial office. or composition of a manuscript. All authors must agree to the order of author listing and must have read and approved Supplemental Material submission of the final manuscript. All authors are respon- While the basic elements of study design and analysis should sible for the accuracy and integrity of the work, and the senior be described in the main text, authors are invited to submit author is required to have examined raw data from any studies Supplemental Material for online publication that helps elab- on which the paper relies that the authors have collected. orate on details of research methodology and analysis of their data, as well as links to related material available online else- Data Publication where. Supplemental material should be included to the extent BSP requires authors of accepted empirical papers to submit all that it helps readers evaluate the credibility of the contribution, relevant raw data (and, where relevant, algorithms or code for elaborate on the findings presented in the paper, or provide analyzing those data) and stimulus materials for publication on useful guidance to policy makers wishing to act on the policy the journal web site so that other investigators or policymakers recommendations advanced in the paper. This material should can verify and draw on the analysis contained in the work. In be presented in as concise a manner as possible. some cases, these data may be redacted slightly to protect subject anonymity and/or comply with legal restrictions. In Embargo cases where a proprietary data set is owned by a third party, a Authors are free to present their work at invited colloquia and waiver to this requirement may be granted. Likewise, a waiver scientific meetings, but should not seek media attention for may be granted if a dataset is particularly complex, so that it their work in advance of publication, unless the reporters in would be impractical to post it in a sufficiently annotated form question agree to comply with BSP’s press embargo. Once (e.g. as is sometimes the case for brain imaging data). Other accepted, the paper will be considered a privileged document waivers will be considered where appropriate. Inquiries can be and only be released to the press and public when published directed to the BSP office. online. BSP will strive to release work as quickly as possible, and we do not anticipate that this will create undue delays. Statement of Data Collection Procedures BSP strongly encourages submission of empirical work that Conflict of Interest is based on multiple studies and/or a meta-analysis of several Authors must disclose any financial, professional, and datasets. In order to protect against false positive results, we personal relationships that might be construed as possible ask that authors of empirical work fully disclose relevant details sources of bias. concerning their data collection practices (if not in the main text then in the supplemental online materials). In particular, we Use of Human Subjects ask that authors report how they determined their sample size, All research using human subjects must have Institutional all data exclusions (if any), all manipulations, and all measures Review Board (IRB) approval, where appropriate.

sponsors

The Behavioral Science & Policy Association is grateful to the sponsors and partners who generously provide continuing support for our non-profit organization.

To become a Behavioral Science & Policy Association sponsor, please contact BSPA at [email protected] or 1-919-681-5932.

behavioral science & policy

where behavioral research meets policy + practice

who we are our mission

The Behavioral Science & Policy Association is a global hub To foster and connect a growing community of interdisciplinary of behavioral science resources, curated by leading scholars practitioners, providing thoughtful application of rigorous and policymakers, aimed at facilitating positive change and behavioral science research for the public and private sectors, innovative solutions to a range of societal challenges. with a simple goal in mind: addressing social change for the benefit of all.

membership call for submissions

There is a growing movement among social scientists and Behavioral Science & Policy is an international, peer-reviewed leaders within the public and private sector, dedicated to journal featuring succinct and accessible articles outlining grounding important decisions in strong scientific evidence. actionable policy applications of behavioral science research that serves the public interest. BSPA plays a key role in this movement, encouraging decisions to be based on evidence. We need you to join BSP journal submissions undergo a dual-review process. Leading us in this effort to make a lasting impact. scholars from specific disciplinary areas review articles to assess their scientific rigor; while at the same time, experts in designat- As a BSPA member, you will receive numerous benefits ed policy areas evaluate these submissions for relevance and including an oniine subscription to Behavioral Science & feasibility of implementation. Policy, early-bird rates for conferences, workshops and briefings, exclusive access to BSPA online webinars and Manuscripts that pass this dual-review are edited to ensure podcasts, waived fees for journal submissions and more. accessibility to scientists, policymakers, and lay readers. BSPA is not limited to a particular point of view or political ideology. Be a leader in our drive for change at This journal is a publication of the Behavioral Science & Policy behavioralpolicy.org/signup Association and the Brookings Institution Press.

We encourage you to submit your manuscript today to Behavioral Science & Policy, at behavioralpolicy.org/journal

Behavioral Science & Policy Association P.O. Box 51336 Durham, NC 27717-1336