Series

America: Equity and Equality in Health 1 Inequality and the health-care system in the USA

Samuel L Dickman, David U Himmelstein, Steffie Woolhandler

Widening economic inequality in the USA has been accompanied by increasing disparities in health outcomes. Lancet 2017; 389: 1431–41 The life expectancy of the wealthiest Americans now exceeds that of the poorest by 10–15 years. This report, part of a See Editorial page 1369 Series on health and inequality in the USA, focuses on how the health-care system, which could reduce income-based See Comment pages 1376 disparities in health, instead often exacerbates them. Other articles in this Series address population health and 1378 inequalities, and the health effects of racism, mass incarceration, and the Affordable Care Act (ACA). Poor Americans This is the first in a Series of have worse access to care than do wealthy Americans, partly because many remain uninsured despite coverage five papers about equity and equality in health in the USA expansions since 2010 due to the ACA. For individuals with private insurance, rising premiums and cost sharing have Department of Medicine, undermined wage gains and driven many households into debt and even bankruptcy. Meanwhile, the share of health- University of California, care resources devoted to care of the wealthy has risen. Additional reforms that move forward, rather than backward, San Francisco, CA, USA from the ACA are sorely needed to mitigate health and health-care inequalities and reduce the financial burdens of (S L Dickman MD); City medical care borne by non-wealthy Americans. University of New York School of Urban Public Health at Hunter College, New York, NY, 5 Introduction research by the economists Piketty and Saez that USA (Prof D U Himmelstein MD, As economic inequality in the USA has deepened, so too revealed levels of income inequality unrivalled since the Prof S Woolhandler MD); and has inequality in health. Almost every chronic condition, stock market bubble of the 1920s. The share of total Harvard Medical School, Boston, MA, USA from stroke to heart disease and arthritis, follows a income going to the top 1% of earners has more than (Prof D U Himmelstein, 6 predictable pattern of rising prevalence with declining doubled since 1970 (figure 1), while most workers in the Prof S Woolhandler) 1 7 income. The life expectancy gap between rich and poor USA have experienced slow income growth. As Correspondence to: Americans has been widening since the 1970s,2 with the measured by the Gini coefficient, a standard metric of Prof David U Himmelstein, City difference between the richest and poorest 1% now income inequality, the USA is now more unequal than all University of New York School of Urban Public Health at Hunter standing at 10·1 years for women and 14·6 years for but three other countries (Chile, Mexico, and Turkey) in College, NY 10035, USA 3 men. The health of poor communities is often neglected: the Organization for Economic Co-operation and [email protected] for example, in Flint (MI, USA), a de-industrialised, Development (OECD). The most equal countries are See Online for infographic impoverished, and predominately African-American city, Denmark, Slovenia, Norway, and Slovakia. www.thelancet.com/ public officials dismissed evidence that children were The surge in top incomes has magnified inequality in infographics/us-health being exposed to toxic levels of lead in the city’s drinking wealth (ie, assets). Since 1986, the top 0·1% of households water for several months.4 (those with assets exceeding US$20 million) has Attention to economic inequality intensified after the Occupy Wall Street movement decried the rising wealth and power of the richest 1%. This movement popularised Key messages • Economic inequality in the USA has been increasing for decades and is now among the highest in developed countries. Search strategy and selection criteria • Differences in life expectancy have been widening, with the wealthiest Americans now We searched PubMed and Google Scholar using the following living 10–15 years longer than the poorest. terms: “health care quality” OR “quality of care” AND “primary • Despite coverage gains from the Affordable Care Act, about 27 million Americans care” OR “specialty care” OR “specialist” OR “hospital” OR remain uninsured—a number that is likely to increase under the reforms advocated by “surgery” OR “cancer” OR “outpatient” AND “inequality” OR Republicans now empowered in Washington, DC. “unequal” OR “disparities” AND “US” OR “United States”; • Both overall and government health spending are higher in the USA than in other “access to care” OR “health care access” OR “barriers to care” countries, yet inadequate insurance coverage, high cost sharing by patients, and AND “inequality” OR “unequal” OR “disparities” AND “poor” geographical barriers restrict access to care for many. OR “poverty” OR “wealthy” OR “income” AND “US” OR • Financing of health care in the USA is regressive, with poor and middle-class “United States”; “regressive” OR “progressive” AND “health individuals paying a larger share of their incomes for care than the affluent, thereby financing” OR “health finance” OR “out of pocket”; and deepening inequalities in disposable income. “medical bankruptcy” OR “medical debt”. Our search included • Rising insurance premiums for employer-sponsored private coverage have eroded articles focusing on the USA published in English between wage gains for middle-class Americans. Jan 1, 2011, and March 31, 2016, prioritising recent research. • Medical indebtedness is common among both insured and uninsured Americans, and We identified additional sources, including important older often leads to bankruptcy. manuscripts, from the reference lists of selected articles and • To achieve health-care equality, a non-market financing scheme that treats health care from consultation with expert colleagues. as a human right is essential.

www.thelancet.com Vol 389 April 8, 2017 1431 Series

20 middle-class families have been bankrupted by illness and medical bills.14 Meanwhile, very wealthy Americans are turning to so-called concierge practices that offer lengthy office visits and unfettered access to specialists. This Series paper examines how the health-care system 15 in the USA contributes and responds to inequality. We focus our attention on the association between inequality and the medical care system. total income (%)

of We first review how social position influences 10 Americans’ access to medical services and the quality of Share those services. The uninsured face the greatest barriers to care, but many insured Americans are also unable to afford medical care because of cost sharing. Although race-based disparities in quality are well documented, the 5 5 0 5 0 low quality scores of doctors and hospitals serving poor 1910 1915 1920 1925 1930 193 1940 1945 1950 1955 196 196 1970 1975 1980 1985 1990 1995 200 2005 2010 2015 communities might reflect patients’ deprived social Year circumstances rather than their providers’ performance. Figure 1: Share of total income received by the top 1% of earners in the USA, 1913 to 2014 We also review how the health-care costs borne by Source: Saez (2016).6 Income is defined as pre-tax market income excluding capital gains and government households—in the form of insurance premiums, taxes, transfers. In 2014, the top 1% included families with annual incomes above $387 810, who collectively received 18% of total income in the USA. and out-of-pocket payments—exacerbate income inequality, forcing many Americans to cut back on food accumulated nearly half of all new wealth, and now and other necessities, and contributing to most personal controls as much wealth as the bottom 90%, whose share bankruptcies. We conclude by discussing the historical has fallen steadily.8 context for today’s health-care inequalities, and propose Wealth inequality between racial and ethnic groups in options for reform. the USA is especially striking, and is several times greater than income inequality. In 2013, median family Inequality and access to care wealth for the non-Hispanic white population was ten Income-related disparities in access to care are far wider times that of Hispanics and more than 12 times that of in the USA than in other wealthy countries.15 Before the African-Americans.9 The racial wealth gap results from 2010 passage of the ACA, which progressively expanded historical factors dating back to slavery—many of which health insurance coverage, 39% of Americans with persist—including legalised racial segregation in the pre- below-average income reported not seeing a doctor for a civil rights era, pervasive job and housing discrimination, medical problem because of cost, compared with 7% of exclusionary city zoning laws, unequal education, and low-income Canadians and 1% of those in the UK.16 inheritance laws that perpetuate past inequalities.10 Inequality in access to care is particularly stark in Although top incomes have risen, so has extreme Southern states. For example, in Texas, Mississippi, and poverty. More than 1·6 million households in the USA, Florida, adults on a low income are more than twice as including 3·5 million children, survive on incomes of likely to face cost-related barriers to care as their less than $2 per person per day—WHO’s definition of counterparts in Maine (a relatively poor New England extreme poverty; this number has more than doubled state) and Massachusetts.17 since the 1990s.11 Disparities in access are largely due to high rates of The health-care system could soften the effects of uninsurance or inadequate health insurance among low- economic inequality by delivering high-quality care to all. income Americans, although Americans with above- Yet the institutions and financing patterns of the health- average incomes probably also have worse access to care care system in the USA—by far the world’s most than do their peers in other countries.16 Today, despite gains expensive12–cause it to fall short of this ideal. Although due to the ACA, 27 million Americans (down from inequalities exist to some extent in every health-care 50 million before the passage of the ACA) remain system, they are particularly stark in the USA. Unequal uninsured. Most of the uninsured have annual incomes access to medical services is likely to contribute to near or below the official poverty line ($11 770 for an disparities in health status, while rising costs (for both individual in 2016). the insured and uninsured) reduce disposable incomes, The uninsured are far more likely than the insured to particularly burdening low-income households. forgo needed medical visits, tests, treatments, and Many patients cannot afford the care they need, and medications because of cost. Cost barriers are especially often forgo medical care altogether. For example, 19% of severe for the millions of uninsured Americans with non-elderly adults in the USA who received prescriptions chronic conditions.18 For example, middle-aged adults in 2014 (after full implementation of the Affordable Care with no coverage for eye care report difficulties in Act [ACA]) could not afford to fill them.13 Millions of reading or recognising a friend across the street

1432 www.thelancet.com Vol 389 April 8, 2017 Series

more frequently than do comparable individuals with Cost sharing and private insurance coverage.19 Uninsured individuals with diabetes spend, In the private insurance market, cost sharing by patients on average, $1446 out of pocket for medical services (ie, through user fees) has increased substantially since each year, and more than 30% do not have a primary the 2000s.29 Many plans now impose co-payments of care provider.20 Similarly, low-income and uninsured more than $30 for primary care and more than $45 for Americans with psychiatric disorders are far more likely specialist visits.30 More than 80% of employer-based than the insured and those on higher incomes are to plans include an annual deductible (the amount a patient experience difficulties obtaining care.21 For poor must pay before insurance covers additional costs), Americans, gaining insurance boosts access to care which averaged $1478 in 2016, an increase of 2·5 times (although it does not fully close the gap between poor since 2006.30 Cost sharing by patients is even higher in and affluent patients), leading to more visits for plans sold through the insurance exchanges created by preventive screenings and greater satisfaction with care the ACA. In mid-level, so-called silver tier plans (which than before gaining coverage.22 account for about 70% of exchange coverage), deductibles averaged $3064 in 2016,30 although some subsidies were Medicaid insurance for low-income Americans available to cushion the deductibles for enrollees with Medicaid, the public insurance programme that covers incomes 100–250% of the poverty level. 58 million low-income Americans, improves health Many private plans also reduce premium costs by outcomes and access to care for its beneficiaries.23 restricting patients’ choice of providers to narrow Gaining Medicaid coverage reduces rates of clinical networks of doctors and hospitals, which often exclude depression,24 financial problems due to illness,24 and academic and cancer referral centres.31 Enrollees who mortality.23 The generosity of the Medicaid programme, seek out-of-network care (either by choice or because of which is largely controlled by state governments, is a medical necessity) generally must pay the entire bill out key determinant of access-related disparities. Before of pocket. the passage of the ACA, most states restricted Medicaid Predictably, patients’ use of care declines as their cost- eligibility to poor children and pregnant women, sharing obligation rises, and people with the worst health disabled people, and the poorest adults with children. are most likely to cut back on care.32 Paradoxically, this Childless adults and parents with incomes above reduction in care-seeking can fail to cut system-wide use, stringent state-specific thresholds (eg, 10% of the instead shifting care from the sick and poor to the healthy poverty level in Alabama) were generally excluded. and wealthy. At least in some cases, when poor patients The ACA’s expansion of Medicaid to all citizens with avoid care, doctors and hospitals fill the empty annual incomes at or below 138% ($16 643 for an appointment slots and beds with patients who are less individual in 2017) of the poverty level promised to price-sensitive33—an example of supply-sensitive cover millions of previously uninsured Americans. demand.34 Strikingly, the USA has the world’s highest However, the US Supreme Court ruled in 2012 that health-care expenditures despite extensive cost sharing states could opt out of the Medicaid expansion, and the by patients. Trump administration is likely to further erode or Additionally, care forgone because of cost sharing repeal it. As of 2016, 19 states (mostly in the South) might ultimately raise costs by increasing downstream have opted out, cutting the number of citizens who health problems. When the Medicare programme (the would otherwise have gained coverage by about public coverage for people aged 65 or older and those 5 million. An additional 5–6 million undocumented with long-term-disabilities) added new co-payments, immigrants do not have insurance because the outpatient visits decreased but hospital admissions ACA specifically excluded them from its coverage increased.35 Among patients who developed a expansion,25 perpetuating major constraints on their myocardial infarction, elimination of medication co- access to care.26 Fortunately, the ACA increased funding payments after the cardiac event increased compliance, for community health centres, which deliver much- and (for racial and ethnic minority patients) led to a needed care to millions of low-income Americans, and 35% reduction in major vascular events and a extended mental health parity regulations aimed at 70% reduction in total health-care spending.36 Similarly, improving insurance coverage for mental health and among children aged 5–18 years with asthma, those addiction treatment. whose insurance required higher co-payments used Although Medicaid improves access to care, specialist fewer medications but had a 41% greater risk of asthma- care is often unobtainable because the programme pays related hospital admissions than did children with low fees to physicians,27 who are free to turn away lower co-payments.37 For nearly a third of children with Medicaid patients. For example, 76% of orthopaedists’ asthma from low-income families with high cost- offices in a nationwide audit study refused to offer an sharing coverage through the Kaiser Health Plan, appointment to a Medicaid-insured child with a parents reported delaying or avoiding outpatients visits, fracture, whereas only 18% refused a child with private and 14·8% reported non-adherence to medications insurance.28 because of cost; 15·6% of all parents (including those www.thelancet.com Vol 389 April 8, 2017 1433 Series

with higher incomes) reported borrowing money or dipped below those of the wealthiest 20%.42 In Canada, cutting back on necessities to pay for their children’s by contrast, the poorest citizens receive the most asthma care.38 medical services, commensurate with their increased health needs.45 Meanwhile, health-care expenditures for Defining underinsurance the wealthiest 20% of Americans accelerated, raising Rising deductibles and other forms of cost sharing by their share of overall health-care consumption. The patients have eroded the traditional definition of ACA, fully implemented in 2014, led to a surge in insurance: protection from the financial harms of illness. health-care expenditures for the bottom 20%, but The term underinsurance describes this problem, but it expenditures for the middle class have flattened while does not have a standard definition. Some studies of health-care consumption by the wealthiest Americans underinsurance have focused on financial vulnerability continues to grow. (eg, measurement of deductibles as a fraction of income),39 others on out-of-pocket costs incurred (either Access problems in absolute dollars or relative to income),40 whereas Geography often affects access to care. Because others have highlighted barriers to care (because of cost physicians are concentrated in cities and affluent or narrow insurance networks).41 No standard quantitative suburbs, many Americans living in rural areas find it thresholds exist for these different concepts. difficult to obtain primary46 and specialty care.47 Many The various definitions of underinsurance highlight rural and Southern states also have a shortage of adequate two related but distinct problems: people with inadequate family planning resources. Texas, for example, has insurance risk financial harm when they receive medical imposed onerous regulations and funding cuts on family services, and they are therefore less likely to obtain planning clinics, causing closure of many48 and a needed care. Despite the absence of consensus on the subsequent increase in unwanted pregnancies.49 Since definition of underinsurance, it is clear that these the closure of the last local abortion clinic in 2013, problems affect many Americans with private coverage women in Lubbock, Texas (population 244 000), are now and have increased. more than 250 miles away from the nearest abortion Between 2004 and 2013, high rates of uninsurance, provider. rising cost sharing (ie, underinsurance), and stagnant Women are also at a financial disadvantage because of incomes all contributed to a decline in overall health- their greater health-care needs (including reproductive care consumption (as measured by the total amount care) than those of men. Although fewer women than spent by insurers and patients) for poor Americans, a men are uninsured, those with insurance have higher trend that was reversed in 2014 when the major out-of-pocket costs. For example, among people with provisions of the ACA came into effect (figure 2).42 For employer-sponsored coverage, women’s out-of-pocket the first time since the 1970s, per-capita medical costs were $233 higher than men’s in 2013;50 among expenditures for the poorest fifth of Americans (who Medicare enrollees, such costs were $640 higher for are, on average, much sicker than the wealthiest 20%) women than they were for men in 2011.51 These costs are especially burdensome because women’s median incomes are 39% lower than those of men.52 6000 Poorest quintile Illness-based disparities are particularly stark for Middle 60% Richest quintile mental illness and substance abuse. Historically, a large 5000 share of psychiatric care was paid for out of pocket or provided in underfunded public institutions. Jails 4000 remain the largest so-called inpatient mental health facilities in the USA. Although the 2008 Mental Health

3000 Parity and Addiction Equity Act mandated that most insurance plans provide equivalent coverage for mental and physical illness, implementation of 2000 this requirement was delayed until 2015, and its 53

Health expenditure per capita (US$) enforcement has proven difficult. Moreover, most 1000 Medicaid programmes (which cover many people with mental disorders) are exempt from these regulations. 0 Psychiatric—and particularly substance abuse— 3 8 3 196 19641966196 19691971197319741976197819791981198319841986198819891991199319941996199819992001200 2004200620082009201120132014 providers are in short supply on a national scale, Year especially in poor and rural areas;53 these areas have been particularly hard hit by the epidemic of drug Figure 2: US health expenditures per capita, adjusted for inflation, by income group, 1963 to 2014 Sources: Dickman and colleagues (2016);42 and the Medical Expenditures Panel Survey43,44 (for years 2013–14). overdoses and self harm, which pushed up the overall Figures for 1996 to 2013 are 2-year moving averages; single-year figures are provided for 2014 to show the effect death rate in the USA in 2015. The ACA, which applied of the Affordable Care Act’s coverage expansions in this year. the parity requirement to the plans sold through the

1434 www.thelancet.com Vol 389 April 8, 2017 Series

exchanges, increased access to mental health, but not to drugs: 27% of low-income Medicare beneficiaries with substance abuse treatment; substantial racial and dementia, hip or pelvic fracture, or chronic renal failure ethnic disparities persist.54 received contraindicated medications compared with 16% of higher-income individuals.17 Poor Americans older than Inequality and quality of care 50 years are also far less likely than their affluent For many conditions, increased quality is implied by, and counterparts to receive recommended influenza and inseparable from, improved access to care. An increased pneumonia vaccinations, and cancer screening tests,17 frequency of primary care visits, for example, is although cost-related barriers might underlie these associated with improved control of diabetes.55 Similarly, differences. among patients who developed an acute myocardial There is strong evidence showing that quality of care is infarction, the uninsured were 38% more likely (and the worse for racial and ethnic minorities,68 although racial underinsured 21% more likely) than the well insured to disparities in the quality of hospital care could have delay seeking emergency care.56 narrowed between 2005 and 2010 as a result of Yet it is unclear whether income-related disparities in improvements among hospitals serving patients from access to care are accompanied by other gaps in global minority backgrounds and more equitable care within all quality, which are harder to measure. Poverty itself hospitals.69 Yet unequal access to care, along with causes ill health, compromises non-medical social institutional racism, remain important drivers of persistent supports and resources that improve medical outcomes, disparities in health-care quality for racial and ethnic and is associated with worse satisfaction with care.57 minorities. For example, although African-Americans tend Hence, differences in the socioeconomic profile of to live closer than white patients to high-quality hospitals, patients, rather than true differences in quality of care, they are less likely to have their surgeries there.70 The might explain why hospitals58 and physicians59 caring for intersection of race, racism, and the health-care system in poor patients score lower on some quality metrics than the USA is reviewed elsewhere in this Series. do health-care providers serving affluent areas. Assessment of quality differences is increasingly Health-care financing inequality difficult because tying quality indicators to financial The USA finances medical care through a complex incentives can induce so-called gaming, which distorts network of public and private insurance programmes, as measurement.60 Nonetheless, payers have implemented well as substantial direct payments by patients. Figure 3 pay-for-performance schemes that reward providers on shows the proportion of Americans covered by the main the basis of proxy measures of quality, and facilities insurance programmes, and the major sources that fund serving poor patients have been disproportionately health care. Taken together, government insurance penalised. For example, safety-net hospitals have seen programmes—principally, Medicare, Medicaid, and their payments reduced under Medicare’s Hospital military health care—account for 42% of personal health- Readmission Reductions,61 Hospital-Acquired Condition care expenditures.72 Yet this figure substantially understates Reduction,62 and Hospital Value-Based Purchasing the government’s share, because it excludes two large, tax- programmes.63 Disturbingly, such programmes introduce funded outlays for private insurance: government agencies’ perverse incentives to avoid poor patients, while shrinking expenditures to purchase private insurance for public- funding for hospitals and physicians continuing to care sector employees (representing 28% of all employer for them. payments for private coverage) and tax subsidies for private In view of the pitfalls of quality measurement, what can firms’ purchase of insurance for their employees. Taking be said about the association between social disadvantage into account these two additional categories boosts the and the quality of medical services? A classic study of public share of total health funding in the USA to 65%.73 patients admitted to hospital in 1984 found that uninsured Total health-care expenditures by the government in the patients were at higher risk (odds ratio 2·35) of receiving USA exceed the total public and private spending per head substandard medical care than their insured counterparts.64 of any other country except for Switzerland.73 In light of However, safety-net hospitals (and hospitals in the this fact, the stark inequalities in health care faced by Veterans Administration [VA] system, which serves mostly millions of Americans seem particularly unjust. non-affluent veterans) have risk-adjusted mortality rates The complexity of health-care financing in the USA for older patients similar to those of other hospitals. By obscures not only the magnitude of public funding but contrast, small hospitals serving isolated rural areas appear also the regressive pattern of who ultimately pays. In fact, to deliver a lower quality of care for medical conditions health care takes a substantially larger share of income than other hospitals do, as measured by both process-of- from the poor than from the wealthy, exacerbating care metrics and mortality.65 Studies of differences in inequalities in disposable income.74 Although com­parative surgical quality and safety are inconclusive,66 although international studies are scarce and mostly old, financing risk-adjusted outcomes appear worse for poor patients schemes in other wealthy countries are generally less across a range of surgical procedures.67 Poor patients are regressive (although cost sharing is rising in some more likely than affluent patients to receive dangerous European countries75). Health-care systems financed www.thelancet.com Vol 389 April 8, 2017 1435 Series

primarily through income taxes, as in Ireland, the UK, and noted previously, insured patients often bear a heavy Portugal, tend to be the most progressive, whereas those (and regressive) out-of-pocket burden for deductibles, co- relying on private insurance and out-of-pocket payments, payments, and out-of-network care. Even older patients, as in Switzerland and the USA, are more regressive.76 almost all of whom are covered by Medicare, face high out-of-pocket costs for their share of the premiums, as The redistributive effect of specific health-care well as co-payments and deductibles,78 a burden that falls financing programmes most heavily on low-income senior citizens. For Medicare Direct out-of-pocket spending is the most regressive enrollees, out-of-pocket medical expenses consume form of health-care financing. The uninsured (who are 11·2% of income among those with incomes above 300% disproportionately poor) pay for much of their care out of of the poverty level, and between 22·7% and 26·8% pocket and, because they do not have insurers’ among those with incomes below 200% of the poverty negotiating leverage, are charged the highest prices.77 As level.51 In an effort to reduce the burden of catastrophic medical bills, the ACA imposed limits on out-of-pocket A Insurance enrolment, USA (millions) medical costs in private plans ($6850 per year for individual plans and $13 700 for families in 2016), Yet these limits, which do not apply to out-of-network and so-called non-essential services, vastly exceed most 33·0 79 14·1 families’ savings. Private insurance premiums are also regressive74 and Employment-based private have risen faster than earnings (figure 4); premiums for 55·9 Direct purchase private Medicare employer-based plans increased by approximately three 175·0 Medicaid times between 1999 and 2016.29 The poorest fifth of Military related Americans spend, on average, 6% of their income on Uninsured private insurance premiums, while the wealthiest fifth 50·5 spend just 3·2%.74 Although employers typically make sizeable contributions to their employees’ premium 46·2 costs, economists believe that this expense is mostly passed on to employees in the form of lower wages. Medicaid is the most progressively redistributive health insurance programme in the USA. It requires little cost B Payers, USA (US$ billions) sharing by patients, is financed through federal and state taxes (with progressive income taxes providing the largest share),81 and most of the benefits go to poor citizens. Medicare is funded largely through federal general Private employer payments revenues and a payroll tax, which remains less for private insurance $338·1 $646·2 Household payments for progressive than Medicaid’s funding base (despite the private Insurance ACA’s extension of the payroll tax to some investment Household out-of-pocket income). Medicare covers both affluent and poor senior $345·3 Medicare Government spending Medicaid citizens, but its high and regressive cost-sharing $545·1 Other public requirements discourage many low-income beneficiaries Government payments for 82 public employees' private from seeking care. Moreover, the growing gap in life $484·8 $341·9 health insurance and for expectancy between the rich and the poor means that subsidies for private coverage under the ACA wealthier Americans will, on average, live to enjoy many $240·1 more years of publicly funded benefits after becoming eligible at the age of 65 years.83 As a result, among men born in 1960, lifetime Medicare outlays are expected to be 28% higher for the wealthiest fifth than for the poorest Figure 3: Proportion of Americans covered by the main insurance programmes (A) and main sources of fifth, a reversal of the pattern 30 years earlier.83 Similarly, health-care funding (B) immigrants (especially the undocumented) collectively (A) Source: US Census Bureau, Current Population Survey, 2015.52 Individuals may have had more than one type of insurance. “Medicaid” excludes individuals with both Medicare and Medicaid. (B) Source: National Health contribute billions more in taxes to Medicare each year Expenditure Accounts, US Centers for Medicare & Medicaid Services, 2015.71 The government spending figure than they receive in benefits.84 excludes $326·2 billion in tax subsidies. “Other public” includes maternal and child health, vocational rehabilitation, the Substance Abuse and Mental Health Services Administration, the Indian Health Service, federal workers’ Medical bills and financial hardship compensation, other federal programmes, public health activities, the US Department of Defense, the US Department of Veterans Affairs, the Children’s Health Insurance Program (CHIP), as well as investment programmes The health-care financing system in the USA leaves (research, structures, and equipment), public and general assistance programmes, school health programmes, and millions of Americans facing medical bills that deplete other state and local programmes. ACA=Affordable Care Act. their assets and drive them into debt. One in four non-

1436 www.thelancet.com Vol 389 April 8, 2017 Series

elderly adults younger than 65 years (and one in three 200 Family coverage: deductibles with annual household incomes <$50 000) reported Family coverage: premiums difficulty paying medical bills in 2015; more than half of 180 Workers’ earnings these individuals owe more than $2500.85 People with 160 deductibles higher than $1500 (or families with 140 deductibles >$3000) and worse health than the overall population are particularly at risk85,86 (figure 5), as are 120 African-Americans and Hispanics.86 100

Medical bills are a major contributor to household debt Growth (%) 80 and bankruptcy,14 comprising more than half of all unpaid personal debts sent to collection agencies87 (figure 6). One 60 in ten families with medical bill problems has declared 40 88 bankruptcy. Although the uninsured are at greater risk 20 than the insured of declaring bankruptcy,39 most medical 14 0 bankruptcies involve debtors who are insured. 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 Financial hardship is especially common among Year people with serious illness.89 Among non-elderly adults Figure 4: Growth in American workers’ earnings, premiums, and deductibles for private insurance coverage, with cancer, more than a third borrowed money or went 2002–14 into debt because of their treatment, and 3% filed for Sources: Medical Expenditures Panel Survey43,44 for years 2002–14 (2007 data unavailable), and US Bureau of Labor bankruptcy.90 Such financial catastrophe appears to Statistics, Current Employment Statistics Survey80 (data from April, 2002, to April, 2014). increase mortality for treatable cancers,91 perhaps because it leads to forgone care. Medicare92 and Medicaid93 A longer lens on health reform coverage provide better (although still imperfect) A century ago, medical care in the USA had little protection from financial hardship. influence on health or wealth. Burgeoning medical Medical bills force families to make difficult choices: capabilities and costs in the 20th century made health 34% of insured Americans with difficulty paying medical care an industry ripe for corporate investment and profit. bills were unable to pay for food, heat, or housing, 15% The transformation from a largely charitable service to a took out high-interest payday loans,94 and 42% took on market-driven enterprise ensued. extra jobs or worked additional hours.85 Most people Markets distribute goods on the basis of purchasing reporting problems with medical bills say they have power, and for mostly non-discretionary purchases, skipped or delayed needed medical care.95 Moreover, such as medical care, this results in particularly lopsided defaulting on medical bills and medical bankruptcies financial burdens. In many other wealthy countries, often has long-term repercussions; these blemishes social democratic and labour parties have successfully remain on credit reports for many years, compromising implemented policies that offset these market access to credit, insurance, housing, and employment. tendencies by creating national health programmes, or Although medical costs impoverish many Americans, by tightly regulating private insurers and health-care this issue is not captured by the US Census Bureau’s providers. The USA does not have such a party, perhaps official measure of poverty. In response, the US Census because the low-income voters most buffeted by the Bureau has introduced alternative poverty measures that market are divided by racial animosities. After the subtract medical costs and other mandatory expenses from 2010 passage of the ACA, congressional stalemates income, and add non-cash government aid (eg, housing blocked incremental reforms that might have restrained vouchers). These alternative measures indicate that more medical markets and democratised care. Republicans, Americans experience poverty than are reflected in official now in control of both houses of Congress and the statistics,96 with medical costs being the largest contributor executive branch in Washington, DC, promise to replace to the difference between the official and alternative the ACA with measures that are even more market- measures, pushing an additional 10 million Americans friendly than the ACA, which would tilt care further below the poverty line.97 towards the wealthy. The medical system in the USA also influences The health-care dilemma in the USA, characterised by inequality as an employer of nearly 17 million Americans. unequal access and unfair financing, echoes that of the Although physicians and nurses are generally well paid, 1950s and 1960s, prior to the passage of Medicare and many other health-care workers are not. The health-care Medicaid, when one in four Americans (and half of older system employs more than 20% of all black female people and minorities) did not have health coverage. workers; more than a quarter of these health-care After 8 years of Republican rule, the Democratic workers subsist on family incomes below 150% of the President Kennedy was inaugurated in 1961. The poverty line, and 12·9% of them are uninsured impetus for reform was bolstered by the popular (Himmelstein DU; unpublished analysis of the 2015 mobilisation for civil rights and enabled by the Current Population Survey). Democratic Party’s landslide victory in the 1964 election. www.thelancet.com Vol 389 April 8, 2017 1437 Series

60 53%

50 47% )

40 37%

30 26% 26% 26%

of survey respondents (% 22% 20% 20 15% 14% Proportion

10

0 Not disabled Disabled Lower High Insured Uninsured ≥$100 000 $50 000– <$50 000 Total, adults deductible deductible 99 999 aged 18–64 years

Disability status Plan deductible Insurance status Household income Characteristics

Figure 5: Share of adults in the USA aged 18–64 years reporting problems paying medical bills in 2015, by selected characteristics Reproduced with permission from the Kaiser Family Foundation/New York Times Medical Bills Survey, 2016.85 High deductible defined as higher than $1500 for an individual or higher than $3000 for a family.

60

52·1% 50

debts (%) 40

30 unpaid personal of

20 17·3% Proportion 8·2% 10 7·3% 7·2%

2·2% 1·6% 1·5% 1·2% 0·6% 0·4% 0·3% 0·1% 0·1% 0·1% 0% 0 l n Retai Unio Utilities Banking or cellular Financial Insurance Unclassified Educational Automotive or health care Government Credit Oil company Cable Rental or leasing Personal services Cheque guarantee Medical Type of creditor

Figure 6: Composition of unpaid personal debts sent to collection agencies in the USA, by type of creditor Source: US Consumer Financial Protection Bureau, 2014.87

The 1965 passage of Medicare and Medicaid vastly students, fairer immigration laws, and the establishment improved access to care and desegregated hospitals. But of the National Endowment for the Arts. The these health-care initiatives were just one part of a broad redistributive agenda, dubbed the War on Poverty, also legislative agenda that transformed American society, vastly expanded non-cash benefits such as food including the Civil Rights and Voting Rights Acts, laws subsidies, free pre-school programmes for poor children, that improved opportunities for women in universities and community health centres, and boosted social and the military, the first major federal aid programmes security benefits, lifting 2·5 million senior citizens out for local (particularly poor) public schools and college of poverty. When President Kennedy came into office in

1438 www.thelancet.com Vol 389 April 8, 2017 Series

1961, 40 million Americans were poor. When his Contributors sucessor President Johnson left office in 1969, that SLD, SW, and DUH developed and drafted the paper, contributed to revisions, and approved the final manuscript. number had fallen to 24 million. In 1972, the gap in remaining life expectancy at the age of 60 years was only Declaration of interests DUH and SW were founders of, and remain active in, Physicians for a 1·2 years between men with incomes above and below National Health Program, a group that advocates for a single-payer the median.2 national health insurance system for the USA. DUH and SW served as Half a century on, neoliberal policies have eroded these unpaid advisors to Senator Bernie Sanders during Sanders’ presidential gains. Today, 43·1 million Americans are poor, many campaign. SLD declares no competing interests. health and social inequities persist, and some have References 1 Woolf SH, Aron LY, Dubay L, Simon SM, Zimmerman E, Luk K. worsened. Even after the ACA’s coverage expansion— How are income and wealth linked to health and longevity? reviewed elsewhere in this Series—27 million Americans April 13, 2015. Washington, DC: Urban Institute, 2015. http://www. remain uninsured and, for many with insurance, access to urban.org/research/publication/how-are-income-and-wealth-linked- health-and-longevity (accessed Jan 12, 2017). affordable care remains elusive. At the same time, 2 Waldron H. Trends in mortality differentials and life expectancy for unneeded and even harmful medical interventions remain male Social Security-covered workers, by socieoeconomic status. common (due, in part, to the fragmented health-care Soc Sec Bull 2007; 67: 1–28. 3 Chetty R, Stepner M, Abraham S, et al. The association between delivery system), bureaucracy consumes nearly a third of income and life expectancy in the United States, 2001–2014. JAMA health spending, and wealthy Americans consume a 2016; 315: 1750–66. disproportionate and rising share of medical resources. 4 Hanna-Attisha M, LaChance J, Sadler RC, Champney-Schnepp A. Elevated blood lead levels in children associated with the Flint drinking water crisis: a spatial analysis of risk and public health Conclusion response. Am J Public Health 2016; 106: 283–90. Many physicians in the USA are working to advance 5 Piketty T, Saez E. Inequality in the long run. Science 2014; 344: 838–43. health-care justice. But increased efforts in this direction 6 Saez E. “Income inequality in the United States, 1913–1998” with Thomas Piketty, Quarterly Journal of Economics, 118(1), 2003, 1–39 are needed. The brave cadre of colleagues who face (Longer updated version published in Atkinson AB, and Piketty T, constant threat for delivering abortion services (which eds, Oxford University Press, 2007). Tables and figures updated to are disproportionately needed by poor women) must be 2015 in excel format. June, 2016. https://eml.berkeley.edu/~saez (accessed Feb 1, 2017). supported and augmented, especially as anti-choice 7 Piketty T, Saez E. Income inequality in the United States, politicians now hold sway in Washington, DC. Doctors 1913–1998. Q J Econ 2003; 118: 1–41. should follow the lead of trainees, such as those in the 8 Saez E, Zucman G. Wealth inequality in the United States since 1913: WhiteCoats4BlackLives movement, who have spoken out evidence from capitalized income tax data. Q J Econ 2016; 131: 519–78. For more on 9 Urban Institute. Nine charts about wealth inequality in America. against the structural racism that still tarnishes many WhiteCoats4BlackLives see Washington, DC: Urban Institute, 2015. http://apps.urban.org/ http://www. medical institutions and policies that deny care to features/wealth-inequality-charts (accessed Feb 14, 2017). whitecoats4blacklives.org immigrants. Physicians should reflect on the ways we— 10 Shapiro T, Meschede T, Osoro S. The roots of the widening racial wealth gap: explaining the black-white economic divide. and the institutions we practice within—embrace or February, 2013. Waltham, MA: Institute on Assets and Social Policy, evade the responsibility to care for the disadvantaged. 2013. http://iasp.brandeis.edu/pdfs/Author/shapiro-thomas-m/ Doctors should also join in demanding reforms that racialwealthgapbrief.pdf (accessed Jan 12, 2017). 11 Shaefer HL, Edin K. Rising extreme poverty in the United States move forward, not backward, from the ACA. Republicans and the response of federal means-tested transfer programs. aspire to roll back the law’s coverage expansions, fully Soc Serv Rev 2013; 87: 250–68. privatise Medicare and the VA, and give state 12 OECD. OECD Health Statistics 2016. June 30, 2016. Paris: Organisation for Economic Co-operation and Development, 2016. governments free rein to cut Medicaid—changes that http://www.oecd.org/els/health-systems/health-data.htm (accessed must be resisted to avoid a public health disaster. Jan 12, 2017). However, Hillary Clinton’s 2016 presidential election 13 Collins SR, Rasmussen PW, Doty MM, Beutel S. The rise in health care coverage and affordability since health reform took effect: defeat suggests that defending the health-care status quo findings from the Commonwealth Fund Biennial Health Insurance cannot win the day. Moreover, proposals for incremental Survey, 2014. Commonwealth Fund Issue Brief 2015; 2: 1–16. steps that could cushion the worst inequities—such as 14 Himmelstein DU, Thorne D, Warren E, Woolhandler S. Medical tighter insurance regulations, allowing a government bankruptcy in the United States, 2007: results of a national study. Am J Med 2009; 122: 741–46. insurance plan to compete in the market, regulation of 15 Schoen C, Osborn R, Squires D, Doty MM. Access, affordability, drug prices, and extension of public coverage to and insurance complexity are often worse in the United States immigrants—failed to excite voters. Until November, 2016, compared to ten other countries. Health Aff2013; 32: 2205–15. 16 Davis K, Ballreich J. Equitable access to care—how the mainstream politicians and pundits deemed such steps United States ranks internationally. N Engl J Med 2014; 371: 1567–70. politically feasible, and more thoroughgoing reform 17 Schoen C, Radley DC, Riley P, et al. Health care in the two unattainable. Now, a more inspiring and egalitarian Americas: findings from the scorecard on state health system performance for low-income populations, 2013. Sept 18, 2013. vision—a health-care reform that address the problems New York, NY: The Commonwealth Fund, 2013. http://www. felt by most insured Americans—seems a more effective commonwealthfund.org/publications/fund-reports/2013/sep/low- rebuttal to the Republican mantra of “Repeal and replace income-scorecard (accessed Jan 13, 2017). [the ACA]”. A bolder step towards health-care equality— 18 Wilper AP, Woolhandler S, Lasser KE, McCormick D, Bor DH, Himmelstein DU. A national study of chronic disease prevalence straight on to universal public insurance—could offer the and access to care in uninsured U.S. adults. Ann Intern Med 2008; best way forward.98 149: 170–76. www.thelancet.com Vol 389 April 8, 2017 1439 Series

19 Li Y-J, Xirasagar S, Pumkam C, Krishnaswamy M, Bennett CL. 41 Centers for Disease Control and Prevention. State-specific Vision insurance, eye care visits, and vision impairment among prevalence estimates of uninsured and underinsured persons— working-age adults in the United States. JAMA Ophthalmol 2013; Behavioral Risk Factor Surveillance System, 1995. 131: 499–506. MMWR Morb Mortal Wkly Rep 1998; 47: 51–55. 20 Garfield RL, Damico A. Medicaid expansion under health reform 42 Dickman S, Woolhandler S, Bor J, McCormick D, Bor D, may increase service use and improve access for low-income adults Himmelstein D. Health spending for low, middle, and high-income with diabetes. Health Aff2012; 31: 159–67. Americans, 1963–2012. Health Aff 2016; 35: 1189–96. 21 Roll JM, Kennedy J, Tran M, Howell D. Disparities in unmet need 43 Agency for Healthcare Research and Quality. Medical Expenditures for mental health services in the United States, 1997–2010. Panel Survey. Table I.D.2. Average total employee contribution (in Psychiatr Serv 2013; 64: 80–82. dollars) per enrolled employee for family coverage at private-sector 22 Berenson J, Doty MM, Abrams MK, Shih A. Achieving better establishments that offer health insurance by firm size and selected quality of care for low-income populations: the roles of health characteristics: United States. Rockville, MD: Agency for Healthcare insurance and the medical home in reducing health inequities. Research and Quality, 2002–14. https://meps.ahrq.gov/mepsweb/ Commonwealth Fund Issue Brief 2012; 11: 1–18. data_stats/quick_tables_results.jsp?component=2&subcomponent=1 23 Sommers BD, Baicker K, Epstein AM. Mortality and access to care &year=-1&tableSeries=1&tableSubSeries=CDE&searchText=I.D.2&sea among adults after state Medicaid expansions. N Engl J Med 2012; rchMethod=1&Action=Search (accessed Jan 16, 2017). 367: 1025–34. 44 Agency for Healthcare Research and Quality. Medical Expenditures 24 Baicker K, Taubman SL, Allen HL, et al. The Oregon experiment— Panel Survey. Table I.F.3. Average family deductible (in dollars) per effects of Medicaid on clinical outcomes.N Engl J Med 2013; employee enrolled with family coverage in a health insurance plan 368: 1713–22. that had a deductible at private-sector establishments by firm size and selected characteristics: United States. Rockville, MD: Agency 25 Sommers BD. Health care reform’s unfinished work—remaining for Healthcare Research and Quality, 2002–14. https://meps.ahrq. barriers to coverage and access. N Engl J Med 2015; 373: 2395–97. gov/mepsweb/data_stats/quick_tables_results.jsp?component=2&s 26 Stimpson JP, Wilson FA, Su D. Unauthorized immigrants spend ubcomponent=1&year=-1&tableSeries=1&tableSubSeries=F&search less than other immigrants and US natives on health care. Text=I.F.3&searchMethod=1&Action=Search (accessed Feb 14, 2017). Health Aff2013; 32: 1313–18. 45 Corscadden L, Allin S, Wolfson M, Grignon M. Publicly financed 27 Rosenbaum S. Medicaid payments and access to care. N Engl J Med healthcare and income inequality in Canada. Healthc Q 2014; 17: 7–10. 2014; 371: 2345–47. 46 Casey MM, Thiede Call K, Klingner JM. Are rural residents less 28 Iobst C, Arango D, Segal D, Skaggs DL. National access to care for likely to obtain recommended preventive healthcare services? children with fractures. J Pediatr Orthop 2013; 33: 587–91. Am J Prev Med 2001; 21: 182–88. 29 Claxton G, Rae M, Long M, Damico A, Sawyer B. 2016 Employer 47 Aboagye JK, Kaiser HE, Hayanga AJ. Rural-urban differences in access health benefits survey. Sept 14, 2016. Menlo Park, CA: Kaiser Family to specialist providers of colorectal cancer care in the United States: Foundation, 2016. http://kff.org/health-costs/report/2016-employer- a physician workforce issue. JAMA Surg 2014; 149: 537–43. health-benefits-survey (accessed Jan 13, 2017). 48 White K, Hopkins K, Aiken AR, et al. The impact of reproductive 30 Rae M, Levitt L, Claxton G, Cox C, Long M, Damico M. Patient health legislation on family planning clinic services in Texas. cost-sharing in Marketplace plans, 2016. Nov 13, 2015. Menlo Park, Am J Public Health 2015; 105: 851–58. CA: Kaiser Family Foundation, 2015. http://kff.org/health-costs/ 49 Stevenson AJ, Flores-Vazquez IM, Allgeyer RL, Schenkkan P, issue-brief/patient-cost-sharing-in-marketplace-plans-2016 (accessed Potter JE. Effect of removal of Planned Parenthood from the Texas Jan 13, 2017). Women’s Health Program. N Engl J Med 2016; 374: 1298. 31 Schleicher SM, Mullangi S, Feeley TW. Effects of narrow 50 Health Care Cost Institute. Out-of-pocket spending trends (2013). networks on access to high-quality cancer care. JAMA Oncol 2016; Issue Brief #9. October, 2014. Washington, DC: Health Care Cost 2: 427–28. Institute, 2014. http://www.healthcostinstitute.org/files/IB%20 32 Brot-Goldberg ZC, Chandra A, Handel BR, Kolstad JT. What does a 9%2010-28-14.pdf (accessed Jan 13, 2017). deductible do? The impact of cost-sharing on health care prices, 51 Noel-Miller C. Medicare beneficiaries’ out-of-pocket spending for quantities, and spending dynamics. October, 2015. Cambridge, MA: health care. October, 2015. Washington, DC: AARP Public Policy National Bureau of Economic Research, 2015. http://www.nber.org/ Institute, 2011. http://www.aarp.org/content/dam/aarp/ppi/2015/ papers/w21632 (accessed Jan 13, 2017). medicare-beneficiaries-out-of-pocket-spending-for-health-care.pdf 33 Enterline PE, Salter V, McDonald AD, McDonald JC. (accessed Jan 13, 2017). The distribution of medical services before and after “free” medical 52 US Census Bureau. Current Population Survey, 2015 Annual Social care—the Quebec experience. N Engl J Med 1973; 289: 1174–78. and Economic Supplement. PINC-01. Selected characteristics of 34 Fisher ES, Wennberg JE. Health care quality, geographic variations, people 15 years and over, by total money income, work experience, and the challenge of supply-sensitive care. Perspect Biol Med 2003; race, Hispanic origin, and sex. https://www.census.gov/data/tables/ 46: 69–79. time-series/demo/income-poverty/cps-pinc/pinc-01.2014.html 35 Trivedi AN, Moloo H, Mor V. Increased ambulatory care (accessed Feb 14, 2017). copayments and hospitalizations among the elderly. N Engl J Med 53 Huskamp HA, Iglehart JK. Mental health and substance-use 2010; 362: 320–28. reforms—milestones reached, challenges ahead. N Engl J Med 2016; 36 Choudhry NK, Bykov K, Shrank WH, et al. Eliminating medication 375: 688–95. copayments reduces disparities in cardiovascular care. 54 Creedon TB, Cook BL. Access to mental health care increased but Health Aff 2014; 33: 863–70. not for substance use, while disparities remain. Health Aff 2016; 37 Karaca-Mandic P, Jena AB, Joyce GF, Goldman DP. Out-of-pocket 35: 1017–21. medication costs and use of medications and health care services 55 Morrison F, Shubina M, Turchin A. Encounter frequency and among children with asthma. JAMA 2012; 307: 1284–91. serum glucose level, blood pressure, and cholesterol level control in 38 Fung V, Graetz I, Galbraith A, et al. Financial barriers to care patients with diabetes mellitus. Arch Intern Med 2011; 171: 1542–50. among low-income children with asthma: health care reform 56 Smolderen KG, Spertus JA, Nallamothu BK, et al. Health care implications. JAMA Pediatr 2014; 168: 649–56. insurance, financial concerns in accessing care, and delays to 39 Collins SR, Rasmussen PW, Beutel S, Doty MM. The problem of hospital presentation in acute myocardial infarction. JAMA 2010; underinsurance and how rising deductibles will make it worse— 303: 1392–1400. findings from the Commonwealth Fund Biennial Health Insurance 57 Kebede S, Shihab HM, Berger ZD, Shah NG, Yeh HC, Brotman DJ. Survey. May 20, 2015. New York, NY: The Commonwealth Fund, Patients’ understanding of their hospitalizations and association 2015. http://www.commonwealthfund.org/publications/issue- with satisfaction. JAMA Intern Med 2014; 174: 1698–700. briefs/2015/may/problem-of-underinsurance (accessed 58 Arbaje AI, Wolff JL, Yu Q, Powe NR, Anderson GF, Boult C. Jan 13, 2017). Postdischarge environmental and socioeconomic factors and the 40 Magge H, Cabral HJ, Kazis LE, Sommers BD. Prevalence and likelihood of early hospital readmission among predictors of underinsurance among low-income adults. community-dwelling Medicare beneficiaries.Gerontologist 2008; J Gen Intern Med 2013; 28: 1136–42. 48: 495–504.

1440 www.thelancet.com Vol 389 April 8, 2017 Series

59 Hong CS, Atlas SJ, Chang Y, et al. Relationship between patient 82 Schoen C, Buttorff C, Andersen M, Davis K. Policy options to panel characteristics and primary care physician clinical expand Medicare’s low-income provisions to improve access and performance rankings. JAMA 2010; 304: 1107–13. affordability.Health Aff2015; 34: 2086–94. 60 Campbell SM, Reeves D, Kontopantelis E, Sibbald B, Roland M. 83 Committee on the Long-Run Macroeconomic Effects of the Aging Effects of pay for performance on the quality of primary care in U.S. Population—Phase II, Committee on Population, Division of England. N Engl J Med 2009; 361: 368–78. Behavioral and Social Sciences and Education, Board on 61 Sheingold SH, Zuckerman R, Shartzer A. Understanding Medicare Mathematical Sciences and Their Applications, Division on hospital readmission rates and differing penalties between Engineering and Physical Sciences, The National Academies of safety-net and other hospitals. Health Aff 2016; 35: 124–31. Sciences, Engineering, and Medicine. The growing gap in life 62 Rajaram R, Chung JW, Kinnier CV, et al. Hospital characteristics expectancy by income: implications for federal programs and policy associated with penalties in the Centers for Medicare & Medicaid responses. Sept 17, 2015. Washington, DC: National Academies Services Hospital-Acquired Condition Reduction Program. JAMA Press, 2015. http://www.ncbi.nlm.nih.gov/books/NBK321304 2015; 314: 375–83. (accessed Jan 13, 2017). 63 Ryan AM. Will value-based purchasing increase disparities in care? 84 Zallman L, Woolhandler S, Himmelstein D, Bor D, McCormick D. N Engl J Med 2013; 369: 2472–74. Immigrants contributed an estimated $115.2 billion more to the Medicare Trust Fund than they took out in 2002–09. Health Aff2013; 64 Burstin HR, Lipsitz SR, Brennan TA. Socioeconomic status and risk 32: 1153–60. for substandard medical care. JAMA 1992; 268: 2383–87. 85 Hamel L, Norton M, Pollitz K, Levitt L, Claxton G, Brodie M. 65 Joynt KE, Harris Y, Orav EJ, Jha AK. Quality of care and patient The burden of medical debt: results from the Kaiser Family outcomes in critical access rural hospitals. JAMA 2011; 306: 45–52. Foundation/New York Times Medical Bills Survey. Jan 5, 2016. 66 Mouch CA, Regenbogen SE, Revels SL, Wong SL, Lemak CH, Menlo Park, CA: Kaiser Family Foundation, 2016. http://kff.org/ Morris AM. The quality of surgical care in safety net hospitals: a health-costs/report/the-burden-of-medical-debt-results-from-the- systematic review. Surgery 2014; 155: 826–38. kaiser-family-foundationnew-york-times-medical-bills-survey 67 Qasim M, Andrews RM. Despite overall improvement in surgical (accessed Jan 13, 2017). outcomes since 2000, income-related disparities persist. 86 Collins SR, Rasmussen PW, Doty MM, Beutel S. Too high a price: Health Aff 2013; 32: 1773–80. out-of-pocket health care costs in the United States. 68 Smedley BD, Stith AY, Nelson AR. Unequal treatment: confronting Commonwealth Fund Issue Brief 2014; 29: 1–14. racial and ethnic disparities in health care. Washington, DC: 87 US Consumer Financial Protection Bureau. Consumer credit National Academy Press, 2003. reports: a study of medical and non-medical collections. December, 69 Trivedi AN, Nsa W, Hausmann LR, et al. Quality and equity of care 2014. http://files.consumerfinance.gov/f/201412_cfpb_reports_ in U.S. hospitals. N Engl J Med 2014; 371: 2298–308. consumer-credit-medical-and-non-medical-collections.pdf (accessed 70 Dimick J, Ruhter J, Sarrazin MV, Birkmeyer JD. Black patients Jan 13, 2017). more likely than whites to undergo surgery at low-quality hospitals 88 Sommers A, Cunningham PJ. Medical bill problems steady for U.S. in segregated regions. Health Aff 2013; 32: 1046–53. families, 2007–2010. Track Rep 2011; 28: 1–5. 71 Centers for Medicare & Medicaid Services. National Health 89 Whitney RL, Bell JF, Reed SC, et al. Predictors of financial Expenditure Accounts. Baltimore: US Centers for Medicare & difficulties and work modifications among cancer survivors in the Medicaid Services, 2017. https://www.cms.gov/research-statistics- United States. J Cancer Surviv 2016; 10: 241–50. data-and-systems/statistics-trends-and-reports/ 90 Banegas MP, Guy GP Jr, de Moor JS, et al. For working-age cancer nationalhealthexpenddata/nationalhealthaccountshistorical.html survivors, medical debt and bankruptcy create financial hardships. (accessed Feb 14, 2017). Health Aff2016; 35: 54–61. 72 Martin AB, Hartman M, Benson J, Catlin A, for the National Health 91 Ramsey SD, Bansal A, Fedorenko CR, et al. Financial insolvency as Expenditure Accounts Team. National health spending in 2014: a risk factor for early mortality among patients with cancer. faster growth driven by coverage expansion and prescription drug J Clin Oncol 2016; 34: 980–86. spending. Health Aff2016; 35: 150–60. 92 Yabroff KR, Dowling EC, Guy EP Jr, et al. Financial hardship 73 Himmelstein DU, Woolhandler S. The current and projected associated with cancer in the United States: findings from a taxpayer shares of US health costs. Am J Public Health 2016; population-based sample of adult cancer survivors. J Clin Oncol 106: 449–52. 2016; 34: 259–67. 74 Ketsche P, Adams EK, Wallace S, Kannan VD, Kannan H. 93 Hu L, Kaestner R, Mazumder B, Miller S, Wong A. The effect of the Lower-income families pay a higher share of income toward Patient Protection and Affordable Care Act Medicaid expansions on national health care spending than higher-income families do. financial well-being. April, 2016. Cambridge, MA: National Bureau Health Aff2011; 30: 1637–46. of Economic Research, 2016. http://www.nber.org/papers/w22170 75 Palladino R, Lee JT, Hone T, Filippidis FT, Millett C. The Great (accessed Jan 13, 2017). Recession and increased cost sharing in European health systems. 94 Bickham T, Lim Y. In sickness and in debt: do mounting medical Health Aff2016; 35: 1204–13. bills predict payday loan debt? Soc Work Health Care 2015; 76 Van Doorslaer E, Wagstaff A, et al. The redistributive effect of 54: 518–31. health care finance in twelve OECD countries.J Health Econ 1999; 95 Baughman KR, Burke RC, Hewit MS, Sudano JJ, Meeker J, 18: 291–313. Hull SK. Associations between difficulty paying medical bills and 77 Anderson GF. From ‘soak the rich’ to ‘soak the poor’: recent trends forgone medical and prescription drug care. Popul Health Manag in hospital pricing. Health Aff2007; 26: 780–89. 2015; 18: 358–66. 78 Cubanski J, Lyons B, Neuman T, et al. Medicaid and Medicare 96 Hutto N, Waldfogel J, Kaushal N, Garfinkel I. Improving the trends and challenges. JAMA 2015; 314: 329. measurement of poverty. Soc Serv Rev 2011; 85: 39–74. 79 Claxton G, Rae M, Panchal N. Consumer assets and patient cost 97 Collins SR. New census poverty measure shows medical expenses sharing. March 11, 2015. Menlo Park, CA: Kaiser Family push 10 million more Americans into poverty. Nov 10, 2011. Foundation, 2015. http://kff.org/health-costs/issue-brief/consumer- http://www.commonwealthfund.org/publications/blog/2011/nov/ assets-and-patient-cost-sharing (accessed Jan 13, 2017). new-census-poverty-measure (accessed Jan 13, 2017). 80 US Bureau of Labor Statistics. Current Employment Statistics 98 Gaffney A, Woolhandler S, Angell M, Himmelstein D. Moving Survey. Washington, DC: US Bureau of Labor Statistics, 2002–14. forward from the Affordable Care Act to a single-payer system. 81 Davis C, Davis K, Gardner M, et al. Who pays? A distributional Am J Public Health 2016; 106: e1–2. analysis of the tax systems in all 50 states, 5th edn. Washington, DC: Institute on Taxation & Economic Policy, 2015.

www.thelancet.com Vol 389 April 8, 2017 1441 Series

America: Equity and Equality in Health 2 The Affordable Care Act: implications for health-care equity

Adam Gaffney, Danny McCormick

Lancet 2017; 389: 1442–52 Inequalities in medical care are endemic in the USA. The Affordable Care Act (ACA), passed in 2010 and fully See Editorial page 1369 implemented in 2014, was intended to expand coverage and bring about a new era of health-care access. In this See Comment pages 1376 review, we evaluate the legislation’s impact on health-care equity. We consider the law’s coverage expansion, insurance and 1378 market reforms, cost and affordability provisions, and delivery-system reforms. Although the ACA improved coverage This is the second in a Series of and access—particularly for poorer Americans, women, and minorities—its overall impact was modest in comparison five papers about equity and with the gaps present before the law’s implementation. Today, 29 million people in the USA remain uninsured, and equality in health in the USA substantial inequalities in access along economic, gender, and racial lines persist. Although most Americans agree Division of Pulmonary and Critical Care Medicine that further reform is needed, the proper direction for reform—especially following the 2016 presidential election—is (A Gaffney MD) and Division of highly contentious. We discuss proposals for change from opposite sides of the political spectrum, together with their Social and Community Medicine potential impact on health equity. (D McCormick MD), Department

of Medicine, Cambridge Health 3,4 Alliance, Cambridge, MA, USA; Introduction them. However, the USA entered the 21st century as an and Harvard Medical School, Inequalities in health are rife in the USA. Disparities in outlier among high-income nations—the only one Boston, MA, USA (A Gaffney, life expectancy between the wealthy and the poor are without universal health coverage. The milestone D McCormick) growing.1 Pernicious race-based health inequalities, the Affordable Care Act (ACA), signed into law by President Correspondence to: consequence of centuries of repression and exclusion, Obama in 2010, marked the beginning of a new era Dr Adam Gaffney, Division of Pulmonary and Critical Care endure. Meanwhile, middle-aged white Americans, of expanded health-care access. As a result of the Medicine, Department of particularly those with less education, have experienced 2016 presidential election, however, it is unclear whether, Medicine, Cambridge Health an unprecedented rise in mortality.2 Although a complex and to what extent, the ACA’s reforms will persist—or be Alliance, Cambridge, MA 02139, array of social factors underlies these inequalities, undone. USA [email protected] equitable access to health care could help ameliorate Unlike landmark health-care reforms in other English- speaking nations, which created new universal, public

Key messages Search strategy and selection criteria • The Affordable Care Act (ACA) has nearly halved the share of Americans without coverage. Disadvantaged groups including the poor, African-Americans, and Hispanics For information on the impact of the Affordable Care Act saw gains, although stark health-care inequalities remain. (ACA) on health-care coverage, access, use, and disparities, • Despite the ACA, financial barriers to care persist and might be worsening. High cost we relied on major health-care survey reports or publications sharing in the form of copayments, deductibles, and co-insurance obstructs access to based on them, including the National Health Insurance care and frequently leads to financial distress, or even ruin. Survey, the Commonwealth Fund Biennial Health Insurance • The ACA expanded Medicaid—a public programme for the poor—improving access for Survey, the Commonwealth Fund Affordable Care Act millions. However, 19 states opted out of the expansion, and limitations in the Tracking Survey, Gallup Healthways, the RAND Health Reform Medicaid programme relegate many enrollees to a lower tier of access. Opinion Survey, the American Community Survey, the Urban • The law improved access to contraception by mandating coverage for insured women. Institute’s Health Reform Monitoring Survey, and the Meanwhile, several states have imposed new restrictions that curtail access to family Employer Benefits Survey of the Kaiser Family Foundation. planning and abortion. We also identified articles published in English about health • The ACA included a number of value-based delivery system provisions. disparities and the ACA in PubMed and the Web of Science by Their equity implications are not yet clear. However, some such pay-for-performance searching for the keywords “Affordable Care Act” in programmes might siphon resources from hospitals and doctors that care for conjunction with one of the following: “disparity”, disadvantaged populations. “disparities”, “inequality”, “inequalities”, or “inequity” for the • The 2016 election altered the political landscape. Republicans have proposed dates Jan 1, 2010, to Oct 4, 2016. We did additional ad-hoc market-based health reforms that would weaken protections for those with so-called searches to locate references relating to gender or sex and pre-existing conditions, decrease federal funding for Medicaid, further privatise racial inequalities and the ACA. We also identified relevant Medicare, and replace the ACA’s subsidies with more regressive tax credits. If references from the reference lists of selected papers, from implemented, these policies would exacerbate health-care inequality. the libraries of the authors, and through the • By contrast, by providing comprehensive coverage to everyone in the USA without cost recommendations of expert colleagues. Owing to limitations sharing, single-payer reform could reduce health-care inequalities. Although the on the numbers of references and the broad scope of the political climate is currently adverse to such a reform, it is supported by most Americans, topic, references were selected after consideration of their and might prove to be the most tenable proposal for change as political winds shift. importance and originality.

1442 www.thelancet.com Vol 389 April 8, 2017 Series

Provision Estimated number affected

Coverage expansion Dependant Family insurance plans must cover dependants up to age 26 years. 3 million people aged 19–25 years gained coverage.9 coverage provision Medicaid Medicaid eligibility expanded to include all people with incomes below 138% of the 13 million people gained coverage.10 expansion federal poverty level.* However, states can opt out as a result of a 2012 Supreme Court decision. 100% of the costs of the expansion paid for by the federal government initially, gradually reduced to 90% by 2020. Individual Individuals without coverage must buy a private plan or pay a tax. Certain exceptions 8·2 million people might become uninsured without this provision.11 mandate apply, including for economic hardship. Employer mandate Employers with 50 or more full-time employees must provide coverage that meets Between 200 000 and 1 million people might lose employer-sponsored certain criteria (see “cost and affordability” below). Those that do not offer coverage insurance without this provision.13 to >95% of their full-time employees and dependant children are taxed at US$2160 per year per employee after the first 30 full-time employees, assuming one or more full-time employees purchase a subsidised marketplace plan. Those that provide plans that do not meet criteria for affordability and minimum value are taxed at $3240 per year for each employee that receives a premium tax credit, but no more than $2160 per employee after the first 30 employees.12 Health insurance market reforms Marketplaces Sets up online marketplaces for the sale of insurance plans to individuals and small 12 million people gained coverage through the health insurance exchanges.10 businesses. Plans sold on the marketplaces must cover ten essential health benefits. Plans purchased on the marketplaces are eligible for premium and cost-sharing subsidies (see “cost and affordability” below). Guaranteed issue Insurers cannot base premiums on health (eg, pre-existing conditions) or gender. Premiums Uncertain of plans can be set by age (up to three times difference in premiums between young and old enrollees), smoking status, benefits tier (see “cost and affordability” below), location, and whether plan is for an individual or a family. Benefit limits Prohibits annual and lifetime limits on benefits. 105 million people had plans with lifetime limits pre-ACA.14 Preventive health Specified preventive health-care services provided without cost sharing. 76 million people became eligible for free preventive services.15 services Provisions Outlaws gender rating, in which women are charged higher premiums than are men. 8·7 million women might benefit from expanded maternity care coverage and impacting gender Maternity care considered an essential health benefit. Some reproductive health-care 45 million women with expanded preventive services access as of 2012.16 equity services (including contraception) considered preventive, and hence must be provided without cost sharing. Cost and affordability Employer- To meet the employer mandate requirement, plans must meet affordability criteria 155 million people had employer-sponsored coverage in 2016.10 sponsored plans (premium ≤9·66% of income), benefit criteria, and minimum value criteria (plan covers at least 60% of total medical expenses on average).12 Affordability criteria pertain only to individual, not family, coverage. Thus, cost of family coverage might exceed 9·66% threshold. Marketplace plans Premium subsidies available on a sliding scale to those earning between 100% and 400% of 10 million people receive premium subsidies.10 7·2 million people receive the federal poverty level. Additional cost-sharing subsidies that reduce copayments and cost-sharing subsidies that increase actuarial value; these plans also have lower deductibles available to those earning up to 250% of the federal poverty level. Sets following out-of-pocket maximums.17 However, a lower court recently ruled in favour of a actuarial values for plans: 90% for platinum plans, 80% for gold plans, 70% for silver plans, Congressional Republican challenge to the legality of these subsidies, a ruling and 60% for bronze plans. Sets out-of-pocket maximums: for 2017, these were $14 300 for that would end these subsidies unless it is overturned by a higher court. family plans and $7150 for individual plans, with lower limits for those earning less than 250% of the federal poverty level. Some employers with less than 25 full-time employees eligible for subsidies for tax credits to buy insurance on small business marketplaces.12 Delivery system reforms Pay-for- Financial rewards and penalties for hospitals based on quality metrics—eg, Hospital Uncertain performance Readmissions Reduction Program, Hospital Value-Based Purchasing Program, and reforms Hospital-Acquired Condition Reduction programme. Accountable care Encourages providers to form ACOs (responsible for a panel of Medicare enrollees) that Uncertain organisations share in savings with Medicare if they reduce expenditures—eg, Medicare Shared Savings (ACOs) Program, Pioneer ACOs. Provisions Increased funding for community health centres. Reduced funding for Disproportionate Uncertain affecting Share Hospital programme. safety-net institutions

ACA=Affordable Care Act.*All undocumented immigrants and many legal immigrants are excluded.

Table 1: Summary of selected ACA provisions

www.thelancet.com Vol 389 April 8, 2017 1443 Series

See Online for infographic programmes, the ACA sought to expand access by Coverage expansion under the ACA www.thelancet.com/ building on the existing financing system. To this end, Universal coverage is a foundation of equitable health infographics/us-health the ACA expanded the Medicaid programme for the poor care. Without it, the uninsured face the unremitting and private insurance for others, while mandating threat of financial ruin from illness or accident. Yet the changes in insurance regulation and provider payment hazard is not only monetary; uninsurance is associated methods. with increased mortality,18,19 poor overall health, and an We explore the reforms of the ACA using the theoretical increase in depressive symptoms and adverse framework of health-care equity: the perspective that care cardiovascular outcomes.20 Moreover, unequal health- should be available to all on an equal basis.5 Health-care care access often exacerbates health inequalities that equity can be construed along three axes: equitable arise from poverty, racial discrimination, and poor access, equitable use (for those with similar needs), and education. equitable quality.5 Because many people do not have The ACA expanded insurance coverage in two major insurance—and because even those with insurance face ways. First, it broadened eligibility for Medicaid, an high costs when they seek care—health-care access has insurance programme for the poor funded by federal and long been grossly inequitable in the USA. In 2010, for state taxes, which is administered by state governments. instance, 48·6 million people were uninsured.6 The Prior to the ACA, many states limited Medicaid eligibility, uninsured were mostly of low or middle income, and covering only very poor people in specific categories disproportionately black or Hispanic;6,7 the ACA was (eg, children and some of their parents, disabled people, expected to particularly benefit these groups.8 and pregnant women). By contrast, under the ACA, all Here, we review the ACA’s coverage expansions, citizens with incomes up to 138% of the federal poverty insurance market reforms, cost and affordability provisions, level became eligible for Medicaid coverage commencing and delivery system reforms (see table 1 for an overview). in 2014. We explore the multifold ways in which the ACA moved Second, the ACA mandated that uninsured citizens the US health system towards increased equity across purchase private insurance (or pay a fine), and offered income, gender, sex, and race and ethnic groups. At the some subsidies to those with incomes between 100% same time, we document the ways in which it has fallen and 400% of the poverty level to offset the costs of short. We conclude with an assessment of the reform’s insurance premiums. The law also established online prospects in light of the 2016 presidential election, and of insurance exchanges or marketplaces through which proposals for change moving forward. individuals could purchase regulated, subsidised, and

100 Number (millions) Percentage 90

80

70

60 uninsured people of 50

40

30 Number and percentage 20

10

0 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 Year Number (millions) 41·0 39·3 38·7 41·3 40·2 41·5 43·6 42·5 41·2 43·6 43·1 43·8 46·3 48·6 46·3 45·5 44·8 36·0 28·6 Percentage 15·4 14·6 14·2 14·9 14·3 14·7 15·2 14·7 14·2 14·8 14·5 14·7 15·4 16·0 15·1 14·7 14·4 11·5 9·1

Figure 1: Number and percentage of people uninsured in the USA, 1997–2015 Uninsured defined as uninsured at time of interview. Two different methods were used in 2004 to calculate the number of insured; the estimates are similar, and this chart uses the first method. These data rely on surveys of civilian, non-institutionalised households. See source21 for further details on methodology.

1444 www.thelancet.com Vol 389 April 8, 2017 Series

standardised plans. Additionally, the ACA imposed an 50 Poor employer mandate that required large employers to offer Near-poor health coverage to their employees or pay a fine.12,13 45 Non-poor Finally, the law required that family insurance policies 40 cover children younger than 26 years of age. 35 Most coverage expansion provisions became effective in 2014, and evidence suggests that they improved health- 30 care equity in several domains (table 1). For instance, the uninsured 25 number of uninsured individuals fell from about 20 48·6 million (16% of the US population) in 2010 to about Percentage 28·6 million (9·1%) in 2015 (figure 1).21–25 Most of this 15 improvement has been the result of the Medicaid 10 expansion. Of the 22 million who gained insurance as a 5 result of the ACA, 13 million were insured through Medicaid.10 At the same time, several measures of 0 2010 2011 2012 2013 2014 2015 access, use, and health outcomes have also improved. Year After implementation of the ACA, more people reported that they could afford medications and had a physician, Figure 2: Uninsurance rates in the USA by poverty status for individuals younger than 65 years fewer reported problems affording care, and self-reported Uninsured defined as uninsured at time of interview. Poor defined as family income less than 100% of the federal poverty level (FPL), near-poor as greater than or equal to 100% and less than 200% FPL, and non-poor as at 23 health status improved. Along similar lines, the least 200% FPL. Data for those with unknown poverty status not shown here. See source6 for details on methodology. percentage of non-elderly (ie, <65 years of age) adults who avoided needed care because of cost fell from 43% in 2012 to 36% in 2014.25 Meanwhile, the percentage of non-elderly undocumented immigrants are excluded from the ACA’s adults who had difficulties with medical bills or medical insurance expansion. debt declined from 41% to 35% over these 2 years.25 Additionally, inequalities by economic class also persist. As had been hoped, insurance coverage and metrics of Among non-elderly adults, 7·6% of the non-poor were access and use particularly improved for the poor uninsured in 2015 compared with 25·2% of the poor and (<100% federal poverty level) and the near-poor (≥100% 24·1% of the near-poor (figure 2).6 Despite afore­mentioned and <200% of the federal poverty level),6 blacks and improvements, many Americans remain unable to afford Hispanics,6,23,26–28 women of childbearing age, and young care. In 2014, 23% of non-elderly adults reported having families.29–31 The requirement that family policies cover skipped a medical visit, and 19% were unable to fill a children aged 18–25 years—a population that has long prescription or undergo a recommended treatment or test had especially high rates of uninsurance—was because of cost.25 Low-income people, unsurprisingly, fare implemented in 2010 and modestly reduced uninsurance even worse.25 In 2015, 44·8% of non-elderly low-income rates32–34 and increased access32 and use35 among young adults had unmet medical needs owing to costs, and adults. However, despite this provision, the uninsurance 24·2% faced difficulties with medical bills.38 rate in 2014 among adults aged 19–25 years was The so-called narrow networks of doctors and hospitals still 17·1%36—substantially higher than that of the overall covered by most marketplace plans also impede population. equitable health-care access. Narrow networks allow Despite this progress, the gains in coverage and access insurers to control costs: they use the threat of exclusion have been modest compared with the size of the gaps prior from networks to extract price concessions from doctors to the law’s implementation. Further gains are not on the and hospitals.39 However, narrow networks can also be horizon: according to the Congressional Budget Office, used by insurers to exclude providers (eg, quaternary even if the ACA were not altered or repealed, 28 million referral centres) that are likely to attract expensively ill people would remain uninsured in 2024 and beyond.10 and hence unprofitable enrollees. In 2015, for instance, Moreover, several metrics indicate that health-care among the dozens of insurance plans offered on access remains inequitable. With respect to racial New York’s marketplace, only one—from a consumer disparities, in 2015 the uninsurance rate among adults cooperative that subsequently suffered financial aged 18–64 years remained far higher for Hispanics (27·7%) collapse—covered care at the city’s leading cancer centre, and non-Hispanic blacks (14·4%) than for non-Hispanic Memorial Sloan Kettering, whereas no plans included whites (8·7%),6 showing that long-standing racial and the prominent New York Presbyterian system.40 ethnic inequalities have been attenuated but by no means Although some people see narrow networks as a eliminated.27,28 Gains in mental health and substance mechanism for cost control and—theoretically—quality abuse care were even weaker. Blacks, for instance, improvement,41 these networks are sometimes too sparse saw no increase in mental health treatment following to provide access to needed specialty care.42,43 implementation of the ACA, whereas no racial group saw The ACA’s Medicaid expansion, which targeted the an increase in substance abuse treatment.37 Finally, poor and near-poor (many of whom are racial or ethnic www.thelancet.com Vol 389 April 8, 2017 1445 Series

minorities), had great potential to improve health equity. expansion, the ACA included provisions intended to However the Supreme Court’s 2012 decision allowing reduce these discriminatory practices (table 1). states to opt out of the Medicaid expansion diminished First, the law outlawed underwriting (ie, increasing its effect. At the time of writing, 19 states44—many of premiums or refusing coverage) on the basis of gender them southern states with large minority populations or health status. Second, among non-group plans and poor records on health-care access and outcomes, (ie, plans purchased by individuals, not provided by particularly for the poor45—have opted out. Studies employers), premiums for older enrollees (≥64 years of comparing non-expansion to expansion states have age) cannot exceed those for young adults (≥21 years of found less improvement in health-care coverage, use, age) by more than a factor of three. Third, the ACA and access among the former.22,46,47 Estimates suggest that required insurance policies to pay for preventive services, state opt-outs will lead to thousands of unnecessary including contraception, without cost sharing (ie, out-of- deaths annually,48 further compounding existing pocket payments) and to cover maternity care.16 Fourth, geographical and racial health inequalities. Notably, the the legislation prohibited annual and lifetime caps on generosity of Medicaid benefits has traditionally varied insurance payments, improving financial protection from state to state. against catastrophic illness. Additionally, even in states that have implemented the Despite these provisions, increasingly subtle forms of Medicaid expansion, the low fees it offers doctors often insurance discrimination against the sick persist. Insurers consign Medicaid-insured patients to second-class status. continue to tailor their benefit packages (and, as described Many physicians will not accept Medicaid,49 and even above, their provider networks) to discourage high-cost when they do, might not offer Medicaid patients equal patients from choosing or remaining in their plan. For treatment. For instance, waiting times for new instance, some insurers require extraordinarily high appointments are longer for Medicaid patients than for patient copayments for essential drugs (eg, antiretrovirals those with private plans.50 Some studies have also found and antipsychotics) needed by expensively ill patients. HIV evidence suggesting pervasive segregation of care by patients enrolled in such plans pay, on average, US$4892 insurance status and race: for instance, some large annually out-of-pocket per drug.55 Such pricing tactics academic centres have maintained separate (and thus effectively skirt regulations prohibiting discrimination invariably unequal) clinic systems for Medicaid and against patients with pre-existing conditions.55 Some major privately insured patients.51 More generally, physicians insurers have also told insurance brokers that they will are significantly less likely to accept Medicaid in areas cease paying commissions to brokers for enrolling with high percentages of non-white people or high levels customers in more comprehensive, low cost-sharing plans of racial segregation than are physicians in other areas.52 because such plans attract patients in poor health To address such disparities, the ACA temporarily anticipating large medical expenses.56 (for 2 years) boosted Medicaid reimbursements for The ACA has improved reproductive health-care access, primary care physicians.53 A secret shopper study yet substantial gaps persist. From 2012 to 2015, the percent conducted before and during this change found an of reproductive-age women who were uninsured fell from increase (from 58·7% to 66·4%) in the proportion of 18·9% to 11·5%.30 The mandate that insurance fully cover practices willing to schedule appointments to Medicaid contraceptives has reduced out-of-pocket spending for enrollees; this increase was an improvement, but still such drugs and devices, yet many insurers still exclude well below the 86% of privately insured callers who were coverage for some items (eg, male condoms and the offered an appointment.53 Regardless, with the expiration morning-after pill).57,58 And whether or not contraception- of the fee boost, this modest improvement will probably related care is actually provided without cost sharing—as be reversed. required by the ACA—can depend on the vagaries of how In summary, although Medicaid offers crucial health physicians code particular office visits.57 coverage to the poor, it often provides an inferior and The ACA’s promise of full coverage for contraceptives separate tier of access to providers. Even if Medicaid were was also eroded by a 2014 Supreme Court decision expanded in all 50 states, the goal of health-care equity excusing employers who claim religious objections from would not be met. paying for such coverage for their employees. Data suggest that this contraception exclusion loophole has Health insurance market reforms left some women unable to obtain intrauterine devices,58 Discriminatory pricing—by age, gender, or health underscoring the risks intrinsic to a system that relies on status—has long characterised the US private insurance employers to provide and oversee their employees’ health market. In the past, many insurers charged older adults coverage.59 (ie, price increasing with age) and women higher These problems have been exacerbated by some state premiums, excluded coverage of maternity care,54 refused governments’ moves to curtail access to contraception to enrol individuals with medical problems (so-called and abortion, particularly for poor women. A number of pre-existing conditions), and set annual or lifetime caps states have imposed onerous requirements on abortion on insurance payments. In addition to its coverage clinics, forcing many to close, and have sharply reduced

1446 www.thelancet.com Vol 389 April 8, 2017 Series

funding for family planning programmes.60 In one such Single coverage Family coverage state—Texas—the cuts resulted in a sharp reduction in the use of long-acting contraceptives and an increase in Marketplace plans births among those reliant on injectable contraceptives.61 Bronze $5731 $11 601 In addition, preliminary evidence suggests the states Silver $3117 $6480 with the most severe abortion restrictions are seeing an Gold $1165 $2535 increase in women investigating and possibly pursuing Platinum $233 $468 self-induced abortions.62 These trends might worsen as a Employer-sponsored plans result of the 2016 election, given Republican proposals HMO $1025 $2758 aimed at reducing abortion access.63 PPO $958 $2012 POS $1230 $2467 Health-care affordability HDHP/SO $2099 $4332 Despite the ACA’s coverage expansions and insurance All plan types $1318 Not applicable market reforms, costs remain a potent impediment to Marketplace plans: deductibles are for 2016 and do not reflect cost-sharing health-care equity, even for people with insurance. High subsidies, which are available to households earning less than 250% of the federal out-of-pocket costs compound both health-care inequality poverty level. Data from HealthPocket.71 Employer-sponsored plans: deductibles and income inequality. With respect to the former, a large are for 2015, and only reflect deductibles for those plans with a general annual deductible. If plans with no deductible are averaged in, the average overall body of research demonstrates that cost sharing deters deductible is $1077 for single coverage plans (vs $1318 for all plan types among the use of clinical services, particularly for low-income plans with a general annual deductible). The employer-sponsored family groups.64–66 With respect to the latter, cost sharing deductible is for plans with aggregate, not per person, deductibles. Data from the 72 exacerbates economic inequality because it takes a greater 2015 Employer Health Benefits Survey. Costs expressed in US$. HMO=health maintenance organisation. PPO=preferred provider organisation. POS=point of proportion of income from the poor than from the service plan. HDHP/SO=high-deductible health plan with a savings option. wealthy67—an issue addressed elsewhere68 in this Series. Some provisions of the ACA do, notably, lower the Table 2: Average annual deductibles for private insurance by metallic tier (marketplace plan) or plan type (employer-sponsored plans) burden of cost sharing for certain services (table 1). As described above, the ACA requires insurers to fully cover certain preventive services69 and this appears to have raising copayments and deductibles in order to keep increased their uptake.70 However, when screening tests premiums below the threshold at which they incur the uncover problems—for instance, a mammographic tax10,74 (Republicans previously proposed replacing the opacity or hyperglycaemia—subsequent work-up and Cadillac tax with something similar—a cap on the tax- treatment can result in onerous cost sharing. Table 2 deductibility of employer-sponsored health insurance displays the average deductibles (the amount patients plans63—though their March, 2017, bill [described below] need to pay before insurance kicks in) for employer- would have just delayed implementation of the tax). sponsored plans and for the plans offered on the ACA’s For people with non-group plans, most of which are now insurance marketplaces. obtained through the ACA’s marketplaces, cost sharing is Regarding employer-sponsored plans, over the past even more onerous. The ACA regulates this cost sharing in decade the average deductible rose by 255%,72 and actual a number of ways. First, the law sets the actuarial value of household outlays for cost sharing rose 77%.73 The ACA’s marketplace plans at four metallic levels: 90% for platinum failure to reverse the trend toward rising cost sharing is plans, 80% for gold, 70% for silver, and 60% for bronze unsurprising for several reasons. First, under the ACA, (table 1). Second, it mandates caps on out-of-pocket the actuarial value—ie, the percentage of all health-care spending (after premiums) for plans sold on the costs paid by the insurer for plan members, on average— marketplace (in 2017, the out-of-pocket maximums were for employer-sponsored plans can be as low as 60% and $14 300 for a family and $7150 for an individual).17 Third, in still meet the requirements of the employer mandate.13 addition to the subsidies that reduce premiums for Thus, employees with a 60% actuarial value plan still pay, enrollees earning less than 400% of the federal poverty on average, 40% of their health-care costs out-of-pocket. level, the law provided a second set of subsidies to offset cost Second, the ACA imposed a new excise tax (the so-called sharing for those earning less than 250% of the federal Cadillac tax) on employer-sponsored insurance plans poverty level.17 These subsidies, whose legality is currently with premiums above a certain threshold. These plans under challenge from Congressional Republicans, are deemed luxurious by virtue of their relatively high effectively increase the actuarial value of the silver plan to premiums and comprehensive benefits, earning them between 73% and 94% (depending on enrollee income).17 the Cadillac label. However, over time, an increasing Despite these regulations and subsidies, those with percentage of typical plans will be subject to the tax: modest incomes sometimes still find themselves by 2028, an estimated 42% of employers will have at least financially squeezed.75 According to one survey,76 almost one insurance plan affected by the tax unless they half of of Americans would have to borrow money or sell a trim costs.74 Although Congress postponed the tax possession to pay a surprise expense of $400—an amount until 2020, employers will probably respond to it by well below the cost sharing resulting from a single test or www.thelancet.com Vol 389 April 8, 2017 1447 Series

hospitalisation under many plans. Similarly, only 37% of private insurers and state-run Medicaid programmes are non-elderly, non-poor households have the liquid assets following Medicare’s lead. necessary to meet the ACA’s out-of-pocket maximums for Proponents of so-called value-based reform typically a given year.77 Among poor families with at least one emphasise paying for value, not volume, using financial uninsured family member, that percentage is even lower.77 risk-sharing arrangements that offer providers bonuses Finally, these out-of-pocket maximums apply only to care for cutting costs and exact fines if they fail to do so, received from in-network providers. Cost sharing for care together with pay-for-performance (P4P) that rewards or obtained out of network—even when medically penalises providers based on their scores on quality appropriate—has no limit. metrics. The common thread linking these ideas is that Cost sharing in marketplace plans is generally higher they are intended to reduce costs by penalising the than in employer-sponsored plans. For example, one provision of excess care, while relying on P4P to ensure study78 estimated that for people with a chronic condition, that cuts in the amount of care do not compromise switching from an employer-sponsored plan to a silver- quality. level marketplace plan would double out-of-pocket New value programmes can be divided into two spending on drugs. Without subsidies, average deductibles overlapping categories: changes in the manner providers for silver plans in 2016 averaged over $3000 for individuals are paid, and a restructuring of health-care delivery.82 The and around $6500 for family plans (table 2).17,71 Even with former consists of the ACA’s new P4P Medicare schemes cost-sharing subsidies, patients face high costs at the for hospitals,83 as well as a subsequent value reform—the point of use; a 40-year-old man earning $25 000 a year, for Medicare Access and CHIP Reauthorization Act of 2015, instance, would still face a median annual deductible which will base 18% of payments to office-based of $2500 and an out-of-pocket maximum of $5000.17 physicians on their P4P scores and cost savings. The Finally, the effect of the ACA on health-care latter category includes incentives in the ACA to form affordability for the poor and near-poor covered by accountable care organisations (ACOs; described below), Medicaid is complex. On the one hand, traditional together with changes in spending on safety-net Medicaid imposes virtually no cost sharing on its providers that care for disadvantaged populations. These impoverished enrollees. Hence, many of those covered programmes have the capacity to affect health-care equity by the ACA’s Medicaid expansion have gained access to in a variety of ways. care without out-of-pocket costs. However, several states The ACA’s P4P initiative involved creating three new refused to expand Medicaid unless the Obama P4P Medicare hospital programmes—the Hospital administration granted them permission to impose new Readmissions Reduction Programme (HRRP), the premiums or cost-sharing requirements.79 For instance, Hospital Value-Based Purchasing (HVBP) programme, Indiana requires that enrollees deposit a monthly and the Hospital-Acquired Condition Reduction payment into a health savings account. Though the Programme— that in conjunction can result in amount is small, it is substantial for those with meagre penalties of 6% of total Medicare reimbursements by incomes.80 The ramifications are serious for those who 2017.83 Although upgrades of quality for all Americans, do not or cannot make these payments. People earning particularly the underserved, is a priority, it is far from between 100% and 133% of the federal poverty level who clear that these programmes will accomplish this. fail to make their payments lose all benefits for Indeed, such P4P programmes could exacerbate health- 6 months.80 For those earning less than poverty, the care inequality by penalising already cash-strapped penalty is less severe: they only lose dental and vision safety-net providers; they serve disadvantaged patients coverage, but are also subsequently obliged to pay whose poor health outcomes could arise from adverse new copayments for doctors’ visits, hospitalisations, social conditions rather than substandard care. P4P and medications.80 Notably, this programme was might also backfire if providers avoid disadvantaged implemented by Governor (now Vice-President) and non-compliant patients who drag down their Mike Pence, on the advice of Seema Verma, the new scores.84 head of the Centers for Medicare & Medicaid Services.81 The overall effect of the ACA’s P4P initiatives on health It might be thus be a harbinger of national changes to equity remains uncertain. A number of early studies, Medicaid. however, raise concerns that two of the ACA’s programmes (HRRP and HVBP) selectively penalise ACA initiatives to maximise value safety-net hospitals.85–88 In addition to expanding coverage and regulating Moreover, it is unclear whether P4P metrics accurately insurance, the ACA aimed to restructure the health-care measure quality. In California, for instance, safety-net delivery system to maximise value; these provisions also hospitals were fined under HRRP because their patients might have implications for health-care equity. Delivery with heart failure, pneumonia, and myocardial infarction system reforms (table 1) initially affected only the federal had elevated readmission rates.88 However, paradoxically, Medicare programme, which has covered virtually all these safety-net hospitals also had lower 30-day mortality persons aged 65 years and older since 1965. However, for all three conditions.88

1448 www.thelancet.com Vol 389 April 8, 2017 Series

Some argue that P4P programmes should statistically Looking ahead: health-care equity in the Trump era adjust quality scores for patients’ socioeconomic status,89 The 2016 election dramatically changed the political although Medicare’s administrators have thus far opposed landscape. Republicans have long vowed to repeal the such adjustment. However, although socioeconomic ACA and replace it with market-based solutions, but with status adjustment might make P4P fairer, capturing the the election of Donald Trump—a billionaire businessman complex impacts of socioeconomic status on P4P whose campaign rhetoric was perceived by many as measures is a formidable task. And even if adjustment for trafficking in crude bigotry—the prospects for repeal socioeconomic status confirmed that safety-net hospitals suddenly improved. The form that the replacement delivered poor quality care, an appropriate response would take was unclear until March 6, 2017, when the might be to increase, not decrease, the resources available Republican House, led by Speaker Paul Ryan, released a to these institutions. bill called the American Health Care Act (AHCA).96–98 Yet Ryan’s bill was to experience a rapid demise. It was The ACA and delivery system change assaulted by hard-right Republicans in the House of The push for structural changes in the delivery system— Representatives for not going far enough, and it lost the second element of the ACA’s value-based reforms— support from Republican moderates after Ryan attempted is embodied in new Medicare programmes that to mollify conservatives by modifying the bill. On encourage the formation of ACOs. These organisations, March 24, Ryan and Trump cancelled the scheduled vote which encompass both hospitals and doctors, contract on the bill. with Medicare to assume financial risk for all of the For the ACA’s supporters, this news was met with health care received by a panel of Medicare enrollees. much applause, although it is not entirely clear that the The ACO strategy has notable parallels to the health AHCA will remain in the dustbin of history. Regardless, maintenance organisation strategy that gained both the AHCA—or something resembling it—will probably prominence and notoriety in the 1980s and 1990s.90 Both serve as the model for right-wing health-care reform in health maintenance organisations and ACOs reward the years to come, and is worth briefly reviewing. providers for reducing their patients’ use of care.90 ACOs The AHCA,96–98 to the surprise of some, would actually also incorporate P4P, which supporters assert will not have maintained much of the overall structure of the only protect patients from incentives to skimp on care, ACA, including insurance regulations such as the one but could also improve quality across the board, thereby protecting those with pre-existing conditions. Like the diminishing health-care inequality.91 ACA, the Republican bill would also have provided tax Others, by contrast, have argued that ACOs could have credits to offset the cost of private health insurance “the unintended consequence of reinforcing health care premiums for those not eligible for employer-paid disparities”.92 For instance, if profitable hospitals and coverage. But because these more regressive credits practices that serve wealthier patients acquire and merge would not have been tied to the price of insurance (unlike with other similarly profitable providers, the segregation those of the ACA), they would have been inadequate to of disadvantaged patients might become only more cover the cost of coverage for many Americans, especially entrenched.92 For now, however, it is too early to judge the those who were older or who had lower incomes.96 equity effects of the ACO strategy. The AHCA would have also transformed the Medicaid Finally, the ACA’s delivery system reforms included programme for the poor. Beginning in 2020, federal two additional measures directed at safety-net providers, funding for the Medicaid expansion would have been with more immediate equity implications (albeit in substantially reduced. At the same time, federal funds different directions). On the one hand, the law earmarked would have been allocated to states using either a per- $11 billion in additional funding for the approximate capita formula or a “block grant”,99 replacing the current 1300 federally subsidised community health centres,93 a matching system that bases the federal contribution on the provision inserted by Vermont Senator Bernie Sanders.94 actual cost of care delivered. These changes would have These clinics, located in medically underserved areas, slashed federal funding of Medicaid by $839 billion over a serve about 24 million patients annually.93 On the other decade,100 forcing states to reduce the comprehensiveness hand, the ACA reduced funding for the Disproportionate of their Medicaid coverage, cut eligibility, or both. Share Hospital (DSH) programme, which provides Finally, in a last-minute maneuver intended to placate supplemental funding for hospitals in which a large hard-right conservatives, a provision was inserted share of patients have Medicaid (which pays low fees and undermining the ACA’s requirement that plans cover is also a marker for hospitals that care for the uninsured).95 “essential health benefits”, which would have promoted The ACA’s drafters assumed that the need for these the sale of bare-bones plans. subsidies would decline as coverage was expanded. Yet Overall, these provisions would have transformed safety-net hospitals continue to care for many uninsured coverage, providing a defined contribution to insurance and underinsured patients, particularly in states that premiums rather than a defined set of covered benefits, refused to expand Medicaid or have many immigrants and they would have slashed Medicaid, effectively so as to whom the ACA’s coverage expansion excluded. fund hundreds of billions of dollars in tax breaks96,100 to www.thelancet.com Vol 389 April 8, 2017 1449 Series

the rich and health-care corporations. They also would cost sharing, a regressive form of funding that is on the have dramatically increased the number of the uninsured rise, often forces non-wealthy families to choose between by an additional 24 million people.96,100 health care and other necessities. And despite the ACA’s By reducing protections for the sick and those with Medicaid expansion aiding millions of Americans, the lower incomes, the AHCA would clearly have exacerbated programme continues to consign poor patients to a health-care inequalities; those who care about health-care separate, and sometimes lower, tier of care. equity will certainly oppose such measures. At the same Change is needed. However, the market-based reforms time, however, the political wisdom of championing the proposed by Republicans would callously cut coverage health-care status quo seems dubious. Although the and exacerbate already pernicious inequities. By contrast, AHCA was profoundly unpopular, it is also clear that a thorough overhaul of US health care—universal, voters want health care to change—a sentiment that comprehensive, tax-funded coverage—remains the best seems reasonable in view of the enormous problems that reform option to close the equity gap. persist despite the ACA’s advances. Hence, equity- Contributors minded advocates might best counter the Republican Both authors contributed equally to the preparation of the manuscript. agenda with a forward-looking health-care agenda of Declaration of interests their own. A single-payer, Medicare-for-all reform— Both authors report research grant support from the National Institutes championed by Senator Bernie Sanders during his of Health, and are active members of Physicians for a National Health Program, an organisation that advocates for single-payer health care in upstart presidential campaign, as well as by many the USA. physicians101 and the nation’s largest nurses union— Acknowledgments would, in our view, best address health-care inequalities. We thank David Himmelstein and Steffie Woolhandler for their Such reform would replace the ACA’s patchwork of assistance with this manuscript. coverage provisions with a tax-funded universal References programme. It would comprehensively cover all US 1 Chetty R, Stepner M, Abraham S, et al. The association between residents without cost sharing (as is currently done in income and life expectancy in the United States, 2001–2014. JAMA 2016; 315: 1750–66. Canada and the UK), eliminating financial barriers to 2 Case A, Deaton A. Rising morbidity and mortality in midlife among care. Importantly, it would also end inequalities in access white non-Hispanic Americans in the 21st century. based on type of insurance, and probably reduce both Proc Natl Acad Sci USA 2015; 112: 15078–83. economic4 and racial102 health-care inequalities. Although 3 Schroeder SA. We can do better—improving the health of the American people. N Engl J Med 2007; 357: 1221–28. passage of such thoroughgoing reform is unlikely in the 4 James PD, Wilkins R, Detsky AS, Tugwell P, Manuel DG. near term, it might be the most feasible option once Avoidable mortality by neighbourhood income in Canada: 25 years political winds shift and politicians catch up with the after the establishment of universal health insurance. J Epidemiol Community Health 2007; 61: 287–96. views of the public: single payer, after all, is already more 5 Whitehead M. The concepts and principles of equity and health. popular than the ACA, favoured by 58% of the US Int J Health Serv 1992; 22: 429–45. public.103 6 Cohen RA, Martinez ME, Zammitti EP. Health Insurance Coverage: early release of estimates from the National Health Interview Survey, 2015. Hyattsville, MD: National Center for Health Statistics, 2016. Conclusion 7 Garfield R, Majerol M, Damico A, Foutz J. The uninsured: a primer. Decades ago, Julian Tudor Hart noted that “the availability Key facts about health insurance and the uninsured in America. of good medical care tends to vary inversely with the Menlo Park, CA: The Henry James Kaiser Family Foundation, 2016. 104 8 Clemans-Cope L, Kenney GM, Buettgens M, Carroll C, Blavin F. need of the population served”. His observation, derived The Affordable Care Act’s coverage expansions will reduce differences largely from experience in the UK, aptly describes US in uninsurance rates by race and ethnicity. Health Aff (Millwood) 2012; health care. 31: 920–30. 9 Sommers BD. Number of young adults gaining insurance due to the In 2010, prior to the ACA, one in six Americans was Affordable Care Act now tops 3 million. June 19, 2012. uninsured.6 Insurers could—and often did—discriminate https://aspe.hhs.gov/basic-report/number-young-adults-gaining- against older and sicker individuals as well as women. insurance-due-affordable-care-act-now-tops-3-million (accessed July 2, 2016). Racial inequities in coverage and access were pervasive. 10 Congressional Budget Office. Federal subsidies for health insurance Illness often led to financial ruin. And limited health- coverage for people under age 65: 2016 to 2026. Washington, DC: care access among the uninsured resulted in tens of Congressional Budget Office, 2016. 18 11 Eibner C, Saltzman E. Assessing alternative modifications to the thousands of unnecessary deaths annually —and Affordable Care Act: impact on individual market premiums and incalculable suffering for many more. insurance coverage. Santa Monica, CA: RAND Corporation, 2014. Through a patchwork of provisions, the ACA has 12 The Henry J Kaiser Family Foundation. Employer responsibility under the Affordable Care Act. 2015. http://kff.org/infographic/ ameliorated the situation, particularly benefiting blacks, employer-responsibility-under-the-affordable-care-act/ (accessed Hispanics, and the poor. Yet serious deficiencies have July 2, 2016). persisted. Many people remain uninsured and 13 McDonough JE, Adashi EY. In defense of the employer mandate: underinsured. Wealth continues to be a crucial factor in hedging against uninsurance. JAMA 2015; 313: 665–66. 14 Musco T, Sommers BD. Under the Affordable Care Act, determining health and access to medical care. Narrow- 105 million Americans no longer face lifetime limits on health network insurance plans bar many from care at leading benefits. Washington, DC: US Department of Health and Human hospitals and from the doctors of their choice. Patient Services, 2012.

1450 www.thelancet.com Vol 389 April 8, 2017 Series

15 Burke A, Simmons A. Increased coverage of preventive services 38 Shartzer A, Long SK, Anderson N. Access to care And affordability with zero cost sharing under the Affordable Care Act. Washington, have improved following Affordable Care Act implementation; DC: US Department of Health and Human Services, 2014. problems remain. Health Aff (Millwood) 2015; published online 16 Arons J. Women and Obamacare: what’s at stake for women if the Dec 16. DOI:10.1377/hlthaff.2015.0755. Supreme Court strikes down the Affordable Care Act. Washington, DC: 39 Howard DH. Adverse effects of prohibiting narrow provider Center for American Progress, 2012. networks. N Engl J Med 2014; 371: 591–93. 17 Collins SR, Gunja M, Beutel S. How will the Affordable Care Act’s 40 Goldberg D. State moves to protect Health Republic customers. cost-sharing reductions affect consumers’ out-of-pocket costs in Nov 8, 2015. http://www.politico.com/states/new-york/albany/ 2016? Issue Brief (Commonw Fund) 2016; 6: 1–17. story/2015/11/state-moves-to-protect-health-republic- 18 Wilper AP, Woolhandler S, Lasser KE, McCormick D, Bor DH, customers-027703 (accessed July 3, 2016). Himmelstein DU. Health insurance and mortality in US adults. 41 Haeder SF, Weimer DL, Mukamel DB. Narrow networks and the Am J Public Health 2009; 99: 2289–95. Affordable Care Act.JAMA 2015; 314: 669–70. 19 Sommers BD, Baicker K, Epstein AM. Mortality and access to care 42 Avalere Health. Exchange plans include 34 percent fewer providers among adults after state Medicaid expansions. N Engl J Med 2012; than the average for commercial plans. July 15, 2016. http://avalere. 367: 1025–34. com/expertise/managed-care/insights/exchange-plans-include-34- 20 McWilliams JM, Meara E, Zaslavsky AM, Ayanian JZ. Health of percent-fewer-providers-than-the-average-for-comm (accessed previously uninsured adults after acquiring Medicare coverage. July 16, 2015). JAMA 2007; 298: 2886–94. 43 Dorner SC, Jacobs DB, Sommers BD. Adequacy of outpatient 21 Ward BW, Clarke TC, Nugent CN, Schiller J. Early release of selected specialty care access in marketplace plans under the affordable care estimates based on data from the 2015 National Health Interview act. JAMA 2015; 314: 1749–50. Survey. Hyattsville, MD: National Center for Health Statistics, 2016. 44 The Henry J Kaiser Family Foundation. Current status of Medicaid 22 Long SK, Karpman M, Kenney GM, et al. Taking stock: gains in expansion decision. http://kff.org/health-reform/state-indicator/ health insurance coverage under the ACA as of March 2015. state-activity-around-expanding-medicaid-under-the-affordable-care- Washington, DC: Urban Institute, 2015. act/ (accessed April 21, 2016). 23 Sommers BD, Gunja MZ, Finegold K, Musco T. Changes in 45 Han X, Nguyen BT, Drope J, Jemal A. Health-related outcomes self-reported insurance coverage, access to care, and health under among the poor: Medicaid expansion vs. non-expansion states. the affordable care act.JAMA 2015; 314: 366–74. PLoS One 2015; 10: e0144429. 24 Carman KG, Eibner C, Paddock SM. Trends In health insurance 46 Sommers BD, Blendon RJ, Orav EJ. Both the ‘private option’ and enrollment, 2013–15. Health Aff (Millwood) 2015; 34: 1044–48. traditional Medicaid expansions improved access to care for 25 Collins SR, Rasmussen PW, Doty MM, Beutel S. The rise in health low-income adults. Health Aff (Millwood) 2016; 35: 96–105. care coverage and affordability since health reform took effect: 47 Wherry LR, Miller S. Early coverage, access, utilization, and health findings from the Commonwealth Fund Biennial Health Insurance effects associated with the Affordable Care Act Medicaid expansions: Survey, 2014. Issue Brief (Commonw Fund) 2015; 2: 1–16. a quasi-experimental study. Ann Intern Med 2016; 164: 795–803. 26 McMorrow S, Long SK, Kenney GM, Anderson N. 48 Dickman SL, Himmelstein DU, McCormick D, Woolhandler S. Uninsurance disparities have narrowed for black and hispanic Health and financial consequences of 24 states’ decision to opt out adults under The Affordable Care Act.Health Aff (Millwood) 2015; of Medicaid expansion. Int J Health Serv 2015; 45: 133–42. 4: 1774–78. 49 Decker SL. In 2011 nearly one-third of physicians said they would 27 Buchmueller TC, Levinson ZM, Levy HG, Wolfe BL. Effect of the not accept new Medicaid patients, but rising fees may help. Affordable Care Act on racial and ethnic disparities in health Health Aff (Millwood) 2012; 31: 1673–79. insurance coverage. Am J Public Health 2016; 106: 1416–21. 50 Bisgaier J, Rhodes KV. Auditing access to specialty care for children 28 Chen J, Vargas-Bustamante A, Mortensen K, Ortega AN. Racial and with public insurance. N Engl J Med 2011; 364: 2324–33. ethnic disparities in health care access and utilization under the 51 Calman NS, Golub M, Ruddock C, Le L, Hauser D. Separate and Affordable Care Act.Med Care 2016; 54: 140–46. unequal care in New York City. J Health Care Law Policy 2006; 29 Shartzer A, Garro N, Pellegrini C, Long SK. Changes in insurance 9: 105–20. coverage, access to care, and health care affordability for women of 52 Greene J, Blustein J, Weitzman BC. Race, segregation, and childbearing age. Washington, DC: Urban Institute, 2016. physicians’ participation in Medicaid. Milbank Q 2006; 84: 239–72. 30 Jones RK, Sonfield A. Health insurance coverage among women of 53 Polsky D, Richards M, Basseyn S, et al. Appointment availability reproductive age before and after implementation of the Affordable after increases in Medicaid payments for primary care. Care Act. Contraception 2016; 93: 386–91. N Engl J Med 2015; 372: 537–45. 31 Karpman M, Gates J, Kenney GM. Time for a checkup: changes in 54 Courtot B, Kaye J, Greenberger M, et al. Still nowhere to turn: health insurance coverage, health care access and affordability, and insurance companies treat women like a pre-existing condition. plan satisfaction among parents and children between 2013 Washington, DC: National Women’s Law Center, 2009. and 2015. Washington, DC: Urban Institute, 2016. 55 Jacobs DB, Sommers BD. Using drugs to discriminate—adverse 32 Wallace J, Sommers BD. Effect of dependent coverage expansion of selection in the insurance marketplace. N Engl J Med 2015; the Affordable Care Act on health and access to care for young 372: 399–402. adults. JAMA Pediatrics 2015; 169: 495–97. 56 Hancock J. Licking wounds, insurers accelerate moves to limit 33 Berk ML, Fang Z. Young adult insurance coverage and out-of-pocket health-law enforcement. Feb 4, 2016. http://khn.org/news/licking- spending: long-term patterns. Health Aff (Millwood) 2016; wounds-insurers-accelerate-moves-to-limit-health-law-enrollment/ 35: 734–38. (accessed Feb 4, 2016). 34 McMorrow S, Kenney GM, Long SK, Anderson N. 57 Politi MC, Sonfield A, Madden T. Addressing challenges to Uninsurance among young adults continues to decline, particularly implementation of the contraceptive coverage guarantee of the in Medicaid expansion states. Health Aff (Millwood) 2015; affordable care act.JAMA 2016; 315: 653–54. 34: 616–20. 58 Bearak JM, Finer LB, Jerman J, Kavanaugh ML. Changes in out-of- 35 Wong CA, Ford CA, French B, Rubin DM. Changes in young adult pocket costs for hormonal IUDs after implementation of the primary care under the Affordable Care Act.Am J Public Health Affordable Care Act: an analysis of insurance benefit inquiries. 2015; 105 (suppl 5): S680–85. Contraception 2016; 93: 139–44. 36 US Census Bureau. Percentage of people by type of health 59 Annas GJ, Ruger TW, Ruger JP. Money, sex, and religion— insurance coverage by age: 2013 and 2014. Current Population the Supreme Court’s ACA sequel. N Engl J Med 2014; 371: 862–66. Survey, 2014 and 2015. Annual Social and Economic Supplements. 60 Gold RB, Hasstedt K. Lessons from Texas: widespread 2015. https://www2.census.gov/programs-surveys/demo/tables/ consequences of assaults on abortion access. Am J Public Health p60/253/table2.pdf (accessed July 1, 2016). 2016; 106: 970–71. 37 Creedon TB, Cook BL. Access to mental health care increased but 61 Stevenson AJ, Flores-Vazquez IM, Allgeyer RL, Schenkkan P, not for substance use, while disparities remain. Potter JE. Effect of removal of planned parenthood from the Texas Health Aff (Millwood) 2016; 35: 1017–21. Women’s Health Program. N Engl J Med 2016; 374: 853–60. www.thelancet.com Vol 389 April 8, 2017 1451 Series

62 Stephens-Davidowitz S. The return of the DIY abortion. 82 Blumenthal D, Abrams M, Nuzum R. The Affordable Care Act at New York Times. March 5, 2016. http://www.nytimes. 5 years. N Engl J Med 2015; 372: 2451–58. com/2016/03/06/opinion/sunday/the-return-of-the-diy-abortion. 83 Kahn CN, Ault T, Potetz L, Walke T, Chambers JH, Burch S. Assessing html (accessed June 2, 2016). Medicare’s hospital pay-for-performance programs and whether they 63 Report from the Health Care Reform Task Force. A better way: our are achieving their goals. Health Aff (Millwood) 2015; 34: 1281–88. vision for a confident America. June 22, 2016. http://abetterway. 84 Casalino LP, Elster A, Eisenberg A, Lewis E, Montgomery J, speaker.gov/_assets/pdf/ABetterWay-HealthCare-PolicyPaper.pdf Ramos D. Will pay-for-performance and quality reporting affect (accessed July 1, 2016). health care disparities? Health Aff (Millwood) 2007; 26: w405–14. 64 Manning WG, Newhouse JP, Duan N, Keeler EB, Leibowitz A, 85 Joynt KE, Jha AK. Characteristics of hospitals receiving penalties Marquis MS. Health insurance and the demand for medical care: under the hospital readmissions reduction program. JAMA 2013; evidence from a randomized experiment. Am Econ Rev 1987; 309: 342–43. 77: 251–77. 86 Ryan AM. Will value-based purchasing increase disparities in care? 65 Brot-Goldberg ZC, Chandra A, Handel BR, Kolstad JT. What does a N Engl J Med 2013; 369: 2472–74. deductible do? The impact of cost-sharing on health care prices, 87 Gilman M, Hockenberry JM, Adams EK, Milstein AS, Wilson IB, quantities, and spending dynamics. Cambridge, MA: National Becker ER. The financial effect of value-based purchasing and the Bureau of Economic Research, 2015. Hospital Readmissions Reduction Program on safety-net hospitals 66 Wharam JF, Zhang F, Landon BE, Soumerai SB, Ross-Degnan D. in 2014: a Cohort Study. Ann Intern Med 2015; 163: 427–36. Low-socioeconomic-status enrollees in high-deductible plans reduced 88 Gilman M, Adams EK, Hockenberry JM, Wilson IB, Milstein AS, high-severity emergency care. Health Aff (Millwood) 2013; 32: 1398–406. Becker ER. California safety-net hospitals likely to be penalized by 67 Evans RG. Going for the gold: the redistributive agenda behind ACA value, readmission, and meaningful-use programs. market-based health care reform. J Health Polit Policy Law 1997; Health Aff (Millwood) 2014; 33: 1314–22. 22: 427–65. 89 Jha AK, Zaslavsky AM. Quality reporting that addresses disparities 68 Dickman SL, Himmelstein DU, Woolhandler S. Inequality and the in health care. JAMA 2014; 312: 225–26. health-care system in the USA. Lancet 2017; 389: 1431–41. 90 Himmelstein DU, Woolhandler S. Global amnesia: embracing 69 Cogan JA, Jr. The Affordable Care Act’s preventive services fee-for-non-service-again. J Gen Intern Med 2014; 29: 693–95. mandate: breaking down the barriers to nationwide access to 91 Berenson RA, Burton R. Accountable care organizations in preventive services. J Law Med Ethics 2011; 39: 355–65. Medicare and the private sector: a status update. Washington, DC: 70 Han X, Robin Yabroff K, Guy GP Jr., Zheng Z, Jemal A. Has Urban Institute, 2011. recommended preventive service use increased after elimination of 92 Pollack CE, Armstrong K. Accountable care organizations and cost-sharing as part of the Affordable Care Act in the United States? health care disparities. JAMA 2011; 305: 1706–07. Prev Med 2015; 78: 85–91. 93 US Department of Health and Human Services. Health Center Fact 71 HealthPocket. 2016 Affordable Care Act market brings higher Sheet. http://bphc.hrsa.gov/about/healthcenterfactsheet.pdf average premiums for unsubsidized. Nov 2, 2015. https://www. (accessed July 1, 2016). healthpocket.com/healthcare-research/infostat/2016-obamacare- 94 Fang L. GOP Officials publicly denounce Bernie Sanders’ premiums-deductibles (accessed Nov 30, 2015). Obamacare expansion, quietly request funding. The Intercept. 72 The Henry J Kaiser Family Foundation, Health Research & July 6, 2015. https://theintercept.com/2015/07/06/gop-senators- Education Trust. The 2015 Employer Health Benefits Survey. 2015. support-sanders-obamacare-expansion/ (accessed Oct 5, 2016). http://kff.org/health-costs/report/2015-employer-health-benefits- 95 Andrulis DP, Siddiqui NJ. Health reform holds both risks and survey/ (accessed Sept 29, 2015). rewards for safety-net providers and racially and ethnically diverse 73 Claxton G, Levitt L, Long M. Payments for cost sharing increasing patients. Health Aff (Millwood) 2011; 30: 1830–36. rapidly over time. April 12, 2016. http://www.healthsystemtracker. 96 Congressional Budget Office. Congressional Budget Office Cost org/insight/payments-for-cost-sharing-increasing-rapidly-over-time/ Estimate: American Health Care Act, 2017. https://www.cbo.gov/sites/ (accessed April 30, 2016). default/files/115th-congress-2017-2018/costestimate/ 74 Claxton G, Levitt L. Issue brief: how many employers could be americanhealthcareact.pdf (accessed March 13, 2017). affected by the Cadillac plan tax. August, 2015. http://files.kff.org/ 97 The Energy and Commerce Committee. Section-by-Section Summary. attachment/issue-brief-how-many-employers-could-be-affected-by- http://energycommerce.house.gov/sites/republicans.energycommerce. the-cadillac-plan-tax (accessed March 9, 2016). house.gov/files/documents/Section-by-Section Summary_Final.pdf 75 Tolbert J, Rudowitz R, Majerol M. Is ACA coverage affordable for low- (accessed March 7, 2017). income people? perspectives from individuals in six cities. April, 2016. 98 Ways and Means Committee. Repeal and replace of health-related tax https://kaiserfamilyfoundation.files.wordpress.com/2016/04/8867-is- policy. https://waysandmeans.house.gov/wp-content/ aca-coverage-affordable-for-low-income-people-perspectives-from- uploads/2017/03/03.06.17-Section-by-Section.pdf (accessed March 7, individuals-in-six-cities2.pdf (accessed April 24, 2016). 2017). 76 Schwarz H. Nearly half of Americans say they can’t afford an 99 Committee on Energy & Commerce and the Committee on Ways and unexpected $400 expense. The Washington Post. June 3, 2015. Means. Manager’s Amendment: Section-By-Section Summary. https://www.washingtonpost.com/news/the-fix/wp/2015/06/03/ https://rules.house.gov/sites/republicans.rules.house.gov/files/115/ tnearly-half-of-americans-say-they-cant-afford-an-unexpected-400- PDF/115-AHCA-SxS-MNGR-Policy.pdf. (accessed March 28, 2017). expense/ (accessed June 2, 2016). 100 Congressional Budget Office. H.R. 1628, the American Health Care Act, 77 Claxton G, Rae M, Panchal N. Consumers assets and patients cost incorporating manager’s amendments 4, 5, 24, and 25. 2017. sharing. March 11, 2015. http://kff.org/health-costs/issue-brief/ https://www.cbo.gov/publication/52516 (accessed March 28, 2017). consumer-assets-and-patient-cost-sharing/ (accessed July 1, 2016). 101 Gaffney A, Woolhandler S, Angell M, Himmelstein DU. Moving 78 Thorpe KE, Allen L, Joski P. Out-of-pocket prescription costs under forward from the Affordable Care Act to a single-payer system. a typical silver plan are twice as high as they are in the average Am J Public Health 2016; 106: 987–88. employer plan. Health Aff (Millwood) 2015; 34: 1695–703. 102 Kovesdy CP, Norris KC, Boulware LE, et al. Association of race with 79 Saloner B, Sabik L, Sommers BD. Pinching the poor? Medicaid cost mortality and cardiovascular events in a large cohort of US veterans. sharing under the ACA. N Engl J Med 2014; 370: 1177–80. Circulation 2015; 32: 1538–48. 80 The Henry J Kaiser Family Foundation. Medicaid expansion in 103 Newport F. Majority in US support idea of fed-funded healthcare Indiana. Feb 3, 2015. http://kff.org/medicaid/fact-sheet/medicaid- system. May 16, 2016. http://www.gallup.com/poll/191504/majority- expansion-in-indiana/ (accessed May 4, 2016). support-idea-fed-funded-healthcare-system.aspx (accessed June 1, 2016). 81 Harper J. Trump picks Seema Verma to run Medicare And 104 Hart JT. The inverse care law. Lancet 1971; 297: 405–12. Medicaid. NPR, November 29, 2016. http://www.npr.org/sections/ health-shots/2016/11/29/503762324/trump-picks-seema-verma-to- run-medicare-and-medicaid (accessed Dec 12, 2016).

1452 www.thelancet.com Vol 389 April 8, 2017 Series

America: Equity and Equality in Health 3 Structural racism and health inequities in the USA: evidence and interventions

Zinzi D Bailey, Nancy Krieger, Madina Agénor, Jasmine Graves, Natalia Linos, Mary T Bassett

Despite growing interest in understanding how social factors drive poor health outcomes, many academics, policy Lancet 2017; 389: 1453–63 makers, scientists, elected officials, journalists, and others responsible for defining and responding to the public See Editorial page 1369 discourse remain reluctant to identify racism as a root cause of racial health inequities. In this conceptual report, the See Comment pages 1376 third in a Series on equity and equality in health in the USA, we use a contemporary and historical perspective to and 1378 discuss research and interventions that grapple with the implications of what is known as structural racism on This is the third in a Series of population health and health inequities. Structural racism refers to the totality of ways in which societies foster racial five papers about equity and equality in health in the USA discrimination through mutually reinforcing systems of housing, education, employment, earnings, benefits, credit, New York City Department of media, health care, and criminal justice. These patterns and practices in turn reinforce discriminatory beliefs, values, Health and Mental Hygiene, and distribution of resources. We argue that a focus on structural racism offers a concrete, feasible, and promising Long Island City, NY, USA approach towards advancing health equity and improving population health. (Z D Bailey ScD, N Linos ScD, M T Bassett MD); Department of Social and Behavioral Introduction view—one that identifies and seeks to alter how such Sciences, Harvard T.H. Chan Racial and ethnic inequalities, including health racism contributes to poor health—is required to School of Public Health, inequities, are well documented in the USA (table),1–5 and understand, prevent, and address the harms related to Boston, MA, USA have been a part of government statistics since the structural racism. There is a rich social science literature (Prof N Krieger PhD, 6–8 8–10,19 M Agénor ScD); and Bard Prison founding of colonial America. However, controversies conceptualising structural racism, but this research Initiative, Annandale-on- 6–8 abound over explanations for these inequities. In this has not been adequately integrated into medical and Hudson, NY, USA report, we offer a perspective not often found in the scientific literature geared towards clinicians and other (J Graves MPH) medical literature or taught to students of health health professionals.9,10,12,13 In this report, we examine Correspondence to: sciences, by focusing on structural racism (panel 1)9–11 as what constitutes structural racism, explore evidence of Dr Mary T Bassett, 9,10,12,13 42-09 28th Street, Long Island a key determinant of population health. To explore how it harms health, and provide examples of City, NY 11101, USA this determinant of health and health equity, we examine interventions that can reduce its impact. Our central [email protected] a range of disciplines and sectors, including but not argument is that a focus on structural racism is essential See Online for infographic limited to medicine, public health, housing, and human to advance health equity and improve population health. www.thelancet.com/ resources. Our focus is the USA. infographics/us-health Although there is growing interest in understanding Structural racism: a brief introduction how social factors drive poor health outcomes,14 and Any account of structural racism within the USA must start directed investigation in social science and social with the experiences of black people and the Indigenous epidemiology into the interconnected systems of people of North America. It was on these two groups that discrimination,9,10,12,13 many academics, policy makers, the initial colonisers of North America (the English, French, scientists, elected officials, and others responsible for defining and responding to the public discourse remain resistant to identify racism as a root cause of racial health Search strategy and selection criteria 9,10,13 inequities. For example, in a Web of Science search An overarching search strategy was not used; instead, we done on Sept 7, 2016, with the term “race” in conjunction drew on our collective experience and specific searches for with “health”, “disease”, “medicine”, or “public health”, different sections to update or amplify the completeness of 47 855 articles were retrieved. However, when “race” was our review of the published literature. To identify review replaced by “racial discrimination”, only 2061 articles articles on racism and health, we searched Web of Science, were located, and only 1996 articles were found when it PubMed, and Google Scholar using the search terms “racism was replaced by “racism”. Furthermore, when “race” was AND health” or “racial discrimination AND health” or replaced by “structural or systematic racism”, only “structural racism AND health”. Only review articles 195 articles were identified (ie, 0·4% of those identified published in English between Jan 1, 2000, and Feb 23, 2016, with the search term “race”). were considered. We identified additional sources by To date, the small body of empirical research on racial performing selected searches in the databases listed above discrimination and health has focused primarily on the and the Google and DuckDuckGo search engines. These stress of perceived unfair treatment as experienced searches were further supplemented from our own 9,10,12,15–18 by individuals (interpersonal racism). Such knowledge of this subject. inequitable suffering matters, but a broad, societal www.thelancet.com Vol 389 April 8, 2017 1453 Series

“man”.6,8,19,23,24 To reconcile this contradiction, the colonists Key messages established legal categories based on the premise that black • Racial/ethnic health inequities in the USA are well documented, but controversies over and Native American individuals were different, less than explanations of these inequities persist. human, and innately, intellectually, and morally inferior— 8,19–21,23 • To date, in the small body of empirical research on racism and health, most studies and therefore subordinate—to white individuals. have focused on interpersonal racial/ethnic discrimination, with comparatively less Buttressing this concept of racial classification has been a emphasis on investigating the health effects of structural racism. long legacy of now discredited scientific theory and inquiry, • Structural racism involves interconnected institutions, whose linkages are historically constructed around the primary assumption that “race” rooted and culturally reinforced. It refers to the totality of ways in which societies was an innate and fixed characteristic and an inherently 6,8,9,19,23 foster racial discrimination, through mutually reinforcing inequitable systems hierarchical category. This manufactured concept of (in housing, education, employment, earnings, benefits, credit, media, health care, race used ostensibly visible phenotypic characteristics and 6,8,19 criminal justice, and so on) that in turn reinforce discriminatory beliefs, values, and ancestry to justify systems of oppression and privilege. distribution of resources, which together affect the risk of adverse health outcomes. Similar processes in other racialised societies, such as • One example of structural racism pertains to the ongoing residential segregation of black those of South Africa and Brazil, have produced country- Americans, which is associated with adverse birth outcomes, increased exposure to air specific racial hierarchies, which ascribe human value on 22 pollutants, decreased longevity, increased risk of chronic disease, and increased rates of the basis of proximity to whiteness. Furthermore, since homicide and other crime. Residential segregation also systematically shapes health-care the 18th century, scientific racism rooted in Aryan or white access, utilisation, and quality at the neighbourhood, health-care system, provider, and supremacy became a blueprint for many other mani­ individual levels. festations of society-specific scientific racism around the 6,22,25 • Several avenues exist for potentially efficacious solutions, including the use of a world. focused external force that acts on multiple sectors at once (eg, place-based multisector initiatives such as Purpose Built Communities, Promise Neighborhoods, The continuing role of ostensibly colour-blind laws and and Choice Neighborhoods), disruption of leverage points within a sector that might policies have ripple effects in the system (eg, reforming drug policy and reducing excessive In the USA, since the passage of the 1960s civil rights 8,20 incarceration), and divorcing institutions from the racial discrimination system laws, government complicity in the promotion of racial (eg, by training the next generation of health professionals about structural racism). discrimination is typically viewed as belonging to the • A focus on structural racism offers a concrete, feasible, and promising approach towards past. Examples of such de jure discrimination include advancing health equity and improving population health. Without a vision of health the legalisation and enforcement of slavery, the Jim Crow equity and the commitment to tackle structural racism, health inequities will persist. laws enacted in the 1870s (which legalised racial discrimination in reaction to the civil rights and social gains attained by the newly freed black population in the Dutch, and Spanish) first promulgated genocide and short Reconstruction period after the US Civil War), the enslavement, and created both legal and tacit systems of forcible removal of Indigenous people from their lands, racial oppression.8,20,21 Our report focuses primarily on the and the forcible transfer of Indigenous children from experiences of black Americans, since most research on their families to punitive so-called boarding schools racism and health has focused on this racialised group. We designed to strip them of their culture.8,19–21,26,27 recognise, however, that Native Americans and other people However, this standard view overlooks the long reach of of colour in the USA—including Latinos, Asian Americans, past practices and the impact of contemporary practices of and Pacific Islanders—have also been the target of health- institutional racism in both the public and private sector; harming racial discrimination, combined with anti- such practices have been and continue to be realised by immigrant and religious (eg, anti-Muslim) discrimination.8 purportedly colour-blind policies that do not explicitly Although issues of immigration and nativism are beyond mention “race” but bear racist intent or consequences, or the scope of this report, our analysis is applicable to the both.28–30 Institutional racism in one sector reinforces it in structural discrimination experienced not only by these other sectors, forming a large, interconnected system of groups but also by societally defined and racialised groups structural racism whereby unfair discriminatory practices in other countries with systems of oppression that have led and inequities in the health and criminal justice systems to health inequities.9,14,16,22 and in labour and housing markets bolster unfair discriminatory practices and inequities in the educational Racial ideology and the categorisation of racialised system, and vice versa.10 One key example, with ongoing social groups intergenerational effects, is the historic Social Security Act As with many other race-conscious societies, the USA has a of 1935, which created an important system of long history as a slaveholding republic and as a colonial- employment-based old-age insurance and unemployment settler nation.8,19–21 The modern concept of “race” emerged at compensation.8,20 The Act also, however, deliberately the cusp of the country’s nationhood, as early European excluded agricultural workers and domestic servants— settlers sought to preserve an economy largely on the basis occupations largely held by black men and women. This of the labour of enslaved African people and their accommodation was made to secure the votes of descendants while upholding the universal rights of Democrats in the South and thus ensure passage of the

1454 www.thelancet.com Vol 389 April 8, 2017 Series

Total White non-Hispanic Asian* Hispanic or Black non- Native American Latino Hispanic† or Alaska Native Wealth: median household assets (2011) $68 828 $110 500 $89 339 $7683 $6314 NR Poverty: proportion living below poverty level, all ages 14·8%; 21·0% 10·1%; 12·0% 12·0%; 12·0% 23·6%; 32·0% 26·2%; 38·0% 28·3%; 35·0% (2014); children <18 years (2014) Unemployment rate (2014) 6·2% 5·3% 5·0% 7·4% 11·3% 11·3% Incarceration: male inmates per 100 000 (2008) 982 610 185 836 3611 1573 Proportion with no health insurance, age <65 years (2014) 13·3% 13·3% 10·8% 25·5% 13·7% 28·3% Infant mortality per 1000 livebirths (2013) 6·0 5·1 4·1 5·0 10·8 7·6 Self-assessed health status (age-adjusted): proportion with 8·9% 8·3% 7·3% 12·2% 13·6% 14·1% fair or poor health (2014) Potential life lost: person-years per 100 000 before the age 6621·1 6659·4 2954·4 4676·8 9490·6 6954·0 of 75 years (2014) Proportion reporting serious psychological distress‡ in the 3·4% 3·4% 3·5% 1·9% 4·5% 5·4% past 30 days, age ≥18 years, age-adjusted (2013–14) Life expectancy at birth (2014), years 78·8 79·0 NR 81·8 75·6 NR Diabetes-related mortality: age-adjusted mortality per 20·9 19·3 15·0 25·1 37·3 31·3 100 000 (2014) Mortality related to heart disease: age-adjusted mortality 167·0 165·9 86·1 116·0 206·3 119·1 per 100 000 (2014)

NR=not reported. *Economic data and data on self-reported health and psychological distress are for Asians only; all other health data reported combine Asians and Pacific Islanders. †Wealth, poverty, and potential life lost before the age of 75 years are reported for the black population only; all other data are for the black non-Hispanic population. ‡Serious psychological distress in the past 30 days among adults aged 18 years and older is measured using the Kessler 6 scale (range=0–24; serious psychological distress: ≥13). Sources: wealth data taken from the US Census;1 poverty data for adults taken from the National Center for Health Statistics,2 and poverty data for children taken from the National Center for Education Statistics;3 unemployment data taken from the US Bureau of Labor Statistics;4 incarceration data taken from the Kaiser Family Foundation;5 data on uninsured individuals taken from the National Center for Health Statistics;2 data on infant mortality, self-assessed health status, potential life lost, serious psychological distress, life expectancy, diabetes-related mortality, and mortality related to heart disease taken from the National Center for Health Statistics.2

Table: Social and health inequities in the USA

Act. This racially motivated exclusion afforded the primarily white recipients additional opportunities to Panel 1: Definitions of structural racism and institutional racism acquire wealth and pass it on to their children, while those Many academics use structural racism and institutional racism interchangeably, but we excluded were unable to do so and instead often became consider these terms as two separate concepts. dependent on their children after retirement, thereby Structural racism refers to “the totality of ways in which societies foster [racial] further curtailing the intergenerational accumulation of discrimination, via mutually reinforcing [inequitable] systems…(eg, in housing, assets.8,20 The net result has been an entrenchment of education, employment, earnings, benefits, credit, media, health care, criminal justice, racial economic inequities that persist to this day.8,10,20,29,30 etc) that in turn reinforce discriminatory beliefs, values, and distribution of resources”, Another example is the War on Drugs and tough-on- reflected in history, culture, and interconnected institutions.9 This definition is similar to crime policies enacted in the 1970s and 1980s (labelled the “über discrimination” described by Reskin.10 “The new Jim Crow”).28 Without ever referring to “race” by itself, these policies stereotyped black Americans as drug Within this comprehensive definition, institutional racism refers specifically to racially addicts—despite similar prevalence of illicit drug use adverse “discriminatory policies and practices carried out…[within and between among white Americans—and disproportionately targeted individual] state or non-state institutions” on the basis of racialised group membership.9 28,30 black people for incarceration. The legacy of these Some of these institutional policies and practices explicitly name race (eg, de jure Jim policies is that the annual rate of incarceration of black Crow laws, which required schools and medical facilities to be racially segregated, and men is 3·8–10·5 times greater than that of white men, restricted certain neighbourhoods to be white-only), but many do not (eg, employer 31 across all age groups; moreover, in 2014, almost 3% of all practices of screening applications on seemingly neutral codes, such as telephone area black men in the USA were serving sentences of at least codes or ZIP codes, because of presumptions about which racial groups live where).11 1 year in prison.31

Structural racism in the private sector in the rental and housing markets against black and Institutional racism also continues unabated in the Latino communities remains pervasive, even though private sector, especially in housing and employment, intentional redlining is no longer legal (the term underpinning the structural racism of the ostensibly redlining is derived from the legal practice initiated in colour-blind policies in the public sector.32–34 In their 1934 by the Federal Housing Administration, which review of the evidence on discrimination in four domains involved marking maps with red lines to delineate (employment, housing, credit markets, and consumer neighbourhoods where mortgages were denied to markets), Pager and Shepherd33 argue that discrimination marginalised, racialised groups to steer them away from www.thelancet.com Vol 389 April 8, 2017 1455 Series

As this brief summary suggests, structural racism is an 9,12,13,16–18 Panel 2: Pathways between racism and health ongoing—and not just historical—concern across Economic injustice and social deprivation8,9,12,32–35 multiple systems. We next consider the implications of Examples include residential, educational, and occupational segregation of marginalised, such systemic racism on population health. racialised groups to low-quality neighbourhoods, schools, and jobs (both historical de jure discrimination and contemporary de facto discrimination), reduced salary for the Health consequences of structural racism: same work, and reduced rates of promotion despite similar performance evaluations evidence and evidence gaps Contemporary scholarship has established multiple 9,36–38 Environmental and occupational health inequities pathways by which racism harms health, involving Examples include strategic placement of bus garages and toxic waste sites in or close to adverse physical, social, and economic exposures, as well neighbourhoods where marginalised, racialised groups predominantly reside, selective as maladaptive coping behaviours and stereotype threats government failure to prevent lead leaching into drinking water (as in Flint, MI, in 2015–16), (panel 2).9,12,13,15–18,21,30,32–50 Typically concurrent, these and disproportionate exposure of workers of colour to occupational hazards exposures can accumulate over the life course and across Psychosocial trauma9,15,16,18 generations. Examples include interpersonal racial discrimination, micro-aggressions (small, often To date, research on racial discrimination and health unintentional racial slights and insults, such as a judge asking a black defence attorney has focused primarily on interpersonal discrimination as 9,16–18 “Can you wait outside until your attorney gets here?”), and exposure to racist media a psychosocial stressor. The strongest evidence in the coverage, including social media scientific literature is for adverse effects on psychological wellbeing, mental health, and related health practices 9,30,39 Targeted marketing of health-harming substances (eg, sleep disturbance, eating patterns, and the Examples include legal substances such as cigarettes and sugar-sweetened beverages, and consumption of psychoactive substances, including illegal substances such as heroin and illicit opioids cigarettes, alcohol, and drugs), as summarised in 9,12,15,16,18,35,51–58 Inadequate health care9,17,40–45 panel 3. Furthermore, growing research is Examples include inadequate access to health insurance and health-care facilities, and linking interpersonal racism to various biomarkers of substandard medical treatment due to implicit or explicit racial bias or discrimination disease and wellbeing, including allostatic load, inflammatory markers, and hormonal dysregulation.16,18 State-sanctioned violence and alienation from property and traditional lands9,21,30,46–48 Here, we focus instead on adverse health effects of Examples include police violence, forced so-called urban renewal (the use of eminent structural racism through two distinct but related pathways domain to force the relocation of urban communities of colour), and the genocide and emphasised in the literature: residential segregation and forced removal of Native Americans health-care quality and access.9,12,13,18 Both of these Political exclusion49,50 pathways include actionable leverage points to reduce Examples include voter restrictions (eg, for former felons and through identification exposure and promote health equity. A third relevant 28,30,35 requirements) pathway, discriminatory incarceration, is only briefly mentioned since it is discussed elsewhere in this Series Maladaptive coping behaviours9,16,18 by Wildeman and Wang.59 Examples include increased tobacco and alcohol consumption on the part of marginalised, racialised groups Residential segregation Stereotype threats15–18 As a reflection and reinforcement of structural and Examples include stigma of inferiority, leading to physiological arousal, and an impaired institutional racism, most residents in the USA have patient–provider relationship grown up in, and continue to live in, racialised and economically segregated neighbourhoods.29,33,34,60 Analysis of 2010 US Census data has found that “the average white person in metropolitan America lives in a white neighbourhoods). Additionally, strong evidence neighborhood that is 75% white”, whereas “a typical from experimental audit studies reveals continued racial African American lives in a neighborhood that is only discrimination in hiring decisions. In one study that 35% white (not much different from 1940) and as much used identical résumés, which differed only in the name as 45% black”.61 The literature on racial residential of the applicant, hiring managers called back those with segregation and poor health32,34,36,37,62–68 examines several traditionally white names (eg, Brad or Emily) 50% more direct and indirect pathways through which structural often than those with traditionally black names (eg, racism harms health, including the high concentration Jamal or Lakisha).33 In another study that used mailed of dilapidated housing in neighbourhoods that people of résumés, white applicants with criminal records were colour reside in,62,63 the substandard quality of the social64 called back more often than were black applicants and built65 environment, exposure to pollutants and without criminal records.33 Ongoing de facto racial toxins,36,37,65 limited opportunities for high-quality segregation in the workforce is partly why black education and decent employment,34,66 and restricted Americans, on average, have lower wages than those of access to quality health care.65 Health outcomes white Americans.35 associated with residential segregation documented

1456 www.thelancet.com Vol 389 April 8, 2017 Series

Panel 3: Dominant approaches to studying racial discrimination as a psychosocial stressor and associated adverse health outcomes, with counterexamples of research on measures of structural racism

Racism and stress metabolism change in response to chronic stressors.15 There is To date, racism has primarily been conceptualised as a burgeoning evidence linking experiences of discrimination to psychosocial stressor in the health science literature, and the biomarkers of disease and wellbeing, including allostatic load, strongest and most consistent evidence of its adverse health telomere length, cortisol dysregulation, and inflammatory effects concerns mental health, as detailed in several markers.9,16,18 comprehensive, systematic reviews.9,12,15,16,18 In one such review,16 Reliance on self-reports of exposure to racial discrimination published in 2015, the authors found that self-reported racism Most of the research on racial discrimination and health has was positively associated with increased levels of negative relied on self-reported measures, although some studies have mental health, including all individual mental health outcomes used vignettes or experimental situations. Evidence suggests except for positive affect (eg, depression, anxiety, distress, that because of well known cognitive biases, including social psychological stress, negative affect, and post-traumatic stress), desirability, self-reported data are likely to provide an and negatively associated with positive mental health (eg, underestimate of actual exposure, leading to underestimates of self-esteem, life satisfaction, control and mastery, and the magnitude of the association of racial discrimination with, wellbeing). After adjusting for publication bias, the association and its impact on, adverse health outcomes.9,18 Some immigrant between reported racism and mental health remained twice as groups, moreover, might be less likely than others to recognise large as that for physical health, which was driven primarily by racist interactions, or less likely to attribute discriminatory obesity outcomes. There is growing evidence that experiences behaviour to racism as opposed to language skills, immigration of racism are associated with poor sleep outcomes, which could status, or chance.9,52 be linked to both mental and physical health.51 Counterexamples of research on measures of structural racism Stress pathways Although small in comparison with psychosocial approaches, Much of the research on interpersonal racism and health has an emerging body of research has begun to investigate the posited that racism is a social stressor that operates through relationship between health and four domains of state-level diverse stress pathways, including physiological, psychological, structural racism: political participation, employment and job and behavioural pathways. Experiences that are perceived as status, educational attainment, and judicial treatment, racist act as social stressors, which can initiate a set of including incarceration.9,12,16,35,53–58 Black people living in states neurobiological and behavioural responses (ie, coping with higher levels of structural racism in these domains were behaviours) that can affect mental and physical health. These more likely than those living in states with lower levels of experiences can be chronic and include everyday hassles of structural racism to self-report a myocardial infarction in the receiving poor service at restaurants, being followed or not previous year; meanwhile, the same association for white helped in stores, and generally being treated with less respect people was null or protective.57Another study that used the and consideration than others. Acute experiences of violence, same measures found a positive association between structural harassment, and other threatening behaviour are also included racism at the state level and the odds of births that were small in this category. However, although such exposures are most for gestational age in both black and white women.58 Such likely to garner media attention, the common, chronic measures could be used to build the evidence base regarding experiences of discrimination are more consistently associated the connections between structural and institutional racism with poor health outcomes than are acute experiences,9,15,16,18 and health, and highlight areas for intervention. Priority should probably reflecting how brain chemistry and general be given to expanding this type of research. among black Americans include adverse birth USA there has been a shift from macrosegregation to outcomes,32 increased exposure to air pollutants,36 microsegregation, whereby “blacks and whites became decreased longevity,34,66 increased risk of chronic more evenly distributed across states and counties during disease,32,34,64 and increased rates of homicide and other the first two-thirds of the twentieth century, [and] … less crime.66,67 These adverse outcomes far outweigh any evenly distributed at the city and neighborhood levels”.60 benefits deriving from social support or political power Highlighting the need to think about smaller geographies, that accrue from the clustering of black Americans (or researchers have also noted that, as income inequality has other oppressed racialised groups) in adjoining increased, people at the top and bottom of the neighbourhoods.63,68 Residential segregation is thus a socioeconomic distribution have increasingly become foundation of structural racism and contributes to spatially isolated,69,70 such that “middle-class blacks are less racialised health inequities. able than their white counterparts to translate their higher Moreover, analysis of residential segregation requires economic status into desirable residential conditions”.34 addressing the intertwined occurrences of residential In recognition of the trend towards microsegregation segregation by both racialised group and class.60,69,70 In the and increased social polarisation, public health www.thelancet.com Vol 389 April 8, 2017 1457 Series

researchers have recently begun to use the Index of criminalised unemployment, vagrancy, and loitering.26 Concentration at the Extremes (ICE).70 This measure was The resultant prison population effectively re-established introduced into the sociological literature in 200169 and free labour for Southern states to rebuild infrastructure.73 was designed to measure economic polarisation—the The effects of mass incarceration, as traced by Wildeman extent to which a population is concentrated into the and Wang59 from the 1970s, are best understood as a extremes of wealth or impoverishment—by taking the continuation of racialised imprisonment8,10,20 rather than difference between the number of affluent and poor as an emergent process.28 Moreover, as noted previously, households in an area and dividing it by the total number strong feedback mechanisms exist between inequities in of households in the area.70 Moreover, these areas can be incarceration, employment, and health on a population measured at multiple levels (eg, census tract, city level.30,35,59 neighbourhood, and county). New innovations include the development of an ICE for racialised economic Health-care quality and access segregation, which uses data on the joint distribution of Interpersonal racism, bias, and discrimination in health- income and race/ethnicity. Research done in New York care settings can directly affect health through poor City, for example, has shown that ICE measures that health care. Almost 15 years ago, the Institute of Medicine captured both income and racialised group yielded larger Report titled Unequal Treatment: Confronting Racial and risk ratios, at both the neighbourhood and census tract Ethnic Disparities in Health Care40 documented systematic levels, for infant mortality, premature mortality, and and pervasive bias in the treatment of people of colour, diabetes mortality than an ICE solely for income or the resulting in substandard care. Evidence continues to poverty level.70 support this finding.41–44 Underscoring the need for explicit analysis of the However, it would be short sighted to view these health burden of residential segregation (regardless of problems solely as a matter of institutional and how it is measured) and neighbourhood disinvestment, interpersonal discrimination within health-care there is evidence to suggest that these structurally driven, settings.17,40–44 Instead, it is essential to understand the place-based exposures harm economic opportunity and, broad context within which health-care systems operate, when coupled with inadequate gun control, contribute to including the potentially disparate settings in which the lethal burden of gun violence and crime in health-care professionals and their patients reside. predominantly black and Latino neighbourhoods71,72 and Specifically, residential segregation systematically shapes in impoverished Native American reservations.21 In turn, health-care access, utilisation, and quality at the the violence and crime in these neighbourhoods neighbourhood, health-care system, provider, and reinforces the intergenerational legacy of racialised individual levels.45 The socioeconomic disadvantage punitive policing,8,20,21,28,31 perpetuating vicious cycles of resulting from systematic disinvestment in public and further community depletion and adverse health private sectors renders it difficult to attract primary-care outcomes.8,9,28,30,31,35,59 providers and specialists to predominantly black neighbourhoods.40,45 Likewise, health-promoting resources Discriminatory incarceration are inadequately invested into these neighbourhoods. The penal institutions that constitute the US criminal Health-care infrastructure and services are inequitably justice system—police departments, court systems, distributed, resulting in predominantly black neighbour​ correctional agencies, parole and probation departments, hoods having lower-quality facilities with fewer clinicians and sentencing boards—have established policies and than those in other neighbourhoods. Moreover, most of practices that are ostensibly colour-blind yet they these clinicians have lower clinical and educational criminalise communities of colour (eg, through day-to- qualifications than those in other neighbourhoods. This day practices such as stop and frisk) and disproportionately inequitable system is likely to disproportionately expose incarcerate black men, women, and children.30 As black residents to racially biased services.45 reviewed in this Series by Wildeman and Wang,59 each component of the criminal justice continuum—from Addressing structural racism to advance health arrest to re-entry—carries various health consequences, equity and a growing body of literature has documented severe Although efforts to counter institutional racism and adverse health outcomes associated with incarceration residential segregation in the housing market and on the individual, their families, and neighbourhoods. medical care system require initiatives focused on these What should not be lost in the explication of these institutions, such initiatives are not sufficient. Also outcomes is their roots in structural racism; the present needed is intersectoral work, especially that which is disproportionate representation of black people in the guided by transdisciplinary frameworks and action. penal system is reminiscent of the Black Codes and Analytical insights derived from a systems perspective convict leasing practices from the colonial period.8,26 New suggest several avenues for efficacious solutions, freedoms afforded to black people following the US Civil including the use of a focused external force that acts on War were promptly undone by laws that selectively multiple subsystems (ie, sectors) at once, disruption of

1458 www.thelancet.com Vol 389 April 8, 2017 Series

leverage points (ie, key points of intervention within a investment (over the course of 30 years before the sector that could be important for maintenance of the project) to US$123 million, property values in the system, both within and outside the particular sector in surrounding area increased, and new grocery stores, question), and divorcing institutions from the racial banks, and other businesses opened.74 The evidence of discrimination system.10 We highlight some promising, changes in the social determinants related to health concrete, intersectoral examples of each of these types of inequities is striking; to date, no health impact solutions, which have the potential to reduce, if not assessment has been done, although it is clearly remove, the burden of structural racism on population warranted. Future place-based interventions should health. build in health equity impact assessments from the start. Two federal initiatives launched in 2010 have Place-based, multisector, equity-oriented initiatives followed similar principles: the US Department of Health and health equity are substantially influenced by Education’s Promise Neighborhood initiative and the For more on Promise the places where people live, work, play, and pray.14 Yet, US Department of Housing and Urban Development’s Neighhorhoods see https://www2.ed.gov/programs/ the USA has high levels of racialised economic Choice Neighborhood initiative. Results of health impact promiseneighborhoods/index. 69,70 segregation. Within this context, multisector, place- assessments are eagerly awaited. html based partnerships focusing on equity can be an effective Short of full-scale community redevelopment, data For more on Choice means of placing pressure on the systems of structural suggest that improvements in housing lead to Neighborhoods see racism operating in a specific geographical region. improvements in health. In New York City, individuals https://portal.hud.gov/ hudportal/HUD?src=/program_ Place-based initiatives create structures for reinvesting in and families on a low income are able to enter lotteries offices/public_indian_housing/ neighbourhoods that have long been sidelined. Several for affordable housing units. Data from the New York programs/ph/cn initiatives have combined public and private partners City Housing and Neighborhood Study,75 which assessed from multiple sectors to achieve community-specific the impact of re-housing on those who won the lottery changes.74 These community-specific, multisectorcompared with those who did not, showed reductions in interventions that seek neighbourhood-wide coverage depression and asthma exacerbations. Although results have thus far focused primarily on predominantly black among adolescents were mixed, findings from the and Latino neighbourhoods, and also on Native American Moving to Opportunity study,76,77 in which vouchers for reservations, that have experienced high levels of poverty, housing were randomly allocated, suggest that housing health-limiting built environments, and substandard mobility policies that enable voluntary movement out of resources for schools and housing as a result of deprived neighbourhoods can result in long-term generations of structural racism. improvements in health and social outcomes. Established in 2009, Purpose Built Communities is Building government and public support for large- exploring the redevelopment of more than 20 high-need scale initiatives to counter structural racism is both neighbourhoods with the use of a model based on their necessary and possible. In May, 2016, the Government original 1995 development site: the East Lake Alliance for Race and Equity (GARE) and the non-profit neighbourhood of Atlanta, GA.74 About 20 years ago, a Living Cities jointly launched Racial Equity Here, a private philanthropist partnered with the president of the $3 million initiative to help five cities (Albuquerque, NM, Atlanta Housing Authority, a resident leader, and several Austin, TX, Grand Rapids, MI, Louisville, KY, and community business leaders to revitalise the area by Philadelphia, PA) improve racial equity, building on razing a violent, poorly maintained public housing approaches such as Seattle’s Race and Social Justice For more on Seattle’s Race and development and rebuilding a new mixed-income Initiative, which has explicitly recognised the links Social Justice Initiative see http://www.seattle.gov/rsji development, which involved temporary displacement of between racial equity and health equity.78 As the Mayor of residents during construction. Unlike other attempts at Austin, Steve Adler, noted, “Government helped create a rebuilding public housing, this development’s planning lot of the inequities, it institutionalized them. It’s and rollout was organised and backed by a dedicated important for the government, the city government to non-profit and focused on high-quality construction and address racial inequity, not just because of the conditions, on safe walkways and streets. The effort included a but also because we helped create it.”78 cradle-to-college educational curriculum, and a combination of facilities, programmes, and services Advocating for policy reform prioritised by community residents to promote healthy With the recognition that mass incarceration is a system behaviours, create jobs, and reduce crime in the short used to subordinate black people,10,28,30 efforts to reduce term, and break the cycle of intergenerational poverty discriminatory criminal sanctions on drug use (a leverage concentrated in this community in the long term.74 point) are also beginning to gain traction. From the 1980s With active involvement of community residents, by to 2010, the federal government sentencing guidelines 2015, crime had declined by 95% (compared with a mandated penalties for crimes related to crack cocaine 50% overall decline in Atlanta), the employment rate (a cheaper formulation more common in black among families in public housing increased from 13% communities than in other communities) that were to 70%, capital investments increased from no 100 times harsher than sentences for crimes involving www.thelancet.com Vol 389 April 8, 2017 1459 Series

the pharmacologically identical substance in powder of race or an analysis of racial inequality in relation to form, effectively targeting black people for prolonged health and other outcomes.85 Although many medical prison sentences.30 In the first sentencing breakthrough schools now include diversity training and provide in decades—the Fair Sentencing Act of 2010—the crack- instruction on cultural competency, such instruction is to-powder penalty ratio was reduced to 18:1, shrinking often brief (and sometimes delivered online). Moreover, the disparity but not eliminating it.30 Meanwhile, the programmes typically focus on individual prescription opioids, which are fuelling the current responsibility to counteract interpersonal discrimination; opioid epidemic among white people, have been relatively the goal is for individuals to increase their sensitivity to, unregulated. It was not until opioid addicts from white and knowledge about, other racial/ethnic groups.87,88 The communities started being incarcerated and dying in emphasis is therefore on “others”, in a way that could large numbers that the national narrative shifted from inadvertently contribute to racial stereotyping, as opposed penalisation to treatment—a clear demonstration of the to critical self-reflection about the participants’ positions racialised nature of the War on Drugs.79 in their societies’ race relations. The past decade has also witnessed new bipartisan By contrast, approaches based on structural efforts, across the country, to reduce the number of competency,83 cultural humility,89 and cultural safety46,90,91— people who are imprisoned. For example, California has which have been implemented in health professionals’ sought to address its unconstitutionally overcrowded training in several countries such as Canada and New prisons through several legislative initiatives, including Zealand—encourage a lifelong commitment to self- Proposition 47.80 This ballot initiative, passed in reflection and mutual exchange in engaging power November, 2014, commutes drug possession felonies imbalances along the lines of cultural differences. These (and a few minor offenses) to misdemeanours. It also approaches emphasise the value of gaining knowledge allows people serving a sentence for an eligible felony about structural racism, internalised scripts of racial conviction to petition the court for resentencing. With superiority and inferiority, and the cultural and power the disproportionate impact of drug arrests, prosecutions, contexts of health professionals and their patients or and convictions on black and Latino men and women, clients. Tying interactions between patients and health- Proposition 47 is likely to reduce racial inequities in care providers to population-level inequalities requires sentencing. Since 2014, more than 4000 people have been skilled instruction and considerable time, far beyond that released under this initiative and California has reduced patched together for short training courses in cultural overcrowding in prisons; however, racial inequities and competency.83 These approaches also require that health health effects have not yet been assessed.81 professionals be informed by scholarship from diverse disciplines about the origins and perpetuation of—as well Training the next generation of health professionals as remedies to counter—structural racism. It remains the Structural racism has developed over centuries and is charge of those committed to exploring and reversing deeply embedded in the thoughts and behaviours of people structural racism to connect how these forms of social in the USA and other countries,6,8,10,22,25 with its influence inequality translate into health and health-care inequities, extending to how health sciences are taught and the within and across generations.9,13,82,86 routine practices of health agencies and health-care Professional education about structural racism after providers.6,7,13,82–85 An analysis of structural racism is graduate school also matters, especially for clinical and required to recognise these problems and change them. public health practitioners whose decisions affect peoples’ Fortunately, a new wave of public health and medical health daily.13,92 As Hardeman and colleagues13 advocate, students, galvanised by protests over police killings and health professionals already practising in the field can the Black Lives Matter movement, have been advocating to still “learn, understand, and accept” the contemporary ensure that medical and public health schools incorporate and historical basis of structural racism in the USA, essential pedagogy about racism and health into standard understand how structural racism shapes our overarching coursework, as one step towards divorcing medical and narrative around inequities, define and call out racism public health institutions from their supportive roles in the when it is present, and contribute to the understanding of system of structural racism.13,82–84,86 Similarly, several public equity through clinical care and health research from the health agencies have begun to reform their institutional perspective of marginalised groups and with a healthy structure and organisational culture. dose of cultural humility. Several local health departments The standard practice for teaching about race and have already incorporated anti-racism training into staff health in medical and public health schools is one in professional development, and introduced internal which race is often discussed, but conversations about reforms to drive organisational change.92,93 For example, racism are sidelined, with scant hours (if any) devoted to in the mid-1990s the Alameda County Public Health social epidemiologists, medical anthropologists, social Department began to place neighbourhood offices in scientists, or historians who focus on racism and areas with poor health outcomes. Over time, these offices health.82–84 Few scientific and medical textbooks include drove changes in the department, including additional discussions of how racism affects the conceptualisation community involvement, staff trainings on anti-racism, a

1460 www.thelancet.com Vol 389 April 8, 2017 Series

new unit and a strategic plan to incorporate equity into commitment, as health professionals, to prevent their work, and an increased presence of the health avoidable suffering, care for those who are unwell, and department in local activism.92 The Boston Public Health create conditions in which all can truly thrive. Commission has also engaged in organisational change, Contributors launching a Racial Justice and Health Equity Initiative All authors contributed to the conceptualisation of the manuscript, that incorporates an anti-racism advisory committee, the literature search, and writing of this report. ZDB, NK, and MTB took the lead in ensuring coherence of the text, including the selection of development of a health equity framework, anti-racism appropriate data, and in data interpretation. training and professional development, and a forthcoming 93 Declaration of interests evaluation of its activities. As institutional reform is We declare no competing interests. closely associated with other models of productive Acknowledgments practices—including quality improvement, collective NK’s work is supported in part by an American Cancer Society Clinical impact, community engagement, and community Research Professor Award. mobilisation—application of an anti-racism lens should References not only be judged on its moral merits but also on its 1 US Census Bureau. Detailed tables on wealth and ownership assets: contributions to organisational effectiveness. We 2011. http://www.census.gov/people/wealth/data/dtables.html (accessed Jan 25, 2017). anticipate that forthcoming evidence will continue to 2 National Center for Health Statistics. Health, United States, support the view that removing racism from institutions 2015: with special feature on racial and ethnic health disparities. is essential to protect and promote the health of our May, 2016. http://www.cdc.gov/nchs/data/hus/hus15.pdf (accessed Jan 25, 2017). increasingly diverse communities. 3 National Center for Education Statistics. Family characteristics of school-age children. May, 2016. http://nces.ed.gov/programs/coe/ Conclusion pdf/coe_cce.pdf (accessed Jan 25, 2017). Since the American colonial period, public and private 4 Bureau of Labor Statistics. Labor Force Characteristics by Race and Ethnicity, 2014. Report 1057. November, 2015. https://www.bls.gov/ institutions have reinforced each other, maintaining opub/reports/race-and-ethnicity/archive/labor-force-characteristics- racial hierarchies that have allowed white Americans, by-race-and-ethnicity-2014.pdf (accessed Jan 25, 2017). across generations, to earn more and consolidate more 5 James C, Salganicoff A, Ranji U, Goodwin A, Duckett P. Putting men’s health care disparities on the map: examining racial and wealth than non-white Americans, and maintain political ethnic disparities at the state level. Menlo Park, CA: Kaiser Family dominance. This structural racism has had a substantial Foundation, 2012. https://kaiserfamilyfoundation.files.wordpress. role in shaping the distribution of social determinants of com/2013/01/8344.pdf (accessed Feb 20, 2017). 6 Hammonds EM, Herzig RM. The nature of difference: sciences of health and the population health profile of the USA, race in the United States from Jefferson to genomics. including persistent health inequities. The stark reality is Cambridge, MA: MIT Press, 2008. that research investigating the relationship between 7 Krieger N, Bassett M. The health of black folk: disease, class, and ideology in science. Mon Rev 1986; 38: 74–85. structural racism and population health outcomes has 8 Omi M, Winant H. Racial formation in the United States, 3rd edn. been scant, and even less work has been done to assess New York, NY: Routledge/Taylor & Francis Group, 2015. the health impacts of the few interventions and policy 9 Krieger N. Discrimination and health inequities. Int J Health Serv changes that could help dismantle structural racism. 2014; 44: 643–710. 10 Reskin B. The race discrimination system. Annu Rev Sociol We can, however, look to history as a guide. Notably, 2012; 38: 17–35. the handful of studies on the impact of the abolition of 11 US Equal Employment Opportunity Commission. Facts about race/ Jim Crow laws have consistently shown improvements in color discrimination. January, 1999. https://www.eeoc.gov/facts/fs- mortality in the black community, and converging race.pdf (accessed Jan 25, 2017). 12 Gee GC, Ford, CL. Structural racism and health inequities. mortality between black and white communities in the Du Bois Rev 2011; 8: 115–32. 15 years after the passage of the 1964 Civil Rights Act.53–56 13 Hardeman RR, Medina EM, Kozhimannil KB. Structural racism We recognise that efforts to implement reforms to and supporting black lives — the role of health professionals. N Engl J Med 2016; 375: 2113–15. dismantle structural racism have repeatedly encountered 14 Marmot M, Friel S, Bell R, Houweling TAJ, Taylor S, on behalf of serious obstacles and backlash from institutions, the Commission on Social Determinants of Health. Closing the gap communities, and individuals seeking to preserve their in a generation: health equity through action on the social determinants of health. Lancet 2008; 372: 1661–69. racial privilege.8,20,26,30 However, as Frederick Douglass 15 Berger M, Sarnyai Z. “More than skin deep’’: stress neurobiology famously said in his 1857 address on the struggle against and mental health consequences of racial discrimination. slavery in the USA, the West India emancipation, and the Stress 2014; 18: 1–10. backlash that ensued: “Power concedes nothing without 16 Paradies Y, Ben J, Denson N, et al. Racism as a determinant of health: a systematic review and meta-analysis. PLoS One 2015; 10: e0138511. 94 a demand.” 17 Paradies Y, Truong M, Priest N. A systematic review of the extent Without a vision of health equity and the commitment and measurement of healthcare provider racism. J Gen Intern Med to tackle structural racism, health inequities will persist, 2014; 29: 364–87. 18 Williams DR, Mohammed SA. Racism and health I: pathways and thwarting efforts to eliminate disparities and improve the scientific evidence.Am Behav Sci 2013; 57: 1152–73. health of all groups—the overarching goals for US health 19 Frederickson GM. Racism: a short history. Princeton, NJ: Princeton policy as enunciated by the official Healthy People 2020 University Press, 2003. For more on Healthy People 2020 see https://www. objectives. The challenge is great, but rising to this 20 Zinn H. A people’s history of the United States: 1492–present. New York, NY: Harper Perennial, 2015. healthypeople.gov/2020/About- challenge lies at the heart of our mission and our Healthy-People www.thelancet.com Vol 389 April 8, 2017 1461 Series

21 Dunbar-Ortiz R. An indigenous peoples’ history of the 48 Chaney C, Robertson RV. Racism and police brutality in America. United States. Boston, MA: Beacon Press, 2015. J Afr Am Stud 2013; 17: 480–505. 22 Loveman M. Making “Race” and nation in the United States, 49 Purtle J. Felon disenfranchisement in the United States: South Africa, and Brazil: taking “making” seriously. Theory Soc a health equity perspective. Am J Public Health 2013; 103: 632–37. 1999; 28: 903–27. 50 Blessett B. Disenfranchisement: historical underpinnings and 23 Higginbotham AL Jr. Shades of freedom: racial politics and contemporary manifestations. Public Administration Quarterly 2015; presumptions of the American legal process, vol 2. New York, NY: 39: 3. Oxford University Press, 1996. 51 Slopen N, Lewis TT, Williams DR. Discrimination and sleep: 24 Waldstreicher D. Slavery’s constitution: from revolution to a systematic review. Sleep Med 2016; 18: 88–95. ratification. New York, NY: Hill & Wang, 2010. 52 Carlisle SK. Perceived discrimination and chronic health in adults 25 Anderson W. Racial conceptions in the global south. Isis 2014; from nine ethnic subgroups in the USA. Ethn Health 2015; 20: 309–26. 105: 782–92. 53 Almond D, Chay KY, Greenstone M. Civil rights, the war on poverty, 26 Foner E. A short history of Reconstruction, 1863–1877. and black-white convergence in infant mortality in the rural South and New York, NY: Harper & Row, 1990. Mississippi. Dec 31, 2006. Cambridge, MA: Massachusetts Institute of 27 Du Bois WEB. Black folk then and now: an essay in the history and Technology, 2006. https://dspace.mit.edu/handle/1721.1/63330 sociology of the Negro race. New York, NY: Henry Holt & Co, 1939. (accessed Jan 25, 2017). 28 Alexander M. The new Jim Crow: mass incarceration in the age of 54 Krieger N, Chen JT, Coull B, Waterman PD, Beckfield J. colorblindness. New York, NY: The New Press, 2010. The unique impact of abolition of Jim Crow laws on reducing 29 Massey DS, Denton NA. The dimensions of residential segregation. inequities in infant death rates and implications for choice of Soc Forces 1988; 67: 281–315. comparison groups in analyzing societal determinants of health. Am J Public Health 2013; 103: 2234–44. 30 Hinton EK. From the war on poverty to the war on crime: the making of mass incarceration in America. Cambridge, 55 Krieger N, Chen JT, Coull BA, Beckfield J, Kiang MV, MA: Harvard University Press, 2016. Waterman PD. Jim Crow and premature mortality among the US black and white population, 1960–2009: an age–period–cohort 31 Carson EA. Prisoners in 2014. NCJ 248955. Sept 17, 2015. analysis. Epidemiology 2014; 25: 494–504. Washington, DC: Bureau of Justice Statistics, Office of Justice Programs, Department of Justice, 2015. https://www.bjs.gov/index. 56 Chay KY, Greenstone M. The convergence in black-white infant cfm?ty=pbdetail&iid=5387 (accessed Jan 19, 2017). mortality rates during the 1960’s. Am Econ Rev 2000; 90: 326–32. 32 Acevedo-Garcia D, Lochner KA, Osypuk TL, Subramanian SV. 57 Lukachko A, Hatzenbuehler ML, Keyes KM. Structural racism and Future directions in residential segregation and health research: myocardial infarction in the United States. Soc Sci Med 2014; a multilevel approach. Am J Public Health 2003; 93: 215–21. 103: 42–50. 33 Pager D, Shepherd H. The sociology of discrimination: racial 58 Wallace ME, Mendola P, Liu D, Grantz KL. Joint effects of structural discrimination in employment, housing, credit, and consumer racism and income inequality on small-for-gestational-age birth. markets. Annu Rev Sociol 2008; 34: 181–209. Am J Public Health 2015; 105: 1681–88. 34 Williams DR, Collins C. Racial residential segregation: 59 Wildeman C, Wang EA. Mass incarceration, public health, and a fundamental cause of racial disparities in health. widening inequality in the USA. Lancet 2017; 389: 1464–74. Public Health Rep 2001; 116: 404. 60 Massey DS, Rothwell J, Domina T. The changing bases of segregation 35 Western B, Pettit B. Black-white wage inequality, employment rates, in the United States. Ann Am Acad Polit Soc Sci 2009; 626: 74–90. and incarceration. Am J Sociol 2005; 111: 553–78. 61 Logan JR, Stults BJ. The persistence of segregation in the 36 Bravo MA, Anthopolos R, Bell ML, Miranda ML. Racial isolation metropolis: new findings from the 2010 census. March 24, 2011. and exposure to airborne particulate matter and ozone in http://www.s4.brown.edu/us2010/Data/Report/report2.pdf understudied US populations: environmental justice applications of (accessed Jan 19, 2017). downscaled numerical model output. Environ Int 2016; 62 White K, Borrell LN. Racial/ethnic residential segregation: framing 92–93: 247–55. the context of health risk and health disparities. Health Place 2011; 37 Mohai P, Pellow D, Roberts JT. Environmental justice. 17: 438–48. Annu Rev Env Resour 2009; 34: 405–30. 63 Kramer MR, Hogue CR. Is segregation bad for your health? 38 Siqueira CE, Gaydos M, Monforton C, et al. Effects of social, Epidemiol Rev 2009; 31: 178–94. economic, and labor policies on occupational health disparities. 64 Kershaw KN, Roux AVD, Burgard SA, Lisabeth LD, Mujahid MS, Am J Ind Med 2014; 57: 557–72. Schulz AJ. Metropolitan-level racial residential segregation and 39 Grier SA, Kumanyika S. Targeted marketing and public health. black-white disparities in hypertension. Am J Epidemiol 2011; Annu Rev Public Health 2010; 31: 349–69. 174: 537–45. 40 Institute of Medicine. Unequal treatment: confronting racial and 65 Landrine H, Corral I. Separate and unequal: residential segregation ethnic disparities in health care. Washington, DC: The National and black health disparities. Ethn Dis 2009; 19: 179–84. Academies Press, 2003. 66 Collins CA, Williams DR. Segregation and mortality: the deadly 41 Shavers VL, Fagan P, Jones D, et al. The state of research on racial/ effects of racism?Sociol Forum 1999; 14: 495–523. ethnic discrimination in the receipt of health care. 67 Krivo LJ, Byron RA, Calder CA, et al. Patterns of local segregation: do Am J Public Health 2012; 102: 953–66. they matter for neighborhood crime? Soc Sci Res 2015; 54: 303–18. 42 Flores G. Racial and ethnic disparities in the health and health care 68 Mendez DD, Hogan VK, Culhane JF. Institutional racism, of children. Pediatrics 2010; 125: e979–1020. neighborhood factors, stress, and preterm birth. Ethn Health 2014; 43 Blair IV, Havranek EP, Price DW, et al. Assessment of biases against 19: 479–99. Latinos and African Americans among primary care providers and 69 Massey DS. The prodigal paradigm returns: ecology comes back to community members. Am J Public Health 2013; 103: 92–98. sociology. In: Booth A, Crouter AC, eds. Does it take a village? 44 Puumala SE, Burgess KM, Kharbanda AB, et al. The role of bias by Community effects on children, adolescents, and families. emergency department providers in care for American Indian Mahwah, NJ: Lawrence Erlbaum Associates, 2001: 41–48. children. Med Care 2016; 54: 562–69. 70 Krieger N, Waterman PD, Spasojevic J, Li W, Maduro G, 45 White K, Haas JS, Williams DR. Elucidating the role of place in Van Wye G. Public health monitoring of privilege and deprivation health care disparities: the example of racial/ethnic residential with the Index of Concentration at the Extremes (ICE). segregation. Health Serv Res 2012; 47: 1278–99. Am J Public Health 2016; 106: 256–63. 46 Browne AJ, Smye VL, Varcoe C. The relevance of postcolonial 71 Burgason KA, Thomas SA, Berthelot ER. Nature of violence: theoretical perspectives to research in Aboriginal health. a multilevel analysis of gun use and victim injury in violent Can J Nurs Res 2005; 37: 16–37. interpersonal encounters. J Interpers Violence 2014; 29: 371–93. 47 Fullilove MT. Root shock: how tearing up city neighborhoods hurts 72 Ulmer JT, Harris CT, Steffensmeier D. Racial and ethnic disparities America, and what we can do about it. New York, NY: New Village in structural disadvantage and crime: white, black, and hispanic Press, 2016. comparisons. Soc Sci Q 2012; 93: 799–819.

1462 www.thelancet.com Vol 389 April 8, 2017 Series

73 Lichtenstein A. Twice the work of free labor: the political economy 85 Morning AJ. The nature of race: how scientists think and teach of convict labor in the New South. New York, NY: Verso, 1996. about human difference. Berkeley, CA: University of California 74 Purpose Built Communities. Purpose Built Communities at the Press, 2011. Chautauqua Institution. Sept 28, 2015. http:// 86 Bassett MT. # BlackLivesMatter—a challenge to the medical and purposebuiltcommunities.org/news-press/purpose-built- public health communities. N Engl J Med 2015; 372: 1085–87. communities-at-the-chautauqua-institution (accessed Jan 25, 2017). 87 Sturm S. The architecture of inclusion: advancing workplace equity 75 Gaumer E, Jacobowitz A, Brooks-Gunn J. The impact of affordable in higher education. Harv J L & Gender 2006; 29: 247–334. housing on the well-being of low-income households. Public Policy 88 Cross TL, Bazron BJ, Dennis KW, Isaacs MR. Towards a culturally Analysis and Management Conference; Miami, FL; Nov 14, 2015. competent system of care: a monograph on effective services for https://appam.confex.com/appam/2015/webprogram/Paper14806. minority children who are severely emotionally disturbed. html (accessed Jan 25, 2017). Washington, DC: CASSP Technical Assistance Center, Georgetown 76 Chetty R, Hendren N, Katz LF. The effects of exposure to better University Child Development Center, 1989. neighborhoods on children: new evidence from the Moving to 89 Tervalon M, Murray-Garcia J. Cultural humility versus cultural Opportunity experiment. Am Econ Rev 2016; 106: 855–902. competence: a critical distinction in defining physician training 77 Ludwig J, Liebman JB, Kling JR, et al. What can we learn about outcomes in multicultural education. J Health Care Poor Underserved neighborhood effects from the Moving to Opportunity experiment. 1998; 9: 117–25. Am J Sociol 2008; 114: 144–88. 90 Darroch F, Giles A, Sanderson P, et al. The United States does 78 Abello OP. 5 cities get support for dismantling systemic racism. CAIR about cultural safety examining cultural safety within Next City (Philadelphia, PA), May 31, 2016. https://nextcity.org/ Indigenous health contexts in Canada and the United States. daily/entry/living-cities-racial-inequity (accessed Jan 25, 2017). J Transcult Nurs 2016; published online Feb 25. 79 Netherland J, Hansen H. White opioids: pharmaceutical race and DOI:10.1177/1043659616634170. the war on drugs that wasn’t. BioSocieties 2016; published online 91 Papps E, Ramsden I. Cultural safety in nursing: the New Zealand Jan 11. DOI:10.1057/biosoc.2015.46. experience. Int J Qual Health C 1996; 8: 491–97. 80 Legislative Analyst’s Office. The 2015–16 budget: implementation of 92 National Association of County and City Health Officials. Proposition 47. Sacramento, CA: Legislative Analyst’s Office, 2015. Expanding the boundaries: health equity and public health practice. http://www.lao.ca.gov/reports/2015/budget/prop47/ Washington, DC: National Association of County and City Health implementation-prop47-021715.pdf (accessed Jan 25, 2017). Officials, 2014. http://www.dialogue4health.org/uploads/resources/ 81 Stanford Justice Advocacy Project. Proposition 47 progress report: Expanding_the_Boundaries_Final_508_091814.pdf (accessed year one implementation. Stanford, CA: Stanford Law School, 2015. Jan 25, 2017). https://www-cdn.law.stanford.edu/wp-content/uploads/2015/10/ 93 Boston Public Health Commission. The Racial Justice and Health Prop-47-report.pdf (accessed Jan 25, 2017). Equity Initiative: 2015 overview. http://www.bphc.org/whatwedo/ 82 Hart A. Anti-racism in public health education: a student-driven health-equity-social-justice/racial-justice-health-equity-initiative/ model for changes in a Master’s of Public Health Program. Documents/RJHEI%202015%20Overview%20FINAL.pdf (accessed American Public Health Association Annual Meeting; Chicago, IL; Jan 25, 2017). Nov 3, 2015. https://apha.confex.com/apha/143am/webprogram/ 94 Douglass F. Two Speeches, by Frederick Douglass: one on Paper323616.html (accessed Jan 25, 2017). West India emancipation, delivered at Canandaigua, Aug. 4th, and 83 Metzl J, Roberts D. Structural competency meets structural racism: the other on the Dred Scott Decision, delivered in New York, on the race, politics, and the structure of medical knowledge. occasion of the anniversary of the American Abolition Society, Virtual Mentor 2014; 16: 674. May, 1857. Rochester, NY: C. P. Dewey, 1857. 84 Anderson W. Teaching race at medical school: social scientists on the margin. Soc Stud Sci 2008; 38: 785–800.

www.thelancet.com Vol 389 April 8, 2017 1463 Series

America: Equity and Equality in Health 4 Mass incarceration, public health, and widening inequality in the USA

Christopher Wildeman, Emily A Wang

Lancet 2017; 389: 1464–74 In this Series paper, we examine how mass incarceration shapes inequality in health. The USA is the world leader in See Editorial page 1369 incarceration, which disproportionately affects black populations. Nearly one in three black men will ever be See Comment pages 1376 imprisoned, and nearly half of black women currently have a family member or extended family member who is in and 1378 prison. However, until recently the public health implications of mass incarceration were unclear. Most research in This is the fourth in a Series of this area has focused on the health of current and former inmates, with findings suggesting that incarceration could five papers about equity and produce some short-term improvements in physical health during imprisonment but has profoundly harmful effects equality in health in the USA on physical and mental health after release. The emerging literature on the family and community effects of mass Department of Policy Analysis incarceration points to negative health impacts on the female partners and children of incarcerated men, and raises and Management, Cornell University, Ithaca, NY, USA concerns that excessive incarceration could harm entire communities and thus might partly underlie health disparities (C Wildeman PhD); Bureau of both in the USA and between the USA and other developed countries. Research into interventions, policies, and Justice Statistics, Washington, practices that could mitigate the harms of incarceration and the post-incarceration period is urgently needed, DC, USA (C Wildeman); particularly studies using rigorous experimental or quasi-experimental designs. Rockwool Foundation Research Unit, Copenhagen, Denmark (C Wildeman); Yale School of Introduction of mass incarceration. According to sociologist Medicine, New Haven, CT, USA In this Series paper, we review research into the effects of David Garland,1 who first used a variant of the term mass (E A Wang MD); and Bureau of Justice Assistance, Washington, mass incarceration on health and health disparities incarceration, it entails historically and comparatively DC, USA (E A Wang) within the USA and between the USA and other extreme levels of incarceration that are so heavily Correspondence to: developed democracies. We first outline the contours concentrated among some groups that incarceration has Dr Christopher Wildeman, Department of Policy Analysis and Management, Cornell Search strategy and selection criteria Key messages University, Ithaca, NY 14853, USA We strove to achieve a complete search of peer-reviewed • In the USA, incarceration is common and concentrated in [email protected] articles and government-funded reports relating to the black community See Online for infographic incarceration and health. Because many of the journals that • Individuals who experience incarceration at any point in www.thelancet.com/ infographics/us-health publish research on family and community effects of mass their life are disproportionately in poor health both incarceration are not indexed by PubMed or PsychInfo, we before, during, and after their incarceration first did a Google Scholar search for peer-reviewed articles and • The physical health of individuals improves in some government-funded reports, including a host of specific domains during incarceration, although the mental health health conditions (such as hepatitis, cardiovascular disease, of individuals generally worsens and major depressive disorder), in addition to the terms • Having been formerly incarcerated is associated with poor “incarceration,” “imprisonment”, “jail”, and “prison” as our mental health and physical health outcomes, as well as search terms. We then searched PubMed and PsychInfo using elevated mortality risk the same terms. We did not use any date restrictions in our • Although little research considers the indirect health search. We also searched the bibliographies of key peer- consequences of incarceration, having a family member reviewed articles and relied on the few other review articles on incarcerated harms the mental and physical health of the topic. Although our exploration was international in non-incarcerated female partners and children scope, we restricted our search to articles and documents • High incarceration prevalence also compromises published in English, with a focus on newer, innovative work. community health, with the strongest evidence We cite the highest-quality works that have contributed the implicating community-level increased incidence of HIV most to this burgeoning field, with special emphasis on • Mass incarceration contributes to racial health disparities in studies using strong research designs making identification of the USA across a range of outcomes because of its direct and plausibly causal relationships possible. Because the goal of our indirect consequences for health, and the disproportionate Series paper was to consider the consequences of mass concentration of incarceration among black communities incarceration for health disparities in the USA, we placed • Because the USA incarcerates many more of its citizens substantially more emphasis on studies within the USA, than do other developed democracies, mass incarceration although we also report research on prisoners’ health in other might have contributed to the country’s lagging developed democracies when appropriate. performance on health indicators such as life expectancy

1464 www.thelancet.com Vol 389 April 8, 2017 Series

become a normal stage in the lifecourse. We then Mass incarceration consider the health effects of current incarceration and On any given day, the USA incarcerates more of its having ever been incarcerated, as well as health disparities citizens (2·2 million) and at a higher level (700 per attributable to these effects. We next review data about the 100 000) than any other country. Yet, for much of its broader health effects of mass incarceration, focusing on history, the USA was no outlier in terms of incarceration. families, communities, states, and nations, as well as As in most developed democracies—the focus of all of health disparities attributable to these effects. Finally, we our comparisons, because these countries are more focus on the next steps for researchers, medical similar to the USA in key ways (such as general standard professionals, and policy makers. Throughout, we are of living, political structure, and core population health careful to note that the teasing out of causal relationships indicators such as infant mortality and life expectancy at between incarceration and health outcomes on the basis birth) than some other counties (eg, China and Russia) of existing research is difficult because there are no that have high incarceration prevalence—the US randomised controlled trials of incarceration relative to incarceration prevalence hovered between 100 per no incarceration in this research area. To overcome these 100 000 and 200 per 100 000 in the mid-20th century.7 obstacles to causal inference, we focus (when possible) on In 1950, for instance, the US incarceration prevalence studies in which confounders were rigorously addressed was roughly 175 per 100 000,8 somewhat lower than through various strategies, including natural experiments. Finland’s (185 per 100 000).9 This prevalence was We find that incarceration is a pressing public health considerably increased for developed democracies, but concern, affecting not only the health of currently and not an aberration. formerly incarcerated individuals but also that of their Starting in the mid-1970s, the US incarceration families and communities.2–4 Because of these myriad prevalence started to spiral upward (figure 1).5 By 1985, negative consequences of mass incarceration for American the USA incarcerated 312 of every 100 000 residents. society, we argue—consistent with some research in this 20 years later, the prevalence had risen to 743 per 100 000. area5,6—that mass incarceration might partly account for Its closest competitors among developed nations were widening health inequality both within the USA and New Zealand (173 per 100 000), Luxembourg (159 per between the USA and other developed democracies. 100 000), and Spain (140 per 100 000).

Australia Austria Belgium Canada Denmark Finland France 800 000

600

400

200

Incarceration per 100 0

Germany Ireland Italy Japan Luxembourg Netherlands New Zealand 800 000

600

400

200

Incarceration per 100 0

Norway Portugal Spain Sweden Switzerland UK USA 800 000

600

400

200

Incarceration per 100 0 1980 1990 2000 1980 1990 2000 1980 1990 2000 1980 1990 2000 1980 1990 2000 1980 1990 2000 1980 1990 2000 Year Year Year Year Year Year Year

Figure 1: Trends in incarceration prevalence in 21 developed democracies, 1981–2007 Calculations based on data from Wildeman (2016).5

www.thelancet.com Vol 389 April 8, 2017 1465 Series

Men born 1945–49, risk of imprisonment Men born 1965–69, risk of imprisonment undoubtedly helped to both launch mass incarceration 10,11 70 White and keep it going. 60 Black Disparities in incarceration by race or ethnicity and 50 education in the USA are marked and have been since the 10,12 40 earliest statistics were collected. Incarceration has 13,14 30 become common for poor men from ethnic minorities. 2·8% of (non-Hispanic) white men born in the late 1960s of imprsonment (% ) 20 and 20·3% of (non-Hispanic) black men from the same Risk 10 15,16 0 cohort spent time in prison by their 30s (figure 2). For black men who did not complete high school, this risk Children born 1978, risk of paternal imprisonment Children born 1990, risk of paternal imprisonment was 57·0%. Moreover, these figures in fact underestimate 70 White the number of men who have experienced incarceration, 60 Black because the data refer only to incarceration in prisons 50 (facilities run by the state or the federal government that 40 hold inmates with sentences in excess of 1 year) and 30 exclude incarcerations in jails (local facilities that hold 20 inmates awaiting trial or sentenced to less than 1 year),

of paternal imprsonment (%) 10 which are far more common. No data are available for the

Risk 0 cumulative risk of total incarceration (in prisons and jails) Total High school dropout Total High school dropout because accurate estimates of the cumulative risk of ever Figure 2: Risk of ever experiencing imprisonment by age 30–34 years for US men by birth cohort, and risk of experiencing jail incarceration in the USA do not exist. ever experiencing paternal imprisonment by age 14 years for US children by birth cohort The conditions of incarceration in the USA are also Sources: Western and Wildeman (2009);15 Wildeman (2009).16 extreme, a fact much less discussed in the literature. For example, although precise estimates are not available for Know imprisoned individual Imprisoned family member 50 the number of individuals in solitary confinement (a form of imprisonment in which an inmate is isolated from any 40 human contact, often with the exception of guards and 30 other members of the prison staff), one study’s investigators estimated that 100 000 prisoners are in solitary confinement 20 in the USA on any given day,17 a figure that suggests that Proportion (%) 10 the USA has more prisoners in solitary confinement than the UK has prisoners overall. 0 Because men who experience incarceration are Imprisoned neighbour Trust imprisoned individual connected to families, their incarceration can have 50 implications for the health and wellbeing of women and 40 children as well. Furthermore, because of the vast racial disparities in the risk of experiencing incarceration, the 30 spillover effects of incarceration for family members 20 could have implications not only among men but also

Proportion (%) among whole communities, divided along racial and 10 ethnic lines. The proportion of black children who will 0 ever have a father imprisoned is high (figure 2). A black White White Black Black White White Black Black child born in 1990 had a 25·1% chance of having their women men women men women men women men father sent to prison;16 for those whose fathers did not Figure 3: Proportion of people in the USA who know individuals currently in state or federal prison, by race finish high school, the risk was roughly double that, and gender at 50·5%. According to the Bureau of Justice Statistics, Source: Lee and colleagues (2015).19 52% of state and 63% of federal inmates reported being parents, to an estimated 1·7 million children (ie, 2·3% of Although the causes of mass incarceration are American children).18 complex, social and criminal justice policies such as the The exposure of black families to incarceration cuts so-called War on Drugs, the deinstitutionalisation of deeper still. Nearly half of black women have a family people with mental illnesses, and punitive sentencing member or extended family member imprisoned policies such as three-strike laws (mandating life (figure 3).19 For white women, the risk is only a quarter as imprisonment for third offences of even relatively high, at 12%.19 Black people are also more likely than the minor felonies) and mandatory minimum sentences overall population to know someone who is incarcerated, (requiring judges to impose long sentences for specific have a neighbour incarcerated, or have a confidante offences, even for some first-time offenders) incarcerated.19

1466 www.thelancet.com Vol 389 April 8, 2017 Series

The pronounced disparities in exposure to incarceration access to illicit drugs and alcohol, and improved health- emphasise the salience of research into its health effects. care access, although the mechanisms are debated.29–32 If incarceration substantially worsens the health of non- However, the decreased mortality for black male prisoners incarcerated family members, mass incarceration could does not hold for other subpopulations of prisoners.29–32 be an important driver of broader health disparities Adjudication between these competing hypotheses is in the USA. Moreover, stark disparities in exposure beyond the scope of this Series paper, but we note that to incarceration probably extend to acquaintances, prisons and jails are some of the only places in the USA neighbours, and confidantes, potentially amplifying the where health care is guaranteed by law (although the contribution of incarceration to health inequities in often-dramatic variation in the quality of health care in the USA.19 correctional facilities undermines the notion that this mandate has been met). In 1976, the US Supreme Court Effects on the health of prisoners ruled in Estelle v Gamble that failure to provide basic A growing number of studies have examined the effects health care in correctional facilities violated the of incarceration on health.2–4 In this section, we review constitutional prohibition against cruel and unusual these effects, which have also been reviewed elsewhere,2,3 punishment. That ruling mandated that prisons and jails including in a 2016 series in The Lancet that explored the provide acute care services, but, as the prison population For the Series on HIV and relationship between incarceration and communicable has aged, prison health-care services have had to provide related infections in prisoners 4 see http://thelancet.com/series/ diseases such as HIV, viral hepatitis, and tuberculosis. increased care for chronic diseases as well. aids-2016 The Series documented the burden of these For many Americans, correctional facilities provide communicable diseases among prisoners,20 as well as incarcerated adults with their first access to preventive options for treatment21 and prevention22 in carceral and chronic medical care.4 An estimated 40% of settings. Importantly—and by contrast with most individuals with chronic medical conditions are research in this area—the Series also considered the diagnosed with a chronic condition while incarcerated,33 implications of communicable diseases for the human and 80% report seeing a medical provider while rights of prisoners23 and in regions where disease incarcerated.34 Unfortunately, the quality of medical care transmission is an especially pressing problem (sub- for chronic disorders in correctional settings is highly Saharan Africa,24 eastern Europe,25 and central Asia25). variable,35 and overcrowding of correctional facilities We consider in more detail the family and community (especially prisons) has even reached the stage at which consequences of mass incarceration, a topic that has judges have mandated the release of prisoners because received little attention in the medical community. the level of overcrowding constitutes cruel and unusual Although we focus on adults, it is important to note that punishment.4 incarcerated young people are at high risk for poor Compared with the non-incarcerated population, physical and mental health.26,27 incarcerated individuals have increased prevalence of Research into the effects of current incarceration on infectious disease (including sexually transmitted health is beset by several shortcomings beyond the diseases, HIV, and hepatitis C), chronic medical obstacles to causal inference mentioned in our conditions (eg, hypertension, diabetes, and asthma), introduction. Scant research has examined objectively substance use disorders, and mental health disorders;34,36 measured health outcomes, and relatively few studies Fazel and Baillargeon2 provide a more exhaustive list of have considered the mental health of current and former differences. While incarcerated, inmates also have a high inmates in the USA.28 Even fewer studies have explored prevalence of vitamin D deficiency.37 However, findings how different durations (eg, months or years) or types from a few studies have shown that incarceration can (eg, prison or jail) of incarceration affect health. In a improve the management of chronic conditions relative similar vein, little research has considered how the to time spent outside of prison, especially in cases of conditions of confinement (eg, solitary confinement) or severe functional limitation38 and HIV.39 However, in the types of criminal justice policies (eg, three-strike laws) time between release and re-incarceration, the probability affect health. Despite these caveats, most evidence of viral suppression declines from roughly 50% to 30%.25 suggests that incarceration has strongly harmful effects Unfortunately, because of data limitations, the effect of on the health of prisoners over their lifecourse. incarceration on many of these disorders is unclear. Overall, physical and psychological wellbeing worsens Effects of current incarceration for inmates, while mortality declines for black inmates. Being incarcerated might, paradoxically, decrease mortality Some study findings show worsening of depressive and physical morbidity in the short term for some groups. symptoms40 and life satisfaction41 during incarceration. Black male prisoners, for instance, have far lower mortality Furthermore, inmates placed in solitary confinement than similarly aged black men in the general population.29–32 suffer greatly,41 and such confinement has serious short- Researchers speculate that the protective effects of current term and long-term repercussions.42,43 For instance, imprisonment for this group might be driven by a inmates in solitary confinement in the New York City jail decreased risk of death by violence or accidents, reduced system had 6·27 greater odds (95% CI 3·92–10·01) of www.thelancet.com Vol 389 April 8, 2017 1467 Series

potentially fatal self-harm (including hanging and many barriers that individuals face after incarceration, it ingesting poison) than those not placed in solitary is unsurprising that they earn 30% less than similar confinement.44 Nonetheless, most research into the never-incarcerated individuals and that some of this mental health of inmates, while acknowledging the high effect is driven by discrimination.10,11 prevalence of mental health problems in correctional Findings from studies of administrative data have populations, has not tested whether mental health shown increased mortality among former inmates, changes as a result of incarceration.36 although the magnitude of this association varies.29,30,52 Of course, the total health effect of incarceration is a Investigators of one study53 that used a quasi-experimental product of time spent incarcerated and time spent free. design to assess whether incarceration caused premature Individuals who experience incarceration spend, on mortality found an effect for women, but not for men, average, far more time out of prison than in it, with after adjustment for confounders measured before much of that time happening after prison release since incarceration to ensure appropriate time-ordering of most individuals experience their first incarceration by confounders, explanatory variables, and dependent their late 30s. For instance, black men who ever variables (such as a history of illicit drug use, low experience prison incarceration spend 13·4% of their education, and pre-existing health problems). The working lives in prison.45 In other words, the average findings of this single study should be tested in further prisoner spends roughly six times as long exposed to the research, especially because it is the sole study to suggest consequences of past incarceration as they do being that prison release might not increase mortality risk. incarcerated. Hence, in considering the lifelong health The evidence that a history of incarceration is associated effects of incarceration, the period after release is of with increased morbidity is somewhat more consistent crucial importance. than the data for mortality, although, again, it remains unclear whether this relationship is indeed causal. Effects of past incarceration However, with the exception of the Coronary Artery Risk Although current incarceration has mixed effects on Development in Young Adults (CARDIA) study54 and the prisoners’ health, past incarceration has a clearly Veterans Aging Cohort study,55 few studies include both deleterious impact on health. Patients with chronic incarceration measures and objective health data. In conditions are often released without medications or a CARDIA, the adjusted odds of left ventricular hypertrophy follow-up appointment in the community.46 Even when (a common sequela of poorly controlled hypertension) provided with a prescription at release, many do not among the ever-incarcerated were 2·7 (95% CI 0·9–7·9) obtain them.47 Recently released inmates are less likely to compared with the never-incarcerated.54 In a matched have a primary care physician, disproportionately use sample, a history of incarceration was associated with emergency departments for health care, and have high 1·8 times increased odds (95% CI 1·147–2·519) of having levels of preventable hospital admissions compared with hepatitis or tuberculosis.56 Studies including less precise the general population.48 Because former inmates are measures of health have also consistently linked previous also at disproportionately high risk of mental health incarceration with poor health.3 Research has also shown problems that can interfere with their ability to follow that the formerly incarcerated have very high prevalence through with care for serious medical conditions,49 these of psychiatric morbidity, with associations especially obstacles to receiving care are even more important. pronounced for dysthymia and major depressive disorder, Before the Affordable Care Act, four-fifths of former and that incarceration is at least partly to blame for this inmates were uninsured at release; even among those increase.40,49 who are insured, many do not have the resources to pay for their care.50 The Affordable Care Act might diminish The direct effects of incarceration on health the health consequences of incarceration, because 10% disparities of the uninsured population has a recent history of Although black populations have high levels of criminal justice involvement.51 Unfortunately, the refusal incarceration, few studies have examined the direct on the part of several states to accept the Act’s expansion effects of incarceration on racial health disparities. The of Medicaid coverage for the poor will probably attenuate scant research in this area supports two conclusions. this benefit. First, racial health disparities among prisoners are muted; Upon release, former inmates often have no housing, differences in mortality and morbidity between black and employment, and family support, and face discrimination white individuals are smaller in prison than in the in finding jobs and housing.10,11 Individuals with health general population.2,38 Second, the post-release effects of issues are also confronted with the responsibility to incarceration certainly contribute somewhat to racial manage these problems, obtain health care, and keep up health disparities, although the magnitude of this effect is with medications and appointments while also meeting unclear. In an analysis, investigators using data from the their basic needs. Individuals convicted of drug felonies National Longitudinal Survey of Youth56 concluded that are also prohibited from accessing safety-net services disparities in incarceration prevalence contributed greatly such as public housing and food subsidies.1 Given the to disparities between black and white men in midlife

1468 www.thelancet.com Vol 389 April 8, 2017 Series

self-reported health, as measured by the 12-Item Short and children of male prisoners. Findings from two Form Health Survey; findings from another study38 that studies have suggested a link between parental used the same data and a measure of self-reported incarceration and child mortality: investigators of a functional limitation (defined as having had any health US study65 found elevated infant mortality, whereas problem that precluded working) showed that findings from a Danish study66 of mortality up to age incarceration explained only 6% of racial disparities in 18 years showed increased mortality among sons but not this measure. Findings from a population-based study in daughters of incarcerated men. A few other studies have New York City57 suggested that disparities in incarceration also shown evidence of gender-specific effects; parental contributed substantially to disparities in asthma incarceration was associated with significantly more prevalence. weight gain67 and higher levels of inflammatory markers Mass incarceration also creates methodological (eg, C-reactive protein) among adolescent girls than problems in documentation of racial health inequities in among boys.68 Yet, given the dearth of research in this prospective longitudinal studies. Because black men area, these findings about gender differences should be have very high levels of incarceration, they are more interpreted with some caution. likely than others to drop out of prospective longitudinal Although very few studies have used physiological surveys. As a result, research based on such surveys measures to assess the health of children of incarcerated could misestimate the magnitude of health disparities if parents, the literature assessing self-reported, caretaker- the health status of black men who experience reported, and teacher-reported outcomes for children is incarceration is worse than those who do not, as most vast. These study findings tell a consistent story: paternal research suggests is indeed the case.58 incarceration is associated with behavioural and mental health problems throughout childhood,69 and a host of The indirect effects of incarceration on health poor outcomes (including increased prevalence of Overview substance misuse70) in adolescence and adulthood.71–73 Until the past 10 years, most research into the health The most wide-ranging assessment of the effect of consequences of incarceration had focused exclusively parental—mostly paternal—incarceration used data on how incarceration affects those who experience it. from the National Survey of Children’s Health,74 showing However, as incarceration has become increasingly links to a host of negative health outcomes among common, researchers have become aware of the broader children, including self-rated health, depression, anxiety, health effects of mass incarceration on families, asthma, and obesity. Findings from a study75 that used communities, and even nations. Because little research data from the National Longitudinal Study of Adolescent has examined the spillover effects of mass incarceration Health (Add Health) underscored that many of the on direct measures of health, our Series paper also negative consequences of paternal incarceration continue encompasses broader studies of wellbeing. In this area, throughout adolescence and early adulthood.75 we are unable to make distinctions between the effects of For maternal incarceration, the story is more current and past incarceration. complicated. A handful of studies have linked maternal incarceration with worse self-reported health,75 Effects of family member incarceration on health educational,76 and criminal justice outcomes77 for Research into the broader family consequences of children. However, other study findings78,79 have shown incarceration suggests myriad channels through which no effects on children after adjustment for factors that incarceration might matter. For example, incarceration are associated with the risk of incarceration and poor decreases the financial contributions individuals can child health, such as low parental education, financial make after release;59 while incarcerated, their financial instability, and criminal activity. Given the paucity of contributions are virtually nil.60 Because keeping in touch studies on this topic, and evidence that maternal with a prisoner is costly,52 incarceration exacerbates incarceration helps some children and harms others,80 financial hardships beyond what would be expected due the net effect of maternal incarceration on children just to decreased earnings. Incarceration also disrupts remains an open question. romantic unions.61 The resulting decrease in adults’ time Fewer quantitative studies (but many qualitative available for household duties might reduce the time ones60,62,63) have assessed how incarceration affects other spent on health-related activities. Having an incarcerated adult family members. Women whose partners are family member—and re-incorporating a former incarcerated experience substantial mental health inmate—is also stressful. Moreover, if the stigma deterioration,81 as well as a host of elevated risk factors attached to incarceration pervades families, as research for cardiovascular disease.82 However, this effect on suggests,62,63 having a family member incarcerated could cardiovascular risk factors was not observed among men reduce the social support available to families.64 in the household.82 We must note that the effect of Although incarceration can also affect prisoners’ incarceration on family violence is unclear. There is little siblings, husbands, boyfriends, and parents, most doubt that incarcerated individuals83 and their families65,84 research has focused on the heterosexual partnerships experience great exposure to violence throughout their www.thelancet.com Vol 389 April 8, 2017 1469 Series

lives. The incarceration of a family member might tuberculosis incidence) and increased multidrug-resistant increase family violence by destabilising already- tuberculosis. Findings from another cross-national study5 disadvantaged homes. Alternatively, the removal of showed that increases in incarceration were associated violent family members from the household might with substantial worsening of life expectancy and infant decrease exposure to violence for the remaining mortality, although the population-level consequences of household members. Existing research provides little incarceration for health were significantly worse in the guidance regarding either possibility. USA than in other developed democracies. This analysis suggested that US life expectancy would have increased Effects of incarceration on communities 51·1% more and infant mortality would have fallen Neighbourhoods with high levels of incarceration are 39·6% more from 1983 to 2005 if incarceration had associated with poor population health, including high remained at the mid-1980s level. Taken together, these prevalence of asthma, sexually transmitted infections, and findings suggest that mass incarceration could contribute psychiatric morbidity;85–90 the challenge is to decipher to both within-country and between-country inequalities whether imprisonment, rather than the factors that lead to in health. imprisonment, is the driver. All studies done so far85–90 Finally, as for longitudinal studies, the US point-in- have tested a linear effect of imprisonment, yet a non- time surveys underlying much of the epidemiological linear relationship between neighbourhood-level and health services research (eg, the National Health prevalence of incarceration and community health is also Interview Survey) exclude inmates,94 resulting in possible. Clear85 proposed a hypothesis of coercive substantial misestimates of disease prevalence and, mobility, suggesting that the crime-fighting benefits of particularly, racial disparities. With so many minority imprisonment at low levels are substantial, but that these men behind bars, their exclusion from almost all benefits fall as imprisonment increases, and that further research provides a fanciful picture of progress in increases in imprisonment raise—rather than reduce— the USA, especially for health inequities between black crime. Testing of this hypothesis is difficult. If true, it has and white populations. profound implications for understanding the effect of incarceration on community health, not only because Conclusions and next steps high levels of violent crime remain one of the most serious Soaring incarceration since the mid-1970s has threats to public health in these communities but also profoundly affected health in the USA, especially in poor because it suggests that the public health consequences of and minority communities. Incarceration might incarceration in these communities could be far larger temporarily improve some physical health outcomes than an additive model would imply. during imprisonment. However, after release and over the lifecourse more broadly, imprisonment seems to Indirect effects of incarceration on states, nations, and worsen both physical and mental health, although we health disparities make this statement with some hesitation because few The indirect effects of incarceration on states and (if any) strong causal tests are available and the health nations, and health disparities more broadly, are most conditions considered so far have been limited. Although readily measured at the population level. Hence, we data are sparse, mass incarceration also probably discuss all three in the same section. worsens the health of the female partners and children Variation at the state level has rarely been used to of inmates. analyse the health effects of differences in incarceration Because of the uneven distribution of incarceration, prevalence, despite the availability of state-level data about these ill effects could be a significant contributor to racial key health outcomes and incarceration. Findings from a health disparities. The criminal justice system has few studies have suggested that states with large numbers become an institution—like the education system—that of former inmates have poorer-quality health-care both reflects systematic and institutionalised racism and systems,91 lower life expectancy,92 and higher incidence of exacerbates existing inequities.3,11,71 Moreover, as some HIV infection93 and infant mortality65 than do states with recent research into the relationship between few former inmates. These state-level studies have also incarceration and population health indicates,5 the shown a link between incarceration prevalence and health uniquely high incarceration prevalence in the USA might disparities. Findings from one study,93 for instance, partly underlie the country’s poor showing relative to showed that mass incarceration explained most of the other developed democracies on population health racial disparities in the incidence of HIV infection. measures over the past 40 years. There is less country-level than state-level research into On a more hopeful note, soaring costs, overcrowding the relationship between incarceration and health. Of of prisons and jails, and a spotlight on overly aggressive these studies, two stand out. Stuckler and colleagues6 policing in minority communities have engendered showed that increased incarceration was linked with agreement that mass incarceration has failed and should increased tuberculosis incidence (a 1% increase in be reversed. There is also increasing recognition, incarceration was associated with a 0·3% increase in although not consensus, that policing should be altered

1470 www.thelancet.com Vol 389 April 8, 2017 Series

Total Arkansas Florida Kentucky 1000

800

600

400

Mortality per 100 000 200

0

Louisiana Maryland Montana New York 1000

800

600

400

Mortality per 100 000 200

0

North Carolina Ohio Oregon South Carolina 1000

800

600

400

Mortality per 100 000 200

0

Tennessee Texas Utah Wisconsin 1000

800

600

400

Mortality per 100 000 200

0 Probation Local jail State Parole Probation Local jail State Parole Probation Local jail State Parole Probation Local jail State Parole prison prison prison prison

Figure 4: Mortality of individuals on probation, incarcerated in local jails, incarcerated in state prisons, and on parole in 15 US states, 2000–12 Calculations based on data from the Annual Parole Survey and Annual Probation Survey;96 and from Noonan and Ginder (2014).97 in key ways (eg, to limit stops to those absolutely quickened with more sweeping reforms in drug necessary and to promote less adversarial contact sentencing, reduced admissions of technical parole between the police and the community than frequently violators, expanded community corrections for those occurs at present). The shift in the nation’s approach to convicted of low-level property and drug crimes, and criminal justice and drug sentencing has led to a small medical paroles for sick and elderly inmates. Those decrease in the prison population, a fall of 2·9% since its concerned about mass incarceration—and health peak in 2009.95 The pace of downsizing could be disparities—should advocate for such reforms, in www.thelancet.com Vol 389 April 8, 2017 1471 Series

conjunction with improved access to health care and contribute to—health inequity, and facilitate undoing the social services for individuals who have been affected by damage it has caused. But research is not enough to stem the criminal justice system. However, even these changes the health effects of mass incarceration on individuals, would cut the penal population by just 30%, because families, and communities, or to mitigate existing health much of the increase in incarceration is due to the inequities. As physicians and researchers, we should adoption of long sentences for violent offenders. engage in conversations about the interplay between Moreover, shrinking the imprisoned population size racism, social control, and health. Such discussions must while expanding the population under social control in also address the health consequences of living in a the community will probably not ameliorate health community subject to overly aggressive policing, and inequity unless the roots of mass incarceration engage communities of colour to build trust, develop are addressed through broader efforts to provide solutions, and ultimately improve health outcomes. opportunities and conditions for people in marginalised Contributors communities to improve their lives. Our analyses of data CW and EAW contributed equally to all components of this Series paper. from the Bureau of Justice Statistics highlight this point. Declaration of interests As shown in figure 4, the crude mortality of probationers We declare no competing interests. and parolees exceeds that of state prisoners and jail Acknowledgments inmates in nearly every state.96,97 These data are CW is supported by the Rockwool Foundation and by a visiting fellowship unadjusted by age, race, and other traits that might from the Bureau of Justice Statistics. EAW is supported by a visiting fellowship from the Bureau of Justice Assistance. The views expressed in account for these differences. However, until US data this Series paper should not be considered those of the Bureau of Justice collection systems are redesigned, we will not know the Statistics, the Bureau of Justice Assistance, the Office of Justice Programs, health risks for the roughly 6 million Americans on or the National Institute of Justice. probation. References Were the USA to return to the levels of incarceration of 1 Garland D. The meaning of mass imprisonment. Punishm Soc 2001; 3: 5–7. the 1970s, at least a half a million people employed by the 2 Fazel S, Baillargeon J. The health of prisoners. Lancet 2011; criminal justice system could lose their jobs, and 377: 956–65. 63 million individuals would still have criminal records. 3 Massoglia M, Pridemore WA. Incarceration and health. The criminal justice system is so deeply rooted in Annu Rev Sociol 2015; 41: 291–310. 4 Dumont DM, Brockmann B, Dickman S, Alexander N, Rich JD. America’s political system and socioeconomic structures Public health and the epidemic of incarceration. that the damage to the health of our communities cannot Annu Rev Public Health 2012; 33: 325–39. be mitigated without addressing the root causes of mass 5 Wildeman C. Incarceration and population health in wealthy incarceration and the forces that inevitably seek to democracies. Criminol 2016; 54: 360–82. 6 Stuckler D, Basu S, McKee M, King L. Mass incarceration can maintain it. explain population increases in TB and multidrug-resistant TB in True understanding of the health consequences of European and central Asian countries. Proc Natl Acad Sci USA mass incarceration and necessary changes in criminal 2008; 105: 13280–85. 7 Blumstein A, Cohen J. Theory of the stability of punishment. justice policy requires investment in improved data, the J Crim Criminol 1973; 64: 198–207. inclusion of several questions about criminal justice 8 Cahalan MW. Historical corrections statistics in the United States, exposure in national health surveys, and the linkage of 1850–1984. Washington, DC: US Department of Justice, Bureau of Justice Statistics, 1986. administrative data about incarceration with data for 9 Pratt J. Scandinavian exceptionalism in an era of penal excess. income, employment, housing, educational outcomes, Part I: the nature and roots of Scandinavian exceptionalism. and health services. We should also invest in a host of Br J Criminol 2008; 48: 119–37. experiments that test interventions to mitigate the 10 Western B. Punishment and inequality in America. New York, NY: Russell Sage Foundation, 2006. negative direct and indirect health effects of incarceration 11 Wakefield S, Uggen C. Incarceration and stratification. and community corrections, and seek perspectives Annu Rev Sociol 2010; 36: 387–406. that facilitate engagement between physicians, 12 Muller C. Northward migration and the rise of racial disparity in epidemiologists, and criminologists.98 There is a special American incarceration, 1880–1950. Am J Sociol 2012; 118: 281–326. 13 Pettit B, Western B. Mass imprisonment and the life course: race and urgency to develop effective interventions that take into class inequality in U.S. incarceration. Am Sociol Rev 2004; 69: 151–69. account the experiences of women and juvenile 14 Sampson RJ, Loeffler C. Punishment’s place: the local offenders. There is also a pressing need to consider concentration of mass incarceration. Daedalus 2010; 139: 20–31. 15 Western B, Wildeman C. The black family and mass incarceration. whether differential processing within the criminal Ann Am Acad Pol Soc Sci 2009; 621: 221–42. justice system of individuals with mental health 16 Wildeman C. Parental imprisonment, the prison boom, and the conditions—wherein individuals with more resources concentration of childhood disadvantage. Demography 2009; are able to gain access to treatment (and avoid jail or 46: 265–80. 17 Resnick J, Kalb J, Baumgartel S, Guilmette C, Nuni J, Porter DE. prison stay) and those with fewer resources are not— Time-in-cell: the ASCA-Liman 2014 National Survey of Administrative could contribute to growing racial, ethnic, and class Segregation in Prison. New Haven, CT: Yale Law School, 2015. disparities in physical and mental health. 18 Glaze LE, Maruschak LM. Parents in prison and their minor children. Washington, DC: US Department of Justice, Office of Improved data would help us to understand how mass Justice Programs, 2008. https://www.bjs.gov/content/pub/pdf/ incarceration has contributed to—and will continue to pptmc.pdf (accessed Feb 2, 2017).

1472 www.thelancet.com Vol 389 April 8, 2017 Series

19 Lee H, McCormick T, Hicken MT, Wildeman C. Racial inequalities 46 Visher CA, Mallik-Kane K. Reentry experiences of men with health in connectedness to imprisoned individuals in the United States. problems. In: Greifinger RB, ed. Public health behind bars. Du Bois Rev 2015; 12: 269–82. New York, NY: Springer, 2007: 434–60. 20 Dolan K, Wirtz AL, Moazen B, et al. Global burden of HIV, viral 47 Baillargeon J, Giordano TP, Rich JD. Accessing antiretroviral hepatitis, and tuberculosis in prisoners and detainees. Lancet 2016; therapy following release from prison. JAMA 2009; 301: 848–57. 388: 1089–102. 48 Frank JW, Wang EA, Nunez-Smith M, Lee H. Discrimination 21 Rich JD, Beckwith CG, Macmadu A, et al. Clinical care of based on criminal record and healthcare utilization among men incarcerated people with HIV, viral hepatitis, or tuberculosis. recently released from prison: a descriptive study. Health Justice Lancet 2016; 388: 1103–14. 2014; 2: 1–8. 22 Kamarulzaman A, Reid SE, Schwitters A, et al. Prevention of 49 Schnittker J, Massoglia M, Uggen C. Out and down: incarceration transmission of HIV, hepatitis B virus, hepatitis C virus, and and psychiatric disorders. J Health Soc Behav 2012; 53: 448–64. tuberculosis in prisoners. Lancet 2016; 388: 1115–26. 50 Cuellar AE, Cheema J. As roughly 700,000 are released annually, 23 Rubenstein LS, Amon JJ, McLemore M, et al. HIV, prisoners, and about half will gain health coverage and care under federal laws. human rights. Lancet 2016; 388: 1202–14. Health Aff2012; 31: 931–38. 24 Telisinghe L, Charalambous S, Topp SM, et al. HIV and 51 Cuellar AE, Cheema J. Health care reform, behavioral health, and tuberculosis in prisons in sub-Saharan Africa. Lancet 2016; the criminal justice population. J Behav Health Serv Res 2014; 388: 1215–27. 41: 447–59. 25 Altice FL, Azbel L, Stone J, et al. The perfect storm: incarceration and 52 Binswanger IA, Stern MF, Deyo RA, Heagerty PJ. Release from the high-risk environment perpetuating transmission of HIV, prison—a high risk of death for former inmates. N Engl J Med 2007; hepatitis C virus, and tuberculosis in Eastern Europe and Central Asia. 356: 157–65. Lancet 2016; 388: 1228–48. 53 Massoglia M, Pare PP, Schnittker J, Gagnon A. The relationship 26 Barnert ES, Perry R, Morris RE. Juvenile incarceration and health. between incarceration and premature adult mortality: Acad Pediatr 2016; 16: 99–109. gender specific evidence.Soc Sci Res 2014; 46: 142–54. 27 Stokes ML, McCoy KP, Abram KM, Byck GR, Teplin LA. 54 Wang EA, Pletcher M, Lin F, et al. Incarceration, incident Suicidal ideation and behavior in youth in the juvenile justice system: hypertension, and access to health care: findings from the a review of the literature. J Correct Health Care 2015; 21: 222–42. Coronary Artery Risk Development in Young Adults (CARDIA) 28 Prins SJ. Prevalence of mental illnesses in US state prisons: Study. Arch Intern Med 2009; 169: 687–93. a systematic review. Psychiatr Serv 2014; 65: 862–72. 55 Wang EA, McGinnis KA, Long JB, Akgün KM. Incarceration and 29 Rosen DL, Schoenbach VJ, Wohl DA. All-cause and cause-specific health outcomes in HIV-infected patients: the impact of substance mortality among men released from state prison, 1980–2005. use, primary care engagement, and antiretroviral adherence. Am J Public Health 2008; 98: 2278–84. Am J Addict 2015; 24: 178–84. 30 Spaulding AC, Seals RM, McCallum VA, Perez SD, Brzozowski AK, 56 Massoglia M. Incarceration as exposure: the prison, infectious Steenland NK. Prisoner survival inside and outside of the disease, and other stress-related illnesses. J Health Soc Behav 2008; institution: implications for health-care planning. Am J Epidemiol 49: 56–71. 2011; 173: 479–87. 57 Wang EA, Green J. Incarceration as a key variable in racial 31 Patterson EJ. Incarcerating death: mortality in U.S. state disparities of asthma prevalence. BMC Public Health 2010; correctional facilities, 1985–1998. Demography 2010; 47: 587–607. 10: 29–37. 32 Rosen DL, Wohl DA, Schoenbach VJ. All-cause and cause-specific 58 Wang EA, Aminawung JA, Wildeman C, Ross JS, Krumholz HM. mortality among black and white North Carolina state prisoners. High incarceration rates among black men enrolled in clinical Ann Epidemiol 2011; 21: 719–26. studies may compromise ability to identify disparities. 33 Wang EA, Hong CS, Shavit S, Kessell E, Sanders R, Kushel MB. Health Aff2014; 33: 848–55. Engaging individuals recently released from prison into primary 59 Geller A, Garfinkel I, Western B. Paternal incarceration and support care: a randomized trial. Am J Public Health 2012; 9: e22–29. for children in fragile families. Demography 2011; 48: 25–47. 34 Wilper AP, Woolhandler S, Boyd JW, Lasser KE. The health and 60 Comfort M. Punishment beyond the legal offender. health care of US prisoners: results of a nationwide survey. Annu Rev Law Soc Sci 2007; 3: 271–96. Am J Public Health 2009; 99: 666–72. 61 Lopoo LM, Western B. Incarceration and the formation and stability 35 Freudenberg N, Heller D. A review of opportunities to improve the of marital unions. J Marriage Fam 2005; 67: 721–34. health of people involved in the criminal justice system in 62 Comfort M. Doing time together: love and family in the shadow of the United States. Annu Rev Public Health 2016; 37: 313–33. the prison. Chicago, IL: University of Chicago Press, 2009. 36 Fazel S, Danesh J. Serious mental disorder in 23 000 prisoners: 63 Braman D. Doing time on the outside: incarceration and family life in a systematic review of 62 surveys. Lancet 2002; 359: 545–50. urban America. Ann Arbor, MI: University of Michigan Press, 2004. 37 Nwosu BU, Maranda L, Berry R, et al. The vitamin D status of 64 Turney K, Schnittker J, Wildeman C. Those they leave behind: prison inmates. PLoS One 2014; 9: e90623. paternal incarceration and maternal instrumental support. 38 Schnittker J, John A. Enduring stigma: the long-term effects of J Marriage Fam 2012; 74: 1149–65. incarceration on health. J Health Soc Behav 2007; 48: 115–30. 65 Wildeman C. Imprisonment and infant mortality. Soc Probl 2012; 39 Meyer JP, Cepeda J, Springer SA, Wu J, Trestman RL, Altice FL. 59: 228–57. HIV in people reincarcerated in Connecticut prisons and jails: 66 Wildeman C, Andersen SH, Lee H, Karlson KB. an observational cohort study. Lancet HIV 2014; 1: e77–84. Parental incarceration and child mortality in Denmark. 40 Turney K, Wildeman C, Schnittker J. As fathers and felons: Am J Public Health 2014; 104: 428–33. explaining the effects of current and recent incarceration on major 67 Roettger ME, Boardman JD. Parental incarceration and depression. J Health Soc Behav 2012; 53: 465–81. gender-based risks for increased body mass index: evidence from 41 Wildeman C, Turney K, Schnittker J. The hedonic consequences of the National Longitudinal Study of Adolescent Health in punishment revisited. J Crim Law Criminol 2014; 104: 133–63. the United States. Am J Epidemiol 2012; 175: 636–44. 42 Gawande A. Hellhole. New Yorker 2009; 85: 36–49. 68 Boch SJ, Ford JL. C-reactive protein levels among U.S. adults 43 Haney C. Mental health issues in long-term solitary and ‘supermax’ exposed to parental incarceration. Biol Res Nurs 2015; 17: 574–84. confinement.Crime Delinquency 2003; 49: 124–56. 69 Wakefield S, Wildeman C. Children of the prison boom: mass 44 Kaba F, Lewis A, Glowa-Kollisch S, Hadler J. Solitary confinement incarceration and the future of American inequality. and risk of self-harm among jail inmates. Am J Public Health 2014; Oxford: Oxford University Press, 2013. 104: 442–47. 70 Roettger ME, Swisher RR, Kuhl DC, Chavez J. 45 Patterson EJ, Wildeman C. Mass imprisonment and the life Paternal incarceration and trajectories of marijuana and other course revisited: cumulative years spent imprisoned and marked illegal drug use from adolescence into young adulthood: evidence for working-age black and white men. Soc Sci Res 2015; from longitudinal panels of males and females in the United States. 53: 325–37. Addiction 2011; 106: 121–32.

www.thelancet.com Vol 389 April 8, 2017 1473 Series

71 Wildeman C, Muller C. Mass imprisonment and inequality in 86 Frank JW, Hong CS, Subramanian SV, Wang EA. health and family life. Annu Rev Law Soc Sci 2012; 8: 11–30. Neighborhood incarceration rate and asthma prevalence in 72 Foster H, Hagan J. Punishment regimes and the multilevel effects New York City: a multilevel approach. Am J Public Health 2013; of parental incarceration: intergenerational, intersectional, and 103: e38. interinstitutional models of social inequality and systemic 87 Khan MR, Wohl DA, Weir SS, Adimora AA. Incarceration and risky exclusion. Annu Rev Sociol 2015; 41: 135–58. sexual partnerships in a southern US city. J Urban Health 2008; 73 Porter LC, King RD. Absent fathers or absent variables? A new look 85: 100–13. at paternal incarceration and delinquency. J Res Crime Delinquency 88 Khan MR, Doherty IA, Schoenbach VJ, Taylor EM, Epperson MW, 2015; 52: 414–43. Adimora AA. Incarceration and high-risk sex partnerships among 74 Turney K. Stress proliferation across generations? Examining the men in the United States. J Urban Health 2009; 86: 584–601. relationship between parental incarceration and childhood health. 89 Thomas JC, Torrone E. Incarceration as forced migration: effects on J Health Soc Behav 2014; 55: 302–19. selected community health outcomes. Am J Public Health 2006; 75 Lee RD, Fang X, Luo F. The impact of parental incarceration on the 96: 1762–65. physical and mental health of young adults. Pediatrics 2013; 90 Rogers SM, Khan MR, Tan S, Turner CF. Incarceration, high-risk 131: e1188. sexual partnerships and sexually transmitted infections in an urban 76 Hagan J, Foster H. Children of the American prison generation: population. Sex Transm Infect 2012; 88: 63–68. student and school spillover effects of incarcerating mothers. 91 Schnittker J, Uggen C, Shannon SKS, Mcelrath SM. Law Soc Rev 2012; 46: 37–69. The institutional effects of incarceration: spillovers from criminal 77 Huebner BM, Gustafson R. The effect of maternal incarceration on justice to health care. Milbank Q 2015; 93: 516–60. adult offspring involvement in the criminal justice system. 92 Wildeman C. Imprisonment and (inequality in) population health. J Crim Justice 2007; 35: 283–96. Soc Sci Res 2012; 41: 74–91. 78 Cho RM. Impact of maternal imprisonment on children’s 93 Johnson R, Raphael S. The effects of male incarceration dynamics probability of grade retention. J Urban Econ 2009; 65: 11–23. on Acquired Immune Deficiency Syndrome infection rates among 79 Wildeman C, Turney K. Positive, negative, nor null? The effects of African American women and men. J Law Econ 2009; 52: 251–93. maternal incarceration on children’s behavioral problems. 94 Pettit B. Invisible men: mass incarceration and the myth of black Demography 2014; 512: 1041–68. progress. New York, NY: Russell Sage Foundation, 2012. 80 Turney K, Wildeman C. Detrimental for some? Heterogeneous 95 Carson EA, Golinelli D. Prisoners in 2012: trends in admissions and effects of maternal incarceration on child wellbeing. releases, 1991–2012. Washington, DC: Bureau of Justice Statistics, Criminol Public Policy 2015; 14: 125–56. 2013. https://www.bjs.gov/content/pub/pdf/p12tar9112.pdf 81 Wildeman C, Schnittker J, Turney K. Despair by association? (accessed Feb 2, 2017). The mental health of mothers with children by recently incarcerated 96 Bureau of Justice Statistics. Annual Probation Survey and Annual fathers. Am Sociol Rev 2012; 77: 216–43. Parole Survey. Probation data available from 1977–2014. Parole data 82 Lee H, Wildeman C, Wang EA, Matusko N, Jackson JS. A heavy available from 1975–2014. Washington, DC: Bureau of Justice burden: the cardiovascular health consequences of having a family Statistics, 2013. https://www.bjs.gov/index.cfm?ty=dcdetail&iid=271 member incarcerated. Am J Public Health 2014; 104: 421–27. (accessed Feb 2, 2017). 83 Western B, Braga AA, Davis J, Sirois C. Stress and hardship after 97 Noonan ME, Ginder S. Mortality in local jails and state prisons, prison. Am J Sociol 2015; 120: 1512–47. 2000–2012 — statistical tables. Washington, DC: Bureau of Justice 84 Wildeman C. Paternal incarceration and children’s physically Statistics, 2014. https://www.bjs.gov/content/pub/pdf/mljsp0012st. aggressive behaviors: evidence from the Fragile Families and Child pdf (accessed Feb 2, 2017). Wellbeing Study. Soc Forces 2010; 89: 285–309. 98 Akers TA, Lanier MM. “Epidemiological criminology”: coming full 85 Clear TR. The effects of high imprisonment rates on communities. circle. Am J Public Health 2009; 99: 397–402. Crime Justice 2008; 37: 97–132.

1474 www.thelancet.com Vol 389 April 8, 2017 Series

America: Equity and Equality in Health 5 Population health in an era of rising income inequality: USA, 1980–2015

Jacob Bor, Gregory H Cohen, Sandro Galea

Income inequality in the USA has increased over the past four decades. Socioeconomic gaps in survival have also Lancet 2017; 389: 1475–90 increased. Life expectancy has risen among middle-income and high-income Americans whereas it has stagnated See Editorial page 1369 among poor Americans and even declined in some demographic groups. Although the increase in income See Comment pages 1376 inequality since 1980 has been driven largely by soaring top incomes, the widening of survival inequalities has and 1378 occurred lower in the distribution—ie, between the poor and upper-middle class. Growing survival gaps across This is the fifth in a Series of income percentiles since 2001 reflect falling real incomes among poor Americans as well as an increasingly strong five papers about equity and association between low income and poor health. Changes in individual risk factors such as smoking, obesity, and equality in health in the USA substance abuse play a part but do not fully explain the steeper gradient. Distal factors correlated with rising Department of Global Health (J Bor ScD) and Department of inequality including unequal access to technological innovations, increased geographical segregation by income, Epidemiology (J Bor, reduced economic mobility, mass incarceration, and increased exposure to the costs of medical care might have G H Cohen MPhil, reduced access to salutary determinants of health among low-income Americans. Having missed out on decades Prof S Galea MD), Boston of income growth and longevity gains, low-income Americans are increasingly left behind. Without interventions University School of Public Health, Boston, MA, USA to decouple income and health, or to reduce inequalities in income, we might see the emergence of a 21st century Correspondence to: health-poverty trap and the further widening and hardening of socioeconomic inequalities in health. Prof Jacob Bor, Department of Global Health, Boston University Introduction to the extent that socioeconomic inequalities in health are School of Public Health, Boston, Income inequality in the USA has increased dramatically avoidable, measurement of health gaps helps to identify MA 02118, USA [email protected] over the past four decades. The share of incomes going to opportunities to target future interventions. See Online for infographic the wealthiest 10% increased from 33% of total earnings www.thelancet.com/ in 1978 to 50% in 2014—a level of inequality not seen Search strategy and selection criteria infographics/us-health since before the Great Depression.1 Incomes for poor and middle-income Americans have barely changed since the We reviewed studies reporting trends in survival inequalities in the 1970s and, adjusted for inflation, have actually declined USA for the period 1980–2015. We searched PubMed and Google since 2000 (figure 1). Given the strong and nearly Scholar for English-language articles published between Jan 1, universal association between socioeconomic status and 1990, and Oct 30, 2016, using the following terms: “((income) or health,2–6 it is natural to ask whether rising gaps in (education)) and ((mortality) or (life expectancy)) and (united income might be associated with widening gaps in health states) and ((longitudinal) or (changes) or (trends) or (over time)).” and longevity between rich and poor Americans. We then added the terms “((inequality) or (disparity))” to the first Our review focuses on income-related and education- search. We manually searched the bibliographies of relevant articles related inequalities in health. Other papers in the Series and used Google Scholar to identify articles citing the relevant address racial inequalities in health—particularly the article. We included articles if they compared survival in a strikingly worse health outcomes of African-Americans— high-socioeconomic status group versus a low-socioeconomic and review the implications of economic inequality for status group, stratified by income or education, at least a decade the US medical care system. apart during the period 1980–2015. Articles were critically appraised to assess socioeconomic strata compared, survival Why socioeconomic inequalities in health matter outcomes measured, whether a period versus cohort Measurement of socioeconomic inequalities in health— perspective was used, whether adjustments were made for also known as health inequalities—is important for several changing composition across strata, and whether the data reasons. Health is a key aspect of wellbeing and health presented reflect measured outcomes or projections. inequalities can compound (or mitigate) existing differ­ A listing of all studies identified in our review and key results ences in wellbeing according to income. Knowledge of the can be found in the appendix. Our review highlights studies distribution of health across income groups offers a more that presented raw survival measures by socioeconomic complete picture of how wellbeing is distributed across group in an early year and a late year (or gaps in an early and society and the effects of policy options. For example, late year), such that changes in survival gaps could be because the poor die younger, they are short-changed by calculated. Findings of studies reporting only relative age-targeted entitlement programmes such as Medicare measures were also generally consistent, and are presented (the public health insurance programme for elderly people) in the appendix and discussed in the text. and the Social Security retirement system.7 Additionally, www.thelancet.com Vol 389 April 8, 2017 1475 Series

divergence in health (by income) due to a direct causal Key messages relationship between individual-level income and 17–19 • Gaps in mortality and life expectancy according to income and education have widened health. Second, income inequality might also have an during the period 1980–2014. indirect effect on health—eg, by strengthening the • Life expectancy has stagnated and even declined in some groups—eg, white women political power of the wealthy with a resulting shift in 20,21 22 without a high school diploma. policy priorities or via changes in social capital (the • Growth in health inequalities has been most pronounced in the bottom half of the shared connections and values between people that income distribution. Since 2001, the poorest 5% of Americans experienced close to encourage advantageous cooperation) or social 23 zero gains in survival. At the same time, middle-income and high-income Americans cohesion. Studies of inequality and health have not have gained over 2 years in additional life expectancy. found a consistent indirect effect of inequality on • Widening gaps in survival across income percentiles since 2001 are only partially aggregate health outcomes; however, this literature is 24,25 explained by falling incomes among the poor. The relationship between income and contested. Still, the largest rise in income inequality survival has also grown stronger, with poverty becoming an increasingly important since the Great Depression offers a case study to risk factor for mortality in the early 21st century. investigate how, if at all, changes in income distribution • The income–survival gradient has become steeper, in part, because of several might be linked to inequalities in health. proximate factors: a shift in the burden of smoking and (to a lesser extent) obesity, In this review, we synthesise evidence on trends in underuse of essential medical care, and increased substance abuse and self-harm in income-related and education-related survival inequalities, lower-socioeconomic status groups. and discuss proximate and distal mechanisms that might • Distal explanations for the rising gradient are harder to evaluate, but might include explain these trends. differential changes in access to and uptake of determinants of health, ranging from fresh fruit and vegetables to information on new risk factors to new medical Growing gaps in survival by income: a review procedures; increasing geographical segregation that exacerbates inequalities in Absolute gaps in survival between rich and poor have access to salutary health amenities; reduced economic mobility leading to greater grown over the past 40 years (table, figure 2; see appendix persistence of poverty; mass incarceration; and, possibly, the erosion of public for a complete list of studies reviewed), with nearly all subsidies that previously shielded the poor from exposure to the market-determined studies reviewed reporting increases in health inequalities prices of health insurance and other health inputs. These factors might have reduced regardless of choice of socioeconomic status measure access to key salutary determinants of health among low-income Americans, (income vs education), choice of outcome (mortality rates exacerbating income losses that have occurred since 2001. vs life expectancy), and time period covered. Increases in • Changes in survival gaps over time might substantially underestimate the growth in survival inequalities were qualitatively large, with life health inequalities across birth cohorts. The gap in cohort life expectancy between the expectancy gaps between high and low socioeconomic top and bottom income quintiles is projected to increase by nearly a decade within a status groups growing by 1–2 years and even more in single generation—from the 1930–60 birth cohort. some demographic groups. With a few exceptions, the • Rising health inequalities should be addressed to avoid the emergence of a smallest absolute increases in survival gaps were observed 21st-century health-poverty trap. in non-elderly populations with lower baseline mortality and in fact represented large relative increases in health inequalities (table).26,33,35 See Online for appendix Health inequalities are particularly pernicious because In addition to general agreement across studies that they can reproduce and reinforce gaps in income and health inequalities have widened, three noteworthy wealth, with negative feedback loops creating a health- patterns emerged. First, we noted consistent differences poverty trap. Poor health can limit economic productivity,8 by gender. Baseline survival gaps were generally larger bankrupt households,9 and impoverish families.10 Poor among men than among women; yet the growth in health in childhood can also have long-term con­sequences survival gaps among women was equal to—or exceeded— for physical and cognitive development,11,12 educational the growth in survival gaps among men, with the result attainment,13,14 and future earnings.14,15 Health deficits that health inequalities increased much faster for women might even have intergenerational consequences,16 than for men in relative terms. Second, life expectancy reproducing inequalities in income, wealth, and human actually fell in some categories, particularly in white capital (ie, the education, training, and health that facilitate women with low income or educational attainment.29,31,32 occupational success) in future generations.12 Evidence on survival trends in black women with low Health inequalities should be understood and addressed socioeconomic status was mixed and more limited.29,31 regardless of whether they are causally linked to income Survival gaps did not increase among Hispanic inequality. However, if the relationship were causal then Americans.31 Third, studies that assessed mortality widening health gaps would be a troubling social cost of experiences across birth cohorts, rather than differences rising income inequality, and income redistribution could across periods, reported much larger increases in survival be an important policy lever to improve population health. disparities.7,27 This finding is to be expected: if cohort Rising income inequality could cause increasing life expectancy is increasing over time, then period life health inequalities through two separate channels. First, expectancy, which averages across cohorts, will under­ divergence in incomes might lead mechanically to estimate those increases. Cohort measures involve

1476 www.thelancet.com Vol 389 April 8, 2017 Series

projections of survival patterns at older ages and thus Top 1% contain substantial uncertainty. At the same time, the 90th percentile 80th percentile period measures reported in most studies might 200 1000

60th percentile Annual household income, substantially underestimate the actual growth in 40th percentile socioeconomic status-related health inequalities from 20th percentile 10th percentile

one generation to the next. top 1% ($1000s) 150 800 Several studies documented widening survival gaps by income levels. Cristia26 linked data from the Survey of Income and Program Participation to tax and Social Security records and assessed trends in survival across 100 600 top 1% ($1000s ) quintiles of long-run earnings, comparing an early period centred at 1990 and a later period centred at 2000. Life expectancy between 35 and 75 years of age increased for 50 400 wealthy men and women, but did not change for poor men and actually declined for poor women (table). The top-to- Annual household income, excluding bottom life expectancy differential across income quintiles 0 200 increased from 2·7 to 3·6 years for men and 0·7 to 1·5 years 1970 1980 1990 2000 2010 Year for women. Cristia further reports that the ratio of mortality in poor versus rich Americans approximately doubled for Figure 1: Inflation-adjusted annual household income at selected percentiles, 1967–2014 35–64 year olds, with smaller increases among 65–75 year All series show percentiles of the distribution except for top 1%, which shows the mean of the top 1%. All income 26 series except for the top 1% are plotted against the left vertical axis, displaying incomes from $0 to $200 000. olds. Cristia’s findings are consistent with those of The top 1% is plotted against the right vertical axis, displaying incomes from $200 000 to $1 000 000. Data are 35 36 Krieger and colleagues and Singh and Siahpush, who from the US Census Bureau Current Population Survey, 1968–2015 Annual Social and Economic Supplements and found widening survival gaps during the 1980s and 1990s World Wealth and Income Database. Income is expressed in 2014 US$. by county-level income and deprivation measures. Income-based survival gaps continued to widen in the for a given quantile as well as changes in the relationship For results of the US Census 2000s. Chetty and colleagues6 linked income tax records to between income and health (panel 1). Bureau Current Population survival information collected by the Social Security system. A natural question is whether the expansion of health Survey see http://www.census. gov/programs-surveys/cps.html Between 2001 and 2014, the life expectancy gradient across inequalities across income quantiles is simply a For more on the World Wealth income quantiles became noticably steeper (table). Men mechanical result of rising income inequality. If the and Income Database see and women in the poorest 5% of households experienced relationship between income and survival were held http://wid.world close to zero net increase in life expectancy during this constant, increased dispersion in the earnings period, whereas women in the top half of the income distribution would be expected to lead to growing gaps in distribution—and men in the top quarter of the survival as the average incomes of the top and bottom distribution—gained more than 2·5 years of life expectancy quantiles diverge. The alternative (and perhaps (figure 3).6 The sharpest divergence in longevity occurred complementary) hypothesis is that health inequalities between poor and upper-middle-income Americans. We across income quantiles might have widened because the found no change in the survival gap between upper-middle- relationship between actual income levels and health (ie, income and high-income Americans. the slope of the gradient) has changed over time. We note that although health inequalities widened To shed light on the contribution of distribution specifically among lower-income and middle-income changes versus slope changes, we decomposed the Americans during the period 2001–14, a very strong quantile-based results from Chetty and colleagues6 using gradient remained between income percentile and health their supplementary data (figure 4). We found that both For more on the data compiled throughout the income distribution, with no evidence of of these explanations have played a part in widening by Chetty and colleagues see a threshold effect; the differences in life expectancy health inequalities. Since 2001, annual inflation-adjusted https://healthinequality.org between the wealthiest 1% and poorest 1% were household earnings have fallen for Americans in the 14·6 years for men and 10·1 years for women. The bottom two thirds of the income distribution (figure 4), importance of income across the full distribution echoes with the largest relative losses experienced in the lower findings from other individual-level studies, most part of the income distribution. From 2001 to 2014, famously the Whitehall study3 of British civil servants. earnings fell by 17% for men and women in households A strength of Chetty and colleagues’ and Cristia’s at the 25th income percentile. Americans at the top of the approach is that they adjust for changing composition by distribution experienced much smaller (or non-existent) assessing survival trends within income quantiles, rather relative losses. The decline in incomes among lower- than levels. This technique reduces the potential for earners and moderate-earners is a new trend since 2000, selection bias that emerges when the proportion of the reversing—and erasing—previous earnings gains during population in different socioeconomic status categories the 1990s (figure 1). shifts over time. At the same time, health trends by Falling incomes among poor Americans were one income quantile might reflect both changes in income explanation for the divergence of life expectancy across www.thelancet.com Vol 389 April 8, 2017 1477 Series

Age Socioeconomic Early Late Adjusts for Measure of survival Change in Change in Survival gap, Survival gap, Absolute change (years) groups compared year year changes in (units) survival, survival, early year late year in survival gap composition low SES high SES (percent change) Income Cristia (2009)26 Men 35–75 1st vs 5th quintile 1983 2003 Yes Period life expectancy 0·0 0·9 2·7 3·6 0·9 (33%) (years) Women 35–75 1st vs 5th quintile 1983 2003 Yes Period life expectancy –0·4 0·4 0·7 1·5 0·8 (114%) (years) Chetty et al (2016)6 Men ≥40 1st vs 4th quartile 2001 2014 Yes Period life expectancy 1·0 2·6 8·4 10·0 1·6 (19%) (years) Women ≥40 1st vs 4th quartile 2001 2014 Yes Period life expectancy 1·3 3·0 4·5 6·2 1·7 (38%) (years) Men ≥40 1st vs 20th ventile 2001 2014 Yes Period life expectancy 0·3 2·3 11·5 13·5 2·0 (17%) (years) Women ≥40 1st vs 20th ventile 2001 2014 Yes Period life expectancy 0·0 2·9 6·3 9·1 2·9 (44%) (years) Waldron (2007)27 Men ≥65 Top vs bottom half 1912 1941 Yes Cohort life expectancy 1·3 6·0 0·7 5·4 4·7 (671%) cohort cohort (years) NAS (2015)7 Men ≥50 1st vs 5th quintile 1930 1960 Yes Cohort life expectancy –0·5 7·1 5·1 12·7 7·6 (149%) cohort cohort (years) Women ≥50 1st vs 5th quintile 1930 1960 Yes Cohort life expectancy –4·0 5·7 3·9 13·6 9·7 (249%) cohort cohort (years) Education* Ho and Fenelon (2015)28 White men ≥50

1478 www.thelancet.com Vol 389 April 8, 2017 Series

Age Socioeconomic Early Late Adjusts for Measure of survival Change in Change in Survival gap, Survival gap, Absolute change (years) groups compared year year changes in (units) survival, survival, early year late year in survival gap composition low SES high SES (percent change) (Continued from previous page) All women ≥25

Results of all studies identified that assessed time trends in absolute survival gaps by income and education. Studies that adjusted for changes in composition accounted in some way for changing population shares across SES categories. We report absolute changes in survival over time in high-SES and low-SES groups, to highlight differences in the evolution of mortality patterns within groups. We report survival gaps37 between high-SES and low-SES groups in the early period and the late period (early year and late year are the first and last years respectively of the early and late periods compared) to capture how health inequalities have changed. We present the absolute change in survival gaps, a difference-in-differences type measure, to establish whether health inequalities are growing. Widening survival gaps are reflected as positive changes, and are presented in the units of the outcome (years for life expectancy; deaths per 1000 person-years for mortality rates). The percent change in survival gaps enables comparison across different outcomes; however, important differences across the studies in measures and methods used make direct comparisons difficult. SES=socioeconomic status. NAS=National Academies of Sciences, Engineering, and Medicine. *University education refers to undergraduate studies.

Table: Changes in survival gaps by income, education, and county socioeconomic conditions, 1980–2014 income quantiles since 2001 (figure 3); but they were not point of income; by 2014, the gradient had increased to the only—and indeed not the primary—explanation. 2·4 years per log-point. (Incomes increase by a factor of When the income–health gradient was held constant at its 2·7 for every log-point—eg, from US$22 000 to $60 000 to 2001 slope, changes in income distribution explained only $163 000.) In men, the gradient increased from 3·2 years about one third of the increase in the top versus bottom per log-point in 2001 to 3·7 years per log-point in 2014. quartile life expectancy gap in men and about one sixth of The gradient for women is clearly steeper in the later the increase in women. More importantly, the income– period (figure 4). In men, the change in slope—although survival gradient changed as well, becoming substantially similar in absolute magnitude—is more difficult to see stronger over time (figure 4). In 2001, female life because the baseline gradient was steeper and non-linear expectancy was on average 1·7 years higher for every log- (figure 4). www.thelancet.com Vol 389 April 8, 2017 1479 Series

700 60 years of age between 1972 and 1998). The relationship between income and age-specific mortality strengthened 600 across birth cohorts (figure 5). Although the relationship between income and mortality could be expected to be 500 attenuated at older ages because of survivorship bias, income-related mortality odds ratios rose across birth 400 cohorts for all age groups observed. Based on projected future trends in age-specific mortality rates, Waldron27 300 estimated that the gap in male life expectancy at age 65 years between the top and bottom half of the income 200 distribution increased from less than 1 year for men Change in survival gap (%) born in 1912 to more than 5 years for men born in 1941 100 (table).27 Using similar calculations, a 2015 report7 by the 0 National Academies of Sciences, Engineering, and

–100 Medicine (NAS) estimated survival curves for the 1930 Income Education County and 1960 birth cohorts based on data from the Health socioeconomic conditions and Retirement Study. Survival was assessed by quintile of long-run earnings (averaged from ages 41 to 50 years). Figure 2: Percent changes in survival gaps, 1980–2014 The study7 projected that life expectancy at 50 years will Chart shows percent changes in income, education, and county socioeconomic condition-related survival gaps reported in the literature reviewed. Each bar represents a row in the table. The vast majority of studies reported stagnate for men in the poorest quintile and decline for widening survival gaps. women in the bottom two quintiles (figure 5). As a result, the gap in life expectancy at age 50 years between the top For more on the Health and Changes in the income–survival gradient differed across and bottom income quintiles was projected to increase Retirement Study see the income distribution. The life expectancy gradient from 5·1 years for men born in 1930 to 12·7 years for http://hrsonline.isr.umich.edu increased sharply in households earning less than $60 000 men born in 1960; the comparable figures for women per year, with the slope rising by 37% in men (from were 3·9 years and 13·6 years. The gap in cohort life 2·5 to 3·4 years per log-point of income) and 77% in expectancy between the richest and poorest 20% of women (from 1·4 to 2·5 years per log-point). However, the Americans is projected to increase by nearly a decade in a gradient did not change for men and women in single generation. households earning over $60 000 per year (figure 4). Estimates of cohort life expectancy should be In summary, the divergence of life expectancy for poor interpreted cautiously because old-age survival has not and middle-income Americans since 2001 (figure 3) yet been observed in recent birth cohorts, necessitating reflects both changes in income distribution and a rise extrapolation from mortality rates observed at younger in gradient between income and health, with the latter ages; for example, the NAS study7 projected cohort life a more important factor. Both distribution changes expectancy at age 50 years even though no members and slope changes were non-linear and specifically dis­ of the 1960 cohort had reached that age by end of advantaged the poor (figure 4). From 2001 to 2014, poverty follow-up. Nevertheless, these two cohort studies7,27 deepened in the USA; at the same time, poverty emerged highlight the possibility that a slow-moving disaster as an increasingly important risk factor for mortality in might be unfolding for the health of lower-income American adults. Americans who entered the labour market after the The studies discussed above presented period post-World War 2 boom and have spent their working (ie, repeated cross-sectional) estimates of survival which years in a period of rising income inequality. The might underestimate—or overestimate—changes in inscrutability of trends in cohort life expectancy gaps survival gaps across actual birth cohorts. A cohort while they are emerging should not lower our vigilance. perspective has the advantage of capturing a population’s Differences in mortality across birth cohorts can still be full exposure and health history from childhood, documented at younger ages (figure 5). More importantly, mapping health more closely to lived experiences policies could potentially be designed that mitigate the (panel 1). The challenge in assessment of trends across very large survival deficits emerging for recent cohorts of birth cohorts is that age-specific mortality rates can be lower-income Americans (figure 5). definitively known only for those ages observed. Cohort life expectancy estimates have to be projected from Growing gaps in survival by education: a review models unless all members of the birth cohort have died. Survival gaps by educational attainment have also Waldron27 used Social Security data on lifetime income widened in the past three decades.28–34,44–54 Although and mortality to assess age-specific mortality rates of estimates differ depending on groups compared, time men with incomes above and below the median for period analysed, ages at which mortality was assessed, cohorts born between 1912 and 1941 (who reached and whether studies adjusted for composition changes,

1480 www.thelancet.com Vol 389 April 8, 2017 Series

Panel 1: Challenges in measuring trends in health inequalities Studies differed with respect to data sources, indicators of To overcome these limitations, studies have increasingly socioeconomic status and health, and statistical measures used analysed income as observed in tax records or Social Security to summarise health inequalities. data.6,7,26,27 Such data reduce reporting bias and can be averaged across many years to minimise the effects of year-to-year Data sources variation. A so-called washout period (eg, 2 years) can be Studies have used three types of data: (1) nationally representative included to reduce the influence of reverse causality from population-based surveys such as the National Health Interview health shocks. Perhaps most importantly, health outcomes can Survey and National Longitudinal Mortality Study, which collect be assessed across quantiles of the income distribution, self-reported education and income data at a single point in time avoiding the compositional complications of educational and have been linked to subsequent national death records;29,30 attainment or fixed income brackets. To the extent that a (2) income histories from individuals’ tax records that have been person’s rank in the income distribution reflects unobserved linked with death records maintained by the Social Security factors that might be correlated with health, assessment of system;6,7,26,27,38 and (3) US death certificates, which have included secular trends within income quantiles holds these factors information on educational attainment since 1989, and can be constant. At the same time, health trends for specific income combined with Census data to enable assessment of survival quantiles might reflect both changes in income for that trends by education.31,33,34 Long-run trends dating to 1960 have quantile as well as changes in the relationship between income been analysed using county-based measures of socioeconomic and health. status.35,36,39 Measures of health and longevity Measures of socioeconomic status Studies assessing trends in survival have made different Studies have used income, education, and area-based choices regarding the age groups included—eg, 25–65 years,35 measures—each of which has its limitations. Incomes vary 35–75 years,26 40–76 years,6 and 50 years and older.7 Due to across the life course. Earnings measured at a single point in differences in age-specific mortality rates, these choices can time might be a poor reflection of lifetime earnings, particularly substantially influence results. Typically, the association for retirees, and are subject to random fluctuations which between low socioeconomic status and mortality is strongest attenuate the relationship between socioeconomic status and on a relative scale at younger ages, but absolute differences health. Health shocks can also negatively affect earnings,40 are greatest at older ages when background mortality is inflating the cross-sectional association between income and higher. Most studies report age-standardised or age-adjusted health. mortality rates. Many report period life expectancy, Measurement of socioeconomic status by educational combining age-specific mortality rates to create a period attainment overcomes some of these limitations, because survival curve from which life expectancy can be estimated. schooling is largely determined by age 25 years and does not fluctuate. However, the coarseness of education data can make Period versus cohort measures inferences difficult. For example, no clear ordinal ranking for Period survival measures (eg, life expectancy in 2015) are the post-university education exists, so most studies lump together basis for international comparisons and are useful to identify all university graduates. Simply counting years of schooling the effect of exposures that simultaneously affect people of all ignores important variation in the quality of education that ages, such as famine. However, period measures might miss might be associated with both earnings and health. important transformations in the risk faced by different birth Self-reported income and education data are also subject to cohorts—for example, AIDS caused particularly high death rates reporting biases, which could be substantial.41 A key threat to among people who were sexually active during the first decades validity is compositional change within education categories of the epidemic. over time. In 1920, failure to complete high school was Several studies compared survival across different birth cohorts common among people who might today be considered as rather than across different periods.7,44–47 Although cohort middle class; over time, it has become a marker of very low approaches more accurately describe the fate of specific socioeconomic status. Hence, analyses that use a static measure populations, the full cohort survival curve can only be assessed of educational attainment to indicate socioeconomic status after all members have died. Hence, only trends in non-elderly might find a decline over time in life expectancy, even if no real mortality are observed for recent birth cohorts.27 Cohort life change in socioeconomic status-based differences in survival expectancy can be projected by extrapolating current survival has occurred.42 Researchers have addressed this issue by patterns and imputing future age-specific mortality rates,7,27 but randomly reallocating people across education categories to such estimates should be interpreted with caution, as they are equalise the distribution of educational attainment across time projections of future events rather than observed estimates. periods.27,29,30,43 Unfortunately, this approach is not guaranteed to eliminate bias.

www.thelancet.com Vol 389 April 8, 2017 1481 Series

Using a smaller sample enabling analysis of trends Men A 28 0·4 beginning in 1980, Ho and Fenelon compared life expectancy at age 50 years for white university-educated (≥16 years of education) Americans versus white 0·3 Americans with a high-school education or less (≤12 years death (years) of education; table). Between 1980 and 2006, the gap in 0·2 life expectancy increased by 1·1 years (26%) for men and 2·1 years (84%) for women.28 Several other studies reported education-specific trends 0·1 in mortality rates that were consistent with these 33 year in expected age at findings. Ma and colleagues assessed trends in age- 0 standardised premature mortality (ages 25–64 years) from 1993 to 2007, by educational attainment (0–12 years Change per –0·1 of education, 13–15 years, and ≥16 years) as reported in 0 5 10 15 20 death certificates. Premature mortality fell by 34% for Mean household income 30 60 101 683 men with a university education and by just 5% for men ($1000s) with a high-school education or less. For university- B Women educated women, premature mortality fell by 29%, but 0·4 increased by 11% in women with 12 years of schooling or fewer.33 The difference in mortality risk associated with low (vs high) education increased by 14% in men and 0·3 55% in women (table).33 Focusing on white Americans death (years) aged 45–84 years, Montez and Berkman32 found that 0·2 between 1986 and 2006, age-standardised mortality fell by 22% among university-educated men and by 8% 0·1 among men with fewer than 12 years of schooling; mortality fell by 11% in university-educated women but year in expected age at increased by 21% in women with fewer than 12 years of 0 schooling. Case and Deaton34 and Cutler and colleagues54

Change per also report widening gaps in mortality rates by –0·1 educational attainment. 0 5 10 15 20 Household income ventile Changes in the composition of education groups (eg, the Mean household income 27 54 95 653 shrinking proportion of Americans failing to complete ($1000s) high school) might confound analyses of trends in 30 Figure 3: Changes in life expectancy across the income distribution, 2001–14 education-based health inequalities (panel 1). Studies Figure shows average annual change in life expectancy by household income correcting for composition shifts have nonetheless found ventile. Shaded areas show 95% CIs. Reproduced from Chetty et al,6 by growing gaps. Meara and colleagues29 found that between permission of the American Medical Association. 1990 and 2000, mortality at ages 25–84 years fell for people with at least some university education, but not for those the studies we reviewed were broadly consistent in without higher education; the education gap in life pointing to widening survival inequalities by educational expectancy at age 25 years increased by 1·6 years (table). attainment (table, appendix). Adjusting for composition, Hendi30 found that the gap in Combining data from death certificates and US Census life expectancy between white women who completed data, Olshansky and colleagues31 compared period life university and those that did not complete high school expectancy at age 25 years for Americans with a university increased by 2·2 years from 1991 to 2005; however, the gap education (undergraduate; ≥16 years of education) versus for white men rose by only 0·1 years (table). Life expectancy Americans with less than a high-school education declined by 0·7 years for white women with fewer than (<12 years of education; table). From 1990 to 2008, 12 years of schooling.30 Although changes in composition education-related life expectancy gaps increased by played a role—explaining 26% of the widening gap in 0·8 years for men and 2·6 years for women. Trends women and 87% of the widening gap in men by varied by race ethnicity, with life expectancy gaps one estimate30—the studies29,30,52 reviewed nevertheless widening by more than 8 years among white people, found consistent evidence of widening survival about 4 years among black people, and holding steady inequalities, even after adjustment for composition. among Hispanic people. During the 1990s and early Several studies have assessed the education–survival 2000s, life expectancy outright declined by 5·3 years for gradient across successive birth cohorts, rather than women and 3·4 years for men among white people with time periods.44–47 Masters and colleagues45 found that fewer than 12 years of schooling.31 the widening gap in mortality across education groups

1482 www.thelancet.com Vol 389 April 8, 2017 Series

A B 0 Women 0·4 Men

–5 0·3

–10

0·2 –15 year in life expectancy (years)

0·1

Change in household income (%) –20 Change per 0 –25 0 20 40 60 80 100 <5 10 25 50 100 250 500 1000 Household income percentile Household income ($1000s)

C Men D Women 2001–02 2013–14 90 90

87

87 84 death (years)

81 84

Expected age at 78

75 81

<5 10 25 50 100 250 500 1000 <5 10 25 50 100 250 500 1000 Household income ($1000s) Household income ($1000s)

Figure 4: Decomposition of trends: changes in life expectancy across the income distribution, 2001–14 Graphs show (A) percentage change in household income from 2001 to 2014, by income percentile, for men (similar trends were observed for women); (B) the average annual change in life expectancy by household income category over the period 2001–14 (whereas figure 3 shows changes by income percentile); and the income–life expectancy gradient in 2001–02 and 2013–14 for men (C) and women (D), shown as a kernel-weighted moving average. Household income is expressed in 2012 US$, based on US tax records, and is presented on a log scale. In (B), income categories were defined in 0·5 log-point bins: 8·5, 9·0, 9·5, …, 14.0; male datapoints are offset for visualisation purposes. Differential changes in life expectancy by income percentile were the result of both (A) falling incomes among the poor and (B–D) a steeper gradient between income and life expectancy. Data are from Chetty et al.6 was attributable entirely to cohort rather than period groups (table, figure 2, appendix). The trends identified exposures, consistent with the theory that trends in have origins that precede 1980. Studies have identified survival inequalities are driven primarily by differences widening survival gaps in earlier decades not covered by across—rather than within—individuals.45 The gains in our literature search. Among white men aged 25–64 years, survival enjoyed by white men and women with at least a the mortality rate ratio between the top and bottom high-school education were not experienced by black income categories rose from 1·8 in 196055 to 2·2 in men and women, suggesting that the health benefits of 1973–78.56 A study57 comparing 1960 and 1986 mortality education might be modified by factors associated with rates by income and education found a similar pattern. By race and ethnicity.45 Studies by Everett and colleagues44 contrast, a study35 of non-elderly mortality by county and Krueger and colleagues47 also identified increased income quintile identified narrowing survival inequalities education–mortality hazard ratios in more recent between the mid-1960s and 1980. Consolidation of a cohorts. A challenge in these studies is that more recent longer, internally comparable individual-level time-series birth cohorts are observed only at younger ages when the and decomposition into longer-run trends are important education–mortality gradient is typically strongest (eg, avenues for future research. figure 5); separation of cohort effects from age-related Finally, education and income levels are not simply effect modification in a regression context without proxies for a single underlying construct. Rather, they are additional assumptions is difficult. distinct factors that might be related to health in different Taken as a whole, the literature suggests a pattern of ways and which, if causally related, would imply different widening gaps in survival across income and education interventions. www.thelancet.com Vol 389 April 8, 2017 1483 Series

health might be increasing over time for reasons A 2·0 60–64 independent of changes in the labour market value of 65–69 education. Increases in education might have facilitated 70–74 the adoption of and adherence to healthful behaviours, 75–79 1·8 such as condom use,61 smoking cessation,62 and medication compliance,63 as new health information and 64 1·6 interventions have become available. Studies of the so- called sheepskin effect—the benefits of receiving a diploma over and above the effect of a single year of (odds ratio) 1·4 schooling—suggest that both labour market and non- labour market factors might explain the growing 51 1·2 association between education and survival. Another possibility is that education, income, and survival are all Mortality in low-income vs high-income men consequences of in-utero and childhood health exposures 1·0 1912 1916 1920 1924 1928 1932 1936 (which in turn bear the imprint of early-life socio­ Birth cohort economic circumstance).65,66 B With these comments on the relationship between income, education, and health, we now turn to the 90 89 88 evidence for different mechanisms. Why have socio­ economic differences in health widened during the past

85 40 years? We consider specific causes of death and 83 individual-level risk behaviours, often described as 82 proximate causes, as well as more fundamental or distal

death (years) 80 causes that might give rise to changes in the distribution of 80 78 78 specific risk factors and in turn to widening health gaps.67 77 77 76 Proximate factors mediating widening Expected age at 75 health gaps Evidence on mechanisms for widening health gaps comes from cause-specific mortality data—available 70 Q1 Q2 Q3 Q4 Q5 Q1 Q2 Q3 Q4 Q5 from death certificates—and studies on the evolution of 1930 1960 individual-level risk factors across different socio­ Birth cohort economic groups. A leading hypothesis is that widening mortality gaps Figure 5: Widening income-related inequalities in survival across birth cohorts (A) Excess mortality (odds ratio) associated with lifetime income in the lower (vs upper) 50% of the income result from differential trends in individual-level risk distribution is shown at different ages for men born in 1912–38. Birth cohorts were grouped into 3-year intervals; factors. Research has focused on the triad of risk factors the start of the interval is labelled on the horizontal axis. The association between income and mortality increased central to clinical recommendations for the prevention 27 across birth cohorts for all observed age groups. Data are from Waldron (2007). (B) The projected life expectancy and control of cardiovascular and metabolic disease in at age 50 is shown for 1930 and 1960 birth cohorts, by income quintile—Q1 (poorest) to Q5 (richest). Life expectancy is displayed as expected age at death, conditional on reaching age 50. Estimates for 1930 cohort are the USA: do not smoke, maintain a healthy weight, and based on observed and projected age-specific mortality rates; estimates for 1960 cohort are based entirely on adhere to prescribed medication.63,64 Smoking prevalence projections. Underlying data for these projections come from the Health and Retirement Study. Projections has fallen much faster among well-educated and higher- presented are for men; similar divergence in life expectancy by income was projected for female birth cohorts. income Americans than among lower socioeconomic Cohort life expectancy for the bottom two quintiles of the 1930 birth cohort was 76·6 years (Q1) and 77·2 years 28 (Q2). Data are from the National Academies of Sciences, Engineering, and Medicine study.7 status Americans, leading to the emergence of substantial differences in the burden of smoking-related Education is a causal determinant of both earnings and disease.28,54 A similar phenomenon is emerging with health,58,59 and evidence suggests that both of these obesity.68 But how much of the widening survival gap relationships have strengthened over time. Economic do these risk factors explain? returns to higher education have increased in the past Trends in causes of death provide some insight. Across 40 years, due to a phenomenon described by economists birth cohorts, education gaps in heart disease and lung as skill-biased technical change.60 New technologies, cancer mortality increased, but gaps did not widen for chiefly the rise of computers in the office and factory non-preventable cancers, suggesting the importance of floor, have increased the wage returns to education- changes in modifiable risk factors and use of medical related expertise, while reducing demand for less skilled care.45 Between 1968 and 1998, male cardiovascular clerical labour. Divergence in earnings across education mortality declined faster in high-socioeconomic status groups is one reason for the growth in income inequality counties than in low-socioeconomic status counties.36 and might be the cause of growing health inequalities. At The absolute gap in diabetes mortality between university the same time, the relationship between education and graduates and people with a high-school education or

1484 www.thelancet.com Vol 389 April 8, 2017 Series

less increased from 1993 to 2007.33 Mortality due to heart related or alcohol-related) together accounted for half of disease, stroke, and cancer declined in all education the total increase in mortality in adults with a high-school groups, yet absolute differences between groups education or less.34 remained approximately constant, thus increasing Because abuse of alcohol and of prescription or illicit relative differences.33 Among white women without a drugs and self-harm behaviours affect morbidity as well high-school education, mortality from smoking-related as mortality, these mortality trends are probably only the causes (including lung cancer and chronic lower most visible signal of larger trends in chronic physical respiratory disease), diabetes, and Alzheimer’s disease pain, mental illness, and addiction. Indeed, Case and increased between 1986 and 2006, accounting for much Deaton34 note very large increases in poor physical health of the divergence in mortality in this category relative to days (28%), poor mental health days (25%), and the higher education groups.69 proportion of people unable to work because of disability Further evidence can be found from trends in risk (49%) among middle-aged white adults. The rise in behaviours. Smoking patterns have diverged, with an adverse coping behaviours among Americans of low increasing burden among people with the least socioeconomic status—namely, abuse of controlled education.28 However, one study54 found that divergent substances including tobacco, alcohol, and prescription smoking rates were not sufficient to explain the widening drugs, as well as self-harm—should thus be seen in the mortality gap. Another study28 found that smoking trends context of growing epidemics of disability, chronic pain, explained 25–50% of the growth in the life expectancy gap and multiple morbidity71 amidst shrinking capacities across education groups for women, but did not contribute to cope.72 to widening survival gaps in men. Additionally, at the Rising mortality due to adverse coping behaviours county level, increases in the association between explains a substantial portion of the divergence in mortality and county income from 1980 to 2002 were survival chances for low-income and poorly educated attributed in part to changes in mortality unrelated to Americans since the 1990s. smoking.35,39 Among adults, trends in obesity and management of hypertension and cholesterol have been Distal mechanisms for widening health gaps similar across education groups over time, suggesting Distal mechanisms that underlie changing health that these factors do not substantially mediate the changes inequalities are difficult to pin down.53 However, an in survival inequalities. Furthermore, the education– exclusive focus on individual-level behaviour as a mortality gradient increased over time—even after mechanism would miss the larger structural factors that adjusting for these individual-level risk factors.54 In sum, might be driving these trends. We highlight several differential smoking trends by socioeconomic status are potential explanations for why health inequalities have an important part of the explanation, but do not tell the increased in an era of rising income inequality: whole story. Obesity has not contributed substantially to (1) differential adoption of technological innovations in historical trends, although it might lead to future gaps in medicine as well as non-medical health inputs (eg, survival.70 information about complex health risks and preventive Substance abuse and addiction might also be behaviours), with differences driven both by diffusion contributing to widening health inequalities. Ma and behaviour and by cost barriers; (2) increasing geographical­ colleagues’ 2012 analysis33 of premature deaths in segregation that creates inequalities in access to salutary 26 states between 1993 and 2007 concluded that the health amenities; (3) reduced economic mobility leading differential increase in accidental deaths for people with to increased persistence of poverty for low-income 12 years of schooling or fewer (relative to university Americans; (4) the repercussions of rising incarceration graduates) was responsible for 23% of the total increase rates; and (5) the erosion of public subsidies for health in the all-cause-mortality gap among women and 49% of insurance and other health inputs, and increasing the total increase among men. Many of these deaths exposure to market prices. These distal hypotheses are not were attributable to poisoning by prescription drugs, intended to be exhaustive. However, they highlight presumably mostly opioids. potential reasons for the changing relationship between Case and Deaton34 found that between 1999 and 2013, income and health at the bottom of the income all-cause mortality in the USA among white, non-Hispanic distribution. Although a full investigation is beyond the adults aged 45–54 years fell from 235·1 to 178·1 deaths per scope of this paper, we note that several of these 100 000 person-years in university-educated adults, but hypotheses could be linked indirectly to growing income increased from 601·4 to 735·8 deaths per 100 000 person- inequality higher in the distribution and its effect on years in those without higher education—a 52% increase technological innovation, housing markets, economic in the education-based mortality gradient in just 14 years.34 opportunity, and political appetite for redistributive During this period, deaths from poisonings increased by policies. four times, leap-frogging deaths due to lung cancer in this Innovations in medical care and new information on age group. The increase in deaths from poisonings, health risks might lead to widening health inequalities if suicide, and liver disease (plausibly intravenous drug- high-income or well-educated Americans are the first to www.thelancet.com Vol 389 April 8, 2017 1485 Series

adopt new healthful technologies and behaviours.63 Health the increased concentration of economic growth in gaps might shrink as innovations fully diffuse, but could specific regions, the rise in housing prices during the persist with continuous innovations. Several factors are at 1990s and early 2000s, which has priced many Americans play here. First, education and income might improve out of specific geographical housing and rental markets, people’s ability to consider and incorporate new and widening income inequality itself. Residential information on health risks (or prevention strategies) into segregation might translate into health inequalities their decision making, from smoking to medication through a range of social and environmental exposures adherence, leading to faster adoption.64 Second, the including crime, gun ownership, access to green non-health benefits of many risk behaviours—eg, the space and supermarkets, air pollution, social capital, temporary relief of stress, pain, and hunger afforded by community norms, and local policies that affect tobacco, opioids, and calorie-dense, nutrient-poor foods— population health. Several studies6,35,36,81 have identified might be valued more by low-income Americans who face rising inequalities in health across geographical units more of these stressors. Third, if medical innovation such as US counties, with richer counties pulling away increases the availability of effective but costly from the rest. interventions (and if these interventions are not fully A third possible pathway is the reduction in economic subsidised), then differences in ability to pay might lead to mobility for low-income Americans, which could lead to widening disparities in access. A similar argument applies the hardening of class boundaries and to the persistence to other health inputs, such as the increased availability of of poverty both within and across generations.82,83 Because healthy, yet costly, food options. Fourth, not all innovation current earnings are correlated with future earnings, is healthful, and even health-promoting innovations such rising income inequality implies even larger gaps in as pain relievers might be marketed for abuse. High- future income. Furthermore, this correlation might have socioeconomic status Americans might have more increased over time with growing inequalities in returns options to mitigate exposure to new health risks. to on-the-job experience in higher (vs lower) skilled The graded diffusion of health inputs by socioeconomic occupations.84 To the extent that health decisions are status is not new. For example, in the early 20th century, made with an eye to the future, increasingly bleak child mortality in the USA declined much faster among prospects for economic mobility might lead to reduced children of professionals than among children of labourers, investment in health and divergence in health behaviours with the discovery of germ theory and hand-washing across income groups. Economic insecurity and information campaigns.73 However, the importance of vulnerability, including anxiety about the possibility of differential adoption has probably inc­reased over time with catastrophic health expenses,9 might also contribute to the rapid growth in accurate health information and stress with direct health consequences.85 The rising effective medical technologies and the relative importance burden of disease associated with substance abuse and of these innovations for health compared with earlier eras. self-harm among low-income Americans might reflect Studies have attributed approximately half of all life lower future optimism and efforts to cope with stressors expectancy gains since 1950 to medical innovations,74 including economic vulnerability. In turn, disinvestments including statins,75 antihyper­tensives,76 and coronary care in health—including use of addictive substances—can units,77 with further gains attributable to changes in health reduce future productivity and earnings.86 The rising risk behaviours such as smoking and diet spurred by the epidemic of opioid addiction, as well as other adverse diffusion of new scientific knowledge.77 Although many of coping behaviours, might signal the emergence of a these innovations pre-dated the period of study, the 1980s 21st century health-poverty trap, a negative feedback loop and 1990s were a period of rapid diffusion, with large between poverty and poor health. declines in smoking prevalence and increases in use of The wave of incarceration that began during the 1970s medication to control cholesterol and hypertension. The is a fourth possible mechanism underlying the widening implications for health inequalities depend on the nature health inequalities between the poor and middle class. of the innovation. New treatments for diabetes and HIV led As discussed in detail in a companion paper87 a large to differential uptake by educational attainment that proportion of poor men, especially black men, are current exacerbated inequalities.63 However, classes of anti­ or former prisoners, and incarceration appears to inflict hypertensives that simplified compliance led to widespread substantial economic and health damage on prisoners, uptake and reduced usage gaps.78,79 their families, and their communities.88 A second potential explanation for rising health A final potential mechanism is the increased exposure inequalities is the increasing geographical segregation of Americans to the prices of health inputs. of high-income and low-income Americans, resulting Technological change has increased the armamentarium in differential access to neighbourhood-level health of effective medication and procedures. Concurrently, amenities. Income segregation across school districts in the proportion of medical costs covered by insurers has the USA increased by 15% between 1990 and 2010, and fallen over time, with rising deductibles and co- within-district segregation across schools increased by payments shifting more of the costs of care onto users. 40%.80 Factors contributing to rising segregation include Although some economists and policy leaders have

1486 www.thelancet.com Vol 389 April 8, 2017 Series

promoted user fees as a way to dampen health-care cost part, the large health gaps between the USA and Europe.100 growth, cost sharing—ie, passing more of the costs for At the same time, health inequalities have persisted—and a given service on to patients—discriminates by ability grown—even in the most redistributive northern European to pay. Since the early 2000s, health spending for countries.101 Exploration of the effects of specific health and wealthier Americans has soared, whereas expenditures social policies on health inequalities is a key area for future have risen modestly for the middle class and actually research. fallen for those with low incomes.89 Medical spending More recently, widespread dissatisfaction with the in the USA was historically highest among the poor, health-care financing system in the USA triggered who have the greatest medical need and whose care is passage of the Affordable Care Act. As discussed in detail subsidised by Medicaid. However, the income– in another paper102 in this Series, that legislation has spending gradient has now reversed—a phenomenon reduced gaps in insurance coverage, although it remains known as the Inverse Care Law.90 Although the health to be seen whether the Affordable Care Act will stop the implications of greater cost sharing are not fully growth in income-related health inequalities. The future understood, cost sharing might possibly have led the of the Affordable Care Act—and access to health poor to avoid needed services, which might have insurance for millions of low-income Americans—is contributed to the widening health gap between poor now in question with the transition in 2017 to a and middle-income Americans. Whether the rising Republican-controlled government that has promised to inequalities in medical spending will translate into dismantle the legislation. even more divergent health outcomes is unknown. Impoverishing medical expenses9 could also lead to Future areas of research worse health outcomes—a health-poverty trap driven In light of existing literature, we highlight several by medical cost. Price exposure to health inputs extends important areas for future research on this topic in beyond medical care, to include healthy food, good panel 2. childcare, safe neighbourhoods, and the wide range of social determinants of health that are—at least in part—priced in the market. Panel 2: Areas for future research Exposure to prices of health care and other health inputs • Further characterisation of the burden of disease associated with poverty and limited can be mitigated—and has been mitigated in the past— economic opportunity in the 21st century, including—but not limited to—epidemics through public subsidies. For instance, in the early of addiction and self-harm. 20th century, massive improvements in the health of low- • Assessment of trends in socioeconomic inequalities in morbidity. Most of the income Americans were achieved through control of literature has focused on mortality. Existing behavioural risk factors such as smoking, infectious disease via publicly funded centralised water drinking, and medication adherence do not fully capture the increased burden of 91 filtration and purification. Elites were willing to fund disease among the poor. Chronic pain and mental health issues are key expensive water infrastructure and other public goods to under-researched aspects of disease burden.34 avoid epidemics of typhoid and cholera that could spread to • Research into the implications of increased geographical segregation by 92 them. Although investments in the control of non- socioeconomic status, including the increased sorting of people by economic communicable diseases cannot be motivated by fear of opportunity and implications for health inequalities. 93 contagion, investments in the health of the poor might • Investigation of geographical predictors of health among the poor, with high levels of nevertheless be driven by political demands. On a global area-related heterogeneity shown in recent research.6 scale, demands for justice and equality led to the • Assessment of the role of early-childhood exposures in explaining widening health commitment to “health for all” at the 1978 International gradients. Although a growing body of literature points to childhood health as an Conference on Primary Health Care in Alma Ata, origin of adult disease, it is notable that mortality gaps are widening among ageing Kazakhstan. In the USA, widespread mobilisation for baby boomers born into a post-World War 2 society that was (at least for white social justice that coalesced during the Civil Rights Americans) more equal than current society. Nevertheless, a full account of childhood 94 Movement led to the introduction of Medicare and exposure could unearth alternative explanations for the divergence in adult health. 95 Medicaid (the health insurance programme for the poor), • Research into the mediating role of public institutions and social programmes and 96 racial integration of hospitals, public investments in their capacity to weaken the link between income, education, and health. 97 community health centres, early childhood investments • Evaluations of policies and programmes using credible causal designs (eg, randomised 98 such as Head Start, and other Great Society and War on trials and natural experiments) to understand policy options to reduce the strong Poverty programmes. This ensemble of reforms led to mid- association between poverty and poor health. century gains in survival among poor Americans, especially • Evaluations of interventions to alleviate poverty and reduce income inequality, with black Americans. Comparison across high-income particular attention to which types of interventions—eg, tax credits, job training, countries reveals evidence that social protection policies employment programmes—are the best for health. mediate the effect of macroeconomic fluctuations on • Evaluations of interventions to reduce the financial burden of ill health, including the health of the vulnerable—eg, reducing incidence of Affordable Care Act and other policies designed to increase access to health coverage, 99 suicides. Welfare-state expansions have been associated as well as health systems reforms that might reduce access to care. with life expectancy improvements and might explain, in www.thelancet.com Vol 389 April 8, 2017 1487 Series

Our review should be interpreted in light of its Although these explanations are necessarily speculative, limitations. We reviewed the extant literature on trends understanding the reasons behind rising health in­ in socioeconomic inequalities in health in the USA equalities is of crucial importance. Estimates of cohort over time. Although we have discussed some potential life expectancy project that the gap in life expectancy reasons for these changes, this literature is by nature between the richest and poorest quintiles will have descriptive and does not support inferences about the increased by nearly a decade within a single generation.7 effects of specific policy interventions. Interventions to Mitigation of these gaps might still be possible. Public prevent or mitigate the 21st-century burden of disease investments in the health of low-income Americans have associated with poverty, including epidemics of addiction led to large health improvements in the past. New and self-harm, are needed. Further research using robust investments in the health of poorer Americans might be causal designs (eg, randomised trials and natural necessary to avoid a 21st century health-poverty trap and experiments) is urgently needed to better understand to ensure that low-income Americans have an equal how we can halt and close the widening gap in life opportunity to lead long, healthy lives. chances between lower-income and upper-income Contributors Americans. We emphasise further the importance of All authors contributed equally to the writing of the manuscript. more finely grained data on morbidity, risk factors, and Declaration of interests causes of death to identify opportunities for intervention. We declare no competing interests. References Conclusion 1 Saez E. Striking it richer: the evolution of top incomes in the United States (updated with 2014 preliminary estimates). Berkeley, The rise in income inequality in the USA in 1980–2015 CA: University of California Berkeley, 2015. has coincided with widening inequalities in health and 2 Preston SH. The changing relation between mortality and level of longevity. Not only do the poor have lower incomes, they economic development. Popul Stud (NY) 1975; 29: 231–48. increasingly live shorter lives than do higher-income 3 Marmot MG, Stansfeld S, Patel C, et al. Health inequalities among British civil servants: the Whitehall II study. Lancet 1991; 337: 1387–93. Americans. 4 Krieger N, Fee E. Measuring social inequalities in health in the Growth in income inequality is most visible in the United States: a historical review, 1900–1950. Int J Heal Serv 1996; soaring incomes of top earners. By contrast, rising health 26: 391–418. inequalities have been observed across the full income 5 Deaton A. Health, inequality, and economic development. J Econ Lit 2003; 41: 113–58. distribution and, in recent years, chiefly among low- 6 Chetty R, Stepner M, Abraham S, et al. The association between income and middle-income Americans. Whereas the top income and life expectancy in the United States, 2001–2014. 1% of earners take home a larger share of total earnings JAMA 2016; 315: 1750–66. 7 National Academies of Sciences, Engineering, and Medicine. than at any time since the Great Depression, their health The growing gap in life expectancy by income: implications for outcomes still represent just 1% of the population. federal programs and policy responses. Washington, DC: National Changes in survival gaps during the period 1980–2015 Academies Press, 2015. 8 Bor J, Tanser F, Newell M-L, Bärnighausen T. In a study of a are not simply a reflection of rising life expectancy population cohort in South Africa, HIV patients on antiretrovirals among top earners. had nearly full recovery of employment. Health Aff 2012; Our review attributes rising survival inequalities to 31: 1459–69. 9 Himmelstein DU, Warren E, Thorne D, Woolhandler S. Illness and two factors. First, since 2001, real incomes have declined injury as contributors to bankruptcy. Health Aff (Millwood) 2005; for poor and middle-income Americans; these income (Suppl Web Exclusives): W5-63–73. losses, experienced by those who can least afford them, 10 Smith JP. Socioeconomic status and health. Am Econ Rev 1998; might very well have implications for health and 88: 192–96. 11 Almond D, Mazumder B. The 1918 influenza pandemic and longevity. Second, the association between income and subsequent health outcomes: an analysis of SIPP data. Am Econ Rev life expectancy has become stronger over time. Factors 2005; 95: 258–62. including technological innovation, increased geo­ 12 Case A, Fertig A, Paxson C. The lasting impact of childhood health graphical segregation, mass incarceration, reduced and circumstance. J Health Econ 2005; 24: 365–89. 13 Miguel E, Kremer M. Worms: identifying impacts on education and economic mobility, and increased exposure to prices health in the presence of treatment externalities. Econometrica 2004; might have contributed to a steepening gradient between 72: 159–217. income and health and to the increasing importance of 14 Bleakley H. Disease and development: evidence from hookworm eradication in the American south. Q J Econ 2007; 122: 73–117. poverty as a risk factor for poor health. Gaps have 15 Smith JP. The impact of childhood health on adult labor market widened in causes of death linked to smoking and other outcomes. Rev Econ Stat 2009; 91: 478–89. adverse coping behaviours, although the aggregate 16 Thompson O. Genetic mechanisms in the intergenerational trends have yet to be explained fully. Importantly, the transmission of health. J Health Econ 2014; 35: 132–46. 17 Lindahl M. Estimating the effect of income on health and mortality increased correlation between health and income might using lottery prizes as an exogenous source of variation in income. contribute to negative feedback loops, making mobility J Hum Resour 2005; 40: 144–168. along either axis more difficult. Interventions to decouple 18 Hoynes HW, Miller DL, Simon D. Income, the earned income tax credit, and infant health. Am Econ J Econ Policy 2015; 7: 172–211. income and health—in addition to interventions targeted 19 Evans WN, Garthwaite CL. Giving mom a break: the impact of solely to improving health and economic prospects— higher EITC payments on maternal health. Am Econ J Econ Policy might be needed. 2014; 6: 258–90.

1488 www.thelancet.com Vol 389 April 8, 2017 Series

20 Lynch JW, Smith GD, Kaplan G, House JS. Income inequality and 46 Masters RK, Link BG, Phelan JC. Trends in education gradients of mortality: importance to health of individual income, psychosocial ‘preventable’ mortality: a test of fundamental cause theory. environment, or material conditions. BMJ 2000; 320: 1200–04. Soc Sci Med 2015; 127: 19–28. 21 de Mello L, Tiongson ER. Income inequality and redistributive 47 Krueger PM, Tran MK, Hummer RA, Chang VW. government spending. Public Finance Rev 2006; 34: 282–305. Mortality attributable to low levels of education in the United States. 22 Kawachi I, Kennedy BP, Lochner K, Prothrow-Stith D. PLoS One 2015; 10: 1–13. Social capital, income inequality, and mortality. Am J Public Health 48 Montez JK, Zajacova A. Trends in mortality risk by education level 1997; 87: 1491–98. and cause of death among US white women from 1986 to 2006. 23 Wilkinson R, Pickett K. The Spirit Level. In: The Spirit Level: Am J Public Health 2013; 103: 473–79. Why more Equal Societies Almost Always do Better. London, UK: 49 Montez JK, Hummer RA, Hayward MD, Woo H, Rogers RG. Trends in Allen Lane, 2009: 215–28. the educational gradient of US adult mortality from 1986 through 2006 24 Lynch J, Smith GD, Harper S, et al. Is income inequality a by race, gender, and age group. Res Aging 2011; 33: 145–71. determinant of population health? Part 1. A systematic review. 50 Miech R, Pampel F, Kim J, Rogers RG. The enduring association Milbank Q 2004; 82: 5–99. between education and mortality: the role of widening and 25 Lynch J, Smith GD, Harper S, Hillemeier M. Is income inequality a narrowing disparities. Am Sociol Rev 2011; 76: 913–34. determinant of population health? Part 2. US National and Regional 51 Hayward MD, Hummer RA, Sasson I. Trends and group Trends in income inequality and age- and cause-specific mortality. differences in the association between educational attainment and Milbank Q 2004; 82: 355–400. US adult mortality: implications for understanding education’s 26 Cristia JP. Rising mortality and life expectancy differentials by causal influence.Soc Sci Med 2015; 127: 8–18. lifetime earnings in the United States. J Health Econ 2009; 52 Hadden W, Rockswold P. Increasing differential mortality by 28: 984–95. educational attainment in adults in the United States. 27 Waldron H. Trends in mortality differentials and life expectancy for Int J Health Serv 2008; 38: 47–61. male social security-covered workers, by socioeconomic status. 53 Gross T, Glied S, Muennig P. Widening mortality disparities by Soc Secur Bull 2007; 67: 1–28. educational attainment among native-born Americans adults over 28 Ho JY, Fenelon A. The contribution of smoking to educational 3 decades of follow up: the mystery deepens. Int J Healthc 2015; 1: 40–47. gradients in US life expectancy. J Health Soc Behav 2015; 54 Cutler DM, Lange F, Meara E, Richards-Shubik S, Ruhm CJ. Rising 56: 307–22. educational gradients in mortality: the role of behavioral risk 29 Meara ER, Richards S, Cutler DM. The gap gets bigger: changes in factors. J Health Econ 2011; 30: 1174–87. mortality and life expectancy, by education, 1981–2000. Health Aff 55 Kitagawa E, Hauser P. Differential mortality in the United States: 2008; 27: 350–60. a study in socioeconomic epidemiology. Cambridge, MA: Harvard 30 Hendi AS. Trends in US life expectancy gradients: the role of University Press, 1973. changing educational composition. Int J Epidemiol 2015; 44: 946–55. 56 Duleep HO. Measuring socioeconomic mortality differentials over 31 Olshansky SJ, Antonucci T, Berkman L, et al. Differences in life time. Demography 1989; 26: 345–51. expectancy due to race and educational differences are widening, 57 Pappas G, Queen S, Hadden W, Fisher G. The increasing disparity and many may not catch up. Health Aff 2012; 31: 1803–13. in mortality between socioeconomic groups in the United States, 32 Montez JK, Berkman LF. Trends in the educational gradient of 1960 and 1986. N Engl J Med 1993; 329: 103–09. mortality among US adults aged 45 to 84 years: bringing regional 58 De Neve J-W, Fink G, Subramanian S V, Moyo S, Bor J. Length of context into the explanation. Am J Public Health 2014; 104: 82–90. secondary schooling and risk of HIV infection in Botswana: evidence 33 Ma J, Xu J, Anderson RN, Jemal A. Widening educational from a natural experiment. Lancet Glob Health 2015; 3: e470–77. disparities in premature death rates in twenty six states in the 59 Lleras-Muney A. The relationship between education and adult United States, 1993–2007. PLoS One 2012; 7: 1–10. mortality in the United States. Rev Econ Stud 2005; 72: 189–221. 34 Case A, Deaton A. Rising morbidity and mortality in midlife among 60 Katz LF, Autor DH. Changes in the wage structure and earnings white non-Hispanic Americans in the 21st century. inequality. Handb Labor Econ 1999; 3: 1463–555. Proc Natl Acad Sci 2015; 112: 15078–83. 61 Walque D De. How does the impact of an HIV/AIDS information 35 Krieger N, Rehkopf DH, Chen JT, Waterman PD, Marcelli E, campaign vary with educational attainment? Evidence from rural Kennedy M. The fall and rise of US inequities in premature Uganda. J Dev Econ 2007; 84: 686–714. mortality: 1960–2002. PLoS Med 2008; 5: e46. 62 Escobedo LG, Peddicord JP. Smoking prevalence in US birth cohorts: 36 Singh GK, Siahpush M. Widening socioeconomic inequalities in the influence of gender and education.Am J Public Health 1996; US life expectancy, 1980–2000. Int J Epidemiol 2006; 35: 969–79. 86: 231–36. 37 Mackenbach JP. Should we aim to reduce relative or absolute 63 Goldman DP, Smith JP. Can patient self-management help explain inequalities in mortality? Eur J Public Health 2015; 25: 185. the SES health gradient? Proc Natl Acad Sci USA 2002; 99: 10929–34. 38 Waldron H. Mortality differentials by lifetime earnings decile: 64 Cutler DM, Lleras-Muney A. Understanding differences in health implications for evaluations of proposed Social Security Law behaviors by education. J Health Econ 2010; 29: 1–28. changes. Soc Secur Bull 2013; 73: 1–37. 65 Case A, Lubotsky D, Paxson C. Economic status and health in 39 Krieger N, Chen JT, Kosheleva A, Watermna PD. Not just smoking childhood: the origins of the gradient. Am Econ Rev 2002; 92: 1308–34. and high-tech medicine: socioeconomic inequities in US mortality 66 Currie J. Healthy, wealthy, and wise: socioeconomic status, poor rates, overall and by race/ethnicity, 1960–2006. Int J Health Serv health in childhood, and human capital development. J Econ Lit 2012; 42: 293–322. 2009; 47: 87–122. 40 Smith JP. Healthy bodies and thick wallets: the dual relation between 67 Link BG, Phelan J. Social conditions as fundamental causes of health and economic status. J Econ Perspect 1999; 13: 144–66. disease social conditions as fundamental causes of disease. 41 Sorlie PD, Johnson NJ. Validity of education information on the J Health Soc Behav 1995; (Extra issue): 80–94. death certificate.Epidemiology 1996; 7: 437–39. 68 Frederick CB, Snellman K, Putnam RD. Increasing socioeconomic 42 Dowd JB, Hamoudi A. Is life expectancy really falling for groups of disparities in adolescent obesity. Proc Natl Acad Sci USA 2014; low socio-economic status? Lagged selection bias and artefactual 111: 1338–42. trends in mortality. Int J Epidemiol 2014; 43: 983–88. 69 Montez JK, Zajacova A. Explaining the widening education gap in 43 Elo IT, Preston SH. Educational differentials in mortality: mortality among US white women. J Health Soc Behav 2013; United States, 1979–1985. Soc Sci Med 1996; 42: 47–57. 54: 165–81. 44 Everett BG, Rehkopf DH, Rogers RG. The nonlinear relationship 70 Stewart ST, Cutler DM, Rosen AB. Forecasting the effects of obesity between education and mortality: an examination of cohort, and smoking on US life expectancy. N Engl J Med 2009; race/ethnic, and gender differences.Popul Res Policy Rev 2013; 361: 2252–60. 32: 893–917. 71 St John PD, Tyas SL, Menec V, Tate R. Multimorbidity, disability, 45 Masters RK, Hummer RA, Powers DA. Educational differences in US and mortality in community-dwelling older adults. adult mortality a cohort perspective. Am Sociol Rev 2012; 77: 548–72. Can Fam Physician 2014; 60: e272–80.

www.thelancet.com Vol 389 April 8, 2017 1489 Series

72 Shippee ND, Shah ND, May CR, Mair FS, Montori VM. 87 Wildeman C, Wang EA. Mass incarceration, public health, and Cumulative complexity: a functional, patient-centered model of widening inequality in the USA. Lancet 2017; 389: 1464–74. patient complexity can improve research and practice. 88 Drucker EM. A plague of prisons: the epidemiology of mass J Clin Epidemiol 2012; 65: 1041–51. incarceration in America. New York, NY: The New Press, 2011. 73 Ewbank D, Preston SH. Personal health behavior and the decline in 89 Dickman SL, Woolhandler S, Bor J, McCormick D, Bor DH, infant and child mortality: the United States, 1900–1930. In: Himmelstein DU. Health spending for low-, middle-, and Caldwell JC, Findley S, Caldwell P, et al, eds. What We Know About high-income Americans, 1963–2012. Health Aff 2016; 35: 1189–96. Health Transition: The Culture, Social and Behavioural 90 Tudor Hart J. The inverse care law. Lancet 1971; 297: 405–12. Determinants of Health. Canberra: Health Transition Centre, The 91 Cutler D, Miller G. The role of public health improvements in Australian National University, 1990: 116–50. health advances: the twentieth-century United States. Demography 74 Bunker JP. The role of medical care in contributing to health 2005; 42: 1–22. improvements within societies. Int J Epidemiol 2001; 30: 1260–63. 92 Troesken W. The limits of Jim Crow: race and the provision of water 75 Taylor F, Huffman MD, Macedo AF, et al. Statins for the primary and sewerage services in American cities, 1880–1925. J Econ Hist prevention of cardiovascular disease. Cochrane Database Syst Rev 2002; 62: 734–72. 2013; 1: CD004816. 93 Costa D. Health and the economy in the United States, from 1750 to 76 Wright JM, Musini VM. First-line drugs for hypertension. the present. J Econ Lit 2015; 53: 503–70. Cochrane Database Syst Rev 2009; 3: CD001841. 94 Card D, Dobkin C, Maestas N. Does Medicare save lives? Q J Econ 77 Cutler D. Your Money or Your Life: Strong Medicine for America’s 2009; 124: 597–636. Health Care System. New York: Oxford University Press, 2004. 95 Goodman-Bacon A. Public insurance and mortality: evidence from 78 Goldman D, Smith JP. Socioeconomic differences in the adoption Medicaid implementation. J Polit Econ (in press). of new medical technologies. Am Econ Rev 2005; 95: 234–37. 96 Almond DV, Chay KY, Greenstone M. Civil rights, the war on 79 Goldman D, Lakdawalla D. A theory of health disparities and poverty, and black-white convergence in infant mortality in the rural medical technology. BE J Econ Anal Policy 2005; 4: 1–30. south and Mississippi. Dec 31, 2006. MIT Department of 80 Owens A, Reardon SF, Jencks C, Owens A. Income segregation Economics Working Paper No. 07-04. DOI:10.2139/ssrn.961021. between schools and school districts. Am Ed Res J 2016; 53: 1159–97. 97 Bailey MJ, Goodman-Bacon A. The war on poverty’s experiment in 81 Ezzati M, Friedman AB, Kulkarni SC, Murray CJL. The reversal of public medicine: community health centers and the mortality of fortunes: trends in county mortality and cross-county mortality older Americans. Am Econ Rev 2015; 105: 1067–104. disparities in the United States. PLoS Med 2008; 5: e119. 98 Ludwig J, Miller DL. Does Head Start improve children’s life 82 Chetty R, Hendren N, Kline P, Saez E, Turner N. Is the chances? Evidence from a regression discontinuity design. Q J Econ United States still a land of opportunity? Recent trends in 2007; 122: 159–208. intergenerational mobility. Am Econ Rev 2014; 104: 141–47. 99 Stuckler D, Basu S, Suhrcke M, Coutts A, McKee M. The public 83 Aaronson D, Mazumder B. Intergenerational economic mobility in health effect of economic crises and alternative policy responses in the United States, 1940 to 2000. J Hum Resour 2008; 43: 139–72. Europe: an empirical analysis. Lancet 2009; 374: 315–23. 84 Looney A, Manoli DS. Are there returns to experience at low-skill 100 Beckfield J, Bambra C. Shorter lives in stingier states: social policy jobs? Evidence from single mothers in the United States over shortcomings help explain the US mortality disadvantage. the 1990s. Upjon Institute working paper; 16-255. Kalamazoo, Soc Sci Med 2016; 171: 30–38. MI: Upjohn Institute for Employment Research, 2016. 101 Mackenbach JP. The persistence of health inequalities in modern 85 McEwen BS. Protective and damaging effects of stress mediators. welfare states: the explanation of a paradox. Soc Sci Med 2012; N Engl J Med 1998; 338: 171–79. 75: 761–69. 86 Halpern MT, Shikiar R, Rentz AM, Khan ZM. Impact of smoking 102 Gaffney A. McCormick D. The Affordable Care Act: implications for status on workplace absenteeism and productivity. Tob Control 2001; health-care equity. Lancet 2017; 389: 1442–52. 10: 233–38.

1490 www.thelancet.com Vol 389 April 8, 2017