The political polarization of health outcomes in the USA

MainA prominent literature in public health investigates ‘social determinants of health’—non-medical factors that shape health outcomes—such as race, ethnicity, income, education, public policy and environmental factors. This article argues that political beliefs, by shaping trust, engagement and adherence to medical advice, have become an important social determinant of health, even beyond coronavirus disease 2019 (COVID-19)-related matters. Liberals and conservatives differ more today in their health outcomes than in decades past, and these differences are poised to persist as attitudes towards seeking and trusting care—factors that affect health outcomes—now divide the political left and right.While previous work has investigated diverging health outcomes along political lines, much of this existing research relies on ecological data or self-reported health outcomes. These two approaches have led to different conclusions: ecological data that (for example) correlate county-level mortality rates and presidential vote show that people who live in ‘redder’ (that is, more Republican) counties have higher mortality rates1,2,3,4, perhaps due to differences in health policies between red and blue (that is, Democratic) places5,6,7,8,9. Conversely, a different set of studies rely on self-reported health outcomes to show that conservative people, for various psychological or political reasons, self-report better health10,11,12. Issues of aggregation or the misperception of one’s own health may well explain these contradictory findings, but their inconsistent directions prevent conclusions about the relationship between individuals’ politics and their health.So far, the clearest evidence on the relationship between health and political beliefs at the individual level comes from studies linking death records to survey results. These studies13,14,15, conducted using data from the end of the twentieth century and the first decade of the 2000s, find modest evidence that conservatives have higher all-cause mortality rates than liberals, but that Republicans have lower mortality rates than Democrats. However, these data leave important questions unanswered: they do not capture causes of death (for example, car crashes versus disease-based deaths), health during the lifetime or changes in health over time, making it difficult to understand the mechanisms behind political differences. What is more, the health and political landscapes have drastically changed in the last 15 years as political coalitions have shifted and health has become increasingly politicized.

New evidence on the relationship between politics and health is urgently needed.This article explores three questions. First, we seek to reconcile the existing mixed findings on the relationship between politics and health using longitudinal, individual-level data. Second, we ask how this relationship has changed since 2010, when previous mortality analyses conclude. Third, we seek potential explanations for these changes, identifying promising candidates for future research.To understand the relationship between politics and health in recent decades, we draw on individual-level medical data and death records from the National Longitudinal Study of Adolescent to Adult Health (referred to as the Add Health survey), which has tracked a nationally representative cohort of people who were adolescents in the 1990s (most born between 1976 and 1982) over the course of their lifetimes16. Unusually among health studies, Add Health includes a measure of political beliefs: self-reported liberal–conservative placement. It thus captures individual political orientation and medically validated health measures (for example, haemoglobin A1c (HbA1c) levels) and their changes over time, and individual-level cause-of-death data that extend into the COVID-19 era. Although Add Health’s only political measure is ideological self-placement (there is no usable measure of partisanship or vote choice), ideology is increasingly correlated with partisanship in recent years17.We find that conservative Americans in this cohort, who were about as healthy as liberals in the early 2010s, experienced worsening health through the 2010s and higher mortality in the early 2020s. Roughly half of this new health gap is due to people changing their ideology over time, with new entrants to the conservative coalition being less healthy than new liberals. But another sizeable share is due to people who were already liberal or conservative diverging more in health over time. Changes in the socio-economics of the liberal and conservative coalitions—including education, income and insurance status—contribute to both processes.By the 2020s, conservatives were dying at significantly higher rates than liberals, with the gap concentrated in internal causes (for example, heart disease, cancer and stroke).

The divide since 2020 is substantial: while only 0.2% of ‘very liberal’ respondents died of internal causes between 2020 and 2022, the probability for people who identified as ‘very conservative’ was 1.14 percentage points higher (P = 0.021; 95% confidence interval (CI), (0.18, 2.11)). This gap is not limited to deaths from COVID-19 and is not reducible to demographic or geographic differences between the groups, nor is it a pure function of ageing: previous cohorts’ death patterns in older data did not show a similar correlation between health and ideology before 2010.We suggest that these growing health gaps are consistent with a mechanism of politically rooted changes in engagement with the health system. Using a large public opinion survey, we find that people on the right, particularly Trump voters and Republicans, express less trust in their personal doctor and are less willing to seek care for non-COVID-19-related health problems18,19. We also find that people on the right with chronic illnesses are more sceptical than people on the left that medicines to treat those illnesses are safe and effective. This political divide in consumption of care may sustain or deepen the health divide that has emerged in recent decades. However, both these findings and those on health outcomes are purely descriptive; more work is needed to uncover causal relationships.Our findings raise serious concerns about the equity of health outcomes between Americans of different political backgrounds: conservatives are becoming a less healthy population, and their growing disinclination towards seeking and following medical advice means that these differences may be difficult to address. Although many institutions have lost the trust of Americans in recent decades, the case of medicine is a particularly stark illustration of the consequences that can follow when politics leads people to divest from institutions that promote their welfare.ResultsHealth dataTo explore the relationship between politics and health, we draw on the Add Health survey. Add Health is an ongoing survey tracking a nationally representative cohort in the USA, first interviewed while in grades 7–12 in 1994. So far, this cohort has been interviewed five times.Add Health interviews include survey questions on demographic and health attributes and measurements of ‘biomarkers’ collected in survey takers’ homes by trained interviewers.

Five biomarkers are measured in both waves 4 and 5, which are our focus here: body mass index, lipids (that is, cholesterol), HbA1c levels, blood pressure and C-reactive protein levels. Add Health has also checked respondents’ vital status against the National Death Index up to 2022 to determine whether respondents have died. Because of their age, mortality is a rare, although important, outcome in this sample: those interviewed in wave 5 were born between 1974 and 1983, with over 92% born between 1977 and 1982. By 2021–2022, the cohort was largely between 40 and 45 years old.In the third (2001–2002), fourth (2008–2009) and fifth (2016–2018) survey waves, Add Health also asked respondents whether they identified as very liberal, liberal, moderate, conservative or very conservative. Although Add Health does not have a usable measure of party identification or vote choice, it is worth emphasizing that its inclusion of any questions about political orientation makes it unusual among health surveys.Due to Add Health’s school-based sampling design, standard errors are clustered at the school level. All data are weighted to be representative of the cohort in the given wave.Ideology and health markers during the lifetimeFigure 1 shows the average number of comorbidities (out of five measured biomarkers; see Methods for details) for liberals and conservatives, as measured in waves 4 and 5. Higher numbers represent more total comorbidities and thus poorer health (Supplementary Section 1 presents results by individual measure). In 2008–2009 (wave 4), no discernible relationship exists between biomarkers (taken in 2008–2009) and ideological identification (as expressed in the 2008–2009 survey). An equivalence test presented in Supplementary Table 4 suggests that there is no difference greater than 0.04 between the most liberal group and any other.

Fig. 1: Biomarkers by ideological self-identification in waves 4 and wave 5.The alternative text for this image may have been generated using AI.Full size imageHealth status by ideological self-identification in wave 4 (2008–2009) and wave 5 (2016–2018) of the Add Health survey. Both ideology and health are measured in wave 4 and wave 5; each represents contemporaneous health and ideological measures. Black points represent the average and black vertical lines are the 95% CI. The thin horizontal grey lines represent quartiles. Density is underlaid. The comorbidity index is a combined measure of the five health indicators for respondents with each measure. P values, in parentheses, are the difference between the category and those identified as very liberal using OLS regression, with two-sided test for significance. Full statistical reporting is provided in Supplementary Section 1.2. P value is not applicable (NA) for very liberal respondents.However, by 2016–2018, the most conservative respondents (as measured in 2016–2018) were the least healthy (as measured in 2016–2018) (mean (M) = 0.305; 95% CI, (0.26, 0.36)). Table 1 shows the differences in the wave 4 and wave 5 cross-sections. The gap in health between the most liberal and conservative respondents in wave 5 represents one-third of the comorbidity scale’s standard deviation, making it statistically and substantively significant (0.08, 95% CI; (0.02, 0.14); P = 0.014). Supplementary Section 1.2 shows full tabular regression results.Table 1 Differences in biomarkers by ideological self-identification between wave 5 and wave 4Full size tableWhat explains conservatives becoming less healthy than liberals between 2008 and 2016?