Patient Trust and Personal Doctors

This paper focuses on trust in one’s personal doctor. What do we mean when we say that a patient trusts a doctor or that a person trusts medicine or the medical industry? A popular sociological definition of trust (Cook, Hardin and Levi Reference Cook, Hardin and Levi2005) holds that one person should trust another only when the former’s interests are ‘encapsulated’ by the latter; that is, that one’s interests are the same as the other’s. While a strict definition, it illustrates an important, more general, point: people trust one another when they believe, even in moments of vulnerability, that their interests will be considered, even prioritized. Most definitions of trust ‘stress the optimistic acceptance of a vulnerable situation in which the truster believes the trustee will care for the truster’s interests’ (Hall et al. Reference Hall, Dugan, Zheng and Mishra2001; emphasis in original). When a patient visits a doctor – or considers doing so – to say they trust the doctor means they believe the doctor will prioritize their health and wellbeing, perhaps even their financial wellbeing. While Cook, Hardin, and Levi’s (Reference Cook, Hardin and Levi2005) definition of trust does not allow for any analogue toward organizations or institutions, other trust scholars (for example, Zucker Reference Zucker1986; Shapiro Reference Shapiro1987) acknowledge the possibility of trust in collections of people, and the implications for the individual are similar: they must believe the organization or institution will prioritize their interests.

Trust in most major American societal institutions has polarized along partisan lines in recent years, and there is support for the idea that this happens as partisans come to the conclusion that these institutions are comprised of out-partisans (Brady and Kent Reference Brady and Kent2022). For example, as Republicans come to believe the press is made up of Democrats, their trust in the institution of the press decreases. Medicine has, for most of the post-1970 period, actually bucked this trend; as of 2019, opinion of medicine (as an institution) was substantially identical across partisan lines (Brady and Kent Reference Brady and Kent2022). Since the beginning of the COVID-19 pandemic, though, opinion toward medicine seems to have joined most other institutions and is now partisan (see further discussion, below).

To the extent that partisans’ belief that institutions are made up of out-partisans determines (dis)trust of those institutions, one possible reason for this polarization of trust in institutions is the actual ideological composition of institutions. Kent (Reference Kent2022) finds that changes in average ideology among constituents of institutions since 1980 do in fact correlate with changes in polarization in trust in institutions over the same period. For example, as the average member of the press has become more liberal, Republicans (and conservatives) have become less trusting of the press. This suggests that Cook, Hardin and Levi’s (Reference Cook, Hardin and Levi2005) definition of trust may in fact be the right way to think about institutional trust: as people who lack shared characteristics become more and more representative of an institution, or at least perceive that to be the case, people trust that institution less.

Doctors and others in medical professions (for example, nurses) have, since 1980 (as measured by campaign contributions), on average shifted to the left ideologically with respect to other institutions (Kent Reference Kent2022), and political ideology differs across medical specialities (Bonica, Rosenthal and Rothman Reference Bonica, Rosenthal and Rothman2014). In 1990, about 60 per cent of campaign donations made by physicians went to Republicans; in 2018, nearly two-thirds of donations made by physicians went to Democrats. Gallup survey data parallel this trend: between 2011 and 2016, the proportion of physicians surveyed who identified as Democratic increased by 7 percentage points and the percentage who identified as Republican decreased by 3 points (Adamy and Overberg Reference Adamy and Overberg2019).

Though not polarizing over most of the post-1970 period, trust in doctors has declined over time (Lipset and Schneider Reference Lipset and Schneider1983; Davies Reference Davies1999; Blendon, Benson and Hero Reference Blendon, Benson and Hero2014). Patients’ trust in doctors is important not just on a societal scale, but seems to affect medical outcomes: higher trust is associated with better self-care among diabetics (Bonds et al. Reference Bonds, Camacho, Bell, Duren-Winfield, Anderson and Goff2004; Lee and Lin Reference Lee and Lin2011; Mancuso Reference Mancuso2010), completion of colorectal screening (Gupta et al. Reference Gupta, Brenner, Ratanawongsa and Inadomi2014), and following doctors’ recommendations to control high blood pressure (Jones et al. Reference Jones, Carson, Bliech and Cooper2012). Furthermore, higher trust in doctors and the health care system in general is positively associated with higher self-evaluations of good health, itself an important correlate of health outcomes (Nummela et al. Reference Nummela, Sulander, Rahkonen and Uutela2009).

All of this suggests an uncomfortable reality: if patients come to believe that the medical industry, doctors in general, or their personal physician(s) are different from them politically, health outcomes may decline. Conversely, if someone perceives their doctor to share their political background, it could improve trust and health outcomes.

Over-Time Trends

We investigate over-time trends towards three elements of trust. First, we ask how much confidence in medicine people have generally. Respondents answered on a three-point scale saying they have ‘a great deal of confidence’, ‘some confidence’, or ‘hardly any confidence’ (full questions and wording are included in Supplementary Appendix 10). Second, we analyze trust in one’s personal doctor. To our knowledge, academics have not studied trust in one’s personal doctor as a function of partisanship. Respondents are asked to what extent they trust their personal doctor on a four-point scale: a) a great deal; b) somewhat; c) not too much; and d) not at all. Finally, we track over time trends in people’s adherence to their doctor’s advice. If people’s trust in their personal doctor is polarizing along partisan lines, people’s adherence to their doctor’s orders might also be polarizing along partisan lines. We asked respondents, ‘how closely would you say you follow your doctor’s advice and treatment recommendations…?’ Respondents answered this question on a five-point scale from ‘extremely closely’ to ‘not at all closely’.

We expected that the COVID-19 pandemic would shape partisan evaluations of doctors, so we analyzed trust and adherence both before and after the coronavirus pandemic. We specifically chose the above questions because publicly available data asked these questions in the years prior to COVID, and those surveys included questions related to partisanship. For our post-COVID-19 survey, we commissioned a poll through the NORC-Amerispeak panel, a random sample of people 18 years of age and older living in the USA (SA 9 has NORC respondent demographics). Our 2022 survey mimicked, as closely as possible, the questions asked in the prior years (for details on the prior surveys that we used as a benchmark, see SA 11).

For trust in one’s personal doctor, we compared our 2022 data to a 2013 survey from AARP (conducted by Social Science Research Solutions; AARP 2013) that, like the NORC survey, included a random sample of people 18 years of age and older living in the USA. We chose this survey because it was the most recent survey that asked about trust in one’s doctor and also asked about people’s political predispositions. Finding a survey to benchmark adherence to medical doctors’ advice proved difficult, but a 2011 AARP survey asked about adherence to a doctor’s advice, though only to people living in the USA aged 50 and over (AARP 2011). While this randomly sampled survey is nationally representative of people 50 years and older, unlike the prior question, it does not represent all adults. Given the age categorization, we compare the 2011 results with those aged 50 and over in our 2022 NORC sample (only for the adherence question do we subset to 50+; the other questions include all respondents). The third question, which asks about confidence in medicine, was benchmarked against survey questions fielded as part of the 2019 Cooperative Election Survey conducted by YouGov. Although not a random sample, YouGov is well respected among social scientists and is nationally representative.

Figure 1 shows the change over time for each of these measures by partisanship. Because the 2011 and 2013 surveys do not ask about which way Independents lean, we categorize partisanship by people who, when asked the first time, say they are either Democrats or Republicans.

Figure 1.

Trust over time, by partisanship.

Note. The graph shows the percentage of people who identify as Democrat or Republican (excluding learners as they are not available on initial surveys) and their trust/adherence towards facets of the medical system. The left-hand panel shows the percentage of respondents who say they have ‘a great deal of confidence’ in medicine. The middle panel is the percentage of people who say they have a ‘great deal’ of trust in their doctor; the right panel are the percentage of respondents (50 and older) who say they follow their doctor’s advice ‘extremely closely’ or ‘very closely’. Lines represent 95 per cent confidence intervals (partisan differences in confidence in medicine and trust in one’s personal doctor are statistically different from 0 at the .05 level in 2022, but not in the initial time frame). See SA 1 for related statistical tests.

The first panel of Figure 1 shows the proportion of respondents who have a ‘great deal of confidence’ in medicine as a whole. In 2019, no partisan divide existed but by 2022, that divide expanded to nearly 26 percentage points (p < .05). As has been documented by others, COVID-19 drove this divide: the medical community’s response became attached to a partisan valence, and trust in the institution became partisan (Blendon and Benson Reference Blendon and Benson2022).Footnote ⁶

However, one’s personal doctor is removed from the broader institution and, perhaps as such, has not received the same attention as partisan trust in medicine as an institution. However, trust in one’s personal doctor has also become partisan. The second panel of Figure 1 shows the percentage of respondents who say they have a ‘great deal’ of trust in their personal doctor. While Republicans were slightly more trusting of their personal doctor in 2013, a partisan divide emerged by 2022 such that Democrats were 12 percentage points more likely than Republicans to say they trusted in their personal doctor a ‘great deal’ (p < .05).Footnote ⁷

Not only does partisanship predict trust in one’s own doctor in 2022, it is robust to controlling for a host of covariates that might be correlated to trusting one’s doctor (see SA 2). Partisanship is not simply a proxy for other baseline factors that influence trust in one’s doctor: among people who have a doctor or place they regularly go when they are sick or need advice about their health, with the exception of age, partisanship is the strongest predictor of trust in one’s doctor in our 2022 NORC sample.Footnote ⁸ Partisanship is more predictive of trust in one’s doctor than income, race, insurance status and education.

The third panel shows the percentage of respondents who say they follow their doctor’s advice ‘extremely closely’ or ‘very closely’. Because the 2011 sample included people only 50 years old and older, the 2022 respondents (in the third graph only) are a subset of that age range as well. Here, the partisan divide is smaller but still represents an appreciable shift. In 2011, Republicans 50 and over were about 4 percentage points more likely than Democrats to indicate high adherence to their doctor’s advice, while in 2022, Democrats were about 4 percentage points more likely than Republicans to exhibit high adherence levels (although the difference in 2022 is not statistically different from 0). While the divide is modest, it represents an eight percentage point swing in eleven years. Furthermore, the divide widens considerably when analyzing by vote choice: 81 per cent of Biden voters (50 and over) say they followed their doctor’s advice extremely or very closely compared to 70 per cent of Trump voters (the 2011 survey does not ask about vote choice).

What Explains the Partisan Divide in Trust Towards Personal Doctors?

What might explain the development of a partisan divide between Republicans and Democrats’ feelings towards personal doctors? One possible explanation is compositional changes among the parties’ supporters. The two parties changed in the 2010s in a way that meaningfully correlates with trust in doctors. Notably, people without a college degree moved toward the Republican Party (Barber and Pope Reference Barber and Popeforthcoming); a demographic that is less trusting of institutions generally and less trusting of doctors specifically. This may account for part of the emerging correlation between partisanship and trust. SA1 presents analysis showing that this is partially the case: Republicans and Democrats without a college degree (both white and non-white) expressed decreased levels of trust between our 2013 and 2022 samples, but the decline occurred equally among both parties. However, Democrats with a college degree became relatively more trusting, and Republicans with a college degree became relatively less trusting. Analysis in SA1 shows that controlling for education modestly attenuates partisanship’s predictive power between 2013 and 2022.

Table 1.

Conjoint table

Note. Respondents were asked to choose between Doctor A and B by attributes listed in the left-hand column. Respondents were shown profiles in which one characteristic per attribute was randomly shown (for example, randomly told the doctor was a Democrat or Republican).

Alternatively, the emerging partisan gap could be explained by growing generalized distrust among Republicans compared to Democrats that extends to include doctors but is not unique to them. Scholars (for example, Brady and Kent Reference Brady and Kent2020; Boon et al. Reference Boon, Salleras, Menchen-Trevino and Wojcieszak2020) document growing Republican distrust in institutions that have been associated with the intellectual establishment including the media, science, academia and government itself. Opposition to ‘experts’ and intellectual institutions has simmered on the political right for decades,Footnote ⁹ and became amplified in the years preceding COVID-19 (Nature Editorial 2010). Conversely, and even prior to the COVID, the political left embraced intellectual establishments. For example, then-President Barack Obama committed in his 2009 inaugural address to, ‘restore science to its rightful place’ (qtd. in Nature 2010). This could mean that Democratic trust in doctors represents shifts in a broader trend that includes other intellectual establishments and their members.

Both the 2013 AARP survey and our 2022 NORC survey (used above) asked about respondents’ trust in their members of Congress, judges, and neighbours using the same wording and scale that we ask about trust in their personal doctor. Table 1 in SA 1 shows Republicans and Democrats have diverged in trust across each of these people. This aligns with the theory that the partisan trust gap includes but is not exclusive to, trust in doctors.Footnote ¹⁰

However, the gap between these surveys – 2013 to 2022 – does not tell us much about the timing of these shifts. Survey companies rarely ask about trust in one’s personal doctor and political measures so it is difficult to know, due to a lack of data, exactly when trust in personal doctors transformed from non-partisan to partisan between 2013 and 2022.Footnote ¹¹

However, the General Social Survey (GSS), a survey fielded every two years, has long asked how much confidence people have in medicine and other institutions. Figure 2 tracks confidence in medicine (similar to the question in Figure 1) along with confidence in the ‘scientific community’ and education – two fields closely related to medicine in that they can be considered ‘intellectual’ institutions. A partisan gap in confidence in the scientific community and education had developed by 2010, with Democrats expressing more confidence. This aligns with others’ work on a growing divide in intellectual institutions over this time (Brady and Kent Reference Brady and Kent2020; Boon et al. Reference Boon, Salleras, Menchen-Trevino and Wojcieszak2020). In this same period, however, medicine remained largely bipartisan although Republicans in the 1990s were slightly more confident in medicine than Democrats and that gap had substantively (and statistically) closed by 2000. Not until the 2021 GSS do Democrats, statistically (p < .05) or substantively, express more confidence in medicine than Republicans.Footnote ¹²

Figure 2.

Confidence in intellectual institutions.

Note. The graph tracks confidence in institutions as measured by the General Social Survey (on a 1–3 scale, higher levels reflect more confidence). Red lines represent average trust among Republicans and blue lines among Democrats. Lines represent 95 per cent confidence intervals. The GSS did not run in 2020. SA 3 shows the difference between Democrats and Republicans with 95 per cent confidence intervals. Between 2000 and 2019, Democrats and Republicans held statistically similar attitudes towards medicine. This was unique among 9 different non-government related institutions during this time period (for example, trust in the press was partisan).

Indeed, confidence in medicine is noteworthy in that it is the only non-governmental institution that remained non-partisan throughout the 2010s.Footnote ¹³ Supplemental Appendix 3 tracks confidence in nine non-governmental institutions asked by the General Social Survey (GSS; Smith et al. Reference Smith, Davern, Freese and Morgan2019), such as organized religion, labour, the military, and the press; each of these institutions had large gaps in confidence between Democrats and Republicans by at least 2010. Medicine remained bipartisan until 2021.Footnote ¹⁴ If trust in institutions and their constituents – including medicine and doctors – could be explained fully by changes in the composition of the parties or secular trends in trust, the divergence between Republicans and Democrats would have emerged in all institutions at approximately the same time; in fact, trust in medicine remained stubbornly non-partisan until just before the COVID-19 pandemic. Thus, while a partisan gap in trust is not unique to medicine, its timing – and prior resistance to partisan trends – are heavily suggestive of the COVID-19 crisis for creating this partisan divide.

Consistent with this finding, the remainder of this section focuses on a third explanation that we highlighted above: as messaging about COVID-19 precautions became increasingly tied to Democrats and liberals, a spillover occurred into areas of health other than COVID. This led to an erosion of trust among Republicans (outside of COVID-related measures). Our expectations are symmetric: we expect that if people perceive doctors to be aligned with the Democratic party, that this will also increase trust and adherence among Democrats. This expectation aligns with the over time trends that COVID-19 did not simply exacerbate an existing divide but created a partisan schism in the first place.

We test this hypothesis using a survey experiment conducted on the Prolific survey platform of people over 18 years of age living in the USA. Prolific is an opt-in convenience sample, so the cross-sectional results do not represent the population at large (see SA 9 for descriptive statistics), but convenience samples are useful for analyzing differences between experimental treatment groups. Because Prolific is heavily Democratic, and Republicans are important to the intervention we are studying, we over-recruited people who reported voting for Donald Trump in the 2020 election. Of the 1,204 survey participants, 1,150 passed a basic attention check and were included in the study (95.5 per cent passage rate, which is a good return for online samples (see Aronow et al. Reference Aronow, Kalla, Orr and Ternovski2020).Footnote ¹⁵

In this experiment, we primed respondents to think about COVID-19 through a partisan lens, as was common during the pandemic. To do this, we showed respondents in the treatment group a headline that was typical of partisan charges during the COVID-19 crisis: that Dr Anthony Fauci, the head of the National Institute of Allergy and Infectious Diseases and the medical community’s public face during the COVID-19 crisis, was a Democrat. The headline, taken from The New York Post reads: ‘Trump bashes Anthony Fauci as ‘a Democrat’ and ‘Cuomo’s friend’.Footnote ¹⁶ We believe this experiment mimics messaging that was repeated many times over the first few years of the COVID-19 pandemic.

Our outcomes of interest are three questions that closely parallel the questions above: trust in one’s personal doctor (answered on a 0–10 sliding scale), whether someone adheres to their doctor’s advice (answered on a 0–10 sliding scale), and whether a person has confidence in the medical system (4-point scale).Footnote ¹⁷

To get a baseline and increase the precision of our estimates, we asked respondents for each of these outcome variables, in addition to a range of demographic variables, in our first survey wave. Two weeks later we reinterviewed the respondents. At this point, half of the respondents that returned were randomly assigned to the treatment group and were shown the headline that Fauci was a Democrat while half were assigned to a pure control group (and saw no prime). Of the 1,150 respondents in our Wave 1 survey, 1,040 returned to Wave 2 (90 per cent return rate; attrition is not correlated with Wave 1 characteristics or outcome variables, see SA 4.)

We expected that Trump voters exposed to the treatment would express lower trust in their personal doctor, a decreased willingness to follow their doctor’s advice, and less confidence in the medical system as a whole than those in the control condition. We expected exposure to the prime would increase trust, adherence, and confidence among Biden voters.

The results, presented in Figure 3, show this to be generally the case. The blue and red points in Figure 3 are the differences-in-means, presented separately among Biden voters (blue dots; n = 542) and Trump voters (red points; n = 287), controlling for Wave 1 answers to the dependent variable. The estimates can be represented by the following equation:

$${Y_{t = 1}} = {\alpha _1} + {\beta _1} \cdot Treat + {\beta _2} \cdot {Y_{t = 0}} + \varepsilon $$

Figure 3.

Experimental prime: Fauci is a democrat.

Note. Each point is the difference-in-means between respondents who were primed by Trump’s charge that Dr. Anthony Fauci was a Democrat and those who were not, controlling for wave 1 measure of the DV. Because we expected heterogeneous treatment effects by political predispositions, we analyzed Biden (blue points; n = 542) and Trump (red points; n = 287) voters separately. Positive (negative) values mean people exposed to the treatment were more (less) likely to express confidence/adherence/trust than those in the control group. Grey points represent the difference between the treatment effect among Biden voters and Trump voters. Lines represent 95 per cent confidence intervals.

Y _t=1 represents the outcome of interest (adherence, trust in the doctor, confidence in the medical system) after treatment, while Y _t=0 represents the outcome of interest before treatment. By controlling for the baseline (Y _t=0 in the above equation), we are controlling for regression to the mean between survey waves. Our estimate of interest is β ₁ which measures the treatment effect for the given outcome measure in Wave 2, controlling for the Wave 1 measure of the dependent variable.Footnote ¹⁸ The grey points are the difference between the red and blue points, which can be thought of as the polarization between Trump and Biden voters in reaction to the prime.

By way of explanation consider the left-most blue point in Figure 3, which is positive. This point is positive because Biden voters in the treatment group expressed more trusting attitudes after the Fauci prime when compared to the control group, controlling for Wave 1 attitudes. The first red point is negative because Trump voters who received the Fauci prime became less trusting than those in the control group (who received no prime). The grey point represents the difference between the blue and red points (differences in treatment effects).

The Fauci prime polarized both trust in one’s own doctor and confidence in medicine (p < .05). It did not change reported adherence among Biden voters, but it did decrease reported adherence to doctor’s advice among Trump voters to modest, although not quite statistically significant levels. This may be explained by a ceiling effect among Biden voters if they already complied at a high rate.

The experimental results presented above support the hypothesis that partisan messaging primed people to think of medicine and doctors as partisan and Democratic.Footnote ¹⁹ This messaging impacted both Democrats and Republicans. Does this shift have any implications for how partisans interact with doctors? In the next section we present evidence that if Americans are given partisan information about doctors, they will use it to decide which doctor to visit.

Consequences of Viewing Doctors as Partisan

The literature on trust in the medical field focuses on patients and their doctors sharing identities as a means to increase trust in one’s personal doctor. These studies, as this paper’s introduction outlines, largely focus on sharing a racial or gender identity. Black people are more trusting of black doctors and women are more trusting of doctors who are women. We wondered if partisanship might have a similar effect. The increasing politicization of medicine, whether due to COVID-19, health care reform, or abortion policy means that people have more opportunities to evaluate their doctor through a partisan lens. Patients may ask their doctor’s political views or come to conclusions about their political allegiances given their public-facing activities, social media, or markers that suggest a political allegiance. Does sharing (or not sharing) a partisan identification influence people’s trust in their personal physician?

Study 1: Conjoint Experiment

To investigate this possibility, we first conducted a conjoint experiment in which respondents selected one of two dermatologists they would choose to visit given a host of covariates of a hypothetical doctor. This experiment (and the next one) assumes that to engage with a healthcare provider, one must trust them in the first place. (This is a related, although distinct concept of trusting a doctor who already provides you care.)

If made plainly clear to patients, does a doctor’s politics affect peoples’ decisions to seek care even after taking other factors into account? We ran this experiment at the end of the first wave of our experiment described in the previous section. The pool of respondents was made up of people living in the USA who are 18 years old or older.Footnote ²⁰

We asked respondents, ‘Suppose you need to schedule a dermatology appointment and are choosing which dermatologist to go to. Which dermatologist are you more likely to schedule an appointment and visit? Doctor A or Doctor B?’ We then presented respondents with a table of randomly assigned attributes across each of the categories detailed in Table 1. Each respondent was asked to do this twice. The order in which respondents saw the attributes varied across respondents, but within respondents the attribute order stayed constant.

We included attributes that we believed were important factors, and often readily available, to people when searching for healthcare providers such as patient ratings, distance to the office, and, because photos are often included in doctor bios, perceived race and gender. In the context of this experiment, and of primary interest, we also labeled the doctor as a Democrat or Republican. While a blunt treatment, it enables the observation of a treatment effect if it exists. Furthermore, as the next section explores, this information is sometimes available through databases for doctors who hold certain political views.

Thus, each respondent saw a table for Doctor A and Doctor B, with attributes randomly assigned. For example, Doctor A was randomly assigned to have a 3.9 star rating, ten minutes away, attended an Ivy League (and so on) while Doctor B had a 4.1 star rating, was thirty-five minutes away, and attended a state school (and so on). Because we randomly assigned attributes, it is possible both Doctor A and Doctor B shared an attribute (for example, both are male). Each respondent was shown two tables for a total of 2,299 choices (1,150 respondents × two tables per respondent; one respondent ranked only one of the tables).

To analyze the experiment, we subset the sample between Democrats (represented by blue dots) and Republicans (represented by red dots) and estimate the average marginal component effect (AMCE) using the following equation (standard errors are clustered at the individual level):

$$\begin{gathered}Docto{r_{A|B}} = {\beta _0}\cdot HighRating + {\beta _1}\cdot MedRating + {\beta _2}\cdot FarAway + {\beta _3}\cdot IvyLeague + {\beta _4}\cdot Black \\ + {\beta _5}\cdot Hispanic + {\beta _6}\cdot Male + {\beta _7}\cdot Democrat + {\alpha _0} + \varepsilon \\ \end{gathered}$$

The AMCE for all respondents in a given party, estimated in the top-left panel of Figure 4, shows the marginal effect of a doctor with a given attribute against the baseline (omitted group in regression). For example, coefficients next to ‘High Rating’ and ‘Medium Rating’ show the marginal effect of doctors with either of these ratings compared to those with a low rating (which is the omitted group), averaged over all other combinations of attributes. As one might expect, doctors with a higher rating are preferred to a lower-rated doctor. Medium-rated doctors are also preferred over lower-rated doctors, but less so than highly-rated doctors.

Figure 4.

Conjoint Experiment: Which Doctor Would You Choose?

Note: Each point is the marginal effect of choosing a doctor with the given attribute relative to the omitted category (averaged over all other combinations of attributes). Blue dots are Democratic respondents, red dots are Republican respondents. Lines represent 95 per cent confidence intervals. For example, the coefficients next to ‘High Rating’ and ‘Medium Rating’, show the marginal effect of doctors with either of these ratings compared to those with a low rating (the omitted category). As one might expect, doctors with a higher rating are preferred to a low-rated doctor as are medium-rated doctors to low-rated doctors, but slightly less so. Of interest in this paper is that Democratic respondents prefer a doctor who is Democrat (positive coefficient on ‘Democrat’ indicator); Republican respondents prefer a doctor who is Republican (negative coefficient on ‘Democrat’ indicator). The panels include all respondents (top-left); women respondents only (top-right); black respondents only (bottom-left); and Latinx respondents only (bottom-right). The omitted category for race is `Whiteʼ; for gender is `Femaleʼ; for distance is `Nearbyʼ and for school type is `State school.ʼ

Of interest here is the coefficient on Democrats among Democratic (blue dots) and Republican (red dots) voters. As expected, Democrats are more likely to say they would choose the Democratic doctor compared to the Republican doctor. The reverse holds true for Republicans. As in the priming experiment, both Democrats and Republicans move in response to the doctor’s partisanship. The difference between Democrats’ and Republicans’ likelihood of choosing the Democratic compared to Republican doctor is 28 percentage points for all respondents (that is the difference between the red and blue points on the variable ‘Democrat’ in the top-left panel), 32 percentage points for female respondents and 29 percentage points for Hispanic respondents (each statistically significant at the 5 per cent threshold).

To benchmark the importance of partisanship, consider the relationship between co-partisanship against sharing a gender or race. Existing literature predicts that patients who share a racial or gender identity with their doctor have increased levels of trust. How does sharing a partisan background compare? The remaining three panels subset respondents by gender and race. Among Democratic women, Black Democrats, and Hispanic Democrats, sharing a partisan label is at least as important as sharing a gender or race. The same is true for Latino Republicans, although the coefficient is not statistically significant (the sample lacks enough Black Republicans to present an analysis). These results are striking; the literature points to the importance of shared identity for increasing trust among women and under-represented minorities. Our conjoint experiment hints that partisanship potentially rivals these markers.Footnote ²¹

Study 2: Trust in Healthcare Providers

The previous section finds that at least in a hypothetical setting, politics can shape which doctor people choose to visit. But do people actually care about their healthcare provider’s political views? Anecdotal accounts suggest that for some, the answer is ‘yes’. To return to the example of Bob Odenkirk’s heart doctor (discussed in the introduction), Odenkirk dismissed the doctor’s advice to take cholesterol-lowering drugs, in part because of the doctor’s conservative political leanings. For Odenkirk, his doctor’s politics affected trust.Footnote ²²

This might not be an isolated example. There are websites that direct (prospective) patients to physicians that align with their political belief system. There are directories for pro-life OBGYNs (https://aaplog.org/) and LGBTQ+ friendly doctors (https://lgbtqhealthcaredirectory.org/). A national directory of conservative professionals, titled simply ‘Conservative Professionals’ (with a URL ‘conservativeprofessionals.com’), seeks to connect conservative people with a ‘therapist, physician, life coach, attorney, or other professional who shares the same beliefs.’ Their website’s homepage (as of 9 December 2023) cites building trust as central to their mission: ‘… it may be difficult to trust that some Liberal professionals will listen to their Conservative clients with an open mind and heart’.

We want to know whether people’s political predispositions shape their willingness to seek healthcare when the source of that healthcare is attached to a political label. The very existence of ‘conservativeprofessionals.com’ suggests that people do in fact care about the politics of their health provider. (Note that although this example focuses on conservatives, we expect both the left and right to be sensitive to political labels.)

To investigate this, we recruited a new sample via Prolific of people living in the USA aged 18 years and older from November to December 2023. A total of 777 respondents passed a simple attention check at the start of the survey (92 per cent passage rate). (Sample demographics are in SA 9 and question wording is in SA 10). We randomly told respondents either about conservativeprofessionals.com or zocdoc.com.Footnote ²³ As discussed above, conservativeprofessionals.com is an online directory that connects patients to conservative healthcare providers and other professionals. Conversely, zocdoc.com is a generic, popular website that connects people to healthcare providers without any explicit mention of politics. Respondents were shown one of the two descriptions below.Footnote ²⁴

There are growing ways to seek healthcare, including online navigators that connect patients to healthcare providers. One such navigator is [‘Zocdoc.com’/ ‘conservativeprofessionals.com’], which connects potential patients to physicians and other health professionals [blank/who hold conservative values].

How likely are you to seek healthcare from websites like [zocdoc.com/ conservativeprofessionals.com]?

Thus, these two groups are given identical surveys, with the exception of being provided information about zocdoc.com (henceforth ZD) rather than conservativeprofessionals.com (henceforth CP). In addition to asking respondents how likely they are to seek healthcare from the given website, we also asked respondents whether they would be willing to share their Prolific email to learn more information about CP/ZD. (Prolific prohibits asking respondents for their actual email, but does allow asking for the email generated for each user on the website.)

For this experiment, we shifted from partisanship to ideology because the actual website is titled ‘conservative professionals’, and we wanted to test concordant political beliefs. Despite the differences between ideology and partisanship, they have become increasingly intertwined in contemporary politics (for example, Fiorina and Abrams Reference Fiorina and Abrams2008) and they both serve as affective political identities (Conover and Feldman Reference Conover and Feldman1981). We expect that the mechanisms that increase trust and engagement in an ideologically concordant doctor would increase trust in a partisan concordant doctor. Furthermore, like Democrats and Republicans, trust in doctors among liberals and conservatives has moved in different directions: in 2013, conservatives were about 7 percentage points more likely to express high trust in their personal doctor, but in 2022, liberals were about 5 percentage points more likely to express trust in their doctor.

Given the ideological divide that has emerged with respect to healthcare, our first hypothesis is observational: we expected that conservatives would be less likely to say they would seek care from ZocDoc.com or be willing to provide an email address to learn more information. Our primary experimental hypotheses expected that conservatives (liberals) assigned to the CP treatment would be more (less) likely to express they would seek care and provide their email address than conservatives (liberals) assigned to the ZD treatment.

Furthermore, we hypothesized that telling people about conservativeprofessionals.com would itself provide new information to respondents: that there are conservative health professionals. We expected that conservatives exposed to the CP treatment would be less likely to perceive healthcare providers to be liberal and that conservative patients would express higher levels of trust in health professionals generally.

Tables 2 and 3 present the experimental results by ideological self-placement. Our first hypothesis, that conservatives would be less likely to engage with ZocDoc.com relative to liberals, is evident in column 2. The most liberal respondents express the highest interest in seeking care from Zocdoc with a fairly consistent drop off as respondents become more conservative. Extremely conservative respondents are nearly a full-scale point less likely to say they would be interested in seeking care from ZocDoc.com and about half as likely than extremely liberal people to provide an email to learn more about ZocDoc. Here, we believe that general distrust in the medical system, or perhaps past mistrust of personal doctors, dampens prospective care.Footnote ²⁵

Table 2.

Results: DV = Seek healthcare from website

Note. Results are presented for the respondent’s answer to the question: ‘How likely are you to seek healthcare from websites like zocdoc.com/conservativeprofessionals.com’. The scale ranges from 0 ‘Extremely Unlikely’ to 6 ‘Extremely Likely’.

Table 3.

Results: DV = Willing to provide email to learn more

Note. Results are presented for the respondent’s answer to the question: ‘Would you be willing to share your Prolific email to learn more information about zocdoc.com/conservativeprofessionals.com.’

Second, there are substantively large differences between ZD and CP conditions by ideological self-identification in expressed willingness to seek care or provide an email, again with the strongest reactions emerging at the political extremes. Liberals are significantly less likely to say they are interested in seeking care from CP compared to ZD. This effect weakens but remains negative all the way through to people who say they are slightly conservative. Those who say they are conservative or extremely conservative express positive interest in seeking care from CP relative to ZD. Because online convenience samples, including this one, severely under-represent conservative respondents, the standard errors are large, but so is the substantive difference. Extremely conservative respondents in the CP treatment express a similar willingness to seek care as extremely liberal people in the ZD treatment, effectively erasing differences between extremely liberal and extremely conservative people’s willingness to seek care.

Similar patterns emerge for willingness to provide an email address. Conservatives are more likely to say they would provide an email address to CP than to ZD; again this is especially pronounced among the most conservative and liberal respondents.

These findings suggest that politics does in fact influence people’s choice to seek healthcare. This is most pronounced among people at the extremes. It is worth noting that the unwillingness to seek care from an out-group, among liberals, is stronger than a willingness to seek care from the in-group (among conservatives). In other words, the treatment effect was stronger among liberals than conservatives. As the other experiments demonstrate, attaching a political label moves people on both the left and right.Footnote ²⁶

The results also suggest that it is not just that people want politics out of the exam room as there is at least marginally more support among conservatives for having a doctor who shares their ideological background compared to a politically ambiguous doctor from ZD. It may be that people strongly dislike someone of the opposing background (as shown by the larger treatment effect among liberals), and are marginally more receptive to someone who shares their background compared to a politically neutral source.Footnote ²⁷ Further research is needed to more fully understand this distinction.

The Comparative Context

While this paper focuses on the USA, the COVID-19 crisis intersects with politics across the world. For example, Backhaus, Hoven and Kawachi (Reference Backhaus, Hoven and Kawachi2023) use survey data from twenty-one countries and find that far-right voters compared to voters who supported centre parties were nearly three times as likely to express vaccine hesitancy. Likewise, the rhetoric of far-right parties across Europe, while mixed, was often sceptical of governmental policies to prevent COVID-19 from spreading. For example, the far-right German party, AfD, which emerged in the mid-2010s, framed its opposition to COVID-19 measures in populist anti-elite and anti-immigrant appeals (Lehmann and Zehnter Reference Lehmann and Zehnter2022). In some cases, where the far-right was out of government, far-right parties organized anti-government lockdowns (Wondreys and Mudde Reference Wondreys and Mudde2022).

Data from the International Social Survey Programme, which surveys respondents across multiple countries, shows that, like in the USA, people who supported far-right parties were less likely to say they trusted doctors compared to supporters of centre parties and far-left parties in 2021 (SA 8; ISSP Research Group 2014, 2024).Footnote ²⁸ In some countries, like Germany, a clear partisan divide emerged over the 2010s: far-right supporters were slightly more likely than average to say they trusted doctors in 2011, but were 13 percentage points less likely than the average voter to say they trusted doctors in 2021. In Italy, on the other hand, far-right supporters were the least likely to say they trusted doctors in 2021, but they were also the least likely to say they trusted doctors in 2011.