New Coronavirus News from 8 Feb 2023

The political polarization of COVID-19 treatments among physicians and laypeople in the United States [pnas.org, 8 Feb 2023]

Authored by Joel M. Levin, Leigh A. Bukowski, Julia A. Minson, and Jeremy M. Kahn

Abstract
In the United States, liberals and conservatives disagree about facts. To what extent does expertise attenuate these disagreements? To study this question, we compare the polarization of beliefs about COVID-19 treatments among laypeople and critical care physicians. We find that political ideology predicts both groups’ beliefs about a range of COVID-19 treatments.

These associations persist after controlling for a rich set of covariates, including local politics. We study two potential explanations: a) that partisans are exposed to different information and b) that they interpret the same information in different ways, finding evidence for both.

Polarization is driven by preferences for partisan cable news but not by exposure to scientific research. Using a set of embedded experiments, we demonstrate that partisans perceive scientific evidence differently when it pertains to a politicized treatment (ivermectin), relative to when the treatment is not identified. These results highlight the extent to which political ideology is increasingly relevant for understanding beliefs, even among expert decision makers such as physicians.

A growing literature on political polarization has documented unexpected links between political ideology and beliefs that are unrelated to the principles of liberalism or conservatism (1, 2). During the COVID-19 pandemic, liberal and conservative Americans have disagreed sharply on matters such as the origins of the virus (3), the severity of the pandemic (4, 5), and the effectiveness of a range of interventions, including masking, distancing, vaccination, and drugs like hydroxychloroquine and ivermectin (3, 5–7). Such disagreements inhibit cooperation, fuel partisan antipathy, and threaten public health.

Prior work offers clues about the roots of this polarization: Partisans consume different information (4, 8, 9), evaluate the same information in different ways (10, 11), and often lack the tools (12) or motivation (13, 14) to discriminate between accurate and inaccurate claims. On these bases, we might expect beliefs about COVID-19 treatments to be dramatically less polarized among people who are particularly informed, trained, and motivated, such as physicians.

In the present work, we examine physicians’ beliefs about treatments for COVID-19, benchmark their polarization against that of lay adults, and provide evidence for two psychological mechanisms that give rise to polarized beliefs among both groups.

Methods

We study two samples: a novel panel of 592 board-certified critical care physicians (“physicians”) and a sample of 900 adults recruited from an online panel (“laypeople”), all based in the United States. We focused on critical care physicians because they are important decision makers in the treatment of severe COVID-19 and because their day-to-day judgments are less influenced by patient preferences, compared to other physicians. Additional information about both samples and other methodological details are in SI Appendix.

We surveyed physicians in three phases between April 2020 and April 2022 and surveyed laypeople concurrently with the final physician survey. In each survey, physicians evaluated a clinical vignette about a severely ill COVID-19 patient and made decisions about which treatments to administer. For each treatment option, physicians reported beliefs about effectiveness and the quality of clinical evidence and made incentivized predictions about the proportion of their peers who made the same decision. Laypeople reported beliefs about treatments but did not make treatment decisions. All participants also reported beliefs about the effectiveness of COVID-19 vaccines and their support for vaccine mandates as well as a range of individual characteristics. Most estimates are based on data from Phase 3 surveys, administered in March and April 2022.

To investigate the role of information consumption on belief polarization, we asked participants in both samples to report how they consume news (e.g., print, social media, television) as well as their preferred cable news source (if applicable). Cable news consumption is a particularly plausible source of variation in public exposure to COVID-relevant information because partisans watch different cable news networks (4), and networks differed markedly in their coverage of hydroxychloroquine, ivermectin, and vaccination (9, 15, 16). Physicians also reported how they engage with scientific research.

To measure bias in the evaluation of information, we embedded an experiment in surveys administered to both samples. Participants read an abridged research abstract (physicians) or a research summary written in a journalistic style (laypeople), both of which reported the results of the TOGETHER trial (17), a well-powered randomized controlled trial that failed to find evidence that ivermectin was effective for treating COVID-19.* Between subjects, we randomized whether the treatment was identified as ivermectin or was anonymized (“GL-22”). We then elicited beliefs about the study’s informativeness, its methodological rigor, and the likelihood that its authors were biased. We expected partisans’ beliefs on all three measures to diverge more sharply when the drug was identified as ivermectin.

We measure political ideology on a 7-point scale bounded by “very liberal” and “very conservative,” with the midpoint defined as “middle of the road” (18). To compare polarization across samples and outcomes, we standardized all outcome variables by subtracting the mean response of political moderates (“slightly liberal,” “middle of the road,” “slightly conservative”) from each response and dividing by the standard deviation.† Our primary analyses control for a range of plausible confounding factors, including demographic, professional, and regional characteristics (Fig. 1A caption).

For simplicity and statistical power, we collapsed measures that were collected at multiple time points after ruling out significant temporal variation (ANOVAs; ps > 0.18).‡ The timing of key elicitations can be found in SI Appendix, Table S1. Full study materials are at https://github.com/pitt-healthsciences/covid_polarization/. A preregistration specifying some elements of our analytical approach can be found at https://aspredicted.org/11M_35D.

Following our prespecified exclusion criteria, we retained 410 physician responses and 882 layperson responses. Research was approved by the University of Pittsburgh Institutional Review Board. All subjects provided informed consent.

Results
We find robust evidence of polarization on eight of ten physician outcomes and all six layperson outcomes (ps < .001; Fig. 1A). To illustrate the magnitude of these effects, conservative physicians were approximately five times more likely than their liberal and moderate colleagues to say that they would treat a hypothetical COVID-19 patient with hydroxychloroquine (Cohen’s h = .37). On average, physicians’ beliefs were less polarized than laypeoples’ (