TRUST IN PHYSICIANS AND HOSPITALS DURING THE COVID-19 PANDEMIC IN A 50-STATE SURVEY OF US ADULTS
Roy H. Perlis, MD, MSc; Katherine Ognyanova, PhD; Ata Uslu, MS; Kristin Lunz Trujillo, PhD; Mauricio Santillana, PhD; James N. Druckman, PhD; Matthew A. Baum, PhD; David Lazer, PhD JAMA Network Open. 2024;7(7):e2424984. Doi
COMMENTARY BY RONALD PETERS MD
No suprisse here. Doctors have become the markting arm of Big Pharma. Drugs, drugs and more drugs in 10 to 15 minute office visits. In medical school training, chronic diseases such as hypertension, anxiety, depression, irritiable bowel syndrome, allergies and more are treated with symptomic medications. Chronic diseases are “managed”, not cured. Fortunately medical training is beginning to include integrative medical practices with nutritional counseling, exercise recommendations, and stress reduction tools. Also, increasing numbers of physicians are recommending vitamins, minerals and other nutrients that make up the natural wisdom of the body.
ABSTRACT
IMPORTANCE
Trust in physicians and hospitals has been associated with achieving public health goals, but the increasing politicization of public health policies during the COVID-19 pandemic may have adversely affected such trust.
OBJECTIVE
To characterize changes in US adults’ trust in physicians and hospitals over the course of the COVID-19 pandemic and the association between this trust and health-related behaviors.
DESIGN SETTING, AND PARTICIPANTS
This survey study uses data from 24 waves of a nonprobability internet survey conducted between April 1, 2020, and January 31, 2024, among 443 455 unique respondents aged 18 years or older residing in the US, with state-level representative quotas for race and ethnicity, age, and gender.
MAIN OUTCOME AND MEASURE
Self-report of trust in physicians and hospitals; self-report of SARS-CoV-2 and influenza vaccination and booster status. Survey-weighted regression models were applied to examine associations between sociodemographic features and trust and between trust and health behaviors.
RESULTS
The combined data included 582 634 responses across 24 survey waves, reflecting 443 455 unique respondents. The unweighted mean (SD) age was 43.3 (16.6) years; 288 186 respondents (65.0%) reported female gender; 21 957 (5.0%) identified as Asian American, 49 428 (11.1%) as Black, 38 423 (8.7%) as Hispanic, 3138 (0.7%) as Native American, 5598 (1.3%) as Pacific Islander, 315 278 (71.1%) as White, and 9633 (2.2%) as other race and ethnicity (those who selected “Other” from a checklist). Overall, the proportion of adults reporting a lot of trust for physicians and hospitals decreased from 71.5%(95%CI, 70.7%-72.2%) in April 2020 to 40.1%(95%CI, 39.4%-40.7%) in January 2024. In regression models, features associated with lower trust as of spring and summer 2023 included being 25 to 64 years of age, female gender, lower educational level, lower income, Black race, and living in a rural setting. These associations persisted even after controlling for partisanship. In turn, greater trust was associated with greater likelihood of vaccination for SARS-CoV-2 (adjusted odds ratio [OR], 4.94; 95 CI, 4.21-5.80) or influenza (adjusted OR, 5.09; 95 CI, 3.93-6.59) and receiving a SARS-CoV-2 booster (adjusted OR, 3.62; 95 CI,
2.99-4.38).
CONCLUSIONS AND RELEVANCE
T This survey study of US adults suggests that trust in physicians and hospitals decreased during the COVID-19 pandemic. As lower levels of trust were associated with lesser likelihood of pursuing vaccination, restoring trust may represent a public health imperative.
INTRODUCTION
Physicians have traditionally represented a key element in public health outreach efforts; most adults will see a physician on a regular basis, and these appointments represent an opportunity to encourage healthy behaviors ranging from diet and exercise1 to smoking cessation2 to seatbelt use3 and firearm safety.4 A 2022 survey reported that US adults had greater trust in physicians and nurses than in any other institution, including the Centers for Disease Control and Prevention,5 a result supported by a Kaiser Family Foundation tracking poll of 2007 adults conducted in 2023.6 Levels of trust increased early in the COVID-19 pandemic; in a Gallup poll, belief that physicians had high or very high ethical standards increased from 65%in 2019 to 77%in 2020.7
However, undercurrents of distrust in medicine are not new in US society—for example, concerns about the health effects of vaccines have persisted long after they were disproven.8 During the COVID-19 pandemic, medicine and public health more broadly became politicized, with the internet amplifying public figures9 and even physicians10 encouraging individuals not to trust the advice of public health experts and scientists.11 As such, the pandemic may have represented a turning point in trust, with a profession previously seen as trustworthy increasingly subject to doubt. By 2023, 1 poll showed perception of physician ethics to be substantially below the prepandemic baseline.7
In the present study, we drew on a 50-state US survey that began early in the COVID-19 pandemic to seek to characterize change in trust in physicians and hospitals over the course of the pandemic, aiming to confirm that trust had decreased. We further investigated the extent to which trust in physicians and hospitals is associated with specific health behaviors, including vaccination and vaccine boosters, to assess the relevance of this trust to public health. As the survey also captured aspects of political preference, it allowed us to distinguish and control for a key potential source of confounding.
METHODS
We used data from 24 waves of a nonprobability internet survey conducted using a commercial vendor, PureSpectrum, which aggregates and deduplicates participants in multiple national panels. PureSpectrum is an online marketplace for survey panel samples working with multiple recruitment vendors. Each of those vendors incentivizes respondents for participation in surveys based on the length of the survey, their specific panelist profile, and target acquisition difficulty. The specific type of incentives for participation varies and may include cash, airline miles, gift cards, redeemable points, sweepstakes entrance, and vouchers. The survey was developed and overseen by a consortium of academic sites, the COVID States Project,12 formed early in the pandemic to understand COVID-19–related attitudes and behaviors. The survey was conducted approximately every 1 to 2 months beginning April 1, 2020, through January 31, 2024, among individuals aged 18 years or older residing in all 50 states and the District of Columbia. A full description of the survey waves and date ranges can be found in eTable 1 in Supplement 1. Participants provided informed consent online. The study protocol was reviewed and approved by the institutional review board of Harvard University as exempt as only deidentified data were used and no participant contact was required. This study followed the American Association for Public Opinion Research (AAPOR) reporting guideline.
To ensure representativeness of US adults, the survey used quotas for gender, age at first survey completion, and race and ethnicity within each state. We included attention checks and openended answers that were used to filter out unreliable or automated respondents (eTable 2 in Supplement 1). This nonprobability sampling approach has previously been shown to yield results that approximate those of probability-sampled designs and administrative data collection.13,14
Individual sociodemographic characteristics were self-reported. Race and ethnicity, as with gender and age, were collected to allow confirmation of the representativeness of the US population and reweighting of the sample. They were identified using a survey instrument that included categories for race and ethnicity. Information on trust in physicians and hospitals and trust in scientists was collected by asking, “How much do you trust the following people and organizations to do what is right?” followed by a list of entities with 4 choices (a lot, some, not too much, or not at all; e Appendix 2 in Supplement 1). In waves prior to August 2022, we asked a variant of this question,
“How much do you trust the following people and organizations to do the right thing to handle the current coronavirus (COVID-19) outbreak?” We also asked about propensity to trust more generally, by asking “Generally speaking, would you say that most people can be trusted, or that you cannot be too careful in dealing with people? Please give your answer on a scale from 1 to 10, where 1 is you cannot be too careful and 10 is most people can be trusted.”
SARS-CoV-2 vaccination status in surveys after the availability of the vaccine was collected by asking individuals whether they had been vaccinated and later by number of prior vaccinations as a means of incorporating information about vaccine boosters. Influenza vaccination status was collected (in survey wave 27, from spring 2023) by asking about prior vaccination or intention to be vaccinated.
STATISTICAL ANALYSIS
Survey results were reweighted with interlocking national weights for age at survey completion, gender, and race and ethnicity, as well as educational level and region, using 2019 US Census American Community Survey data,15 via the survey package in R, version 4.0 (R Project for Statistical Computing),16 a standard approach for nonprobability samples.17
We first used survey-weighted ordinal logistic regression to examine the association between physician trust score and a range of sociodemographic features, drawing on 2 of the survey waves; for these analyses, if a respondent completed both waves, we selected only the index response. We then applied survey-weighted logistic regression models with vaccination against SARS-CoV-2 or influenza as the outcome and incorporated trust in physicians and hospitals as well as sociodemographic features as independent variables. To examine the association between trust in physicians and hospitals reported on the prior survey wave and likelihood of vaccination against SARS-CoV-2 among individuals not previously vaccinated, we also used logistic regression, adjusting for the same covariates as in prior models. This analysis began with wave 18 (early summer 2021) as the first wave that occurred after vaccination was widely available in all US states. All P values were from 2-sided tests, and results were deemed statistically significant at P < .05.
On 1 survey wave (the first of the 2 examined cross-sectionally), we also asked the respondents an open-ended question to identify factors associated with different trust levels: “You said you trust doctors and hospitals [amount of trust]. Can you tell us why that is?” As standard topic modeling approaches do not perform well in brief text,16 we instead used a large language model to summarize themes and example text, to identify factors associated with poor trust (ie, the lowest level of trust). Specifically, we applied a large language model (Generative Pre-trained Transformer 4 turbo, gpt-4-1106-preview; OpenAI) with the temperature hyperparameter that controls the model’s randomness set at 0.7 to identify the 4 major themes reflected in the uncoded data from the open ended responses .We used the following prompt: “Each of the following survey responses explains why someone does not trust doctors or hospitals. From them, without using any other knowledge, identify 4 main themes that can be explained in a brief phrase, plus an ‘other’ category for comments that don’t fit in any theme. Then provide 5 examples of responses in each category.” The model was presented with the full list of responses (via OpenAI’s Python API), then asked to characterize each response into the best-fitting category with a second prompt: “Please place each response in one of these categories that best fits,” to allow an estimate of the proportion reflected in each theme.
RESULTS
The combined data had 582 634 responses across 24 survey waves, including 443 455 unique respondents. The unweighted mean (SD) age was 43.3 (16.6) year; 288 186 (65.0%) reported female gender, and 155 269 (35.0%) reported male gender; 21 957 (5.0%) identified as Asian American, 49 428 (11.1%) as Black, 38 423 (8.7%) as Hispanic, 3138 (0.7%) as Native American, 5598 (1.3%) as Pacific Islander, 315 278 (71.1%) as White, and 9633 (2.2%) as other race and ethnicity (those who selected “Other” from a checklist).
Figure 1 illustrates the proportion of individuals who reported a lot of trust in physicians and hospitals in each survey wave over time, subdivided by gender, race and ethnicity, and age. Overall, the proportion of adults reporting a lot of trust for physicians and hospitals decreased from 71.5% (95%CI, 70.7%-72.2%) in April 2020 to 40.1%(95%CI, 39.4%-40.7%) in January 2024. (In light of the shift in wording for this question, we examined 4001 respondents from June and July 2023 randomly selected to answer both forms of the question; responses to the 2 versions were strongly correlated (Spearman ρ = 0.76; 95%CI, 0.74-0.78). eFigure 9 in Supplement 1 illustrates the change in proportion of responses at each level of trust at each survey wave.
We then focused on 2 waves in spring and summer 2023 (from April 5 to May 5, 2023, and from June 29 to August 1, 2023), indicated by a gray box in Figure 1; characteristics of this cohort are summarized in the Table. eFigure 1 in Supplement 1 illustrates state-by-state proportions of individuals reporting high levels of trust (“a lot”) and low levels of trust (“not at all” or “a little”).
In these 2 waves, we examined associations between individual sociodemographic features and levels of trust in physicians and hospitals in ordinal regression models (Figure 2). Characteristics independently associated with decreased trust included being 25 to 64 years of age, female gender, lower educational level, lower income, Black race, and living in a rural area. Adding self-reported political affiliation did not meaningfully change these associations (eFigure 2 in Supplement 1).
We next examined the association between trust and COVID-19 vaccination status during these 2waves. In logistic regression models, higher levels of trust were associated with a greater likelihood of being vaccinated in unadjusted models (a little trust vs none: odds ratio [OR], 1.63 [95%CI, 1.40-1.90]; some trust vs none: OR, 3.38 [95%CI, 2.95-3.88]; and a lot of trust vs none: OR, 7.59 [95%CI, 6.59-8.75]) and models adjusted for sociodemographic features (a little trust vs none: OR, 1.38 [95% CI, 1.16-1.65]; some trust vs none: OR, 2.48 [95%CI, 2.12-2.90]; and a lot of trust vs none: OR, 4.94 [95%CI, 4.21-5.80]) (Figure 3). Associations were not meaningfully different with further inclusion of political affiliation (eFigure 3 in Supplement 1).
Results were similar when considering SARS-CoV-2 vaccine boosters as the outcome, rather than any vaccination, in unadjusted models (a little trust vs none: OR, 1.54 [95%CI, 1.27-1.86]; some trust vs none: OR, 3.29 [95%CI, 2.77-3.92]; and a lot of trust vs none: OR, 5.96 [95%CI, 5.02-7.09]) and adjusted models (a little trust vs none: OR, 1.23 [95%CI, 1.00-1.52]; some trust vs none: OR, 2.22 [95%CI, 1.84-2.68]; and a lot of trust vs none: OR, 3.62 [95%CI, 2.99-4.38]) (eFigure 4 in Supplement 1). Inclusion of party affiliation was associated with similar results (eFigure 5 in Supplement 1).
We then repeated these analyses for influenza vaccination, available in the first of the 2 waves, to assess whether trust generalized beyond COVID-19 to other health-related behaviors. Once again, higher levels of trust were significantly associated with vaccination status in unadjusted models (a little trust vs none: OR, 1.40 [95%CI, 1.06-1.83]; some trust vs none: OR, 3.48 [95%CI, 2.72-4.46]; and a lot of trust vs none: OR, 7.43 [95%CI, 5.79-9.53]) and adjusted models (a little trust vs none: OR, 1.21 [95%CI, 0.91-1.61]; some trust vs none: OR, 2.63 [95%CI, 2.03-3.40]; and a lot of trust vs none: OR, 5.09 [95%CI, 3.93-6.59]) (eFigure 6 in Supplement 1). As in the other analyses, inclusion of political party affiliation yielded similar results (eFigure 7 in Supplement 1).
We also considered whether trust in physicians and hospitals was explained by other forms of trust, by adding overall sense of trustworthiness of other people, as well as trust in scientists, to the logistic regression model for vaccination status. Including these terms, greater trust in physicians and hospitals was significantly associated with SARS-CoV-2 vaccination (a little trust vs none: adjusted OR, 1.03 [95%CI, 0.84-1.26]; some trust vs none: adjusted OR, 1.37 [95%CI, 1.13-1.66]; and a lot of trust vs none: adjusted OR, 1.94 [95%CI, 1.59-2.36]) (eFigure 8 in Supplement 1). Although the survey design does not allow us to examine causation directly, we next considered lagged trust as a factor associated with vaccination status at the following wave, among individuals who responded to 2 consecutive waves. We estimated a logistic regression model at each survey wave beginning with wave 16 (January 2021), each with SARS-CoV-2 vaccination status as outcome
READ THE ORIGINAL ARTICLE FOR GRAPHS OF DATA ANALYSIS AND REFERENCES