The purpose of this study is to use recently released nationally representative data to estimate the association between perceptions of health information on social media and level of trust in multiple sources of information about cancer. We hypothesized that perception of a lot of health mis- and disinformation on social media would be associated with lower levels of trust in cancer information sources. By extension, this study evaluated the interaction effect between race and ethnicity of the participants, perceptions of social media health mis- and disinformation, and trust in cancer information. We hypothesized that the association between perceptions of a lot of mis- and disinformation on social media and trust in cancer information sources would vary by race and ethnicity. The results of this research have implications for effective communication about cancer in public health education campaigns, especially for racial and ethnic minoritized populations.

This study used cross-sectional data from the Health Information National Trends Survey 6 (HINTS 6), which is a nationally representative survey of civilian, noninstitutionalized adults aged 18 years and older living in the United States. HINTS 6 provides data on adults’ knowledge of cancer risk factors, attitudes toward cancer screening, and cancer prevention and screening behaviors. HINTS 6 used a 2-stage probability sample of residential addresses. Mail and online surveys were administered to household members from March 7 to November 8, 2022, with a response rate of 28.1% [28]. The data are publicly available and deidentified. Further details about the survey methodology and recruitment procedures are available from the HINTS 6 Methodology Report [28].

Given the focus of this study was perceptions of false or misleading health information on social media, persons that reported that they did not use social media were excluded. There were 4710 cases with complete data for the dependent and independent variables. After using listwise deletion for 573 cases with missing data for the control variables, the final analytical sample consisted of 4137 adult social media users.

Our dependent variables were measured by asking participants, “In general, how much would you trust information about cancer from...” Responses included the following: “a doctor,” “family or friends,” “religious organizations and leaders,” “government health agencies,” “charitable organizations,” and “scientists.” The response options were dichotomized into low levels of trust (“not at all” or “a little”) versus high levels of trust (“some” or “a lot”).

The primary independent variable was perceptions about health mis- and disinformation on social media, which was assessed by the following question: “How much of the health information that you see on social media do you think is false or misleading?” HINTS had not measured perceptions about social media mis- and disinformation in prior iterations of the survey. However, this measure did not differentiate between people’s perceptions of mis- versus disinformation. The original response categories were “a lot,” “some,” “a little,” and “none.” We dichotomized this as “less than a lot” (including “some,” “a little,” and “ none”) versus “a lot.” Race and ethnicity were self-reported by the participants in 5 categories: “non-Latino White,” “non-Latino Black,” “Asian American,” “other,” and “Latino.”

Demographic control variables included age (18-34, 35-49, 50-64, and ≥65 years), sex (male and female), marital status (married or cohabiting, formerly married, and never married), residence in a metropolitan versus nonmetropolitan county as designated by the United States Department of Agriculture in 2013, education (high school or less, some college, and college degree or higher), full-time employment status, and feelings about household income (finding it very difficult on present income, getting by on present income, and living comfortably on present income). It should be noted that age was not collected as a continuous variable in HINTS 6, which limited the age categories that could be analyzed. In addition, we controlled for frequency of visiting social media sites (never, monthly/weekly, and daily) in the past 12 months and personal and family (first- or second-degree biological relatives) history of cancer.

All analyses accounted for survey weights and design using jackknife replicate weights for variance estimation. Statistical significance was set at α<.05. The descriptive statistics for the study sample were calculated as survey-weighted percentages accompanied with the raw sample size for each variable. The bivariate relationship between level of trust in cancer information and perceptions of mis- and disinformation were calculated with cell percentages and adjusted Wald P values. Then, multivariable logistic regression models were calculated for each dichotomous outcome. In addition to the main effect, we also tested the interaction effect between perceptions of health mis- and disinformation on social media and participants’ self-reported race and ethnicity. To facilitate interpretation of the interaction effect, we calculated predicted marginal effects from the multivariable logistic regression models.

For this study, the primary focus was perceptions of information on social media. Therefore, we conducted a sensitivity analysis in which we excluded 257 adults who had not visited a social media site in the past year or reported that they did not use social media (n=3880 were included). After excluding these participants, the results were similar, as shown in Multimedia Appendix 1, Table S1. In addition, we conducted a sensitivity analysis for an ordinal measurement of the dependent variables (“a lot,” “some,” “a little,” and “not at all”) using ordered logit regression, and we found that the results were replicated with this alternative measurement, as shown in Multimedia Appendix 1, Table S2. Another sensitivity analysis included participants that did not use social media (n=4986). After including participants that did not use social media, the results were similar, as shown in Multimedia Appendix 1, Table S3. In Multimedia Appendix 1, Table S4, we tested an alternative measurement of the independent variable in which perception of “a lot” of social media mis- and disinformation was compared with respondents that reported “some” and “none” or “a little.” For this sensitivity analysis, we combined “none” and “a little” because only 108 participants chose “none” for this measure. We replicated the main result using this alternative measurement of the independent variable.

The University of Texas Southwestern Medical Center institutional review board determined that the study was exempt from review because it used publicly available data without personal identifiers.

We found that trust in cancer information from doctors or scientists did not vary based on perceptions of health mis- and disinformation on social media. This suggests that people view doctors and scientists as credible sources of cancer information. However, we found that perception of a lot of mis- and disinformation was associated with reduced levels of trust in cancer information from family or friends, government health agencies, charitable organizations, and religious organizations and leaders. This finding supports other studies that found that mis- and disinformation is associated with reductions in trust in media but extends this prior literature by finding an impact on trust in other sources of cancer information [14-19]. Moreover, this finding is consistent with the Comprehensive Model of Information Seeking, which identifies trust as a mechanism linking mis- and disinformation to health behaviors [12,13].

There were notable variations in the relationship between trust in cancer information sources, perceptions of false or misleading health information, and race and ethnicity. For instance, we found that Black participants who perceived a lot of health mis- and disinformation on social media were less likely to have high trust in cancer information from religious organizations and leaders compared to Black participants who perceived less than a lot of health mis- and disinformation on social media. Another contribution of our study is that White participants who perceived a lot of health mis- and disinformation on social media were less likely to have high trust in cancer information from government health agencies and family or friends compared to White participants who perceived less than a lot of health mis- and disinformation on social media. There have been mixed findings on trust in cancer information sources by race and ethnicity in the recent literature, with one study finding higher trust among Black participants for several sources of cancer information compared to White participants and lower trust in doctors among Latino participants compared to White participants [21-25]. However, another study found that trust in cancer information from government health agencies and family or friends declined among Black participants after the COVID-19 pandemic [26]. Our study adds to this literature by identifying that the effect of mis- and disinformation on trusting information sources may vary among racial and ethnic minoritized populations.

We were able to replicate the findings of the study using several different sensitivity analyses, as shown in Multimedia Appendix 1. However, the results should be interpreted within the constraints of the cross-sectional data. First, this study cannot be used to determine the causal relationship between perceptions of mis- and disinformation and trust in social institutions. Second, the 2022 wave of the HINTS survey was the first time that the public’s perceptions of mis- and disinformation were measured. If this measure is collected in subsequent iterations of HINTS, then analyses may be able to detect changes in the association between mis- and disinformation and trust in information sources over time. We note that perceptions of mis- and disinformation may not be an accurate measure of objective exposure to social media mis- and disinformation. Further, this measure does not differentiate between people’s perceptions of mis- versus disinformation. Another limitation is that the focus of this study was on social media mis- and disinformation rather than all media, such as traditional television and print, and therefore the results should be interpreted for this specific form of media. Finally, this study focused on trust in cancer information, and the findings might not apply to trust in other types of health information. By extension, levels of trust in government information may differ between federal and state government health agencies, which were not differentiated in our study [29,30].

Certain sources of cancer information may need enhanced support from the threat of mis- and disinformation, such as government health agencies, charitable organizations, religious organizations and leaders, and family or friends. Moreover, there were notable variations in the relationship between trust in cancer information sources (government health agencies, family or friends, and religious organizations and leaders), perceptions of false or misleading health information, and race and ethnicity. One positive finding is that perceptions of mis- and disinformation were not associated with levels of trust in credible sources of cancer information such as doctors or scientists overall or by race and ethnicity. In prior work, researchers have suggested that interventions should be focused on improving trust in science [1]. Although bolstering trust in science or doctors is important, our findings indicate that other sources of cancer information may be more susceptible to the threat of mis- and disinformation. Moreover, interventions should partner with racial and ethnically minoritized populations that are more likely to have low trust in certain cancer information sources associated with mis- and disinformation on social media.