Abstract

Purpose:

Determine factors associated with perceived amount of health misinformation when using social media.

Methods:

Secondary analysis of the Health Information National Trends Survey 7 (HINTS 7). US adults 18 years and older completed surveys in 2024 (N = 7,278). Participants with complete data for all variables were included for analysis (N = 4,741). Univariable and multivariable logistic regression models were used to identify correlates of interest for how much health misinformation participants saw when using social media. “A lot” and “some” responses were compared to “a little” and “none.” As a sensitivity analysis, “a lot” was compared to all other responses combined.

Results:

Higher odds of perceiving “a lot/some” misinformation were observed among adults aged 50–64 years and 65–74 years (vs. 18–34 years), college graduates (vs. non-college graduates), and incomes of $50,000–$99,999 and $100,000 or more (vs. $0–$19,999). Lower odds were identified in Black and Hispanic (vs. White) participants and those who agreed (vs. disagreed) their social media network had the same views on health. Odds of perceiving “a lot” of misinformation were higher in adults aged 50–64 years and college graduates. Lower odds were found in those who: were Black and Hispanic, agreed their social media network had similar health views, had higher trust in the healthcare system, and disagreed (vs. neutral) to a strong sense of ethnic group belonging.

Conclusion:

Perceived health misinformation on social media varies by sociodemographic characteristics, trust in the healthcare system, and alignment of health views within social media networks.

Introduction

In July 2021, during the COVID-19 pandemic, the US Surgeon General’s Advisory issued a public statement to draw attention to health misinformation and provided advice on how individuals and organizations could reduce the spread of misinformation (1). Health misinformation was described as “information that is false, inaccurate, or misleading according to the best available evidence at the time” (1). Social media has been a prominent channel through which misinformation about COVID-19 and COVID-19 vaccination has circulated, and is a common way health misinformation spreads (2–7). For example, Stephan et al. (8) found that most TikTok videos about female pelvic floor conditions were of poor informational quality and that 18 percent contained misinformation. Kaya et al.’s (9) evaluation of YouTube videos on hypertension management determined that nearly half (49 percent) were misleading. Dhanoya et al. (10) reported that among fertility-related content on Twitter/X and Instagram, 74 percent of posts did not cite sources or academic references and 45 percent contained inaccurate information. Loeb et al. (11) examined 150 widely viewed videos on YouTube about prostate cancer screening and treatment, and found that 115 (77 percent) contained biased or poor-quality information. These patterns raise concern given the widespread use of social media among US adults. According to a 2025 Pew Research Center survey, 84 percent of US adults reported using YouTube and 71 percent reported using Facebook, with substantial use of Instagram (50 percent) and TikTok (37 percent) (12).

Evaluating how individuals perceive health misinformation on social media is essential for identifying vulnerable populations, understanding how belief systems shape interpretation and trust, and informing interventions that strengthen the public’s ability to recognize misleading information. A systematic review by Wang et al. (13) emphasized the importance of examining how susceptibility to health misinformation varies across sociodemographic groups and how belief systems contribute to its spread. Prior research suggests that perceptions of information credibility are shaped by a combination of individual characteristics, trust in institutions, information-processing capacity, and social context (13, 14). From an information processing perspective, individuals may rely on both analytic and heuristic processes when evaluating health information (15–17). These dual-processing modes have distinct implications: analytic processing involves systematic and effortful thinking, whereas heuristic processing relies on mental shortcuts and peripheral cues. In complex or unpredictable communication environments such as social media, heuristic cues such as trust in the source of information or consistency with prior beliefs may shape judgments of credibility and increase susceptibility to misinformation (15).

Individuals’ positions within the social structure, including education, income, and race/ethnicity, may influence access to and evaluation of health information, as described by communication inequality frameworks (18, 19). Trust in the healthcare system is a particularly important factor, as it may influence how individuals interpret health information and which sources they rely on for high-quality health information. Lower trust has been associated with greater skepticism toward authoritative information and increased susceptibility to misinformation (20, 21). Moreover, individuals may be more likely to accept health information that aligns with their beliefs, values, and identities rather than evidence that is factual or objectively derived, consistent with motivated reasoning (22). Social media environments can reinforce existing beliefs through exposure to like-minded networks and algorithm-driven content, which may contribute to confirmation bias and differences in perceptions of misinformation (23–25). In parallel, individuals’ capacity to find, understand, evaluate, and use health information, often conceptualized as health literacy or digital health literacy, affects their ability to distinguish credible information from misleading content (26–29).

The present study is a secondary analysis of the 2024 Health Information National Trends Survey (HINTS 7) data. Informed by concepts from information processing theory, communication inequality frameworks, and motivated reasoning, we selected variables identified in the literature as being associated with health misinformation, including: (1) sociodemographic characteristics and health-related factors; (2) health literacy and information-seeking capacity; (3) trust in the healthcare system; and (4) social context, including alignment of health views within individuals’ social networks. Given the breadth of potential influences and the use of secondary survey data, this analysis was exploratory in nature, aiming to identify patterns that may inform future hypothesis-driven research. This analysis extends prior work using HINTS data (30–33) and may inform the development of targeted interventions for groups less likely to perceive substantial amounts of health misinformation on social media.

Materials and methods

HINTS is administered by the National Cancer Institute (34). HINTS 7 data were collected from March to September 2024. Non-institutionalized US adults aged 18 years and older completed surveys via two modes: paper and web. The sampling strategy had a two-stage design: (1) selecting an address from a stratified sample of addresses, and (2) selecting one adult from each sampled household. Participants received a $2 unconditional incentive during the first mailing and were offered a bonus of $10 Amazon electronic gift card to respond on the web. Full methodology of the HINTS 7 (2024) can be found on the HINTS website (35). A total of 7,278 participants completed the HINTS 7 surveys. After including only those with complete data for all variables of interest, 4,741 participants were included for this study’s analyses.

Perceived frequency of health misinformation on social media

Participants were asked, “How much of the health information that you see on social media do you think is false or misleading?” with response options of a lot, some, a little, and none. For the primary analysis, we compared “a lot/some” vs. “a little/none.” As a sensitivity analysis, we further compared “a lot” vs. all other responses (some, a little, and none).

Sociodemographic characteristics

We evaluated the following age groups in years: 18–34 (reference), 35–49, 50–64, 65–74, and 75+. Participants were asked “What sex were you assigned at birth, on your original birth certificate?” and had the following response options: male (reference), female, and don’t know. Regarding education, we compared those who were college graduates and those who were not college graduates (reference). Race/ethnicity was evaluated with the following categories: non-Hispanic White (reference), non-Hispanic Black/African American, Hispanic, non-Hispanic Asian, and non-Hispanic other. Participants were asked “Thinking about members of your family living in this household, what is your combined annual income, meaning the total pre-tax income from all sources earned in the past year?” For the analysis of this study, the following categories of household annual income were compared: $0 to $19,999 (reference), $20,000 to $49,999, $50,000 to $99,999, and ≥ $100,000.

Frequency of internet use

Participants were asked, “About how often do you use the Internet, either on a computer, laptop, smartphone or any other device?” Response options were about once per day, a few times a week, less than once per week, rarely, and never. We categorized responses as the following: “about once per day,” “a few times a week” and “less than once per week” combined, and “rarely” and “never” combined (reference).

Frequency of social media use

Participants were asked in the past 12 months how often they did the following: (1) visited a social media site, (2) interacted with people who have similar health or medical issues on social media or online forums, and (3) watched a health-related video on a social media site (for example, YouTube). Responses options for each of these three questions were: almost every day, at least once a week, a few times a month, less than once a month, and never. For questions on (1) visiting a social media site and (2) interacting with others who have a similar health issue, we compared “almost every day” to all other responses combined (reference). For the question about (3) frequency of watching a health-related video, we compared those who watched a health-related video (i.e., almost every day to less than once a month) to those who responded “never” (reference).

Self-rated general health

Participants were asked if they would say their health is excellent, very good, good, fair, or poor. We compared those who responded excellent, very good, and good (reference) to those who responded fair and poor.

Frequency of going to healthcare provider

Participants were asked, “In the past 12 months, not counting times you went to an emergency room, how many times did you see a doctor, nurse, or other health professional to get care for yourself?” We evaluated responses as the following categories: none (reference), 1–4 times, and 5 or more times.

Trust in healthcare system

Participants were asked “How much do you trust the healthcare system (for example, hospitals, pharmacies, and other organizations involved in healthcare)?” Response options were a lot, some, a little, and not at all. We compared to those who responded that they had “a lot” with those who responded some, a little, and not at all responses combined (reference).

Tell whether health information is true/false on social media

Participants were asked how much they agree or disagree that they “find it hard to tell whether health information on social media is true or false.” We compared those who responded strongly agree and somewhat agree with those who responded strongly disagree and somewhat disagree (reference).

Confidence in filling out medical forms

Participants were asked “How confident are you filling out medical forms by yourself?” Response options were very, somewhat, a little, and not at all. We compared those who responded “very” with all other responses combined (reference). This item is a proxy for limited health literacy skills (36).

Search skills for health information on internet

Participants were asked how much they agree or disagree with the statement that they have the skills to find the health information they need on the Internet. We compared those who responded strongly agree and somewhat agree with those who responded strongly disagree and somewhat disagree (reference).

Depression and anxiety

Participants were asked over the past 2 weeks how often they have been bothered by the following problems: (1) little interest or pleasure in doing things, (2) feeling down, depressed, or hopeless, (3) feeling nervous, anxious, or on edge, and (4) not being able to stop or control worrying. Response options for all questions were: nearly every day, more than half the days, several days, and not at all. We assigned a numeric score for each response for each of the 4 questions: nearly every day as 3, more than half the days as 2, and several days as 1, and not at all as 0. The sum of scores for the first two questions was used to evaluate depression. The sum of scores for the last two questions was used to assess anxiety. For each condition, possible scores ranged from 0 to 6. These items are a valid tool to assess depression and anxiety (37).

Social media network has same views about health

Participants were asked how much they agree or disagree that “most of the people in my social media networks have the same views about health as me.” We compared to those who strongly agree and somewhat agree to those who strongly disagree and somewhat disagree (reference).

Talk with friends/family about health

Participants were asked “Do you have friends or family members that you talk to about your health?” Response options were yes and no (reference).

Political viewpoints

Participants were asked, “Thinking about politics these days, how would you describe your own political viewpoint?” We evaluated political viewpoints as liberal (i.e., responses of very liberal, liberal, and somewhat liberal), moderate, and conservative (i.e., responses of very conservative, conservative, or somewhat conservative).

Belonging with own ethnic group

Participants were asked how much they agree or disagree that they “have a strong sense of belonging to my own ethnic group.” Response options were strongly agree, agree, neither agree nor disagree, disagree, and strongly disagree. We evaluated responses as the following: strongly agree/agree, neither agree nor disagree (reference), and strongly disagree/disagree.

Data analysis

Logistic regression models were used to determine factors associated with perceiving a lot/some health misinformation. As a sensitivity analysis, additional models were created to determine factors associated with perceiving a lot of health misinformation. Factors with an overall p-value of less than 0.10 in univariable models were considered in multivariable models. Backward elimination was used to arrive at a parsimonious model given the large number of candidate predictors and the exploratory nature of this analysis. Independent variables were sequentially removed until all remaining variables were statistically significant at p-value < 0.05. Analyses were conducted in R version 4.5.1 with the following packages: haven, dplyr, survey, srvyr, and broom (38). A survey design object accounting for jackknife weighting and replicate weights was created with the function “as_survey_rep” (38). Logistic regression models using this survey design object were created using the function “svyglm” (38). Weighted percentages, adjusted odds ratios (aOR), 95% confidence intervals (95% CI), and p-values are reported.

ResultsCharacteristics of the analytical sample

Supplementary Table 1 summarizes the characteristics of the 4,741 included participants, including both weighted and unweighted percentages. Based on weighted estimates, nearly 60% of participants were between ages 18 and 49, with slightly more males than females (51%), a majority of participants were non-college graduates, and representation across race/ethnicity and income levels. Most participants (84%) reported excellent, very good, or good general health.

Primary analysis comparing a lot/some vs. a little/none perceived health misinformation

Weighted row percentages and univariable logistic regression models for perceiving a lot/some health misinformation on social media are in Table 1. Exactly 81.6% (weighted percentage) perceived lot/some health misinformation and 18.4% perceived a little/no health misinformation on social media. The multivariable logistic regression model results for perceiving a lot/some health misinformation are in Table 2. Compared to the 18–34 years age group, participants who were 50–64 years (OR = 1.67 [95% CI: 1.15, 2.42]) and 65–74 years (OR = 1.89 [95% CI: 1.25, 2.86]) had greater odds of perceiving a lot/some misinformation. College graduates (OR = 1.53 [95% CI: 1.08, 2.18]) had greater odds of perceiving a lot/some misinformation compared to those who were non-college graduates. Black/African American (OR = 0.54 [95% CI: 0.36, 0.81]) and Hispanic (OR = 0.49 [95% CI: 0.34, 0.69]) participants had lower odds of perceiving a lot/some misinformation compared to White participants. Those with an annual household income of $50,000–$99,999 (OR = 1.63 [95% CI: 1.02, 2.61]) and $100,000 or more (OR = 1.75 [95% CI: 1.03, 2.98]) had greater odds of perceiving a lot/some misinformation than those in the $0–$19,999 range. Participants who agreed (OR = 0.64 [95% CI: 0.48, 0.84]) that people in their social media network have the same views on health as them had lower odds of perceiving a lot/some misinformation compared to those who disagreed.

“How much of the health information that you see on social media do you think is false or misleading?”CharacteristicsA lot/some % (SE)A little/none % (SE)Unadjusted OR [95% CI]‡P-valueNumber of participants (weighted %)3,925 (81.6%)816 (18.4%)Age groupOverall < 0.00118–3474.8 (2.3)25.2 (2.3)Reference–35–4982.7 (1.7)17.3 (1.7)1.61 [1.13, 2.31]0.01050–6486.3 (1.6)13.7 (1.6)2.12 [1.47, 3.06]< 0.00165–7487.2 (1.6)12.8 (1.6)2.30 [1.59, 3.33]< 0.00175+77.5 (6.1)22.5 (6.1)1.16 [0.51, 2.68]0.714Birth sexOverall = 0.072Female83.4 (1.2)16.6 (1.2)1.27 [1.00, 1.61]0.052Male79.9 (1.4)20.1 (1.4)Reference–Don’t know62.7 (21.2)37.3 (21.2)0.42 [0.05, 3.46]0.413EducationNot college graduate78.2 (1.4)21.8 (1.4)Reference–College graduate87.4 (1.2)12.6 (1.2)1.94 [1.46, 2.58]< 0.001Race/ethnicityOverall < 0.001Non-Hispanic White86.0 (1.0)14.0 (1.0)Reference–Non-Hispanic Black or African American75.5 (2.9)24.5 (2.9)0.50 [0.35, 0.73]< 0.001Hispanic71.1 (2.8)28.9 (2.8)0.40 [0.29, 0.57]< 0.001Non-Hispanic Asian80.3 (6.8)19.7 (6.8)0.66 [0.26, 1.71]0.389Non-Hispanic other78.7 (4.4)21.3 (4.4)0.60 [0.34, 1.08]0.089IncomeOverall < 0.0010 to $19,99971.6 (3.7)28.4 (3.7)Reference–20,000 to $49,99974.9 (2.3)25.1 (2.3)1.18 [0.74, 1.89]0.47250,000 to $99,99983.9 (1.5)16.1 (1.5)2.07 [1.31, 3.29]0.003≥ $100,00087.3 (1.5)12.7 (1.5)2.73 [1.69, 4.43]< 0.001Frequency of internet useOverall = 0.021More than once a day82.6 (0.9)17.4 (0.9)1.69 [0.38, 7.51]0.483About once per day and less than once per week70.3 (4.9)29.7 (4.9)0.85 [0.18, 4.07]0.828Rarely and never73.8 (12.6)26.2 (12.6)Reference–Frequency of visiting social media siteAlmost every day81.9 (1.1)18.1 (1.1)1.10 [0.85, 1.41]0.460Not almost every day (at least once a week to never)80.5 (1.6)19.5 (1.6)Reference–Interact with people with similar health issue on social mediaAlmost every day67.3 (9.3)32.7 (9.3)0.46 [0.19, 1.13]0.087Not almost every day (at least once a week to never)81.8 (1.0)18.2 (1.0)Reference–Watching health-related video on social mediaWatched (almost every day to less than once a month)81.9 (1.0)18.1 (1.0)1.09 [0.85, 1.40]0.483Never80.5 (1.8)19.5 (1.8)Reference–General healthFair and poor80.4 (2.3)19.6 (2.3)0.91 [0.66, 1.27]0.580Excellent, very good, and good81.8 (1.0)18.2 (1.0)Reference–Frequency of visiting health professionalOverall = 0.003None75.2 (3.0)24.8 (3.0)Reference–1 to 4 times80.7 (1.4)19.3 (1.4)1.38 [0.95, 2.00]0.086≥ 5 times86.4 (1.3)13.6 (1.3)2.09 [1.37, 3.20]0.001Trust in healthcare systemA lot82.4 (1.9)17.6 (1.9)1.09 [0.78, 1.51]0.618Not a lot (some, a little, and not at all)81.1 (1.2)18.9 (1.2)Reference–Hard to tell whether health information on social media is true or falseStrongly/somewhat agree82.1 (1.2)17.9 (1.2)1.10 [0.84, 1.45]0.490Strongly/somewhat disagree80.6 (1.7)19.4 (1.7)Reference–Confidence in filling out medical formVery84.6 (1.2)15.4 (1.2)1.79 [1.27, 2.52]0.001Not very (somewhat, a little, and not at all)75.3 (2.3)24.7 (2.3)Reference–Search skills to find health information needed on the internetStrongly/somewhat agree81.8 (0.9)18.2 (0.9)1.24 [0.72, 2.15]0.432Strongly/somewhat disagree78.4 (4.4)21.6 (4.4)Reference–Score on depression scale (median [IQR])0 [0, 2]1 [0, 2]0.92 [0.84, 1.00]0.051Score on anxiety scale (median [IQR])0 [0, 2]0 [0, 2]0.96 [0.89, 1.04]0.364People in my social media network has same views on health as meStrongly/somewhat agree78.8 (1.4)21.2 (1.4)0.70 [0.54, 0.91]0.008Strongly/somewhat disagree84.1 (1.3)15.9 (1.3)Reference–Have friends or family to talk about healthYes82.1 (0.9)17.9 (0.9)1.20 [0.86, 1.69]0.276No79.2 (2.7)20.8 (2.7)Reference–Political viewpointOverall = 0.140Conservative81.0 (1.8)19.0 (1.8)1.11 [0.79, 1.57]0.548Moderate79.4 (1.7)20.6 (1.7)Reference–Liberal84.9 (1.9)15.1 (1.9)1.46 [1.00, 2.13]0.052Strong sense of belonging to my own ethnic groupOverall = 0.060Strongly agree and agree79.9 (1.5)20.1 (1.5)0.85 [0.63, 1.15]0.282Neither agree nor disagree82.4 (1.5)17.6 (1.5)Reference–Strongly disagree and disagree86.7 (2.0)13.3 (2.0)1.39 [0.97, 2.01]0.076

Weighted percentages and univariable logistic regression model results for perceiving “a Lot/Some” health misinformation on social media (n = 4,741).

*Row percentages are weighted.

†

SE is weighted standard error.

‡

OR is odds ratio and 95% CI is 95% confidence interval.

“How much of the health information that you see on social media do you think is false or misleading?”CharacteristicsAdjusted OR [95% CI]*P-valueAge groupOverall = 0.01618–34Reference–35–491.28 [0.87, 1.86]0.20250–641.67 [1.15, 2.42]< 0.00165–741.89 [1.25, 2.86]0.00475+0.91 [0.38, 2.17]0.831EducationNot college graduateReference–College graduate1.53 [1.08, 2.18]0.019Race/ethnicityOverall = 0.001Non-Hispanic WhiteReference–Non-Hispanic Black or African American0.54 [0.36, 0.81]0.004Hispanic0.49 [0.34, 0.69]< 0.001Non-Hispanic Asian0.72 [0.28, 1.82]0.474Non-Hispanic other0.69 [0.37, 1.28]0.229IncomeOverall = 0.008$0 to $19,999Reference–$20,000 to $49,9991.06 [0.65, 1.74]0.806$50,000 to $99,9991.63 [1.02, 2.61]0.044≥ $100,0001.75 [1.03, 2.98]0.039People in my social media network has same views on health as meStrongly/somewhat agree0.64 [0.48, 0.84]0.002Strongly/somewhat disagreeReference–

Multivariable logistic regression results for perceiving “a Lot/Some” health misinformation on social media (n = 4,741).

*OR is odds ratio and 95% CI is 95% confidence interval.

Sensitivity analyses comparing “a Lot” to all other categories of perceived health misinformation

Weighted row percentages and univariable logistic regression models for perceiving a lot of heal