Assessing the Impact of Mainstream Media Misinformation on Intent to Undergo Colorectal Cancer (CRC) Screening Among Marginalized Communities: Results From an Online Randomized Controlled Trial (RCT)

Introduction: While the NordICC trial showed that colonoscopy reduced CRC incidence (31%) and mortality (50%) in per-protocol analysis, many media outlets focused on the intention-to-screen findings (i.e., no change in mortality) with headlines questioning the effectiveness of colonoscopy. We hypothesized that misleading news reports on the NordICC trial may have disproportionately affected perceptions of CRC screening in marginalized communities. To test our hypothesis, we performed a secondary analysis of data from an online RCT assessing the impact of two media articles (one rated high- and one low-quality in terms of scientific accuracy) on intention to undergo colonoscopic CRC screening. Methods: We recruited a representative nationwide sample of unscreened Americans aged 45-75 years at average risk for CRC to complete the survey. Participants were randomized 1:1 to review a high- or low-quality media article on the NordICC trial as determined by 19 gastroenterologists using the Quality Index for Health-Related Media Reports scale. Individuals answered questions on their intent to undergo colonoscopic CRC screening before and after reading the article. The primary outcome was a negative change in intent to undergo colonoscopy (Figure 1). To assess whether the low-quality article differentially affected screening perceptions in certain groups, we conducted regression models with interaction terms between the assigned article and age, sex, race/ethnicity, education, and income. Results: Of the 2,013 individuals surveyed, 1,531 (76.1%) stated they planned to do a colonoscopy or were undecided before reading their assigned article. After reading the article, 10.6% of people no longer intended to do a colonoscopy; no difference was seen between the high- and low-quality articles (adjusted P=0.20). Table 1 shows data from the regression models with interaction terms between the assigned article and age, sex, race/ethnicity, education, and income. The low-quality article did not disproportionally affect screening perceptions across the various sociodemographic groups (all adjusted P >0.05). Conclusion: We found that a low-quality article on the NordICC trial did not differentially impact people’s attitudes towards colonoscopic CRC screening in traditionally marginalized groups. While this is reassuring, coverage of the NordICC trial still led ∼10% of people to change their mind on colonoscopy, underscoring the importance of responsible reporting of clinical research by the media.Figure 1.: Definition for a Negative Change in People’s Plans to be Screened. Table 1. - Estimated Marginal Probabilities From Regression Analysis on Having a Negative Change in Intent to Undergo Colonoscopic CRC Screening After Reading the Assigned News Article Variable Negative change in intent to undergo colonoscopic CRC screening Low-quality article % (95% CI) High-quality article % (95% CI) P Age(years) 0.06 45-49 14% (6%-32%) 8% (3%-19%) 50-64 9% (4%-21%) 12% (5%-25%) 65-75 10% (3%-26%) 12% (5%-28%) Sex 0.10 Female 13% (5%-29%) 9% (4%-21%) Male 9% (4%-22%) 12% (5%-25%) Race/ethnicity 0.85 Non-Hispanic White only 11% (4%-24%) 9% (4%-20%) Non-Hispanic Black only 14% (5%-36%) 11% (4%-28%) Hispanic 16% (6%-38%) 11% (4%-27%) Non-Hispanic Asian only 5% (1%-23%) 10% (2%-33%) Other/multiracial/prefer not to say 12% (3%-34%) 12% (3%-34%) Educational attainment 0.08 High school degree 9% (3%-22%) 17% (7%-35%) Some college education 11% (4%-26%) 10% (4%-23%) College degree 12% (4%-27%) 9% (3%-20%) Graduate degree 13% (4%-33%) 8% (2%-23%) Total household income ($) 0.63 < 50,000 12% (5%-26%) 9% (4%-19%) 50,000–100,000 16% (7%-32%) 10% (4%-22%) ≥100,001 13% (5%-30%) 14% (5%-32%) Prefer not to say 5% (1%-34%) 10% (2%-36%) The models included interaction terms between the assigned news article and age, sex, race/ethnicity, educational attainment, and income. P-values are calculated using the likelihood-ratio test between the two models (full model vs full model with the article X variable interaction term). The regression models adjusted for all variables in the table as well as prior exposure to other articles that covered the NordICC trial, marital status, employment status, health insurance, usual source of care, self-reported health status, number of medical comorbidities, political party, main source of news, US region, having a non-first degree relative or diagnosed with CRC, belief on whether is susceptible to CRC, and belief on whether CRC screening is beneficial.