Two related analyses of social-media health-information judgement and self-reported health-decision use: one within 2024 and one comparing independent survey cycles in 2022 and 2024.
Research-development work · Observational findings · Not a peer-reviewed or AI-specific study
Adjusted odds ratio (95% CI 1.42–2.34); complete-case n=5,183.
Cross-sectional survey-weighted logistic regression using the HINTS person weight and 50 JK1 replicate weights. Association does not establish causation.
Methods and validationSurvey-year interaction OR (95% CI 0.83–1.73); p=0.330; pooled n=10,838.
Repeated cross-sectional model with 100 merged Rizzo replicate weights and 98 design degrees of freedom. No established change in association.
Methods and validationWeighted percentage reporting use of social-media information for personal health decisions, by cycle and reported difficulty judging truthfulness:
| Cycle | No reported difficulty | Reported difficulty |
|---|---|---|
| 2022 · HINTS 6 | 13.4% (95% CI 9.8–17.0%) | 18.8% (95% CI 15.8–21.7%) |
| 2024 · HINTS 7 | 15.1% (95% CI 12.5–17.8%) | 24.2% (95% CI 21.8–26.6%) |
The descriptive gap is numerically larger in 2024, but the formal survey-year interaction is not statistically significant (p=0.330). A within-person change cannot be inferred from independent cross-sectional samples.
Population survey weights and replicate-based standard errors were used instead of treating observations as a simple random sample. The pooled survey follows NCI merging-tool logic.
Perceived difficulty judging social-media health information is measured with a single survey item. It should not be presented as a validated measure of digital health literacy.
Neither design establishes that judging difficulty causes information use. Temporality, residual confounding and reporting effects remain unresolved.
Which information-judgement processes could explain this association, and how might a prospective or experimental design distinguish the possible explanations?
Project repositories contain code, variable definitions, saved outputs and study-level validation reports. The 11 October 2026 cross-project desk audit reviewed records; it did not independently rerun the raw public-use survey files.
Research-development analyses in Catherine Ge's portfolio. No peer-review claim or unverified claim of independent intellectual contribution.