Inferential Drift and the Table 2 Fallacy in Peruvian Health Literature: When Association Is Mistaken for Explanation

Vera-Ponce V.J.; Zuzunaga-Montoya F.E.; Bustamante-Rodríguez J.C.; García-Lara R.A.; Gutiérrez de Carrillo C.I. · International Journal of Statistics in Medical Research (2026) · DOI: 10.6000/1929-6029.2026.15.36

Resumen

Observational health studies may report multivariable associations without specifying whether the inferential task is descriptive, predictive, or causal, allowing conclusions to exceed what the design and analysis can support. This study aimed to characterize scientific production on associated factors, risk factors, determinants, predictors, or correlates in Peruvian populations or using Peruvian data; to estimate the prevalence of inferential incoherence and level-2 Table 2 fallacy patterns in a full-text sample; and to examine article characteristics associated with the prevalence of inferential incoherence. We conducted a cross-sectional bibliometric and methodological study of Scopus-indexed articles published from 2000 to 2025. A stratified random sample of 400 full texts was coded using prespecified rules; inter-coder reliability was assessed in a random 20% subsample, and adjusted prevalence ratios were estimated using modified Poisson regression with robust variance. Of 1,264 eligible articles, 866 (68.5%) were published during 2020–2025. In the full-text sample, 241/400 (60.3%) were classified as inferentially incoherent and 175/400 (43.8%) as showing a level-2 Table 2 fallacy pattern. Cross-sectional design was associated with a higher prevalence of incoherence (adjusted PR, 1.38; 95% CI, 1.16–1.65). The Scopus-indexed Peruvian health literature on associated factors has expanded substantially, but objective, method, and conclusion were frequently misaligned. The level-2 Table 2 pattern was a frequent prespecified mechanism of this drift, although it partially overlapped operationally with the primary outcome. Editorial and training practices should require authors to state the descriptive, predictive, or causal task before interpreting multivariable coefficients.

Instituciones: Universidad Ricardo Palma; Universidad Nacional Toribio Rodríguez de Mendoza de Amazonas