NIJAMSS peer-reviewed article

Bias in Epidemiological Research and Its Consequences for Public Health Decisions: A Critical Appraisal of Selection Bias, Information Bias and Confounding

Daniel A. Otwori

Abstract: Bias is a systematic deviation from the truth that can arise during study design, participant selection, measurement, analysis, interpretation or publication. Because epidemiological findings guide prevention, clinical practice and allocation of public resources, even modest bias can produce large population-level consequences. This critical narrative review examines selection bias, information bias and confounding, with comparative attention to high-income settings and resource-constrained contexts such as Kenya. Evidence was synthesised from foundational epidemiology texts, reporting guidelines, recent methodological scholarship, World Health Organization materials and studies of routine health-information systems. The review shows that bias rarely occurs as an isolated error. Selection mechanisms can alter the distribution of confounders; measurement error can be differential; and analytical adjustment can introduce additional distortion when investigators condition on colliders or use poorly measured covariates. Contemporary causal-inference literature has refined selection bias as the difference between the target-population effect and the effect estimated in the selected sample. In low-resource settings, methodological bias is compounded by structural sources of error, including incomplete diagnosis, under-reporting, weak record linkage, changing denominators and unequal access to care. Kenya has made important progress through digital reporting and surveillance reform, yet fragmented systems, manual delays and uneven data quality remain relevant risks. The paper proposes a bias-control framework covering design, measurement, analysis, reporting and health-system strengthening. It concludes that bias cannot be eliminated completely, but it can be anticipated, diagnosed, quantified and transparently reported through rigorous design, directed acyclic graphs, validation studies, sensitivity analysis, equity-focused reporting and sustained investment in surveillance quality.

Keywords: Causal inference; Confounding; Epidemiology; Information bias; Kenya; Selection bias; Surveillance

Journal

Nexus International Journal of Applied Science, Medicine and Social Sciences

Publication date

May 2026

Volume / Issue

Vol. 1, Issue 1

Paper number

001

Article type

Critical Narrative Review

Published paper ID

NIJAMSS-V1S1-MAY2026-001

Publisher

Nexus Academic Press

Access

Open Access

Full text PDF: https://www.nijamss.org/archives-download/nijamss-vol-1-series-1-may-2026-paper-001.pdf