Collider Bias and Infectious Disease Studies: A  Review

Pratap Kumar  Jena; Jugal  Kishore; Sabyasachi  Shukla; Debashis  Mohapatra; Arpita  Patel

Pratap Kumar Jena Associate Professor, School of Public Health, Kalinga Institute of Industrial Technology (KIIT) Deemed to be University, Bhubaneswar, India
Jugal Kishore Director-Professor, Department of Community Medicine, Vardhman Mahavir Medical College & Safdarjung Hospital, New Delhi, India
Sabyasachi Shukla PhD Scholar, School of Public Health, Kalinga Institute of Industrial Technology (KIIT) Deemed to be University, Bhubaneswar, India
Debashis Mohapatra 4PhD Scholar, School of Public Health, Kalinga Institute of Industrial Technology (KIIT) Deemed to be University, Bhubaneswar, India
Arpita Patel Department of Community Medicine, Kalinga Institute of Medical Sciences, Bhubaneswar, India

Keywords: Collider Bias, Infectious Diseases, HIV/TB; COVID-19, Epidemiology, Causal Inference

Abstract

Collider bias is the distortion of the observed association between an exposure and an outcome that occurs when analysis conditions (by design or adjustment) on a variable that is a common effect of the
exposure and outcome, or their causes, thereby opening a noncausal pathway that induces a spurious association or alters effect estimates. This review explains the mechanism, distinguishes collider bias from
confounding and broader selection bias, and illustrates its impact in infectious disease research, particularly in the context of COVID-19, HIV, tuberculosis (TB), and malaria. During the COVID-19 pandemic, for example, studies restricted to hospitalised patients suggested smoking might protect against severe disease, an artefact later understood as collider bias. Similar issues have emerged in HIV and TB research, where restricting analyses to patients in care or clinical trials created misleading associations, and in malaria studies, where hospital-based sampling distorted genetic findings. We conducted a structured narrative review of the literature published in the past decade, identifying methodological and applied papers that described collider bias in
infectious disease contexts. Key findings show that collider bias can produce false protective effects, reverse expected associations, and either exaggerate or attenuate true effects. Practical solutions include the use of directed acyclic graphs (DAGs) to identify colliders, application of inverse probability weighting, sensitivity analyses, and reweighting with external population data. The broader implications extend beyond
methodology: collider bias can shape public perception, misinform clinical practice, and obscure health disparities. Recognising and addressing collider bias is therefore essential for generating accurate,
equitable, and actionable evidence in infectious disease research.
Keywords: Collider Bias, Infectious Diseases, HIV/TB; COVID-19, Epidemiology, Causal Inference

DOI: https://doi.org/10.24321/0019.5138.202619

How to cite this article:
Jena P K, Kishore J, Shukla S, Mohapatra D, Patel A. Collider Bias and Infectious Disease Studies: A Review. J Commun Dis. 2026;58(1):145-152.

References

Holmberg MJ, Andersen LW. Collider Bias. JAMA. 2022 Apr 5;327(13):1282–3. doi:10.1001/jama.2022.1820.

Mansournia MA, Nazemipour M, Etminan M. Interaction Contrasts and Collider Bias. Am J Epidemiol. 2022 Sep

;191(10):1813–9. doi:10.1093/aje/kwac103.

Tattan-Birch H, Marsden J, West R, Gage SH. Assessing and addressing collider bias in addiction research: the

curious case of smoking and COVID-19. Addiction. 2021 May;116(5):982–4. doi:10.1111/add.15348.

Griffith GJ, Morris TT, Tudball MJ, Herbert A, Mancano G, Pike L, Sharp GC, Sterne J, Palmer TM, Davey Smith

G, Tilling K, Zuccolo L, Davies NM, Hemani G. Collider bias undermines our understanding of COVID-19

disease risk and severity. Nat Commun. 2020 Nov 12;11(1):5749. doi:10.1038/s41467-020-19478-2

Westreich D. Berkson’s bias, selection bias, and missing data. Epidemiology. 2012 Jan;23(1):159–64.

doi:10.1097/EDE.0b013e31823b6296