International Clinical and Medical Case Reports Journal (ISSN: 2832-5788) | Volume 5, Issue 11 | Case Report | Open Access
John Croucher*
Stephnie Hon1, John Croucher2*
1International College of Management Sydney, Sydney, Australia
2Macquarie Business School, Macquarie University, Sydney, Australia
*Correspondence to: John Croucher
Fulltext PDFSimpson’s paradox is a statistical phenomenon in which a trend that appears in several separate groups of data either disappears or reverses when the groups are combined. That is, the relationship between two variables can look one way within each subgroup but look completely different when all the data are aggregated. Although widely reported in epidemiology and other branches of health care, the paradox remains a major source of misinterpretation in medical research. This paper examines Simpson’s paradox and its consequences for analysing clinical outcomes and treatment effectiveness. It outlines the paradox’s mathematical basis and shows how confounding, selection bias, and inappropriate aggregation can produce misleading conclusions. If Simpson’s paradox is not recognised, it may lead to incorrect treatment recommendations and incorrect policy decisions. However, understanding the paradox helps identify hidden sources of bias and interpret statistical results within their clinical context. Accounting for Simpson’s paradox is essential for all healthcare researchers, as it emphasises the need to look for subgroup effects before drawing conclusions from aggregated medical data. Another source of confounding variables is spurious correlation, where two variables may seem related only because they are each related to another variable that may not have been considered. For both Simpson’s paradox and spurious correlation, case studies are provided.
Simpson’s paradox; Health care; Spurious correlation; Epidemiology; Confounding variables
Stephnie Hon, John Croucher. The Illusion of Association: Simpson’s Paradox in Clinical and Epidemiological Studies. Int Clinc Med Case Rep Jour. 2026;5(11):1-7.