Skip to content
Grad Coach

Diagram · CC BY 4.0

Simpson’s paradox example: kidney stone treatments

Treatment A does better for small stones and for large stones, yet worse when the two are combined.

Simpson's paradox in a real medical study: success rates of two treatments for kidney stones (Charig et al. 1986, split by stone size as in Julious and Mullee 1994). Treatment A is open surgery (orange); treatment B is percutaneous nephrolithotomy (gray). Bars run from 0% and are labeled with the success rate, with successful treatments / patients treated under each bar: Small stones: A 81/87 = 93%, B 234/270 = 87%. A is higher; Large stones: A 192/263 = 73%, B 55/80 = 69%. A is higher; All patients: A 273/350 = 78%, B 289/350 = 83%. B is higher.

Swipe sideways to see the whole diagram, or download the PNG.

Simpson’s paradox in kidney stone treatment data. Derek Jansen, Grad Coach, CC BY 4.0.

What it shows

Simpson’s paradox in data from a real medical study: success rates of two kidney stone treatments (Charig and colleagues, 1986), split by stone size as in Julious and Mullee (1994). Treatment A is open surgery; treatment B is percutaneous nephrolithotomy.

Treatment A has the higher success rate for small stones and for large stones, yet the lower rate when the two groups are combined. The reversal arises because A was given mostly for large stones, which are harder to treat.

How to read it

  1. Compare the bars within each stone size first: A is higher for small stones (93% against 87%) and for large stones (73% against 69%).

  2. Then look at the combined bars: B is higher, 83% against 78%.

  3. Check the counts under the bars: 263 of A’s 350 patients had large stones, against 80 of B’s.

  4. Stone size is the confounder: it affected both the choice of treatment and the chance of success.

The data

  • Small stones: A 81/87 = 93%, B 234/270 = 87%.
  • Large stones: A 192/263 = 73%, B 55/80 = 69%.
  • All patients: A 273/350 = 78%, B 289/350 = 83%.
  • The totals are those Charig and colleagues (1986) report; the counts by stone size follow Julious and Mullee’s (1994) split and sum to the published totals.

Ways to use it in teaching

  • Show only the combined bars first and ask which treatment students would choose, then reveal the split.

  • Ask what random assignment would have done to the link between stone size and treatment.

  • Set it beside the correlation vs causation diagram: stone size plays the confounder’s role.

Sources

  • Charig, C. R., Webb, D. R., Payne, S. R., & Wickham, J. E. (1986). Comparison of treatment of renal calculi by open surgery, percutaneous nephrolithotomy, and extracorporeal shockwave lithotripsy. BMJ, 292(6524), 879–882. https://doi.org/10.1136/bmj.292.6524.879
  • Dunn, P. K. (2021). Scientific research and methodology. https://bookdown.org/pkaldunn/Book/
  • Julious, S. A., & Mullee, M. A. (1994). Confounding and Simpson’s paradox. BMJ, 309(6967), 1480–1481. https://doi.org/10.1136/bmj.309.6967.1480
  • Simpson's paradox. (n.d.). In Wikipedia. Retrieved October 18, 2026, from https://en.wikipedia.org/wiki/Simpson%27s_paradox

Inference and evidence

Related diagrams

All open educational resources

Need something else?

Tell us which diagram, template or worked examples would help your teaching, and we’ll do our best to add it to the collection.