Stang · European journal of epidemiology 2022 · Historical data re-analysis and methodological review · n=?

A twenty-first century perspective on concepts of modern epidemiology in Ignaz Philipp Semmelweis' work on puerperal sepsis.

Cited 10 times in the scientific literature.

Level 5 - mechanism / opinion, no new human data

Historical methodological review and quantitative bias re-analysis of historical observational data (Level 5 by design analogy).

PubMed 35486338 · doi:10.1007/s10654-022-00871-8 · record verified 2026-08-28

What was done

The authors reviewed Ignaz Semmelweis's complete published work and data on puerperal sepsis mortality (as compiled by von Györy in 1905) through the lens of modern epidemiological methods. They assessed mortality definitions, identification of bias sources, and modern causal concepts, and conducted quantitative bias analyses addressing selection bias from loss to follow-up and information bias from outcome measurement error.

What was found

The abstract reports no numerical values or effect sizes. Methodologically, quantitative bias analysis indicated that differential loss to follow-up is an unlikely explanation for Semmelweis's findings, and outcome misclassification would only matter if it varied across time periods. Confounding by health status could not be quantitatively addressed. Semmelweis recognized cause-specific mortality principles, applied potential outcome reasoning to estimate preventable deaths, and refuted alternative causal theories, although his hypothesis that clinic overcrowding was a risk factor was incorrect.

Why it matters

This paper demonstrates that foundational concepts of modern epidemiology—such as counterfactual reasoning and bias assessment—were inherently present in early clinical observation, methodologically validating the robustness of Semmelweis's conclusions on antiseptic handwashing.

Limits

The abstract provides no sample size or numerical metrics. The evaluation relies on historical records with potential unmeasured confounding by patient health status that could not be quantitatively adjusted.

Cited by