Environment or…? Establishing Causality

Abstract

The identification of environmental hazards to health is rarely, if ever, a process of experimentation, but one of observation. However, it is insufficient merely to conduct a suitable epidemiological study and obtain a statistically significant association to prove a causal relationship between an environmental hazard and a disease. This issue arises in all areas where statistics based on observations are analysed. A close association was once measured between the salaries of Presbyterian ministers in Massachusetts and the price of rum in Havana1. No causal association has been claimed for this association, although it is based on accurate data, correctly analysed, and is statistically significant. This extreme example illustrates that evidence of association is a long way from proof of causality – rarely, however, will the association be so obviously spurious. Many suggested associations between variables measured over a long period are likely to be due to the effects of general economic progress independently on both variables. Thus, an association between the decline in excess winter mortality in England and Wales and the number of houses with central heating may not be causal2 – economic progress may have contributed to the ability of households to install central heating but may also have affected society’s and individuals’ ability to overcome the health effects of winter in several other ways.