Reporting of attributable and relative risks, 1966–97

Evidence-based medicine has excited health-care professionals. It applies to clinical and public-health settings, which need different problem analyses. The goal of public health is to decrease or prevent diseases in the population. Relative risks (RR) and odds ratios (OR) estimate the strength of association between diseases and risk factors; whereas attributable risk (AR) estimates the amount of disease attributable to a certain factor. A risk factor may be strongly related to a disease, but may contribute less to the problem of that disease in the population if its prevalence is low (low AR). Conversely, a risk factor with a weak association may contribute more if its prevalence is high (high AR). Control of the latter factor would reduce burden of disease in the population more effectively than control of the former. AR can link causality with public health action. AR can be best derived from cohort studies; its estimation from case-control studies is also possible if controls represent the general population. Despite repeated advocacy of its importance for public health, AR has received less attention than RR. We counted the number of papers reporting RR with AR which appeared on MEDLINE from January, 1966, to January, 1998. The “Textword” procedure was used in searching for: “relative risk” or “odds ratio” for retrieving articles with RR or OR, and “attributable risk”, “attributable fraction”, or “aetiological fraction” for articles with AR. There was one report with AR for 31 with RR