The prevalence of coffee drinking among hospitalized and population-based control groups.

Data on coffee-drinking habits obtained from a case-control study conducted in Detroit were used to compare the proportions of coffee drinkers in a hospital and a population control series. The comparison was based on interviews with 262 hospitalized controls and 427 population controls. The overall proportion of coffee drinkers in the total hospital control group was similar to that in the population control group. However, the proportion of moderate-to-heavy coffee drinkers among controls hospitalized for conditions that may have caused them to alter their diet (eg, gastrointestinal disorders and cardiovascular disease) was lower than that among population controls. In contrast, the proportion of moderate-to-heavy coffee drinkers among controls hospitalized for conditions that probably did not cause a change in diet (eg, fractures) was almost identical to that among population controls. These results suggest that, in hospital-based case-control studies of the effects of coffee consumption, it would be prudent to restrict the referent group to those patients hospitalized for conditions that probably did not cause a change in diet. The magnitude of bias resulting from failure to exclude controls hospitalized for diet-altering conditions will depend on two factors that may vary between studies: (1) the distribution of diet-altering conditions among the hospital controls, and (2) the relationship of these diseases to coffee consumption.