Steven Stellman - Smoking prevalence in neighborhood and hospital controls: implications for hospital-based case-control studies

Version 1

      Publication Details (including relevant citation   information):

      Morabia, A., Stellman, S. D., Wynder, E. L. 49 (8)   885-9-

      Abstract: It is widely believed that the   prevalence of smoking among hospital patients is greater than   that of the general population because many conditions for which   patients are hospitalized are caused by or associated with   smoking, and that this increased prevalence may bias results of   case-control studies of tobacco-related diseases. For this   reason, many authors have suggested excluding from the control   series patients hospitalized for tobacco-related illnesses. The   present study investigated potential selection bias for hospital   compared to neighborhood controls in studying tobacco-related   diseases. The 709 cases from six U.S. cities had tobacco-related   cancers or myocardial infarction. They were individually matched   to one hospital control and to one neighbor. After excluding   patients with tobacco-related diseases, hospital controls were   less often current smokers and more often former smokers than   neighborhood controls. Among male ever smokers, hospital controls   tended to smoke more cigarettes per day than neighborhood   controls. Compared with the U.S. population, there was an   overrepresentation of smokers in neighborhood controls rather   than an under-representation of smokers in hospital controls.   Relative risk estimates varied according to type of control.   Choosing between hospital and neighborhood controls in   case-control studies should be dictated by criteria related to   the study hypothesis, participation, or cost. In particular,   exclusion of hospital controls with diseases known to be   tobacco-related seems to be a successful strategy for reducing   selection bias.

      Address (URL): http://www.ncbi.nlm.nih.gov/pubmed/8699208