Selection bias in the study of spouse similarity for psychiatric morbidity in clinical samples

The existing literature on spouse similarity for health variables can be divided into two fields of application, high blood pressure and other coronary risk factors and psychiatric morbidity. In the former field, almost all studies have been conducted on general-population samples, whereas in the psychiatric field, most of the investigations studied clinical samples, i.e., where one or both members of the couple were under psychiatric treatment. In this report, we demonstrate that studies of spouse similarity for psychiatric morbidity are subject to selection bias whenever they are conducted on clinical samples. This selection bias has a double origin. First, some bias arises from the fact that the probability of a couple being drawn for a clinical sample is dependent on whether one or both spouses are afflicted with the illness in question. A second source of bias lies in the likelihood that each spouse's help-seeking behavior is not independent of the other's. It would appear then that, owing to this selection bias, it is preferable to restrict the study of spouse similarity for psychiatric morbidity to general-population samples.