A method for estimating bias introduced into epidemiological investigations by those who refuse to participate.

Several methods have been suggested for estimating the error introduced into epidemiological investigations by potential participants who decline participation. As these could not be applied to our present epidemiological case-control breast cancer study, a new and simple method was developed based on information gathered from those controls who participated only relunctantly. A formula is given for calculating an adjusted mean for the sample and corresponding p-value. Our own data were used to illustrate the method. The calculations then indicated that those who declined participation were slightly selected in some aspects, but not to such a degree as to obscure or cause any significant differences between patient and control groups.