We included seven cohort studies with high quality, which minimized the likelihood of recall interviewer and selection biases that can be a concern in epidemiological studies.
VZ supervised all aspects of the field work, including translation/back translation, training and monitoring of interviewers, and selection of population controls.
In the first phase, interviewers selected one to two eligible persons per household by doorstep screening using a computer-generated probability selection method.
Instead of introducing a fixed effect for each interviewer into the selection equation, we assume that interviewer effectiveness is a random effect.
Since they were not RCTs (prospective cohort study [ 23], retrospective cohort study [ 24]), recall bias, interviewer bias, and selection bias were always a concern.
Interviews are lousy selection mechanisms because interviewers are so biased and make decisions on the basis of first impressions: far better to give candidates a task similar to the one they will have to perform and see how well they do.
Interviewers received training on household selection, interviewing techniques, and the survey tool, and were provided with a survey guide.
We enrolled persons closest in age as controls, not to control confounding by age, but rather to provide simple standardized guidelines to the interviewers for control selection.
As to selection of interviewers, interviewing is a skill in itself, and it should not be assumed that health personnel such as nurses, medical students or, especially, doctors are the best qualified to do it.
Following the household screening and selection procedures, interviewers visited each selected woman at her home to administer the survey.
