Threats to Validity in Epidemiological Studies
An essential feature of epidemiological studies is a comparison of two groups in terms of presence or absence of exposure or presence or absence of disease. For the measurements to be com-parable, the investigator should ensure absence of bias. Biases can be divided into three general types: 1) selection bias; 2) infor-mation or observation bias; and 3) confounding bias.
Selection bias can arise when the sampling procedure is influ-enced a priori by the disease or the exposure.
Another example, referred to as self-selection bias, occurs when subjects who have been exposed to an event are more likely to participate in a study if they have the disease or prodromal stages of the disease under study. A similar type of selection bias can occur when subjects are solicited from newspaper or other similar advertisements
In case–control studies, information bias occurs when the details about prior exposure are obtained in a noncomparable manner or are subject to poor recall. To minimize such bias, exposure data should be collected without knowledge of disease status. This procedure is known as blindness. However, because of selec-tive recall, when the sole source of information is the affected individual, this type of bias sometimes presents insurmountable problems.
Confounding bias results when a third factor that is a cause of the disease under study is also associated with the exposure. A con-founding factor is a cause of the disease under study independent of its association with the exposure