Ascertainment bias arises when data for a study or analysis is collected (or surveyed, screened, or recorded) such that some members of the intended population are less likely to be included than others. The resulting study sample becomes biased, as it is systematically different from the intended population. Ascertainment bias is related to sampling and selection bias.
It can happen when there is more intense surveillance or screening for the outcome among exposed individuals than among unexposed individuals, or differential recording of the outcome.
Ascertainment bias can occur in screening, where take-up can be influenced by factors such as cultural differences. It can occur in case-control studies in the initial identification of cases and controls, which can be skewed by relevant exposures, leading to biased relationships.
In a clinical trial, if allocation concealment and blinding are lacking, then outcome ascertainment can be overtly influenced by knowledge of the allocation.
In a Cochrane systematic review “Helmets for preventing head and facial injuries in bicyclists”, the authors examined information from five case-control studies. These studies obtained information from cyclists who had crashed or fallen. Cases were cyclists suffering head, or facial injuries and controls were cyclists suffering other injuries. In this situation, ascertainment bias could occur if information on helmet wearing was obtained differently for cases and controls.
In describing how to avoid ascertainment bias, the authors write:“In order to study facial injuries, cases should be limited to serious injuries (lacerations and fractures) that would result in an emergency department visit whether or not a head injury was also present” People with minor facial injuries may be identified because bicyclists seek care for head injuries.” This example illustrates the notion of biased screening for the outcome leading to ascertainment bias.
A database analysis found that rates of Alzheimer’s disease were lower among higher users of statins; a letter commenting on this study suggested that some or all of this observed association may be due to ascertainment bias: Alzheimer’s diagnosis requires contact with a doctor and statins use might be associated with frequency of visiting the doctor.
We did not identify studies reporting the impact of ascertainment bias.
Appropriate inclusion of cases and controls within case-control studies is essential to avoid ascertainment bias. In cohort studies, knowledge of exposures of interest should be kept separate from screening, identification and recording of relevant outcomes of interest. In clinical trials, allocation concealment and blinding of all study staff are critical to maintaining a low risk of bias throughout the trial.