Let’s say that study researchers created two truly random groups of participants and blinding was successful (no one knew who was in which group).
But, by the end of the study, some data are missing or incomplete for many of the study participants. Some people skipped a few of their appointments or just disappeared completely.
That could lead to bias.
If you examine the participant data in a study and see a meaningful difference between the types of people who have missing data or drop out of a study and those who stayed in the study, there’s the possibility that the authors overestimated or underestimated the effect of the treatment because the people who did not do well stopped participating.
– In a study of 50 men and 50 women, 35 of the women stopped participating in the study.
– Half of the participants in the placebo group stopped showing up for their follow-up appointments.
Missing participant data is an unavoidable consequence of conducting research with humans. It doesn’t necessarily indicate a high risk of bias, but it can.
The risk of attrition bias for a study with missing participant data depends on:
1. How much information is missing
2. Who dropped out of the study