Attrition is the loss of study units from a sample. In practice, attrition occurs in studies or research when data are missing or study participants do not complete the research program.
What Causes Attrition?
There are several causes of attrition in a research setting, including lack of consent, inconsistent or missing data, dropping out of the study, inability to locate, refusal to participate, and absenteeism.
For example, many studies use questionnaires and surveys to obtain information from a target audience. However, if the participants don’t answer all questions or give enough information, attrition can happen, causing the study’s results to be biased or incomplete.
Before considering attrition bias, you must identify whether the types of attrition are systematic or random.
Systematic (Non-Randomized) Sample
Attrition bias occurs when participants drop out of a non-randomized study. Some researchers call this a systematic sample.
In some studies, participant groups are hand-selected (i.e., not randomly selected) because each participant has unique characteristics. These may be characteristics that the study measures or observes. Therefore, when participants drop out of a study uniquely designed for those participant groups, attrition bias occurs because there are differences in the starting and ending samples.
You’re conducting a study to compare literacy among children aged 7 to 12 in a private and public school. You have 15 students from the private school and 15 students from the public school complete a simple survey in class, asking how often they read, whether their parents read with them, and how many books they have at home.
When you review your data, you see that several students from the public school were absent on the day the survey was conducted in class (perhaps the school buses didn’t run that day or a bus broke down, so some students couldn’t attend). Additionally, you find that several students from both schools didn’t answer all the questions or only partially answered some questions. After the survey, teachers from both classes informed you that the language in the questions was too challenging for some students to understand.
As a result, you are missing data from both student groups, and the sample size from the public school is smaller than that from the private group, creating an attrition bias in your research results.
Attrition bias can still occur in a study or research program that uses randomized samples. The validity or possible attrition bias of a randomized sample study will greatly depend on your results.
You have a randomized sample of 25 participants and are using online interviews to collect information on how 25 participants use social media over a four-week period. You conduct a pre-interview before the four-week time frame, one interview per week during the study timeframe, and a final post-study interview.
Over the course of those six interviews, you lose 15 participants, leaving you with a sample of 10 when you started with 25. This will affect your results’ validity, because there’s a significant difference in your initial and final sample sizes. However, assuming there’s no systematic pattern among participants who dropped out versus those who completed the study, there’s probably little or no attrition bias.
After reviewing your results, you see that the 15 participants who dropped out of your study initially reported low-to-average social media use. However, after the first week, you note that the data from those 15 participants showed they had the highest social media use compared to the remaining 10 participants. Although the study was randomized, this is a systematic difference that creates attrition bias in your results.
Find this useful?
Subscribe to our newsletter and get writing tips from our editors straight to your inbox.
Over the course of the study, two participants dropped out. When reviewing the results, you don’t identify any patterns between the participants who dropped out and those who completed the study. As a result, you can assume there’s little to no bias in your work and that the results have strong validity.
There’s some debate as to what a good attrition rate is, but most scholars seem to agree that an attrition rate of less than 5% usually indicates no concern or low bias, while an attrition rate of more than 20% is a cause for concern and negatively affects your research results.
Using the above examples of social media use among 25 randomized participants, we can calculate a simple attrition rate using the following equation:
No. of Participants Lost in a Given Timeframe
x 100 = Attrition Rate
Total No. of Participants in the Same Timeframe
15 out of 25 starting participants were lost, therefore:
15 / 25 x 100 = 60% attrition rate
2 out of 25 starting participants were lost, therefore:
If you’re conducting a study or longitudinal research using non-randomized samples, there are several precautionary steps you can take:
● Create a project identity.
● Offer cash or other incentives to participants.
● Develop a tracking system for your participants to avoid losing contact.
● Keep follow-up interviews brief.
● Collect detailed contact information for all participants.
● Use emails, postcards, or telephone reminders to retain participants in the sample.
● Conduct a pilot study to identify and correct unforeseen issues or gaps in your study method.
Conducting research is a daunting task for veteran and beginner researchers. However, when it comes to reporting your results, there are experts here to help. Proofed will proofread your first 500 words for free to ensure error-free writing, so that your results can shine.