Is a study with increasingly high participant dropout rates at every interval an example of attrition bias, self-selection bias, observer effect, or reporting bias?

Attrition Bias is the Correct Classification for Increasingly High Participant Dropout Rates in a Study

Increasingly high participant dropout rates at every interval in a study is a clear example of attrition bias. 1 This type of bias occurs when systematic differences exist between study groups in the number and way participants are lost from a study.

Understanding Attrition Bias

Attrition bias is a form of selection bias that occurs when participants drop out of a study over time, potentially compromising the study's internal validity. According to the MELODEM guidelines, this is particularly problematic because:

• Dropout rates that increase at each interval create systematic differences between those who remain in the study and those who leave 1
• The bias is not necessarily toward the null and can sometimes reverse the direction of association, making harmful exposures appear protective or protective exposures appear harmful 1

Why This Is Attrition Bias (Not the Other Types)

Distinguishing from Other Biases:

1. Self-selection bias: This occurs primarily at enrollment when participants decide whether to join a study based on certain characteristics. While dropout may involve self-selection, the specific pattern of increasing dropouts over time is classified as attrition bias 1

2. Observer effect: This refers to changes in participant behavior due to being observed, not to their dropout from the study

3. Reporting bias: This involves selective revealing or suppression of information by participants or researchers, not participant dropout 1

Impact of Attrition Bias on Study Results

Attrition bias can significantly distort study findings:

• It can lead to overestimation or underestimation of treatment effects 2
• Estimates of means become biased even at low attrition rates 3
• Associations between variables become biased when attrition depends on both baseline and follow-up variables 3
• The probability of dropping out is often higher for those not achieving desired benefits, systematically eliminating treatment failures from analyses 2

Factors Associated with Attrition

Several factors predict higher dropout rates:

• Lower educational level 3
• Higher tobacco consumption 4
• Diagnosed cardiovascular disease 4
• Higher anxiety and depression scores 4
• Difficulty using technology in technology-based interventions 1
• Finding data entry too burdensome 1

Methods to Address Attrition Bias

To minimize the impact of attrition bias:

• Over-recruitment at the beginning of the study 5
• Use of sampling weights and tailored replenishment samples 5
• Sensitivity analyses to illuminate the robustness of findings 1
• Clear definition of the research question and population of interest 1

However, it's important to note that when data are missing not at random (MNAR), no method provides unbiased estimates at attrition rates of 25% or higher 6. In such cases, researchers should acknowledge this limitation.

Best Practices for Reporting

When evaluating studies with high attrition:

• Compare characteristics of completers versus dropouts 1
• Report attrition rates for both intervention and control groups separately 1
• Conduct analyses to determine if dropout is related to the intervention, outcome, or both 1
• Consider whether the reason for dropout is related to the intervention or outcome 1

Attrition bias remains one of the major methodological problems in longitudinal studies and requires careful consideration when interpreting study results, especially when dropout rates increase over time.