Question

Consider the following methods and results taken from an undisclosed source abstract (so that we don't give away the answer). However, EMS = Eosinophilia-myalgia syndrome and LT = L-tryptophan
METHODS:
A random sample of 813 physicians practising in the United States and Canada was obtained. Physicians were asked to provide diagnoses for 6 case vignettes having diverse resemblances to EMS. Six weeks later, participants were asked to provide diagnoses for a complementary series of cases described in identical text except for different data regarding LT use.
RESULTS:
Physicians who responded (N = 227, 28%) were more likely to diagnose EMS when LT exposure was present compared to the same case without LT use. In the most striking difference, EMS was diagnosed by 48% of physicians when the case was described in a man using LT, but by only 8% of physicians for the same case without LT use.

This is a clear cut example of:

Berkson's bias

Healthy worker bias

Recall bias

Detection or diagnostic bias

Answer 1

Answer to the question)
A brief idea about all four bias is as follows:

• Berksons bias is when the sample is chosen not from original population. But a sub population is formed and then the sample is selected from this subpopulation
• Healthy workers bias is: when the comparison being done of the treatment is not done correctly or the correction group is not suitable for the given scenario
• Recall bias : This occurs due to error in recalling all the responses correctly due to the research
• Detection bias: this bias occurs by the people who record the results or final outcomes
• .

In this scenario the true population is: All physicians

out of that 816 are selected

from these 816, 6 case vignettes are chosen from each of the physicians

Hence the sample is chosen from the subpopulation rather than the true population

Hence this bias is called: Berkson's bias