Prevalence estimates for COVID-19-related health behaviors based on the cheating detection triangular model

Area of Science:

• Public Health
• Survey Methodology
• Behavioral Science

Background:

• Conventional direct questioning (DQ) surveys struggle with sensitive topics like COVID-19 behaviors due to social desirability bias (SDB), leading to inaccurate data.
• SDB can cause nonresponse and untruthful answers, threatening the validity of prevalence estimates for unknown behaviors.
• Indirect questioning techniques (IQTs) can mitigate SDB by ensuring response confidentiality.

Purpose of the Study:

• To assess the validity of the Cheating Detection Triangular Model (CDTRM), a novel IQT, for estimating COVID-19-related health behaviors.
• To evaluate the CDTRM's effectiveness in accounting for participants who disregard survey instructions ('cheaters').

Main Methods:

• An online survey of 1,714 participants in Taiwan was conducted.
• CDTRM prevalence estimates were generated using an Expectation-Maximization algorithm for three COVID-19 behaviors of varying sensitivity.
• CDTRM estimates were compared against DQ estimates and official statistics from the Taiwan Centers for Disease Control.

Main Results:

• For low-sensitivity behaviors, CDTRM and DQ estimates aligned with official statistics.
• For medium and high-sensitivity behaviors, CDTRM estimates were higher and presumed more valid than DQ estimates.
• The estimated cheating rate increased with the sensitivity of the health behavior.

Conclusions:

• The CDTRM effectively controlled for SDB in surveys on COVID-19 health behaviors.
• The CDTRM shows promise for improving estimation validity compared to DQ for sensitive health behaviors and attributes.
• This IQT is a valuable tool for obtaining more accurate prevalence data when SDB is a concern.