Comparing different methods to detect and correct nonresponse bias in postal questionnaire studies

Area of Science:

• Dental Public Health
• Epidemiology
• Survey Methodology

Background:

• Postal questionnaires are common in health research, but nonresponse can introduce bias.
• Estimating population parameters accurately requires addressing potential biases from nonrespondents.
• Previous methods for bias correction may not fully account for the characteristics of nonresponders.

Purpose of the Study:

• To compare different methods for detecting and correcting nonresponse bias in postal questionnaire data.
• To evaluate the effectiveness of using mailing stage and socioeconomic status as predictors of nonresponse.
• To determine the most accurate method for estimating population means in the presence of nonresponse.

Main Methods:

• A clinical survey of 5-year-old children collected data on decayed, missing, and filled teeth (dmft).
• Postal questionnaires were sent in three stages to target nonresponders, with data available for all children.
• Estimates of population dmft were generated using methodologies comparing dmft with response stage and area socioeconomic status measures.

Main Results:

• A response rate of 80.9% was achieved, but nonrespondents had a significantly higher mean dmft (2.41) compared to responders (1.27).
• Significant linear trends in dmft and socioeconomic status were observed across mailing stages.
• The regression method using mailing stage provided the most accurate adjusted dmft estimate (1.42), outperforming socioeconomic status-based methods (1.31-1.32).

Conclusions:

• Nonresponse bias can impact study findings even with acceptable response rates (e.g., 80.9%).
• Researchers should implement and report analyses to detect and correct for nonresponse bias.
• The mailing stage of response is a valuable covariate for adjusting estimates in postal survey research.