Regression calibration of self-reported mobile phone use to optimize quantitative risk estimation in the COSMOS study

Abstract

Overview

The Cohort Study of Mobile Phone Use and Health (COSMOS) involves extensive data collection on mobile phone use through self-reports and operator records. These data are pivotal in studying the health impacts associated with mobile phone usage.

Methodology

To address data limitations and potential biases in self-reports, COSMOS researchers evaluated various statistical approaches for constructing mobile phone exposure histories:

• Four regression calibration methods: simple, direct, inverse, and generalized additive model for location, shape, and scale
• Complete-case analysis
• Multiple imputation

The analyses were applied to data from Denmark, Finland, the Netherlands, Sweden, and the UK, collected between 2007 and 2012.

Findings

Comparison of methods revealed that regression calibration approaches were superior, showing less bias and lower mean squared errors relative to complete-case analysis or multiple imputation.

Conclusion

Integrating self-reported and operator-recorded data using regression calibration techniques significantly improves the accuracy of estimating the relationship between mobile phone use and health outcomes. The prospective design of COSMOS and enhanced exposure assessments are expected to yield more reliable conclusions about the health impacts of mobile phone use.