Predictors of follow-up and assessment of selection bias using inverse probability weighting

Abstract

Dropouts in cohort studies can introduce selection bias; however, inverse probability weighting (IPW) can evaluate this potential bias. Objectives: To assess: a) predictors of retention in a cohort, b) whether differential follow-up introduced selection bias. Design and Methods: Baseline assessment and follow-up were done using mailed questionnaires. We compared retained participants with drop-outs after two years. We used Cox regression to estimate the relative risk (RR) of hypertension for obesity. We used IPW to adjust for confounding and selection bias, thus evaluating whether differential follow-up affected RR estimates. Results: We recruited 9,907 participants from December 1999 to January 2002. The 2-year follow-up questionnaire was answered by 87% of them. Missing information at baseline, younger age, smoking, marital status, and overweight were associated with attrition. The univariate RR of hypertension for obesity was 6.4; 95% confidence interval (CI) 3.9–10.5). Adjustment for confounding using IPW attenuated the estimate (RR = 2.4, 95% CI: 1.1–5.3). Additional adjustment for selection bias did not modify it. Conclusion and Discussion: Follow-up through mailed questionnaires of a geographically disperse cohort in Spain is possible. Despite existing differences between retained and lost to follow-up participants, this does not necessarily have an important impact on the RR estimates.