Selection bias was reduced by recontacting nonparticipants.

摘要: Abstract Objective One of the main goals health examination surveys is to provide unbiased estimates indicators at population level. We demonstrate how multiple imputation methods may help reduce selection bias if partial data on some nonparticipants are collected. Study Design and Setting In FINRISK 2007 study, a population-based study conducted in Finland, random sample 10,000 men women aged 25–74 years were invited participate. The included questionnaire collection examination. A total 6,255 individuals participated study. Out 3,745 nonparticipants, 473 returned simplified after recontact. Both participants followed up for death hospitalizations. follow-up allowed check assumptions missing mechanism, tailored used handle data. Results Nonparticipation strong predictor mortality five-year follow-up. However, recontact response does not predict or morbidity among when adjusted age sex. result suggests that respondents can be as proxy all nonparticipants. comparison raw reveals clear differences estimated prevalences smoking heavy alcohol usage. Conclusion All efforts collect likely useful even rate remains low. Statistical analysis provides an indication extent bias, studies where available assumptions.