Assessment of selection bias due to dropouts in the follow-up of the Scania Public Health Cohort. Issue 4 (June 2021)

• Record Type: Journal Article
• Authors: Canivet, Catarina; Nilsson, Anton; Björk, Jonas; Moghaddassi, Mahnaz; Östergren, Per-Olof
• Is Part Of: Scandinavian journal of public health. Volume 49:Issue 4(2021)
• Page Start: 457
• Page End: 464
• Publication Date: 2021-06
• Subjects: Population-based study -- register data -- bias -- selection -- epidemiology -- follow-up studies -- cohort studies -- public health; Public health -- Periodicals; Social medicine -- Periodicals; 614
• DOI: 10.1177/1403494820919544
• Languages: English
• ISSNs: 1403-4948

Abstract

Aims: To investigate potential differences between participants and dropouts in the 2005 follow-up of the Scania Public Health Cohort Study regarding the prevalence of commonly studied health determinants and whether these factors had differential associations with three health outcomes: all-cause mortality and purchase of prescribed cardiovascular or psychotropic drugs during a 10-year follow-up period. Methods: The Scania Public Health Cohort was initiated in 1999/2000, with randomly invited participants aged 18–80 years from the general population (58% participation). Questionnaire data from 10, 462 participants and 2576 dropouts in the 2005 follow-up (80% participation) were linked to public registers on mortality and purchase of prescribed drugs. Results: Age, male gender, being born abroad, low educational level, low self-rated mental and general health and daily smoking were all related to dropping out. The 10-year mortality was higher among dropouts (13.4% versus 11.9%; age-adjusted hazard ratio 1.6, 95% confidence interval: 1.4–1.8). In 13 out of 18 analyses, similar associations between health determinants and outcomes were found across participants and dropouts. However, being born outside of Sweden was associated with higher risks for all three poor health outcomes among participants, but not so among dropouts. Conclusions: Despite selective participation at follow-up, there was little evidence of selection bias, insofar as estimated associations were generally similar across participants, dropouts and the whole cohort. This finding is important for the assessment of the validity of prospective findings from this cohort and similar ones, where the loss of individuals at consecutive follow-ups of exposure is non-negligible.