Figure 2.

Heat map of evidence of association between lifestyle factors and diagnosis-specific sickness absence

Strength of association: rate ration (RR) lower than 1·1 (low), 1·1–1·49 and significant (moderate), and 1·5 or higher and significant (high). Consistency: I² values greater than 50% and significant (low), 25–50% (moderate), and lower than 25% (high). Robustness to serial adjustments and multiple testing: RR not robust to adjustments (low); robust to adjustments, but not to multiple testing (moderate); and robust to adjustments and multiple testing (high). Population attributable fractions (PAF) on the basis of exposure prevalence estimates obtained from European countries (PAF_external): greater than 10% (high), 5–10% and significant (moderate), and lower than 5% (low). Although causal associations can be strong and weak, strong multivariable-adjusted associations are less likely to be confounded than weak associations. For example, an RR of 1·3 between a single confounder and sickness absence could explain a weak 1·05 times increase in risk of sickness absence associated with the lifestyle factor; the corresponding RR required to explain a strong 1·5 times increased association between the lifestyle factor and sickness absence would be as high as 2·4. Details of PAF calculations are provided in the appendix (p 6). *Overall rating is indicated as: 0 (at least one low rating in strength of association, consistency, or robustness), + (high or moderate rating for strength of association, consistency and robustness, and moderate PAF); or ++ (high or moderate rating for strength of association, consistency and robustness, and high PAF). †Data available from Finnish Public Sector study²⁰ and Health and Social Support study.²¹ Not estimated=non-significant association or negative PAF.