Appendix Table 3.
E-value analyses for different risk ratio thresholds
E-values* for depression | E-values for suicidal ideation | |||||
---|---|---|---|---|---|---|
Reference risk ratio | 1.0 | 1.27 | 2.28† | 1.0 | 1.27 | 2.28 |
GEE | 2.61 (1.25) | 3.51 (1.98) | 6.82 (4.18) | 2.66 (1.00) | 3.58 (1.14) | 6.94 (3.05) |
Cox | 2.97 (1.50) | 3.96 (2.21) | 7.61 (4.56) | 3.68 (1.00) | 4.85 (1.66) | 9.17 (3.68) |
MSM | 4.19 (1.77) | 5.50 (2.51) | 10.33 (5.07) | 3.11 (1.00) | 4.13 (1.34) | 7.91 (3.24) |
GEE = generalized estimating equations logistic regression model; Cox = Cox proportional hazards regression model; MSM = marginal structural model
Please see Table 2 for the results of the three models. E-values are given in the following format: e-value for estimate (e-value for upper confidence interval). The E-value denotes the minimum strength of association, on the risk ratio scale, that would be required for an unobserved confounder to have with both the exposure and outcome, conditional on the measured covariates, in order to move an estimate to a certain risk ratio or the upper limit of a confidence interval to encapsulate this risk ratio.
Estimated adjusted hazard ratio for suicidal ideation or attempted/completed suicide in Mollan et al (2).