Figure 2

Relationship of administration format with exposure, outcome and other variables. A hypothetical scenario is represented by a causal diagram. Investigators may be interested in comparing health-related quality of life (HRQL) between HIV-infected patients with and without AIDS. Both interviewer and self-administration are available. Patients with the AIDS-defining illnesses cytomegalovirus (CMV) retinitis or brain toxoplasmosis are more likely to require interviewer administration because of visual or cognitive impairment, respectively. Administration format is not a confounder since it is on the causal pathway from exposure to outcome and does not cause CMV retinitis nor brain toxoplasmosis. The table shows three scenarios. In the first scenario the administration format is restricted to self-administration and the difference in HRQL is 20 units. In the second and third scenario, both interviewer and self-administration are available and it is assumed that patients with AIDS are more likely to require interviewer administration because of CMV retinitis or brain toxoplasmosis. The effect of interviewer-administration is ± 10 units in the second and third scenario, respectively, which has an effect of ± 2 units on the between-group comparisons.