Table 2 Logistic regression and generalized estimating equation models

Model 1

Participation of physician: yes/no

Physician characteristics: Location of hospital/practice (urban/rural), size of catchment area (number of community codes the responding patients of each dermatologist belong to), median age of all patients assigned to the physician, median tumor stage of all patients assigned to the physician

Model 2

Participation of patient: yes/no

Patient characteristics: hospital-based physician: yes/no, age, sex, UICC (stage), place of residence (urban/rural)

Model 3

Missing values in study questionnaire: yes/no

Patient characteristics: Age, sex, early−/late-responder, time since diagnosis, social stratum, place of residence (urban/rural), recruited by hospital/office-based physician

Physician characteristics: Hospital/practice, location of hospital/practice (urban/rural), size of catchment area (number of community codes the responding patients of each dermatologist belong to), median age of all patients assigned to the physician, median tumor stage of all patients assigned to the physician, recruitment proportion, number of contacts before participation