From estimand selection to adaptive and robust estimation for causal inference: applications in clinical epidemiology