Evaluation of the effect of methodological assumptions on estimates of adherence to antipsychotics: a real-world data study