Decision making may differ by subjects in a research study and real-world clinical care setting. Because we were conducting a secondary analysis of our existing dataset from this trial, we had not conducted inter- views of study participants to purposefully identify themes or factors not available in the care manage- ment records dataset. The potential for selection bias exists because primary care physicians reviewed their patient lists generated from administrative data prior to the mailing of study invitation materials and could exclude individuals from being invited to participate in the study. However, we believe this potential bias is small because the specified criteria for excluding names was that the patient did not have a diagnosis of dementia, and only a small proportion of all names generated from the administrative databases for review by these physicians are estimated to have been excluded from the mailing of study invitation materials.