We agree with Dr. Harewood that clinicians play an important role in the process of screening for colorectal cancer and that interventions directed primarily at the patient are probably not in themselves adequate to achieve high screening rates. In our randomized trial of a patient-directed decision aid, we found that the decision aid could increase rates of screening by providing patients with information about colorectal cancer screening and with a means of discussing that information with their physicians. We did not direct any systematic intervention toward the providers, nor did we implement changes in office systems that have been shown to increase screening rates, such as reminder systems (1). The rates of screening test completion we observed (37% for intervention patients vs. 23% for controls) could probably be increased by providing further support for providers and by improving office systems. We plan to test an integrated approach in future studies and are preparing a ew version of the decision aid that will also allow patients the option of screening colonoscopy.

Dr. Harewood is concerned that the patients we studied were more motivated than those who did not participate, creating a potential problem with the generalizability of our findings. Of the 1006 patients who did not participate, only 42 reported being uninterested and 236 reported "not having time to come in early." Many were simply unable to be contacted (n = 233), did not arrive early enough (n = 201), or were called back early for their appointments (n = 183). Although it is possible that patients who agreed to come in early and actually did so were more motivated than other patients, it is also possible that our selection of only patients who had not been recently screened actually created a more recalcitrant group. In either case, this selection process preceded randomization and thus may have influenced generalizability; however, it probably did not influence effect size. Only further testing of routine patients in everyday pract e can tell us the extent to which our findings are generalizable.

Our study design made blinding unfeasible because providing information to the provider about eligibility for screening through our chart marker was an integral part of the intervention. We hope our study will lead to further research that will shed light on how best to help primary care medical practice achieve optimal levels of screening.