We are all choice architects, influencing our patients by describing options in different contexts. This is devilishly difficult given the time constraints created by our present payment structure, the imperfect nature of the data we have at our disposal, the varying preferences and educational backgrounds of our patients, and our own biases hidden from our ability to appreciate their influences. This past week there was yet another major study published in the NEJM regarding the effectiveness of screening mammography on prevention of breast cancer. It was another case of looking at the glass and saying whether it was partially full or partially empty.
My sentiments on this align with Gilbert Welch who wrote an accompanying editorial. The basic take home message was that in order to save one life the authors found they needed to screen 2500 fifty year old women for 10 years in order to save one life. These efforts would generate 1000 false positive tests and likely expose 15-20 women falsely diagnosed with breast cancer to various forms of treatment including surgery and chemotherapy. Within a cohort of 1000 women, on average 4.5 would die of breast cancer in 10 years and this would drop to 4 with deployment of mammography.
My question is, now that this and other studies supporting similar numbers have been completed and published, how many women will be given this data as part of their routine discussions to have mammography? I suspect next to none.
Medicine is filled with these types of issues. Basically healthy people with few if any symptoms come to us for some sort of service. They want reassurance. We have become increasingly equipped to offer an array of services to prevent some sort of undesirable future outcome. Unfortunately we as physicians don't understand the numbers particularly well and we are not encouraged to present them in an understandable and transparent format to our patients.
In Dr. Welch's piece that accompanies the NEJM article, he raises the issue that physicians are being judged on quality metrics that include participation of their patients in various activities such as screening mammography. This gives clinicians a vested interest in getting patients to comply with various, purportedly beneficial activities. However, just how beneficial are these activities and what information should be shared with patients before they commit to any given course of action?
I came across an article by D Misselbrook and D Armstrong. 276. British Journal of General Practice, April 2001 where they looked at patient responses to the following scenarios:
The results are quite startling.
Depending on the data presented and the framing, anywhere from two thirds to less than twenty percent definitely would agree to treatment. Those who likely would decline went from about twenty percent to well over half.
What are we to do with such information? What is the correct approach to informing patients? I have little doubt that using the NNT or PPB approaches will result in fewer patients electing for certain interventions. Is that reason enough not to share this information? I do not think so. In fact, it seems to me that the quality of the informed consent process could be directly assessed by the number of patients who actively decide against specific measures. Too high a level of compliance with certain measures might equate with evidence of an inadequate informed consent process and a lower level might be a good measure of the quality of the informed consent.