Two recent posts here and there had the EMRandHIPAA.com site all ablaze for a few days with comments, pro and con, about clinical decision support. Neil Versel popped out a rather racy post, titled IBM’s Watson Addresses Errors of Diagnosis. In it, he describes IBM Healthcare’s attempt to move more into the EMR field with a clinical decision support software cleverly named “Watson”. This, of course, received quite a verbose reception. Now, I suppose I wouldn’t be a good blogger if I didn’t jump on the chance to put my, albeit very late, two or three cents in.
1. Drug-drug interactions, allergies. Right now, there are a ton of these potential interactions that pop up during office visits. During the course of doctors’ busy days, there is often little time for the delays that come with taking these popup annoyances too seriously. There are simply too many. For example, it sometimes happens that when I go to prescribe or refill insulin in a patient with 10 or more medications, that I get a popup box of checkboxes and drop-down menus asking me if I really want give insulin with drug #1, drug #2, drug #3, drug #4, drug #5, drug #6, drug #7, etc. Luckily there is a checkbox for “Do not ask me about this interaction in the future”. Each override requires a check box and a drop-down selection of why I wish to prescribe it anyway. Was it because I will carefully followup the patient’s condition? Was it because the patient has tolerated the combination before? Was it… etc, etc. How will a computer be able to call some drug-drug interaction not important? There’s really not a logic rule that allows me as a human to blow on by most of those interactions and check off “not a problem”. I wonder how Watson would handle that.
2. Example of tetracycline contraindication in pregnancy. So IBM gives an example of a contraindicated drug in pregnancy, something that medical students are trained to pick out on a multiple-choice exam as a no-no, and then somehow Watson is supposed to contribute to the easy stuff? Why not just display the pregnancy class letter (i.e., A, B, C, etc) next to a drug you’re about the prescribe? Isn’t that much cheaper than investing in a supercomputer? I think they’ll need to give a better example to convince buyers of the value here. Perhaps it would be more convincing if they used an example of medical knowledge that only a seasoned MD would know after all of their training, residency, fellowship steps are completed.
3. Will the input be optional for doctors or required by a given EMR system? This one pretty much supercedes my previous two considerations above since, if I could turn on and off the CDS function as I saw fit, then there would be less of a chance for it to get in my way and slow me down. I would probably only turn it on if I was really stumped and wanted some help. Since this usually only happens when I find myself in a rare situation that I have little experience with (maybe once or so a week), I think this would be the most logical time to use CDS. Then, I would genuinely welcome the chance to be helped by a CDS system.
Dr. West is an endocrinologist in private practice in Washington, DC. He completed fellowship training in Endocrinology and Metabolism at the Johns Hopkins University School of Medicine. Dr. West opened The Washington Endocrine Clinic, PLLC, as a solo practice in 2009. He can be reached at email@example.com.