Tonight, some health IT, with contributions from our very own Matt Wills, GA.

El-Gayar, Deokar, Wills (2008), “Current Issues and Future Trends of Clinical Decision Support Systems (CDSS).” written for an encyclopedia…

Matt says the paper was written mainly to map the field and identify gaps.

Serious issues in safety and access that CDSS can address! Scarce resources, spiraling costs — these are all issues that drive better CDSS. Matt mentions the 1999 Institute of Medicine report that found 48K-98K die each year as a result of medical error. The financial cost of those errors: $6B per year.

ULAM: a liver allocation model! Ew!

Dang! Matt says there’s a lack of research that says CDSS resuls in clear clinical benefits. The research isn’t saying CDSS produces no results; the problem is develping clear metrics and baselines and generalizing results. The existing research is mostly anecdotal.

One hindrance to CDSS research: clinician throughput! They don’t want participation in an experiment with some new tech tht would cut down the number of patients they could process each day.

Another problem: physician autonomy. They don’t want a machine telling them what to do. So maybe there’s an opening for CDSS in medical school first, as a pedagogical tool. Test it out in the low-risk area, with interns who still have doctors supervising them, with students who are confronting simpler problems with clearer answers. Apply the CDSS to simulations, benchmark with established student norms and medical knowledge.

Physicians need transparency in the CDSS! They want to know how the diagnosis or recommendation was arrived at.

Matt says we need better Natural Language Processing (NLP) technology, especially to read all the “unstructured text” that physicians generate — just another example of how this is a very young field with a lot remaining to be done.

Matt would like to see more research on collaboration in the clinical setting. Lots of interesting telemedicine issues to be studied here — Matt finds a gap in the lit there!