The Challenge of Deploying Decision Support Systems (DSS) in Pathology
Posted Sep 06 2011 12:00am
This is a guest blog by Dr. Glenn Edwards. He is currently Medical Director, St John of God Pathology, Wembley , Western Australia. By way of disclosure, he was formerly CEO and Medical Director of Pacific Knowledge Systems .
Jim Harrison’s recent summary of pathologists’ visions about their view of the future from an IT perspective deserves close attention (see: Key Changes in AP and CP during the Next Five Years; Relevance of IT ). Many pathologists believe that decision support and an enhanced consultative role, including better interpretation of test results, are the keys to a sustainable future for pathology. I fully concur with this view. However there has been very little movement towards this goal, and in my view we are steadily heading backwards.
There are many reasons for our lack of progress, not least of which is the current crop of outdated and inflexible LIS technologies. Meanwhile, and as a result, much of the discussion of decision support systems (DSS) in clinical pathology (CP) - and in healthcare generally – centers on what we could or should be doing. Most of the hard work of detailing and refining our ideas, based on hands-on experience, is yet to be done.
I had the good fortune to work with a DSS (the LabWizard suite) designed specifically for pathologists. These tools are in routine operation in busy automated labs, mostly in Australia, enabling enhanced interpretative reporting of CP reports. We have learned many hard and valuable lessons from this experience.
: A rule-based DSS represents and reasons with expert knowledge. As such it will soon conflict with human experts doing their own reasoning to make sound clinical decisions. If the clinicians don’t believe they have control over the system, and don’t understand its origins, its content and its behaviour, then they will not trust it or use it. Where we have succeeded in pathology, this success has been founded directly on local pathologist control over the design and construction of the DSS. In contrast, if (for whatever reason) we can not secure pathologist buy-in, the project is dead in the water at the outset.
: Most systems are designed with a modest group of rules to cover common situations. This causes problems when a patient doesn’t neatly fit the predetermined criteria. This, of course, happens often. The conversation then usually centers on the Procrustean problem of how to make context-independent guidelines useful in a myriad of specific contexts. And this will always be a tough case to make for a busy resident who needs to make rapid, sensible and safe clinical decisions.
? The answer is “to do any more is too hard”. And therein lies the problem with conventional approaches to rules-based systems. When I see a pathologist really excited by his or her DSS, it is because of patient-specific contexting. A DSS that accurately identifies context, in sufficient detail to match the human expert, becomes a valued tool supporting the busy clinician or pathologist. This speaks directly to some major issues of design for rules-based systems, most of which are not capable of this level of patient-specificity. In the case of labs, it is the reason we have been so reliant on “canned comments” – which are so generic and bland as to be utterly useless.
In an upcoming set of guest blog notes, I will explore a number of these DSS issues in more detail. I hope this might stimulate some discussion around the big questions posed by Jim Harrison on the future of pathology.
Dr Glenn Edwards MBBS MD MAACB FRCPA FACHI
Medical Director, St John of God Pathology PO Box 646 Wembley WA 6913 Tel: (08) 9284 8174 Mob: 0414 793 784