Clinical decision making is a hot topic in medical informatics. Just how many decision can be made, correctly, by a computer? Is there potential that someday the physician will be replaced? Do they really speed up workflow?
There are two types of decisions dealt with in the beginning of the chapter: diagnostic and management. Diagnostic means which questions to ask, tests to order or procedures to perform. Management decisions are after the diagnosis is made, knowing when and how to treat. I think clinical decision support systems can be useful in both of these areas. Given a patient's chief complaint, it might be helpful to enter that into a system and get guidance to arrive at a diagnosis. Also, it can be helpful to use when, given a diagnosis, it could point you to what the treatment should be.
An important thing to keep in mind with clinical decision support is that there are always many variables that cannot be figured into whatever algorithm or artificial intelligence the system is using. That is why I believe physicians will not be replaced for quite a while by computers.
The chapter chronicles many of the systems developed to help make clinical decisions.
I like the Leeds Abdominal Pain System. It is useful for working up a patient to calculate the probability of seven possible explanations: appendicitis, diverticulitis, perforated ulcer, cholecytitis, small-bowel obstruction, pancreatitis and non-specific. This system was correct in 91.8% of cases, where clinicians were correct in 65-80%. It is limited, though in that it only considers 7 conditions. So it would be useful as an initial diagnostic help, but then if it proved to be incorrect one would need to use other methods to arrive at a diagnosis.
System developed is the HELP system. This is for generating alerts "when abnormalities in the patient record are noted." This is a great advantage to physicians, and there are many more systems like this in place. Medicine is getting so complicated, that all providers cannot keep up to date with all of the med interactions and contraindications. Systems like this provide a safety net for the patients. Most docs I've seen are very grateful for things like this. It saves people's lives.
The chapter also notes that after a diagnosis is reached there are many more factors to be analyzed in treatment, costs and benefits being one of them. This is another area where algorithms may not be able to help much as the physician brings a lifetime of experience in many areas to reach a conclusion. Many of the social issues in medicine cannot be addressed with decision support systems.
These systems can give advice in different ways, passively or actively. I've seen many physicians get frustrated with too many warnings that are not personalized to the situation. This is dangerous, and the alert fatigue generated callouses the user to the prompts. I also wrote a little last time about this subject. There is the consulting model verses the critiquing model. Obviously, most physicians would prefer to consult decision support systems, maybe only having critiquing systems in place for serious situations.
All in all, this chapter was an interesting read, and educated me about the different types of systems out there. I haven't had a whole lot of experience with them, but I will be much more willing to try them when situations arise.
No comments:
Post a Comment