Algorithms

[S@M]

I am looking to scope the international perspective on taking systematic reviews of clinical effectiveness, or evidence based guidelines and developing algorithms, or decision trees from them. Is this currently being done anywhere, or is there any discussion/debate that you are aware of? The particular focus of my interest is in taking complex documents like systematic reviews and developing simplified pathways to facilitate implementation, or informed decision making. In doing so, I fully intend that the algorithm/tree/flow chart link back to the primary source. There may be some additional issues/concerns/questions the list may like to raise. I would be interested to hear  a range of viewpoints.

I am starting a review of international evidence-based organisations web sites, and a Cochrane/Medline search today but am keen to hear from others who may have been down this path before. Obviously, my area of interest is health care, but if your in sciences etc and use algorithms, then I'd be interested in hearing about that as well.

cheers and thanks

[Dave]

That's a mouthfull

[mjrpes]

That's a mouthfull

"developing simplified pathways to facilitate implementation"

ahhhhh... the academia-speak is burning the roof of my mouth!

[Dave]

True.. it's summer vacation. I'm only supposed to do that kind of thing 9 months out of the year

[[xeno]Julios]

i've asked my dad about this, so waiting for a reply.

btw i'm not clear on what exactly this algorithm is supposed to do.

Surely it can't parse through all the literature reviews and come to an informed decision based on this.

Don't humans (for now) have to do this work and analyze the data and interpretations, and then make a decision based on this?

[S@M]

Lols Dave, mjrpes - sorry for the buccal cavity damage!

Julios, people do need to do the work (thats my job) its not about having a computer read text and develop the algorithm (although that will come with time). Its about providing the core ideas of a long, complex document in a simplified form to guide clinical decision making. The trend is to develop care pathways etc from guidelines, but I want to go back a step or two and look at methods potentially outside the health domain; and systematic reviews are substantially different in content and structure to guidelines, so the standard approach would be a poor fit (I think).

[[xeno]Julios]

I see.

So the algorithm would be highly tailored for a specific domain/specific forms of research within that domain.

But there may be more general structural algorithms that can then be refined for a given task right?

I'm guessing there would be a hierarchy of values, each with a weight that is assigned based on either user-input, or deeper criteria.

And then based on the place within the hierarchy (or it may be another structure instead of a hierarchy), the values are processed.

And then you can define in advance what output options you want...

can you give me a concrete example, coz i'm startin to flounder in the abstract...

[S@M]

as an example: split thickness skin graft donor site post skin harvest management (the bit of the body they take skin from to put somewhere else)

Core issue 1 - size, location of wound, degree of exudate/risk of haemorrhage. Pressure dressing is applied perioperatively (but usually leaks within 8 hours). A systematic review recommends:

Remove pressure dressing @ 48 hours
Monitor secondary dressing daily
Remove secondary dressing when wound healed

however, this is complicated by excessive exudate/haemorrhage; wound colonisation; clinical infection etc.
These complications need to be factored in to an algorithm to guide decision making. So the primary decision "tree" has side branches that need to feed in to the process. (is this better??)

Im not sure at this stage what to call them though, decision trees, care maps, pathways or algorithms. Nor am I set on a particular method.

okay, websearch for clinical algorithms found this link which helps illustrate. Although I dont want to get locked in to a medical/health care model/approach when industry or other areas might have useful methods also

http://www.aafp.org/afp/980700ap/legler.html#al19

[[xeno]Julios]

this is reply from my dad:

I think the closest would be the Cochrane collaborations. Also look at the the UK national institute for clinical excellence (NICE). There is also some literature in the priority setting field that may relate to this. Also look at the WHO website for a recently launched program called Health Metrics Network.

[S@M]

ahhh WHO - had not considered them. Thanks Jules, I'm quite familiar with Cochrane and NICE etc, will add the Health Metrics Network to my list.

thanks again