By Laurie Gelb
Can decision support consist of more than threats, promises and stiff upper
Here's where domains, measures and thresholds come in.
Here's where the rubber hits the road.
In one study, sufferers, clinicians and payors were asked how they would
measure the value of a drug for a condition for which disease- modifying
therapy did not yet exist. The same methodology works whether options are
plentiful, mediocre, whatever. But in this case--
Physicians highlighted clinical results in one or more domains, all of which
have a demonstrable impact on quality of life.
Patients focused on being able to experience things they have not been able
to experience recently.
Payors wanted to see statistically significant differences from placebo on
some objective measures, not really caring which -- the FDA's job.
Every stakeholder was able to specify domains (pain being one, just so we're
clear on what a domain is) that were relevant to him, and whether or not an
improvement in that particular domain would in itself justify
prescribing/taking/reimbursing therapy. Obviously, not all domains were
salient to every stakeholder.
Every stakeholder was able to specify how improvements in salient domains
would be measured (numerically and/or categorically) as well as her
threshold for that improvement -- what number or value or outcome would
constitute sufficient reason to act.
But the answers were different for everyone. (So were the questions, of
course -- computer-assisted interviewing uses previous answers to frame
So, when you're doing stakeholder research, instead of dragging out a stack
(real or virtual) of static scenario cards for tradeoff analysis and
sorting, instead of asking about abstractions like preference and
satisfaction that aren't used in real life, what if you asked about:
Domains that are salient
Measures that are used to measure change or value in those domains
Thresholds applied to those measures to justify action
Bear in mind, these are studies that run (very) low five figures and a few
weeks, all told -- this is a framework for frequent studies, not once a
decade. So you can track how the findings change as the environment does.
What next? You might design decision support that makes very clear...
What domain(s) are affected by the intervention you recommend or wish
considered What measures show change when the intervention is used, in whom,
and how frequently. how predictably To what extent any particular threshold
of change can be predicted, guaranteed or even hoped for
Of course, you update this as the data come in and time goes on.
Presto! User-centric decision support can be yours.
And it can be theirs.
If you do all this on the Web, kiosk or CD-ROM, you can develop a "wizard"
that enables the user to "buy in" to their choice using their own criteria.
The decision that's owned is the result that's achieved.
We're not just talking about justifying or avoiding therapy -- this is about
staff/physician recruitment/retention, open enrollment and a thousand other
Effective decision support reduces and supports the burden of choice.
How is yours doing?
Any stories to share?