Transitioning To Speech Recognition Editing: A Case Study
I came across a very interesting article published in
ADVANCE for HIM's sister publication, ADVANCE for Health Information
Executives. The article, entitled,
Breaking
the Productivity Glass Ceiling, is a description of one medical
facility's transition from traditional transcription to speech recognition (SR)
editing using Nuance's eScription "computer-aided medical transcription"
(CAMT) platform.
The story of Seattle Children's Hospital's transition to SR
editing is one that is quite familiar to me, having been involved in a number
of such endeavors personally and speaking with many other MTs and managers who
have done likewise. One of the
unfortunate shortcomings of this article is that it makes little mention of the
difficulties likely to be encountered by all parties involved when making this
kind of transition, other than a passing comment that "Some...MTs picked up
[SR editing] skills faster than others..."
There is no discussion, for instance, of the problem of dictators who,
for one reason or another, simply are not good SR candidates.
As a matter of fact, the authors of the article assert that
the percentage of dictators whose dictation is voice recognized "is now
stabilized at about 80 percent" and that "the other 20 percent can be
attributed to our residency program, in which we have providers who are new to
the system rotating in and out every few months." This leaves the impression that 100% of the
permanent physicians' dictations are being successfully recognized by the
system. If this is true, I suspect the
eScription recognition threshold for this facility has been set fairly low, as in
my experience with the eScription platform, I've never seen this level of
successful implementation, ever. I
certainly am willing to stand corrected, if there are any eScription users out
there who can provide evidence to the contrary, but until then I view this
claim with skepticism. The problem with
setting the recognition threshold too low in order to recognize all dictators,
in my experience, is that a significant percentage of the resulting SR drafts
are going to be so bad it will take longer to edit the report than to simply
type it from scratch.
Also of interest in the article, the authors report after
transitioning to SR editing in January 2008, Seattle Children's SR editors have
realized a 61% overall increase in productivity. This is not out of line with what I've seen
across the industry. The article's
authors also state that the hospital has been able to reduce outsourcing from
30% to 10%, with the 10% necessary only because of an increase in dictation
volume. The hospital has also been able
to eliminate chronic overtime for their MT department. This supports a contention I've been making
for years now, which is that a good SR platform will pay for itself by virtue
of increased productivity alone, without the need to reduce MT editors'
compensation.
The article does not mention how or if editors' rate of
compensation has been adjusted as part of the transition to SR editing. This consideration is probably of more
concern to MTs working on production, and I get the impression that the Seattle
Children's MTs are employees working for hourly wage. But for MT editors working on production
especially, the level of increased productivity is only half the story; it's
how their rate of compensation is adjusted that
makes all the difference. For instance,
in this particular situation, if editors worked on production and their rate of
compensation had been decreased by 50% (that is, cut in half), a common
occurrence across the industry, MTs would have realized a net loss
in total compensation because they aren't seeing a 100% (i.e., doubling) of
their productivity.
I don't mean to give the impression that I'm opposed to the
use of SR technology in the healthcare industry. On the contrary, I believe SRT can be a great
boon to MTs and enable them to be more productive while saving a lot of
physical wear and tear on the hands and wrists.
The technology isn't the problem.
My concern relates to how the technology is marketed, that is,
whether or not the people who write the checks have realistic expectations
going in, and how compensation is tied to the use of the technology. Skilled medical transcriptionists who make
the transition to SR editing should not be penalized with decreased
compensation in order to help pay for a technology platform that was sold with
unrealistic expectations.