Is voice recognition a high-tech con game?

notownbuffΤεχνίτη Νοημοσύνη και Ρομποτική

17 Νοε 2013 (πριν από 3 χρόνια και 8 μήνες)

63 εμφανίσεις

Diagnostic Imaging ,

PACS web


Is voice recognition a high
tech con game?

Costs go one way and benefits another as more work is dumped on radiologists and more
of the bottom line is shifted to them

by Ernest Camponovo, M.D., MBA

Dec 03, 2004

I'm not a fan of voice recognition transcription, so I read Eric Trefelner's column, "Voice
recognition misses a few beats" (May, page 104), with special interest. I'm also not naive
enough to believe individual radiologists can, or should, stand in the w
ay of this
remarkable technology.

Many of us went into radiology at least in part because we were comfortable with
technology and wanted to apply the latest innovations to advance the health of our
patients. But voice recognition could become just the lat
est high
tech method of
increasing our cost of doing business.

Trefelner touches on how radiologists should approach issues of cost awareness, a
concept of critical importance to all radiologists. What concerns me is the insidious,
incremental passing on
of costs to radiologists, while the benefits of voice recognition go
to hospitals or imaging center owners. We must be vigilant over costs even when they
appear in the guise of innovations necessary to maintain competitive advantage.

Increasing consumer a
nd competitive pressure requires radiologists to constantly evaluate
where they "add value" to the patient's healthcare experience. The "radiology value
chain" represents the entire cascade of decisions and events that constitutes the patient's
encounter w
ith imaging services. From the moment the referrer makes a decision to
obtain a radiologic exam until we provide a timely, accurate, and clinically useful
interpretation, we have the ability to add value to the patient's care. This includes the
ranscription process, which is too often downplayed by radiologists as someone
else's problem. But we are often judged by our reports, and, therefore, the
dictation/transcription problem has to become our problem.

All of us have experienced different syst
ems, each with its strengths and weaknesses. My
favorite experience occurred long ago when a group of us military radiologists were
moonlighting at a local community hospital. The after
hours transcriptionists were also
one in her last semeste
r of a Ph.D. program in English. This was the one
time in my life when my reports came back better than I dictated them. Too often,
however, we are satisfied just to see reports that reasonably reflect what we meant.

The problem with voice recognition tra
nscription is that the strengths add value and the
weaknesses add costs, and the value and costs go to different entities. The two major
strengths frequently touted for voice recognition are timeliness of reports and cost

But accurate and timel
y report production is not yoked to voice recognition.
My experience with several hospitals with a dedicated, stable, in
house transcription crew
that knows the radiologists and their idiosyncrasies is that very high quality reports can
be turned out and a
utofaxed to the referrers easily within four hours, and often within one

Even in today's highly charged consumerist medical marketplace, this has to be
considered more than adequate for most reports.

As for the cost savings, it would serve radiologi
sts well to recall the saying, "There's no
such thing as a free lunch
." As Trefelner says, the huge savings in labor advertised for
voice recognition systems are really just shifted costs, typically from the hospital to the

The costs saved fro
m eliminating transcription jobs resurface as decreased
radiologist productivity. Unfortunately, because the radiologist is the most expensive link
in the radiology value chain, even a minor decrease in his or her productivity results in a
ly high cost to the whole practice, and the competitive advantage of rapid
reports can be quickly negated. When the radiologist is not a hospital employee, however,
this cost is invisible to administrators, who are often eager to accept the trade.

Over th
e past 10 to 20 years, the costs of doing business have been steadily climbing for
radiology practices, and reimbursements have failed to keep up. A practice's success is a
complex interplay of revenues and costs that the group hopes will result in net pro

Keeping close tabs on costs is critical but increasingly difficult, as many costs are
insidious, incremental, or otherwise hidden.
Voice recognition, separate from any
potential benefit, is clearly a cost to radiologists in decreased productivity, a
nd any net
benefit must be realized at least in part by the radiologists bearing the real costs.


Radiology is performed more efficiently and more accurately when done in batches.
Voice recognition systems using the rad
iologist as the editor cause interruptions with
every case and decrease the radiologist's productivity. Some studies show up to a 36%
decrease in productivity with voice recognition.

Anecdotally, voice recognition adds 30
minutes to three hours to the rad
iologist's work day.

My group recently took over a reading contract with three MRI centers using voice
recognition. Each radiologist had to spend two to three days of training for the system
and these were all U.S. radiologists with English as a primary l
anguage and no accents.
All of this time represents costs borne by the radiologists, while the savings benefit the
center owners. Even proponents of voice recognition agree that costs are shifted to

One challenge to radiologists is to show a
dministrators how the radiology
value chain depends on all links working together and that decreased radiologist
productivity can lead to decreased imaging volumes for the facility.

Instead of blindly allowing ourselves to be entranced by the seeming magi
c of voice
recognition, we need to carefully and objectively evaluate it on a cost
benefit basis as we
would any proposed change to our practice/business structure.

Keeping in mind the
problem with voice recognition systems
that the costs and benefits do n
ot necessarily
accrue to the same entities
what are the radiologist's options?

First, as Trefelner pointed out in an earlier Diagnostic Imaging column, "Don't put a
radiologist up against the wall until you mean it" (January 2002, page 56), radiologists
hould realize that they do not have to cater to administrators' whims. The national
shortage of radiologists has given the specialty market power that we can exercise in
evaluating activities and opportunities
what we will and won't do and what we will and

won't accept. Radiologists should be aware of their rights and obligations under their
contracts. They should have rights that at a minimum allow input into decisions regarding
report generation: not only if it is done but how it is done. Radiologists sho
uld demand
that they be allowed to continue batch reading and that there be at least one full
time in
house editor with primary responsibility for editing reports.

My comments are not made to condemn voice recognition, but rather to focus attention
on how

decisions should be made regarding the anticipated costs and rewards, and who
pays and who benefits.

Certainly, an increasing number of facilities and practices across
the country are incorporating voice recognition technology with at least superficial
ccess. But radiologists should be aware that even if the decision is made after careful
analysis, the course can be treacherous. At Children's Hospital of Boston,

it took two
separate attempts in four years to finally implement voice recognition. Success
achieved only after a dedicated multidisciplinary team of radiologists, hospital
management, IT specialists, and vendor experts approached the problem in stages.
Interestingly, their chronicle of the implementation emphasizes the importance of cost
efit analysis but does not address the issue of who incurred the costs and where the
benefits were credited. Radiologist productivity was not addressed at all.

Ultimately, the challenge for radiologists is to work with administrators to develop true
egic partnerships with aligned interests. With voice recognition, realigning interests
may be as simple as Trefelner's suggestion that some of the savings be spent in the
radiology department. Radiologists need to convince administrators that when the
ology value chain works to maximum utility, providing excellent customer service,
the facility benefits. Conversely, any impediment to a highly performing value chain,
such as decreased radiologist productivity and ability to handle workload, can create
lays, dissatisfaction, and even lowered quality. This, in turn, has a negative impact on
the facility's profitability. With any new technology, program, or procedural change,
radiologists should expect a true partner's share of input into the decisions and

a fair share
of benefits commensurate with the costs they bear.


1. Trefelner E. Voice recognition misses a few beats. Diagnostic Imaging May

2. Heilman R. Voice recognition transcription: Surely the future, but is it ready?
ographics 1999;19:2.

3. Mehta A. Voice recognition technology. Accessed June 2004.

4. Sperry K. Study looks at the pros and cons of voice recognition.
42203.php. Accessed June 2004.

5. Bankhead C. Voice recognition adds time to radiology reports. Accessed June 2004.

6. Antiles S, Hornberger C, Weis M, et al. Implementing speech recognition at C
Hospital Boston.
19.asp. Accessed June

7. Camponovo E. The business of radiology: cost accounting. JACR, in press.

Dr. Camponovo is a practicing diagnostic radiologist and nuclear medicine physician

central Pennsylvania. He now serves as chairman of The Heart of Lancaster Regional
Medical Center and Carlisle Regional Medical Center.