Saturday, May 05, 2007

Speech Recognition...Again



There are a few ways to get noticed on an AuntMinnie.com discussion. Mentioning GE Centricity always draws a lot of attention, but these days, a thread talking about speech recognition will be pounced upon like Howie Mandel after some poor "Deal or No Deal" contestant opens the case containing $0.01.
A recent AuntMinnie front page article reported the results of a study from the U.K. and Germany that discovered

. . . speech recognition technology eliminated "significant delays" in report transcription and typing issues, leading to an improvement in service to both clinicians and patients. Researchers at Hammersmith Hospital and Charing Cross Hospital also concluded that a well-planned organization-wide implementation of speech recognition technology "can have a dramatic impact on radiology service delivery. . ."

In 2004, emergency room and primary care reports took three to five days to complete at Hammersmith Hospital. Finished inpatient and outpatient reports ranged from a low of three days to as many as eight days for inpatient reports in August 2004. With the advent of speech recognition technology, the time from report generation to completion declined to one to two days from January 2005 to February 2006.

This article prompted a number of responses on an accompanying AM.com discussion. A few were positive about SR, but many were doubters, including yours truly. You may remember my earlier post about SR, and I haven't changed my mind. I haven't seen any evidence to suggest that SR has improved. Articles like the above don't do much to help SR's reputation. I mean really, three to five days! That is totally unacceptable under any circumstances. Did our British friends have only one transcriptionist available? Did she have to use the loo for 30 minutes of every hour of her 4 hour daily shift? If you set the bar that low in other venues, I could be the Steeler's next quarterback! I will have to admit that SR might be able to improve the situation if things are that bad.
Fortunately, here in my little corner of the boonies, things are pretty good. Our turnaround time with Human Transcription (HT) is usually from 10-20 minutes, with STAT's coming out within 5 minutes. Of course, if there is reason to do so, we type in a prelim, leave a voice-clip, or (gasp) even pick up the phone to call the clinician. Why would we ever want to switch to SR?
Look, I love technology more than anyone else I know, and I am absolutely fascinated by the concept of the computer listening to me and converting my uttered pearls of wisdom to text, or to commands, etc. Wonderful. But if this is such a wonderful idea, why hasn't it caught on elsewhere, such as among the administrative types that are pushing it on us rads? I think the ultimate reason is this: A machine will never understand the nuances (pun intended for the Powerscribe set) and the outright inherant inconsistancies of human speech. I can dictate this: "The left shoulder, I mean the right shoulder shows no fractures" or some equally flawed speech, and it will come out OK with HT, but not with the machine. Those for whom SR works well have not only trained the machine, but moreover, they have trained themselves to dictate in a completely different way than they learned as youngsters. I personally am not that big a fan of technology that forces me to adapt to it.
I have to wonder if some of the more ardent supporters of SR on the various AM threads work for Agfa or Nuance, or if they are just low-level administrator types taking some heat for spending $300K on something that doesn't work very well.
Has anyone really added up the cost of SR vs. HT (speech recognition vs. human transcription)? I suspect that once you pay the $300K for the initial set up, add in training time and costs, cost of upgrades, cost to replace the darn thing in 3-5 years, etc., there won't be much savings with SR. That does not, of course, take into account the radiologists' time, effort, pain, and potential malpractice suits caused by using SR. But then, the administrative types that usually make this decisions AREN'T PAYING FOR THE RADIOLOGISTS so they don't really care about that. But we rads DO care. If SR is so bloody wonderful, why don't administrators as well as the occupants of the C-suite use it? Apparently, it's good enough to generate reports upon which life-or-death decisions are made, but not good enough to replace the secretaries of any of the above. Think about it, folks....

3 comments :

Anonymous said...

Hey Dali
I just stumbled upon your site while doing some web research and wanted to do a counter point to your argument.
It sounds like your facilities transcriptionists are hold well on the turnaround time at your facility but I would ask what your average study volume is. At a site with over 120K studies, I have personally witnessed report turnaround times drastically change. We have gone from two to five days down to our sought after benchmarks of 4 hours for inpatients and 12 hours for outpatients (not neglecting the Stat floor and ER studies which average a 10 minute turnaround time). I can only name one radiologists out of 20 that I personally trained that didn’t adapt self-edit within days of first using the system.
While I agree that the smaller facilities probably still can’t justify the cost of a voice recognition dictation system. I think you miss the mark on this one.
Like the site – Have a good one!
-john


----------------------------------------------------
John Sole RT (R)(CT) - CPSA
PACS Administrator Supervisor
Swedish American Health System
1401 E State St
Rockford, IL 61104-2315
815-489-4738
jsole@swedishamerican.org

Anonymous said...

My hospital system. Over 1 year after VR launch to 3 hospitals we have a 0% useage at the largest facility, 54% at the facility that supposedly "loves" VR and 27% at the third facility. Aweful numbers. But then again our turn around time was never bad like some have. After speaking with a group that has a 100% useage the hospital is coming to the grim reality that they totally botched the launch and have and continue to grossly undersupport the software and the radiologists.

I am not sure we will ever turn it around. I'll keep you posted.
1) poor planning
2) Poor execution
3) Really poor training
4) Spotty support
5) No reason for radiologists to buy in
6) Customer service of VR company very bad, blames the rads for "crashes" and missed words
= DISASTER

Anonymous said...

Nice site Dalai, but your SR comments are delightfully olde worlde! Yes our turnaround times were hopeless before Speech rec (and had you heard us live you would have heard us tell you this) and SR was a godsend (all to do with the poor wages paid to transcriptionists in our expensive city blah blah). And yes I know it does change a little bit about how you report (it trains us a bit, but I guess we use English language perhaps more fluently over here!!) But haven't new technologies trained us all along? Was it so difficult to get used to looking at a monitor (there are still those who would love to lift the bottom of the film up, but it just isn't so good on a monitor). Sure, we ring clinicians occasionally but for the most part they know the report will be on the HIS and often beat the patient back to the ward or to the clinic.
I would love you to come and see what we have done in person! None of us work for any company, but simply slog away in a department producing over 230 000 exams per year. The university hospital in Nijmegan in Holland have had exactly the same experience as us. It is all about how you put it in, get that wrong and you are stuffed!By the way, it is spectacularly good for training too, but I haven't got time to explain just how right now. Seriously, anyone reading the adverse comments...come and see in person. Open invite! We'd love to show you, and no radiologist's jaw has failed to drop to the floor when they see it in action!