Voice Recognition services offered VS Transdyne’s Transcription service.
There has been a lot of talk about voice recognition replacing or reducing the cost incurred in having medical dictations transcribed. Voice recognition technology has evolved and achieved dramatic improvisations over the years, yet serious cost savings or benefits are yet to be realized.
This page is not a review on the technology capabilities of voice recognition but more a review on what Transcription vendors offering voice recognition based transcription have to offer.
This review is based on a proposal made by one of the major voice recognition hardware and medical transcription services vendor (Lets Call them VRTSO)to a real hospital.
The synopsis of the proposal: We are not privy to the actual rates quoted.
 FEATURE 1 :
VRTSO offers a dictation platform for the doctors to call into and make dictations.
There is a per-minute charge associated with each dictation made. This charge might be a multiple of the actual long distance expense incurred by the VRTSO. Cost should include the expense of owning and operating and supporting the Dictation platform. This cost could be in the range of 3 to 10 cents per minute or more based on what the VRTSO is doing with the dictations.
If VRTSO is simply handing back the dictation voice files back to the client, this cost could be as low as 3 cents per minute, on the other hand if the VRTSO is running these dictations through their backend VR engine and handing over the raw transcribed files to the client, this charge could be much more.
FEATURE 2 :
Dictations are run through the VRTSO’s voice recognition engine and raw transcripts are handed to the client for further editing.
In addition to paying a premium for generating the raw transcripts, client has to use additional resources either in-house or contracted to edit the raw transcripts to generate usable reports of acceptable quality. The quality of the raw transcript is unpredictable and so is the cost in editing these raw transcripts.
1) Client needs to have tools to access and edit dictation voice files and raw transcripts. VRTSO sells or leases these tools for a set fee.
2) For achieving any reasonable accuracy levels, client’s users (physicians) have to expend time and effort to adapt and learn to dictate in a new style that the VR can process efficiently. This is usually referred to as training the VR system to the users style of dictation, but is more like training the user to adapt to the VR systems needs. VRTSO’s pitch is that VR’s efficiency improves with time.
3) Training users to dictate poses it self as a major issue when it comes to facilities like hospitals as the number of users is quiet high and usage is infrequent. Every physician associated with the hospital will need to train and maintain a profile on the VRTSO’s system. Hospitals should also consider the resources expended in training new residents/interns being brought into the hospital system every year.
4) There is considerable amount of additional effort needed from the users in dictating into a VR system, as they no longer are able to dictate freeform. Every little bit of formatting, grammar and punctuation has to be thought out and dictated.
5) Mid dictation corrections would be a night mare as the user has to base his corrections based on how well he could recollect what was said earlier in the dictation.
6) It is usually assumed that every minute of free form dictation needs 3 minutes to be dictated into a VR to be able to achieve efficient VR.
FEATURE 3 :
Dictations are run through the VRTSO’s VR engine and raw transcripts are edited by the VRTSO Transcription Labor.
10 to 20 cents per line. It is hearsay that this particular vendor charges about 13 cents per line. Note that VRTSO charges this per line edited.
1) What constitutes a line?
a. The vendor proposal on which this review is based defines a line as a “Gross line with a minimum of 4 characters and a maximum of 65 characters which includes text in headers and footers, text in templates, text in tokens and spaces………..”. The list goes on.
b. This line count methodology itself is a money spinner for the vendor.
c. An analysis of the clients estimated 200000 lines a month based on a 65 character line is meaning less and would translate to anything between 300000 to 400000 lines based on the above line count methodology. Lets call them funny lines.
2) How are edited lines counted?
a. Just to start off with the VRTSO made sure that the line count has been adjusted favorably for the VRTSO by coming up with this great line count methodology, the bigger issue is how is the number of lines edited determined.
b. Obviously a lot depends on what the VR engine generates in the form of raw transcripts. VRTSO’s tend to throw a range X% to Y% accuracy in raw transcripts and put the onus on the quality of dictation made by the user, well with the rider that it would improve over time.
c. Lets assume that out VRTSO was bold enough to vouch that the raw transcripts would be 60% accurate, what does this mean to the client. Will 40% of each document need to be edited or 60% of the documents will be letter perfect and only 40% of them need to be edited? This issue is not as simple as it appears above, but for the sake of simplicity lets assume that these are the only two possibilities and another wishful possibility.
i. 40% of each raw transcript needs to be edited: This would mean each and every raw transcript has to be edited!, may be some by just a few words, some more extensively, but the client would have to pay for the entire line count as being edited. 300000 to 400000 funny lines.
ii. Only 40% of the raw transcripts need to be edited: If this is the case, our client has lucked out and would only pay for 40% of the funny line count (300k to 400k), which would be about 120k 160K lines.
FEATURE 4 :
Monthly rental equipment. VRTSO provided equipment for client to mange work flow and let internal resources edit the raw files.
VRTSO is generous enough to let the client pay for every minute of usage and then pay some in the form of Monthly rental equipment. As per my information its about 800$ per month for 10 sets of equipment and licenses.
Based on the above information, it is quiet reasonable to stay away from the VRTSO or for that matter any other vendor who would define line the way they are defined in this example. Other issues are
1) Extremely expensive line count methodology.
2) Undefined Raw transcript quality.
3) Added burden on the physicians.
Word is that it takes seven minutes to transcribe one minute of freeform dictation.
It takes 3 minutes to perform VR friendly dictation for the same one minute of free form dictation. It probably would take the same 7 minutes to edit the raw transcript generated from VR and probably cost a lot more to have it edited this VRTSO.
To satisfy your documentation needs would you
1) Pay for 1 minute of your doctors time + 7 Minutes of your Transcriptionist time
2) Pay for 3 minutes of your doctors time + 3 to 7 Minutes of your Transcriptionist time + VRTSO’s charges
3) Pay for 1 minute of your doctors time + a 10 cents per line fee of TransDyne which will be at least 30% less than what you would pay for 7 minutes of your Transcriptionist.
The situation of this client is in some what like this, Let me buy (lease) into an expensive system (cus it sounds good) to simplify my transcription issues and later buy a more expensive service to use the newly acquired system.
Book mark this page and check it frequently as this page will be updated soon
|
