Tuesday, August 30, 2011

GIDI-Up!
...and OUT!

A golden oldie post from 2007. We now have integrated digital voice recording, which magically crashed following server upgrades. Seems the XML codes from iRecorder aren't being heard by IMPAX. My partner Dr. Killer longs for the days of the GIDI. How nostalgia clouds our memories...



Image courtesy of: http://www.kinkelder.org/


Just about every medical student has read Sam Shem's novel, "The House of God" and most can recite the 13 Laws by heart:
    1. GOMERS DON’T DIE.
    2. GOMERS GO TO GROUND.
    3. AT A CARDIAC ARREST, THE FIRST PROCEDURE IS TO TAKE YOUR OWN PULSE.
    4. THE PATIENT IS THE ONE WITH THE DISEASE.
    5. PLACEMENT COMES FIRST.
    6. THERE IS NO BODY CAVITY THAT CANNOT BE REACHED WITH A #14 NEEDLE AND A GOOD STRONG ARM.
    7. AGE + BUN (Blood Urea Nitrogen) = LASIX DOSE.
    8. THEY CAN ALWAYS HURT YOU MORE.
    9. THE ONLY GOOD ADMISSION IS A DEAD ADMISSION.
    10. IF YOU DON’T TAKE A TEMPERATURE, YOU CAN’T FIND A FEVER.
    11. SHOW ME A MEDICAL RESIDENT WHO ONLY TRIPLES MY WORK AND I WILL KISS HIS FEET.
    12. IF THE RADIOLOGY RESIDENT AND THE MEDICAL STUDENT BOTH SEE A LESION ON THE CHEST X-RAY, THERE CAN BE NO LESION THERE.
    13. THE DELIVERY OF GOOD MEDICAL CARE IS TO DO AS MUCH NOTHING AS POSSIBLE.

I can attest to the fact that these laws are really fairly accurate. But I wanted to bend one of the laws a bit as an introduction to today's topic, the GIDI.
The GIDI is an example of Dalai's version of the 11th law: SHOW ME A PROGRAM OR TECHNOLOGY DESIGNED BY ENGINEERS FOR I.T. WITHOUT PHYSICIAN INPUT THAT ONLY TRIPLES MY WORK AND I WILL KISS IT. This will make sense shortly.

You are naturally asking, "What is a GIDI?" Well, the official title of this software extravaganza is "Generic IMPAX DTMF Integration," and it was designed to help us communicate reports to the outside world. It was a good idea, with all the right intentions, but we all know what the road to Hell is paved with, don't we?

Here's the problem the GIDI was designed to solve: We have Agfa Impax PACS, as if you didn't know, and we also have an older Dictaphone analogue voice dictation system. As we dictate happily along, there must be a way to enter the patient into the system for the transcriptionist to know whose report she is typing. Prior to the GIDI, we had the option of simply reading the patient ID, or the accession number, or something like that into the mike as the beginning of the dictation. There is also the option, if it is set up to do so, of keying in that same number with the pad on the dictaphone head-unit. From the Agfa GIDI manual:

GIDI was written because tones describing the images must be manually dialed into the dictation system. This is slow, tedious, and error prone.
I'll agree with that. So, Agfa was commissioned to write the GIDI!
GIDI introduces a more efficient workflow. The application automatically starts at boot time with Impax and waits for a user to press the "start dictation" button on Impax. GIDI is aware that the button has been pushed and uses the modem to dial into the dictation system a series of tones based on Impax study information and user specified settings. The radiologist then makes their recording. GIDI is also aware when the user presses the end dictation tone, which causes the Impax study to be updated such that the dictation is complete.
If you're interested, this is the control panel for the GIDI set-up.
On our particular Dictaphone implementation, The GIDI buys us two things: first, we don't have to enter the accession number (although we usually dictate it anyway), and second, the voice dictation can be called up by telephone from the system if you know the patient's ID. This allows for instant gratification when transcription turn-around just isn't good enough.
To this point, everything sounds great. Now, here's how it works in practice. A SoundBlaster board is placed in your workstation computer if there wasn't already one there, and it is connected to a modem output. The modem AND your dictaphone are connected in parallel to the telephone line. You start the GIDI by double-clicking its icon, if it didn't start by itself, or if it shut off, and then you activate the Dictaphone, which dials into the tank. When you begin a dictation by clicking the proper button, Mr. GIDI transmits the tones over the phone line, Dictaphone gives a voice prompt, and off you go. When done, you click the "Dictated" button, the termination tone is sent through, just like pressing "end" on the mike, and you are ready for the next victim.

Sounds pretty straight-forward, right? This should be a labor-saving device for all involved. Except for one little problem. It is a true pain in the backside to use. It was overengineered and undertested. In actual use, the tones are transmitted VERY SLOWLY. Yes, an extra 6 seconds per study is no big deal, unless you are reading 200 studies a day as we do some weekends. That would be 1200 seconds, or 20 minutes. Let's see...even an extra 10 minutes per day, working, say, 200 days per year, adds up to 2000 minutes or 33 hours. That would be just under four 9-hour working days per year. You get the idea. All that time wasted because of a program that is supposed to be helping me.

As Ron Popiel says, "Wait! That's not all!" The thing has a bad habit of hanging up the dictaphone, crashing IMPAX, or decoupling itself with the dictation system, requiring restarting of the GIDI, redialing into Dictaphone, and sometimes rebooting the computer altogether. So, add another 10-20 minutes to the daily toil. The system only allows dictating one exam at a time, so we have to re-GIDI every study on a multiple trauma patient. (And I'm talking about a CT of the head, C-spine, T-Spine, L-spine, chest, abdomen, pelvis, and at least one extremity, as well as CR's of all of the above, so there could be 10-20 different clicks for one patient.)

And, there's more! Occasionally, when the GIDI blows up, the study that was on the screen disappears, and if you weren't paying very close attention in those first few seconds, it is rather difficult to know who just left your screen. Thus, a number of studies are accidentally lost from the worklist, only to be rediscovered hours or days later.

The punch-line is that we don't even need the damn thing. Our human transcriptionists are rarely more than 10-20 minutes behind our dictations, and they are very good about getting STAT reports online, well, STAT! And, we type in our preliminaries for the ER, so they are not in desparate need of this technological tour de farce either. But no one will let it go.

Sadly, we have in the GIDI yet another example of a solution designed without regard to the actual workflow of the actual human being that actually uses it. Again and again and again, we are delivered products that the engineers (keep in mind, I am an engineer by training as well) think we should like, but never asked us if we actually do. Maybe it all boils down to communication. Somehow there needs to be dialogue between the end-users of these things and the folks that create them. Don't go looking for something you can patent, guys, but try to come up with something that will truly help me do my job. And please listen to me when I tell you how I do my job, don't just show me newer and shinier gadgets that you think will enhance my workflow. Most of the time, they just get in my way. Like the GIDI. Which really has to go. Soon. Very, very soon. Like today.

2 comments :

mbecker said...

is your problem really GIDI...
or is that you are using an antiquated dictation system and trying to make it work with a modern pacs? that technology must be 15 years old...

i do not mean to be facetious (probably cant spell it) but audible tones? modern digital dictation is not that expensive relative to the price of your pacs...

we have agfa 5.2 and dvi digital voice dictation (not voice rec...just a 5 year old computer based digital dictation)...you push the agfa start dictate button, it starts the dictaphone with the correct patient number inputed and your are ready to go...push end dictation on the mike and the agfa pacs automatically moves to the next case in the work list...a beautiful thing that does not occur on other pacs we use...have to go back to the worklist...choose the patient...

Mark said...

Hey Dalai..
We're also in a small-town in the southeast. My wife is an "indepentent sonographer" an IDTF offering mobile general, abdominal, and vascular ultrasound, and hopefully echo here in the near future. I'm a software engineer for a defense contractor, focusing mainly on the Military Health System(MHS). Now, she started by providing mobile ultrasound services to rural practices in our area. She would take the equipment to the site, set up, and see a block of patients for a set fee. This service was initially provided for OB/GYNs, who interpreted there own exam results, meaning there was no need for telerad system to shoot the studies over to a nighthawk or whatever for reading. Recently, she has considered expanding the services to internal medicine and family practice physicians. Obviously, In order to do that we had to contract with a radiologist to read and interpret the studies.
I began looking for a PACS that would serve as the DICOM server and the storage infrastructure. I quickly found that the commercial systems were extremely expensive, much to expensive for small independent providers like my wife. I began to research my own solution. After all, TCP/IP is just a protocol, and packets are just packets right? I found some open source DICOM objects and developed my own DICOM listener (SCU), which captures and stores the DICOM files to one of several network attached storage devices with rule based logic depending on which modality is sending the files. It is all set up behind a secure intranet VPN, which is connected via broadband to our radiologists PACS system a couple of states north. I spent less than $1000 US dollars on my alternative.

The system works, but it is painfully slow. Again, even if we spent eighty thousand dollars for a commercial PACS system and bumped up to a ISDN or T1 line, the infrastructure is identical (packets are packets right?) It is simply not designed to support telemedicine. It may work well in a hospital, but is limited if the reading radiologist is three states over, or in India for that matter. Obviously, she cant just send files containing private health information in email. FTP is not a secure option either, and secure VPN over TCP/IP is limited to the infrastructure between the source and destination. **BUT** A system utilizing a SFTP appliance http://accellion.com/replace-ftp-sftp-accellion-sfta-solutions.html designed specifically to handle large files could have a potentially huge impact on this whole teleradiology thing. I've been speaking with Accelion corporation, about this idea of using secure file transfer appliances end to end to encrypt and transfer DICOM between remote system. I spoke candidly with their chief technology officer in Sinapore and explained to him my idea of enabling the device as a dicom SCU, listening on say port 104, then marketing the device as a significantly faster, more secure, and compliant method of sharing DICOM files. Im trying to get them to "loan" me a couple of appliances to set up a test scenario. I doubt I can make it happen, but if your interested I keep you in the loop.

Mark Poe
markpoe@swidiagnostics.com