There has been some discussion recently on Dr Crippin’s blog regarding voice recognition and the particulartly the accuracy of radiology systems.
I thought I’d take this opportunity to let you see some real world examples.
The following items are all live reports from our RIS along with the corrected version.

1)
“There is a calcifcation and three in the right breast.
Report: No dlatation view was confirmed presence of calcification was do not have adequate diameter with a year time shown. The right. The follow-up times six -1 years time or consider the centimetre seen. The end of the right.”

“Indication: Mocrocalcification R3 in the right breast.
Report: Magnification views confirm the presence of the calcifications but do not add a great deal to the evaluation. We could either follow-up in six months or one years time or consider stereotactic biopsy now.”

2)
“There is cells: The patient was of a questionable area in the right breast.
Report: left shoulder was shown to indicate the presence of a focal mass or gross tissue distortion or any other features. I think the four the appearances probably due to composite shadowing. The calcification to the”

“Indication: Compression views of a questionable area in the right breast.
Report: The compression views are reassuring and do not indicate the presence of a focal mass or localised tissue distortion or any other worrying features. I think therefore the appearances are probably due to composite shadowing (Reclassification R2)”

3)
“Clinical indication: Line with the fall. Pain in right hip.
Report there is a left T8 chang ascites. No bony injury seen.”

“Indication: Unwitnessed fall. Pain in right hip.
Report: There is a left THR in situ. No bony injury seen.”

4)
“Report: No bony injury seen to skin fold. The radial metastasis or sclerotic although there is a fracture has been identified in this region. Follow-up film in seconds 10 days would be useful if the patient has ongoing symptoms.”

“Report: No bony injury seen to the scaphoid. The radial metaphysis is a little sclerotic although no definite fracture has been identified in this region. Follow-up films in 7-10 days would be useful if the patient has ongoing symptoms.”

5)
“For details for the injury to the persistent sweeling and cellulitis and right knee is a moderate effusion. The venous cruciate ligament collateral ligaments are normal. No abnormality of the articular surfaces are sub-articular dense any abnormalities of the soft tissues to indicate inflammatory changes around the body and there is no evidence of thickening. Cause of the joint effusion is clear scan no other shown.”

“Indication: Blackthorn injury to knee. Persistent swelling and cellulitis.
Report: MRI Right Knee: There is a moderate effusion in the joint. The mensci and cruciate ligament are normal. Patella ligament is normal. Collateral ligaments are normal. There is no obnormality of the articular surfaces or subarticular bone. I don’t see any abnormalities of the soft tissues to indicate inflammatory changes around foreign body and there is no evidence of synovial thickening or synovitis. The cause of the joint effusion is therefore unclear from the scan as no abnormality has been show.”

6)
“For details of rectal tumour MRI pelvis there is a radial head from the right hip replacement but this does not the perirectal tissues. Scans show approximately 12cm above the dentate line in the neoplasm extends at least extent is most wall of the. On the axial images there is a presacral lymph node in the 1 o’clock position of the rectum. There is no measure 13ml in diameter and has a transverse internal texture. The is in contact with the 7-French or section margin. Thee is further smaller perirectal at the 7 o’clock position and 10 o’clock position also having intravenous internal texture and probably represents a focal lung metasteses. The para-sagittal images of O group of along the inferior recent Eric vessels between the aortic bifurcation. The largest tear measures 16ml and again slightly projects texture around the likely to represent metastases. The primary tumour itself extends through the muscularis into the surrounding area rectal fat indicating invasion is seen on scan is nine 779 the invasion is in the lateral position there is no patient to the peritoneum at this level. Assess for his AST 3 and one tumour.”

“Intication: Rectal tumour
Report: MRI Pelvis: There is a great deal of artefact from a right hip replacement but this does not seriously affect the perirectal tissues. Sagittal scans show an upper rectal neoplasm approximately 12cm above the dentate line and the neoplasm extends at least 6cm and is almost all above the peritoneal infection. On the axial images there is a presacral lymph node in the 1 o’clock position relative to the rectum. This node measures 13mm in diameter and has a hetrogeneous internal texture. It is in contact with the circumferential resection margin. There are further smaller perirectal nodes at the 7 o’clock position and 10 o’clock position also having heterogeneous internal texture and probably represent local lymph node metastases. The sagittal images also show a group of nodes along the inferior mesenteric vessels between the aortic bifurcation. The largest node here measures 16mm and again has a slightly heterogeneous texture and a rounded outline likely to represent a metastasis. The primary tumour itself extends through the muscularis into the surrounding perirectal fat indicating vascular invasion here and this is seen on scans 7 to 99 but the invasion is in the lateral position so there is no invasion into the the peritoneum at this level. This therefore is a T3, N1 tumour.”

Corecting some of these took longer than it would have taken to type it it the first place. Unfortunately the corrections are part of the learning process so highlighting the lost and typing it from scratch isn’t an option.
The system is getting better but sometimes ‘Metal Mickey’ has off days or funny five minutes – five perfectly readable reports are followed by one of total garbage.
The worrying thing is that in at least two of the above the meaning implied in the machine transcribed report is the exact opposite. “left shoulder was shown to indicate the presence of a focal mass or gross tissue distortion” became “The compression views are reassuring and do not indicate the presence of a focal mass or localised tissue distortion”. “Cause of the joint effusion is clear” became “The cause of the joint effusion is therefore unclear”.
Thats before we get to confusion between hypo- and hyper- and the odd ‘no’ getting missed (as in “no evidence of…”).

Something to think about.

All is rosey in the garden of Connecting for Health isn’t it? No one complaining; no bad news.

But maybe there is another reason….

Managers have attacked the Connecting for Health IT project for ‘bullying’ people into talking down problems on the ground.

West Herts primary care trust IM&T service manager Roz Foad was among speakers at an IT conference who criticised the scheme to create an NHS-wide clinical computer system.

She told HSJ: ‘There is a bullying aspect to Connecting for Health.’ Local staff felt unable to voice their concerns, she added. ‘We are not allowed to put out anything that is not spin”

(HSJ 9/11/06)

For a year and a half our RIS has quite happily sucked information from PAS as required. Any address changes automatically came over to RIS.
New PAS – a bit slow, other problems with the interface but, still information seemed to come over. Within two weeks of PAS we noticed an increase in the number of DNAs into the department.
Dug a bit deeper. All appointment letters are generated by RIS using the patient demographics. Some letters were going out with partial addresses- no post towns, counties our postcodes.
Looked further. RIS and PAS were talking but every time PAS received updates from the Spine (the data system carrying patient information for the whole of England [1]) the address details were amended, and mapped seemingly to random fields in the address, leading to the problems previously mentioned.
So the spine is wrong; it needs modifying.
But no, they can’t do that. It’s a national system. It can’t be modified.
So, instead, we have currently applied 11 patches (and counting) to our interface to cope with the various permutations of address coming over.

DNAs are down again…but what a bunch of twats!

[1] Remember it’s not a national health service as far as computing goes – politics gets in the way – but that’s another story

Lorenzo is now talking to the spine!

The interface is up!

Still need to check the old PAS for anyone not on the new one.

 But, as Meatloaf said – Two out of three ain’t bad

Good news!

Well we went live with the new PAS this weekend and updated all our hosital numbers to the new 10 digit format [1].
Now PAS won’t talk to RIS – a problem with the interface. This means we can’t pull patient details over from PAS. We’re OK if they’ve been in Radiology before, but it all falls apart if they haven’t.
Firstly we need to search the new PAS and hope they’re there, but we only imported patients who had been treated by the hospital over the last two years [2] so there’s a fair chance they’re not..
Next we have to search the old PAS and try there. If we find them then we have to book the patient onto the new PAS using the number from the old PAS prefixed with our four figure hospial ID.
Still with me? If the patient is on neither PAS [3] we have to contact the central ‘data spine’ to find the patient then register then on the new PAS. This was taking up to 20 minutes per patient over the weekend [4].
Once we have a new hospital number we pull the patient details over and book them.
Like we could do in 10 seconds on Friday!

Still think the ID card system will be workable?

“Sorry to keep you sir. Just join the crowd behind me while I wait 20 minutes to retrieve your biometrics”.

…And the radiologists couldn’t retrieve images to report for several hours which, as this is what are paid to do, seems a tad wasteful.

[1] Perhaps someone in the health infomantics world could explain why acute trust patient numbers begin with ‘R’ while PCT numbers begin with ‘5′; contrywide, not just locally. There must be some form of logic, but it escapes me.

[2] And anyone under 25 or with a history of cancer

[3] Note I don’t say PAS  system like everyone seems to. That would means Patient Administration System system – for some reason that makes me irrationally annoyed.

[4] According to an email from the IT dept BT were looking into it. When I left to go home we had no notification that it had been sorted. If it takes BT as long as it did to fix my ISDN line … but that’s another story.