Chris, Autofile doesn't need to be in the directory containing the docs, and it should not be!
You have to create a series of rules so that it knows where to scan on the page.
I've bumped the build up to 17 now so that it can deal with g3 raw fax formats.
Graham http://www.synapsedirect.com/ Synapse - the EMR for smart users
DrM,
How does it identify which patient the forms should be attached to? OCR is good (even great) for typed items, but a written out name or social security isn't going to do very well with OCR. Our problem also is with a lot of international folks that's first name could be "Nu" "Nu ee" or "Nuee" and getting a name like that to match using OCR would be quite the challenge. So, how does Synapse do this automatically?
My EMR and HIPAA Blog
My (Unrelated) TV Blog Network
Autofile is only going to work with typed names unless they have very neat handwriting!
techguy: How does it identify which patient the forms should be attached to? OCR is good (even great) for typed items, but a written out name or social security isn't going to do very well with OCR.
How does it identify which patient the forms should be attached to? OCR is good (even great) for typed items, but a written out name or social security isn't going to do very well with OCR.
It certainly won't. The use is for labs, forms, or consults .... that have a predictable layout.
In this respect, I've never seen a lab form that has the patient's name handwritten.
I am not sure which forms would be handwritten either. It would be much better to print out the form ... WITH the patient's name on it. or if you want to get fancy, a bar code.
Our problem also is with a lot of international folks that's first name could be "Nu" "Nu ee" or "Nuee" and getting a name like that to match using OCR would be quite the challenge. So, how does Synapse do this automatically?
Interesting Twist. My feeling is that this would be challenging for determining Name based on just the OCR. To solve this for Synapse, I would love to have Autofiling query the database to do some checking .... ie. for me, if the patient's Health Number was detected, never mind the name, file the document under the patient that has that Health Number. I understand that a National Health Number doesn't exist in the US, so Date of Birth and the patient's name might "team up" to better clarify the patient.
Another note: if Autofile has problems, it just leaves the file alone so you can do it manually. Note to Graham: maybe standalone Autofile could Name as much as it can detect, but make sure it is still goes in a "not entirely recognized" directory. Whereas, a failed Synapse import, should just wait for the secretary to do it manually.
The Nu / Nu ee / Nuee problem is sufficiently rare (in 99%) of practices, that it probably wouldn't pass the 90/10 rule, so bloating Autofile to accomodate rare events might not be worth it.
Neat situation though. I think it would affect the standalone, but the tightly integrated Synapse version of Autofile would have much more data at it's disposal and thus would likely have a great "Successful Autofile" percentage.
DrMurdoch: The Nu / Nu ee / Nuee problem is sufficiently rare (in 99%) of practices, that it probably wouldn't pass the 90/10 rule, so bloating Autofile to accomodate rare events might not be worth it.
I don't see a problem. The OCR will pick up the characters and will identify Nuee as the first name, or Nu Ee as the first and second given names.
DrMurdoch:Note to Graham: maybe standalone Autofile could Name as much as it can detect, but make sure it is still goes in a "not entirely recognized" directory. Whereas, a failed Synapse import, should just wait for the secretary to do it manually.
It already does rename as much as possible .. but since you're getting 100% you won't see that!
Build 18
Build 22 of Autofile
A brief benchmark.. Best case scenario - no skewed scans, and all of the same type with rule preselection.
6 scans took 4 mins to correctly identify the patient name, nhi and test date.
Here's a Jing screencast of how to create an Autofile rule
Just wondering if I should print the patient's NHI number on the side of the letters I send the GPs. A vertically aligned bar code ... do you think anyone would appreciate it?
Like this