Partners In Health has used plastic ID cards and paper ID cards (with printed labels) to help identify patients at registration. We are currently discussing pros/cons about using biometrics (fingerprints, voice scanning, retinal scanning or iris scanning), ID cards (with or without photo), health passport, etc.
Could you share successes and challenges with patient identification? Cost and accuracy are essential pieces.
One of our challenges has been the initial cost and maintenance of card printers. Perhaps this should be expected when we have printed 75K cards over the past 3+ years at one health center, and many more at a hospital.
Hi Ellen, At HAS, we used labels. In the legacy system it was printed with a dot-matrix printer; before the popularity of barcodes. In OpenMRS, we printed barcode labels with Dymo label printers. HHF was set to also print labels, but with Zebra label printers. My goal at both sites was to put the labels on the vaccine cards. At HAS, the patients almost always carried their vaccine cards because the label on it proved if they were in-district out out-district. (Which made the whole census system break as people would buy their way into the district.) As I mentioned at the OpenMRS Implementer meeting on this subject, HAS was supposed to pilot an eye scan technology. If you need more info on this, let me know and I will dig up the old emails.
They labels always seemed to hold up better than the vaccine cards!
The biggest challenge in using fingerprints was increase in enrollment time. We found that we needed to take at least 4 scans during the patient registration to generate a template that was robust. It depends somewhat on the type of fingerprint reader, its quality and algorithm to generate templates from minutiae. But I assume for the very best backlit reader and homogeneous population, you will still need 3 scans.
If the load is not heavy and can be split across multiple registration queues, I think fingerprints work very well as identifier. At least in our example with close to 200,000 samples, it was definitive in uniquely identifying a patient in less than a second.
Thanks, James. Did you have patient photos or try laminating cards? Did you use labels for lab samples and dossier/files? Did patients try to use another patient’s card?
@ball, This was in production at one large district hospital in Northern India with an outpatient load of 1200+ patients in 5hrs. We implemented only registration and printed the registration/visit summary for the outpatients. We had 3 queues for registration/revisit, but if you do the math we couldn’t spend more than 30 sec/patient. No outpatient records were recorded into OpenMRS. The server burnt down due to power fluctuations and we were kicked out of the hospital because they didn’t plan resources for the implementation.
@ball I did my masters work on patient matching , and another additional barrier we faced was the poor performance of the probabilistic algorithm in identifying kenyan names - especially where there is transposition of names - thats why we are trying out the biometrics. I will update this page when we have our earliest results in a few weeks on the pilot of the biometric system.
I am curious as to what are the costs of printing the cards – We haven’t looked at that cost ourselves but i think that it would be worthwhile to find out
Dear @ball, At HAS, we took pictures of employees with the Patient Image Module, and then generated a BIRT report (in the form of an employee badge) and outsourced printing employee badges. I think we ended up negotiating a bulk rate of 50 Gourdes / badge, which was cheaper than we could buy the printer and cards for 400 employees, but turn-around was slow. It didn’t really matter since it was just for employees.
We put labels on the patient’s paper chart and on their vaccine card. The lab didn’t use barcodes or labels.
We have tested the use of Bar Code printed on Plastic ID card. The bar code contained information about the patient especially their unique ID. On clinic visit, the bar code is scanned with a hand-held bar code scanner to display patient details. We are completing plans for a pilot in one of our high volume site i.e >15k patients. The card will be completely useless if stolen or misplaced since the card has no other information except bar code, this reduces stigmatization issue.
We used the DigitalPersona fingerprint module. There are a number of devices that use that module. The device used is a backlight - DigitalPersona U.are.U 4500. We used the DigitalPersona ID SDK, which was particularly good for 1:n matching. DigitalPersona was bought by CrossMatch and they just have one SDK now. So that’s a good thing.
