Using heuristics to guarantee patient uniqueness

[Reggie Vivekananda]

Hi All,

I had an idea that sounds good but needs more discussion of its validity.
What can you tell about using heuristics to match a patient record with another patient record?
As you know, one of the problems that the project can (will) face is how to keep a unique record of a patient. If you’re unaware of the scenario, the scenario is:
- There are medical facilities that maintain a local database with attended patients data
- Patients can migrate from one medical facility to another
- There’s no communication between the medical facilities
- From times to times, a local database is synchronized with a central server

And the problem is: how to guarantee that a patient that migrated between the facilities has only one record in the database?
I know that biometrics like fingerprints and iris scans are an approach, but I want to discuss another approach: the use of heuristics over patient data to try to match duplicate records.
I’ve started by thinking what characteristics could be used to identify a person, like:
- name, living place, preferred nickname
- anatomical characteristics like eyes color, size of the legs and other body proportions
I also thought about what characteristics would differentiate two physically equal persons that not lived the same life, and then social relations came to my mind.
So I think that using social data like:
- son of, father/mother of, friend of, ...
- studied at, lived at, worked at, ...
would help a lot to match similar patient records.

What do you think about that?
What personal, physical, anatomical and social data would be more relevant to identify a person?
How reliable would be an heuristically calculated match?
Can you point to some articles related to this kind of thing?
Do you know something about anatomy that want to share with us?

Thanks for the help,
--
Reggie Vivekananda

[Rodrigo Gidra]

Hi Reggie, take a look at:

 https://wiki.openmrs.org/display/docs/Patient+Matching+Module

[ ]`s
Rodrigo Gidra

[Mário Areias]

Hi Reggie,


Are you talking about sync patient information through different facilities? Is that a requirement at all? Sync patient information is quite complicated and there are many, many problems related to that. It also depends of what strategy you are using to sync this information, you might not have this problem at all! In Mirebalais, we had a syncing running from time to time, so if a patient was transferred to another hospital the data was already there.

Before discuss any solutions, it would be better to get a good understanding of what the requirements are :)

Also, be careful when using OpenMRS modules, there are many which are not in active development anymore and wouldn't work in recent OpenMRS versions. The last release from this one was in 2012 and was released using OpenMRS core 1.7. In Buendia, we are on 1.10, which pretty means this module is not gonna work for us :/

[Leonardo Lima de Vasconcellos]

You could use face recognition also.

[Dan Cunningham]

Mario,

We do need to solve the problem of syncing from one server to another.

For Nutrition, a typical project set up is:

- Inpatient Therapeutic Feeding Programme (ITFP) being run from a clinic / hospital setting in a large city

- 5-10 satellite Ambulatory (outpatient) Therapeutic Feeding Programmes (ATFPs) which are visited once per week by a team of clinicians - usually nurses - who set up a temporary clinic for the day in a village maybe 2-4 hours away by 4WD vehicle (sometimes just one truck, often two trucks, for safety).

Patients sometimes transfer from the mobile sites to the hospital, and vice versa.

Three common scenarios are:

1. Patient in ATFP is deteriorating and needs urgent transfer to ITFP. They get driven back to the hospital in one truck while the other truck stays at the village (it will return later in the day). At the moment they simply take the paper patient chart to the hospital. Options we've been considering for this are:

a. take a tablet back, and when it gets in range of the hospital wi-fi it'll sync to that

b. have a dedicated mobile smartphone app specifically for taking a back up and re-syncing when you get back to the ITFP

c. plug a USB stick into the ATFP server which automatically backs up the database onto the USB stick, take the USB stick back to the hospital and plug into the ITFP server to do a sync. That's a bit of a fiddly process for clinicians to remember to do, in what might be a quite chaotic environment, though

d. take a printer to the ATFP so you can print a back up of the patient record and take that with you. Seems excessive to take a printer though.

e. any other ideas?

2. Patient discharged from ITFP because they are out of critical condition, and will continue for a few weeks in AFTP. Simplest solution to this case seems to be that every night the ATFP server syncs with the ITFP server before you take it out to the village.

3. Readmission of a patient - which could be in a different location from where they were admitted before.

What does everyone think about options a-d, pros and cons of each?

Any other ideas?

- Dan

--
Project Buendia
http://www.projectbuendia.org
Source code and docs at https://github.com/projectbuendia/buendia/wiki
Twitter @ProjectBuendia
---
You received this message because you are subscribed to the Google Groups "Buendia developers" group.
To unsubscribe from this group and stop receiving emails from it, send an email to buendia-dev...@googlegroups.com.
To post to this group, send email to buend...@googlegroups.com.
Visit this group at http://groups.google.com/group/buendia-dev.
To view this discussion on the web visit https://groups.google.com/d/msgid/buendia-dev/09dbaa6d-ca12-42af-bd51-17de815cdb6a%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

--

Dan Cunningham

d...@possibility.design

+447964404897

Skype: djcunning

@dancunningham

linkedin.com/in/dancunningham

[Dan Cunningham]

Reggie,

I love how thoroughly you are thinking through this problem of patient identification!

These are all good ideas.

I just wanted to share a few thoughts and challenges to each of these, based on what I've heard from Ivan and clinicians we've been interviewing:

Fingerprints - promising, as I don't believe they change much in pattern after a certain age (key question: what is that age? these programmes are for children 6mo - 5 years old) or at least the change is just in size and there should be matching algorithms that are invariant to size. Would need dedicated fingerprint scanning hardware.

(idea: we could match the mother instead of the child, then ask the name of the child - only issue would be when a different carer brings the child in e.g. a different family member)

Iris scan - would it be possible to do this just with smartphone/tablet camera in macro mode? would be cool not to need any dedicated hardware!

Names - tricky for several reasons:

- there may be lots of similar names - especially in Muslim countries where one patient might be called "Mohamed Adam Mohamed" and another might be called "Adam Mohamed Mohamed"

- they sometimes give their names in different orders

- with Ebola they would sometimes give fake names due to the stigma (same may apply with Tuberculosis and HIV)

- sometimes families don't name their children until they are a few years old, as there is a high chance of losing them - really sad :-(

Anatomical characteristics - these will change a lot with young children, and would also be influenced by sickness.

Location names - these are often really hard to match. There may be lots of different local names for the same place, sometimes people might just describe where they are from as "on the other side of the river, walk for some time until you see a big Mango tree". Ivan can share a lot more about that problem!!

It seems to me simpler (and easier to understand and use) to have one method of identification that's relatively reliable - fingerprint or iris scan seem most promising - rather than a combination of various factors.

Of course in the best case they come in with their discharge note or wristband from last time, with the Patient ID on it. That's only going to happen rarely though!

Dan

[Mário Areias]

Now we have real problems! Before having an opinion, I have some questions.

What's the infrastructure for the mobile sites? They will have only local network connection or they will have any internet access by chance?

Will ITFP and ATFP sites use the buendia server? If they do, is there any guarantee they will use the same version?

How often these transfer happens? Is it common or exception? Is possible to a patient being transferred from one ATFP to another ATFP? Is there any specific law regarding patient data store? Or patient data transfer?

Cheers,

Mário

You received this message because you are subscribed to a topic in the Google Groups "Buendia developers" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/buendia-dev/oWaKhbnSfXg/unsubscribe.
To unsubscribe from this group and all its topics, send an email to buendia-dev...@googlegroups.com.

To post to this group, send email to buend...@googlegroups.com.
Visit this group at http://groups.google.com/group/buendia-dev.

To view this discussion on the web visit https://groups.google.com/d/msgid/buendia-dev/CAJ9s0WgYWe0X-D08ejk4a%2BdKTeejPZ6tvaRATtA5XvYt2%3DM3qQ%40mail.gmail.com.

[Dan Cunningham]

Responses below!

On Fri, 28 Aug 2015 at 14:54 Mário Areias <mario.s....@gmail.com> wrote:

Now we have real problems! Before having an opinion, I have some questions.

What's the infrastructure for the mobile sites? They will have only local network connection or they will have any internet access by chance?

Only local. We should assume no internet.

Working plan would be:

- 1 Edison + 1 wi-fi router - running on battery (with back-up to run of car battery if necessary)

- 1 backup Edison + 1 backup wi-fi router

- probably 4-5 tablets

I guess there is a chance the remote team would have a VSAT (satellite internet) unit, but that is very expensive to use and really just for emergencies. It would be good to come up with a solution that doesn't rely on that, and I'm not sure they'd always have one.

I'm assuming there is a chance there would be mobile cellular data in some cases, but in most cases there would not. In a few more cases there would be SMS (I know it's possible to do data over SMS) - but again I don't think we can rely on that.

Even in the main hospital there might only be very unreliable internet. 

Will ITFP and ATFP sites use the buendia server? If they do, is there any guarantee they will use the same version?

Yes, and yes, we can instruct staff to only update the servers together. We are thinking the ATFP servers would all go back to the same hospital (ITFP site) every evening.

 

How often these transfer happens? Is it common or exception?

Common.

I have some data for an ITFC in Chad for 2014 which shows:

- 1208 total admissions

- 120 were transfers from ATFC - so about 2 per week

What I am not sure of is how often the transfer is expedited i.e. the patient would leave the ATFC site before the day's work is finished vs. the patient just goes back with the rest of the staff at the end of the day. I can ask that question to clinicians we are meeting over the next week.

In the other direction:

- 834 out of 1204 discharged from ITFC were transferred to an MSF ATFC

Is possible to a patient being transferred from one ATFP to another ATFP?

I think this is very unlikely, anyway if we sync overnight this wouldn't cause any problems.

Is there any specific law regarding patient data store? Or patient data transfer?

I suspect that it's likely that MSF's Ethical Guidelines are stronger than local laws in a lot of cases; someone at MSF is looking into those for us to let us know what we need to follow.

[Rodrigo Gidra]

Whats wrong with using "Location UUID" (or a "Site UUID") + "Patient UUID" as a compound key to identify a unique patient?
we just have to check for duplicates (and then use the "euristics") each time a patient is created.

am i being too simplistic?

[ ]'s
Rodrigo Gidra

[Reggie Vivekananda]

Thanks for the info, Rodrigo!
I gave a look into that module and it seems to go in  the same direction, but I think it could in a more automated way.
I think this would be a 5-steps solutions:
    1) local databases synchronizes with central database
    2) central database tells synchronized databases of possible matches
    3) local databases confirms a possible match
    4) local databases synchronizes with central db
    5) central db spreads the match confirmation during local dbs synchronization.

    During synchronization, the central server would try to match a patient entry using some heuristics, like "this name, age, height and reach are very similar" and ranking the match with some grade. If the match grade is high enough, then the server creates an entry in another table which will mean that 'oh, those entries can probably be of the same patient' and pass the new 'can be' entry to the each local database that synchronizes.

    When the patient returns to a local db that have the 'can be' entry, the attendant would be able to confirm if that the entries are really of the same patient, and spread this new information during new synchronizations.

     The process above would be much more useful if false positives and false negatives could be reduced to a minimum or even be avoided. It's here where I think that anatomical and social data could really help.
     False negative occurs when a patient record belongs to a same patient that appears in another record and the system says they aren't the same patient. How would you avoid this situation?
     False positive occurs when a patient record doesn't belong to the same patient that appears in another record and the system says they the same patient. How would you void this situation?
     In fact, what I'm asking is: what data shall be used to distinguish to patient?
     I didn't read all the data that the matching module is using, but they seem to be normal stuff (name, address, phone number, etc.) and I think that some of that cannot even be applied with African villages. If they cannot be applied which ones could be?
     Also, the weighting points given in the module heuristics probably must be tweaked to achieve better results with Africa population.
     Can you see the examples for false positives and false negatives? I can explain better in another post, if you want.

    Best regards,
--
    Reggie Vivekananda

[Reggie Vivekananda]

Hi Dan,

    I'm with when you say it's better to have one reliable method of identification instead of many.
    Some authors says that iris scanning is better than fingerprint because of the number of points used on recognition (200 and 20, respectively), so I've searched for iris scanning and found an iris recognition project (http://projectiris.co.uk) based on OpenCV for C++/QT platform.
    I will try it later in my tablet to have an insight how good the software is, and since there's an OpenCV port for Android, we can think about porting the C++ code to Android (not easy and not scary).
    My tablet is a low cost device, so, if the software works good, it'll be really promising.

    Best regards,
---
    Reggie Vivekananda

  

Em sexta-feira, 28 de agosto de 2015 10:42:20 UTC-3, Dan Cunningham escreveu:

Reggie,

I love how thoroughly you are thinking through this problem of patient identification!

These are all good ideas.

I just wanted to share a few thoughts and challenges to each of these, based on what I've heard from Ivan and clinicians we've been interviewing:

Fingerprints - promising, as I don't believe they change much in pattern after a certain age (key question: what is that age? these programmes are for children 6mo - 5 years old) or at least the change is just in size and there should be matching algorithms that are invariant to size. Would need dedicated fingerprint scanning hardware.

(idea: we could match the mother instead of the child, then ask the name of the child - only issue would be when a different carer brings the child in e.g. a different family member)

Iris scan - would it be possible to do this just with smartphone/tablet camera in macro mode? would be cool not to need any dedicated hardware!

Names - tricky for several reasons:

- there may be lots of similar names - especially in Muslim countries where one patient might be called "Mohamed Adam Mohamed" and another might be called "Adam Mohamed Mohamed"

- they sometimes give their names in different orders

- with Ebola they would sometimes give fake names due to the stigma (same may apply with Tuberculosis and HIV)

- sometimes families don't name their children until they are a few years old, as there is a high chance of losing them - really sad :-(

Anatomical characteristics - these will change a lot with young children, and would also be influenced by sickness.

Location names - these are often really hard to match. There may be lots of different local names for the same place, sometimes people might just describe where they are from as "on the other side of the river, walk for some time until you see a big Mango tree". Ivan can share a lot more about that problem!!

It seems to me simpler (and easier to understand and use) to have one method of identification that's relatively reliable - fingerprint or iris scan seem most promising - rather than a combination of various factors.

Of course in the best case they come in with their discharge note or wristband from last time, with the Patient ID on it. That's only going to happen rarely though!

Dan

[Reggie Vivekananda]

Hi Mario,

    Yes, we're discussing both syncing and identification problems... maybe we should open another topic to discuss only syncing.
    I'm afraid of being too simplistic when approaching the syncing problems, because I didn't face so many problems with syncing before.
    Can you tell us the problems you see with syncing?

    Best regards,
--
    Reggie Vivekananda

[Mário Areias]

Hi Dan

Thanks for The answers! Although, I forgot to ask other important questions: How many patients we have to sync? How many facilities needs to be synced? And is it possibly to a patient being transferred from an ITFP to other ITFP?

Reggie,

Yeah, there are lots of challenges when doing sync of patient data. Later today, I will describe some of them in depth :)

--

Project Buendia
http://www.projectbuendia.org
Source code and docs at https://github.com/projectbuendia/buendia/wiki
Twitter @ProjectBuendia
---
You received this message because you are subscribed to the Google Groups "Buendia developers" group.
To unsubscribe from this group and stop receiving emails from it, send an email to buendia-dev...@googlegroups.com.
To post to this group, send email to buend...@googlegroups.com.
Visit this group at http://groups.google.com/group/buendia-dev.

To view this discussion on the web visit https://groups.google.com/d/msgid/buendia-dev/306bcb56-416d-4f98-bbc9-2eff43bbcd44%40googlegroups.com.

[Mário Areias]

Now I have more time about why I am so concerned about sync.

It is quite complicated

To illustrate that, I'll give you an example. Most hospitals in US do not share patient data in an automated fashion. They use fax to do it. Can you believe it? They do because sync patient data is hard, here is why:

1) Legislation: Patient data is extremely confidential and every country treat this slightly different from another. In some countries you need to get explicitly authorisation from patients to be able to share their information with anyone. Other countries gives you lots of constraints in how and where to store.

2) Security: Because the patient data is extremely confidential, security should be treated with extreme care. If any information is leaked, it can have serious impact on the hospital/organisation.

3) Scale: If you are talking big countries, you have lots and lots of data. One patient can have all sorts of data stored, everything from basic patient information to diagnoses, visits, orders, drugs, doctors and etc. So get all that data and multiply by thousands or even millions of people. If you use Brazil and US as an example this is already huge, but in countries like India this is extreme.

Also, there is another aspect of scale: facilities. How many facilities will have access to this information? All of them will have access to all patient data? Is there any control? If so, how is it implemented? How facilities will have access to that information? Will they sync in their local database? How many copies of this confidential data you will have across the country? Is this secure? Access to an central database is an option? As you can see, there are many and many implications regarding scale.

4) Standards: Although, in healthcare informatics there are some established protocols, there is no existing protocol to sync patient data (HL7 should be able to do fix this, but I haven't heard yet people using it for this). So if facilities have different applications, how will they be able to share data?

These are some of the challenges you need to have in mind, when thinking about how to sync patient data. That's why is so hard to sync patient information, I've heard that in UK they have a good system to do that and I bet other countries might have something in place. However, this problem is quite complicated even for developed countries.

Specifically for Buendia, it seems we have some of these problems, but others we don't have, like the Standards. Everyone will use the same system, which helps quite a lot when doing sync. But we do have other outstanding problems, because in Buendia we don't have big and nice hospitals in a major city in a developed country. Infrastructure is a MAJOR problem in this kind of settings like Dan described in previous e-mails. Budget and staff are other problems as well.

Phew, it was a long e-mail, but that's what I mean about syncing patient data is quite complicated. At first glance, it seems we can build our own (and hopefully) sync mechanism if the scale and use cases are not very complicated. If they are complicated, I would suggest to leverage something else to tackle this problem, like OpenHIE.

Let's see how this conversation continues :)

[Nailson Landim]

Thanks for the long e-mail Mario. It's a hard problem, but as you said, it will be done in a single system.

I think iris recognition would be better because it has no physical contact, so there's no common surface to spread a contaminant.

With this recognition, would exist a single way to identify someone, and would be easier to sync and compare patient data. 

Dan, does the time is synced between tablets and servers?

On Thursday, August 27, 2015 at 3:13:19 PM UTC-3, Reggie Vivekananda wrote:

[Leonardo Lima de Vasconcellos]

I just loved the idea of Iris recognition. I even did some research myself and I'm positive that we can have this working on android with only one limitation: the camera has to be a good one. I read somewhere that should be at least an 8MP camera.

Take a look at this:

https://www.youtube.com/watch?v=G95Qm8Fv5z0

Leonardo Lima de Vasconcellos

Sistemas ▪ E-Commerce ▪ WebSites ▪ Hospedagem

--

Project Buendia
http://www.projectbuendia.org
Source code and docs at https://github.com/projectbuendia/buendia/wiki
Twitter @ProjectBuendia
---
You received this message because you are subscribed to the Google Groups "Buendia developers" group.
To unsubscribe from this group and stop receiving emails from it, send an email to buendia-dev...@googlegroups.com.
To post to this group, send email to buend...@googlegroups.com.
Visit this group at http://groups.google.com/group/buendia-dev.

To view this discussion on the web visit https://groups.google.com/d/msgid/buendia-dev/b5f5acc4-ae9c-41d3-8c35-54fa0e6103a0%40googlegroups.com.

[Mário Areias]

I don't know much about iris recognition. But have in mind, OpenMrs doesn't have any feature related to that. Also we need to understand if it is doable to scan iris for every single patient. I assume that's not easy when a patient is unconscious.

I have great concerns in increase complexity from the beginning. I would rather start with something more simple like Rodrigo suggested and then move to iris scan later on, if that makes sense.

To view this discussion on the web visit https://groups.google.com/d/msgid/buendia-dev/CAP8EwAPFy%2B1BV-OQw%3D0o_nsakuCnCMQGMp75RuGLi2oHrxNJwQ%40mail.gmail.com.

[Leonardo Lima de Vasconcellos]

You've mad good points but I beg to differ:

1. To take a picture of an unconscious person just open his/her eyelid. Medical doctors do this all the time. Maybe the globe will not be positioned on the right position but the picture can be taken after the person regain consciousness. But, sure, we cannot assume nothing. We have to make sure and validate this with MSF doctors.
   
2. Iris recognition should not be the primary identification value of a patient because some people just don't have eyes. The iris information would be an added value to identification to help matching records alongside with other information.
3. The solutions mentioned before are more complex than implement a well known algorithm, but of course we need to look into it a bit more. That being said, I know squat of iris recognition and to discuss this further I'm going to read a bit more (hehehe).

  

[Leonardo Lima de Vasconcellos]

You made good points* 

[Reggie Vivekananda]

Thanks for your thoughts, Mario!

Based on what you wrote, there are some main areas that we need to attack:

- regulatory/legislation: where legislation and adopted standards matter;

- financial/funding: how much money the project will need/have and how to get it;

- computing resources: hardware and software for each pilot sites and centralizer

I will add my comments in your e-mail below.

--

Reggie Vivekananda

Em sáb, 29 de ago de 2015 às 06:09, Mário Areias <mario.s....@gmail.com> escreveu:

Now I have more time about why I am so concerned about sync.

It is quite complicated

To illustrate that, I'll give you an example. Most hospitals in US do not share patient data in an automated fashion. They use fax to do it. Can you believe it? They do because sync patient data is hard, here is why:

1) Legislation: Patient data is extremely confidential and every country treat this slightly different from another. In some countries you need to get explicitly authorisation from patients to be able to share their information with anyone. Other countries gives you lots of constraints in how and where to store.

 [Reggie] In this area, we'll need someone that can understand the local laws and which protocols are allowed that can serve as a bridge from technical side to regulations side. Who is able to play this role?

2) Security: Because the patient data is extremely confidential, security should be treated with extreme care. If any information is leaked, it can have serious impact on the hospital/organisation.

[Reggie] For syncing purposes, the data can be encrypted in a PKI way. If there's no PKI locally, we can think about providing services like that. We can also put the keys on crypt chips (or smart cards, usb dongles, ...) to accelerate the math and act like a key vault. In my opinion, there's no technical issues here, just regulatory ones. What says the law about patient data being transferred with medical authorization?

3) Scale: If you are talking big countries, you have lots and lots of data. One patient can have all sorts of data stored, everything from basic patient information to diagnoses, visits, orders, drugs, doctors and etc. So get all that data and multiply by thousands or even millions of people. If you use Brazil and US as an example this is already huge, but in countries like India this is extreme.

[Reggie] I agree with you, the data can become too large for simple infrastructures. In this case, I think the problem is budget, because I think that technically it's possible to deal with such amount of data (Google, Amazon, Facebook are good examples). Maybe one of that big companies, like Google, could help us?

  

Also, there is another aspect of scale: facilities. How many facilities will have access to this information? All of them will have access to all patient data? Is there any control? If so, how is it implemented? How facilities will have access to that information? Will they sync in their local database? How many copies of this confidential data you will have across the country? Is this secure? Access to an central database is an option? As you can see, there are many and many implications regarding scale.

[Reggie] Again, I think this is more a regulatory kind of problem. If the law allows us, we can figure a way to sync the data, even if an office boy have to get an encrypted copy of the data to sync and bring that data to another facility.

 

4) Standards: Although, in healthcare informatics there are some established protocols, there is no existing protocol to sync patient data (HL7 should be able to do fix this, but I haven't heard yet people using it for this). So if facilities have different applications, how will they be able to share data?

[Reggie] If we can read and write HL7 records, we can sync them, even if HL7 doesn't have (yet) any syncing protocol. Regarding to different facilities, what kind of standards they use? If they don't have an obligatory standard, we can propose one along with a test version of the system to corroborate the goodness of the standard.

[Reggie Vivekananda]

I've some updates on the iris recognition software of http://projectiris.co.uk :
Good: the software is compilable, well organized and works (worked with my cheap tablet 1280x960 front camera)
Bad: it requires user intervention (a little bit hard) to tell where in the image the iris shall be recognized
Conclusion: this sw is, for me, a good start point for studying but it would require lots of changes to be good for in-field use

I will look the source code of some eye tracking software to have an idea of how hard would be to mix both functionalities (discover where iris is / identify that iris).
Also, I will look other iris recognition software to see how they could fit into a mobile environment.

Best regards,
--
Reggie Vivekananda

[Rodrigo Gidra]

Is there a "parent openRMS" database where we can data pump "child openMRS" databases?

Iris recognition is cool but useless in the project if we dont have a unique way to identify
patients along geographically separated facilities.

[ ]'s
Rodrigo

On Thursday, August 27, 2015 at 3:13:19 PM UTC-3, Reggie Vivekananda wrote:

[Reggie Vivekananda]

And how about using iris (or a hash over iris data) to identify the patients?
--
Reggie Vivekananda

[Rodrigo Gidra]

Sounds cool, but how do we scan iris from people in hazmat suits?

[ ]'s
Rodrigo

[Reggie Vivekananda]

That's a good question, Rodrigo!
What you are asking me is: "How good is a iris scanner when there's some translucid/transparent material between the sensor and the iris?" and the answer can only be given with experimentation and/or advices from more experienced people, since I don't have (yet) experience with iris scanning.
If iris scanning can reliably identify the 95% of the people who won't be in using hazmat suits or belonging to some other exception group (e.g.,people who born with eyeball disorders), then iris scanning is still a very valid option.

Best regards,
--
Reggie Vivekananda

[Nailson Martins Dantas]

I think that a person on a hazmat suit is properly identified, or am I wrong?

Enviado de um celular

---

De: Reggie Vivekananda
Enviada em: ‎02/‎09/‎2015 14:18
Para: Buendia developers
Assunto: [buendia-dev] Re: Using heuristics to guarantee patient uniqueness

--
Project Buendia
http://www.projectbuendia.org
Source code and docs at https://github.com/projectbuendia/buendia/wiki
Twitter @ProjectBuendia
---
You received this message because you are subscribed to the Google Groups "Buendia developers" group.
To unsubscribe from this group and stop receiving emails from it, send an email to buendia-dev...@googlegroups.com.
To post to this group, send email to buend...@googlegroups.com.
Visit this group at http://groups.google.com/group/buendia-dev.

To view this discussion on the web visit https://groups.google.com/d/msgid/buendia-dev/efcccc20-e938-486a-9a51-3a35e0866d81%40googlegroups.com.

[Rodrigo Gidra]

Yes Nailson. but every system with sensitive information needs an authentication and authorization mechanism.
You cant leave sensitive information open to any user. what if a patient is using the tablet? what if someone loose a tablet with all that info...

the system needs to know who's accessing it so that it can provide the right information or deny it according to the user role/authentication status.

[ ]`s
Rodrigo

[Leonardo Lima de Vasconcellos]

Iris recognition would not be a security feature for the app but a unique identifier for the patients.
To secure the tablet use Android's own security features.

[Rodrigo Gidra]

By the way... how are you going to attach a reader to a "buendia device"? it would require to redesign the plastic cover.

[ ]`s

[Leonardo Lima de Vasconcellos]

Using the tablet's camera. Have you read this thread at all?

--

Project Buendia
http://www.projectbuendia.org
Source code and docs at https://github.com/projectbuendia/buendia/wiki
Twitter @ProjectBuendia
---
You received this message because you are subscribed to the Google Groups "Buendia developers" group.
To unsubscribe from this group and stop receiving emails from it, send an email to buendia-dev...@googlegroups.com.
To post to this group, send email to buend...@googlegroups.com.
Visit this group at http://groups.google.com/group/buendia-dev.

To view this discussion on the web visit https://groups.google.com/d/msgid/buendia-dev/7dfa8368-0cf1-4d35-b7ec-05fb0137b23d%40googlegroups.com.

[Rodrigo Gidra]

i did, in my phone... but first is kinda of strange to think of iris recognition withouth thinking in authentication/authorization
ok i got it, you want to use it as a UUID for patients and well i dont remember its possible to capture iris
with a tablets camera....

but, i think you have a very nice ideia.

[]'s

[Rodrigo Gidra]

All the solutions i saw to iris scan in tablets/phones requires additional hardware.
Samsung is about to create a tablet with built in iris recognition, but then again
it requires a iris scanner instead of a tablets camera.

[ ]`s