It sounds like we are converging on something here, and potentially actually agreeing on an approach.
Ben - your statement that it will be possible to migrate relatively easily from person-less providers to person-full providers with a Liquibase script between 1.9.0 and 1.9.1 or whenever makes me feel much better. This makes sense - I _definitely_ do not want to go back to a beta for 1.9 at this point given that we have dependencies on 1.9 for our June release in Rwanda and need it out the door before then. Ben - the delegation you describe about provider.name and provider.person.name is not currently happening. This _should_ be fixed as a bug prior to 1.9.0, if possible. Can we take this onto an upcoming design call in the next few weeks with an aim to agreeing upon a design and create some tickets? Mike ________________________________________ From: [email protected] [[email protected]] On Behalf Of Ben Wolfe [[email protected]] Sent: Friday, April 20, 2012 8:16 AM To: [email protected] Subject: Re: [OPENMRS-DEV] Providers and Persons PIH has asked for birthdate to be null for patients recently: https://tickets.openmrs.org/browse/TRUNK-3016 Use-case: when importing patients from past records that have missing date. Mark: if the provider is linked to the person, provider.name<http://provider.name> is cleared and should defer to provider.person.name<http://provider.person.name>. We need a simple API/pojo method for getting the name if it doesn't exist already. Any changes decided upon here will not make it into 1.9.0. Any change to the underlying datamodel or API at this point will force us to re-beta the 1.9.x line. If we change our minds and make provider.person non null in 1.9.1 it is not the end of the world. Creating person records for each null provider.person would be relatively easy using liquibase. Mark's Provider dashboard would simply require that point release and would work more "completely". I'm on the side of requiring providers to have a person record. person.gender doesn't have to be nullable: We could add an "unknown" gender to prevent NPE backwards compatibility. person.birthdate could be nullable and make it an API/UI option if it is required on the person/patient/provider creation screens. (with 10000 extra rows in the person_name table we will need to implement the lucene searches soon though: https://tickets.openmrs.org/browse/TRUNK-425) Ben On Fri, Apr 20, 2012 at 12:13 AM, Burke Mamlin <[email protected]<mailto:[email protected]>> wrote: Actually, allowing persons to exist without names, genders, or birthdates will probably have broader implications on existing code (possibly introducing a wave of NPEs) than whether or not providers, who aren't being used anywhere yet, are required to be linked to a person record. Would the allowance for person records without names, genders, or birthdates be allowed for persons linked to patients and/or users as well? We don't want patients without names, genders, or birthdates do we? I could imagine (and think folks might appreciate) creating user accounts without requiring genders/birthdates, but I'm not thrilled about supporting users without names. So, we end up with something like this (with provider, like user & patient, requiring a person record): * Person stubs (not linked to user, patient, or provider) and Patients (i.e., person records linked to a patient) still require name, gender, and birthdate. * Person linked to a user or provider (but not patient) must at least have a name (gender & birthdate become optional) That would mitigate a large number of the potential NPE, meaning that only code assuming that all persons have known genders or birthdates be refactored. FWIW, I'll try to chat with Shaun tomorrow about the implications for de-duplication efforts. -Burke On Thu, Apr 19, 2012 at 11:00 PM, Michael Seaton <[email protected]<mailto:[email protected]>> wrote: Thanks Burke, Sounds like a a decent approach, I'd be interested to hear what Shaun thinks about the de-duplication issue. That is also something I care a lot about. I would hope API and UI-level restrictions would be able to mitigate this, when appropriate. Obviously there is little chance that this conversation has any impact on our 1.9 release. Unfortunately, it will be harder to make a previously nullable column not-nullable down the road than it would to start out with the more restrictive option. Of course, I could have weighed in on this 9 months ago (or more) when it was actually being designed, but I don't think I would have had the same perspective then. In any event, this isn't really a deal-breaker for us, but it will have an impact on what providers can be managed by the module we are developing. I'm happy to discuss it further on a design call if my proposal has merit. If not, we can work with it as is. Cheers, Mike On 04/19/2012 04:44 PM, Burke Mamlin wrote: Mike, I'm just concerned about our de-duplication efforts. If we make all demographics optional, then we may go take de-duplication from a difficult task to an impossible one. That said, if the de-duplication effort can take linked objects into consideration – i.e., know that a person is linked to a provider – then perhaps an anonymous person record isn't so terrible… as long as it's linked. In other words, perhaps the compromise here is the best solution: only un-linked person records require basic demographics. In other words, you cannot create a nameless, genderless, ageless person record that stands on its own; rather, it would have to be linked to a patient, user, and/or provider. Likewise, you could not unlink an person record lacking name+gender+age from all patient/user/provider objects without also voiding it. Of course, the API would need to enforce this. Maybe we can discuss with Shaun regarding side-effects on de-duplication efforts. BTW… I know I have a reputation as a complicator (that's why Win avoids me). I'm not thrilled about it & I'd much rather be seen as a simplifier, but I don't believe it's always "simpler" when an easy solution now just ends up deferring the difficulty or shifting it to somebody else. Usually, un-debated solution (even my own) are weaker than those that have been beaten around a few heads. In fact, the best solutions I've seen are compromises. If I'm making things complicated without good reason, please continue to call me on it. -Burke On Thu, Apr 19, 2012 at 3:53 PM, Michael Seaton <[email protected]<mailto:[email protected]>> wrote: Hi Burke, I understand your point. But is it all or nothing with regards to constraints? If we remove the data model constraints on Person here, it gives us the flexibility to address each use case as we see fit: * We allow Providers to be simply links to mostly-empty person records, with a single Person name that contains a title and a surname. * We allow Users to also be genderless and ageless (why should we care about this here either)? * We force Patients to have age and gender specified if entered via certain service methods * We leave the door open for Patients without gender or birthdate if needed (eg. if I want to import from EPI Info, and some patients are missing these fields) As it is now we will have a provider management module that will manage _some_ providers, but not all. This isn't a deal-breaker, but it would be better if it weren't so. We also have API methods that we have to deal with (and haven't done so well as of yet) - for example, in trunk currently we have Provider.getName() return simply the name property of the object. It does not take the underlying Person into consideration at all. Yes, this could (and should) be fixed, but it's just added complexity to now have 2 places where names may be stored for a Provider. It just seems to me that we are over-complicating things unnecessarily, and I want to put out this alternative design out for consideration... Mike On 04/19/2012 03:23 PM, Burke Mamlin wrote: Mark, If we have (very useful) information about all providers in the district and they contain title & identifiers like "Dr. Ochieng (1234-5)", are you saying that we should not require gender, at least estimated birthdate, and at least two names on Person to allow this? The key difference between a provider and a user is that a provider does not have to have anything whatsoever to do with the EMR beyond being references (e.g., capturing a referring provider from an external facility). Defining a person without knowing their name, whether they're male vs. female, or whether they're two years old vs. 50 years old would make Person solely a linking tool – i.e., akin to auto-generating 100,000 nameless, genderless, ageless records in the person table and picking the next one when you want to link a patient to user, provider to user, etc. This is directly counter to our effort to leverage Person as the one location within which we focus our efforts to avoid and address duplication errors. We could require that all providers are persons, but I don't believe we should make all hopes at identification for person optional to do it; rather, we'd need to come up with a solution for those folks who get the list of providers like "Dr. Ochieng (1234-5)" and are missing the demographics to generate a person. What's driving you to this conclusion? Is there some reason that you can't limit yourself to providers linked to person records? Where is the assumption that provider must be a person being made? -Burke On Thu, Apr 19, 2012 at 2:46 PM, Michael Seaton <[email protected]<mailto:[email protected]>> wrote: Hi all, I brought up an idea buried in an email during the Provider Attribute thread a few weeks ago, but it was lost a bit in all of the many other points being made, so I wanted to raise it here again explicitly. I think making the link between Provider and Person optional is a mistake. Here is my reasoning: * A Provider in OpenMRS is not meant to be a subclass. It is meant to be a particular role in the health care system that a Person can have. And a person may have many of these. * The most similar thing we have in OpenMRS to this is User. No users are allowed in OpenMRS without being associated with a Person. User is probably more accurately thought of as a User Account. Clearly a single Person can hold multiple User Accounts. * The only real reason I can see for making Person optional is so that we can avoid having to require things like "gender", "birthdate", etc. on Provider that have traditionally been required of Person at the database and API levels. For example, if we are uploading a list of 10,000 providers from a national registry, and all we have is a String name, and no gender or birthdate data. However, I think a more valid solution to this problem is just to remove those data model and API restrictions, and to make Person not require this data. Then, we just move those validation constraints onto the Patient object if we desire (though to be honest we don't always have this information here either, particularly if we are importing data from a legacy system). So, I realize we are at RC3 for 1.9 and all, but my proposal would be to: * Make Person not null on Provider * Make gender and birthdate nullable on Person This is particularly relevant to our team at PIH as we are currently building a Provider Management module on top of 1.9. Mark has had to bake in the assumption that all Providers have Person records (at least those that can be managed with this module), because this is required for many of our use cases - in particular the notion that Providers are linked to their Patients and to other Providers via Relationships. I would be interested to hear others thoughts on this or if any of my assertions / assumptions are just wrong. Thanks, Mike _________________________________________ To unsubscribe from OpenMRS Developers' mailing list, send an e-mail to [email protected]<mailto:[email protected]> with "SIGNOFF openmrs-devel-l" in the body (not the subject) of your e-mail. [mailto:[email protected]<mailto:[email protected]>?body=SIGNOFF%20openmrs-devel-l] ________________________________ Click here to unsubscribe<mailto:[email protected]?body=SIGNOFF%20openmrs-devel-l> from OpenMRS Developers' mailing list ________________________________ Click here to unsubscribe<mailto:[email protected]?body=SIGNOFF%20openmrs-devel-l> from OpenMRS Developers' mailing list ________________________________ Click here to unsubscribe<mailto:[email protected]?body=SIGNOFF%20openmrs-devel-l> from OpenMRS Developers' mailing list ________________________________ Click here to unsubscribe<mailto:[email protected]?body=SIGNOFF%20openmrs-devel-l> from OpenMRS Developers' mailing list _________________________________________ To unsubscribe from OpenMRS Developers' mailing list, send an e-mail to [email protected] with "SIGNOFF openmrs-devel-l" in the body (not the subject) of your e-mail. [mailto:[email protected]?body=SIGNOFF%20openmrs-devel-l]
