Stream: implementers
Topic: IPS Development effort?
Iryna Roy (Mar 03 2021 at 15:18):
Hello dear implementers, how many days of effort do you think is needed to implement the International Patient Summary exchange? I am trying to estimate. My current guess based on what I know is 2-4 months - am I right or too off? Thank you! Based on 1 developer , FHIR server , tools etc.
Lloyd McKenzie (Mar 03 2021 at 15:51):
That depends a lot on starting point. What experience to the developers have with XML/JSON? How closely aligned is current data to IPS constraints? How much terminology mapping is needed? Are you using a language that will let you take advantage of one of the reference implementations to help with parsing/serialization?
Iryna Roy (Mar 03 2021 at 15:52):
HAPI FHIR, Java. The plan is to do a pilot, a reference implementation.
Iryna Roy (Mar 03 2021 at 15:55):
for example, we don't need or already have terminology. We need to accept prepared IPS, query the latest IPS or previous version of IPS.
Lloyd McKenzie (Mar 03 2021 at 15:59):
Are you just looking to render the document or do you need to do something specific with the discrete data? Accepting, rendering, querying and displaying should take less than a week if you don't care about the discrete data. How long the discrete data will take depends on how much you want to extract and how complex the structure and terminology mapping is to what your internal model is. (Caveat - I don't usually write code and can't estimate worth beans... :smile: )
Iryna Roy (Mar 03 2021 at 16:05):
That's already helpful, Lloyd :)
Jens Villadsen (Mar 03 2021 at 19:01):
It depends on your internal model - and how much of HAPI FHIR you are using (JPA, plain or simply as parser) and how much of the IPS you are going to support. Only MS or the whole shabang? Is the developer familiar with HAPI? What level of code quality are you aiming for? How much/little coverage are you willing to accept?
Rob Hausam (Mar 03 2021 at 19:15):
All of these are good points. And I think there aren't really very clear answers in the community (yet). But my gut feeling is that you are probably in the ballpark from a technical perspective (although it may be between this and Lloyd's suggestion of a week). Determining the related policy considerations may, of course, take much longer. ;)
Peter Jordan (Mar 03 2021 at 20:47):
Good points from @Jens Villadsen . In addition, my experience to date, mapping to the IPS from an existing data model (based on the US CCR) that's currently rendered into CDA suggests that much of the development effort (which I'm doing, as a reality check, alongside the model mapping) will go into handling data quality issues. Ultimately, as nearly always with interoperability solutions, the major constraint and time-consumer is the quality of the source data in endpoint systems. I also expect that reaching consensus with all stakeholders on data filtering issues will be a lengthy process.
Iryna Roy (Mar 03 2021 at 21:18):
Thank you! Yes, I agree, the effort to map something to generate a valid IPS can be significant. I am aiming to pilot the server side of it: accepting the IPS composition into HAPI FHIR/FHIR-compliant server and storing information, retrieving a list of compositions or a particular composition by id. Am I missing anything from the server side? I am planning on building profiles for the validation of incoming data but I don't see a huge development effort. Correct?
Peter Jordan (Mar 03 2021 at 21:24):
@Iryna Roy - why don't you use the existing IPS Profiles for validation purposes? There are also testing tools available, such as MITRE's Inferno, that can perform validation against those Profiles?
Iryna Roy (Mar 03 2021 at 23:37):
@Peter Jordan that's exactly what we are planning to do but we have our national profiles, we need to do the gap analysis and create profiles compatible with IPS but national-based. Thank you for pointing out the tools. Do you mean Inferno can generate the IPS to send against the end point?
Peter Jordan (Mar 04 2021 at 00:06):
@Iryna Roy AFAIA, Inferno can currently be used to test IPS instances (Bundles) against the IPS Profiles. However, I do believe that there are plans to generate IPS examples - @Matt Rahn and @Rob Hausam might have more information.
Giorgio Cangioli (Mar 04 2021 at 12:56):
@Iryna Roy if you are interested in IPS examples , @dr Kai U. Heitmann is working on a project aiming to generate a large amount of IPS examples.
Last updated: Apr 12 2022 at 19:14 UTC
