OpenMRS Interoperability and Integration Updates

The presentation of OpenMRS distro HIS in the TAC call went well.

Some discussion ongoing on Talk: https://talk.openmrs.org/t/integrating-openmrs-offering-a-practical-pathway-to-implementers/42840?u=mksrom

Next step:

• Create GH the repository: openmrs-distro-his. → Mekom

• Create the Jira project: OpenMRS Distro HIS [HIS] → Mekom

• Initialize the repo using the Ozone Maven Archetype → UCSF to create the ticket, Mekom to work on it.

  • Adjust the Group ID to be org.openmrs

  • Artifact ID to openmrs-distro-his

  • Remove Ozone white labelling to keep the default Ref App theme.

  • Update the O3 Ref App dependency to point to latest next version of O3 (in the pom + Docker images)

  • Enable EPRNext integration

  • Disable other integrations (Lab integration should be enabled when OpenElis Global integration is ready)

  • Disable “embedded ERP” features:

    • Billing & Stock OMODs dependencies

    • Billing & Stock ESMs

• QA/testing of the new distribution.

• Start a dev server for OpenMRS Distro HIS: dev.his.openmrs.org (test.his.openmrs.org / demo.his.openmrs.org, alternatively dev-his...)

• Documentation:

  • OpenMRS Atlassian doc:

    • The repository

    • Vision/strategy

    • The specific configuration (components enabled…)

    • Use cases to cover / patient flows

    • Adoption of OpenMRS Distro HIS:

      • How to use that distribution and replace O3 Ref App with your own EMR distro.

Next call:

• What about analytics?

Romain Buisson Kevin Ngari, Eudson Bambo Wamathaga Kamau

Billing and stock interfaces in OpenMRS (in progress): Interface specs points mostly identified now. Need to write this down in developer-friendly format. (this process is started by Romain Buisson ).

• Looking specifically at the hybrid EMR/ERP screen (order basket):

  • Price

  • Stock availability

Providing the actual FHIR APIs:

• Case 1: ERP with OMOD: Extend the FHIR module API from Billing or Stock Management OMODs to bring the ERP-specific domains.

• Case 2: ERP with dedicated ERP software: Camel routes to expose the FHIR API.

  • This is short/mid term solution: Ideally this should be part of the product itself (Odoo/ERPNext…)

Choice of ERP → choosing ERPNext:

• has a strategic impact in the region.

• Is a bit inferior for initialization

Stephen Senkomago Musoke mentions a possibility to have a bare metal version as an alternative to Docker Compose.

Product & Services terminology management:

• Product catalog VS OCL integration?

  • to be discussed next week.

Gilbert Muthee Arthur Mugume Romain Buisson Amos Laboso Kevin N. G Ngari

Pre-analysis of user stories for Billing and Stock Management done by Paul Adams. Results to be shared by Romain Buisson

• Data points needed for the integration between EMR & ERP (dedicated ERP or embedded ERP modules).

• Used as support for building the APIs on both ERP solutions.

Draft concept note started by Gilbert Muthee to be shared.

Regarding choosing between ERPNext VS Odoo: both integration will be available with OpenMRS FHIR (OpenMRS EIP component) - both are a viable option.

Lab Integration:

• Laboratory app is already implemented in Ref App (Lab results entry).

  • On the same principle of the ERP integration:

    • Laboratory backend module should provide a FHIR interface.

    • Take out the EMR-focused frontend components should be in their own app.

    • Make sure those new frontend components rely on the FHIR API so that the backend can be swapped between the embedded Lab modules and dedicated fully-fledged LIMS.

    • Paul to check the user stories behind that work (initiated by UgandaEMR - also used by Palladium) - Romain Buisson

Billing integration:

• ⚠️ KenyaEMR/KenyaHMIS still using a different version of the Ref App. This needs to be aligned.

SSO:

• Might not be mandatory, but a nice to have.

Initialization of ERP/LIMS… components:

• Metadata (eg, list of products) needs to be aligned within the HIS.

  • As a first step, probably fine to manually align the products list (via each system initializers)

    • Odoo comes with Odoo Iniz

    • ERPNext comes with an initialization mechanism with some limitations. See in Ozone: https://github.com/ozone-his/ozone/blob/main/scripts/erpnext/data_import/generate-import-script.groovy . This only assumes being able to create new metadata or edit existing metadata, but not both at once.

    • SENAITE initialization must be done by starting the system, manually configuring it, and exporting the config. This config file can then be fed to the Plone Initializer.

    • OpenELIS Global - Romain Buisson to identify how OEG can be initialized.

  • Longer term solution:

    • Product catalogue: ERP might be the one driving the list of products & services. Appropriate products & services would be broadcast to their corresponding component (EMR, LIMS…)

Deployment:

• Probably use the Ozone deployment stack to facilitate the deployment (ready-to-use Ozone Docker Compose). This is a standard Docker Compose project to enable and start the various HIS components.

For next meeting:

• Backup & restore of the various HIS components

• Error handling in message routing (integration framework)

• Manual troubleshooting of errors (integration framework)

• Adoption & Documentation

• More detailed scope for a potential hackathon.

Gilbert Muthee Romain Buisson Michael Bontyes

• Scope of the group: facility level

• What is out of this group:

  • MPI/CR Integration squad

  • FHIR channel

  • Offline discussions

• Recurrence of call? Weekly in May, then to revisit later.

• Gilbert/UCSF use case:

  • Looking for a unified UI to integrate ERP data, but also looking to be able to switch from “integrated OpenMRS module” vs. “dedicated ERP” like Odoo - Context statement

  • Common features, no matter the approach and whatever the backend/flavour is (ex. OpenMRS ERP module vs. dedicated ERP system integration). For example:

    • Stock levels upon ordering + moving goods and commodities around

    • Billing features

• Next:

  • Planning integration Hackathon/Workshop in collaboration with Mekom in ~3 weeks - preparing concept note for it (to be sorted out: Goals, type of devs needed)

    • Started: Identify the boundaries/cut-off point, scope, and features of both flavours (ex. what do users get using billing module vs dedicated?)

    • To be started prior to the workshop:

      • Paul A. to look into overlaps and identify what should live where - also by reviewing the OpenMRS data model

      • Prepare the re-architecture scope

      • Pick tools for dedicated ERP. Ex. Odoo vs. ERPNext. Lab would be later.

      • To consider on the long term: how feasible would be the transition from the “light”flavour to the “dedicated “flavour” - data migration + ERP configurations/metadata to consider.

      • Product Catalogue: where will be the source of truth for the products between OpenMRS and the ERP. Middleware example: OpenConceptLab or Odoo. Considering pulling products from higher levels, that can be consolidated with local products = cascade approach. Limit current scope to facility level for now.

    • Workshop scope:

      • Dev and re-achitecture work for that scope to be “optional” + FHIR for the ERP readiness (format and endpoints)

      • Documentation, guideline-style to facilitate other to reuse - A BA could help

      • Current User Stories for billing and Stock

    • Not in scope:

      • FHIR facade - would be a next step

Romain Buisson, Michael Bontyes

• Discussed the architecture - currently working on a diagram that will be shared here soon for everyone to contribute

Eudson Bambo, Romain Buisson, Michael Bontyes, Antony Ojwang

• Objectives of the group:

  • Share practices and learn from each other

  • Investigate a robust interoperability layer and tech stack at facility level - how to enrich data from connecting systems together

• Antony - Palladium

  • In house OpenMRS module to exchange data between EMR <> Centralized Messaging (SMS reminder for Apt - w/ route to SMS provider) 

  • Exchange with Pharmacy Mgmt for HIV-ARV dispensing

  • Data exchange with National Data Warehouse - used for ML training

  • Most KenyaEMR hosted on premises = monthly sync need from facilities to central warehouse

  • Intero with Lab systems (referral labs), VL testing, influenza 

  • HIE is sometimes not under Palladium/KenyaEMR control (ex. need to use existing Rest APIs), but often using HL7 format

  • The OpenMRS module also generates Patient IDs

  • Now using OpenHIM instead of Interoperability Layer (custom solution) - also used for patient referrals

  •  HIE with SHR

  • For consolidation > above sites solutions pulling data from facilities (work in progress)

  • Registrations centralized in client registry

  • Difficulty when not in control of tech architecture

• ICRC

  • Using OpenMRS DB sync = DB to DB, can work offline using queues

• Describe architecture and outline tools

  • Assumptions and considerations

    • Reusable and modular integration of OpenMRS

    • Connectivity issues

    • Both cloud and on premises infra

    • Limited ownership on architecture and infrastructure decisions

    • Limited ownership of data (integrity and regulation aspects)

    • Variety of technical specifications and performance (laptop, edge devices, or cloud)

  • International level - beyond country level

    • Reporting 

  • National level

    • Shared Health Record

    • Master Patient Index

    • Client registry

    • Provider registry

    • Facility registry

    • Reporting

      • DHIS2

    • Machine Learning

    • Insurances

  • District level - often similar needs to National level

  • Implementing Partner level = Cross-facilities level

    • Reporting

  • Facility level

    • Centralized user management (including roles)

    • EMR

      • Bahmni, OpenMRS

    • Lab

      • Senaite, OpenELIS

    • Product catalog

    • ERP

      • ODOO

  • Community-Outreach

    • Data capture

      • ODK, Android FHIR SDK, OpenSRP, Comcare

    • eCDSS

      • Almanach (ICRC), eCare (MSF), CHT
        

• Comparaison of tech stacks

  • Data flattening

    • Mekom: Debezium

  • Data streams

    • Mekom: Flink with Ozone Analytics

    • Palladium: OpenHIM

    • MSF: OpenFN with OpenMRS<>OCL<>DHIS2

  • Aggregation of patient files

    • Mekom: Camel + OpenMRS EIP for event bus (DB Sync)

  • Conflicts Management

    • Mekom: Can view report in DB sync - not resolution
      

• Next call:

• Group per nature/similarities of intero (ex. MPI vs analytics pipelines)

• Horizontal-vertical

• Authentication

• Anonymized, aggregated

• Terminology + mapping management

• Frequency and quantity

• Location and device client of data consumption

• Level of connectivity

• How to reflect interoperable data in UX - Ex. level of insurance coverage

Eudson Bambo, Samuel Lubwama, Romain Buisson, Michael Bontyes

Expectations: share common ground and efforts, learn from each other, explore architecture options, share reusable components like routes. 

Context

• Shared interested expressed from Mekom, UCSF, MSF/Madiro, METS

• How to balance between backed-in development and integration with existing

• All looking for vertical and horizontal interoperability (hori: LAB, billing, stock, and verti: MPI, SHR, reference lab system)

• Samuel:

  • UgandaEMR is using UgandaEMR Sync module to send out FHIR data, which includes a UI for profiles/resources/sync/logs.

  • Scope: HIV surveillance, ART access, LAB requests ALIS, eCBSS, HTS, ART regimen, TB, HIV exposed infant, Client registry, SHR, Cross border 

  • Roadmap: not receiving data at the moment, only sending from facilities. Data consolidation/conflict resolution is pending too.

  • Be careful with scoping, first with the knows, then the unknowns. Need to scope out what belongs to OpenMRS and what belongs to OpenHIE.  Ex: intero with OpenMRS or others? 

• Romain:

  • Using Apache Camel with XML routes, no UI at the moment, can also write JAVA routes.

  • Questions around using OpenMRS as a start point. Ideally looking at a more centralized architecture, looking at OpenMRS as a FHIR client. Need a place to handle FHIR messages. Ex. FHIR → ODOO or FHIR → ERP Next. 

  • Approach: prioritizing Event bus (like EIP) rather than scheduler. Using API to get events + notifications. Also looking for queues services for handling flows, load of data, statuses, conflicts. 

• Eudson:

  • What about more synchronous use cases like a patient needing medication → inform pharmacy in SENAITE and billing in ODOO. 

  • Be careful of the meaning of the data, ex. drug order. 

• Michael: 

  • Early stage, looking at common use cases and architectures

  • Scope: lab tests requests and results (within and outside facility), referrals between facilities, sync with master patient index

Next steps

• Schedule a call next week to:

  • What: Map/list elements of the ecosystem, ex. OpenMRS vs OpenHIR vs Terminology platform, etc.

  • Why: Map use cases and gaps/challenges - contextualize with examples from facilities and concrete scenarios

  • How: Explore approach and architecture options, incl. async/sync cases + what can we share? which components?

• Share visibility in Slack #FHIR channel + notes here.