Upload
others
View
3
Download
0
Embed Size (px)
Citation preview
Bridging the Gap: Applying Lessons from API Development to Healthcare Enterprise Integrations
Shelby Switzer | @switzerlyHealthify | @HealthifyUS | www.healthify.us
F o r r e f e r e n c e o n l y . D o N O T d i s t r i b u t e .
3
Who I Am
Open Data
APIs
@switzerlyF o r r e f e r e n c e o n l y . D o N O T d i s t r i b u t e .
ProductWe build community health infrastructure to connect underserved populations with the services they need to thrive.
ServicesWe provide network building services to foster relationships between healthcare organizations, local government, and community services.
4
HealthifyOur mission is to build a world where no one’s health is hindered by their need.
CustomersPayors, providers, government entities
UsersSocial workers, community health workers, program managers
Integration partners7 live integrations with EHRs, HIEs, Case Management Systems
F o r r e f e r e n c e o n l y . D o N O T d i s t r i b u t e .
World*
9
The Integration GapHealthcare
* SaaSy startup world, at least * For the optimists: HL7 vs FHIR
Greatest Common Denominator JSON over HTTP(S) Pipe-delimited files via email/server
Most debated topic REST vs GraphQL SFTP vs FTPS*
Culture Standardization Customization
Naming Conventions Puns and animals Obscure acronyms
Modus Operandi Products Projects
F o r r e f e r e n c e o n l y . D o N O T d i s t r i b u t e .
10
But still...Your users have problems.
And they’re not going away.
You are not an island.
Neither is your product nor your integrations team.
The future is real.
And you need to plan for it.F o r r e f e r e n c e o n l y . D o N O T d i s t r i b u t e .
11
But still...Your users have problems.
And they’re not going away.
You are not an island.
Neither is your product nor your integrations team.
The future Change is real.
And you need to plan for it.F o r r e f e r e n c e o n l y . D o N O T d i s t r i b u t e .
Design-first product thinking
12
1. Know your users and their problems.
2. Iterate on the design.
3. Define success and measure it.
F o r r e f e r e n c e o n l y . D o N O T d i s t r i b u t e .
1. Know your users and their problems.F o r r e f e r e n c e o n l y . D o N O T d i s t r i b u t e .
Key integration user groups
14
Group At Healthify Problems
End users Social workers, community health workers
High on volume, low on time. Wants to find quality services and coordinate care.
Business / admin users Program managers, executive stakeholders
Wants to understand outcomes and value of SDOH investment. Wants to surface relevant data to other hospital staff via EHR.
Patients Underserved communities (for you, may be same as end user)
Does not trust or know how to access community services. Wants help navigating the system. Doesn’t touch the software but they’re why we’re doing this.
Developers CMS dev team, contractors, EHR analysts
Does not know what the workflow is, why they’re building this, or who you are. May not have clear project management on their end.
F o r r e f e r e n c e o n l y . D o N O T d i s t r i b u t e .
15
EHRs and Healthcare
They don’t know (or care?) why they’re here
Workflow mock-ups, explicit data + message instructions
Shared drive
Explain the why behind every decision
Weekly check-ins with written follow-ups
Speak their language (and to their egos!)
Act as their project manager as soon as you see it’s needed
Be prepared for hand-holding and relationship management as your core responsibilities
SaaS APIs
They know why they’re here
User-friendly documentation and tutorials
Code snippets and sandboxes
Open source tools and client libraries
Events and hack days (free pizza)
Dev newsletter
Guide product use through collateral and dev marketing
The Developer User GroupF o r r e f e r e n c e o n l y . D o N O T d i s t r i b u t e .
16
Case study: Health System X- We pushed back too hard on their long timeline estimate
Then the IT director de-prioritized the build and we started 6 months later than planned
- We didn’t step up with project management for their team after the second meeting
when their team hadn’t delivered progress
The build dragged on for 6 months, when it should’ve been finished in 12 weeks
- We didn’t supply written info about the project’s data needs or workflow requirements
early enough or follow up on decisions to all stakeholders
We spun our wheels regularly debating topics we’d already discussed, and they had to add
scope late in the game to account for data requirements F o r r e f e r e n c e o n l y . D o N O T d i s t r i b u t e .
Iteration in the wild
18
EHRs and Healthcare
Focus is on workflow and data:
Ask for screenshots of client’s EHR and make mockups (we use InVision)
Make sure IT is part of workflow finalization
Store everything to make future features easier
Use a mock server (Redox’s dashboard) to build against
Look at real data as early as possible
Understand how the Redox models relate to the workflow
“Contract” is the workflow, but even that may change.
SaaS APIs
Focus is on the interface
Design first, build later
Start with an API definition (e.g. API Blueprint or OpenAPI Spec)
Get feedback on the design, update it
Use a mock server as mutual test-bed for design iteration (we use Apiary)
Determine final design (this is the “contract”) and build.
F o r r e f e r e n c e o n l y . D o N O T d i s t r i b u t e .
Don’t get into data fights with the EHR.
19
- Be liberal with what you accept, conservative with what you send- Don’t enforce [many] data validations- Empower the user by displaying info about history and origin of the data
F o r r e f e r e n c e o n l y . D o N O T d i s t r i b u t e .
21
Case study: Health System Y- We ignored Redox models that we didn’t support in our own data model and didn’t
think we needed
We discovered mid-build that the client expected Visit information attached to the media
message we were sending. We didn’t have a concept of Visit or Encounter...
- We didn’t see real insurance data come in until late stages.
After seeing real data, we realized our assumptions about insurance data quality were
wrong, and we had to add build to account for that (with discord about who would do it).
- We assumed everyone could support (and wanted) a PDF.
We found out PDFs are real hard for Epic customers, and the customer had to go to Epic
directly for blob server support, when they’d rather have discrete data anyway.F o r r e f e r e n c e o n l y . D o N O T d i s t r i b u t e .
Product success
23
● Usage/activity● Conversion rates (e.g. integration transactions that led to a
referral completion)● NPS or other user satisfaction ratings
● Did each identified user group achieve their end result faster orwith better quality? Was the problem solved?
F o r r e f e r e n c e o n l y . D o N O T d i s t r i b u t e .
Project success
24
● Reusability and cost efficacy of integrations● Time to deploy● Dev time spent per integration
● Is this less of a project and more of a product?
F o r r e f e r e n c e o n l y . D o N O T d i s t r i b u t e .
25
Case study: Healthify- We built reusable interfaces for EHR integrations using Redox.
Our most recent integration used 3 already built interfaces (ADT, MDM, Flowsheet ORU) and
dev staff time needed was only 5 hours total.
- We implemented reporting for integrations to understand product success.
All clients with integrations have over 66% engagement (much more variable without).
Referral activity related to integrations workflows is at least 3 times higher than activity
without.
- Our integrations roadmap aligns with product roadmap.
This makes it easier to know what to say no to versus what new build to take on.
F o r r e f e r e n c e o n l y . D o N O T d i s t r i b u t e .
Design-first product thinking
26
1. Know your users and their problems.
2. Iterate on the design.
3. Define success and measure it.
RECAP
F o r r e f e r e n c e o n l y . D o N O T d i s t r i b u t e .