#AMIA2020
There are three powerful ideas behind medicine as a moral force:

- Structural competency (https://structuralcompetency.org/)
- Cultural humility
- Ecosocial approach

#AMIA2020
structural competency
New medicine for inequalities that are making us sick
structuralcompetency.org
January 26, 2025 at 7:59 AM
Q: Some tips on the design process?
A: Get users on board as soon as possible (preferably on the ideation phase). Don't work in silos. Moreover, don't design on your own. All these are two sides of a coin (faces of a dice...?)

#AMIA2020
January 26, 2025 at 11:01 AM
Q: What was an important bottleneck you faced?
A: Limitation of actionability - are the resources there?

#AMIA2020
January 26, 2025 at 10:56 AM
Time for Q&A.

Important remarks:

- When talking about actionability we should also consider what it means for interventions.
- Again, actionability != importance. This actionability should be tied to a (realistic) intervention - what will actually cause change

#AMIA2020
January 26, 2025 at 10:50 AM
They also performed usability testing with true end-users. This allowed them to evaluate effectiveness, efficiency, and satisfaction. They also incorporated workflow analysis and observations

(They even used eye-tracking to evaluate the interface - cool!)

#AMIA2020
January 26, 2025 at 10:42 AM
Preview of the prototype. They considered physical organization (e.g., negative space), visual cues and consistency (e.g., fonts, word placement), color (contrast, alarming data), and even 3D effects (consideration of color blindness)

#AMIA2020
January 26, 2025 at 10:36 AM
#CDS should address:

- Technical components
- Thresholds
- Alarm fatigue/missed alarms (both are bad)
- Data analytics (what are you evaluating?)
- Performance of code (could you recreate it? - yep, talking about reproducibility)

#AMIA2020
January 26, 2025 at 10:24 AM
A:

- Community based participant research is crucial
- Trickle results over time (instead of huge dump at the end)
- Consider context on how results are presented (focus on the when and where)
- Co-create information flow pipelines

#AMIA2020
January 26, 2025 at 10:19 AM
#CDS is essential for identifying processes and patterns invisible to the human eye, follow evidence-based practice for optimal clinical management (hey #Oncoguide!), and establishing a meaningful patient-provider relationship

#AMIA2020
January 26, 2025 at 10:19 AM
Q: How should the information be returned to participants in the context of low graph literacy and high cognitive impairment (e.g., chronically ill older patients)?

#AMIA2020
January 26, 2025 at 10:08 AM
Talking about #CDS, there are five rights that #HealthIT should encompass:

#AMIA2020
January 26, 2025 at 10:08 AM
A: Key ingredients to make #ML/#AI a success in this field:
- They should be scrutinized in their own way
- More work on validation
- Faster pipelines that show real value for stakeholders
- Make sure these tools represent the right groups of people

#AMIA2020
January 26, 2025 at 10:01 AM
Time for @kmillerdrph on "Predictive Models: the Human Factor)

#AMIA2020
January 26, 2025 at 10:01 AM
Now for some Q&A:

Q: Do you think technology like #ML/#AI will move the #PROs research forward or it will hold it back due to people not fully trusting them?

#AMIA2020
January 26, 2025 at 9:51 AM
This is how the final prototype looks like

#AMIA2020
January 26, 2025 at 9:51 AM
More lessons:

- Investigators want real-time data (again, make things go fast! Fill in the knowledge gap as quick as possible)
- The sense of community can be a powerful ally for pretty much everyone (researchers, participants, sponsors, etc.)

#AMIA2020
January 26, 2025 at 9:43 AM
They switched to high-fidelity prototypes and tested on colors, amount of information, how this information was conveyed, etc.

(I love seeing the evolution and refinement of the prototype)

#AMIA2020
January 26, 2025 at 9:43 AM
Some lessons learned. Participants:

- Care about privacy, security, and respect
- Want convenience and flexibility (ease of use is king)
- Want to know what is in it for them (this ties very much with what was discussed yesterday in W07 regarding #FAIR data)

#AMIA2020
January 26, 2025 at 9:37 AM
There were several prototypes. Using permutation importance, they identified important variables. However, that visualization was deemed to complex

#AMIA2020
January 26, 2025 at 9:37 AM
Here's where user-centered design comes in. Again, see how this is a very iterative approach

#AMIA2020
January 26, 2025 at 9:27 AM
Current observational/experimental research is expensive, slow, and labor-intensive. This results in systems that learn slowly, leaving many questions unanswered.

The digital transformation should be the corner stone of a #RapidLearning healthcare system

#AMIA2020
January 26, 2025 at 9:20 AM
Now, how to decide what to show? A global explanation of the model (left) or user-specific (right)? Think of the end-user!

#AMIA2020
January 26, 2025 at 9:20 AM
(For a more detailed discussion on this, check this paper: "People-powered data collaboratives: fueling #DataScience with the health-related experiences of individuals" https://academic.oup.com/jamia/article/26/2/159/5255652. Will definitely give it a read later myself)

#AMIA2020
January 26, 2025 at 9:09 AM
Congratulations to all winners of #AMIA2020's Student Paper and Doctoral Dissertation Awards 🥳!
January 26, 2025 at 9:09 AM
Now's the turn of @MeghanTurchioe. They developed a #ML model to predict post-partum depression (PPD) within 1 year of childbirth. It uses a set of clinical features related to mental health history, comorbidities, etc. It used a logistic regression with L2 reg.

#AMIA2020
January 26, 2025 at 9:09 AM