Developing Clinically-Useful Data Visualizations is Too Expensive and Takes Too Long

*I may be compensated when you click on my links, but you will not be charged.

Because I’m an epidemiologist and a data scientist, I wanted to see what was new in optimizing data visualizations for use in the clinical setting – you know, between clinician and patient. So much pontification has gone on about wearables that collect health data, building visualizations onto medical records and patient portals, and soliloquizing about dashboards, I wanted to see if as of 2022, we had developed anything useful.

So I attended a webinar on October 19, 2022 put on by the United States (US) Agency for Healthcare Research and Quality (AHRQ). As you can read in the explainer on my blog, AHRQ is a federal agency that funds grants to study ways to improve healthcare quality in the US. So at the webinar, they had three of their grant recipient teams presenting on what they had developed in their visualizations.

Use-Cases

I cannot do justice here to all the work these teams put into their visualizations, and I encourage you to watch the 1.5 hour recorded webinar yourself to learn all the details. They did iteration after iteration, gathered mixed-methods data, and did their best to develop dynamic, data-based images that effectively served a variety of audiences in clinical care.

• One presenter described the difficulty of tracking patient-perceived health for rheumatoid arthritis (RA), much less visualizing it so that it was useful to both the patient and the clinician.
• Another presenter provided a very compelling visualization of systolic and diastolic blood pressure over time (that looked strangely familiar to me…).
• A final presenter talked about visualizations for decision-making in “drug-drug interactions” – and to be honest, I had no idea what he was talking about the entire time. If I had a drug-drug interaction, the only decision I’d make would be to go to the emergency department!

What They Did was Too Expensive and Takes Too Long

Whatever these teams did, it was too expensive, and it took too long. The one that was probably the biggest masterpiece was the RA one. It is awesome: evidence-based, visually-appealing, and super intuitive. But if we were to add up all the time, effort and money that went into that, and bring it to a competent digital health business like Athena Health, they would probably point out that there is literally no business model behind this type of development.

How to Make Clinical Dashboard Design Cheap and Short

Below, I am reprinting an image from a book chapter I co-authored with Natasha Dukach titled, “Framework to Evaluate Level of Good Faith in Implementations of Public Dashboards”. AHRQ’s funded projects are not technically “public dashboards”, but they were developed with public funds ostensibly so that the results could serve the public.

As you can see from the graphic, there are a few very basic steps in a generic dashboard development process.

1. Vaporware: First, you have nothing – vaporware. You only have design.
2. Alpha: Then once you make an “alpha” prototype, you have something that you can play with. It may not be fully functional – or it might just even barely function. But enough of it exists so you can get some feedback on it.
3. Beta: You make as many alphas as possible so you can create a beta. A beta is what you actually put in the field to field test. Maybe you never actually launch a beta to everyone, but you actually try to do real-life work with it to see if it actually works.
4. Launch a real dashboard: Once a beta is developed that is ready to be turned into an app to be launched, you launch it (and start versioning/upgrading it).

Shorten the Process and Hire the Right People

To minimize the cost and time to get to step four, you just want to do each of these steps with the fewest resources and in the shortest time – but still be effective at reaching the goal of an effective dashboard that accomplishes whatever goal you established for it.

Here is what that takes: Actual design talent. Actual mastery at designing interfaces, processes, studies to get feedback, and so on. And that is what most people in healthcare lack: design talent. I design studies as an epidemiologist, but I used to be a fashion designer. With the basic design tools I learned in fashion, I design apps, databases, dashboards, study protocols, online courses, you name it. I have not met anyone in the healthcare field who has studied hardcore design like I have. I have met a lot of people who like to sew and do crafts, but that’s totally not the same as having an academic background in design.

Would you hire an architect, graphic, clothing, or interior designer who only had an academic background in healthcare subjects?

So if we want to get serious about clinical data visualizations, we will stop hiring the people with the wrong talent to do these things. One of the presenters showed a slide with about 20 people on it, and said something like, “Yes, it really takes that many people to do this!” She is probably correct – if you hire the wrong people.

Monika M. Wahi, MPH, CPH is a public health data scientist and educator. Are you a career public health professional looking to improve your data science skills at an accelerated rate through a structured mentoring program? If so, please sign up to participate in a market research 30-minute Zoom call to provide feedback about a new program Monika is offering: https://buff.ly/3UnLqmq ???