A Medtech client needs a new data analysis tool

Big data often poses challenges for targeted reporting. This was the case for our medical technology client, who provides high-tech, micro-invasive treatment devices for various surgical procedures, helping many patients in Europe regain their health. Over the last few years, the client has built a technical infrastructure, including databases and recording devices, and was ready to start sharing the available data from all treatments with its local workforce, the clinical specialists. This is the perfect starting condition for a focused data product design. They asked us to design and build a Power BI App that provides instant insights into daily processes, efforts, and product performance.

Ingredients: strategy for goals - workflow for results

What we've learned about data product design over the years is that achieving challenging goals requires a valid strategy. Clearly define the 'why', 'what', and 'how' of your work right from the start. Then, ensure you have a performing workflow for gathering user requirements. Our tool of choice for this is the Chart Doktor's Data Question Map.

That's why we initiated the design process with several online workshops. We invited stakeholders and end-users of the dashboard we planned to create.

The goal was to understand the users' needs and perspectives and compile a list of well-defined data questions.

For all the pros out there, here's a takeaway: if you have a well-defined data question at hand, you can directly derive your visual - the data answer - from it. This is one ingredient of the secret sauce for successful data visualization 😎.

Unraveling the Real-World Chaos

In the workshops, we discovered not just one, but many ways the available data could assist clinicians with their daily tasks.

As is always the case with real-world scenarios, there's a big, chaotic jumble of all things that may be relevant for this dashboard 😅. We needed to gain a deep understanding of the users' workflows and processes, the patient situation, and how different products enable various treatments for them.

This understanding was necessary to define all the jobs-to-be-done of the data and to identify the users' questions.

It took us a while, but we finally succeeded in distilling the main data questions. The last step was prioritizing them to establish a clear roadmap and to know where to start.

The power of the Chart Doktor Data Question Map lies in how easy it is to stay focused and achieve results!

Here is the highest-priority data question that we wanted to address first:

"How is a specific hospital performing? How does its performance compare to similar hospitals? What are areas of improvement?"

For measuring performance, the durations of procedures and procedure steps were the most relevant.

Now that we had a north star to lead us through the design, it was time to start chart ideation!

Time to start sweating - it's the data visualization ideation moment.

With the details and requirements in hand, it's necessary to take a step back, observing the big picture and the details simultaneously. The challenge now is one of imagination - the left and right brain working in harmony.

How can we display actual procedure step durations and compare them with benchmark data in a simple and intuitive way? How can we reduce the mental load of deciphering a chart?

Of course, this could have been visualized with simple bars. However, this proved to be difficult to read - the visual signals emphasized the duration differences between the steps, instead of the deviations from the benchmark. We wanted more signals and less noise.

It became evident that standard chart types (bar, line, pie) wouldn't suffice for this task. We needed a custom chart type for it to work effectively.

Thus, we allowed the magic to unfold and trusted our experiences and visual intuition, all while continually checking with our north star: How well does this chart answer the data question?

A Coxcomb-chart - the solution we came up with

As mentioned the challenge was to not only visualise the current appointment data, but also to visualise the according benchmarks to compare. This is where the so-called Coxcomb chart entered the stage. Small hint: we show an example here, original data were obfuscated.

But would we be able to build this with Power BI? That was not clear from the beginning and after some trials with Charticulator we decided to implement it with Deneb, an open-source visual from the AppSource Power BI visuals, provided by the fabulous Daniel Marsh-Patrick. 

How do you read a Coxcomb chart?

It's all about procedure steps (access, diagnosis, treatment and evaluation) and their duration in minutes. You need to go through clockwise starting at the top.

Each slice of the Coxcomb chart stands for one step. The size of each slice shows how much time that step takes compared to the total time of the procedure. The longer a step is, the bigger the arc (or angle) of its slice in the chart. This part is analogous to how a pie chart would function.

Now comes the special thing: The average time for each step in the country is shown as a circle line in the chart. The distance from the center of the chart to the edge of each slice (the radius) shows how the time for that step compares to the average.

If a step takes less time than the average, the slice will not reach the circle. If it takes more time, the slice will go beyond the circle.

Putting things together - the hospital performance dashboard

The overview page of the dashboard provides a view of the three main procedures (appendectomy, gland resection, and kidney stone removal) for all hospitals. It displays key figures like the number of cases, duration, and goals. On the right-hand side, there is a filter panel that allows users to narrow down the view by country, hospital, and surgeon. (Please note that dummy data was used for this case study.)

The next image below shows the dashboard in action, with a specific client (Hospital Central de Valencia) selected.

You can immediately see how many procedures the hospital has performed in the selected time period and how efficiently the hospital is performing each step of each process compared to the national average. On this basis, meetings can be held with the surgeons to discuss the causes of deviations, the influence of different products and possible actions.

In this case, for example, there are significant deviations from the average in the area of diagnosis and also in the assessment of the appendectomy procedure. The next step would be to understand where these deviations come from. This evaluation is crucial for discussing the data with local surgeons.

What's the outcome? Doctors are finally talking with clinicals about data  

After the testing phase of the dashboard design with real users, we realized that they quickly grasped how to interpret the chart, even faster than we anticipated. Medical professionals often have a strong aptitude for working with data! After deploying the dashboard to hundreds of users in the field, we received tremendous feedback. They could finally leverage their meticulously collected data without having to spend their evenings creating charts in Excel. The dashboard has a fresh and appealing appearance and can be presented as is to their clients, the surgeons, providing evidence of successful workflow adaptations and stimulating discussions on how to enhance efficiency and effectiveness with our medtech client's high-end products. 

During the presentations at an internal but international conference with more than 100 participants, it became crystal clear: the users in the individual countries discuss the results from product usage with the staff of the hospitals using our dashboard and derive detailled actions from this. Our dashboard has owned its place in daily work at the client and started to enable the intended change.