3 Keys to Health Data Visualization Success
Scott Plewes | October 10, 2019 | 5 Min Read
While from a cost perspective, data visualization may be a relatively low investment compared to other aspects of data science, getting this aspect “wrong” can undermine all the other efforts. It can lead to missed insights, poor decisions and misunderstood recommendations or predictions, which can negate the potential benefits of all the other work.
Data Visualization is a key aspect of data science that enables your data analysis to be operationalized. Simply put, operationalization is the idea that you’re going to do something with the data at hand (after analysis and modeling) such as drawing a conclusion or taking an action. As long as a person is going to be interpreting and taking action on the data, the analysis modeling, followed by visualization, will be key to translating what was found by the analysis into something that a person can understand clearly and use.
However, visualization is not just about taking the data analysis and presenting it “correctly”. Sometimes, it involves going back into the raw data and understanding what needs to be visualized based on the needs and goals of both the user and the operations. As MedTech adoption of data science becomes more expansive, this becomes an increasingly important aspect.
There are a ton of subtleties to getting data visualization “right”, and it would take a whole book to go into all of them in detail, but here are a few key things teams tend to overlook about data visualization that you should be aware of if you want to get the most out of your data analysis effort.
Context of the Data Visualization
It’s More Than What You See, It’s When You See It
Many articles on visualization focus on, for example, when to use a bar graph versus a pie chart (it’s usually better to use a bar graph in most situations, by the way) or considerations of that nature that discuss the form of the visualization. People put a lot of effort into how to display data, as they should. However, data is always viewed in context. There is often a specific workflow involved. Even when “exploring” data, there are predictable relationships in terms of where the user might want to start and what they might want to “dig into”. Getting this right prevents a slew of common problematic situations like overwhelming dashboards filled with data that your user doesn’t really need – or at least doesn’t need all at once.
Scale of The Data Visualization
Choosing Which Data or Perspective to Emphasize
The data is the data, but what you pick as the scale, the primary dimensions, the secondary dimensions and so forth make a huge difference in how people can gain or fail to gain insight from it. There’s no guaranteed way to do this, but there are two factors that always come into play. One is circling back to the question you were trying to answer in the first place. This is where you need to refer back to the questions you formalized in your data strategy and how they tie back to the business problem you’re trying to solve. The other is just experimenting with different axes, color schemes and so on.
Visualizing Data vs. Visualizing The Mathematical Model
When you look at the visualization, this subtle difference is often there, although not always explicitly apparent. We’re not sure why this difference is rarely discussed, but it’s worth knowing because not recognizing it can lead to problems.
An example should help explain the point:
If you plot various data points using two axes, you have plotted your data. You are done describing it in its “raw form”. You could then describe various characteristics of the data such as the mean. This is literally just a description of some aspect of the data and although that might be very important, it’s a description – nothing more.
The true power of data science though, at least sometimes, comes from finding a pattern. Patterns in data science emerge from mathematical models. Consider now data plotted with two variables and you draw a straight line through it – you’ve just applied a mathematical model. More specifically, you are implicitly saying there is a linear relationship between the two variables you have plotted. You’re saying there is a pattern, with that pattern being the linear relationship. At this point, you’ve moved beyond a description of the data to mapping a pattern.
Being able to identify patterns is important because they can allow your team to make predictions, for better or worse. Patterns also always have some assumption or sets of assumptions built into them. In the example we just described above, one of the assumptions is a linear relationship.
What the analysis and/or modeling is doing is to describe properties (averages for instance) or patterns (linear relationships) in the data and make those clear via visualizations. In the case where the patterns are too complex to visualize – machine learning that identifies a broken bone for instance – it’s important to be clear about the inference being made; the bone is broken, or not.
All this being said, any time your data scientist is using something like K-Clustering or Logistical Regression, for example, to make predictions, then she/he is applying a model with assumptions, limitations, as well as a ton of potential value when it comes to predictions. As one very senior data scientist told us “sometimes data scientists can fall in love with their models”. Like any model, it’s good to both appreciate the skill that went into it and also retain some healthy skepticism and look for ways to test it against reality. In the end, it’s a model, not a crystal ball.
Getting Data Visualization Right
While from a cost perspective, data visualization may be a relatively low investment compared to other aspects of data science, getting this aspect “wrong” can undermine all the other efforts. It can lead to missed insights, poor decisions and misunderstood recommendations or predictions, which can negate the potential benefits of all the other work. Data science is only as good as the “fit” of the mathematical model to the reality it’s trying to reflect. It’s not always obvious how good that “fit” is and/or how a poor visualization can undermine even the most accurate model.
Download: Applications of Data Science in Medical Devices
In this eBook we demystify the domain of data science, articulate what’s involved, dive into the two main approaches to applying data science, and review the most common pitfalls experienced by teams that lead to failed data science initiatives.
Health Information System Integration
In this webinar, we discuss interoperability in healthcare and answer attendee questions on Health Information System integration. Download the webinar Now.Read More
Achieving Interoperability in your Healthcare Organization
The challenge with integrating healthcare organizations is that they are just that – separate organizations. Each organization provides a healthcare service, but they provide that service using their own process.Read More
Using Contextual Data in Connected Device Design
With more physical things being connected to the internet, the amount of data collected by these devices continues to grow. When designing an experience for an IoT device, considering context becomes increasingly important to avoid agitating users.Read More