As the Chief Information Officer at Massachusetts Eye and Ear (MEE) in Boston, I have been privileged to work with an amazing group of health care providers and have access to a large repository of clinical, operational and research data they generate. It has the potential to be tremendously valuable, if you have the ability to visualize it in ways that can assist leadership, providers and researchers in their work. This information has been collected over decades of patient care. Under my direction, my team implemented a Business Intelligence platform at MEE over two years ago. In the process, I’ve found that there are a set of principles you need to follow to allow for the successful implementation of this type of project. These relate to 1) choosing the right platform for your organization, 2) ensuring that you are working with good quality data, 3) keeping the data definitions consistent, and 4) having a good understanding of the needs of the consumers of the visualizations.
Choosing the right platform to visualize your data can be daunting. There are a lot of great options on the market right now. Many of them provide similar functionality, but differ in price and scalability. You need to determine where your data will reside before you choose your visualization platform. Some solutions require that the data be hosted in the cloud, where others are geared more towards onsite repositories. In healthcare, we typically try to keep our data locally hosted for a number of reasons (i.e., speed, security, accessibility, etc.). You should also consider a platform that provides the security structure that best fits your organization. I’ve found that integration into a local Active Directory can help simplify user management through the use of Active Directory security groups. This approach is much easier to maintain than one that requires local security administration within the platform itself. You will also want to establish a governance structure to ensure that requests for data and visualizations are appropriate and prioritized. If you have an existing structure in place, you should be able to leverage the same model. (i.e., an IRB for research requests, an Ambulatory Council for outpatient needs, etc.).
"Standardize on your naming conventions and you will be poised for success"
Prior to beginning the visualization of your data, you must understand the quality of the data in your possession. Understanding the schema of the data and the contents of the fields will be critical in providing a quality visualization that can be used to make decisions. Stay away from utilizing fields that are either missing values or where the values don’t make logical sense. Over time, data collection methodologies change. Although the fields are initially populated with good data, fields can be deprecated due to newer collection methods. As a result, you may need to normalize the data set prior to visualization. Assembling a comprehensive data set prior to attempting to visualize the data is critical providing an accurate visualization of the data. Ensuring that the first presentation of data is accurate, will be critical to leadership confidence in the platform and will allow you to grow the platform rapidly.
Defining the values you are presenting sounds like a straight forward process, but it can make or break a visualization project. The goal with any reporting is consistency across visualizations. To accomplish this, you must be very precise and consistent in how you define and refer to your values, whether they be calculated or native in the data set. An example of this is how you define “New Patient” in healthcare. If you are in a sizable organization, a simple label can mean very different things to different viewers of the data. For example, is the patient new to the provider; new to the specialty; new to the organization; or the billing definition of “new” (3 years since last treated in the organization)? As you can see, this gets very confusing, very quickly. Standardize on your naming conventions and you will be poised for success.
The last principle to understand is how your consumers of the visualizations prefer to work. Depending on your organization, you will most likely find a number of users who are very comfortable with dynamically accessing the visualizations to leverage the filtering, drill down and other capabilities that the platform provides; where others will prefer a static PDF delivered to them. We utilized a bolt-on tool that worked with our platform to schedule the delivery of provider specific dashboards to the individual attending physicians on a monthly basis. This allowed for a greater distribution of the visualizations without the consumption of additional licenses for the platform. This hybrid approach has served us well in providing information to all who need it. Educating the consumers on which type of visualizations are available will help gain buy-in and expedite the acceptance and usage of the platform. To further gain buy in and to justify the purchase of the platform, try to quickly determine the pain points in your organization. Some of the first visualizations we developed were focused on “missing charges.” By understanding where this was occurring, we were able to work with the staff involved to recover what would have been lost revenue and educate them so that their workflow could address the gaps. The amounts recovered from these visualizations alone have essentially paid for the platform.
These are the basic principles that I’ve found can help with the successful implementation and adoption of a Business Intelligence/Analytics platform to provide visualization into your data. In all of the principles described above, remember to keep the lines of communication open with the business side of the organization. Meet with senior leadership to ensure you are given direction on where to focus, but also engage with the front line workers to understand the data that they are entering into the systems. This is what you will need to drive change and lead to success within any organization.