Uniqueness in Healthcare Data Management or Healthcare Data Management Challenges
I have had the opportunity to work in data management for large corporations most of my career. I have thoroughly enjoyed learning about all of the different aspects of business and how businesses can understand more about their products and services from analyzing their data. Industries such as travel, transportation, retail, finance, and others. When I first started working in healthcare, it quickly became apparent there were many differences in the nature of the systems and data they produce. Here is an outline of some of the significant challenges I have seen:
Healthcare’s unique data challenges start with the capture of data. While this subject is worthy of a complete book in itself, let’s put our attention to the nature of these challenges. Most businesses have core capture systems that capture transactions that define the nature of their business. Their sales and marketing departments use that data to adjust strategy, increase revenue, margin, and ultimately, profitability. In healthcare, the nature of the services provided has required many different systems be built to accommodate the business processes. Specific service areas often require quite different processes; from the Emergency Room to a clinical setting to surgery to labs. The medical practice areas also can be thought of separately; information about patients in oncology, orthopedics, ophthalmology, and other areas all require different data and data types.
Noting the need for many levels of data capture, healthcare has been busy over the years building many of these individual systems. Thus the focus on an integrated system has undeniably been desired, but ultimately slow to develop. Simplistically, the Electronic Medical Record (EMR) is the term used for a system that can capture comprehensive patient and visit data. While many EMR vended solutions have existed, they have proven very costly and difficult to implement successfully. This in most cases is driven by the requirement to change business processes to fit the technical solution and it’s flow. Many healthcare organizations have yet to implement an EMR. Of those who have, very few, if any, leverage it as a single system of record.
All businesses have complexities within the data that is collected that go beyond what most of us think of when we initially explore them. It would be simplistic to think of banking requiring only information about money that flows in and out of accounts. Upon initial investigation it is easy to see the need for significant information about their clients, demographics, and operations, just to name a few. As businesses look for that leg up on their competition, it continually comes back to understanding more aspects of the influencers that impact their specific business.
In healthcare, there is incredible complexity just in the identification and storage of transactional data.
First and foremost is the concept of the clinical note. Think of this as a comments area where the patient visit is documented verbally. Each provider has their own style in recording what they put into the note. More importantly, each approach to how a note is structured can also fluctuate. On the surface, it would seem that healthcare would be best served if they could adopt a system of drop downs, check boxes, and the like to standardize the capture of data. In late 2005 at the Healthcare Data Warehousing Association annual conference, a physician spoke on the value of these notes. He noted “We don’t want structured data, until we have to have it”. This seemed so profound to me that I instantly wrote it down. Specifically what he was getting at was that medical professionals have to be able to leverage the information around clinical events such that they can analyze and learn from it. However, they can’t be restricted to the kind of structured system that would have them entering data into some simplistic form. I felt like his directive was to tell IT to hurry up and figure out how to transform unstructured clinical notes into queryable data because providers cannot easily change how they enter it in the foreseeable future.
Unstructured data also includes medical images – xrays, MRIs, ECGs, and many others. As medical technology evolves, more and more automation and recording of procedures will require additional and possibly more complex unstructured data.
Many businesses have transaction volumes that create challenges in processing 24 hours of data in a 24 hour window just so they can put all of their data into an enterprise data warehouse. This “depth” of data is truly a significant challenge. For many large healthcare providers, they also have that “depth” of data to process. While RFID is in it’s infancy, it is easy to see that all industries will be challenged by the “depth” of data relating to RFID in the future.
In addition to “depth”, most providers have a “breadth” of data that I have not seen in other industries. Like customer data, patient data exists about every aspect of who a patient is; their surroundings and demographics, family history, insurance coverage, etc. Consider all of the various tests, lab results, medications, procedures, treatments, etc. that are captured for every patient. Each of these consists of many fields of surrounding data. One piece of information that caught my attention that providers would need to know is the angle of the patient’s bed after a specific procedure has been performed.
Healthcare also has several other critical criteria to consider when designing a data management program.
- The need to reproduce results as they looked at the time of the analysis requires that data be versioned to a point in time.
- Meta data in healthcare takes on additional meaning from most other industries. We do not have the ability to search across actual xrays, images, tissue samples, and other pieces of unstructured data that healthcare has captured. Yet, we can search across the meta data of that unstructured data.
- Consistent terminology – there are many standards and terminologies applicable to healthcare. The challenge is exactly that, there is no single terminology that accommodates all of healthcare, to the business and to the patient. An enterprise data model is necessary to mitigate this issue. From my discussions with other healthcare organizations, most recognize this need, but the time, resources, and experience to develop it does not comply with the business demands for solutions that are already past needed.
- Working with this incredible number of possible variables, it is easy to see why medical research is an enormous field starving for more information but challenged with navigating all possible options.
- The future holds even more challenges. When we think of genomic and proteomic data, these are examples of emerging pieces of knowledge that healthcare will continue to learn about and ultimately figure out how best to capture, store, and analyze.
All industries have their own unique needs that make data management challenging. Never before had I heard of a company that had over 1000 source systems. Through the mountains of systems that exist in healthcare, achieving a mature data management environment and practice may seem unreachable. So where do you focus? That depends on many factors that should be investigated at length. If you are charged with building solutions for healthcare, make sure to focus on two things: 1) The unique challenges you face in what is needed by your organization before you design a solution 2) Leverage proven experience in data management to help you get started.
Bruce D. Johnson
Unit Head of the Enterprise Data Trust for Mayo Clinic
About The Author
Bruce is a 1986 graduate of DeVry Institute of Technology in Columbus, Ohio. In my 20+ years of IT experience I have focused on application architecture, design, project management, and IT management, mostly relating to Data Warehousing.
Throughout most of my IT career I have focused on large database projects on a variety of technologies and platforms. The last 10 years I have mostly focused on IT management and data warehousing. My first exposure to data warehousing came in 1990 at Burlington Northern Railroad as a project lead for a significant component of a very large project to define all corporate data. From there I wrote my own dynamic SQL generator, received significant training on data warehousing, and managed/architected the efforts to bring data warehousing to a large corporation, and consulted on other efforts.