Late-Binding™ vs. EMR-based Models: A Comparison of Healthcare Data Warehouse Methodologies

Written by . Posted in Analytics in Healthcare, Data Warehouse

Two groupsHealthcare organizations putting together an analytics strategy have many things to consider, not the least of which is a healthcare data warehouse. I know that many organizations are taking a wait-and-see approach with analytics solutions provided by EMR vendors and other out-of-the-box solutions. In this post, I’ll compare two models of a data warehouse: Late-Binding™ and EMR-based. I’ll outline things to consider when planning for long-term success in data warehousing and analytics.

Healthcare Analytics: Make the Right Decision First

Analytics is the most important technology competency that healthcare organizations will need to master in the ever-changing healthcare environment. Organizations that can make decisions, improve care, and change behavior based on data will be the organizations that will succeed in the face of decreased payments and increasing scrutiny from the government.

Some organizations believe that they can delay an enterprise analytics solution for a short-term solution. These organizations are not considering the significant risks involved in changing technologies. It takes time to build a culture of analytics and train your knowledge workers in an analytics platform. It takes time to build trust in an organization’s data. This trust is hard-earned and a breach of trust can lead to damaged reputations and further analytics budget reductions. Changing technologies puts the analytics strategy at risk, but it also puts your organization at risk by delaying robust analytics capabilities that healthcare organizations need to be successful today.

Core Competencies of Healthcare Data Warehouse Vendors

EMR vendors know EMRs. They understand how to capture data efficiently and how to build interfaces that easily fit into physician workflows. They do not typically have a great deal of expertise in data warehousing and analytics. The distinction must be made because there are now best practices based on decades of experience in healthcare analytics that when thoughtfully applied, greatly increase an organization’s chances for success in analytics. These best practices include technology, governance, and deployment.

Health Catalyst knows healthcare data warehousing and analytics. Catalyst currently has over 500 years of experience in healthcare data warehousing and analytics. The Catalyst team has been focused on making analytic use of healthcare data for about as long as healthcare analytics has existed! Just as important as our deep understanding of technology, Catalyst understands how to deploy and govern a data warehouse to ensure success in building a culture that leverages one of the greatest assets of many healthcare organizations: data.

Data Warehouse Comparison: Healthcare Data is Complex

Many of the out-of-the-box data warehouse solutions are built on a dimensional data model. The dimensional model has been used extensively in manufacturing and retail domains. When compared to retail and manufacturing data, healthcare data is far more complex. Each patient encounter contains thousands of data points linking the patient to elaborate clinical workflows, diagnoses, procedures, complex financial transactions, among many other data points. In order to map all of these data points into the dimensional model, complex transformations of the data are required to conform complex data to the relatively simple data model of a dimensional schema. Systems that require heavy transformation of data as it is loaded typically do not scale well because the cost of including new data points and maintaining multiple complex feeds quickly consume the team assigned to maintain and develop the data warehouse.

Still other solutions are based on the enterprise data model approach. In this approach, a database model is defined for the data warehouse that seeks to capture every data point possible in a platform agnostic database structure. This approach is fraught with many of the same issues outlined in the dimensional model approach: maintenance of complex feeds and longer development times. There are many examples of this approach never making out of the development phase in healthcare.

In contrast, Catalyst’s Late Binding approach is a data warehousing methodology built for healthcare. Our model limits transformation of data as it is loaded into the data warehouse. As data is loaded into the model, Catalyst’s tools allow creation of data flows that conform table and field names to global standards. This gives all of the data in the data warehouse a uniform look and feel, but facilitates faster incorporation of new data sources due to the lightweight nature of the transformation. It also results in more flexible use of the data for various data marts within the data warehouse. Subsequent data marts are created that transform data more heavily, but are sourced only from the minimally transformed Source Marts. This separation between loading the data and modeling the data is a proven technique in healthcare and results in more maintainable and coherent load scripts and metadata while enabling a variety of analytic techniques on the data warehouse.

Organizations that opt for an unproven short-term solution based on initial cost will be at a disadvantage when the time comes to systematically deploy an enterprise-wide solution.   When reviewing different data warehouse and analytics products, make sure to ask about successful deployments and proven results. A history of success is the best predictor for success in your organization.

What differences do you see in healthcare data warehouse models? What do you look for when comparing your option?

HC-IASWDW-ebook-banner

Tags: , , , , , , , , , , , , , , ,

Trackback from your site.

Eric Just

Eric Just

Eric Just joined the Health Catalyst family in August of 2011 as Vice President of Technology, bringing over 10 years of biomedical informatics experience. Prior to Catalyst, he managed the research arm of the Northwestern Medical Data Warehouse at Northwestern University's Feinberg School of Medicine. In this role, he led the development of technology, processes, and teams to leverage the clinical data warehouse. Previously, as a senior data architect, he helped create the data warehouse technical foundation and innovated new ways to extract and load medical data. In addition, he led the development effort for a genome database. Eric holds a Master of Science in Chemistry from Northwestern University and a Bachelors of Science in Chemistry from the College of William and Mary.

Leave a comment

    Collect key Knowledge Center content to share with colleagues or yourself by dragging it here.