Want to know the best healthcare data warehouse for your organization? You’ll need to start first by modeling the data, because the data model used to build your healthcare enterprise data warehouse (EDW) will have a significant effect on both the time-to-value and the adaptability of your system going forward. Each of the models I describe below bind data at different times in the design process, some earlier, some later. As you’ll see, we believe that binding data later is better. The three approaches are 1) the enterprise data model, 2) the independent data model, and 3) the Health Catalyst Late-Binding™ approach.
Learn more about Steve Barlow
Mr. Barlow is a co-founder of Health Catalyst and former CEO of the company. He oversees all development activities for Health Catalyst's suite of products and services. Mr. Barlow is a founding member and former chair of the Healthcare Data Warehousing Association. He began his career in healthcare over 18 years ago at Intermountain Healthcare, and acted as a member of the team that led Intermountain's nationally recognized improvements in quality of care delivery and reductions in cost. Mr. Barlow holds a BS from the University of Utah in health education and promotion.
Read articles by Steve Barlow
The state of healthcare information technology and analytics has evolved to the point where a revised executive structure is advisable in the C-suite. This new structure calls for a Chief Data Officer (CDO) to focus on extracting data from systems and on mining value from that data, rather than getting data into systems, which is the responsibility of the CIO.
This article makes the case for the CDO, explains how the need for this emerging role evolved, outlines its responsibilities, advises on how to recruit and budget for this position, and details its domain in eight critical business areas:
Governance and standards
Data architecture and technology
Meeting regulatory demand
Creating business value
The phrase “data shopping” should conjure up images of crowded stores, out-of-stock items, long lines, and cranky sales clerks. This scenario is similar to that of your data users and analysts when they are trying to operate without a strict data management policy and without a unified data platform. Many healthcare institutions attempt to operate with data stored in multiple locations, accessible in different ways. Too much time is spent by users looking for the one source of truth and too much time is spent by analysts attempting to gather data to fulfill user requests. Not enough time is spent analyzing data and generating improvements. Data shopping is dangerous and organizations caught up in the spree need to consider a cleanup on aisle 9 (that’s analytic-speak for “consider an enterprise data warehouse”)
It’s no secret that the failure rate of data warehouses across all industries is high – Gartner once estimated as many as 50 percent of data warehouse projects would have only limited acceptance or fail entirely. So what makes the difference between a healthcare data warehouse project that fails and one that succeeds? As a former co-founder of HDWA, Steve details six common reasons: 1) a solid business imperative is missing, 2) executive sponsorship and engagement is weak or non-existent., 3) frontline healthcare information users are not involved from start to finish, 4) boil-the-ocean syndrome takes over, 5) the ideal trumps reality, and 6) worrying about getting governance “perfect” immobilizes the project.
Healthcare informatics has come a long way since its founding visionaries saw a way to use technology to extract healthcare data to improve patient care. But a new era has arrived and health systems are now facing the new challenge of maintaining massive amounts of powerful data that’s sitting unused in expensive storage. The next phase of healthcare informatics is for health systems to move from data acquisition to data extraction, so they can use the insights of the data to prioritize which areas will benefit the most by using data to improve quality and reduce costs.
The most common types of data architectures for EDWs are: the enterprise data model, the summary data mart (also called the star schema), and the adaptive or Late-Binding™ data model. These three models differ in when they bind (early to late). So what does it mean to bind data late? It doesn’t mean to bind everything late, but the model recognizes that rules and vocabularies that are volatile should be bound later. Once a clear analytics use case calls for the data, then bind it. The binding option for a late-binding architecture are: Bind in the source system. Bind during ETL to the source mart. Bind in the source mart. Bind during ETL to the customized data mart (SAM). Bind in the SAM. Bind in the visualization layer.
An EDW is the only viable solution for driving healthcare analytics. This fact has resulted in many BI tools and visualization solutions being marketed as cloud data warehouses, promising quick, user-friendly answers. While they do a great job of visualizing data and exposing it to end users, these tools cannot replace an EDW for 5 reasons in particular:
i. BI tools don’t optimize healthcare data- optimizing data and exposing data-quality issues represents a significant chunk of the initial stages of an EDW project. BI tools just can’t offer this functionality.
ii. BI tools can’t handle large amounts of healthcare data- one patient encounter can general hundreds of rows of data, meaning that reports from BI tools will be slow to generate and inefficient.
iii. BI tools don’t work well with healthcare data at different levels of granularity- Some tools have difficulty displaying the one-to-many and many-to-many data relationships required in healthcare.
iv. BI tools can’t optimize healthcare data for multiple user types- Applying logic against the data so it is understandable at multiple levels for different audiences is something BI tools simply cannot do.
v. BI tools don’t provide for modularity, understandability, and code reuse
The star schema approach to data warehouses is simple and straightforward. Its design is considered best practice for a wide variety of industries. But it lacks the flexibility and adaptability necessary for the healthcare industry. A Late-Binding™ approach, on the other hand, is designed specifically for the analytics needs of healthcare providers. It offers the flexibility to mine the vast number of variables and relationships in healthcare data effectively and leave room for the inevitable future changes.