Early- or Late-binding Approaches to Healthcare Data Warehousing: Which Is Better for You?
Since the 1980s, organizations wanting to better understand their operations through analytics use enterprise data warehouses (EDWs) to organize and make sense of their data. And while these EDWs are commonplace in most industries, the technology has been a challenge for healthcare. Health systems find it difficult to implement and maintain EDWs.
Healthcare Data Warehousing Challenges
There are several reasons healthcare organizations have struggled to successfully implement and use EDWs. For instance, organizations need to put clinical data from the EMR into the data warehouse, but much of this relevant clinical data is unstructured. Nursing notes and physician progress notes are two examples of unstructured data. This type of data has very few content limits or restrictions placed on it, making it difficult or impossible to use for analytical purposes without a specially adapted technology solution. Complicating matters is the fact that unstructured clinical data can reside in several systems within the healthcare organization.
Healthcare analytics and healthcare quality improvement programs also require the ability to integrate clinical data with financial data and patient satisfaction data. Only by integrating all of the organization’s sources of data into a single source of truth can the health system create meaningful, actionable analytics.
What Is Early Binding?
With the goal of meaningful analytics in mind, let’s look at how organizations have approached this task in the past. One approach is called early binding. Binding is the process most data warehouses use to extract data from source systems and connect that data to business rules. In doing so, the data warehouse optimizes data for analysis and retrieval. Early-binding approaches to data warehouse development opt to optimize, through the application of business rules or data cleansing routines, very early in the data warehouse development lifecycle.
In practice, the decision to bind early can have a huge, often negative, impact on the success of your data warehousing projects.
Using this approach, teams spend anywhere from six months to several years mapping their organization’s information systems to a single, enterprise-wide, data model. The data model, in order to account for the complexity of the organization’s business processes, can be enormous in scope and complexity. Oftentimes, the early-binding approach results in project manager burnout and may not deliver what leadership expected in terms of functionality.
Where Early Binding Makes Sense
Early-binding approaches are not necessarily or inherently bad. Earlier binding is appropriate for business rules or vocabularies that change infrequently or in cases where the organization needs to lock-down data for consistent analytics.
However, there are some characteristics that make it less attractive as a data model for healthcare. It takes a long time to start using healthcare analytics due to the effort and work involved in binding the data and agreeing upon business rules. In addition, early binding has limited agility to adapt to new business rules and variation, which are common in healthcare.
Binding later means delaying the application of business rules (such as data cleansing, normalization, and aggregation) for as long as possible until a clear analytic use case requires it. Late binding is ideal for “what if” scenario analysis and is best suited to the ever-changing world of healthcare data.
The result is faster time-to-value. Instead of spending months and even years to implement a data warehouse, health systems can launch their EDWs in weeks.
Late-binding data warehouses are also more scalable and adaptable to the industry-specific problems healthcare organizations are trying to solve. Healthcare is undergoing changes to business rules and vocabulary at an unprecedented rate. The Late-binding data warehouse provides not only faster time-to-value, it also enables the agility required for today’s healthcare analytics demands.
What are your experiences with healthcare data warehouses? Have you used early binding approaches before?