How Can Healthcare Providers Improve Their Data Integrity to Achieve Better Insights?

If you’re like most healthcare executives, you need — but don’t always get — solid, reliable data. The terms may be inconsistent; longitudinal data may be broken or incomplete; summary data, skewed and difficult to use. How do you move forward and acquire data that you can trust?

The first step, after recognizing that you have a problem, is to become a more informed user. Then you can consider taking a few preemptive fixes.

The four layers that determine your data integrity

Becoming more informed means understanding that data can lose its integrity at several points along its journey, beginning with applications that generate clinical data. A critical task at this stage, too often missed, is to establish common definitions. Their absence can lead to ‘data confusion.’ Data quality can also deteriorate when collected from different systems, for instance, when individuals pull data manually (instead of through an automated pipeline) or when a lack of internal checks fails to identify incomplete or broken data sets.

At the aggregation layer, where data is summarized for statistical analysis and aligned with a model, incorrect rules could lead to inaccurate hierarchies and values. The data model itself could be flawed, too. Finally, at the consumption layer, data may be presented in ways that fail to meaningfully inform the end user or lack the transparency required by data scientists who may want to build enhanced AI-driven tools without introducing skewed outcomes.

Preemptive fixes to improve your data-driven approach

To improve the quality of your healthcare data, keep in mind the four layers — application, integration, aggregation and presentation — and consider taking the following steps:

Harmonize. Make sure that data at landing is mapped correctly and has the same meaning across systems. “There are analytics that can be done at the starting tier itself, which falls around data governance,” NTT DATA Intelligence & Automation Advisor Priti Malkan says. “It’s about applying textual or semantic analytics and making sure we are all saying the same thing and agree on what kind of values we are looking at.”

Check quality. Make quality management part of the process. That’s the right approach at all layers, but especially with respect to data integration. Internal logic checks can help provide that data flows are accurate, complete, consistent and valid — and that they measure up to all other expectations.

Model correctly. At the aggregation layer, the key is to work with a data model that advances your goals. Start with the end — say, identifying which higher-complexity patients could most benefit from intensified services — and work backwards. Make sure that you and your team build out the correct internal rules and hierarchies needed to reach your objective.

Set the table. To use a restaurant analogy, a customer expects the food they ordered, but also needs the right silverware. In the case of data consumption, healthcare executives may only need the right tools to complement their expertise, or a data scientist may need to implement AI modeling.

Improve your healthcare data integrity today

Advanced AI-driven applications are powerful and appealing. But until you trust your data, resist the temptation to buy and deploy. There's much you can do today to improve your healthcare data and make them better fit your current business needs. Read our latest point of view, Data-Driven Healthcare: Building Trust and Gaining Insight, for a more detailed discussion on how your organization can use data to improve ROI and achieve better outcomes.