An agency where better data makes a bigger difference
Ìý
Having the most accurate and up-to-date data is critical in every industry. Whether manufacturing vehicles or forecasting crop production, the better the data, the better the outcomes for all.Ìý
Ìý
The United States Department of Veterans Affairs (VA) knows this more than most — the federal cabinet-level agency provides near-comprehensive healthcare services to eligible military veterans at VA medical centers and outpatient clinics throughout the country. In addition, several non-healthcare benefits include disability compensation, vocational rehabilitation and education. VA has a mission to provide life-long healthcare services to eligible military veterans and improve their health outcomes. While VA uses a wide variety of data and analytics countless times a day, an important use case is for the Medication Possession Ratio (MPR) calculation.
Ìý
The MPR calculation provides clinicians a view into how well a patient is adhering to a medication regimen — critical for veterans taking controlled medications like opioids. For example, if a Veteran stops taking medication, there could be adverse side effects such as an increased risk of suicide. At first glance, this might not seem challenging to track correctly. However, the number of veterans and the complexity of historical data makes it challenging to analyze. With information such as drug identifiers, Veteran's prescription regimes, and refills changing, the information needs aggressive tracking over time. Additionally, various groups across VA were using different calculations to measure MPR, creating the risk of inconsistencies across the system.Ìý
Ìý
VA needed to bring all these groups together with a solely approved and validated expert calculation. They also wanted to move this data, which sat in an on-premise SQL server and spanned over 20 years, to the Microsoft Azure cloud. But the team maintaining it was not versed in product engineering and could only do it on a subset of patients going back two years. Healthcare providers, analysts and researchers across VA need to see the full picture of Veterans and their health data to analyze information and create better outcomes. So, VA chose ºÚÁÏÃÅ to assist them with their data needs.
Ìý
VA meets Data Mesh
Ìý
VA selected ºÚÁÏÃÅ for our agile, hands-on approach and our previous work with them on COVID Patient Manager, VANotify and COVID-19 Chatbot. We introduced the concept of "treating data as a product," a core principle of Data Mesh. Data doesn't sit together in a centralized pool in a Data Mesh. Instead, it's broken down into distinct "data products" owned and managed by the domain teams closest to them. These bespoke products, a relevant, trusted subset of cross-cutting data that contains what a particular group of consumers want, make data highly accessible to the teams that need it.Ìý
Ìý
ºÚÁÏÃÅ paired with the Office of Mental Health and Suicide Prevention to enhance their current MPR calculation to a fully-managed data product, leveraged by multiple customers within the community. Additionally, the data product approach enabled implementation patterns and ensured reliability and consistency across the entire Veteran population.
Ìý
Building the MPR data product provided a verified, trusted single source of truth for calculating MPR across VA. The processing time decreased by 88% and the iteration time to deploy new features was reduced by moving from a manual process to an automated, Continuous Delivery based process, allowing for new features to be delivered quickly and easily. The MPR data product acts as a template for potential new data products in the future.Ìý
Ìý
A new data-driven mission emerges
Ìý
With the new MPR data product in place, another challenge came to light: the lack of a central repository where a user could search and find data assets for reporting or analysis. Data products need to be discoverable by data analysts and scientists to enable data-driven decision-making, and data stewards need an integrated view to manage metadata.Ìý
Ìý
To address these needs, VA engaged ºÚÁÏÃÅ to implement a data catalog for their Health Data and Analytics Platform (HDAP), which ingests data assets to run machine learning models. A data catalog enables consumers to easily find and interpret the data, content, metadata and data models available for analysis to extract business value, thereby improving patient health outcomes.Ìý
Ìý
ºÚÁÏÃÅ worked closely with VA stakeholders to define the need for a data catalog and carved out a thin slice for launching the initial proof of concept. This initial proof of concept included the addition of an existing data product in HDAP to the data catalog and capturing all the minimum metadata requirements as governed by the VA Data Governance Council and the Department of Defense.Ìý
Ìý
ºÚÁÏÃÅ approached the problem with research on a third-party data catalog tool, Collibra, that VA had already procured. Collibra enables organizations to find, analyze and trust their data — it also brings visibility into an organization's data landscape by bringing metadata into a central location, making data discoverable for analysis and business decision-making. Then, we assessed the Gartner Metadata management core capabilities alongside desired VA functionality outlined and evaluated that Collibra can satisfy these functions.
Ìý
New products and a platform for the future
Ìý
The new MPR data product became the first data product inside the HDAP Azure platform. Consumers can now come in, request their own space in the environment, pull data in and run analytics and machine learning models. Additionally, the MPR data product and data catalog enable VA to create a searchable inventory of enterprise data assets. Now, users of the data catalog experience reduced time to discover insights, they understand the quality of the metadata, and they can easily add new metadata to the catalog. ºÚÁÏÃÅ will continue onboarding additional data products and other critical data assets while expanding Collibra's functionality to include data lineage and other features.
Ìý