Over the last two years, public health data systems have been recognized for their ability to provide valuable insight into the pandemic. However, there are many other critical data assets held within state governments. More and more, health information exchanges, Medicaid enterprise systems, all-payer claims databases and others are being included in states’ efforts to modernize data and increase interoperability in order to improve care coordination and quality.
Now is the time to assemble a strong data governance framework that can move beyond point of care and deliver important public and population health analytics in the future.
The process of designing and refining a data governance model should always start with convening, or bringing partners together to solve a problem. Engage a neutral party to ensure every partner has a voice, can define the problem to solve, can identify their own benefits and motivators, and understands the risks and benefits of possible solutions.
This is a critical opportunity to build trust, especially around health equity. Systemic racism has led to distrust around data sharing for many populations. State agencies have been siloed. By doing the important and challenging work of engaging systemically disadvantaged populations with multiple state agencies and other community partners before addressing the specifics of a data governance model, a neutral facilitator can help state agencies and their partners demonstrate their trustworthiness.
Convening takes time. It can be uncomfortable. There will be disagreements. But without a foundation of trust, any data governance model will underperform at best, and at worst, will fail entirely.
Next, examine use cases as well as human and technological drivers and barriers, and define what is in and out of scope for a given data governance model. For example, data categories and data value sets (e.g., enumeration of ethnicities and race) often have external requirements that limit flexibility, so redefining data categories can be out of scope.
Data uses are well within scope of most data governance frameworks, and increasingly include tools like artificial intelligence, machine learning and algorithms (and algorithmic bias). In all data uses, data owners and users are responsible for ensuring their analytics don’t contribute to disparities. For example, should race data be used with claims and clinical data to predict which patients will be readmitted to the hospital, or to prescribe which patients receive additional care coordination services? Using historical data that reflect inequities in health care utilization and quality for predictive or prescriptive analytics can perpetuate health inequities, while comparison between populations can illuminate otherwise hidden disparities.
With a trusting collaborative of partners in place and use cases identified, it is now time to design a data governance framework. I do not recommend a particular structure because there is no single best solution. Whatever model you use should specify the partners with a seat at the table and define each partner’s role(s) in the strategic, tactical and/or operational responsibilities. The best governance models maintain a degree of flexibility around the architecture of the data. At its core, any data governance framework should center the community and people with lived experience.
This moment of data modernization and interoperability is an opportunity to establish strong, collaborative data governance frameworks that center patients and enable important public and population health analytics in the future. While there is no single best model, every successful model will start with convening to build trust and a thorough examination and understanding of use cases.