We need national health data protocols—not another national health network

As I passively scroll through aspirational LinkedIn posts about the promising future of TEFCA, AI evangelists, and newsletters perpetuating the same hot topics (What to expect in 2024), my frustration persists.

The fervor for yet another national healthcare network fuels my amorphous perplexation as I ask why there is no focus on better leveraging the data in existing networks. Is there a need for another network? A recent survey of Health Information Organizations (HIOs) stated that 93% already work with at least one national network (e-Health Exchange, DirectTrust...etc), many with at least two. Additionally, a GAO study recently cited that over 50% of rural based clinics and small hospitals in the US remain unconnected to any HIO network, leveraging fax and mail—what will more networks accomplish for them?

Perhaps we should abate the approach of adding more systems on top of other systems and this one size fits all mentality. Imagine if cellular providers said, everyone needs to buy a new phone since we switched the network to 5G. Even the Internet had to progress from v4 to v6, and do so in a way that did not require the world to build a ‘new’ Internet, although many wish they could since internet protocols were never intended to be used in the ways they are today (requiring the need for security layers, explicit routing...etc).

Many HIOs have advocated for Public Health (PH) partnerships and TEFCA allows for PH access as a use case. However, the most significant challenge PH staff would face retrieving data from HIOs is one of accountability. PH agencies must be able to make decisions on behalf of their populations based on methodically curated data sets. Regardless of how forthcoming or prevalent HIOs may be, there are just so many decisions involving what data to include, or exclude, that can influence insights. Furthermore, the larger the data sets, and the more exchanges data passes through, the more distorted it will become.

So what about data protocols?

A few years ago I ran a clinical lab in California. In Vitro Diagnostics (IVD) laboratory tests are considered medical devices by the Food and Drug Administration (FDA). There is a regulated process to validate laboratory tests which requires an organization to submit performance results for approval to the FDA. Furthemore, the laboratory environment must conform to specific standards (CLIA) to even conduct the test.

The takeaway from this process is that no national laboratory network needs to be built to submit test data across different networks. A COVID test processed by Quest, LabCorp, or any CLIA lab is enough for a provider to make a consistent, confident diagnosis.

Now—if you received disease data rates from every state HIE, would you have that same level of confidence?

If you’re in Public Health, you are probably going to ask about data sources and how it was transformed—and that’s exactly the point.  There is no one process to transform data, it depends on organizational policies, training, and data quality. So, if we think about TEFCA, or any other large distributed network, where could these transformative inconsistencies occur?

• Patient identity management: The minimum data sets to match and merge patients can differ depending on the vendor. The way organizations merge and exclude data can lead to significant differences in patient totals, observations, and demographics. A study conducted using HIE identifiers attempted to estimate disease prevalence in health systems in Denver, CO and noted 3.5% duplication in a population set of just ~218k, although this number could vary based on reconciliation strategy.
• Patient provided data: One of the major challenges with demographic data is the quality and completeness of race and ethnicity data. A study conducted in a New York data set of 2.4 million suggests that 86% of patient's self-report their race and ethnicity compared to the 43% that is recorded in EHRs. Does there need to be better mechanisms and procedures to account for this data, particularly with the increase of Patient APIs.
• Comprehensiveness: In any volunteer participation model there needs to be a way to account for the percentage of participation from some base unit. For instance, if we consider a county in a state a base unit, we could then know that 12 county Public Health organizations are reporting out of 15. This process ensures that data can still be informative, even if everyone is not on a network, since you can still account for rates within given reporting areas.
• Timeliness: If we think about data flow from a provider to a state registry, or Public Health organization to a regional HIO, and perhaps a HIO to an even larger HIO (Perhaps QHIN), there will be a reporting lag. This lag is unavoidable, but where we draw the line is critical. For instance, designating data pulls at midnight Eastern Standard Time (EST).  The important concept is maintaining consistent reference points for analysis purposes.

While a comprehensive healthcare data ecosystem is an aspirational goal, we do not need to wait to have aggregated, quality data reported at the national level. We just have to have better processes and protocols in place for the data we do collect, particularly for Public Health where data transparency will continue to be critical to ensuring public trust. Lastly, this is not something that needs to be overengineered (although it likely will be), there are dozens of ways to record business finances, the key is consistency for year over year analysis. The order of operations for mathematics is hardcoded in our systems allowing for the universal, consistent application of equations. Thinking of data protocols like laboratory tests and HIOs like qualified labs ensures that HIOs still have the flexibility to utilize their proprietary data transformations and features, but also have the fundamental infrastructure and procedures to support data aggregation for key national data requirements.

Or...we just build another network.