CDT Report on Distributed Models for Health Data Analysis | Markle | Advancing America's Future
CDT Report on Distributed Models for Health Data Analysis | Markle | Advancing America's Future

CDT Report on Distributed Models for Health Data Analysis

The Center for Democracy & Technology describes privacy-protective approaches for analyzing health data.

It is now well understood that tracking and aggregating population-level health data can help improve decisions on public safety, cost, quality, and outcomes of care. While bringing together and analyzing information on a large scale can be very informative, there are also potential privacy risks associated with collecting and processing personal data maintained in large centralized databases.

To reduce these privacy risks, the  Markle Connecting for Health Common Framework for Private and Secure Health Information and the Markle Decision-Making for Population Health “First Principles” emphasize an approach in which detailed personal data remains local with the data holders and is shared based on the core tenets of Fair Information Practice Principles (FIPPs). These principles, among other things, require that the purpose of the data being shared is specified and only the minimum necessary data is shared. In this distributed approach, whenever possible, de-identified data is shared across a network to answer specific population health questions.

In the Center for Democracy & Technology’s recent report Decentralizing the Analysis of Health Data the authors describe technical approaches that, in conjunction with a FIPPs-based policy approach, align with many of the Markle “First Principles” for population health. For example, one technical approach involves researchers querying distributed data sources that subsequently provide answers through structured responses—including aggregate information—rather than provide access to complete copies of patient information. The paper highlights various examples where such distributed models are being tested.

Ultimately the paper recommends that these models should continue to be developed, tested and implemented where appropriate and effective—a recommendation also made by Markle’s managing director, Carol C. Diamond, and Farzad Mostashari and Clay Shirky, in their 2009 Health Affairs article titled Collecting and Sharing Data for Population Health: A New Paradigm. As stated in the article, “We believe that this sort of experimentation is not only valuable but vital. However, networked architectures are not a silver bullet. Some problems will be of such national and speculative scope that advance aggregation of information into a single database is the best technical fit. But when networked architectures can be deployed, this can lower costs, make better use of existing technical infrastructure, and help avoid the most significant challenges to the current centralized model. To make such a federated environment work, we will need a set of clearly delineated information policies and technical standards derived from the attributes of the Connecting for Health Common Framework that will lay the foundation for an environment of trusted population health data sharing.”