This post was adapted from a commentary published on Medium.
When the California Health Care Foundation launched the California Health Data Project last winter, my fellow Sacramento Health Data Ambassador, Joel Riphagen, and I were still brainstorming ways to connect our local health community with efforts by the California Health and Human Services Agency (CHHS) to open its data.
At the CHHS Data Fest in March, we heard from many open-data evangelists who spoke of efforts to free health data across the globe. One of them was Damon Davis, director of the US Health and Human Services Agency (HHS) Health Data Initiative. Davis works in HHS's IDEA Lab, an incubator for innovation, and shared with us the Demand-Driven Open Data project led by David Portnoy, one of their entrepreneurs-in-residence.
Portnoy describes the approach this way: "Demand-Driven Open Data (DDOD) is a framework of tools and methodologies to provide a systematic, ongoing, and transparent mechanism for you to tell public data owners what's most valuable."
This elegant concept perfectly captures our role as health data ambassadors. We're connecting data users with data publishers to help those publishers prioritize the data that's most important to users. And DDOD shows us how to engineer an open-data feedback loop for maximum access and transparency.
How It Works
Like any user-centric approach, DDOD starts with user needs. Requests for data are not framed in terms of specific data sets but rather as use cases that articulate the value of a data set. Because DDOD is a highly structured process, Portnoy developed a template to delineate use cases in the following format:
- Use case summary
- Current data and limitation
- Short-term workaround
- Long-term implementation
Those use cases are then tracked through a set of milestones that moves the use case from a posted need to an implemented (and documented) solution:
Each milestone belongs to one of three roles: the data user, the DDOD administrator, and the data owner. In the DDOD framework, the DDOD administrator plays the role of mediator between the data user and the data owner.
How does DDOD work in practice? Procured Health, a company that makes software for hospital supply chains, submitted a use case based on its need to have a "single source of truth" about medical device recalls. The Food and Drug Administration (FDA) caused a problem by using two different data sources to report medical device recalls. The two sources were inconsistent and difficult to reconcile with one other because the datasets were missing unique identifiers that would allow them to be linked, the company said. Following the DDOD process, Portnoy as DDOD administrator helped solve this problem. The FDA is scheduled to release a solution this summer, and best of all, a transparent activity log allows users to track the progress of requests.
Portnoy deployed free and open source tools to pilot the DDOD process, including the use of a wiki, Google Docs, and GitHub Issues. These off-the-shelf technologies can empower innovators to quickly prototype their ideas, test them with real users, and iterate in an agile way. This process of rapid learning helps validate assumptions about the value of a product or process prior to investing in custom software development. Ultimately, however, DDOD will need a more integrated, user-friendly approach in order to scale.
Adopting DDOD in California
One of the more appealing parts of DDOD is that it moves us closer to a world in which interfaces to government can be agnostic as to which agency, department, or level of government is responsible for a given program (or data set in this case). That is, we can easily imagine a DDOD software solution that allows users to enter their use cases on the front end and have those requests routed to the responsible agency on the back end, regardless of the level of government. Local health data stakeholders have consistently expressed that health data being warehoused in myriad locations on the Internet is a source of friction that interferes with putting data to good use.
In fact, Portnoy has posed the question, "What improvements in population health or acceleration of medical discoveries are possible with . . . federal plus state plus local data?" to spur more interagency collaboration around open health data. In response, we asked a local health data stakeholder and nonprofit consultancy, Valley Vision, to submit a use case: The publication of community health needs assessments in the Sacramento region. And our colleagues at CHHS have expressed interest in piloting DDOD at the state level.
As local health data ambassadors, we stand ready to help move these conversations forward and provide the technical assistance necessary to design the technology platforms to make DDOD frictionless. And who knows — this could open doors to something even more ambitious, such as two-way open-data publishing (state to local, local to state) on a single portal. We do know this: A demand-driven approach will be focused on user needs, and in a world of increasingly abundant data, that's something to celebrate.