Anonymizing Health Data

Chapter 6. Longitudinal Events Data: A Disaster Registry

Complex health data sets contain information about patients over periods of time. A person’s medical history can be taken as a series of events: when they were first diagnosed with a disease, when they received treatment, when they were admitted to an emergency department. Our case study for this chapter is the World Trade Center (WTC) disaster registry. The collapse of the Twin Towers at the World Trade Center is one of those unforgettable disasters that has deeply affected many people, both locally and abroad. The Clinical Center of Excellence at Mount Sinai has been following WTC workers and volunteers for years in an effort to provide them with the best treatment for the physical and mental health conditions related to their WTC service.^[63] As the saying goes, you can’t manage what you don’t measure.

Note

We’ll be using the term “events” to refer to a patient’s medical information in this discussion because of the structure of the WTC data. But the methods we explore here apply equally well to other data sets, such as discharge data or claims data. When we get to the methods themselves, we’ll point out how they apply to these other scenarios, if the differences need to be highlighted.

In looking at this longitudinal disaster registry, we need to revisit the assumptions we make about what an adversary can know to re-identify patients in a data set. We started with an adversary that knows all, then relaxed some of our assumptions ...