Identification versus anonymity in e-government STATISTICAL INFORMATION IN THE HEALTH SECTOR

Statistical information in the health sector

The health sector is a very interesting field for collecting and processing information. There is an enormous amount of information generated by patients treated in hospitals, which bill the respective costs to central insurance agencies, etc. Clearly, most of these data is very sensitive, as personal information is almost always integrated therein. Typically, we have data tuples that indicate which person (identified by some number, name, etc.) has had which treatment by which specialized person when and where, and whether the treatment was successful, etc.  

These data are useful for the parties generating them, but also for other parties, for example for health insurers. They typically need substantial amounts of data for different purposes; in particular, identifying data are essential to charge the correct party. 

Another typical application focuses on statistical evaluation of patient data for general purposes. In this context, the identifying data are typically no longer needed. The concept of data mining is usually the central focus, used for example to provide new information computed from the data, or to make forecasts for planning purposes based on both current and historical data. 

Yet often in this very context of data mining and statistics, there is a need to be able to “follow” the same person through different treatments in time and/or space. This is, for example, necessary to gain insight into the success rate of long-term treatments or to compute statistics about the reasons for changing one’s personal medical doctor. One wants to find a possible correlation between certain treatments and either complications or health recovery. Clearly, the simplest implementation of this “following” requirement (linkability) is to keep the identifying parts of the data in the collection. 

At first glance, providing anonymity as well as linkability could seem to be a contradiction. Pseudonyms make it possible to meet both objectives. In this context, we consider pseudonyms calculated by means of cryptographic protocols, as we will show below.