Automated system enables anonymization of COVID-19 patient data

April 14, 2020.– Over the past two weeks, the Innovation and Technology Transfer Office of the Millennium Institute for Foundational Research on Data (IMFD) has been working full time to develop software that would allow health authorities to automatically anonymize the sensitive data collected from patients and suspected COVID-19 cases.

This program fulfills one of the institute’s commitments as part of its work on the data task force: creating a tool that guarantees the availability of information for the study of the pandemic, while protecting people’s private information.

The system is ready to be used by the health authority and works, broadly speaking, as follows: as with other software, a configuration file must first be created, through which the different sets of required data are defined. This is known as “information aggregation levels.”

Then, the original database collected by the surveillance system is loaded and, using the defined configuration, a dataset is generated. Since different configurations can be created—that is, with different levels of information aggregation—the program will generate a dataset for each one. Finally, each dataset is evaluated to measure the probability that a patient could be identified, which indicates the risks involved in publishing the result.

In addition, operating protocols were defined to ensure that the software operator has no way of leaking the data. For example, the program would be run on a computer with no internet connection, which would receive the encrypted data through physical means such as a USB drive or external hard drive.

The program was developed specifically to meet the needs of the Ministry of Health during this crisis, using information provided by epidemiologists about the fields collected by the surveillance system, and for now will not be adapted for other purposes. Its goal is to support the work of health workers by automating a process that—if done manually—would require a significant investment of human resources’ time, which is scarce in times of crisis like the one we are experiencing, along with a necessary review of the anonymization level of the results.

Martín Ugarte, CEO of the Innovation and Technology Transfer Unit, states that “it is important to clarify that this is not a very complex program; it is a tool that removes certain fields from a database and then calculates classic de-identification measures to ensure people’s privacy. I am certain there are many teams in Chile capable of creating similar tools. Our commitment was to create it as quickly as possible, and our team worked to have it ready in case it was needed.”

Models and visualizations

The reports generated by this software could be used by research centers that have developed models seeking to project the behavior of the coronavirus in Chile.

Some of these, for example, need spatiotemporal information—that is, when and in which municipalities infection peaks have emerged—in order to predict the spread of the virus. The software would immediately provide the data they need without compromising patient privacy.

IMFD leads one of the interdisciplinary teams working on this front, made up of researchers from the fields of epidemiology, public policy, and data science. With the databases that the anonymization program can provide, this team will be able to test virus behavior models and enrich the parameters to obtain more accurate estimates.

Along with this, the Technology Transfer Office is working on developing interactive visualizations of public information about the coronavirus, so that society can directly access relevant statistics and data regarding the COVID-19 outbreak in the country. People would be able to select variables of interest to them and learn, through interactive graphs, for example, the percentage rates of virus incidence at the municipal level or the ages of those affected within certain time periods.

The image shows, as an example and with fictitious data, what this platform would look like; it will be available with higher levels of information in the medium term.

—–

Information: innnovacion@imfd.cl