EHRs Can Leak Personal Health Data

By Christine Kern, contributing writer

Gender, race, age, and admission information are all susceptible to leaks.

In light of growing pressure to transition to EHRs and increase interoperability, more and more healthcare providers are putting patient information into electronic databases, and those databases and the information they contain are becoming increasingly valuable to hackers.

To date, emphasis to secure this data has been on the encryption of it to prevent the compromise of protected health information. However, EHR databases may still be in danger of leaking personal information, even when using encryption methods, according to a recently released Microsoft study.

Using patient data from 200 U.S. hospitals, the study found that gender, race, age, and admission information were all susceptible to leaks. The authors determined this by conducting an experiment using four distinct types of cyberattacks.

Using a cumulative attack, researchers recovered data on disease severity, mortality risk, and age, length of stay, admission month, and admission type of at least 80 percent of patients of at least 95 percent of the largest hospitals. With a sorting attack, they were able to recover the admission month and mortality risk of 100 percent of patients for at least 90 percent of the largest hospitals.

The authors also suggest potential exists for more information to be extracted from such databases, highlighting their vulnerability. The researchers also found deterministic encryption could recover more than 60 percent of certain attributes (sex, race, and mortality risk) of patient records.

The authors suggest the amount of recoverable data should be viewed as the lower threshold of what could be recovered, stating, “[One] reason is that the attacks only make use of leakage from the [encrypted database] and do not exploit the considerable amount of leakage that occurs from the queries to the [encrypted database].”

The study concludes the types of encrypted database systems included in the experiment should not be used to store protected health information, given the reality of frequent cyberattacks on networks hosting EHRs and other patient data.

The results will be presented at the Association for Computing Machinery's Conference on Computer and Communications Security next month, according to iHealthbeat.