Assuming that the differential privacy parameter epsilon is less than 0.001, would you be willing to release health information to researchers?