Two de-identification methods, k-anonymization and adding a 'fuzzy factor,' significantly reduced the risk of re-identification of patients in a dataset of 5 million patient records from a large cervical cancer screening program in Norway.