Patient De-Identification Viewer

MIS 752 Lab 2 — Building a patient cohort, de-identifying patient information, and measuring re-identification risk.

Cohort Funnel

Filter	Patients Remaining
All patients	5,000
Female	2,470
Female, age 65+	622
Female, age 65+ with diabetes/CKD	3

De-Identification

Direct identifiers were removed from the analytic data. Exact age was generalized into a 10-year age band, and ZIP code was reduced to the first three digits.

Original Information	De-Identified Version
Patient ID	REDACTED
Name	REDACTED
Birthdate	REDACTED
Exact age: 66	Age band: 60-69
Full ZIP code	ZIP3: 232

Re-Identification Risk

Quasi-identifiers: AGE_BAND, GENDER, RACE, ZIP3

Minimum k: 1
Uniquely exposed records: 980 of 5,000
Exemplar patient's k: 2

AT RISK (k < 5)
The exemplar shares these quasi-identifiers with only one other individual. Additional generalization would be needed to reach the lab's k ≥ 5 safety threshold.

Privacy Takeaway

Removing direct identifiers does not automatically make a dataset anonymous. Combinations of generalized characteristics can still make an individual distinguishable, demonstrating the trade-off between preserving useful analytic information and protecting patient privacy.