# Decades-old anonymized medical data may cause AI misdiagnoses, study finds

Digest AI · Research · published 2026-09-18T12:00:53Z

Canonical: https://digestai.news/story/decades-old-anonymized-medical-data-may-cause-ai-misdiagnoses-study-fi

## Summary

A new collaboration between German and UK researchers warns that anonymized patient records from decades ago could lead AI systems to treat a 2026 patient as if they were still in the year the record first entered the dataset. The paper, titled *Memorisation bias in medical AI*, explains that a model that has seen a patient’s historical data may bias predictions toward that past health state, potentially producing false positives for conditions that have resolved or masking new illnesses.

The authors evaluated four large medical datasets – MIMIC‑CXR, MIMIC‑IV‑ED, MIMIC‑ECG and the HEEDB collection – covering from tens of thousands up to more than 1.8 million patients. In some cases, predicted probabilities shifted by more than 70 percentage points, and the effect persisted for over 25 years in the HEEDB ECG data. Tests showed lower sensitivity for new conditions, while accuracy improved when the patient’s health had not changed. Applying differential privacy during training reduced the bias, and patient‑level differential privacy almost eliminated it, though it cannot be used on already anonymised data.

The researchers note that as model capacity grows under scaling laws, the absolute number of individuals affected by memorisation bias could rise sharply, raising safety and regulatory concerns for medical AI deployments that rely on nationally sourced data.

## Key points

- Four datasets (MIMIC‑CXR, MIMIC‑IV‑ED, MIMIC‑ECG, HEEDB) covering up to 1.8 million patients were tested.
- Predicted probability changes exceeded 70 percentage points and persisted over 25 years in some cases.
- Patient‑level differential privacy almost eliminated memorisation bias, while standard DP only reduced it.

## Why it matters

If memorisation bias spreads as models scale, medical AI could misdiagnose patients, undermining safety and compliance with emerging AI regulations.

## Sources

1. [Decades-Old Anonymized Medical Data May Cause AI Misdiagnoses Now](https://unite.ai/decades-old-anonymized-medical-data-may-cause-ai-misdiagnoses-now) (Unite.AI, 2026-09-18)

## Cite

Digest AI, "Decades-old anonymized medical data may cause AI misdiagnoses, study finds", 18 September 2026, https://digestai.news/story/decades-old-anonymized-medical-data-may-cause-ai-misdiagnoses-study-fi

---

Written by Digest AI's editorial model from the linked sources; the sources are the record. Headlines, digests and key points are written by Digest AI and may be quoted with a link to the story page. Linked articles belong to their publishers. Terms: https://digestai.news/terms#reuse
JSON: https://digestai.news/story/decades-old-anonymized-medical-data-may-cause-ai-misdiagnoses-study-fi.json