Privacy-by-design: Case studies in interactive record linkage using a hybrid human-computer system
Bibliographic record
Abstract
OBJECTIVE: High-quality patient matching from several sources without a common identifier (ID) requires interactive record linkage (RL) using a hybrid human-computer system. MiNDFIRL (MInimum Necessary Disclosure For Interactive Record Linkage) is a hybrid prototype software system that facilitates maximizing linkage accuracy while minimizing information disclosure. We present and evaluate MiNDFIRL using two real-world case studies. MATERIALS AND METHODS: Two user studies were conducted linking 10,000 data pairs from EHR data and 18,240 unique patient IDs from patient generated data. After automated RL, manual review was conducted by three teams of four reviewers (12 total) using MiNDFIRL to resolve potential matches that required human judgment. Reviews for matches were conducted independently and disagreements were resolved through consensus. The teams then participated in a group discussion about MiNDFIRL using a semi-structured interview format. RESULTS AND DISCUSSION: The best algorithm, Random Forest, found 388 and 539 matches each for EHR and patient generated data algorithmically, but also output an additional 303 and 187 potential pairs that required manual review. 232 and 84 more matches were confirmed manually from these uncertain pairs respectively. Among the full uncertain pairs, only 30% of available identifying information was needed in MiNDFIRL to separate out 77% (232/303) and 45% (84/187) true linkages respectively. When available, first names and emails were the most frequently used fields in making RL decisions. CONCLUSION: On-demand access and masking techniques along with risk quantification through a hybrid human-computer system can significantly reduce disclosure while still minimizing false positives and false negatives in real-world RL.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.422 | 0.486 |
| Meta-epidemiology (narrow) | 0.001 | 0.002 |
| Meta-epidemiology (broad) | 0.001 | 0.002 |
| Bibliometrics | 0.004 | 0.004 |
| Science and technology studies | 0.005 | 0.010 |
| Scholarly communication | 0.009 | 0.010 |
| Open science | 0.006 | 0.010 |
| Research integrity | 0.007 | 0.002 |
| Insufficient payload (model declined to judge) | 0.003 | 0.001 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".