Probabilistic Matching of Deidentified Data From a Trauma Registry and a Traumatic Brain Injury Model System Center: A Follow-up Validation Study.
Raj G Kumar, Zhensheng Wang, Matthew R Kesinger and 5 others
PMID 29557888WHAT IT FOUND
A trauma registry and TBI rehabilitation database matching algorithm linked most pairs as the same person, but it missed many people who should have linked.
It validates data linkage, not a treatment.
Key findings
01With the middle matching rule, 98.2% of linked training pairs and 97.4% of linked validation pairs were the same person, but 74.1% and 76.3% of true matches were found.
02With the strictest rule, 99.3% of linked training pairs and 98.0% of linked validation pairs were correct, while only 66.6% and 70.7% of true matches were found.
03The study validated a data-linkage algorithm at one independent medical center, not a clinical treatment or patient outcome.
STILL TO COME
How it was doneWhat they found
Read the rest of this summary
You get three full summaries a month, free, and we do not ask for a card. Search, the TL;DRs and your library stay unlimited either way.
What it does not show
The study used one independent medical center, so the matching values may not transfer to other sites. The linked records came from 1999 to 2012, and some matching variables, such as systolic blood pressure and respiratory rate, are no longer collected in TBIMS after 2013. The algorithm missed a substantial share of true matches under the reported tiers. The paper reports data-linkage accuracy metrics, not patient outcomes, treatment effects, or guidance for therapists. The study had known true match status because medical record numbers were available, which will not be true in a national deidentified merge. Other variables with better data quality or lower random agreement may exist but were not tested.
The easy way to misread this
Do not read this as evidence that linked trauma and rehabilitation records can guide individual patient care. The study tested a matching algorithm at one site with known true match status, and it did not measure patient outcomes.