Probabilistic Matching Approach to Link Deidentified Data from a Trauma Registry and a Traumatic Brain Injury Model System Center.
Matthew Ryan Kesinger, Raj Gopalan Kumar, Anne Connelly Ritter and 2 others
PMID 27088479WHAT IT FOUND
Trauma and brain injury rehabilitation records were linked at a single center with high match accuracy and very few false matches.
This is a research method, not a patient treatment finding.
Key findings
01In the algorithm generation set, requiring weight greater than 5 and highest in cluster gave sensitivity 0.89 and positive predictive value 0.98.
02In the validation set, requiring weight greater than 5 and highest in cluster gave sensitivity 0.87 and positive predictive value 0.97.
03In the validation set, also requiring cluster weight difference greater than 5.34 gave sensitivity 0.83 and positive predictive value 0.99.
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 algorithm was developed and validated at a single center using 241 TBIMS cases, so it may not perform the same way in other trauma or rehabilitation databases. Medical record numbers were used to know true matches, but the intended future use is deidentified databases without such identifiers. Restricting the trauma registry to head injury by ICD-9 codes excluded 19 TBIMS patients (7.8%) and made true matches impossible for those cases. The cause of missing data was not determined, and missingness can lead to false negatives. Data quality and coding may differ across centers because registrars and input methods differ.
Declared interests
Research support is listed as U.S. government, non-PHS, and NIH extramural. The supplied text does not state author conflicts of interest.
The easy way to misread this
Do not read this as evidence that linked trauma and rehabilitation records have already changed patient care. The study validated a linkage method at a single center using medical record numbers, and it did not test clinical outcomes or treatment effects.