Methods to accurately link records pertaining to the same individual in different source files have increasing application in comparative effectiveness research (CER) and patient-centered outcomes research (PCOR). Linking data from a CER/PCOR study with additional externally collected data can enable researchers to measure additional covariates and thereby reduce confounding in observational studies and can capture long-term outcomes without the expense of following patients longitudinally. The main technical challenge of such linkage is the absence of error-free unique identifiers with which to identify records belonging to the same individual across source files. Powerful computer-based techniques have been developed to link records on the basis of error-prone indirect identifiers, but customized methods and guidance for linking data in CER/PCOR studies are lacking. Because linkage requires the collection and storage of potentially identifying patient information, the risks to patient privacy are necessarily greater compared with CER/PCOR studies using completely deidentified patient data. Both the primary beneficiaries and the bearers of the associated risks of CER/PCOR use record linkage, so there is a need to understand the patient perspective on its risks and benefits.
