Imperfect Data Linkage

One of the assumptions of capture-recapture methods is that there is no matching error. Imperfect data linkage includes erroneous linking of notifications on different individuals (false positive matches) or failure to identify the notifications of the same individuals (false negative matches). False negative matches will lead to the overestimation of the case counts while false positive matches will cause the correction for under ascertainment procedure remain incomplete (Brenner). Usually the tools to eliminate the false negative matches will increase the number of false positive matches, and vise versa. However, additional information will always help to minimize the problem. To reduce the false negative matches, never use exact matching procedure alone as typing errors or coding mistakes usually occurred during the data collection procedure. When use probabilistic matching, appropriate weight selection for each variable could also minimize the imperfect data linkage problem. In the area when personal identification is not available or it changes from time to time (e.g. in Tanzania or other developing countries), the investigators are encouraged to collect as much information as possible and apply probabilistic matching technique for better data linkage.