The feasibility and accuracy of anonymized record linkage to estimate shared clientele among three health and social service agencies.

This study was designed to test the effectiveness of three computerized methods of record linkage, using minimal identifiers. The goal of the linkage was to determine the proportion of persons receiving Public Health Nursing Services (n = 5,749) who also receive income assistance from the region (n = 38,800) or the province (n = 16,741). The first linkage method was based on full agreement of a number of variables (deterministic matching) and the second was based on the likelihood of the match (probabilistic matching). The third method combined information from methods 1 and 2. Method 3 proved to be the most effective linkage technique, with a specificity of 0.98 and sensitivity of 0.94. The linkage showed that only 17% of the Public Health clientele received income assistance from the region or province, much lower than expected. Factors which have probably led to an underestimate are discussed.