| Background: The process of identifying record pairs that represent the same entity (duplicate records) is technically known as record linkage and is one of the essential elements of data cleansing. This paper proposes a fast and efficient method for linkage detection within the healthcare domain. The features of the proposed approach are an embedded fast blocking method with a string matching function that accounts for keystroke mistakes made during data entry of a patient's name and the addition of a module that dynamically generates blocks of possible associated and unique records. |
 |
Background: The process of identifying record pairs that represent the same entity (duplicate records) is technically known as record linkage and is one of the essential elements of data cleansing. This paper proposes a fast and efficient method for linkage detection within the healthcare domain. The features of the proposed approach are an embedded fast blocking method with a string matching function that accounts for keystroke mistakes made during data entry of a patient's name and the addition of a module that dynamically generates blocks of possible associated and unique records. Methods: The proposed methods for achieving our goal are as follows. The first step is to standardise the data in the database using SQL (Structured Query Language) statements. The second is to match similar pair records, relying on the discretion of a human expert and observation inferred from the decision support system. The final step is to organise records into match and non-match status. The system was developed in Unified Modeling Language and was coded in Java Programming Language. Results: It was observed that the linkage algorithm with embedded string matching function catering for subtle errors in strings, performs better in duplicate detection (40% recall and 60% precision) than linkage algorithm without a string matching function (90% recall and 10% precision). Conclusion: In this paper we have described an improved record linkage system based on an embedded fast blocking method with a string matching function. The system can be used to improve the data quality and integrity of electronic healthcare records. |