Purpose - This paper investigates, through the use of data mining techniques, patterns in HIV/AIDS patient data. These patterns can be used for better management of the disease and more appropriate targeting of resources. Design/methodology/approach - A total of 250,000 anonymised records from HIV/AIDS patients in Thailand were imported into a database. IBM's Intelligent Miner was used for clustering and association rule discovery. Findings - Clustering highlighted groups of patients with common characteristics and also errors in data. Association rules identified associations that were not expected in the data and were different from traditional reporting mechanisms utilised by medical practitioners. It also allowed the identification of symptoms that co-exist or are precursors of other symptoms. Originality/value - Identification of symptoms that are precursors of other symptoms can allow the targeting of the former so that the later symptoms can be avoided. This study shows that providing a pragmatic and targeted approach to the management of resources available for HIV/AIDS treatment can provide a much better service, while at the same time reducing the expense of that service. This study can also be used as a means of implementing a quality monitoring system to target available resources. |