Repository | Book | Chapter
(2015) Artificial life and computational intelligence, Dordrecht, Springer.
Detecting anomalies in controlled drug prescription data using probabilistic models
Xuelei Hu , Marcus Gallagher , William Loveday , Jason P. Connor , Janet Wiles
pp. 337-349
Opioid analgesic drugs are widely used in pain management and substance dependence treatment. However, these drugs have high potential for misuse and subsequent harm. As a result, their prescribing is monitored and controlled. In Queensland, Australia, the Medicines Regulation and Quality Unit within the state health system maintains a database of prescribing events and uses this data to identify anomalies and provide subsequent support for patients and prescribers. In this study, we consider this task as an unsupervised anomaly detection problem. We use probability density estimation models to describe the distribution of the data over a number of key attributes and use the model to identify anomalies as points with low estimated probability. The results are validated against cases identified by healthcare domain experts. There was strong agreement between cases identified by the models and expert clinical assessment.
Publication details
DOI: 10.1007/978-3-319-14803-8_26
Full citation:
Hu, X. , Gallagher, M. , Loveday, W. , Connor, J. P. , Wiles, J. (2015)., Detecting anomalies in controlled drug prescription data using probabilistic models, in M. Randall (ed.), Artificial life and computational intelligence, Dordrecht, Springer, pp. 337-349.