A Bayesian Analysis on Historical Clinical Data Concerning Treatment Change for HIV/AIDS Patients

Kasmiran, K. A., Zomaya, A. Y., Mazari, A. A. and Garsua, R. J.

The advances of biomedical informatics have allowed considerable amounts of historical data to be recorded about an HIV/AIDS patient. By a Bayesian approach coupled with a statistical hierarchical analysis, we evaluate the significance of historical viral load and CD4 data on the HIV treatment change decision. We show that, while there is correlation between the decision and the historical data up to 3 years before the decision, historical viral load data at 12, 18 and 24 months before the decision and historical CD4 data at 30 and 36 months before the decision provide little additional predictive power to what is provided by more recent data. Our results suggest that in a resource-constrained situation, an abridged historical viral load and CD4 archive can provide comparable prognostic support to a fully-populated archive with significantly reduced storage costs. .

Cite as: Kasmiran, K. A., Zomaya, A. Y., Mazari, A. A. and Garsua, R. J. (2012). A Bayesian Analysis on Historical Clinical Data Concerning Treatment Change for HIV/AIDS Patients. In Proc. Australasian Workshop on Health Informatics and Knowledge Management (HIKM 2012) Melbourne, Australia. CRPIT, 129. Butler-Henderson, K. and Gray, K. Eds., ACS. 41-48

(from crpit.com) (local if available)