advertisement
Abstract #24649 Published in IGR 11-4
The application of discrete event simulation to quantitative risk benefit analysis
Maguire A; Douglas I; Blak BTValue in Health 2009; 12: A453
OBJECTIVES: To date, quantitative risk benefit has mainly involved the translation of Cost-Effectiveness techniques or utility adjusted epidemiological statistics. We aim to describe how Discrete Event Simulation nullDESnull offers the possibility of modelling the occurrence of several adverse events and beneficial events simultaneously whilst accounting for competing events. METHODS: Firstly, a longitudinal patient database is used to identify the target patient population. Secondly, incidence rates for the outcomes are calculated from the entire database, thereby providing the necessary granularity in terms of the predictive factors for the outcomes. The annual probability for each outcome is then assigned to each patient in the cohort and DES generates time to each event. Thereby the expected events for an unexposed patient cohort is created to which relative risks are applied to model drug exposure. An example using glaucoma patients is presented using data from The Health Improvement Network. RESULTS: We obtained data on 17,652 glaucoma patients who were known to be receiving glaucoma therapy at January 1, 2007. Patients were characterised according to the principal determinants of the outcomes (heart failure, asthma/COPD exacerbation). The same database provided general population incidence rates for the outcomes which were assigned to each patient according to their characteristics. National statistics provided death rates. The expected events over one year for a cohort of 10,000 glaucoma patients were: HF = 95; asthma/COPD = 143; deaths = 605. CONCLUSIONS: These expected numbers represent the occurrence of events in the natural history cohort. They were obtained by summing the outcome probabilities across the patient group. They do, however, represent the first step in creating a comprehensive method for risk-benefit quantification via DES and large patient databases; benefit can be modelled if expressed as the occurrence of an event. The method will need to incorporate uncertainty in all the input parameters and to update the probabilities after an event has occurred.
A. Maguire. United BioSource Corporation, London, LondonUnited Kingdom.
Classification:
14 Costing studies; pharmacoeconomics
