Calibrating Microsimulations with Multicriteria Optimization Algorithms

• Activity Number: 323
• Type: Invited
• Date/Time: Tuesday, August 3, 2010 : 10:30 AM to 12:20 PM
• Sponsor: WNAR
• Author(s): Chung Yin Kong*+
• Companies: Massachusetts General Hospital
• Keywords: simulated annealing ; lung cancer

Abstract

To calibrate our Lung Cancer Policy Model (LCPM), we applied two different approaches, weighted-sum and Pareto frontier. Using the weighted-sum approach, two automated parameter search algorithms, Simulated Annealing (SA) and Genetic Algorithm (GA), were independently applied to perform the parameter search. Both search algorithms obtained total GOF scores below 95 within 1,000 search iterations. Results indicated SA outperformed GA in locating lower GOF values. Using the Pareto frontier approach, we investigated whether the incorporation of bronchioloalveolar carcinoma (BAC) as a subset of adenocarcinoma would improve calibration fits. Results indicated that the LCPM produced a better fit to the age-specific lung cancer incidence rate with incorporation of BAC. Concomitantly, adding BAC did not affect the fit of the histological cell type distribution.