Modeling the impact of alternative immunization strategies: using matrices as memory lanes

Journal article

Existing modeling approaches are divided between a focus on the constitutive (micro) elements of systems or on higher (macro) organization levels. Micro-level models enable consideration of individual histories and interactions, but can be unstable and subject to cumulative errors. Macro-level models focus on average population properties, but may hide relevant heterogeneity at the micro-scale. We present a framework that integrates both approaches through the use of temporally structured matrices that can take large numbers of variables into account. Matrices are composed of several bidimensional (time×age) grids, each representing a state (e.g. physiological, immunological, socio-demographic). Time and age are primary indices linking grids. These matrices preserve the entire history of all population strata and enable the use of historical events, parameters and states dynamically in the modeling process. This framework is applicable across fields, but particularly suitable to simulate the impact of alternative immunization policies. We demonstrate the framework by examining alternative strategies to accelerate measles elimination in 15 developing countries. The model recaptured long-endorsed policies in measles control, showing that where a single routine measles-containing vaccine is employed with low coverage, any improvement in coverage is more effective than a second dose. It also identified an opportunity to save thousands of lives in India at attractively low costs through the implementation of supplementary immunization campaigns. The flexibility of the approach presented enables estimating the effectiveness of different immunization policies in highly complex contexts involving multiple and historical influences from different hierarchical levels.

Authors: Alonso, WJ; Rabaa, MA; Giglio, R; Miller, MA; Schuck-Paim, C

• Publication status: Published
• Peer review status: Peer reviewed
• Publisher: Public Library of Science
• Journal: PLoS One
• Volume: 10
• Issue: 10
• Article number: e0141147
• Publication date: 2015-10-28
• Acceptance date: 2015-10-04
• DOI: 10.1371/journal.pone.0141147
• ISSN: 1932-6203
• Language: English