| Printable Copy |
 |
| Printable Copy |
|
Am J Trop Med Hyg. 2018 May;98(5):1489-1497. doi: 10.4269/ajtmh.17-0218. Epub 2018 Mar 8.
Graham M12 Suk JE34 Takahashi S5 Metcalf CJ65 Jimenez AP34 Prikazsky V34 Ferrari MJ78 Lessler J2.
We report on and evaluate the process and findings of a real-time modeling exercise in response to an outbreak of measles in Lola prefecture Guinea in early 2015 in the wake of the Ebola crisis. Multiple statistical methods for the estimation of the size of the susceptible (i.e. unvaccinated) population were applied to weekly reported measles case data on seven subprefectures throughout Lola. Stochastic compartmental models were used to project future measles incidence in each subprefecture in both an initial and a follow-up iteration of forecasting. Measles susceptibility among 1- to 5-year-olds was estimated to be between 24% and 43% at the beginning of the outbreak. Based on this high baseline susceptibility initial projections forecasted a large outbreak occurring over approximately 10 weeks and infecting 40 children per 1000. Subsequent forecasts based on updated data mitigated this initial projection but still predicted a significant outbreak. A catch-up vaccination campaign took place at the same time as this second forecast and measles cases quickly receded. Of note case reports used to fit models changed significantly between forecast rounds. Model-based projections of both current population risk and future incidence can help in setting priorities and planning during an outbreak response. A swiftly changing situation on the ground coupled with data uncertainties and the need to adjust standard analytical approaches to deal with sparse data presents significant challenges. Appropriate presentation of results as planning scenarios as well as presentations of uncertainty and two-way communication is essential to the effective use of modeling studies in outbreak response.
PMID: 29532773 PMCID: PMC5953353 DOI: 10.4269/ajtmh.17-0218