DisplayTitle

A Statistical Model to Detect Asymptomatic Infectious Individuals with an Application in the Phytophthora alni-Induced Alder Decline

First, third, fourth, fifth, and sixth authors: INRA, UMR1136 Interaction Arbres/Microorganismes, F-54280 Champenoux, France and Nancy-Université, UMR1136 Interaction Arbres/Microorganismes, F-54000 Nancy, France; and first and second authors: INRA, Agro-ParisTech, UMR1092 LERFoB, F-54042 Nancy, France.

In some diseases—in particular, tree root infection—stages of infection and inoculum production level and timing are not readily observable because of uncertainty or time lags in symptom appearance. Here, we pose a criterion, based on relative hazard of disease symptoms, to discriminate between healthy and asymptomatic infected individuals. We design a statistical procedure to estimate the criterion for a 6-year survey of alder decline along a northeastern French river. Individual tree symptom hazard was modeled with Cox's regression model, taking estimation of local infection pressure as a risk factor. From an inoculum production experiment, we thereafter assessed the inoculum production level of target trees, including symptomatic and asymptomatic trees ranked according to their symptoms hazard. Using receiver operating characteristic methods, we first evaluated the criterion performance and determined the discrimination threshold to sort out asymptomatic individuals into healthy and infected. Then, we highlighted the fact that the infected asymptomatic trees were among the major inoculum producers whereas severely declining and dead trees were found to be poor inoculum sources.

Additional keywords: spatial point pattern analysis, survival analysis.