Questions and Answers

Q. Do biomodels have a problem, because is it possible to know everything that's required for an accurate prediction of disease spread?

A. This would be a relevant question if the disease spread hadn't already occurred, but where it has already occurred the task is to observe the biology that's taken place. FDI (first day incidence) and FFI (first fortnight incidence) have highly significant correlations with disease prevalence and can be monitored accurately. This is true for subclinical disease but not for the other form of disease ie. acute disease. Where biomodels are multi-faceted, they also accommodate the acute disease. In the same data sets that originally identified FDI, environmental temperature (as a stressor had no correlation with subclinical disease prevalence, yet) did have a good correlation with acute disease prevalence. Hence, by monitoring additional correlations, the predictive accuracy can be increased. Significant biological correlations overcome the need to guess infectious periods, incubation periods or biologically meaningless mathematics. Accurate correlations separately can provide accurate prevalence predictions, and accurate correlations together can provide accurate simulations. In turn, applying different control measures to an accurate simulation, will show their relative effectiveness, and thereby indicate which to select.

Q. If a biomodel can simulate what is likely, then what happens when the next epidemic turns out to be unlikely?

A. The randomness of stochasticism has little value in simulations. Hence, it remains important that subclinical disease has already occurred, because variables don't matter once the subclinical transmission has already taken place ie. the variables become history.

Q. Clinical signs could be subclinical in a partially immune or healthy susceptible population, but the disease could be acute in a population with a different concurrent infection. How can a disease biomodel be accurate with so many variables?

A. Variables such as concurrent infections, environmental relative humidity or temperature, or disease challenge levels and population immunity levels, will all change. None of those matter with subclinical disease because the subclinical disease transmission has already occurred ie. what happens before clinical signs appear is the disease history that gets measured as FDI or FFI. FDI and FFI are correlated with prevalence in vaccinated herds as well as in unvaccinated herds, irrespective of this added variable. The part that isn't subclinical ie. acute disease, can be measured against another correlating factor (ie. temperature or whatever else is found to be useful). Generally however, it's the acute disease that's diagnosed early and where it's controlled effectively there's minimal disease, so prevalence from acute disease is largely irrelevant and doesn't need to be predicted. The acute disease needs to be controlled rather than predicted.

Q. The level of (subclinical) disease transmission will vary between different diseases, and underlying conditions in one epidemic may not be relevant for a different epidemic or a different disease. How can a disease biomodel cater for different diseases?

A. Each disease requires its own biomodel, with observed correlations for prevalence according to what becomes relevant. [The initial correlation with prevalence will be subclinical disease for infectious diseases.] Baylis [22] identified a biomodel in measuring rainfall and the spread of the midge vector against AHS prevalence. Tatem [23] identified a biomodel for Bluetongue. The way to prove the accuracy of a biomodel (and thereby gain administrators' confidence) is not to validate it by looking at the past, but to use it to accurately predict the future.