A former employee of the Institute for Animal Health (England) commented:
"If you've got a problem with the biology of a disease, you don't go throwing maths at it - you look at the biology."
A couple of relevant questions have been mooted with respect to biomodels:
"If biomodels are models, and models are maths, and maths should be used cautiously and carefully in biology... then what exactly are biomodels, and how are they being used for disease control?"
The explanation stems from the different ways in which (1) first generation mathematical disease models are created, compared to the ways in which (2) second generation biomodels are created.
(1) Mathematical (or first generation) disease models divide a population of animals or people, according to their disease states (ie. recovered, infected, susceptible, etc.) [20]. The model then moves individuals between those various disease states, and repeats that movement for each time period (ie. every day, every week, every two weeks, etc.). The dynamic movement of individuals between disease states (and every time period) creates a simulation of the epidemic. Some models add locations or maps to the disease spread. The inaccurate guessing commences when the number of individuals are moved each time period, and the first guess puts a numerical value to a theoretical 'Reproductive Rate' (Ro), or an 'Effective Contact Rate' (ECR). Ro and ECR determine the rate at which individuals pass between the disease states (at each new time period). Ro and ECR are mathematical guesses, and are produced from additional multiple guesses about the biology, such as the length of the infectious period for a disease, and the incubation period for a disease, and the level of population immunity against a disease. It would be correct to conclude that this inaccurate guessing is cumulative.
How then are the first generation models able to mimic what an epidemic looked like [examples:
] in terms of the shape of their (incidence and prevalence) graphs? The answer comes from carefully manipulating the values of Ro, ECR, the incubation period, infectious period, immunity factors, etc., and if it's not possible to create a good match, more 'factors' are added until a good graphical match is created. [The inclusion of these additional factors can be justified later.] It is not difficult to match first generation models to past epidemics. The difficulty for first generation models is to predict what will happen in the future, because these manipulations will unfortunately become useless for predictions. A veterinary surgeon involved with the 2001 UK FMD epidemic noted:
"Veterinary administrators have been duly mesmerized by the aesthetics of modelling graphics, even succumbing to the idea that mathematics is difficult, therefore mathematicians are intelligent, therefore mathematicians have the solutions. That's nonsense, and fortunately someone's proved it [1][24][25]. Epidemiologists have the solutions for epidemiological problems."
(2) Biomodels dispense with disease states, or moving individuals between diseases states, and go directly to the end of an epidemic [22], where prevalence and duration are measured directly: this is possible because the biology has already occured, so there's no need to guess the biology, and it doesn't matter what's occurred within the complex biology [23]. Effectively, guessing has become redundant. The ability to predict prevalence (ie. the final number of infected individuals) and likewise the duration of an epidemic, stems from biomodels factors such as subclinical disease. Subclincial disease spreads before the clinical signs are shown, so when the signs are shown, the prevalence and duration of an epidemic has mostly been predetermined. The control of disease spread still remains possible for individuals that have not become infected with subclinical disease.
By measuring the level of subclinical disease it's possible to both predict prevalence and duration, and then simulate what will happen during an epidemic, before it has happened. This is because the disease has already spread, but the clinical signs have not yet appeared [8]. Biomodel factors such as subclinical disease can also be incorporated into mathematical simulators, to replace Ro or ECR - this allows model simulations to test whether certain disease control measures or contingency plans will be effective in the future.
Measuring subclincal disease (or similar disease factors) to construct biomodels, means that guesswork with mathematics is no longer necessary. Biomodels represent a second generation of disease modelling.
