64 News & Comment TRENDS in Ecology & Evolution Vol.17 No.

2 February 2002

Letters

neither easily defined nor easily measured. independent of the size of the population. Prev.
Modelling McCallum et al. [1] discuss density
4
Vet. Med. 23, 163–172
Greenwood, M. et al. (1936) Experimental
transmission: mass dependence as though these problems do not
exist. Can transmission ever be modelled
Epidemiology, MRC Special Report Series
5 Anderson, R.M. and May, R.M. (1979) Population
action and beyond as depending on numbers-per-unit-area in biology of infectious diseases: Part I. Nature
a reliable way? Ignoring the discussion on 280, 361–367
6 De Koeijer, A. et al. (1998) Modelling the spread of
In a recent review in TREE [1], McCallum how density is measured and introducing
Phocine Distemper Virus among harbour seals.
et al. say that they are confused by the ad hoc (i.e. not mechanistically derived) Bull. Math. Biol. 60, 585–596
terminology related to mass action in transmission functions based on density 7 Dietz, K. (1982) Overall population patterns in the
disease transmission that we introduced does not necessarily help us to better transmission cycle of infectious disease agents. In
in Ref. [2]. In retrospect, we see that our understand transmission. One mechanistic Population Biology of Infectious Diseases (Anderson,
R.M. and May, R.M., eds), pp. 87–102, Springer-Verlag
description of the expression ‘βSI’ (where argument discussed by McCallum et al., the
8 Heesterbeek, J.A.P. and Metz, J.A.J. (1993) The
S is the number of susceptible hosts, I is Holling type argument, was first introduced saturating contact rate in marriage and epidemic
the number of infected hosts and β is the into an epidemic context by Dietz [7], and models. J. Math. Biol. 31, 529–539
transmission coefficient) using the words shown to be problematic in this same 9 Diekmann, O. et al. (1998) A deterministic epidemic
‘pseudo mass-action’ (thus implying that context by Heesterbeek and Metz [8] model taking account of repeated contacts between
the same individuals. J. Appl. Prob. 35, 448–462
the description is not a correct rendering (quoted in Ref. [1] for another reason). 10 Diekmann, O. and Heesterbeek, J.A.P. (2000)
of mass action) might be confusing when The key issue is how to model the Mathematical Epidemiology of Infectious
one does not restrict this, as we did, to a number of contacts with other individuals Diseases: Model Building, Analysis and
case where S and I are numbers. per unit of time per individual. Instead of Interpretation, John Wiley & Sons
The relationship between the intensity modifying the transmission function, it
of contacts between individuals on the one could be worthwhile to look at other
hand and population size on the other, is ways in which contact structures can be
a complicated issue that deserves more incorporated in models. For example, for Modelling
attention and a reviewing progress, as farmed animals, the whole study population transmission: mass
in Ref. [1], is therefore to be applauded. can be viewed as a metapopulation of
Several fundamental problems with density groups of animals, where within-group action and beyond
were, however, neglected, particularly the transmission is modelled best by mass
question of how to define ‘density’for a action and between-group transmission by Response from McCallum, Barlow and Hone
natural population. Behavioural patterns, explicit use of connection matrices [9]. See
as well as heterogeneity of the environment, also Ref. [10] for elaborations on several of In our recent article in TREE [1], we were
create nonhomogeneous distributions of the issues raised here. very careful to say that De Jong et al. [2]
individuals that are not accurately had been ‘widely interpreted as claiming
characterized by simply dividing the Mart C.M. De Jong* that βSI did not represent “true mass
number of animals by the total habitable Annemarie Bouma action”’. It is the way in which their paper
area. In populations that are fluctuating in Quantitative Veterinary Epidemiology has been interpreted, rather than the paper
size, one is also faced with the question of Group, ID-Lelystad PO Box 65, itself, that is the problem. Their paper is
how changes in numbers are related to 8200 AB Lelystad, The Netherlands. restricted to situations in which host
changes in density (i.e. does population *e-mail: m.c.m.deJong@id.wag-ur.nl population size changes, but density
growth come with an increase in area?). remains constant and, in that special case,
Previously [2,3], we discussed two Odo Diekmann βSI (where S is the number of susceptible
points related to modelling transmission Dept of Mathematics, Budapestlaan 6, hosts, I is the number of infected hosts and
and we take this opportunity to explain 3584 CD Utrecht, The Netherlands. β is the transmission coefficient) is indeed
them again in a different way: not a correct representation of the mass
(1) Often in transmission models, host Hans Heesterbeek action assumption. The problem has been
populations are compared that differ in Quantitative Veterinary Epidemiology Group, that this is a special case, one that we think
number, but not in spatial density of animals Faculty of Veterinary Medicine, PO Box 80151, is quite unusual in free-ranging animal
[3–5]. Moreover, variations in number but 3508 TD Utrecht, The Netherlands. populations, although it can be generated
not in density also occur for farmed animals experimentally, and can occur in farmed
References
where the same housing system in farms of animals. As we noted [1], seal colonies are
1 McCallum, H. et al. (2001) How should pathogen
different size will lead to different numbers of transmission be modelled? Trends Ecol. Evol. one of the few natural situations where it is
animals that are, however, kept at the same 16, 295–300 likely that transmission is better modelled
density. In more natural situations, density 2 De Jong, M.C.M. et al. (1995) How does with population numbers rather than
might be determined by animal behaviour transmission of infection depend on population densities per unit area, and this is the
size? In Epidemic Models: Their Structure and
that is independent of area size [6]. example [3] of a wild population referred to
Relation to Data (Mollison, D., ed.), pp. 84–94,
(2) Modelling density dependence Cambridge University Press
in De Jong et al.’s [2] response to our paper.
could be considered a logical next step in 3 Bouma, A. et al. (1995) Transmission of In almost all other natural populations, an
modelling transmission. However, density is pseudorabies virus within pig populations is increase in host population size will lead to

http://tree.trends.com 0169-5347/02/$ – see front matter © 2002 Elsevier Science Ltd. All rights reserved.