Effects of abundance on infection in natural populations: field voles and cowpox virus.

Detailed results on the dynamics of cowpox virus infection in four natural populations of the field vole, Microtus agrestis, are presented. Populations were sampled every 4 weeks (8 weeks in mid-winter) for 6 years. The purpose was to examine the relationships between overall or susceptible host abundance (N, S) and both the number of infected hosts (I) and the prevalence of infection (I/N). Overall, both I and I/N increased with N. However, evidence for a threshold abundance, below which infection was not found, was at best equivocal in spite of the wide range of abundances sampled. Cross-correlation analyses reflected annual and multi-annual cycles in N, I, S and I/N, but whereas N was most strongly correlated with contemporary values of I and I/N, in the case of S, the strongest correlations were with values 1 to 2 months preceding the values of I and I/N. There was no evidence for a 'juvenile dilution effect' (prevalence decreasing with abundance as new susceptibles flush into the population) and only weak evidence of a time-delayed effect of abundance on the number infected. We argue that these effects may occur only in systems with characteristics that are not found here. Transfer function analyses, which have been neglected in epidemiology, were applied. These models, with ln(S) as the input parameter, in spite of their simplicity, could be linked closely to conventional formulations of the transmission process and were highly effective in predicting the number infected. By contrast, transfer function models with ln(N) as the input parameter were less successful in predicting the number infected and/or were more complex and more difficult to interpret. Nonetheless, overall, we contend that while monitoring numbers susceptible has most to offer, monitoring overall abundance may provide valuable insights into the dynamics of infection.