Assessing infections at multiple levels of aggregation

Abstract

The patterns of sero-prevalence of antibodies to four infectious diseases, representing a broad range of pathogens (bacteria: brucellosis; mycoplasma: contagious bovine pleuropneumonia; viruses: infectious bovine rhinotracheitis; protozoa: trypanosomosis) were investigated at three levels of organization (farm, area and district). Three contrasting districts in Kenya were compared: an arid and pastoral area (Samburu); a tropical highland area (Kiambu), and a tropical coastal area (Kilifi). Cattle in three districts were selected by two-stage cluster sampling between August 1991 and 1992. Schall's algorithm, a generalized linear mixed model suitable for multi-level analysis, was compared to ordinary logistic regression (OLR), which ignores clustering of responses; generalized estimating equations (GEE) or Jacknife, to account for clustering at the farm level; SAS VARCOMP, which provides normal-theory random-effects models. Schall's algorithm provided similar estimates to GEE (regression effects) and Jackknife (standard errors) for farm-level clustered data. Extending Schall's procedure for additional district and area-withindistrict random effects usually provided additional information. In general, models that included only a farm-level random effect consistently provided larger estimates of farms' variance components than did models with additional district and area random effects. The four type diseases exhibited various amounts of clustering. Brucellosis had moderate farm clustering plus some area and district clustering. Contagious bovine pleuropneumonia had only a small amount of clustering, mostly by area. Infectious bovine rhinotracheitis exhibited a large amount of clustering, primarily at the farm level. Trypanosomiasis antibody prevalence varied by district, area and farm. We believe that patterns of disease clustering identified by multi-level analysis can be used to better target high-risk units for disease control and guide research to understand disease transmission factors.