Giardia spp

Hillman, Alison E., Ash, Amanda L., Lymbery, Alan J. & Thompson, R. C. Andrew, 2019, Anthropozoonotic significance, risk factors and spatial distribution of Giardia spp. infections in quenda (Isoodon obesulus) in the greater Perth region, Western Australia, International Journal for Parasitology: Parasites and Wildlife 9, pp. 42-48 : 43-44

2.4. Analysis of risk factors for Giardia spp . in quenda

Univariable logistic regression was undertaken using Stata 14 software (StataCorp, 2015) to assess association between Giardia spp . infection and a number of environmental and host risk factors. Environmental factors considered in the univariable risk analyses were trapping site (animal trapped in bushland compared to urbanised environment) and season of sampling. Host factors considered were maturity (adult or subadult), sex, the presence of an active pouch in adult females (active pouch = lactating, inactive pouch = not lactating), concurrent gastrointestinal parasitic infections (assessed on a presence/absence basis) and concurrent ectoparasitic infections (tick, flea and mesostigmatan mite infection intensities were assessed on an ordered categorical scale, and trombiculid mite infections were considered on a presence/absence basis) ( Table 1).

For each univariable assessment, data clustering by trap site was assessed using the likelihood ratio test. The only evidence of clustering occurred with Entamoeba spp. — this association was therefore tested using mixed effects logistic regression, including trap site as a random effect. Ordinary logistic regression was used to assess all other variables. Results were reported with Wald p-values.

An explanatory approach to multivariable analysis was undertaken (Shmueli, 2010), to test the putative causal role of various risk factors in Giardia spp . infection. A multivariable risk factor model was built for each putative risk factor with at least weak evidence (p ≤ 0.10) of an association with Giardia spp . infection. All other independent variables with at least weak evidence (p ≤ 0.10) of an association with Giardia spp . infection, and maturity, were considered as putative confounders for each multivariable risk factor model. Independent variables were sequentially added to the models — they were retained if their inclusion altered the odds ratio (OR) of the risk factor of interest by ≥ 10%. Testing for interaction between each retained confounder and the risk factor of interest was undertaken using the likelihood ratio test — the interaction term was retained in the model if there was at least weak evidence (p ≤ 0.10) of interaction. Variance inflation factors of retained independent variables were checked to ensure they were <10.

2.5. Mapping and investigation of geographical clustering of Giardia spp . infection risk in quenda

Giardia spp . infection was mapped using QGIS v2.18 (QGIS Geographic Information Systems, 2018), using the GPS point of the trap in which the quenda was caught. On mapping, GPS points were displaced by 0.01, and uninfected quenda icons were set at 50% transparency, to improve visibility of the relative distribution of infected quenda.

The risk of Giardia spp . infection was investigated spatially using Kulldorff' s spatial scan statistic (SaTScan version 9.5, www.satscan. org). Data were run as an unfocused Bernoulli model, scanning for clusters of both increased and decreased risk of Giardia spp . infection, utilising Gumbel approximation in the significance testing. Clusters with at least weak evidence (p ≤ 0.10) of geographical clustering were mapped using SaTScan shapefile output in QGIS v2.18 (QGIS Geographic Information Systems, 2018).