Evaluation of the risk of neighbourhood infection of H7N1 Highly Pathogenic Avian Influenza in Italy using Q statistic.

Exposure to the risk of neighbourhood infection was estimated for the H7N1 Highly Pathogenic Avian Influenza (HPAI) epidemic that affected Northern Italy between 1999 and 2000. The two most affected regions (Lombardy and Veneto) were analyzed and the epidemic was divided into three phases. Q statistics were used to evaluate exposure to the risk of neighbourhood infection using two measures. First, a local Q statistic (Qikt) assessed daily exposure for each farm as a function of the number of neighbouring infected farms that were in their infectious period, weighted by the distance between farms. This allowed us to identify the daily time course of risk for each farm and, at any given time, local groups of farms defined by high risk. Second, for each farm a summary statistic of exposure risk within each phase (Qiph) was obtained by summing Qikt over the duration of each phase. This allowed identification of farms defined by persistent, high exposure risk within each phase of the epidemic. Statistical significance was evaluated using conditional Monte Carlo simulation, and significant values of Qiph were mapped to assess the variation of the risk of neighbourhood infection through the phases. Qikt was larger for farms in Lombardy and the reduction of exposed farms was more marked for Veneto. Although the highest value of Qiph was observed in Veneto, in each phase most of the significant values were in Lombardy. In the last phase of the epidemic, a large reduction in the number of farms significantly exposed to the risk of neighbourhood infection was observed in the Veneto region, along with generally low values of Qiph. This may be explained by differences in control measures in the two regions, including pre-emptive slaughtering of farms considered at high risk of infection. The Q statistic allowed us to quantify geographic, time-dynamic variations in exposure to neighbourhood infection, and to generate hypotheses on the efficacy of control measures.