Institute for Applied Ecology, University of Canberra, Canberra, ACT, Australia.
Epidemiol Infect. 2023 Sep 20;151:e165. doi: 10.1017/S0950268823001553.
Bovine tuberculosis (bTB) is prevalent among livestock and wildlife in many countries including New Zealand (NZ), a country which aims to eradicate bTB by 2055. This study evaluates predictions related to the numbers of livestock herds with bTB in NZ from 2012 to 2021 inclusive using both statistical and mechanistic (causal) modelling. Additionally, this study made predictions for the numbers of infected herds between 2022 and 2059. This study introduces a new graphical method representing the causal criteria of strength of association, such as R, and the consistency of predictions, such as mean squared error. Mechanistic modelling predictions were, on average, more frequently (3 of 4) unbiased than statistical modelling predictions (1 of 4). Additionally, power model predictions were, on average, more frequently (3 of 4) unbiased than exponential model predictions (1 of 4). The mechanistic power model, along with annual updating, had the highest R and the lowest mean squared error of predictions. It also exhibited the closest approximation to unbiased predictions. Notably, significantly biased predictions were all underestimates. Based on the mechanistic power model, the biological eradication of bTB from New Zealand is predicted to occur after 2055. Disease eradication planning will benefit from annual updating of future predictions.
牛结核病（bTB）在包括新西兰（NZ）在内的许多国家的牲畜和野生动物中普遍存在，该国的目标是在 2055 年之前根除 bTB。本研究使用统计和机械（因果）模型评估了与 2012 年至 2021 年期间新西兰 bTB 牲畜数量相关的预测。此外，本研究还对 2022 年至 2059 年期间感染牲畜的数量进行了预测。本研究引入了一种新的图形方法来表示关联强度的因果标准，例如 R 和预测的一致性，例如均方误差。机械模型预测的无偏性（4 次中有 3 次）比统计模型预测的无偏性（4 次中有 1 次）更为频繁。此外，动力模型预测的无偏性（4 次中有 3 次）比指数模型预测的无偏性（4 次中有 1 次）更为频繁。机械动力模型加上年度更新，具有最高的 R 和最低的预测均方误差。它还表现出最接近无偏预测的近似值。值得注意的是，所有显著有偏的预测都是低估的。基于机械动力模型，预计新西兰的 bTB 将在 2055 年后从生物学上被根除。疾病根除规划将受益于对未来预测的年度更新。
