WHO Collaborating Center for Infectious Disease Epidemiology and Control, School of Public Health, Li Ka Shing Faculty of Medicine, University of Hong Kong, Hong Kong Special Administrative Region, China.
WHO Collaborating Center for Reference and Research on Influenza, Peter Doherty Institute for Infection and Immunity, Melbourne, Victoria, Australia.
Am J Epidemiol. 2018 Feb 1;187(2):378-388. doi: 10.1093/aje/kwx270.
Influenza viruses are associated with a substantial global burden of morbidity and mortality every year. Estimates of influenza-associated mortality often vary between studies due to differences in study settings, methods, and measurement of outcomes. We reviewed 103 published articles assessing population-based influenza-associated mortality through searches of PubMed and Embase, and we identified considerable variation in the statistical methods used across studies. Studies using regression models with an influenza activity proxy applied 4 approaches to estimate influenza-associated mortality. The estimates increased with age and ranged widely, from -0.3-1.3 and 0.6-8.3 respiratory deaths per 100,000 population for children and adults, respectively, to 4-119 respiratory deaths per 100,000 population for older adults. Meta-regression analysis identified that study design features were associated with the observed variation in estimates. The estimates increased with broader cause-of-death classification and were higher for older adults than for children. The multiplier methods tended to produce lower estimates, while Serfling-type models were associated with higher estimates than other methods. No "average" estimate of excess mortality could reliably be made due to the substantial variability of the estimates, partially attributable to methodological differences in the studies. Standardization of methodology in estimation of influenza-associated mortality would permit improved comparisons in the future.
每年，流感病毒都会给全球带来大量的发病率和死亡率。由于研究环境、方法和结果衡量标准的不同，流感相关死亡率的估计值在不同的研究中往往存在差异。我们通过对 PubMed 和 Embase 的检索，对评估人群中流感相关死亡率的 103 篇已发表文章进行了综述，发现研究中使用的统计方法存在很大差异。使用流感活动指标的回归模型研究采用了 4 种方法来估计流感相关死亡率。这些估计值随着年龄的增长而增加，范围很广，从儿童每 10 万人中 0.3-1.3 例和成人每 10 万人中 0.6-8.3 例呼吸道死亡，到老年人每 10 万人中 4-119 例呼吸道死亡。荟萃回归分析确定，研究设计特征与观察到的估计值变化有关。估计值随着死因分类的扩大而增加，老年人的估计值高于儿童。乘数方法往往产生较低的估计值，而 Serfling 型模型与其他方法相比，其估计值较高。由于估计值的差异很大，无法可靠地得出超额死亡率的“平均”估计值，部分原因是研究方法存在差异。流感相关死亡率估计方法的标准化将有助于未来进行更好的比较。
