FlyTitle: Coronavirus statistics

Tourism flows and death rates suggest under-reporting of covid-19 cases

旅客流量和病死率显示新冠肺炎病例存在漏报

经济学人双语版-追踪隐秘杀手 Tracking the stealthy killer

IN RECENT WEEKS covid-19, a deadly new disease, has slowed in China but spread widely elsewhere. China’s strict quarantine has led to a 90% decline in new infections, whereas outbreaks in Italy and Iran have grown rapidly. In the last week of February 70% of new diagnoses were outside of China. As covid-19 reaches countries unwilling or unable to monitor it, officials must use educated guesswork to track its evolution.

最近几周,新型致命疾病新冠肺炎的疫情在中国放缓,却在其他国家广泛传播。中国的严格隔离措施使得新增感染病例下降了90%,而意大利和伊朗的疫情却迅速加剧。在2月最后一周,70%的新增确诊病例都在中国境外。随着新冠肺炎扩散至那些不愿或无法监控疫情的国家,官员们必须运用有理据的猜测来追踪疫情进展。

The number of cases each country reports depends both on the number of infections and on how many people get tested. By March 1st South Korea had tested over 100,000 people; America just 472.

每个国家公布的病例数取决于感染人数及实际接受过检测的人数。截至3月1日,韩国已经检测了超过十万人,而美国仅检测了472人。

To estimate the number of undetected cases, scholars can make use of patterns in more complete data. One model, built by a team at Harvard, used the number of people flying from Hubei province in China, where the outbreak began, to various countries to predict imported cases. Such data are less relevant now, because Hubei has been locked down for a month.

学者们可以利用较完整的数据所呈现的模式来估计未检出病例的数量。哈佛大学的一个研究团队构建了一个模型,利用从疫情最初爆发地中国湖北省飞往各国的人数来预测输入病例。但鉴于湖北省已经封锁了一个月,这些数据现在已不太适用了。

To derive fresh estimates, The Economist built a similar model. We tested the link within the OECD—a club of mostly rich countries, which should have strong detection capacity—between Chinese tourism in 2019 and confirmed covid-19 cases. As expected, OECD states that swapped lots of tourists with China, such as Switzerland, tend to report higher infection rates than do ones with small flows, like Belgium.

为了得出新的估计,本刊构建了一个类似的模型。我们测试了经合组织(其成员主要为富裕国家,应该具有较强的检测能力)国家在2019年来往中国的旅客数字与新冠肺炎确诊数字之间的关联。正如所料,与中国有大量旅客往来的经合组织国家(如瑞士)报告的感染率往往高于旅客往来量小的国家(如比利时)。

Applied worldwide, our model finds big outliers. The outbreaks in Iran, Italy and South Korea, where the virus is spreading internally, are bigger than tourist flows suggest. At the other extreme, countries like Singapore may have fewer diagnoses than expected because of strong containment efforts. But the Philippines, Russia, Myanmar and Indonesia have lots of people and tourism to and from China, and just eight confirmed cases in total. Thousands more have probably gone undetected.

应用于全球范围时,我们的模型发现了较大的异常值。在病毒正在内部传播的伊朗、意大利和韩国,确诊病例超过了根据旅客流量得出的预测。而另一个极端是,像新加坡这样的国家可能因为采取了强有力的防控措施而使得确诊数字少于模型预测。但菲律宾、俄罗斯、缅甸和印尼与中国的人员和旅客往来庞大,却总共只有八宗确诊病例。实际可能还有几千个病例未被检测出来。

Another pattern bolsters this finding. South Korea and China test regularly. In both places—excluding Hubei, where the virus began claiming lives before authorities formulated a response—0.5-1% of people who have tested positive have died. In other countries with at least one death, this rate is five times higher. Deaths are easier to count than infections are. The most likely explanation for this gap is that for every person diagnosed in these countries, four more do not know they are infected. ■

另一个数据模式支持了这一结论。韩国和中国经常性地检测排查病例,两国(除去湖北省,该省在官方制定防控措施前已经发生死亡病例)确诊病例的病死率都在0.5%至1%之间。但在至少出现一起死亡病例的其他国家中,病死率却是这个数字的五倍。病死人数比感染人数更易统计。对于这种差距的最可能解释是,在这些国家,每一个确诊病例意味着还有四个未知的已感染病例。