截至2022年2月21日,在仅仅两个月时间澳洲新冠确诊累计人数从30万激增至310万(今年一月份之后的确诊人数统计口径上包括了RAT快速抗体检测的数据)。然而,因为感染新冠而去世的累计死亡人数只是从2200人增加到了4900人;自二月初以来,因为感染新冠的住院人数也一直在稳步下降。因此,澳洲联邦政府决定,在关闭了边界已超过整整两年时间后,从2月21日起澳洲重新开放国境,欢迎已完全接种疫苗的国际游客到访澳洲。
在2020年六月的时候我写了一篇用随机点过程模型模拟新冠传染机制的文章,其中对比了英国与澳洲从2020年3月1日到6月1日这两个月这两个国家新冠确诊人数的增长情况。在3月1日的时候英国与澳洲分别各有约10例新冠活跃病例,但两个月后,英国累计新冠确诊人数达到了17万,澳洲只有不到7千人-英国的确诊人数是澳洲的26倍之多。再看截至2022年2月21日的统计数字,英国累计新冠确诊人数达到了1860万,澳洲为310万 - 英国的确诊人数正好是澳洲的6倍;英国累计因感染新冠死亡人数为16万,澳洲为4900人 - 英国的累计死亡人数是澳洲的32倍还多。考虑到英国的人口是澳洲人口的三倍不到(6750万对比2500万),这真是一个挺有意思的比较结果。
2021年中之前,澳洲的新冠疫苗接种进展缓慢,在中国已全民疫苗接种了50%的时候,澳洲接种率只有20-30%。但很快,绝大多数澳洲百姓都积极响应政府的号召去打疫苗,现在16岁以上人群两剂接种率已超过95%,打了加强针的比例也达到了50%。过去两年大多数高校员工都改为在家远程办公的模式。去年我有三分之一的时间都是在家办公(我喜欢在办公室办公,在家里办公搞得常常不知道自己究竟是在上班呢还是在放假,不伦不类的。所以只要允许我都尽量回办公室上班)。我们大学新学期学生都重新恢复面对面教学模式,只是员工都要求打三针的完全疫苗接种,每个人还要在自己的电子人事档案记录上报告疫苗接种状况并由直接领导检查核实。
新冠两年对澳洲的旅游业冲击最大,没有了外国游客,旅游收入的大头就没有了。对现在联邦政府的重新开放国境的决定,澳洲旅游业当然是最为欢迎啦。新冠也重创了澳洲的高等教育产业,多年持续稳步增长的国际留学生产业突然崩塌,留学生经济效益的急剧减少引发各高校大裁员(甚至连带使悉尼、墨尔本的房地产出租业,以及与留学生生活需求相关的其它服务业遭重创)。
据澳洲高校联合会(Universities Australia代表39所澳洲高等院校)的估计,总数约30万的高校教职员工有约10%的人因为新冠丢掉了工作。我们学校从2020年7月1日开始了机构调整及裁员(restructure due to COVID-19),直到最近才完全结束。我的part-time的行政助理就因此被裁员,现在我是学术与行政事务一肩挑。新冠发生之前,每年我都要旅行(自己开车或坐飞机)到我们学校的不同校区(各自相隔100 至数百公里)做不同的统计专题培训班讲课,过去两年也全改为网上模式了,效果当然不如面对面的好。不但学校的印刷所撤销了,据说原来庞大的车队也大部分卖掉,小部分则分配到各个院系自行管理使用与负担费用。新冠使我们都不得不改变了原有的工作模式。
希望这次我们真正地能对新冠病毒说再见了。大家珍重!
以下附上一篇我2020年8月写的关于澳洲维多利亚州当时新冠确诊病例的研究的文章作为澳洲澳洲管控新冠疫情的一个案例供大家参考(此文为自娱自乐以及验证我2020年6月在nature scientific reports上发表的随机点过程病毒传染过程模型而写,写成后只是放在了我的researchgate的个人网页上:MODELLING THE SPREAD OF COVID-19 | Gang (John) Xie | Research Project (researchgate.net))
The COVID-19 second wave attack in Victoria, Australia – a story to tell
The Victoria state of Australia is experiencing the COVID-19 second wave attack since late July. In this report we have used the simulation model as proposed in “A novel Monte Carlo simulation procedure for modelling COVID‑19 spread over time” [1] to analyse the recorded confirmed COVID-19 cases for Victoria over the period 1 March to 9 August 2020 [2] so that we can have more important information that we may need for a better understanding of what happened and the reasons of the stage 4 restriction measures imposed by Victoria state government [3].
As a very brief description of the simulation model we used in this reprot, the model speficication and the meaning of the parameters are given below.
Table 1: Definitions of the argument terms for the simulation function TransSimu()
Input parameter | definition of the input parameter |
days | Observation period (in days) of a simulation study |
nd | Simulation period (in days) of a simulation study |
Rt | Average reproduction/infection rate (i.e., the expected number of secondary cases that each existing infectious case will generate); Rt is a function of time/day. |
muT | Average/expected number of days for an existing infectious person to infect a susceptible person in the population; mean parameter for the negative binomial distribution[4, 5]. |
sizeV | The dispersion parameter for the negative binomial distribution so that variance = muT + (muT)2/sizeV [4, 5]. |
limit | The study/target population size |
pp | The proportion of people with immunity in the population |
n0 | The initial number of infectious persons prior to the observation/simulation period. |
The infection rate pattern for Australia was specified/estimated as
rr = c(rep(2.4,5), rep(2.6,4), rep(2.2,4),rep(3,5), rep(2.4,2),rep(1.7,5), rep(0.6,4), rep(0.4,6),rep(0.6,20), rep(1.4,8), rep(0.7,22), rep(1.4,28), rep(1.9,6), rep(2,9), rep(1.5,8),rep(1.3,6), rep(1.2,11),rep(1.1,3), rep(0.8,3),rep(0.7,8),rep(0.6,8), rep(0.5,5))
and the simulation model was specified as TransSimu(nd=170, Rt=rr, muT=4.4, sizeV=0.8, n0=3). All analyses were performed using the free statistical software ware R [6]. The analysis outcomes are presented and discussed below.
As shown by Figures 1 and 2 below, the first wave of the COVID-19 attack happened in late March and the second wave peaked in early August. Compared with the first wave, the second wave peaked much higher (6+ times more) and still at a very high level at the time of writing. If we look at these two graphs carefully, a very small bump occurred in early May. The difference between Figure 1 and Figure 2 is the former shows the recorded daily cases (bars) and superimposed by the estimated daily numbers (the red lines), the latter shows the estimated daily infectious active cases pattern. Therefore, the latter figure has much higher number values than the former one but the patterns match with a few days lag (the peaks or trouchs in Figure 1 ahead of Figur 2 for a few days because of the cumulative effect in Figure 2). According to news media report [7], the second wave in Victoria started from the middle of May. "’It is likely that the large majority - I said in my statement approximately 90 per cent or more - of COVID-19 infections in Victoria can be traced to the Rydges Hotel.’" “Dr Alpren said the Rydges outbreak started with a family of four returning from overseas on May 9 and moved into the hotel on May 15 after displaying coronavirus symptoms.”[7]
Figure 1: The daily confirmed COVID-19 cases in Victoria state, Australia over the period of 1 March to 9 August 2020 (162 days): bars represent the observed numbers and the red lines are the model estimated numbers.
Figure 2: The pattern of the model estimated number of the infection active cases over the observation period: the bold black curve is the median level, the thin green curve is the 25th percentile level, and the thin red curve is the 75th percentile level out of 1,000 bootstrap simulation runs.
The figure 3 shows the cumulative confirmed case patterns. The near perfect fit of the model estimates (the black curve) and the observed numbers (dot points) shows the applicability of the proposed simulation model. As of 9 August, the observed total confirmed cases is 14957 and the model estimated total confirmed cases is 14754.
A more informative analysis outcome is presented in Figure 4, the estimated infection rate patterns over the observation period (162 days starting from 1 March). According to the model, the infection rates (the Rt values) that caused the first wave were higher than those caused the second wave (2 < Rt ≤ 3 for the first wave; 1 < Rt ≤ 2 for the second wave). However, the rising and dropping of the infection rates during the first wave completed in a much shorter period in comparison with the second wave which is still yet to complete its dropping trend. Another important feature worth noting is the time lag for the peaks and troughs between the Figure 1/Figure 2 and Figure 4. The formers are somehow 10 to 30 days lagged behind as expected from the virus transmission mechanism because the peak of the confirmed cases could only happen after a long enough period of high infectious rate transmission process. For example, the peak of the infection rate was about 10 to 12 days before the recorded peak of confirmed cases for the first wave; the peak of the infection rate was about four weeks before the recorded peak of confirmed cases for the second wave. Even though the highest infection rate for the second wave is lower than the first wave, the above one infection rate maintained for nearly two month. This finally results in a much higher peak of the daily confirmed cases for the second wave in Victoria as shown in Figures 1 and 2.
Figure 3: The cumulative number of the confirmed cases over the observation period: the bold black curve is the model estimated median level, the thin green curve is the 25th percentile level, and the thin red curve is the 75th percentile level out of 1,000 bootstrap simulation runs; the dot points are the actual recorded number of cases.
Fiogure 4: The model estimated infection rate patterns (Rt values) over the observation period (1 March to 9 August 2020).
Based on the resulting simulation model parameters, we are also able to estimate the duration period of a person of the confirmed cases who is infectious of COVID-19 to other people by a simple simulation study. Figure 5 is a graphic representation of the distribution of the duration period of a person of the confirmed cases who is infectious of COVID-19 to other people (i.e., the duration period in days of an infectious active person as defined in the paper). The analysis outcome shows that 75% of the confirmed cases would be infectious in the first 9 days after one catches the COVID-19 (e.g., after that one may still be COVID-19 positive but no longer infectious to others). This result agrees with literature reported conclusion (e.g., infectious period was reported as most likely lasting 7 to 14 days) [8-10]. However, the model says that there is the possibility that an infected person can still be infectious for as many as 54 days.
Final Remarks
I believe the recorded confirmed COVID-19 cases provide a true picture for the infection situation in the population of Victoria state because of 6.4 million population there are more than 1.85 million tests being completed. Therefore the rate of the confirmed cases is well below 1% in most of the days of the observed period of 162 days. The worst case is less than 5% (roughly 700 cases out of 20000 tests). The data analysis outcomes presented in this report confirm the applicability of the newly proposed simulation model that was detailed in [1]. The above analysis results are fully reproducible with the data and r code associated with this report as provided in the attached files.
Fiogure 5: a histgram of the infectious duration period (in days). The corresponding numeric summary: minimum = 0, first quartile = 2, median = 5, mean = 6.35, third quartile = 9, and maximum = 54 (days).
References:
1. Xie, G., A novel Monte Carlo simulation procedure fo rmodelling COVID-19 spread over time. Scientific Reports, 2020.
2. Health and Human Services, Cornavirus COVID-19 in Victoria. 2020, Victoria State Government: https://app.powerbi.com/view?r=eyJrIjoiODBmMmE3NWQtZWNlNC00OWRkLTk1NjYtMjM2YTY1MjI2NzdjIiwidCI6ImMwZTA2MDFmLTBmYWMtNDQ5Yy05Yzg4LWExMDRjNGViOWYyOCJ9.
3. Victoria State Government, Stage 4 restrictions. 2020.
4. Casella, G. and R.L. Berger, Statistical Inference. Second edition ed. 2002: the Wadsworth Group, Thompson Learning Inc.
5. Krishnamoorthy, K., Handbook of Statistical Distributions with Applications. 2006: Chapman & Hall/CRC Press.
6. Team, R.C., R: A Language and Environment for Statistical Computing (version 3.6.2). R foundation for statistical computing, Vienna, Austria, 2019.
7. Kolovos, B. family traced as source of Vic virus wave. 2020.
8. Lauer, S.A., et al. Estimated Incubation Period of COVID-19. 2020.
9. housen, T., A.E. Parry, and M. Sheel How long are you infectious when you have coronavirus? 2020.
10. WebMD Coronavirus Incubation Period. 2020.
转载本文请联系原作者获取授权,同时请注明本文来自谢钢科学网博客。
链接地址:https://wap.sciencenet.cn/blog-3503579-1326873.html?mobile=1
收藏