The effective reproductive number has taken a central role in the scientific, political and public discussion during the COVID-19 pandemic, with numerous real-time estimates of this quantity routinely published. Disagreement between these estimates can be substantial, and may lead to confusion among decision makers and the general public. In this work we compare different estimates of the effective reproductive number of COVID-19 in Germany during the time period from October 2020 through September 2021. We consider agreement between estimates from the same method but published at different time points (within-method agreement) as well as retrospective agreement across different approaches (between-method agreement). The former is based on an archive of real-time estimates compiled from public repositories of various academic groups. While for some approaches, estimates are very stable over time and hardly subject to revisions, others display considerable fluctuations. To assess between-method agreement, we reproduced the estimates generated by different groups using a variety of statistical approaches, standardizing analytical choices in order to assess how they contribute to the observed disagreement. These analytical choices include the data source, data pre-processing steps, assumed generation time distribution, statistical tuning parameters and temporal alignment of estimates. We find these user choices to be at least as important as the choice of statistical method among the growing number of available options. They should thus be communicated transparently along with published estimates.

Estimating the impact of the COVID-19 pandemic on mortality has been a topic of considerable interest to many scientists and politicians. To quantify this impact, some authors have added up the remaining life expectancies of people who have died with a diagnosis of COVID-19, reaching for example a total of 20.5 million years worldwide in 2020. Although useful for comparing the burden of different diseases, this quantity is however difficult to interpret at face value, due to the lack of a denominator and because it cannot be compared to zero and it is not obvious how to obtain a sensible reference value. In fact, a remaining life expectancy is necessarily greater than zero, even at an advanced age. Another potential issue is that it is based on diagnoses that might be unreliable. This is why many authors have finally attempted to quantify the mortality burden of COVID-19 by simply comparing the (period) life expectancy calculated during the pandemic (e.g. in 2020) with its pre-pandemic level (e.g. in 2019) based on all-cause mortality data (official statistics). This would correspond to the average numbers of years that a hypothetical cohort of people would lose if they lived their entire life under the mortality conditions of 2020 (i.e. with COVID-19) rather than 2019. Given that COVID-19 is expected to soon disappear (or at least become much less virulent), this indicator probably greatly exaggerates its real impact on mortality.

In this presentation, we propose a novel statistical indicator, called “population life loss”, which informs on the average life lost by real (not hypothetical) populations of people living in 2020. This indicator is based on all-cause mortality and demographic data, and can take on positive or negative values, so zero will be a natural reference value here. We calculated population life loss in 2020 for women and men living in 27 countries with available data from the Human Mortality Database. While we could confirm the significant impact of COVID-19 on mortality in 2020 in most countries, it turned out that the estimated population life losses could be counted in days rather than years. For example, while life expectancy loss in 2020 in the United States was of 2.1 years for men and 1.6 years for women, population life loss amounted to 10.1 and 6.7 days, respectively. This should be a useful piece of information from a public health perspective, e.g. to contribute to the delicate debate on the appropriateness of the various restrictive measures taken by governments to fight the pandemic.

Reference: Rousson V, Locatelli I (2022). On the impact of the COVID-19 pandemic on mortality: Lost years or lost days? Frontiers in Public Health 10: 1015501.

Real-time surveillance data are a crucial element in the response to infectious disease outbreaks. However, their interpretation is often hampered by delays in data collection and reporting, which bias recent values downward and can obscure current trends. Statistical nowcasting can be employed to correct these biases and enhance situational awareness. This talk summarizes a pre-registered real-time assessment of eight nowcasting approaches, applied by independent research teams to German 7-day hospitalization incidences. All methods were applied from 22 November 2021 to 29 April 2022, each day issuing probabilistic nowcasts for the current and 28 preceding days. Nowcasts were collected in a public repository and displayed in a dashboard. Moreover, mean and median ensembles were generated. All compared methods were able to remove a large part of the biases due to delays. Most teams underestimated the importance of long delays, though, resulting in nowcasts with a slight downward bias. Also, the uncertainty intervals of most methods were too narrow. Averaged across horizons, the best performance was achieved by a model using case incidences as a covariate and accounting for longer delays than the other approaches. For the most recent days, which are often considered particularly relevant, the mean ensemble performed best.

Introduction

At the time of writing, the Covid 19 pandemic, prevalent since late 2019, is still having an immense impact on healthcare systems and nations worldwide. In relation to this implementation of a wide variety of preventative measures such as social distancing requirements and mask-wearing guidelines. Due to measures and the medical effects, enormous economic consequences could be observed in many European economies, including the German one (Destatis, 2021). Therefore, health care and policy makers have been consistently confronted with the trade-off between different prevention strategies since the beginning of the pandemic.

The aim of this work was therefore to systematically identify publications that evaluate individual measures or combinations of these from both a medical and economic perspective. In a further step, the findings of the included primary studies were then summarized using defined and widely used synthesis methods to capture and present overarching effects.

Methods

A systematic database search was performed using PubMed and WebofScience. The search terms primarily targeted cost-effectiveness analyses (CEAs), net-benefit analyses and cost-per-death-averted analyses (CDA). The results of the systematic literature search were summarized within a systematic review using the "vote counting" method, using qualitative scoring based on the observed direction of effect. Here, a classification of the prevention measures into the four measure categories masks, testing, hygiene and "public health" took place.

Results

Of the total of over 4,000 screening publications, 21 studies met the established inclusion and exclusion criteria. Because some studies looked at different combinations of measures, a total of 66 scenarios resulted in the synthesis. It should be noted that many of the studies were based on mathematical modelling rather than experimentally collected data. For the net benefit studies, the decision rule was a net benefit greater than 0 as a positive effect direction. For the CEA, an empirically based effectiveness threshold of €74,159 (price level: 2010) per QALY was used. For the remaining CDAs, the assumed QALY-loss of a Covid-19-related death was multiplied by the previous effectiveness threshold.

Of the 66 scenarios, 40 (61%) showed a positive and 26 (39%) a negative direction of effect. When the scenarios were divided into two groups, sorted according to the underlying infection incidence (mostly according to the reproduction factor), it became apparent that the prevention measures were primarily cost-efficient, especially in the case of high infection incidence. Only the measure category "public health" was cost-efficient regardless of the infection incidence.

Discussion

A broad but relatively heterogeneous selection of primary literature was found, regarding the different measures considered as well as their chosen method of analysis. Nevertheless, some suitable studies could be identified, which overall, independent of further factors, primarily stated a cost-effectiveness of the prevention measures against Covid-19. Infection incidence was identified as an important parameter. Specifically, the cost-effectiveness varied primarily based on the assumed reproduction factor. However, a dependence on the economic level was also found. Macroeconomic net benefit studies were majority cost-effective, whereas CEAs presented a split picture. Interestingly, as the pandemic unfolded, assumptions about the underlying infection patterns also changed.

COVID-19 has led to excess deaths around the world, however it remains unclear how the mortality of other causes of death has changed during the pandemic. Aiming at understanding the wider impact of COVID-19 on other death causes, we use an Italian dataset that consists of monthly mortality counts of different causes of death staring from pre-COVID-19 era to June 2020. Due to the high dimensional nature of the data, we developed a model which combines the conventional Poisson regression with tensor train decomposition to explore the lower dimensional structure of the data. We take a Bayesian approach and impose priors on model parameters. The posterior inference is made using an efficient Metropolis-Hastings within Gibbs algorithm. Our method provides informative interpretations that conform to our hypothesis of the relationship between COVID-19 and other causes of death in addition to the Poisson regression.