Measuring the impact of COVID-19 in Italy and a global perspective

INTRODUCTION

The COVID-19 emergency started in China and quickly spread to Europe, with northern Italy serving as the epicentre for its global transmission [1]. The pandemic has caused substantial morbidity and mortality worldwide, with a death toll that exceeds 15 million people [2]. The severity of the pandemic in terms of hospitalizations and fatality varied greatly across countries. Italy experienced a severe first wave in March-April 2020, when the country did not have the resources and medical equipment to provide adequate care for the thousands of patients looking for medical assistance [3, 4].
The number of COVID-19 deaths is an important indicator to monitor the impact of the pandemic. How­ever, it probably underestimates (especially in the first phase and in countries with limited diagnostic capabilities) or even overestimates (in the subsequent phases when tracing and diagnostic capabilities had improved) the real impact of the pandemic. In this regard, excess total mortality, i.e. mortality from any cause, is a better indicator for assessing the overall impact of the pandemic. This metric takes into account underreporting of COVID-19 cases or overdiagnosis in cases where SARS-CoV-2 infection was not the primary cause of death, and it also covers the indirect effects of the pandemic on the management of other conditions [5].
In this study, we compared the trend in COVID-19 deaths in Italy with that reported globally and estimated the total excess deaths in Italy until April 2023.

MATERIAL AND METHODS

We obtained the number of COVID-19 deaths in Italy and globally from the John Hopkins University database [6]. This database provides a comprehensive and regularly updated record of COVID-19 deaths worldwide from 20 January 2020, to 9 March 2023. The data were presented graphically using the 7-day rolling mean of daily deaths.
To estimate the number of excess deaths, we compared the number of observed deaths during the pandemic with the number of expected deaths had the pandemic not occurred. Expected deaths were obtained from 2 over-dispersed Poisson regression models fitted separately for men and women on the historical daily number of deaths in a pre-pandemic period starting from 1 January 2011 to 31 December 2019. Daily numbers of deaths were obtained from the archives of the Italian National Institute of Statistics (ISTAT) [7]. The data span from 1 January 2011, to 30 April 2023.
The following model was used to estimate the expected number of deaths for the age group (i), year (j), and day of the week (k):
ln (E (deaths_{i, j, k})) = b₀ + b₁ (age_i) + b₂ (year_j) + f (day_k) + ln (pop_{i, j, k}),
where: β₀ is the intercept, β₁ represents the effect of age on mortality rates, β₂ captures the annual trend in mortality rates, f(day_k) is a natural spline function of the day of the year to model seasonal variations in mortality rates, ln(pop_i,j,k) serves as an offset term to account for demographic changes in population size and age structure.
The number of knots of the spline function was determined using the quasi-Akaike Information Criterion (QAIC), testing up to 10 knots with equal spacing. The model’s coefficients were estimated using daily mortality data from 1 January 2011 through 31 December 2019.
The 95% confidence intervals (CI) were derived using a Monte Carlo simulation. We sampled 10,000 model parameters values from a multivariate normal distribution using the parameter estimates and the variance-covariance matrix. The variance was then calculated by subtracting the number of observed deaths from the expected deaths obtained from each iteration. The 95% confidence interval was then computed based on the quantiles of the standard normal distribution.

RESULTS

Figure 1 illustrates the trends in the daily number of COVID 19 deaths globally (Panel A) and in Italy (Panel B). Globally, COVID-19 deaths reached their peaks at the beginning of 2021, in the spring of the same year, and in the initial months of 2022, followed by a gradual decrease. In Italy, the trend was characterized by a huge peak in the months of March 2020 and subsequent peaks in October–November 2020, in the spring of 2021, and in the initial months of 2022. Similarly to what has been observed globally, the number of COVID-19 deaths progressively decreased during the 2022 and the first 4 months of 2023. Still in April 2023 approximately 150-200 COVID-19 deaths per week were registered in Italy.
Table 1 provides our estimates of the excess deaths over the course of the pandemic in Italy. We estimated around 226,000 excess deaths from the onset of the pandemic in Italy in March 2020 to the end of 2022 and no excess deaths between January and April 2023.
Figure 2 gives the excess mortality expressed as the percentage difference compared to the expected number of deaths during the whole of 2022 and up to April 2023 in Italy. During 2022, excess mortality ranged between 4.3% in September to 26.3% in July 2022. In July 2022, part of the excess can be attributed to the extreme temperatures registered in Italy during that month.
Figure 3 shows a comparison between our estimates of excess deaths and the registered COVID-19 deaths in Italy. Between the beginning of the pandemic and the end of April 2023, Italy recorded 189,796 COVID-19 deaths, whereas our estimate of excess deaths for the same period amount to approximately 226,000 deaths. Excess deaths remarkably exceeded COVID-19 in March-April 2020, November 2020, and in July and December 2022. In contrast, while mortality was lower than expected in January–April 2023, 4878 COVID-19 deaths were registered in that period.

DISCUSSION

The study used the officially registered COVID-19 deaths and excess deaths from all causes to measure the impact of the COVID-19 pandemic in Italy and found a substantial disparity between officially registered COVID-19 deaths and excess total mortality up to April 2023. Excess deaths exceeded COVID-19 deaths by approximately 36,000. Our findings also highlight important differences in the timing and magnitude of the impact of COVID-19 in Italy as compared to that observed globally.
A different impact of the pandemic across countries became evident since the first year of the outbreak. In 2020, there was a reduction in life expectancy in several countries, which exceeded 2 years in men and women in Russia, equalled 2.3 years among men in the United States and 1.6 among women, and was generally higher in East as compared to North and South Europe [8]. In 2021, life expectancy rebounded from the 2020 losses in Belgium, Switzerland, Spain, Italy, France, England and Wales, Sweden, and Slovenia. Conversely, in the US and in several Eastern European countries, such as Bulgaria, the Czech Republic, Estonia, Greece, Croatia, Hungary, Lithuania, Poland, and Slovakia, life expectancy dropped further in 2021 [9].
These disparities probably arose from the complex interplay of various factors, such as the timing and the spread of the different SARS-CoV-2 variants during the pande­mic, the age distribution of the population, the level of preparedness of the country, the quality and accessibility of the healthcare system, the level of hospital saturation, the vaccine availability and uptake, and the government responses to contain the spread of infection.
Furthermore, there is an important issue regarding the validity and comparability of official statistics on COVID-19. The ability to identify cases varied across countries as well as the definition of COVID-19 deaths used in official statistics. Moreover, there are concerns about the validity and completeness of reporting. In fact, an important degree of underreporting is expected in many low- and middle-income countries, including China and India. In China around 120,000 COVID deaths were reported since the beginning of the pandemic, but independent estimates ranged between 1 and 1.5 million deaths [10, 11]. However, these estimates are subject to a high degree of uncertainty, depending on the prevalence of infection at different ages, the fatality rate, and mainly the extent to which the elderly have been protected after abandoning the “zero COVID” policy [12]. Similar doubts on the validity of the information on COVID-19 deaths can be raised in India, where official data indicate around 530,000 deaths, but the actual figure could be up to 10 times higher [13, 14].
These considerations highlight the limitations of relying solely on COVID-19 deaths as a measure of the pandemic impact and suggest that total excess deaths provide a better and more accurate measure to quantify the overall health impact. However, it is important to note that while civil registration coverage and completeness approach 100% in the majority of high-income countries, in some low- and middle-income countries, including China and India, civil registration covers only a small share of the population (4% in China and 8% in India) with suboptimal completeness (61.5% in China and 10% in India) [15]. As a result, there remains a substantial degree of uncertainty in estimating excess deaths in these countries.
Regardless of the measure we choose, the significant protection offered by vaccines becomes evident starting from the second half of the 2021 in Italy when most of the population was vaccinated. However, a resurgence in the number of deaths can be observed at the beginning of 2022, attributed to the rapid spread of the Omicron variant in the country [16]. The variant demonstrated higher infectiousness compared to previous variants and partially evaded natural and vaccine-induced immunity [17]. In July 2022, Italy faced an important excess mortality, which was partly attributed to the pandemic and partly a consequence of the heat waves observed in the country [18, 19]. In 2023, although over 5000 COVID-19 deaths were registered between January and April, the number of deaths from any cause was lower than expected. This may be a consequence of a compensatory effect resulting from the depletion (“harvesting effect”) of the vulnerable share of the population during the previous pandemic waves.
On a global level, the effects of the vaccines can be only observed during the second half of 2022, probably due to delay in vaccine delivery, shortages, and inefficient vaccine campaigns in certain countries [20].
The underestimation of COVID-19 deaths is not the sole explanation for the discrepancy between total excess mortality and COVID deaths. In fact, important excesses were registered also for specific causes of death, including cardiovascular diseases and diabetes in several countries [21]. This suggest that COVID-19 contributed to worsening the clinical condition of patients affected by those diseases, or that during the acute phase of the pandemic, health care systems were unable to provide adequate care to patients.
The methodology used to estimate excess deaths is the main strength of our study. By calculating the number of lives lost throughout the entire pandemic period, we were able to provide an estimate of the overall impact of the pandemic in Italy. However, this task is not easily achievable for many countries due to the absence of standardized and efficient data collection systems. This introduces a high degree of uncertainty in global estimates.
Finally, the model showed good reliability in a validation study where it was used to estimate the difference between observed and expected deaths in the years preceding the pandemic [22]. Our estimates aligned closely with those form other sources. The WHO estimated a 12% excess mortality in Italy in 2020 and 2021 combined [23], while Our World in Data reported an approximately 15% excess mortality for Italy in 2020, gradually decreasing but remaining above 10% throughout 2021 and 2022 [24]. These differences probably stem from differences in methodology, particularly in the choice of the baseline period considered to estimate expected deaths [25].

CONCLUSIONS

Excess deaths from all causes is a better metric to measure the overall impact of the pandemic as compared to officially registered COVID-19 deaths. The suboptimal coverage and completeness of civil registrations in certain countries make the global estimates provided questionable [26].

FUNDING

This research was supported by EU funding within the NextGenerationEU-MUR PNRR Extended Partnership initiative on Emerging Infectious Diseases (project no. PE00000007, INF-ACT).

CONFLICTS OF INTEREST

The authors declare no conflict of interest.

References