English (pdf)
Article in xml format
Article references
Automatic translation
Send this article by e-mail
Cited by SciELO 
Similars in
SciELO 
Permalink
INTRODUCCIÓN
The global spread of a new SARS-CoV-2 virus has so far caused a worldwide pandemic with more than seven million deaths in around 240 countries1. The overall impact of the COVID-19 pandemic on mortality at the population level is of great concern to public health and the reporting of available evidence for public policy is a valuable tool for decision-makers and the general public2,3, especially in countries and times with limited resources. The sudden emergence of the coronavirus was a major challenge for the healthcare system in many countries, including Paraguay. Schneider, for example, reported a case fatality rate of about 2.5% for Paraguay4, however, this does not reflect the risk of dying from COVID-19. Not just direct effects, such as the measures imposed by the health authorities to combat the epidemic through vaccination, isolation and the wearing of face masks, have an influence on all-cause mortality. Other critical outcomes are access to hospitals, in particular to primary care, not related to COVID-19 infections. Individual fear of contracting the SARS-CoV-2 virus leaving stay-at-home orders to seek primary care or the fear of encountering overloaded hospital admissions may impact the decision of care seekers and lead to increased deaths from non-COVID-19 causes5.Therefore, crude deaths counts are not a good indicator of a pandemic burden6,7. Measuring the excess of all-cause mortality encompasses direct and indirect effects5,8,9 and is an important metric in tracking the impact of a pandemic, within and between countries8, however, estimating excess mortality depends on the baseline model used10. Interestingly, the most frequent methods used are regression and simple averages, such as the five-year average11; however, simple averages is the least recommended method since it ignores trend and seasonal effects, leading to underestimations of excess mortality10,12-15. This study uses advanced statistical methods to quantify the impact of the COVID-19 pandemic in Paraguay between 2020 and 2022.
METHODS
The monthly all-cause deaths count data at population-level for the period 2013 - 2022 were obtained from the Public Ministry of Health through a freedom of information request. The data contained 21 different causes of death without patient information, aggregated according to the International Statistical Classification of Diseases and Related Health Problems, 10th Revision (ICD-10). The national population figures were obtained from the national statistics office (INE) and were based on census data from 2015. As national population figures are only available annually the author first age-standardized the yearly data according to the WHO world population standard and then interpolated the result to obtain monthly population data.
The study took into account several covariates mentioned in the literature, such as cardiovascular disease death rates, diabetes death rates, cerebrovascular death rates, respiratory system death rates, and national daily average temperature16-18. The covariates were selected based on the strength of association and availability. COVID-19 data were sourced from the COVID-19 Data Hub19. As the focus of this study is on quantifying the impact of the pandemic on all-cause mortality and not on a country comparison, we used leap-year adjusted counts to model excess mortality. Using excess mortality or excess deaths allowed us to estimate additional deaths in a given time period, compared to what we would have expected, and does not depend on how COVID-19 deaths are recorded6.
In addition, this study also computed undercount factors, that is, the ratio of excess mortality to the officially reported COVID-19 death counts of the same year20, as well as P-scores(%), which indicates the relative excess mortality (i.e. observed deaths - expected deaths / expected deaths) for the pandemic years under consideration. Larger values indicate increasing levels of excess deaths3. Excess mortality estimates, P-scores, and undercount factors were computed starting from March, 2020 reflecting the official start of the pandemic in Paraguay21.
The author used counterfactual reasoning5 to estimate excess death from all-cause mortality data, that is, what would have happened if the pandemic has not occurred. A custom Bayesian Structural time series approach with 5000 Markov chain Monte Carlo (MCMC) simulations was used to train the data on pre-pandemic years until 2018, and then further validated to predict values for 2019. The final model included autoregressive (AR1) and seasonal components, to predict estimates of counterfactuals with 95% Bayesian credible intervals for the COVID-19 pandemic period from March, 2020 to December, 2022.
All analyses were performed using R statistical software with R version 4.4.122.
RESULTS
The overall death rate for Paraguay was 4.76 per 1000 population. All-cause mortality was already showing a slow upward trend up to 2019 (Figure 1). Paraguay reported 122,360 deaths from all causes between 2020 and 2022 (2020: 29,673; 2021: 53,073; 2022: 39,614), compared to an expected 88,988 (Bayes 95% CI: 84,044 to 93,627) had the pandemic not occurred. This counterfactual translates to an increase of 33,372 excess in all-cause deaths (95% CI: 28,733 to 38,316), a relative difference of 38% (95% CI: 31%; 46%). Interestingly, the P-score (%) was higher in 2022, suggesting higher COVID-19 prevalence, which can also be observed visually in Figure 1.
Figure 1. Monthly all-cause mortality predictions (dashed line) vs. fitted counts (black line). The Shadowed area indicates 95% Bayesian credible intervals. The vertical line indicates the start of the pandemic. The below plot shows the cumulative excess deaths.
Paraguay showed a relative accurate reporting of COVID-19 related death; however, it is likely that Paraguay undercounted COVID-19 deaths during the first year of the pandemic as indicated by the undercount factor of 1.2 (r = 0.7). Excess deaths have more than doubled when comparing 2021 to 2020. Nevertheless, in 2022 the undercount factor decreased to 0.7.
The following table (Table 1) gives an overview and summary statistics for excess deaths, P-score (%), and undercount factor for the time under consideration.
Table 1. Summery statistics for excess death, P-score (%), and undercount factor
| Year | Expected | Observed | Excess | P-score (%) | Cumulative Covid-death | Undercount factor |
|---|---|---|---|---|---|---|
| 2,020 | 26,868 | 29,673 | 2,805 | 10.4 | 2,292 | 1.2 |
| 2,021 | 36,360 | 53,073 | 16,713 | 46 | 16,624 | 1 |
| 2,022 | 25,760 | 39,614 | 13,854 | 53.8 | 19,688 | 0.7 |
Decomposing the time series (Figure 2), the seasonal component (middle-panel) indicated a clear seasonal pattern, while the top-panel shows the autoregressive (AR1) term used in the final model.
Figure 2. Components of the final model with Trend (top-panel), Seasonal (middle-panel) and Regression (bottom-panel) component and their respective fluctuations.
Looking further at the seasonal component, Figure 3 indicates that all-cause mortality increased with decreasing temperature during winter (April - July) with a clear peak in June.
DISCUSSION
This study uses Bayesian Structural time series analysis to quantify excess mortality in times of the COVID-19 pandemic in Paraguay. The country has excess death in all three pandemic years under consideration and likely undercounting COVID-19 related deaths in the first year, probably due to a lack of testing capacities, reporting challenges, awareness, or a general reduction or access to health care services, and, in particular primary health care3,23,24. A notable drop in the undercount factor in 2022 could be due to not only confirmed but also suspected Covid cases, as well as deaths from other causes such as respiratory or cardiovascular diseases, especially with the sometimes rapidly falling temperatures during the winter months. A slight upward trend in all-cause mortality can already be observed before the country was hit by the COVID-19 pandemic, which may indicate a slow detorization of the healthcare system in general.
A few studies cover countries on the South American continent, including Paraguay20,25,26. At a regional level, it is known from the literature that countries such as Peru, Bolivia and Brazil have high excess mortality rates20,26. Although these studies are not directly comparable, as they use different methods, e.g., linear regression or over dispersed Poisson regression models14, for estimating the expected counterfactuals and different time periods, they are generally consistent with the results of this study. Wang et al26, for example, estimate 22,500 excess deaths for the period 2020-21, using a weighted ensemble technique of six different models, while Karlinsky and Kobak20 estimate 9,600 excess deaths using a linear regression with data from March, 2020 until the end of May, 2021, and Schumacher et.al25 estimate about 1,000 excess deaths for 2020, and about 15,000 for 2021.
On a regional level undercount factors are, on average, between 1.44 (2020), 1.27 (2021), and 1.34 (2022), which, in general, indicates good quality of all-cause mortality data for Paraguay3 (see Table 1). Excess mortality, in an international context of a cross-national study of 49 Western countries, reached, on average, a P-score of 11.4% in 2020, 13.8% in 2021 and 8.8% in 202227, therefore, figures for Paraguay should be a cause for concern for the national authorities.
While the results presented in this paper extend the knowledge on the impact of COVID-19 on all-cause mortality in the context of Paraguay, the study is not without limitations. The author acknowledges that a gender and age-specific analysis would shed more light on the overall impact of the pandemic and might identify inequalities among different population groups. While variations in estimating excess mortality are small between weekly and monthly data, it might be better to use weekly data10, as a faster way to identify patterns and changes of a pandemic outbreak. Furthermore, deaths from the COVID-19 pandemic itself are not accounted for in this study. These limitations should guide further research on quantifying the impact of COVID-19 on all-cause mortality.
To the best of the author’s knowledge this is the first national study to evaluate the impact of COVID-19 on all-cause mortality covering three years of the pandemic using advanced statistical methods.
