Contribution of age-related effects to differences in crude fatality ratios due to COVID-19 in two Argentine provinces between March and August 2020

Contribuição dos efeitos relacionados à idade para as diferenças das taxas de fatalidade bruta por COVID-19 em duas províncias argentinas entre março e agosto de 2020

Abstract

Background:  The ongoing COVID-19 pandemic has affected Argentina with uneven intensity. The force of the mortality attributed to COVID-19 may vary across populations by age structure, which has to be considered when analyzing the impact of the pandemic.

Objective:  This paper presents net estimations of case fatality ratio (CFR) attributed to COVID-19 for two provinces in Argentina (Jujuy and Buenos Aires), using public data provided by the Argentine Ministry of Health.

Method:  After some sensitivity analyses, we opted to make a comparison between two given jurisdictions, Buenos Aires and Jujuy, by applying the Kitagawa decomposition procedure, separating rate ("net" fatality) and structure components (age-attributable effects) from CFR estimations in those provinces along standardized rates.

Results:  After the decomposition we observed that, while overall CFR differences between provinces are slight on average, the magnitude of structure and rate effects tends to go into different directions across age groups, indicating more premature mortality in Jujuy due to COVID-19.

Conclusions:  While comparing the CFR across populations may be a difficult task, we found a great deal of heterogeneity that exists across age groups when comparing two given fatality ratios.

Keywords:
mortality; coronavirus infections; mathematical concepts; public health; epidemiology

Resumo

Introdução:  A pandemia COVID-19 em curso afetou a Argentina com intensidade desigual. A força da mortalidade atribuída à COVID-19 pode variar entre populações por estrutura etária, que deve ser considerada ao se analisar o impacto da pandemia.

Objetivo:  Este artigo apresenta estimativas brutas da razão de mortalidade por casos (CFR) atribuídas à COVID-19 para duas províncias da Argentina (Jujuy e Buenos Aires), utilizando dados públicos fornecidos pelo Ministério da Saúde argentino.

Método:  Após algumas análises de sensibilidade, a fim de fazer comparações justas entre jurisdições, aplicamos uma série de medidas exploratórias e, posteriormente, o procedimento de decomposição de Kitagawa, tentando separar a taxa (fatalidade "bruta" e os componentes estruturais (efeitos atribuíveis à idade) das estimativas da CFR nessas províncias, além das taxas padronizadas.

Resultados:  Após a decomposição, observamos que, enquanto as diferenças globais de CFR entre as províncias são leves em média, a magnitude da estrutura e dos efeitos da taxa tende a ir para diferentes direções entre as faixas etárias, indicando mortalidade prematura em Jujuy atribuída a COVID-19.

Conclusões:  Embora comparar a CFR entre populações seja uma tarefa difícil, encontramos grande heterogeneidade que existe entre as faixas etárias ao compararmos duas determinadas taxas de fatalidade.

Palavras-chave:
mortalidade; infecções por coronavírus; conceitos matemáticos; saúde pública; epidemiologia

INTRODUCTION

The pandemic caused by the coronavirus disease or COVID-19 has arguably been one of the most relevant events worldwide in the last hundred years. To date, different efforts have been made in various disciplines to have a better understanding of the scope of the pandemic and its effects, not only on the health of populations, but in different areas of society, such as economics, education, leisure, among many others (arguably all). Because the process is still ongoing, many of the efforts to glimpse the magnitude of the consequences of the pandemic are transient in nature and should not be thought of as complete processes. In Argentina, the first case identified as positive by COVID-19 was on March 3^rd, 2020. By the end of August of the same year, more than 400 thousand positive cases had been detected and nearly 8 thousand deaths counted, although this process has accelerated significantly in the previous months along with the population's testing capacity: as of June 1^st, nearly 16,800 cumulative positive cases and a total of 539 deaths were reported. These numbers may underestimate the burden of COVID-19 mortality given the difficulties of conducting massive tests in the countries of the region¹, but this work does not seek to question "the right number of cases" (although it is a relevant question in any estimation on the matter) in the country but to provide a descriptive and approximate picture of the CFR attributed to the pandemic and to determine the specific distorting effects of age in such estimations.

Traditionally, age has been one of the determining factors when considering the negative health effects of most diseases. The case of COVID-19 is no exception, as particular emphasis has been placed on age as the main determinant of the risk of death caused by the disease, with men being more likely to die from COVID-19 than women. This is because the observed CFR of the virus has mostly been concentrated in adult men over 65 years of age². In both Argentina and other Latin American countries, there are populations with a more rejuvenated age structure when compared to Europe, but with a higher prevalence of noncommunicable diseases caused by their own unequal epidemiological transition process³, coupled with greater social inequality⁴, more precarious housing conditions, less social protection, among other unfavorable structural situations that have rendered these countries particularly vulnerable to disease and its consequences¹,⁵. As is the case for other countries, in Argentina the demographic and epidemiological transition recorded different durations and sequences according to socio-economic sectors, in the urban and rural spheres, as well as in the geographical regions that make up the national territory⁶. The early beginning of Argentina's demographic transition was associated with the growth of the Pampas region and particularly in the vicinity of the region of what now makes up the space of Greater Buenos Aires (comprising the entire Autonomous City of Buenos Aires and a part of the territory of the Province of Buenos Aires, creating one of the largest Metropolitan Regions in Latin America, with about 20 million inhabitants). The results of this process are such that, today, the Autonomous City of Buenos Aires has the oldest population pyramid and one of the highest life expectancies in the country. The provinces of the North and Northeast of the country have the youngest population structures, having presented later social development in the aforementioned transitions⁷,⁸. It is then to be expected that these various structural effects could affect subnational estimates of the net CFR caused by COVID-19. In other words, it is possible that some or all of the fatality estimates caused by the disease in Argentina may be hiding effects related to the age structure of the affected population⁹, either by diminishing or exaggerating some differential data presented to date. In this paper, after some preliminary observations and given the complicated nature of making feasible comparisons on this topic (mainly due to lack of testing and positive cases reports), we opted to compare two very different jurisdictions: the Buenos Aires province, which has the largest population in the country and therefore provides a good approximation to the national average; and Jujuy, a province of the north that borders with Bolivia and Chile and has human development indicators below the country level. These two jurisdictions presented the most closely related proportion of positive cases among the population (Figure 1), making lack of testing potentially less problematic for an analysis of the fatality rates due to COVID-19. Therefore, comparing them could be useful to visualize an example of such age-structure differences in cases and deaths and to determine the age groups in which those effects are stronger.

Figure 1
Percentage of positive COVID-19 cases by ten age groups in given Argentine provinces, March–August 2020

This work aimed to present an estimate of the case fatality ratio caused by COVID-19 in two jurisdictions of Argentina from its arrival in the country until the end of August 2020, through the separation of effects linked to the age structure of cases recorded through mathematical decomposition techniques. These inputs sought to provide a clearer picture of the fatality ratio of the virus at the general level, but also to identify at which ages the differences are greater between provinces, detecting possible foci of premature mortality.

METHODS

An ecological-type study was conducted on the aggregated data to conduct a first exploration of differentials at the population level, performing an exploratory analysis in the first part of the study and mathematical decomposition procedures in the second. To do this, we took advantage of the daily report of data provided by the Ministry of Health of the Nation of Argentina (or MSAL, given the initials in Spanish), with information collected online through the Integrated Health Information System (SISA). In the case of anonymous, open and public data, it was not necessary to obtain an informed consent nor approval by an ethics committee for analysis. The report used was the count until August 31, 2020⁹, in the middle of the first wave of the pandemic. In these records it was possible to observe a number of basic sociodemographic characteristics of those who had been reported as suspected cases of COVID-19, as well as the date when the disease was detected for the reported positive cases. In addition, the records also informed whether that person had died, and the date of death. However, it should be noted that such information may contain errors resulting from manual loading of data, as well as ex post verifications and corrections that may have been carried out by the health authorities. That is, while these data are constantly updated, they are not those provided and harmonized by the Office of Health Statistics and Information (DEIS), a bureau that deals with Argentina's yearbooks of vital statistics, and should therefore be taken as provisional in nature. (Until the beginning of March 2022, we had not received any official vital statistics for the year 2020. However, vital statistics are not presented on a monthly basis, but rather yearly, which is counterintuitive for the purposes of this study, which focuses only on the effects of the first months of the pandemic). On the other hand, not every death that corresponds to a positive case implies that the cause of death was COVID-19. That is why emphasis is placed on the provisional nature of this study and the estimates obtained. It is not the object of this analysis to discern the "true" impact of the total CFR caused by COVID-19 on the Argentine population, but to make an approximation based on what is known to date and using the available sources. That being said, it also has to be noticed that case-detection (testing) is a very important aspect of this work: sometimes differences in CFR may result from differences in socio-sanitary conditions, but it is also probable that those differentials are due to testing capacities. While we cannot clearly establish the cause of a given CFR difference between two jurisdictions, we can assume that, in the provinces that have larger differentials for the positive cases, those differences are more likely a result of undertesting, and that similar percentages of positive cases on the overall population indicate that the provinces are suitable for comparison. Therefore, part of the exploratory analysis consisted in visualizing the proportion of positive cases detected in the overall population.

So far, the CFR has traditionally been presented in epidemiological analyses as an indicator of the strength of mortality of the pandemic, understood as the relationship between deaths (D) and positive cases (C) of the virus (Equation 1).

(1) $$CFR=\frac{D}{C}1000$$

Where information on deaths distributed by age group is available, the CFR can also be expressed as the weighted sum of the different proportions of death (P) in the different five-year age groups (e) (Equation 2):

(2) $$CFR={\displaystyle \sum P_e*C_e}$$

However, it is well known that dissimilar structures in the affected population can affect phenomena such as mortality and CFR, and can mask and confuse the magnitude of the effects on the results obtained. Typically, standardized rates are used for this type of analysis. But the technique has an important limitation: the results are expressed according to an arbitrary standard that does not allow to see the net effects of the phenomenon in question. To complement direct standardization, based on the sum of the age-specific cases by jurisdiction as the reference population⁹, we applied the technique of decomposition of effects suggested by Kitagawa¹⁰. This procedure (also known as Oaxaca-Blinder Decomposition) serves to separate, in a difference of two rates corresponding to two groups, G1 and G2, how much of the difference can be explained by the net effect corresponding to the incidence of the phenomenon in question (also known as "rate effect", or in this case "RE") and how much of that difference responds to a compositional effect (attributable to the age structure of the groups, known as "composition effect" or "structure Effect", or "SE" in this case) (Equation 3).

(3) $$\Delta \text{CF}{\text{R}}_{\text{G1,G2}}=RE+SE$$

This method is very useful for breaking down effects between groups/populations for which only one observation is taken at a given time, and above all it has already been used satisfactorily to analyze the differences in CFR caused by COVID-19 in other countries¹¹. In addition, like general decomposition methods, it allows to disaggregate the different effects in age groups (ten age groups in this case, considering the available cases and deaths), identifying those in which these differences expand and contract, allowing to identify potentially vulnerable groups where the "net" CFR is greater. After quality checks, a minimum number of cases (less than 1%) was noticed for which the province or sex was not registered. With regard to province, eight deaths were excluded from the fee analysis, while 36 deaths with missing sex information, distributed across the Buenos Aires Province, were imputed at different ages and sexes with a simple criterion of proportionality¹². To explore the positivity pattern, apart from Buenos Aires and Jujuy, we originally added some additional jurisdictions in order to visualize anomalies in the case detection patterns (Autonomous City of Buenos Aires or CABA, Chaco, La Rioja, Río Negro and Mendoza). We also considered other provinces to extend the findings of our study, but sensitivity analyses at the time made other comparisons unsuitable and potentially misleading. Similarly, only the deaths of individuals between the ages of 30 and 99 were considered to avoid an excess weight of the low CFR of the virus in children and young people and to avoid potential "noises" with centenaries. The socio-economic and health heterogeneity presented by the selected jurisdictions are worth mentioning: on the one hand, Jujuy is a jurisdiction that is in the lower half of the distribution of the geographical gross product per capita in Argentina, as opposed to Buenos Aires⁸. Likewise, Jujuy is located in the lower quartile of life expectancy at birth in Argentina, while Buenos Aires is in the third quartile⁸. It also has to be mentioned that Buenos Aires’ population is slightly older than Jujuy's, although differences in proportion are rather small for the analyzed age groups (in ^{Appendix 1} there is the proportion of population for the major age groups). Therefore, these heterogeneities and disparities at the regional level are also worth considering when analyzing the results of this work.

RESULTS

Sensitivity analysis

Table 1 presents the distribution by jurisdiction of cases and deaths among individuals between 30 and 99 years of age (and also adding the five aforementioned provinces for exploratory purposes). According to population projections by the National Institute of Statistics and Census (INDEC), the Province of Buenos Aires and the Autonomous City of Buenos Aires (CABA) concentrate 45% of the population of Argentina as of 2020¹³, so it is not surprising that these are the jurisdictions that have the greatest numbers both of cases and deaths, followed by Jujuy, Chaco, Río Negro, Mendoza and La Rioja, respectively. Of the total deaths in all seven jurisdictions, 55.7% are male, and also in all of them this indicator exceeds 50%. Only in CABA some parity in deaths could be considered given the data.

Table 1
Distribution of deaths recorded by jurisdiction as of August 31, 2020.

Province	Cases	Deaths	% Male Deaths
Buenos Aires	264,681	8,454	56,2
CABA	97,917	3,040	52,2
Jujuy	9,005	327	64,1
Chaco	5,502	217	62,7
Río Negro	5,876	182	62,5
Mendoza	7,419	163	64,4
La Rioja	2,088	80	59,9
Total	277,090	12,463	55,7

Before delving deeper into the decomposition, Figure 1 provides the proportion of positive cases by population in each province, and we can confirm that Chaco, Mendoza, Río Negro and La Rioja have the lowest proportions of positive cases detected (which could suggest an important lack of detection tests that could make any comparison unfeasible). And CABA, the capital city, has, by large, the highest proportion of positive cases, which means that the rates may also be misleading when compared with those of the other provinces (because any decomposition procedure will overestimate both rate effects and the age-structure effect). After visual inspection, we confirm that the most feasible comparison for a decomposition analysis at this time, given the data, should be between the Province of Buenos Aires and Jujuy.

General results

Figure 2 indicates the evolution of CFRs recorded by sex and age in the two selected provinces, that apparently tend to show similar patterns. It is observed, on the one hand, that CFR grows exponentially with age, slowing only at advanced ages, past 80 years. On the other hand, it is confirmed that, in the age groups analyzed, the male CFR is greater than the female CFR.

Figure 2
COVID-19 case fatality ratios (by thousand) by sex and ten age groups in Buenos Aires and Jujuy Provinces, March–August 2020.

Decomposition and standardization results

Table 2 presents the results of the standardized fatality rates (SFR) and the Kitagawa decomposition. We are using Buenos Aires as the Province that serves as a reference for comparisons with Jujuy (remember that differences in CFR rates are expressed as "ΔCFR" and "SE+RE"). The results indicate that, while the overall CFR for both provinces is practically the same, there is a slight difference by structure, both in the SFR and the decomposition: if both provinces had the same age structure between positive cases, the difference between Buenos Aires and Jujuy would be slightly larger (favoring the former), and age-structure components explain almost half of the observed CFR difference, even though they are small in magnitude.

Table 2
Case fatality ratio — CFR (by 1,000) in ages 30–99 and results of Kitagawa decomposition for COVID-19 CFR in Buenos Aires and Jujuy provinces, March–August 2020.

Province	CFR (*1,000)	SFR * 1,000	ΔCFR *1,000	RE*1,000	SE*1,000	RE (relative)	SE (relative)
Buenos Aires	45.00	44.98	-	-	-	-	-
Jujuy	44.95	45.34	0.05	-0.28	0.33	54%	46%

As mentioned above, it is also possible to visualize the different contribution of components by age groups, as presented in Figure 3, again using Buenos Aires as a reference. We can see an important difference in effect distribution, that ends up being compensated: after the age of 70 it becomes clear that there is a strong age-structure effect that diminishes the observed CFR difference, but before that point it seems the other way around. Furthermore, after the age of 80 there is an important net difference in CFR that indicates that differentials should be larger than observed if both provinces were equally exposed to the pandemic. However, the negative gradient indicates that, for the younger age groups, Jujuy presented a higher net mortality than Buenos Aires.

Figure 3
Contribution of components by age between Buenos Aires and Jujuy in COVID-19 case fatality ratio differences, March–August 2020.

DISCUSSION

This work aimed to illustrate a descriptive and comparative picture of mortality by COVID-19 in Argentina, emphasizing the jurisdictions that were most affected by the pandemic between March and August. After an exploratory analysis, regrettably the case-detection capacities in the different provinces make most of them unsuitable for analysis. However, a comparison is possible between the Buenos Aires province (the largest of the country in population size) and the northern province of Jujuy. In both provinces, CFR and testing patterns seem similar, CFR grows exponentially with age, and CFR is higher for males than females. Furthermore, the decomposition analysis showed that differences between the jurisdictions, albeit small in magnitude, would be slightly larger if both provinces had the same age-structure for the positive cases. By presenting component differences in age groups for different jurisdictions, we understand that the small magnitude differences in CFR are actually a result of the components going into opposite directions that, on average, tend to nullify themselves. This suggests a higher premature mortality in Jujuy, which is consistent with a higher penetration of infectious diseases in more disadvantaged populations, but also suggests a higher net CFR in Buenos Aires in older age groups (possibly due to selection effects, given their more advanced epidemiological transition stage), along with differences in the compositional effects. This may suggest that COVID-19 actually presented a higher excess mortality in younger and middle age groups rather than in older age groups in Jujuy, as a result of its arguably more disadvantaged socioeconomic conditions and health infrastructure.

Also worthy of note are the limitations of this work: there were reasons in the scientific and health community to think that Latin America might be one of the places most hit by the COVID-19 pandemic¹. While this is still a fledgling phenomenon, Argentina appears to be no exception: to date there are no signs indicating a slowdown in mortality due to the virus. Therefore, the estimates presented here are of a partial type, which must be taken into account when analyzing them. Second, the CFR is a simple but inaccurate indicator of attributable mortality in a population. While some of these effects can be corrected with mathematical procedures (decompositions, standardizations), there are other situations that the indicator cannot account for, such as asymptomatic infections on the population, which are not likely to be recorded as positive cases. While the CFR may account for early trends in a pandemic, perhaps in the long run other indicators will be more sophisticated¹¹,¹⁴,¹⁵. There may also be deaths attributed to COVID-19 that actually correspond to other causes and vice versa, as well as delays in reporting cases and deaths. It is also worth remembering that we are working here with registration instruments that are not intended for demographic analysis. Therefore, once vital statistics data become available with the disclosure of cause and time of death, parsing the information presented would be desirable. In addition, the distribution of positive cases by age itself may be affected by the different age structures of the population (which was not controlled for in this work). On the other hand, the small number of cases, while allowing an overall analysis of the components of CFR, does not allow detailed distinctions to be made by sex when breaking down the age and rate effects caused by differences in CFRs.

CONCLUSIONS

Despite these limitations, this work succeeded in establishing that the effects attributable to the age structure only explain a portion of the COVID-19 CFR differences established between two jurisdictions in Argentina, identifying that in middle age groups these differences are greater (indicating a greater excess mortality), and illustrating a simple mathematical decomposition procedure to obtain the different CFR rates in two provinces of Argentina.

• Funding:
  none.

REFERENCES

• ¹ Nepomuceno MR, Acosta E, Alburez-Gutierrez D, Aburto JM, Gagnon A, Turra CM. Besides population age structure, health and other demographic factors can contribute to understanding the COVID-19 burden. Proc NaCFR Acad Sci. 2020;117(25):13881-3. https://doi.org/10.1073/pnas.2008760117
  » https://doi.org/10.1073/pnas.2008760117
• ² Davies NG, Klepac P, Liu Y, Prem K, Jit M, Pearson CAB, et al. Age-dependent effects in the transmission and control of COVID-19 epidemics. Nat Med. 2020;26:1205-11. https://doi.org/10.1038/s41591-020-0962-9
  » https://doi.org/10.1038/s41591-020-0962-9
• ³ Frenk J, Frejka T, Bobadilla JL, Stern C, Lozano R, Sepúlveda J, et al. La transición epidemiológica en América Latina. Bol Sanit Panam. 1991;111(6):485–96.
• ⁴ Ravallion M. Income inequality in the developing world. Science. 2014;344(6186):851-5. https://doi.org/10.1126/science.1251875
  » https://doi.org/10.1126/science.1251875
• ⁵ Marmot M. Social determinants of health inequalities. Lancet. 2005;365(9464):1099-104. https://doi.org/10.1016/S0140-6736(05)71146-6
  » https://doi.org/10.1016/S0140-6736(05)71146-6
• ⁶ Redondo N. Población y bienestar en la argentina. Del primero al segundo centenario. In: Torrado S (Coord.). Población y bienestar en la argentina. Del primero al segundo centenario. Buenos Aires: EDHASA; 2007.
• ⁷ Belliard M, Massa C, Redondo N. Análisis comparado de la esperanza de vida con salud en la Ciudad Autónoma de Buenos Aires. Población de Buenos Aires. 2013;10(18):7-29.
• ⁸ Grushka C. Casi un siglo y medio de mortalidad en la Argentina. Rev Latinoam Población. 2014;8(15):93-118. https://doi.org/10.31406/relap2014.v8.i2.n15.4
  » https://doi.org/10.31406/relap2014.v8.i2.n15.4
• ⁹ Naing NN. Easy way to learn standardization: direct and indirect methods. Malays J Med Sci [Internet]. 2000 [cited Apr 22, 2021];7(1):10-15. Available at: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3406211
  » https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3406211
• ¹⁰ Kitagawa EM. Components of a difference between two rates. J Am Stat Assoc. 1955;50(272):1168-94. https://doi.org/10.2307/2281213
  » https://doi.org/10.2307/2281213
• ¹¹ Argentina. Ministerio de Salud. COVID-19. Casos registrados en la República Argentina [Internet]. Ministerio de Salud, Dirección Nacional de Epidemiología y Análisis de Situación de Salud (AR); 2020 [cited Aug 27, 2020]. Available at: http://datos.salud.gob.ar/dataset/covid-19-casos-registrados-en-la-republica-argentina
  » http://datos.salud.gob.ar/dataset/covid-19-casos-registrados-en-la-republica-argentina
• ¹² Dudel C, Riffe T, Acosta E, van Raalte A, Strozza C, Myrskylä M. Monitoring trends and differences in COVID-19 case-fatality rates using decomposition methods: Contributions of age structure and age-specific fatality. PLoS One. 2020;15(9):e0238904. https://doi.org/10.1371/journal.pone.0238904
  » https://doi.org/10.1371/journal.pone.0238904
• ¹³ Preston SH, Heuveline P, Guillot M. Demography: Measuring and modeling population processes. Blackwell; 2000.
• ¹⁴ Instituto Nacional de Estadísticas y Censos. Estimaciones y proyecciones de población 2010-2040. Serie Análisis Demográfico Ciudad Autónoma de Buenos Aires Nº 35. Buenos Aires: Instituto Nacional de Estadística y Censos; 2013.
• ¹⁵ Trias-Llimós S, Bilal U. Impact of the COVID-19 pandemic on life expectancy in Madrid (Spain). J Public Health. 2020;42(3):635-6. https://doi.org/10.1093/pubmed/fdaa087
  » https://doi.org/10.1093/pubmed/fdaa087

Appendix 1 Proportion of population structure for individuals aged 30 and above in Buenos Aires and Jujuy for July 2020.

Province Buenos Aires Jujuy Age Group Males Females Males Females 30–59 0,73 0,67 0,75 0,72 60–85 0,26 0,30 0,23 0,26 85 and above 0,02 0,03 0,01 0,02

Province	Buenos Aires		Jujuy
Age Group	Males	Females	Males	Females
30–59	0,73	0,67	0,75	0,72
60–85	0,26	0,30	0,23	0,26
85 and above	0,02	0,03	0,01	0,02

Publication Dates

• Publication in this collection
  07 Apr 2025
• Date of issue
  2025

History

• Received
  28 Nov 2020
• Accepted
  03 July 2022