Policy Responses Against the COVID-19 Pandemic in Latin America

The COVID-19 pandemic is spreading rapidly worldwide. Latin America, the region with the highest income inequality, remains as one of the worst-hit areas worldwide. Despite accounting for 8.4% of the global population, Latin America has witnessed 20.3% of the total SARS-COV-2 cases and 30.2% of the COVID-19 deaths to date. Several countries in the region are among the worst-hit worldwide. Brazil has had more than 11 million SARS-COV-2 cases and Mexico, Argentina and Colombia have exceeded the 2 million cases each. Similarly, the five most populated countries in the region (Brazil, Argentina, Mexico, Colombia and Peru) exceed 600,000 SARS-COV-2-related deaths. The pandemic reached Latin America later than other continents, and the first case of COVID-19 in the region was reported in Brazil on February 26, followed by a case in Mexico on February 28, 2020 and subsequently spreading throughout the region during March 2020.

Policy responses to COVID-19 in Latin America have sought to reduce viral spread, increase the capacity of the health system response, mitigate negative consequences, and strengthen governance. Effectiveness studies of social distancing policies in China, India, European countries, the United States and worldwide have shown that these appear to be effective to reduce viral transmission.

Despite the heavy burden of the COVID-19 in Latin American countries, there have been few studies examining the effectiveness of COVID-19 policies. Likewise, few studies have explored variation at the local level in the effectiveness of COVID-19 policies. Inequalities in policy effectiveness can arise due to within-country differences at the local level due to their geographical, sociodemographic, mobility patterns, and governance differences. In Latin America, high levels of poverty, urban density, household crowding, lack of safety nets, unemployment and precarious work cluster geographically and coexist with structural inequities in governance and built environments, thus creating barriers for effective compliance with preventive recommendations and for the implementation of well-functioning contact tracing and isolation mechanisms. Understanding the effectiveness of policies at the local level and exploring potential explanations for effect heterogeneity is essential to reduce the burden of the ongoing COVID-19 pandemic and inform the preparedness for future pandemics.

In this study, the investigators aim, first, to estimate the effectiveness of nonpharmaceutical interventions on SARS-COV2 transmission and COVID-19 mortality in Latin America; second, to examine the effect heterogeneity of transmission and mortality at the local level. Third, assuming there is evidence of moderate to substantial heterogeneity at the local level, the investigators aim to explore potential explanations for this heterogeneity. The study will use an interrupted time series method to estimate their effects in each local government, and random effects meta-analysis and meta-regression to obtain pooled effects, heterogeneity estimates and potential explanations.

Methods Design and setting: Natural experiment exploiting the variation in the temporal and spatial implementation of policy interventions, aimed to reduce the spread and mortality of COVID-19 in Latin America. The unit of analysis are local governments, i.e. third-tier administrative levels such as municipalities, districts or cantons.

Eligibility criteria: See below. To date, eligible countries are Argentina, Brazil, Chile, Colombia, Costa Rica, Guatemala, Mexico, Paraguay, and Peru. These countries represent 80.9% of the population in Latin America, and the vast majority of SARS-CoV-2 cases and COVID-19 deaths.

Interventions: Interventions include (i) policies aimed at reducing viral transmission, (ii) policies aimed at increasing the capacity of the health system's response, and (iii) policies aimed at mitigating the negative consequences of the epidemic and potential adverse effects of interventions. We will use the PoliMap taxonomy to categorise the examined policies.

Comparator: Counterfactual outcome defined as the projection of the pre-intervention trend to simulate what would have happened if the policy had not occurred.

Data sources: COVID-19 cases and deaths data, as well as the covariates, from official government sources, such as the Ministry of Health and Ministry of Science and Technology. The intervention information will come from legal documents, official statements, and quantitative accounts from trustable sources.

Covariates: First model at the local level does not include covariates (see below). Second model (i.e. the meta-analysis), we will examine the change in heterogeneity after adjusting for several covariates at the local level. Local level covariates include projected population size in 2020, demographic density, age-structure of the population, household density and socioeconomic status. We will use data from official sources of information, primarily the latest national population census in each included country.

Statistical analysis: See the Statistical Analysis Plan for details on the modelling assumptions. The study will use an interrupted time series design, where each local government acts as its own control. The main strength of this design is its capacity to distinguish the effect of the intervention from secular change. The study will use a Poisson regression to model the count data (for both outcomes) and accounting for overdispersion and secular trends. A full discussion on potential biases and violations of assumptions can be found in the Statistical Analysis Plan.

In a second stage, the investigators will use random effects meta analysis to pool the effect estimates for each intervention or combination of interventions. This analysis informs whether any implemented intervention was effective to reduce COVID-19 cases and deaths and the degree of heterogeneity between the effects at the local level. If there is evidence of moderate to high levels of heterogeneity (defined as higher than 50%), the investigators will also use standard meta-regression techniques to assess whether local level determinants (see Covariates) can explain the observed heterogeneity. The investigators will build the models and test the analytical strategy using publicly available data on COVID-19 cases and deaths from Finland and Sweden from January 1 to March 31.