Understanding how socioeconomic inequalities drive inequalities in COVID-19 infections

Abstract

Across the world, the COVID-19 pandemic has disproportionately affected economically disadvantaged groups. This differential impact has numerous possible explanations, each with significantly different policy implications. We examine, for the first time in a low- or middle-income country, which mechanisms best explain the disproportionate impact of the virus on the poor. Combining an epidemiological model with rich data from Bogotá, Colombia, we show that total infections and inequalities in infections are largely driven by inequalities in the ability to work remotely and in within-home secondary attack rates. Inequalities in isolation behavior are less important but non-negligible, while access to testing and contract-tracing plays practically no role because it is too slow to contain the virus. Interventions that mitigate transmission are often more effective when targeted on socioeconomically disadvantaged groups.

Figure 1

Visual representation of the theoretical model. An initial infection A potentially infects two other individuals, called B and C. (a) A successfully infects B and C. A gets tested upon experiencing symptoms, and isolates upon receiving a positive test result. This begins a process of contact tracing, through which C (but not B) is tested. B does not infect anyone else; the only person they potentially infect is immune. C infects two other people before being isolated. Once C is isolated, she does not come into contact with a potential infection outside the household, but still infects an individual in the same household. Individuals in the model may or may not be symptomatic, get tested, be contact traced, and they may isolate for a variety of reasons. (b) The infection tree summarises the “branching process” in the model, i.e. the first and second generation potential infections caused by A.

Figure 2

Estimations of incidence rate using data and baseline simulations. Panels (a), (b), and (c) show the per capita incidence over the previous 2 weeks based on confirmed cases (those who test positive) for each SES at each date. Panel (a) is based on the administrative data from the HSB on the number of confirmed cases at each date. Panel (b) is calculated using the number of infected individuals that test positive in the model simulation with no mobility change, while panel (c) uses the same calculation for the model simulation that allows for mobility to change over the course of the epidemic (in a way that best predicts total detected cases). Panels (d), (e) and (f) show the cumulative per capita incidence (including both confirmed and unconfirmed cases) by the 3rd March 2021 (the most recent date for which the CoVIDA data is available). Panel (d) uses positivity in CoVIDA data to calculate incidence, see Section S1.1. Panel (e) and (f) includes all infections in the versions of the model without and with mobility change respectively. All model results are calculated by taking the median value over 50 simulations. Model and actual dates are aligned by taking the model time period for which the model-predicted 2 week total per capita incidence is the same as the actual value on June 1st 2020, and setting this time period to be June 1st 2020.

Figure 3

Upward Adjustment Scenarios. Baseline indicates the model with the parameters of Table 1 and no adjustment. The panels in columns 2 to 6 are the results of upward adjustment scenarios. In the top row of columns 2 to 6 (100% adjustment), the set of parameters indicated in the column heading is adjusted so that all SES have the same value as SES 5&6. In the bottom row (50% adjustment), all SES other than 5&6 have their parameters adjusted to move halfway to the value of 5&6. Parameters adjusted in each set are as follows: out of home (number of contacts outside the home), within home (within-household SAR, household size), isolation behavior (probability of isolating conditional on observing symptoms, testing positive, being contact traced, and probability of quarantining as a household), testing & tracing (probability of self testing, delay in test consultation, delay in test results, and probability of being contact traced). Point estimates denote the median of 50 simulations. Error bars indicate the 0.025 and 0.975 quantiles of the 50 simulations. When the error bars are close to 0, this indicates that some simulated epidemics die out in the very early stages, leading to incidence close to 0.

Figure 4

Mean-preserving reduction in inequalities. Describes the effect of reducing inequalities in all parameters simultaneously while preserving the mean of all parameters. The value of parameter k for an SES j in the baseline simulation can be written as $v_{\mathit{jk}}={\overline{v}}_k+{\epsilon }_{\mathit{jk}}$, where ${\overline{v}}_k$ is the (population weighted) mean value for the parameter across all groups, and ${\epsilon }_{\mathit{jk}}$ is some deviation. The graph plots the results of adjusting all parameters to the value $v_{\mathit{jk}}^{\ast }(\lambda )={\overline{v}}_k+(1-\lambda ){\epsilon }_{\mathit{jk}}$. The outcome variable is the median cumulative per capita incidence across all SES over the course of the entire simulated epidemic in 50 models with no mobility change. Error bars indicate the 0.025 and 0.975 quantiles of the 50 simulations.

Figure 5

Policy-style scenarios. In “Untargeted” scenarios, policy adjustments affect all groups equally. In “Targeted” scenarios, only the parameters of SES 1&2 are adjusted, but adjustments in this group are greater, such that the mean adjustment across the whole population is the same as in the untargeted scenario. “10% initially vaccinated”: 10% of the population are immune to the virus from the start of the epidemic. “Reduce outside-home contacts by 1”: mean reduction of 1 in contacts outside the home. “Increase isolation by 20 p.p.”: mean increase of 20 percentage points in probability of isolating conditional on being symptomatic and being contact traced. “Increase self-testing by 30 p.p.”: mean increase of 30 percentage points in the probability of being tested after observing symptoms. “No testing”: probability of self-testing and of being contact traced are set to 0. “Fast testing”: all tests have a consultation delay and a results delay of 1 day. Outcome variable is the median cumulative per capita incidence across all SES for 50 simulated epidemics with no mobility change. Error bars indicate the 0.025 and 0.975 quantiles of the 50 simulations.