Cumulative COVID-19 incidence indicators should account for the population-at-risk dynamics

Risk estimators for the COVID propagation are essential for assessing the real risk to the population posed by the different COVID waves (1). These estimates usually rely on the 14-day cumulative incidence rate of new COVID-19 cases (IR₁₄ in the following), defined as the cumulative number of reported COVID-19 cases over 14 days divided by the population at risk per 100,000, following the classical epidemiological definition of incidence rate (2):

The IR₁₄ metric has been particularly useful to monitor the dynamics of the pandemic as well as to set thresholds used to trigger mitigation strategies during the pandemic (3). A common assumption within COVID-19 risk estimates has been to use the total population as the population at risk in Eq.[1]. However, the risk of being infected should take into consideration those who were not yet infected instead of the total population as the population at risk. Failing to account for the dynamics of the exposure population could, importantly, bias downwards real transmission levels. This bias may delay mitigation policies and complicate the understanding of the pandemic’s evolution, as it prevents a correct interpretation of the comparison of IR₁₄ over time (comparing levels of different waves) and across space (comparing the same waves across countries or regions regardless of the unique previous evolution of the pandemic in each area). This correction is especially important during the first waves of the COVID-19 pandemic when reinfection was nearly unattainable (4).

We propose to correct the IR₁₄ by accommodating the estimates to the situation where a non-negligible part of the population was not exposed to the risk (already immunized in the first pandemic stages). In doing so, we take into account that the exposed-to-risk population (those who can get infected) is lower than the total population when part of it has already had the infection. Holding constant the numerator, a reduction of the denominator represents an overall higher risk.

To illustrate this potential bias, we collected the daily number of COVID-19 reported cases in Spain during the pre-vaccine period, between 15/01/2020 and 11/07/2021, and calculated the IR₁₄ for each day using the total Spanish population as a denominator. Within the pre-vaccine period (between 15/01/2020 and 11/07/2021), we assume the infection provided 6, 8, or 10 months of immunity after the disease is contracted, which is a conservative assumption in the first stages of the pandemic (4). We do so by modifying the IR₁₄ as:

We observed that the population at risk as defined in our formula was decreasing over time as COVID-19 infections were increasing, and represented around 93% of the total Spanish population in Spring 2021 (Figure 1) when accounting for a 10-month immunity period.

• Download figure
• Open in new tab

Figure 1.

Percentage (%) of the population at risk of COVID-19 infection relative to the total Spanish population during the pre-vaccine period, from 15th January 2020 to 11th July 2021, assuming the immunity after the disease is contracted may last for 6 (dot-dashed), 8 (dashed), 10 (solid) months.

• Download figure
• Open in new tab

Figure 2:

14-day cumulative incidence rate of new COVID-19 cases calculated using Eq.[1], dot-dashed line, compared to its modified version calculated using Eq.[2], solid line, assuming 10 months of immunity. The threshold of 200 new cases is shown with a dashed line.

This decline in the denominator represents, as expected, a higher incidence rate of new COVID-19 cases, particularly in Spring 2021 for the Spanish case. For example, our estimates suggest that assuming 10 months of immunity, the threshold of 200 new COVID-19 cases per 100,000 inhabitants surpassed on 10/4/2021 would have been surpassed 5 days earlier if the population at risk only accounted for those individuals at risk of infection.

Our approach could be considered conservative as not all cases were reported, especially during the first wave, when there was a significant lack of testing for asymptomatic and mildly symptomatic individuals (5). We estimated that around 3-4% of the population had been infected by December 2020, which contrasts with higher estimates based on a seroprevalence study (9.9%) (6). Therefore, within the pre-vaccine period, using the total population as a denominator for the COVID-19 cumulative incidence rate represents an underestimation of the actual COVID-19 transmission risk.

We only considered the pre-vaccine period since we acknowledge vaccines could have altered significantly the population at risk, decreasing it with the massive vaccination plans and then increasing it with the subsequent loss of immunity after several months (7). Furthermore, the risk of reinfection also increased substantially over time with the appearance of new variants (8,9). These factors can potentially significantly impact both the population at risk and the rate of new COVID-19 cases and will need to be carefully modeled in further studies.

In conclusion, correcting the incidence rate of new COVID-19 cases by the fact that the population at risk is not the total population, but a changing fraction of it provides higher and more accurate estimates of COVID-19 risk. Therefore, the population at risk should be carefully considered when monitoring epidemic data.

The data underlying this article come from the Centro Nacional de Epidemiología and the Instituto de Salud Carlos III, and are available at https://rubenfcasal.github.io/COVID-19/.