Monitoring the proportion of the population infected by SARS-CoV-2 using age-stratified hospitalisation and serological data: a modelling study

Abstract

Background: Regional monitoring of the proportion of the population who have been infected by SARS-CoV-2 is important to guide local management of the epidemic, but is difficult in the absence of regular nationwide serosurveys. We aimed to estimate in near real time the proportion of adults who have been infected by SARS-CoV-2.

Methods: In this modelling study, we developed a method to reconstruct the proportion of adults who have been infected by SARS-CoV-2 and the proportion of infections being detected, using the joint analysis of age-stratified seroprevalence, hospitalisation, and case data, with deconvolution methods. We developed our method on a dataset consisting of seroprevalence estimates from 9782 participants (aged ≥20 years) in the two worst affected regions of France in May, 2020, and applied our approach to the 13 French metropolitan regions over the period March, 2020, to January, 2021. We validated our method externally using data from a national seroprevalence study done between May and June, 2020.

Findings: We estimate that 5·7% (95% CI 5·1-6·4) of adults in metropolitan France had been infected with SARS-CoV-2 by May 11, 2020. This proportion remained stable until August, 2020, and increased to 14·9% (13·2-16·9) by Jan 15, 2021. With 26·5% (23·4-29·8) of adult residents having been infected in Île-de-France (Paris region) compared with 5·1% (4·5-5·8) in Brittany by January, 2021, regional variations remained large (coefficient of variation [CV] 0·50) although less so than in May, 2020 (CV 0·74). The proportion infected was twice as high (20·4%, 15·6-26·3) in 20-49-year-olds than in individuals aged 50 years or older (9·7%, 6·9-14·1). 40·2% (34·3-46·3) of infections in adults were detected in June to August, 2020, compared with 49·3% (42·9-55·9) in November, 2020, to January, 2021. Our regional estimates of seroprevalence were strongly correlated with the external validation dataset (coefficient of correlation 0·89).

Interpretation: Our simple approach to estimate the proportion of adults that have been infected with SARS-CoV-2 can help to characterise the burden of SARS-CoV-2 infection, epidemic dynamics, and the performance of surveillance in different regions.

Funding: EU RECOVER, Agence Nationale de la Recherche, Fondation pour la Recherche Médicale, Institut National de la Santé et de la Recherche Médicale (Inserm).

Figure 1

Description of seroprevalence and hospitalisation data (A) Estimates of seroprevalence by age group in the Île-de-France and Grand Est regions, in May to June, 2020 (median date May 14). (B) Cumulative number of hospitalisations per 100 000 population, in Île-de-France and Grand Est, from March 1 to May 6, 2020. (C) Estimates of infection–hospitalisation ratio by age group in Île-de-France and Grand Est. The y-axis is displayed in logarithmic scale. (D) Daily number of hospitalisations by age group in metropolitan France, from March 1, 2020, to Jan 30, 2021.

Figure 2

Reconstruction of the proportion infected in metropolitan France (A) Scatter plot of the seroprevalence in regions estimated with our model on May 11, 2020 (x-axis) and in seroprevalence studies in May, 2020 (y-axis), obtained from the SAPRIS serosurvey and EpiCov database. Data from the SAPRIS serosurvey in Île-de-France and Grand Est (triangles contoured in red) were used to calibrate the model. Bars represent the 95% CIs of the seroprevalence estimated by the model. (B) Proportion infected among adults in metropolitan France between March 1, 2020, and Jan 24, 2021. Timing of infection was reconstructed from the daily number of hospitalisations for COVID-19 and the delay from infection to hospital admission. The grey area represents the 95% CI. (C) Proportion infected in metropolitan France and in the 13 regions of metropolitan France, by date. (D) Geographical distribution of the proportion infected on Jan 15, 2021. (E) Proportion infected by age group and date. ARA=Auvergne-Rhône-Alpes. BFC=Bourgogne-Franche-Comté. BRE=Bretagne. COR=Corsica. CVL=Centre-Val de Loire. GES=Grand Est. HDF=Hauts-de-France. IDF=Île-de-France. NAQ=Nouvelle-Aquitaine. NOR=Normandie. OCC=Occitanie. PAC=Provence-Alpes-Côte d'Azur. PDL=Pays de la Loire.

Figure 3

Proportion infected in the regions by age group and over time (A) Estimates for the 13 regions of metropolitan France are shown on three dates. (B) Relative risk of infection of younger (<50 years) versus older (≥50 years) individuals. ARA=Auvergne-Rhône-Alpes. BFC=Bourgogne-Franche-Comté. BRE=Bretagne. COR=Corsica. CVL=Centre-Val de Loire. GES=Grand Est. HDF=Hauts-de-France. IDF=Île-de-France. NAQ=Nouvelle-Aquitaine. NOR=Normandie. OCC=Occitanie. PAC=Provence-Alpes-Côte d'Azur. PDL=Pays de la Loire.

Figure 4

Proportion of infections detected by surveillance over different periods between June, 2020, and January, 2021 Bars represent 95% CIs.

Figure 5

Sensitivity analysis (A) Proportion infected on Jan 15, 2021, assuming different sensitivities of the serological tests. (B) Proportion of infections detected by surveillance between June, 2020, and January, 2021, assuming different sensitivities of the serological tests. In our baseline analysis, we consider a sensitivity of the test of 85%.