Using influenza surveillance networks to estimate state-specific prevalence of SARS-CoV-2 in the United States

  1. Justin D. Silverman
  2. Nathaniel Hupert
  3. Alex D. Washburne

Reviewed by

Read the full article See related articles

Listed in

Log in to save this article

Abstract

Analysis of influenza-like illness surveillance data estimates that most SARS-CoV-2 infections in the United States went undetected in March 2020.

Article activity feed

  1. NCRC

    Our take

    In this study, available as a preprint and thus not yet peer reviewed, using influenza-like (ILI) surveillance systems, authors estimated that large surges observed in ILI beginning in March 2020 were due to COVID-19, but by the end of March, the case detection rate across the United States was only 12.5%. Analyses of morbidity and mortality doubling times suggest that SARS-CoV-2 has rapidly spread throughout the country since its introduction on January 15, 2020. Although ILINet is a robust surveillance system, authors make several assumptions in their study and model that contribute to uncertainty, and these results should not be used for hospital surge projections or other healthcare demand forecasts.

    Study design

    modeling-simulation

    Study population and setting

    Leveraging existing influenza-like illness (ILI) surveillance systems in the United States (ILINet) and data from the past 10 years, this study modeled the excess number of non-influenza ILI to estimate the true prevalence of SARS-CoV-2 infection. Using excess ILI estimates as a proxy for COVID-19 burden, authors then identified surges in non-influenza ILI, estimated ILI admission rates and prevalence of SARS-CoV-2 from the surge, estimated the rate at which patients who are positive for SARS-CoV-2 and have ILI symptoms are positively identified as having COVID-19, and estimated epidemic growth rates (e.g., doubling time, or the length of time it takes for an epidemic to double in size) using an SEIR model.

    Summary of main findings

    Authors identified surges in non-influenza ILI that exceeded expected levels beginning in early March 2020. Some states reported additional non-influenza ILI in excess of up to 50% higher than any previously reported level. Authors were able to assess care-seeking for ILI in New York City, and found that both care-seeking and admission to hospital emergency departments for ILI increased during the month of March, suggesting that care-seeking for mild ILI actually decreased. The rate at which COVID-19 patients were identified using ILI surveillance varied largely by state and over time, but increased over the month of March (from approximately 1% to 12.5% across the United States). Authors estimated that COVID-19-associated deaths doubled every 3.01 days across the United States during the month of March. Assuming the ILI surge is due entirely to COVID-19, authors estimated the slowest doubling time for cases across the United States, starting January 15, is 4 days.

    Study strengths

    ILINet is a robust syndromic surveillance platform, and authors were able to estimate background non-influenza ILI levels using ten years of historic data.

    Limitations

    Authors assumed the patient population reported by providers to ILINet is representative of their entire state. However, ILINet is limited to approximately 2,600 voluntarily enrolled outpatient providers across the United States and may not be representative of statewide populations, and likely underestimates the number of influenza and/or COVID-19 cases. To account for this latter limitation, authors scaled the data in order to compare ILINet data to confirmed COVID-19 case counts per state. Authors also assumed that SARS-CoV-2 is entirely responsible for the excess non-influenza ILI (i.e., the identified surge). Although this is likely, not accounting for increases in other circulating viruses, such as other coronaviruses or RSV, contributes to uncertainty in the model results. However, this assumption does become more robust as the prevalence of SARS-CoV-2 continues to increase. Asymptomatic infections would not be captured in ILINet. The SEIR models were US-wide, and as such were unable to capture regional variations in transmission or the effects of interventions.

    Value added

    This study is the first to leverage ILINet, a pre-existing and robust influenza-like illness surveillance system, to quantify the prevalence, detection rate, epidemic growth rates, and clinical rates of SARS-CoV-2 in the United States.

  2. Version published to 10.1126/scitranslmed.abc1126
  3. ScreenIT

    SciScore for 10.1101/2020.04.01.20050542: (What is this?)

    Please note, not all rigor criteria are appropriate for all manuscripts.

    Table 1: Rigor

    NIH rigor criteria are not applicable to paper type.

    Table 2: Resources

    No key resources detected.

    Results from OddPub: Thank you for sharing your code and data.

    Results from LimitationRecognizer: We detected the following sentences addressing limitations in the study:
    Our study has several limitations. First, the observed ILI surge may represent more than just SARS-CoV-2 infected patients. A second epidemic of a non-seasonal pathogen that presents with ILI could confound our estimates of ILI due to SARS-CoV-2. Alternatively, it is also possible that our use of ILI data has underestimated the prevalence of SARS-CoV-2 within the US. While early clinical reports focused on cough and fever as the dominant features of COVID [5], other reports have documented digestive symptoms as the complaint affecting up to half of patients with laboratory-confirmed COVID [19], and alternative presentations, including asymptomatic or unnoticeable infections, could result in ILI surges underestimating SARS-CoV-2 prevalence. Additionally, our models have several limitations. First we assume that ILI prevalence within states can be scaled to case counts at the state level. This is based on the assumption that the average number of cases seen by sentinel providers in a given week is representative of the average number of patients seen by all providers within that state in a given week. Errors in this assumption would cause proportional errors in our estimated case counts and syndromic case detection rate. Second, our epidemic models are crude, US-wide SEIR models varying by growth rate alone and as such do not capture regional variation or intervention-induced changes in transmission. Our models were used to estimate growth rates from ILI for testing with COVID ...

    Results from TrialIdentifier: No clinical trial numbers were referenced.

    Results from Barzooka: We did not find any issues relating to the usage of bar graphs.

    Results from JetFighter: We did not find any issues relating to colormaps.

    Results from rtransparent:
    • Thank you for including a conflict of interest statement. Authors are encouraged to include this statement when submitting to a journal.
    • Thank you for including a funding statement. Authors are encouraged to include this statement when submitting to a journal.
    • No protocol registration statement was detected.

    About SciScore

    SciScore is an automated tool that is designed to assist expert reviewers by finding and presenting formulaic information scattered throughout a paper in a standard, easy to digest format. SciScore checks for the presence and correctness of RRIDs (research resource identifiers), and for rigor criteria such as sex and investigator blinding. For details on the theoretical underpinning of rigor criteria and the tools shown here, including references cited, please follow this link.

  4. Version published to 10.1101/2020.04.01.20050542v3 on medRxiv
  5. Version published to 10.1101/2020.04.01.20050542v2 on medRxiv
  6. Version published to 10.1101/2020.04.01.20050542v1 on medRxiv