Active surveillance documents rates of clinical care seeking due to respiratory illness

Marta Galanti, Devon Comito, Chanel Ligon, Benjamin Lane, Nelsa Matienzo, Sadiat Ibrahim, Atinuke Shittu, Eudosie Tagne, Ruthie Birger, Minhaz Ud‐Dean, Ioan Filip, Haruka Morita, Raúl Rabadán, Simon J. Anthony, Greg A. Freyer, Peter Dayan, Bo Shopsin, Jeffrey ShamanInfluenza 2020 · DOI

TL;DR

Traditional respiratory illness surveillance, relying on patients seeking medical care (often defined as ILI), significantly underestimates the true burden of viral infections. This study reveals that a large proportion of infections are asymptomatic or mild, and even when people seek care, their symptoms don't always meet ILI criteria. Active, longitudinal surveillance is crucial for a more accurate picture of respiratory virus circulation and for developing effective public health interventions.

What Problem Does This Paper Address?

Globally, respiratory viral infections impose a substantial burden on public health and economies. However, current surveillance methods, primarily healthcare-based and often relying on the "influenza-like illness" (ILI) definition, have significant limitations. These methods only capture individuals who seek medical attention, missing a large segment of the population with asymptomatic or mild infections. Furthermore, the ILI definition itself is not a perfect proxy for actual viral infections, as many medically attended respiratory illnesses don't meet ILI criteria, and conversely, many ILI cases may not be caused by specific viruses. This leads to a skewed understanding of which viruses are circulating and their true impact, potentially biasing the cost-effectiveness assessments of interventions and control strategies. The paper aims to address this gap by using active, longitudinal surveillance to provide a more comprehensive view of respiratory virus infections and associated healthcare-seeking behaviors.

Methods

The study employed a multi-pronged approach combining a prospective longitudinal cohort study with existing surveillance data.

  • Longitudinal Cohort: A cohort of 214 healthy individuals in New York City participated in daily symptom diaries and weekly nasopharyngeal specimen collection for two years. This allowed for detailed tracking of symptoms, healthcare-seeking behavior, and laboratory-confirmed viral infections.
  • Pediatric Emergency Department (ED) Data: Data from 761 children and teenagers visiting three NYC pediatric EDs were used for comparison, specifically focusing on virus prevalence in a medically attended population.
  • NYC Department of Health and Mental Hygiene (DOHMH) Surveillance Data: This included syndromic (ILI) and viral surveillance data from NYC laboratories and EDs, used to compare with the cohort's findings.

The core innovation lies in the active, longitudinal sampling combined with detailed daily self-reporting. This contrasts with passive, healthcare-based surveillance. Statistical analysis involved defining infection events, analyzing medically attended illness (MA), sick days (HOME), and medicine uptake (MEDS) in relation to viral events. A key analytical step was using Bayes' theorem to "rescale" the virus distribution observed in the pediatric ED to better reflect the general population, based on the probability of seeking care for each virus type.

Datasets

  • Longitudinal Cohort Data: Collected from 214 healthy individuals in NYC from October 2016 to April 2018. This dataset includes daily symptom reports, healthcare-seeking behavior, and weekly nasopharyngeal swab results tested for a range of respiratory viruses.
  • Pediatric Emergency Department (ED) Data: Collected from 761 children and teenagers visiting three NYC pediatric EDs from August 2016 to June 2018. This dataset provided viral testing results for patients presenting with respiratory complaints.
  • NYC DOHMH Surveillance Data: This included weekly ILI data from NYC EDs and viral surveillance data from NYC laboratories for the 2016/2017 and 2017/2018 cold/flu seasons.

This study did not introduce a new benchmark dataset but rather used existing surveillance data for comparison with its novel, actively collected cohort data.

Key Insights and Ideas

  • ILI is a Poor Proxy: The study strongly demonstrates that ILI is an unreliable indicator for the true burden of respiratory viral infections and influenza. A significant portion of medically attended ILI cases are not associated with a positive viral test, and conversely, a vast majority of actual viral infections go undetected by ILI surveillance.
  • Healthcare-Seeking is Virus-Dependent: The likelihood of seeking medical attention varies significantly by virus type. Viruses like influenza and metapneumovirus are more likely to prompt a doctor's visit compared to RSV or coronaviruses.
  • Asymptomatic Infections are the Norm: The majority of respiratory infections are asymptomatic or subclinical, meaning standard surveillance methods that rely on symptomatic individuals will inherently underestimate the total infection burden.
  • Surveillance Skews Towards Severe Pathogens: Healthcare-based surveillance data tends to overrepresent viruses that cause more severe symptoms (like influenza and HMPV), leading to a distorted view of the overall circulation of less severe but more common viruses (like HRV and coronaviruses).
  • Active Surveillance is Essential: Longitudinal, active surveillance, which collects data directly from individuals regardless of whether they seek care, is vital for a comprehensive understanding of respiratory virus dynamics, including asymptomatic infections and varied healthcare-seeking behaviors.

Key Results

  • Under-detection of ILI: 71% of self-reported ILI episodes in the cohort went unattended by healthcare providers.
  • ILI Mismatch: Only 2/3 of medically attended ILI cases were associated with a positive respiratory viral test in the same week.
  • Influenza Underestimation: Influenza accounted for only 18% of medically attended ILI cases.
  • Low Detection Rate: Among cohort participants with a positive viral test and respiratory symptoms, only 5% were medically attended and classifiable as ILI.
  • Infection Rates: The cohort experienced an estimated 1 ILI event per person per year, but multiple respiratory infections (3-4 per person per cold/flu season) were common.
  • Virus-Specific Care Seeking: The probability of seeking medical attention varied from 4% (coronavirus, RSV) to 20% (influenza, HMPV).
  • Skewed Hospital Data: In pediatric EDs, HRV and coronaviruses represented 76% of positives, while influenza and HMPV were only 6% and 4% respectively. After rescaling, influenza and HMPV were found to be overrepresented in hospital data compared to the general population.

Limitations and Future Work

  • Sample Size: The authors acknowledge that the study's sample size is relatively small, limiting the confidence in precise estimates for individual viruses and the ability to distinguish healthcare-seeking behaviors across different age groups. Larger longitudinal studies are recommended.
  • Convenience Sampling: The cohort was recruited via convenience sampling, meaning certain high-risk groups (infants, elderly, immunocompromised) were not represented. This could bias the findings.
  • Interconnectedness: Some participants were connected (e.g., family members, colleagues), which could lead to infections being transmitted within the study group, potentially biasing transmission dynamics.
  • Detection Window: The weekly sampling might miss infections with very short durations or those where viral shedding is intermittent.
  • Limited Viral Panel: The study could not account for viruses not included in the tested viral panel.

Future work could involve larger, more diverse cohorts to better characterize age-specific behaviors and risks. Distinguishing healthcare-seeking behaviors across different demographic groups and exploring the impact of socioeconomic factors and healthcare policies would also be valuable. Quantifying inappropriate antibiotic use in relation to respiratory illness, as hinted at in the results, is another promising avenue. The authors emphasize the need for active, longitudinal sampling to complement standard surveillance for a more accurate understanding of respiratory virus prevalence and for developing more robust control measures.

Summary updated

Abstract

Standard, healthcare-based respiratory surveillance has multiple limitations. Specifically, ILI is an incomplete metric for quantifying respiratory disease, viral respiratory infection, and influenza infection. The prevalence of respiratory viruses, as reported by standard, healthcare-based surveillance, is skewed toward viruses producing more severe symptoms. Active, longitudinal studies are a helpful supplement to standard surveillance, can improve understanding of the overall circulation and burden of respiratory viruses, and can aid development of more robust measures for controlling the spread of these pathogens.
BibTeX
@inproceedings{galanti2020active,
  author = {Marta Galanti and Devon Comito and Chanel Ligon and Benjamin Lane and Nelsa Matienzo and Sadiat Ibrahim and Atinuke Shittu and Eudosie Tagne and Ruthie Birger and Minhaz Ud‐Dean and Ioan Filip and Haruka Morita and Raúl Rabadán and Simon J. Anthony and Greg A. Freyer and Peter Dayan and Bo Shopsin and Jeffrey Shaman},
  title = {{Active surveillance documents rates of clinical care seeking due to respiratory illness}},
  booktitle = {Influenza},
  year = {2020},
  doi = {10.1111/irv.12753},
}
Figure