By Cody J. Benthin, MD

Staff Physician, Pulmonary and Critical Care Medicine, Northwest Permanente, Portland, OR

Dr. Benthin reports no financial relationships relevant to this field of study.

SYNOPSIS: Surrogate markers of influenza severity, specifically trends in ICU use, were collected and revealed differences from current influenza reporting.

SOURCE: Baker AW, Edmond MB, Herwaldt LA, et al. Real-time surveillance of influenza morbidity: Tracking intensive care unit resource utilization. Ann Am Thorac Soc 2017;14:1810-1817.

The annual seasonal severity of influenza varies and results in yearly fluctuation of healthcare system use, including ICU resources. Current real-time reporting of influenza morbidity and mortality within the United States is limited to hospitalization and influenza plus pneumonia composite mortality as presented by the CDC’s weekly update.1 These data do not predict the need of hospital resources precisely, including ICU beds, ventilators, or use of extracorporeal membrane oxygenation (ECMO). The CDC’s online Influenza Hospitalization Surveillance Network database provides a more comprehensive description of severity, including ICU use (but only retrospectively).

The aim of this study was to determine if a surveillance system using sentinel hospitals within the United States that serially recorded days of ICU admission, mechanical ventilation, and ECMO use on a weekly basis could provide an accurate and opportune description of influenza severity and associated resource use. Baker et al developed a pilot study to collect data retrospectively from a 36-week period of the 2013-14 influenza season at three tertiary care hospitals, which included 2,408 beds in three separate states. Subjects included were all patients requiring inpatient admission diagnosed with influenza based on a positive respiratory sample polymerase chain reaction (PCR) or who had received an influenza-related diagnosis code (ICD-9).

The weekly number of influenza admissions, patients started on mechanical ventilation, or ECMO were recorded. The days accrued in the ICU, on mechanical ventilation, or on ECMO also were tracked. Data from Hospital A were extracted through electronic health record (EHR) queries, whereas manual chart review was completed at Hospitals B and C. Overall, 431 patients were identified to be hospitalized with influenza from Aug. 4, 2013, through April 12, 2014. Influenza admissions represented 0.6% of the total 76,968 admissions, and 81% of the patients were adults (> 18 years of age). The majority were identified by positive PCR (83%).

Twenty percent of patients required mechanical ventilation, averaging 12.5 days on a ventilator. Four percent of patients required ECMO, averaging 13.5 days. Deaths varied between hospitals (2.8% at Hospital B and 12.5% at Hospital C).

Weekly trends in admissions and ICU use varied between hospitals, with Hospital A demonstrating an earlier peak of resource use, which coincided with hospitalization and death. However, Hospitals B and C showed a later peak of hospitalization, which was followed by later peaks in ICU resource use. The combined network influenza hospitalization rate declined rapidly following its peak; however, rates of ICU, mechanical ventilation, and ECMO use remained elevated.


The findings of this study demonstrate that trending weekly rates of influenza-specific ICU resource use, including beds, mechanical ventilation, and ECMO, is feasible, and suggests the possibility that multi-center surveillance of influenza outbreaks in this manner could add to existing real-time severity updates. Tracking trends in ICU beds and mechanical ventilation use may be superior to death given the relatively low rate of the latter endpoint. As demonstrated in the data collection of this study, advances in EHR allow for robust data queries within health systems. These may require advanced coding techniques initially, but subsequent queries may be substantially more efficient. We know that there are limitations in our current traditional healthcare-based systems for reporting influenza activity, and internet-based data sources such as Google have been used to produce activity estimates ahead of these systems.2,3 As seen in this study, regional differences of peak influenza rates would be expected to produce similar differences in ICU use. In fact, there are efforts underway to develop accurate local (city or regional) influenza monitoring and forecasts based on these advances.4

Influenza continues to be a major challenge affecting health systems, requiring periods when ICU resources are in higher demand. These outbreaks vary depending on the timing and severity of each season, the virulence of circulating strains, and the populations affected. Given these points, the real-time monitoring and reporting of ICU use may contribute significant knowledge, which would aid in planning for anticipated resource need.


  1. Centers for Disease Control and Prevention. Overview of influenza surveillance in the United States. Available at: Accessed April 5, 2018.
  2. Yang S, Santillana M, Kou SC. Accurate estimation of influenza epidemics using Google search data via ARGO. Proc Natl Acad Sci U S A 2015;112:14473-14478.
  3. Xu Q, Gel YR, Ramirez Ramirez LL, et al. Forecasting influenza in Hong Kong with Google search queries and statistical model fusion. PLoS One 2017;12:e0176690.
  4. Lu FS, Hou S, Baltrusaitis K, et al. Accurate influenza monitoring and forecasting using novel internet data streams: A case study in the Boston metropolis. JMIR Public Health Surveill 2018;4:e4. doi: 10.2196/publichealth.8950.