An Evaluation of the Influenza Hospitalization Surveillance Network

Importance

Annually, influenza disseminates worldwide, causing widespread illness and, in severe cases, death. In the 2014–15 season for the United States, laboratory-confirmed influenza-associated hospitalizations reached upwards of approximately 65 cases per 100,000 persons, 30 in 2015–16, 60 in 2016–17, and 102 in 2017–18.¹ Influenza-associated hospitalization cases are organized by age, underlying medical conditions, virus subtype, and cumulative/weekly rates.^1,2 Severity is indexed by accumulating influenza-associated hospitalization case counts and calculating cumulative and weekly (unadjusted) incidence rates using population estimates from the National Center for Health Statistics (NCHS) to estimate hospitalization rates in the United States.¹

The inequities of influenza infection result in time away from work and other societal obligations. The economic losses from the effects of influenza are considerable and the cost of hospitalization because of influenza is substantial. A study published in June of 2018 estimated the average annual total economic burden of influenza to the healthcare system and society was $11.2billion. Direct medical costs were estimated to be $3.2billion, and indirect costs $8.0billion.³ Influenza infection can be largely, but not completely, prevented by vaccination. The CDC’s 2017–18 influenza season vaccine effectiveness study showed that for children between 6 months and 8 years old, there was 68% less incidence of influenza (subtype A or B) in those vaccinated compared to those unvaccinated, while in the elderly population (>65 years old), there was only a 17% reduction of influenza in those who were vaccinated compared to those unvaccinated.⁴ The contents (or viral subtype targets) of influenza vaccines are based on recommendations by the WHO that carefully analyze sentinel surveillance of viral genotyping each year.⁵ Influenza can only be prevented through vaccinations; there is no cure for the infection outside of physician-prescribed antiviral drugs and basic symptom management. Influenza surveillance benefits the public by outlining the severity of each influenza season in an approximation of real time to help drive public health entities’ intervention strategies within the United States.

Purpose

The purpose of the IHSN within the EIP of the CDC is to conduct population-based surveillance for laboratory-confirmed influenza-associated hospitalizations.⁵ The objectives of the IHSN are to determine the time and location of where influenza activity is occurring, track influenza-related illness, determine which influenza virus subgroups are circulating, detect influenza virus mutation events, and measure the influence influenza has on hospitalizations and deaths in the US population.⁴

IHSN-gathered data is used to estimate age-specific hospitalization rates on a weekly basis and display characteristics of persons hospitalized with influenza. Cases are identified by reviewing hospital laboratory and admission databases and infection control logs for patients hospitalized during the influenza season with a documented positive influenza test (ie, viral culture, direct/indirect fluorescent antibody assay [DFA/IFA], rapid influenza diagnostic test [RIDT], or molecular assays, including reverse transcription-polymerase chain reaction [RT-PCR]).⁴ There is no legal requirement to submit influenza-associated hospitalization data to the CDC because it is not a nationally notifiable disease;⁷ however, participation is conditional for each participating state to receive funding from the CDC. The IHSN facilitates integration with other systems by aggregating data collected from individual EIP state surveillance systems (Figure 1).

• Download figure
• Open in new tab
• Download powerpoint

Figure 1.

The IHSN data flow from site location to the CDC, where the data is then inputted into FluView for public use. Additional information from laboratory-confirmed influenza cases provided to the CDC include patient ID number, surveillance site, hospital admission date, patient DOB, influenza test methodology, and identified influenza subtype (A or B).⁸

• Download figure
• Open in new tab
• Download powerpoint

Figure 2.

An example of a 2 × 2 table used to calculate sensitivity and PPV. Test 1 is the method of interest and Test 2 is the method used for reference. The sensitivity calculation is TP/(TP + FN). The PPV calculation is TP/(TP + FP). ²⁵ Note: FN, false negative; FP, false positive; TN, true negative; TP, true positive.

The IHSN conducts surveillance on the individual populations of the 10 EIP-participating states. Data is collected annually and published weekly starting in the beginning of October and ends as late as May. Each of the EIP states have designated counties that contribute data to the IHSN.⁴ Among the 10 states, there are approximately 70 counties whose hospitals contribute data to the IHSN. The IHSN accumulates data from 267 acute care hospitals and laboratories in counties varying in socioeconomic status within the 10 EIP sites. All sites within the EIP are geographically distributed throughout the United States, and encompass approximately 27 million people.⁸ Surveillance officers (usually through EIP-participating public health departments) are trained to collect laboratory-confirmed influenza cases from laboratory logs, infection control practitioner logs, weekly calls to data collection sites (hospitals), or (depending on the state) state-reportable condition logs.⁶ Data is then compiled and sent on a weekly basis to the CDC for analysis and eventual input into the FluView application.1,2 Patient information is recorded with each case in all EIP-participating states. This is because, in contrast to the CDC’s notifiable conditions, laboratory-confirmed influenza (subtype A) is a reportable condition in all EIP states (Table 1) and that same information is required for use at the CDC (Figure 1). However, unique patient information (name, date of birth [DOB], patient identification [ID] number) is encrypted and securely sent, and is not published in weekly surveillance reports, nor is it inputted into the FluView application.

Resources Used

The IHSN is primarily financed by core funding for operation and personnel training provided to the EIP by the CDC.^8,9

Usefulness

Through the FluView Interactive application, the IHSN uses laboratory, hospital admission database, and infection control logs to capture hospitalized cases with a documented positive influenza test result during the regular influenza season.^1,2 This is a comprehensive approach for accumulating data. The IHSN addresses the variability of testing methods by outlining the Food and Drug Administration (FDA)-cleared, or the Clinical Laboratory Improvement Amendment–waived influenza testing method that includes, but is not limited to, viral culture, DFA/IFA, RIDT, or nucleic acid–detecting molecular assays.²

Simplicity

FluView application allows for real-time data access and can differentiate cumulative rates based on age group, EIP state, and influenza season. Data is gathered by weekly reports to the CDC Influenza division by each EIP-participating state (Figure 1). The 10 states participating in the EIP that contribute data to the IHSN FluView application are California, Colorado, Connecticut, Georgia, Maryland, Minnesota, New Mexico, New York, Oregon, and Tennessee. Georgia, Maryland, and Tennessee only require that influenza subtype A be reported to the state health department. All other states require all hospital-confirmed influenza cases to be reported to their state health department authorities (subtypes A and B).^{11⇓⇓⇓⇓⇓⇓⇓⇓-20}

Flexibility

Influenza can undergo antigenic drift, which are changes made (through mutation) to its varying subtypes. Because of antigenic drift, previous vaccination targets (subtypes) are then less effective at preventing infection in the population, making influenza difficult to control each year.²¹ Considering the unpredictable nature of influenza, The IHSN has a high degree of flexibility between influenza seasons. The IHSP can adjust to each influenza season by adding additional reporting sites outside of the EIP states (sites).⁶ The 2009–10 H1N1 pandemic prompted this change in the IHSP’s surveillance capacity. Additionally, the IHSP can also remove sites as needed. This has potential to compromise the longitudinal validity of data gathering and analysis. Each EIP-participating state has their own unique criteria for reportable conditions (Table 1), which can also compromise the validity IHSN data. However, aggregation of data at the CDC level is simplified because of their strict criteria for each case report (Figure 1).⁸

Data Quality

Consistent surveillance officer training at EIP sites mitigates variability of the data accumulation process at the state level. The IHSN uses NCHS data to form population estimates used in rate calculations when calculating weekly and cumulative influenza-associated hospitalization rates.¹ However, each test method outlined within the CDC’s “Information for Clinicians on Influenza Virus Testing” has variable sensitivity and PPV measures (Table 2).²² This variability has the potential to compromise the overall reliability of rate calculations used in the FluView application via underreporting caused by inaccurate test results (false negatives).

Acceptability

For the IHSN EIP sites to receive funding from the CDC, they are required to comply with basic reporting standards of the CDC’s national notifiable conditions. By having trained surveillance officers for collection of relevant information (and paying them to do so), this allows EIP sites to participate in the IHSN, ensuring as much data is provided as possible. Apart from 3 participating sites (Table 1), laboratory-confirmed influenza (A and B subtypes) is a state-reportable condition ensuring compliance at the site level. Failure to report a “reportable” or “notifiable” condition by a hospital or physician office subjects them to potential revocation of individual medical licenses or operating licenses of the institutions (hospitals) at fault.²³

Sensitivity and Positive Predictive Value

Table 2 includes a compilation of 3 tests each selected from the “Available FDA-Cleared Rapid Influenza Diagnostic Tests (Antigen Detection Only)” and the “FDA-Cleared Nucleic Acid Detection Based Tests for Influenza Viruses” pages on the CDC’s website, ^22,24 and the sensitivity/PPV calculations for each test. Test selections were made by numbering each test in each table and submitting them into a random number generator. Calculations were performed using “Nasopharyngeal Swab” sample data.

The clinical sensitivity of all 3 nucleic acid testing methodologies ranges from 90% to 100%, while for antigen detection methods, they range from approximately 84% to 97% for influenza subtype A. The confidence that a detected positive value is actually positive within the patient for nucleic acid testing methods are all almost universally 100%, whereas antigen detection tests only had a range of approximately 75%–93% confidence in positive values for influenza subtype A.

The IHSN is heavily reliant on the accuracy of influenza testing methods at the individual laboratories within the EIP states’ participating counties. Sensitivity and positive predictive values were determined at individual testing levels to address this at the IHSN level. There are currently no criteria for confirming positive influenza tests within the IHSN. Confirmation testing for positive results is left to the discretion of the EIP-participating states. Table 1 indicates only 3 EIP-participating state health departments require confirmation testing on all positive influenza tests. The lack of confirmation testing could lead to an inflation of false positive test results on methods with a lower positive predictive value. Table 2 outlines the differences in sensitivity and positive predictive values between the 6 selected tests. It is noted that there is a lot of variability in sensitivity and specificity among the different test types.

Representativeness

The IHSN has a high degree of representativeness in terms of geographic distribution of counties within the EIP-participating states and of the EIP states themselves. This allows for a stratified approach to IHSP data collection, which helps published data to be more generalizable to the rest of the United States.

A key challenge is accurate representation of a grossly underreported disease like influenza.^32,33 The CDC has struggled for decades to adjust and refine their models to determine epidemic thresholds and determination of seasonal severity. This is because of changes in diagnostic technology, access to diagnostics, and modeling techniques.^34⇓⇓-37 It is important to note that population-based estimates of influenza are based on census data, which is also based on statistical models that have evolved over the decades as well. The dichotomy of having more cases reported may result in stimulating media reporting, which in turn stimulates patient demand that stimulates healthcare providers to order influenza testing. Because of an increase in influenza molecular testing options, increased access of testing options to physicians can cause them to overscreen, which can lead to an artificial inflation of positive influenza cases that may or may not be contributing to patient hospitalizations.³⁸ The IHSN counts all hospitalizations that have a laboratory-confirmed positive influenza test. Artificial inflation of positive cases in the form of overscreening, combined with the IHSN case definition, can lead to a misrepresentation of the population’s influenza-associated hospitalization rates. This raises concerning questions regarding the scientific basis upon which we claim severity: is it based on antigenic shift (ie, a pandemic), or more accurate statistics for an underreported disease?

Timeliness

Each EIP IHSN state has variable reporting conditions and timelines for influenza (Table 1). All participating states require all laboratory-confirmed influenza cases to be reported to the state health department. The reporting timeframe for influenza in each state ranges from immediate to reporting “within 7 days” (Table 1). The CDC estimates there to be a median 7-day lag time from the time a case is identified to when the CDC receives the report for the IHSN.⁶ It is unclear as to whether the IHSN inputs influenza cases using the identification date at the laboratory level or the date the CDC received the data. However, a 7-day lag time between identification and reporting to the CDC is fairly rapid considering the geographical distribution of EIP sites and frequency of influenza cases.

Stability

There have been no significant events or available evidence that suggest the stability of the IHSP and their FluView application have ever been compromised. The IHSP provide weekly updates and there have been no notable delays in updates as of 2018.