Win the adventure of a lifetime!

UNITED24 - Make a charitable donation in support of Ukraine!

Homeland Security


IV. Ensuring Adequate Data for Decision Making: Surveillance Systems


Decisions about how to respond to the fall resurgence of 2009-H1N1 will have to be made quickly in response to rapidly evolving information about the epidemic. The quality of decision-making will depend on reliable and timely estimates of the number of people and specific subgroups that are infected, ill, seeking medical care, being hospitalized, requiring intensive care, and dying from 2009- H1N1; changes in the virus; stresses on health systems; and effectiveness of various medical and public health interventions.

CDC, in close coordination with local and state public health departments, supports a number of impor­ tant systems for surveillance of influenza activity. These systems have provided valuable data through the spring and summer, but they have shortcomings that will limit their ability to provide the data needed to make informed decisions. While it is not possible to remedy all of these limitations before Fall 2009, there are a number of short-term steps that could be taken to significantly improve the data available for decision making.

The Working Group believes there is an important opportunity to upgrade national surveillance systems in time for Fall 2009 by knitting together and extending existing systems. We are aware that CDC is developing plans along these lines and strongly support these efforts.


Decisions about how to respond to the fall resurgence of 2009-H1N1 will have to be made quickly in
response to rapidly evolving information about the epidemic. The quality of decision making in response
to the 2009-H1N1 pandemic will depend on accurate and timely data to answer six key sets of questions:

  1. Approximately, how many people are becoming infected, experiencing illness, seeking medical care, being hospitalized, requiring intensive care, and dying from 2009-H1N1? These data allow esti­ mates of severity, which help determine the intensity of response that is justified. A subsidiary but important challenge is to estimate the same numbers for seasonal strains of influenza that may be circulating over the same period.

  2. How are these numbers changing over time? Are they increasing or decreasing, and how quickly?

  3. Who is becoming infected and who is at greatest risk of severe outcomes (i.e., hospitalization, ICU admission, death)? Specifically, what are the ages, underlying conditions, and other risk factors for infection and severe outcome?

  4. How is the virus changing? Most important, are there changes in illness severity, antigenic char­ acter (which could compromise immunity acquired from natural infection or a vaccine), or drug resistance of the circulating virus?

  5. Are the medical and public health systems able to respond adequately? Is there adequate capacity in physician offices, emergency rooms, hospitals, ICUs, morgues, points of dispensing (PODs), and other public health venues set up for the pandemic?

  6. How well do medical and public health responses work? Does the vaccine protect against infec­ tion or severe outcome? Is the vaccination strategy (e.g., mass vaccination clinics) able to target effectively the recommended population groups? Does antiviral treatment reduce severity? Do social mitigation measures reduce transmission?

Federal decision-makers need data that answer these questions to inform policies and recommenda­ tions about the priority groups for vaccination and treatment, to calibrate the intensity of social mitiga­ tion interventions, and to provide guidance to clinicians about appropriate treatment and prevention. State and local decision-makers need the data for the same reasons, but they also need to understand the situation in their communities, which may differ from the national average. Clinicians need data especially related to questions 3 and 5 in order to target scarce treatment to the appropriate patients, improve clinical treatment, and implement surge capacity plans in the event of increased demands on the health care system. The general public needs to understand the size and severity of the epidemic and be motivated to comply with social mitigation measures. Historically, compliance improves when the epidemic is perceived to be severe. All of these data are needed as close to instantaneously as possible.

Existing Data Streams

CDC, in close coordination with local and state public health departments, supports a large number of systems for surveillance of influenza activity. The output of many of these systems is summarized publicly on FluView, and includes measures (some more nationally representative than others) of outpatient consultation for influenza-like illnesses ( ILI ), hospitalization for influenza, pediatric deaths from influenza, population-wide deaths from pneumonia and influenza, and virus characteristics (subtype and drug resistance). Federal decision-makers have relied on these systems as the main source of data on trends in case numbers, age distribution, and virus characteristics.

A second key source of data on the epidemic in the early days of the spring wave of 2009-H1N1 was the relatively detailed reports from state and local health departments describing individual confirmed cases, noting (with varying completeness) key variables such as age, underlying conditions, and outcome (i.e., recovery, hospitalization, ICU, death). By early May, this level of reporting had become unsustainable; most jurisdictions stopped testing most mild cases for 2009-H1N1 virus and ceased detailed reporting of individual cases. Local authorities in many communities continued gathering data on the most severe cases, but these data were not systematically reported to CDC. Thus the clinical picture of confirmed infection at the national level is relatively static, based on the first case reports in the epidemic, and it has not been possible to track the evolution of the epidemic in the United States .

Shortcomings of Current Data Streams

While the data collected about 2009-H1N1 thus far have been extremely valuable, they have a number of limitations. The key shortcomings of existing data streams are:

  • Some key data are not updated continuously . Since individual-level case reporting ended in early May, there has been no systematic way to update national data on high-risk groups (i.e., according to age and predisposing conditions) for confirmed infection and severe outcome at the national level. Some of these data exist locally but are not being aggregated into a national picture now that reporting to CDC is not at the individual level. Up-to-date information on these variables is needed, for example, to inform decisions on who should receive priority for vaccination and antiviral treatment.

  • Current systems are geographically limited . Influenza activity is geographically heteroge­ neous, as was apparent in the spring wave of 2009-H1N1 and as is known for seasonal influenza. Responses, therefore, should vary locally, but they can do so only with local information. For national decision-makers, geographic coverage is important to ensure a nationally represen­tative picture of the epidemic. Many of the most detailed data feeds, such as the influenza- confirmed cases at hospitalization monitoring sites funded by CDC through the Emerging Infections Program (EIP), are geographically limited. By chance, during the spring none of these EIP sites was in an area with a heavy burden of 2009-H1N1 disease.

  • Current systems do not provide reliable estimates of influenza morbidity and mortal­ity . For many purposes it is critical to know, for example, approximately how many people are infected or hospitalized, measured as total numbers of people or numbers per 100,000 popula­ tion. Most of these systems do not answer that question, but instead measure what proportion of visits to health care providers or emergency departments are for ILI , and what proportion of ILI cases that undergo viral testing are positive for 2009-H1N1.

  • No systematic approach yet exists to monitor the capacity of the health care system to respond . Although many jurisdictions monitor emergency department volume, national inte­ gration of these data is geographically spotty. For total burden of hospitalizations and intensive care admissions due to influenza, few if any representative data are available. Such a system, called “HAvbed,” is planned by DHHS but has not yet been implemented.

  • Laboratory capacity to confirm diagnosis and isolate viruses for further characterization is limited . Most public health laboratories now restrict virus testing to patients with severe disease and many laboratories will be unable to maintain even this practice if the number of cases grows much higher in the fall. Commercial testing for pandemic 2009-H1N1 and other viral respiratory pathogens is not widely available or widely used, so 2009-H1N1 infection in most patients may not be confirmed in fall 2009; as a result, diagnosis will be based empirically on clinical symptoms and knowledge of which respiratory viruses are circulating in each community.

  • Current systems cannot monitor the burden of mild illness that does not come to medical attention . Reliable estimates of this burden are needed to understand the severity of illness— the more people are becoming infected without coming to medical attention, the smaller we expect the overall burden of morbidity, mortality, and health care system stress to be for a given prevalence of infection. At present we do not know this number.

  • Current systems for reporting and analyzing adverse events associated with vaccination may not be well suited to challenges likely to arise during a vaccination campaign for 2009-H1N1 . The Working Group has identified two concerns in this area. First, adverse event surveillance and analysis depends to a large degree on the ability to link vaccination to pos­ sible adverse events via medical records, but the administration of vaccine in settings other than traditional medical care may circumvent this linkage. Second, high-risk groups that are prioritized for vaccination also are likely to experience adverse health events at high rates. Existing systems may not be able to rigorously evaluate elevated rates of such common events in high-risk groups, e.g., spontaneous abortions (miscarriages) in pregnant women or various complications in neurologically impaired patients.

In a rapidly growing pandemic wave, most state and local health departments do not have the capacity to count every hospitalization or death without depleting limited public health resources. Therefore, more efficient and sustainable surveillance methods are necessary to obtain the key data needs during a moderate or severe pandemic, including a qualitative assessment of local influenza activity combined with virologic sampling of a representative number of viral isolates.


It is not possible to address all of these limitations before the autumn wave of 2009-H1N1. (In Chapter 8, we recommend long-term measures to erect a comprehensive and population-based influenza sur­ veillance system that would address data needs for decision making in seasonal influenza and future pandemics.)

The Working Group believes, however, that CDC can take a number of steps in the coming weeks to significantly improve critical data for decision making.


We recommend that DHHS take rapid advantage of available opportunities to upgrade national surveillance systems to improve decision making during the fall resurgence The critical surveil­ lance information for decision making includes data on influenza-like symptoms in the population, emergency room admissions, health system utilization, hospitalized patients, and adverse events.

Below, we suggest several specific measures that may improve situational awareness and decision making through the autumn wave. These recommendations attempt to balance the need for improved data with the practical constraints of assembling systems to acquire these data in a short time frame. We recognize that efforts to address many of these needs, and many other aspects of surveillance, are ongo­ ing; we highlight here aspects that appear to be both urgent and addressable within a short time frame.


We recommend that CDC rapidly assemble an integrated system, by combining syndromic sur­ veillance and emergency department data from existing local and state surveillance systems into a geographically representative national network, that rapidly reports total and ILI -related emergency visits.

Needs/gaps in existing systems and possible approaches: Most states and many large cities have imple­ mented their own syndromic surveillance systems in emergency departments. These electronic systems provide valuable information on ILI trends based on symptoms that bring patients to medical care cen­ters. These systems often collect data within 12 to 24 hours of patient visits. However, these state and local systems currently are not integrated, making it difficult to obtain regional or national situational awareness of influenza activity based on reports from individual centers.

For example, the International Society of Disease Surveillance has implemented a simple and flexible integrated system and collects aggregate counts of ILI syndromes by age group in order to monitor and compare ILI activity throughout the United States very quickly (e.g., see the International Society of Disease Surveillance’s DiSTRIBuTE system). However, only nine jurisdictions (a mix of cities, counties, and states) participate. This system could form a natural template for additional data feeds. We believe it may be feasible to expand this or other systems in the coming weeks and we are aware of efforts by CDC to do so.

Expected benefits : This system would provide a national picture, with some local resolution, of trends in the numbers of patients visiting emergency departments, the percentage of such patients with ILI , and the distribution of ILI by age. Such information would allow Federal, state, and local officials to obtain a better sense of the trajectory of the outbreak (in scale, scope, and pace) in different regions of the United States over time. Systems of this type already are proving useful for evaluation of local control measures, although additional information is required to assess the severity of disease associated with the 2009-H1N1 virus (e.g., cases requiring hospitalization or case fatality rates).


We recommend that CDC implement a system to measure the burden of ILI on a weekly basis. Although nationally representative data would be valuable, it may be beneficial for these surveys to oversample in jurisdictions that have relatively robust surveillance plans in place for tracking influenza-related primary care visits, hospitalizations, and deaths in order to more accurately monitor changes in rates of more severe illness over time.


Needs/gaps in existing systems and possible approaches: Existing systems do not establish the number of ILIs occurring in place and time as a rate per 100,000 population. This precludes estimates of severity because the severe cases, which are better ascertained, cannot be related to overall levels of infection.

This may be accomplished through web-based or telephone-based surveys.

Expected benefits : These studies would provide approximate denominators of mild and medically attended illness against which more detailed data on hospitalizations and fatalities can be compared. Such denominators are especially important for estimating severity of infection and consequently for predicting peak burdens on health care: for a given number of severe outcomes, the overall severity is much less if there are many cases of mild illness or of symptoms that do not cause a patient to seek medical attention. Data from random public surveys would reduce, although not eliminate, the uncer­ tainties cited above concerning overall severity. In addition, the surveys would provide an independent measure of the number of people affected by illness that may be attributable to 2009-H1N1 and to the rate of change in these numbers. The interpretation of ILI activity due to 2009-H1N1 will depend on the proportion of 2009-H1N1 compared to other circulating respiratory viruses in each community where surveillance is taking place. In the spring, 2009-H1N1 was more prevalent; but in the fall, other viruses will likely be circulating, such as respiratory syncytial virus (RSV) and seasonal influenza. Thus, these numbers will be best interpreted in conjunction with virologic data.


Because hospital facilities may become dangerously scarce in the fall, we recommend that DHHS implement an integrated system to monitor health care system utilization overall and attributable to respiratory infection, with an emphasis on incidence and prevalence of cases occupying hospital beds, ICU beds, and mechanical ventilators.

Needs/gaps in existing systems and possible approaches: Hospital and intensive-care utilization are not routinely monitored in the United States . Southern Hemisphere countries are reporting stress on ICUs from 2009-H1N1 illness even during a period of school holidays, and the epidemic probably has not yet peaked. As noted above, DHHS is developing the HAvBED system, which may be expanded to meet present needs. An alternative or complementary approach may be to integrate existing state and local systems, such as the New York State Health Emergency Response Data System (HERDS). In any system, it would be valuable for such data to be immediately available to state and local providers. Since most hospitals maintain such information on a daily basis, the key is to implement a simple system that allows defined information to be regularly uploaded.

Expected benefits: Acute stress on ICUs or increased demand for ventilators may be a trigger for resource reallocation from less affected areas and/or for intensifying community mitigation measures. Accurate measures of health care system utilization would facilitate more efficient sharing of resources.


We recommend that CDC define a mechanism to gather timely clinical, epidemiologic, and virologic data on a representative sample of patients hospitalized for respiratory illness and ensure that those data are available to inform national recommendations to clinicians, public health officials, and the public. Such data could be gathered by assembling a network of participating sites, such as sites currently specializing in influenza surveillance; healthcare systems with appropriate electronic record-keeping systems; and states and localities interested in participating.

The data ideally would include:

    A. Results of systematic testing of patients hospitalized for respiratory infection to determine the presence of respiratory viruses including 2009-H1N1. To improve representativeness of data, such testing would ideally be done within a defined population according to prospective
    criteria rather than according to clinician discretion.

    B. Clinical data—including age, predisposing conditions, course of hospitalization (whether admitted to ICU or ventilated), duration of hospital/ICU stay, and resolution (death, discharge), vaccine status, presence/absence of bacterial secondary infections, and identity and timing of antibiotic and antiviral administration—should be reported for a representative sample of hospitalized cases of 2009-H1N1 infection.
In addition, it would be valuable for CDC to define explicitly the most important clinical studies needed to guide response during the autumn wave, gain Institutional Review Board approval, identify and fund sites to perform these studies during the early autumn, and put in place a mechanism for rapid dissemination of results

Needs/gaps in existing systems and possible approaches: There is an important gap in our ability to assess the clinical features of pandemic influenza infections in an ongoing way to inform treatment and pre­ vention decisions. CDC’s Emerging Infections Program (EIP) reports population rates of infection with confirmed influenza. These data are valuable but are limited by variation in the sensitivities of immuno­ logical and nucleic-acid-based assays and by clinician discretion regarding whether to test. Adequate personnel and funding should be available so that EIP sites have capacity to perform PCR -based tests (which are more sensitive) and to test systematically rather than at clinician discretion.

For clinical information, existing data streams are limited and state and local health departments are unable to follow up most hospital admissions to determine clinical course. Such data are particularly critical and may change over time as the pandemic progresses, either due to changing susceptibility in the population or changes in the virus. While it is not feasible to obtain clinical information for all hospitalized patients, sentinel hospitals or EIP sites could be used to gather detailed clinical data in a standardized fashion. In addition to these standard data, clinical studies—for example, on optimal management of severe cases that do not respond to antiviral therapy—will be needed, and little time is left to ensure that they will be ready to commence early enough to have maximal impact. In addition, waiting for peer-reviewed results to be published will likely diminish the value of any findings, as a manuscript submitted in mid-September might not be published until November, after the possible peak of the epidemic.

To address these needs, CDC should work with existing sites that specialize in influenza surveillance, or research centers, to prospectively monitor for changes in the clinical or epidemiologic characteristics of the virus over time. Other states or locales that have interest and capacity to participate should be included, when possible, to improve geographic representativeness. These “sentinel sites” should use standardized protocols and data collection instruments to ensure that timely and up-to-date clinical, epidemiologic, and virologic data on patients hospitalized for respiratory illness are available to inform national recommendations to clinicians, public health officials, and the public. Adequate funding will be needed to support these sites.

Expected benefits: Such data streams, and CDC’s guidance based on them, would be of primary benefit to clinicians and to vaccine planners in targeting prevention and treatment to groups at high-risk of severe disease. Changes in risk groups or changes in clinical spectrum (e.g., more rapid progression to death, increasing need for ICU care or ventilation among hospitalized cases) may be early signs of changes in the virus or in other factors, such as bacterial superinfection, that would warrant changes in control measures or clinical management. Such changes are not observable now because of the lack of ongoing clinical characterization of severe cases. A rapid means to disseminate clinical data and results of key clinical studies would provide clinicians with needed information while it is most valuable.


We recommend that DHHS ensure the adequacy of surveillance systems and signal evaluation systems for vaccine-associated adverse events (VAE), with particular focus on the risk of common adverse events that are likely to occur at high rates in high-risk populations (e.g., pregnant women) and whose association with vaccination may be difficult to assess rapidly.

In addition, we believe it would be valuable for DHHS to assess the adequacy of existing systems for VAE reporting to detect rare events in settings of nontraditional vaccine distribution (e.g., in public settings, such as malls) and take steps to improve these systems where needed.

Needs/gaps in existing systems and possible approaches: Existing VAE detection systems and surveillance planned for the fall focus on detection of rare complications, such as Guillain-Barré syndrome. In an atmosphere of heightened public concern, common adverse events occurring in high-risk groups likely to be early candidates for vaccination (e.g., spontaneous abortions) may be expected to occur frequently among early vaccine recipients, even if the vaccine is perfectly safe. A mechanism is needed to evaluate the possible contribution of vaccine to such common adverse events to address public concerns, even if the plausibility of such associations is low.

Major existing adverse event detection systems such as CDC’s Vaccine Safety Datalink rely on linked medical records, including vaccination and adverse events for the same persons. If public distribution of vaccine occurs, these systems might not accurately record vaccination status, hence may be unable to function as normal to detect and evaluate signals of adverse events.

Expected benefits: Systems to address vaccine safety are crucial to the success of any vaccination program but will be of particular importance this fall given likely heightened awareness of such issues during a pandemic and during a rapid mass vaccination campaign.


Given the short time until the expected resurgence of 2009-H1N1, it is not feasible to create entirely new surveillance systems. Nonetheless, we believe that it may be feasible to significantly improve surveil­ lance capabilities by upgrading existing systems. Such improvements could have the following direct benefits for decision making.

  • Continuously updated clinical information will provide a basis for national recommendations to physicians, with reliable data on who is at highest risk and which treatments are most effec­ tive for such patients.

  • Emergency department surveillance, combined with a system to monitor demand on hospitals, can provide a considerably stronger basis for decisions about resource allocation to overtaxed areas and for assessing the need for enhanced community mitigation measures to slow demand on the health system.

  • Emergency department surveillance and population-based surveys will inform estimates of the current stage of the epidemic and its trajectory.

  • Adequate reporting and analysis of adverse events is crucial to ensuring vaccine safety and to maintaining public acceptance of the vaccine.

We are aware that CDC is developing plans to expand its surveillance efforts for fall 2009 and we strongly support such efforts.


Join the mailing list