Using national pandemic influenza planning documents, CDC guidance, and the recent medical literature, we developed a telephone survey questionnaire (Table 2) that asked “Do you have the flu?” and questions about symptoms such as fever, cough, and sore throat. This design allows comparison of the survey data with the SPN’s definition of ILI (i.e., fever [>100°F (37.8°C)] and cough and/or sore throat in the absence of a known cause other than influenza). Possible risk factors for exposure to novel avian influenza strains through live poultry (7), foreign travel (8,9), and cats (10) were also addressed in the survey.

To explore the feasibility of using this telephone survey, we partnered with Zogby International (Utica, NY, USA) (11), a private public opinion research company experienced at collecting extensive demographic and socioeconomic information through nationwide telephone surveys. A protocol to perform telephone surveys was submitted January 11, 2006 for institutional review board (IRB) review, subsequent questionnaire modifications were approved March 20, 2006, and the first survey was completed the week of April 24, 2006, ≈110 days later. Monthly, ≈1,205 randomly selected adults, who were already providing demographic and socioeconomic information for other Zogby International telephone surveys, were asked to participate in a research study about influenza. Participants answered follow-up questions over 5-day intervals ending on each of the following dates in 2006 (corresponding to CDC influenza surveillance calendar weeks) (12): April 24 (week 17); May 16 (week 20); June 6 (week 23); July 25 (week 30); August 15 (week 33); September 14 (week 37); and October 12 (week 41). These dates were selected as a convenience sample based upon having received necessary IRB approvals and having the availability of telephone survey capacity by our telephone surveying partners. Telephone numbers for the surveys were randomly selected from commercially available national data sets of residential directory-listed telephone numbers. The probability of selection for the telephone survey was adjusted to be proportional to population sizes within area codes and telephone exchanges. As many as 6 calls were made to reach a sampled phone number, and the procedures did not result in the same telephone numbers selected to be surveyed for each subsequent month.

One adult from each randomly sampled household was asked to answer the survey questionnaire by telephone. All interviews were conducted by Zogby International’s general interviewers, who are monitored by supervisors to assess adherence to surveying standards. For this study, the interviewer-to-supervisor ratio was 12:1. Quality control checks of interviewer performance were conducted on 10% of all calls. The survey was performed according to a protocol created by medical researchers at the University of Texas Health Science Center at Houston and approved by the university’s Institutional Review Board. Survey results were tabulated, analyzed, and correlated by the study authors with nationwide influenza surveillance reports from CDC (Table 1) (2) for the 2005–06 and 2006–07 influenza seasons (1,2,12).

Of 8,449 adults contacted, 7,268 (86%) agreed to participate in the survey. Participants were from representative socioeconomic and racial/ethnic groups (Table 3) across the United States. The overall monthly participation rate was 83%–87%. Participation rates for men and women were similar.

Of the 7,268 adults surveyed by telephone during 7 surveys (CDC weeks 17–41), 2,337 (32%) said that they had cats in the home, 184 (2.5%) lived in close contact with chickens, and 230(3.2%) answered “yes” when asked “Do you have the flu?” Of the 230 adults who answered yes to having the flu, only 49 (21%) reported having received an annual influenza vaccine, 68 (30%) reported having fever or abnormal body temperatures, 93 (40%) reported having a cough, and 89 (39%) reported having a sore throat (Table 4). According to CDC influenza surveillance reports for weeks 17, 20, and 41 (Table 5), influenza activity nationwide peaked in early March 2006 (at approximately week 10); influenza B virus was the most commonly isolated influenza virus during weeks 17 and 20, and influenza A virus was the most frequently isolated virus in week 41 (12). Data for weeks 21–39 in 2006 were not reported by CDC in the 2005–06 and 2006–07 influenza seasons (12).

As shown in Table 4, our surveys of 7,268 adult respondents provided nationwide, prepandemic baseline information about household contact with cats (10) (28%–37%) and chickens (1%–4%), factors that may be relevant to the spread of currently circulating strains of avian influenza A (H5N1) virus. The data collected in 7 surveys over a 7-month period of low influenza activity were consistent with surveillance data gathered and reported by SPN.

More than 80% of persons in the populations and demographic subgroups surveyed nationwide agreed to participate in our study; this rate was similar for all regions of the country. The rate of self-reported illness from “flu” ranged consistently from 2% to 5% over the 7-month survey period, a time of low influenza activity and without reported human illnesses from highly pathogenic avian influenza A (H5N1) in the United States. Because only 30% (68/230) of respondents who self-reported the flu also self-reported fever, it follows that only a very small fraction of all respondents (0.9%, 68/7268) may have met CDC’s definition of ILI (i.e., fever [>100°F (37.8°C)] and cough or sore throat) (12). Similarly, during the same time frame, the SPN reported outpatient ILI rates of 1.2% and 0.9%, and the other CDC surveillance systems (Table 5) reported only regional and sporadic ILI in most states; moreover, none of the CDC-measured indicators suggested that influenza-related illness or death were excessive on the dates our surveys were conducted.

A comparison with CDC surveillance data suggests that the household telephone survey produced plausible, reproducible, and accurate results during a period when the circulation of influenza (predominantly influenza B virus) in the community was minimal. These results, when interpreted in the context of all other applicable surveillance reports, suggests that a direct telephone survey of adults at home could improve situational awareness of an influenza pandemic.Had unexpected trends or inconsistent rates of illness or apparent geographic disparities been identified, an analytic study could have been conducted to suggest possible risk factors or further investigations.

According to the DHHS Pandemic Influenza Plan, “Some states are considering the use of systematic phone surveys to supplement SPN data during a pandemic by providing estimates of local cases and affected households. CDC will explore the utility and feasibility of conducting this type of survey on a national level” (2). In a recent national telephone survey of 2,075 persons in Sweden, responses to the question “Did you have the ‘flu’ last week?” provided useful public health information regarding ongoing influenza disease activity (13). The survey took only 125 working hours to complete and cost approximately €3,250 (US $4,150). On the basis of our results, and assuming appropriate IRB approvals are already in place, we estimate that a new questionnaire could be developed in conjunction with public health officials and the telephone survey partners within 1 to 2 days, and the surveys could be conducted over the next 2 days with results delivered essentially immediately to public health officials. The direct cost of conducting each of the 2- to 3-day telephone surveys involving 1,200 adults and asking 10 questions was estimated to be $14,000; the 7-survey project in Tables 2–4 had a direct cost of ≈$98,000. In the United States, CDC recently adapted an ongoing national household telephone survey project, the Behavioral Risk Factor Surveillance System (BRFSS), to measure and monitor influenza vaccination coverage during the 2004–05 influenza season (14). CDC surveyed 26,526 adults during February 1–27, 2005 (14), and reported that ≈6,363 (24%) had been vaccinated. This CDC study also showed that influenza vaccination coverage among adults through January of the 2004–05 influenza season was greatest among persons >65 years of age (62.7%). In the telephone survey reported here, 21% (49/230) of those adults who self-reported the flu had received an annual influenza vaccine. The rates of influenza immunization reported in our August, September, and October 2006 surveys were lower than those in the early months of 2006, possibly because the 2006–07 trivalent immunizations were not yet available through healthcare providers at that time; however, this trend in coverage was similar to the monthly coverage trends reported by CDC’s BRFSS in 2004–05 (14).

Commercial polling agencies already elicit personal information from the public on an ongoing basis. Recently, as demonstrated by its adaptation of the BRFSS system, CDC was able to obtain information in response to a new public health problem by adding a few questions to an existing telephone survey. Although this adaptation of the survey, data analysis, and reporting infrastructure of the BRFSS appears to have been successful for determining influenza health status, it has apparently not yet been translated into a permanent surveillance system or ongoing capability. Moreover, the BRFSS might be overwhelmed in a pandemic emergency and, on very short notice, be forced to use additional telephone surveys that exist in the private and academic sectors. The adaptation of a private opinion survey company’s capabilities in partnership with an academic medical center through approved protocols could create a feasible, safe, inexpensive, flexible, and acceptable way of deriving public health information in emergencies and improving situational awareness.

Our study has several limitations. First, the telephone surveys were conducted only during months with low influenza activity in 2006. Second, although 7,268 adults were asked if they had “the flu,” only those who responded affirmatively (230 adults) were asked further questions about their influenza vaccination status and the presence of fever, cough, and sore throat. Biases may have been introduced based on the sequential approach to these questions, because some persons with fever, cough, or sore throat may have been misclassified by answering “No” to have the flu; additional follow-up questions about these ILI-related symptoms were not asked of all those surveyed. Third, this study had limitations similar to those of CDC’s BRFSS: 1) being a land-line telephone–based survey, our study excluded adults in households without telephones and adults who use only cellular telephones; and 2) because the data were self-reported and subject to recall bias, especially for questions that required recall over a longer period, frequency estimates might be less precise for some conditions or behaviors. Fourth, our estimates of the proportion of adults with nondirectory-listed telephone numbers including those who had ILI (12) were based on unmeasured self-reported temperatures (13) rather than on direct observations by healthcare providers, as in the SPN intended for ambulatory populations in healthcare facilities and clinics. Fifth, other illnesses and other viral infections besides influenza can cause ILI and can be accompanied by fever, cough, or sore throat. Finally, there was no laboratory confirmation of influenza or ILI in our survey participants.

Nonetheless, we were able to analyze our data in terms of self-reported fever, cough, or sore throat, thereby enabling us to estimate the proportion of respondents who met CDC’s definition of ILI (12,13). This capability might be useful during an actual pandemic, when it might be more desirable to assess the health status of patients by telephone (15) rather than exposing these and other patients and healthcare workers to the risk for healthcare-associated transmission of respiratory pathogens in clinical settings. Although the SPN’s nationwide estimate of ILI, determined using CDC criteria, was similar to our telephone survey-based estimate for the periods of low influenza activity in 2006, more data are needed to assess the performance of these methods during seasonal influenza peaks and during a pandemic. Others have suggested that sentinel surveillance studies typically underestimate influenza in a population; thus, telephone surveys may prove to be an increasingly important component of influenza surveillance (13).

Telephone surveys might offer a practical solution to addressing the gaps in knowledge of influenza health status that might arise during a pandemic. Further telephone surveys should be performed during the peak influenza season to determine whether such an approach to surveillance would be a useful addition to ongoing influenza surveillance systems. The usefulness of the telephone survey to gain influenza immunization history and current ILI information on all people at risk should also be explored.