Description and Evaluation of the 2009–2010 Pennsylvania Influenza Sentinel School Monitoring System

Abstract

Objectives. We described and evaluated the 2009–2010 Pennsylvania Influenza Sentinel School Monitoring System, a voluntary sentinel network of schools that report data on school absenteeism and visits to the school nurse for influenza-like illness (ILI).

Methods. Participating schools provided daily absenteeism and ILI data on a weekly basis through an online survey. We used participation and weekly response rates to determine acceptability, timeliness, and simplicity. We assessed representativeness by comparing participating schools with nonparticipating schools. We compared monitoring system data with statewide reports of laboratory-confirmed influenza.

Results. Of the 3244 Pennsylvania public schools, 367 (11%) enrolled in the system. On average, 79% of enrolled schools completed the survey each week. Although the peak week of elevated absenteeism coincided with the peak of statewide laboratory-confirmed influenza cases, the correlation between absenteeism and state data was nonsignificant (correlation coefficient = 0.10; P = .56). Trends in ILI correlated significantly with state data (correlation coefficient = 0.67; P < .001).

Conclusions. The school-based sentinel system is a simple, acceptable, reliable device for tracking absenteeism and ILI in schools. Further analyses are necessary to determine the comparative value of this system and other influenza surveillance systems.

The emergence of pandemic influenza A (pH1N1) in spring 2009 had a disproportionate impact on school-aged children.^1,2 More than half of the reported cases of pH1N1 in the United States were among children younger than 18 years.³ Although preexisting immunity in older age groups may have contributed to this pattern of disease,⁴ schools may also have played an important role in amplifying transmission of the virus.⁵

School-aged children are known to play an important role in the epidemiology of influenza.^6–8 A typical ill student infects an estimated 2.4 (95% confidence interval [CI] = 1.8, 3.2) other children within a school.⁹ Children may also serve as the means of introduction of influenza infection into households during seasonal influenza outbreaks.^10–13 Attack rates in adults who reside with school-aged children may be 2 to 3 times higher than are attack rates in similar adults who do not reside with school-aged children.⁸

Because pH1N1 heavily affected school-aged children, and because children play an important role in the transmission of flu, it is important to closely follow the occurrence of influenza and related indicators in this population. In fact, influenza activity in schools may be an important early indicator of activity in the general population,¹⁴ and schools are of particular importance for controlling influenza.^7,15,16 Although it is not possible to directly monitor influenza in all schools, using sentinel school absenteeism data as a proxy for data on influenza prevalence in the community may be warranted.^14,17,18

Recognizing the atypical epidemiology of pH1N1 and the importance of early influenza detection, local, state, and federal public health agencies made schools an early focus of monitoring, prevention, and control activities. In May 2009, the Pennsylvania Department of Health (PADOH) recommended that all schools monitor and report to public health authorities excess absenteeism that may be related to influenza. In a follow-up to this recommendation, for the 2009–2010 school year the PADOH established a voluntary sentinel network of elementary, middle, and high schools around the state to report weekly counts of student absenteeism and visits to the school nurse for influenza-like illness (ILI). After 1 school year of collecting surveillance data from schools, we evaluated this monitoring system.

METHODS

In September 2009, we sent the superintendents of all Pennsylvania public school districts letters and e-mails inviting them to participate in the Pennsylvania Influenza Sentinel School Monitoring System. The main purpose of the monitoring system was to track influenza-related measures, including absenteeism and ILI, in school-aged children. To participate, districts identified schools willing and able to provide data. Schools interested in becoming sentinels registered by completing an online enrollment form. Enrolled schools identified a primary and secondary contact person to receive weekly e-mails and other correspondence.

Data Collection

During the school year, we sent out 36 surveys covering the period between September 14, 2009, and June 4, 2010. We sent a weekly e-mail to the sentinel schools’ contact persons on Monday morning that contained a link to a secure online survey. Although both the primary and secondary correspondents received the weekly e-mail reminder, only 1 of the 2 correspondents was to complete the online survey each week. Participants had until 12:00 pm the next day to complete the survey. Correspondents reported the following numbers for each day of the previous week: (1) the number of students absent (for any reason) from school, and (2) the number of students seen in school for ILI by the school nurse or a representative. We defined ILI according to Center for Disease Control and Prevention guidelines, which included having a fever (> 100°F) or being feverish and having a cough or sore throat. Schools were also able to report any unusual patterns of illness. We did not request any individually identifiable student data.

Analysis and Dissemination of Data

Once the weekly online survey was closed, we downloaded the data, examined them for duplicate responses and other data problems, and appended them to a cumulative SAS version 9.1.3 (SAS Institute, Cary, NC) data set. We categorized schools by grade level according to the highest grade in the school, with schools ending at grades K through 6 categorized as elementary school, schools ending at grades 7 through 9 as middle school, and schools ending at grades 10 through 12 as high school. Analyses performed on a weekly basis included the following:

1. number of schools reporting and total enrollment of reporting schools;

2. number of schools reporting elevated levels of absenteeism or ILI;

3. percentage of schools exceeding an arbitrary absenteeism threshold by grade level;

4. percentage of schools exceeding an ILI threshold by grade level;

5. weekly average absenteeism rate by region and grade level, and

6. weekly average percentage of students seen for ILI by region and grade level.

We disseminated weekly findings to Pennsylvania public health field personnel. We distributed summaries to participants in December and at the end of the school year.

Evaluation Criteria

We evaluated several of the Pennsylvania Influenza Sentinel School Monitoring System attributes as recommended by the Center for Disease Control and Prevention's guidelines for the evaluation of public health surveillance systems.¹⁹ For our evaluation, we assessed the system's simplicity, flexibility, acceptability, representativeness, timeliness, stability, and data quality. We assessed representativeness by comparing characteristics of participating schools with nonparticipating schools. We used participation rates and weekly response rates to determine schools’ overall acceptance of the system. In interviews with key personnel, we assessed the basic structure of the system and its ease of operation, which we used to appraise simplicity, stability, and timeliness. At the end of the school year, we also invited correspondents to complete a brief online survey assessing their experience with the monitoring system.

We also examined the relationship of community levels of influenza as measured by laboratory reporting to levels of school ILI and absenteeism. In Pennsylvania, clinical laboratories are required to report cases of laboratory-confirmed influenza within 1 business day of test completion (028 Pa. Code § 27.22). We used data from September 25, 2009, through June 4, 2010, to assess the correlation between the number of laboratory-confirmed influenza cases in the state and the number of schools with elevated levels of absenteeism and ILI visits to the school nurse. We did not include data from December 14, 2009, through January 1, 2010, because schools were on winter break and the surveillance system did not collect data during this period. We used baseline data from the first few weeks of the school year to determine whether a school was experiencing higher than usual levels of absenteeism at any time during the school year.

We defined elevated absenteeism as an 80% or greater change in absenteeism over the school week with the highest absenteeism rate over the 90th percentile for the grade level (8% for elementary school, 9% for middle school, and 11% for high school). We arbitrarily defined elevated ILI as 2% or greater of students seen with ILI on any day, with a minimum of 6 students seen in a single day. We measured the association between the laboratory reporting of influenza and school monitoring data using Spearman rank-order correlation coefficients (SAS version 9.1.3).

RESULTS

The results of our evaluation of the Pennsylvania Influenza Sentinel School Monitoring System and its relationship to influenza testing reporting are as follows.

System Attributes

Representativeness of system.

During the 2009–2010 school year, 367, or just more than 11%, of the 3244 Pennsylvania public schools were enrolled in the monitoring system. More than half of the sentinel schools were elementary schools (58%, n = 214), 19% (n = 69) were middle schools, and nearly 23% (n = 84) were high schools (Table 1). These proportions were similar to the proportion of elementary, middle, and high schools in Pennsylvania as a whole (P = .71; elementary schools, 56%; middle schools, 19%; high schools, 25%). Approximately 19% (n = 95) of the state's 500 school districts, 70% (n = 47) of the 67 Pennsylvania counties, and all 6 Pennsylvania public health regions had at least 1 school enrolled in the monitoring system. However, some regions were under- or overrepresented because the percentage of sentinel schools in each public health region was not proportional to the total number of public schools in each region (Table 1; P ≤ .001).

TABLE 1.

Characteristics of All Pennsylvania Public Schools and Schools Enrolled in the Pennsylvania Sentinel School Monitoring System: September 2009–June 2010

Characteristic	All Pennsylvania Schools (n = 3244)	Enrolled Schools (n = 367)
Total no. of students	1 801 971	214 685
Grade level, %
Elementary	56	58
Middle	19	19
High	25	23
Geographic region, %
North central	6	12
Northeast	11	13
Northwest	9	17
South central	14	18
Southeast	36	25
Southwest	24	15

Simplicity of system.

The surveillance system was relatively inexpensive and easy to establish and maintain. It took less than 3 weeks to create the enrollment survey, contact and recruit schools, and establish the data collection surveys and processes. We used an automated e-mail merge system to contact correspondents and direct them to the weekly online data surveys. The only cost for running the surveillance system was the license for the Internet-based survey tool. However, the online survey allowed users to submit data for only 1 school at a time. Therefore, the reporting process was more time consuming for individuals who were responsible for submitting data for multiple schools.

Flexibility of system.

The surveillance system was rather flexible because it required little additional time to enroll and collect data from additional schools. Schools that were interested in becoming sentinels after the initial enrollment period were able to contact the PADOH and be added to the system. However, few schools enrolled after the initial enrollment period.

Stability of system.

The surveillance system required minimal maintenance and management because any person in the PADOH Bureau of Epidemiology with access to the school contact list and survey link could send out the weekly e-mail.

Timeliness of system.

Data collected each week reflected the schools’ experiences from the previous week. Thus, when the collection surveys closed the data were already 4 to 8 days out of date. By contrast, influenza test results were generally reported to PADOH within 1 day of completion and often on a real-time basis.

Acceptability of system.

Of the 147 correspondents who completed the year-end evaluation, 85% (n = 115) were primary correspondents, 14% (n = 19) were secondary correspondents, and the remainder were unsure. When asked whether they would be interested in continuing the system next year, 25% were very interested, 55% were somewhat interested, and 20% were not interested. Several correspondents expressed interest in continuing if given the ability to submit information on more than 1 school at a time, whereas some did not think it was necessary to continue reporting after flu had subsided.

Data completeness and quality.

On average, 79% of enrolled schools completed the survey each week (range 56%–92%). Approximately 70% (n = 259) of the enrolled schools reported data at least 27 (75%) of the 36 weeks for which we requested data. Of these “regular reporters,” 62% were elementary schools, 17% were middle schools, and 21% were high schools. These proportions were similar to the percentage of total elementary, middle, and high schools enrolled in the system.

Correspondents had to manually enter their school name and identification code each week, resulting in occasional errors. Also, some schools occasionally submitted duplicate records. Substantial effort was required each week to clean the data before analysis. The online data collection tool forced correspondents to answer all questions, so there were no unknown or blank responses.

System Results and Comparison With Influenza Test Reporting

School absenteeism and ILI visits to the school nurse.

On average, the percentage of schools with elevated levels of absenteeism was 14.4% per week. Absenteeism was highest during the week ending October 30, 2009, when 36% (107 of 297) of schools reported elevated levels of absenteeism. On average, the percentage of schools with elevated levels of ILI was 2.7%. Visits to the school nurse for ILI were highest during the week ending October 23, 2009, when 21.7% (77 of 313) of schools reported elevated levels of ILI. When we stratified ILI and absenteeism data by school level, elementary and middle school absenteeism peaked during the same week

Absenteeism in high school peaked during the end of the school year (May 28, 2010). The peak week of ILI was the same for all school levels.

Comparison of trends.

Figure 1 shows the relationship between the number of laboratory-confirmed cases of influenza in the state and the number of schools with elevated levels of absenteeism and ILI. Visual inspection of the figures indicates that school absenteeism and ILI data correlated with state data. The number of laboratory-confirmed influenza cases in Pennsylvania peaked during the week ending October 30, 2009, when there were 8576 cases. The peak week of elevated absenteeism coincided with the peak of laboratory-confirmed influenza cases and the peak week of ILI visits occurred 1 week before. However, absenteeism also displayed several other peaks throughout the year that did not correspond well to laboratory reporting data. Similar patterns were seen when restricting laboratory data to confirmed influenza cases among those aged 5 to 19 years (Figure 2). Results remained the same after graphing the percentage of schools with elevated absenteeism or ILI and the number of laboratory-confirmed influenza cases in Pennsylvania (results not shown).

FIGURE 1.

Laboratory-confirmed influenza cases, schools reporting excess levels of absenteeism, and schools reporting excess visits to the school nurse for ILI per week: Pennsylvania, September 2009–June 2010.

Note. ILI = influenza-like illness.

FIGURE 2.

Laboratory-confirmed influenza cases among students aged 5–19 years, schools reporting excess levels of absenteeism, and schools reporting excess visits to the school nurse for ILI per week: Pennsylvania, September 2009–June 2010.

Note. ILI = influenza-like illness.

The correlation between absenteeism and laboratory reporting data was nonsignificant (Spearman correlation coefficient = 0.10; P = .56). By contrast, trends in ILI correlated well with laboratory reporting data (Spearman correlation coefficient = 0.67; P ≤ .001). Results remained the same after restricting state data to confirmed influenza cases among those aged 5 to 19 years (absenteeism: Spearman correlation coefficient = 0.09; P = .58; ILI: Spearman correlation coefficient = 0.67; P ≤ .001). The correlation between absenteeism and ILI visits to the school nurse was nonsignificant (Spearman correlation coefficient = 0.28; P = .12).

DISCUSSION

The Pennsylvania Influenza Sentinel School Monitoring System fulfilled its primary objective by collecting, organizing, and distributing information on levels of absenteeism and ILI in schools. The surveillance system was inexpensive and relatively easy to establish, and sentinel schools generally represented schools in the state. Strengths of the system included its simplicity, low maintenance, regular data collection, and accessibility to participating schools. Schools across the state expressed strong interest, and weekly participation rates were high throughout the entire school year. In fact, the average weekly response rate of 79% indicates that schools were continuing to participate and complete the weekly survey even after flu had largely subsided in the community and media attention surrounding pH1N1 had waned.

The high levels of enrollment and participation seen during the 2009–2010 school year may simply reflect the intense interest in influenza in educational settings resulting from pH1N1. Similar findings and interest might not extend to a more typical influenza season. Still, at the end of the school year when school correspondents were asked whether they would be interested in participating in the system next year, 80% responded that they were very or somewhat interested in continuing. This finding suggests that schools recognize the importance of influenza surveillance and the effect influenza has on their educational activities. An indirect benefit of this system was to help strengthen the working relationship between Pennsylvania's educational and public health systems, which will hopefully continue into subsequent influenza seasons.

A comparison of the epidemic curves of absenteeism and ILI data indicated a close similarity between influenza activity detected by the school-based system and statewide influenza laboratory data. Although the peak week of school absenteeism coincided with the peak week of community influenza cases, school-based ILI peaked 1 week before. If school-aged children develop influenza before the general population²⁰ and thus seek care via the school nurse, our results are as expected. Additionally, our finding that the peak week of elevated absenteeism coincided with the peak of statewide laboratory-confirmed influenza cases is consistent with the findings of other studies that compare school absenteeism data and other community influenza surveillance tools.

During the 1975–1976 winter, Peterson et al. piloted a school-based influenza surveillance system in all 1900 public schools in Minnesota and compared the system with other state surveillance data.¹⁴ The purpose of the new system was to discover whether school-based surveillance data could be used to determine when and where cases of influenza were occurring.¹⁴ The school-based surveillance system detected influenza at the same time that specimens submitted by physicians to a state laboratory grew influenza virus.¹⁴ The authors concluded that the school-based influenza surveillance system was superior to other recently used statewide indirect influenza surveillance systems and could be implemented to enhance laboratory surveillance to obtain reliable information on the presence and extent of influenza in the community.

Lenaway and Ambler evaluated a school-based surveillance system as an effective measure of epidemic influenza in the community using 5 influenza seasons worth of surveillance data (1988–1989 through 1992–1993) from the Boulder County Health Department in Colorado.²⁰ They compared the school-based system, which covered 30 elementary schools, 9 middle schools, and 5 high schools, with a preexisting sentinel communicable disease surveillance system. In general, the school-based system closely followed the rise, peak, and fall of epidemic illness as measured by the preexisting system. As reported by the authors, 3 of the 5 epidemic peaks matched exactly, whereas the other 2 peaks showed the school surveillance system peaking 1 week before the sentinel system.

In a study conducted using data from 2005 to 2007 from 6 primary schools in London, Schmidt et al. reported that peaks of school absenteeism, measured using electronic absence records, coincided with peaks of influenza A and B, measured by laboratory reports.²¹ Similar to our findings, Schmidt et al. found that several high peaks of absenteeism were not associated with influenza.

Using school absence records as a surveillance tool may have the advantage that absences are already documented at the school level on a daily basis. However, because most schools are not able to break down absenteeism by cause, schools report absenteeism for any reason. Therefore, absenteeism is only a crude approximation of illness. In fact, some of the absenteeism spikes detected with the Pennsylvania system were not flu related but instead related to events such as fairs, holidays, hunting season, vacations, or snow. Further, because respiratory infections resulting from other pathogens are also common among school-aged children, school absence data may have a low specificity for influenza, similar to other public health syndromic surveillance tools that record the number of visits because of respiratory symptoms. Still, if a child does not seek medical care outside the school, peaks of absenteeism in individual schools may actually be because of influenza outbreaks unnoticed by other surveillance systems.¹⁸

Although absence data coded for illness could be a more sensitive tool for disease surveillance,^17,20 having schools collect illness data on all students absent from schools is probably unrealistic, especially in larger schools, which may have high daily counts of absenteeism year round. Finally, school surveillance systems are limited because data cannot be collected on weekends or during school breaks.

In this study, we evaluated a sentinel school influenza surveillance system. As our results suggest, an online tool is a relatively simple, acceptable, and reliable device for tracking absenteeism and ILI in schools. Overall participation and completeness rates were high; however, the system was established and operated during a period of high interest in pH1N1, and therefore, school-based reporting may not be as effective during a normal influenza season. Further analyses may be necessary to determine the comparative value of such a system and other influenza surveillance systems. Data from a larger and possibly more geographically representative cohort of schools and multiple influenza seasons would help establish baseline activity and would allow more extensive analysis and comparisons. Such data could additionally be used to identify any larger-than-normal increases in ILI in school-aged children during a particular time of year and could serve as an early warning system of increased flu activity at the local level. Recognizing such patterns in schools would allow public health officials to implement influenza control measures in schools and communities, which would help reduce the individual and societal burden of the disease.