Influenza and School-Based Influenza-Like Illness Surveillance: A Pilot Initiative in MARYLAND

On June 11, 2009, the World Health Organization (WHO) declared a pandemic in response to the 2009 influenza A (H1N1) virus. Enormous pressure was placed on public health infrastructure worldwide to minimize the adverse potential of the pandemic, with particular emphasis on surveillance measures, including that of schoolchildren aged 5–18 years.^1,2 Further, recognition of the disproportionate amount of illness and hospitalization among younger people caused by the 2009 H1N1 virus intensified the relevance of influenza surveillance in the school-age population.^3,4

School-based influenza surveillance is based on the premise that monitoring the school-age population provides enhanced early detection of influenza epidemics.^5–7 Epidemiologic evidence suggests that influenza illness occurs first among school-age children and that, once infected, this population efficiently transmits the virus among family members and then to the general community.^8–13 Previous studies have validated the use of school-based influenza surveillance as a marker of general community influenza-like illness (ILI).^6,7,9,13,14

Numerous examples of existing school-based influenza surveillance systems^15–22 can be identified from an Internet search of both international and U.S. health departments, but wide heterogeneity and absence of established data parameters limit applicability and standardized implementation. Varying methods of school-based influenza surveillance include detection based on total absenteeism, disease-specific absenteeism, school health room visits, and/or a combination of these. Of the listed methods, only total absenteeism has been demonstrated as a validated indicator of possible influenza activity/epidemic.^7,18,20 A wide range of school absenteeism (8%–47%)^7,9,10,23 is reported during influenza epidemics, but an absenteeism rate of 10% is traditionally used (and accepted) as a valid indicator.^24–27 Additional surveillance information is often used, including determination of absenteeism based on illness type and, more specifically, ILI. However, no set guideline exists for interpretation of these data.

The Anne Arundel County Department of Health (AACoDOH) in Maryland developed a School-Based ILI Surveillance System (SILISS) to identify 2009 influenza illness trends in the schools through monitoring absenteeism rates and the number of school health room visits due to ILI symptoms. We evaluated the ease of administering the surveillance system and determined its accuracy as a tool for detecting community influenza activity.

METHODS

Definitions

The following definition for ILI—fever >100°F and cough and/or sore throat—was used based on accepted Centers for Disease Control and Prevention (CDC) and WHO 2009 H1N1 case definitions.^28,29 No diagnostic tests were performed.

SILISS

SILISS was implemented at the start of the Anne Arundel County school year in August 2009. School nurses from the county's 80 public elementary schools (children aged 5–12 years) were mandated to collect the following data on a daily basis: (1) percent total absenteeism, (2) total number of health room visits, and (3) number of health room visits secondary to ILI symptoms. The data were submitted by the school nurses using the Web-based software SurveyMonkey™³⁰ and reviewed daily by AACoDOH.

Surveillance of only public elementary schools was chosen because (1) infection precedes the general community in this population, (2) elementary students may provide more reliable information regarding absenteeism (vs. truancy), and (3) students of this age group may be more likely to seek medical attention via school health room for symptom verification.^7–14

To determine the accuracy of SILISS as a tool for detecting community influenza activity, we compared weekly data with the Electronic Surveillance System for the Early Notification of Community-Based Epidemics (ESSENCE) for Maryland. ESSENCE is a Web-based syndromic surveillance system based on hospital emergency department visits designed for the early detection of disease outbreaks, suspicious patterns of illness, and public health emergencies.^31–33 ESSENCE was reported to have effectively detected and tracked changing ILI trends during pandemic H1N1 2009. For ESSENCE, a person with ILI is defined as someone who presents with a chief complaint of either fever and cough or fever and sore throat or influenza/flu. ESSENCE is based on International Classification of Diseases, Clinical Modification, 9th Revision diagnostic codes; ILI cases do not necessarily represent laboratory-confirmed influenza cases. ESSENCE methods have been published elsewhere.³⁴

SILISS user feedback

A short feedback questionnaire was sent to all school health nursing staff through SurveyMonkey 15 weeks after SILISS implementation. The purpose of the questionnaire was to determine time required for SILISS and ease of administration.

Statistical analysis

SILISS' weekly total ILI visits, percent ILI health room visits, and mean absenteeism for the county were compiled and compared with weekly data from ESSENCE. Spearman rank correlation (two-tailed) was calculated for all measured variables in both surveillance systems for the entire school year (week ending August 29, 2009, to June 19, 2010), and for weeks when influenza activity in Maryland was “widespread” (defined by CDC flu activity reports and methods as “outbreaks of influenza or increases in ILI cases and recent laboratory-confirmed influenza [viral culture, direct/indirect fluorescent antibody assay, reverse transcription-polymerase chain reaction, or a rapid influenza diagnostic test] in at least half the regions of the state with recent laboratory evidence of influenza in the state”) for the week ending September 5 to November 28, 2009.³⁵ Correlation was computed only for weeks when data were available for SILISS (data were unavailable for the weeks ending November 28 and December 26, 2009; and for the weeks ending January 2, February 13, and April 3, 2010). Statistical analysis was performed using SPSS® version 16.0.³⁶

RESULTS

SILISS

After one school year (week ending August 29, 2009, until June 19, 2010) of data collection, the weekly response rate from schools ranged from 88.7% to 96.8%. The average response rate from schools during the school year was 94.1% (95% confidence interval [CI] 93.5, 94.7) of schools reporting per week. The average weekly ILI visit proportion across all schools for the year was 0.78% (95% CI 0.45, 1.10). The average weekly absentee rate across all schools for the year was 4.42% (95% CI 4.03, 4.80). Weekly mean absenteeism peaked at 7.76% for the week ending October 24, 2009. This peak corresponded to a peak value for both weekly mean percentage of ILI health visits (4.2%) and total number of ILI visits (n=388) (Figures 1 and 2).

Figure 1.

SILISS mean absenteeism vs. SILISS number of ILI visits by week in Anne Arundel County, Maryland: 2009–2010^a

^aAnne Arundel County, Maryland, SILISS data comparing mean absenteeism (bars) with number of ILIs (dashed line) by week end date: August 29, 2009, to June 19, 2010. The overall correlation is low (ρ=0.088); however, mean absenteeism correlates strongly (ρ=0.832, p=0.01) with number of ILIs during the first part of the school year, when Maryland state influenza activity was defined as “widespread.” Source: Centers for Disease Control and Prevention (US). Flu activity and surveillance: reports and surveillance methods in the United States: week ending September 5 to November 28, 2009 [cited 2011 Mar 1]. Available from: URL: http://www.cdc.gov/flu/weekly/fluactivitysurv.htm

SILISS = School ILI Surveillance System

ILI = influenza-like illness

Figure 2.

Percent of ILI visits in Maryland: ESSENCE vs. SILISS, August 29, 2009, to June 19, 2010^a

^aMaryland ESSENCE is represented by bars and Anne Arundel County, Maryland, is represented by the dashed line. SILISS percent ILI visits correlate strongly with ESSENCE number of ILI visits throughout the time period (ρ=0.777, p=0.01).

ESSENCE = Electronic Surveillance System for the Early Notification of Community-Based Epidemics

SILISS = School ILI Surveillance System

ILI = influenza-like illness

SILISS intra-variable correlation

To analyze SILISSS intra-variable correlation, data were used from the week ending August 29, 2009, until June 19, 2010. Weekly SILIS data were unavailable for weeks when school was closed (November 22–28 [Thanksgiving break]; December 20–26 and December 27–January 2 [Christmas break]; February 6–13 [snowstorm school cancellation]; and March 30–April 3 [spring break]). The SILISS intra-variable correlation for the entire school year (n=37) was the following: number ILI visits vs. percent ILI visits (ρ=0.956, =0.01) and vs. mean absenteeism (ρ=0.088) (Figure 1); and percent ILI visits vs. mean absenteeism (ρ=0.150). When Maryland influenza activity was widespread, the intra-variable correlation was the following: number ILI visits vs. percent ILI visits (ρ=0.979, p=0.01) and vs. mean absenteeism (ρ=0.818, p=0.01); and percent ILI visits vs. mean absenteeism (ρ=0.832, p=0.01).

SILISS vs. ESSENCE

To compare SILISS variables with ESSENCE variables, weekly data were used from the week ending August 29, 2009, until June 19, 2010 (n=37). The highest level of correlation found between ESSENCE and SILISS was ρ=0.805 (p=0.01; percent ILI visits) (Figure 2). The lowest level of correlation was ρ=0.234 (p=0.16; ESSENCE percent ILI visits vs. SILISS mean absenteeism).

When influenza activity in Maryland was widespread (n=12), the highest levels of correlation between SILISS and ESSENCE data were found for SILISS mean absenteeism vs. ESSENCE percent ILI visits (ρ=0.979, p=0.05) and vs. ESSENCE number of ILI visits (ρ=0.965, p=0.01).

SILISS user feedback

Forty-four surveys were completed (100% response rate). Forty-four nurses reported spending an average of 11.9 minutes per day (range: 1–60 minutes) completing the survey, including collecting and submitting data. Forty-three nurses also spent an average of 13.3 minutes per day (range: 5–60 minutes) of health room assistant time. Respondents reported that SILISS was useful for both the health room staff and health department, was not burdensome, was executed with clear instructions, and had a website that was easy to use (data not shown).

DISCUSSION

SILISS was a strong predictor of local influenza activity and demonstrated ease of administration. For the entire school year, comparison of SILISS data with Maryland ESSENCE data demonstrated a high level of correlation for ESSENCE ILI data vs. SILISS ILI data, but not for SILISS mean absenteeism. When influenza activity was at its highest, all of the variables from SILISS correlated strongly with the variables measured by ESSENCE; however, SILISS mean absenteeism outperformed both SILISS ILI percent and number of visits.

Peterson and colleagues validated school-based influenza surveillance in the mid-1970s by comparing total absenteeism with viral cultures, demonstrating that sustained excess school absenteeism (defined as levels that are twice as high as normal) was an adequate indicator of the presence and extent of influenza illness.⁶ On this premise, Lenaway and Ambler described a school-based influenza surveillance system based on total absenteeism (using 7.5% as a reporting threshold), demonstrating a general correlation with sentinel measures (i.e., sentinel physician-reported number of ILI visits).⁷ Although a number of health departments have incorporated additional information into their school-based influenza surveillance systems, including ILI-specific health room visits, there is no validation of these data in the literature. Our data suggest that mean absenteeism is an excellent measure of community influenza activity, but only when influenza activity is very high (e.g., during an epidemic). For example, it was noted through SILISS that mean absenteeism increased on school days proximal to long weekends or holidays. School health room ILI data appear to provide an accurate measure of community influenza irrespective of influenza activity. Given the specific nature of the ILI definition (in contrast with absenteeism) and the community measure used in this comparison (emergency department ILI visits), this finding is intuitive.

School-based influenza surveillance also provides important data for public health analysis and decision-making, particularly in situations of uncertainty. This is evident in the example of school closure, which is often considered as a nonpharmaceutical intervention during influenza epidemics. Current literature suggests that the evidence for the efficacy of school closure for community mitigation purposes is contradictory,^37–42 and that it also poses serious economic repercussions.^43,44

In summary, the AACoDOH school-based ILI surveillance initiative demonstrates a resource-practical and cost-efficient model using both absenteeism and ILI data that correlate highly with external measures of community influenza illness.

Limitations

Limitations of the model included that it required both school nurse and school administrator cooperation. Given the publicity of the 2009 H1N1 pandemic, increased awareness may have influenced children and parent behaviors with regard to health room visits and keeping children home, as well as staff willingness to submit required information. In addition, ILI data were collected only on school days and did not include symptomatic schoolchildren identified outside of the school setting, including those who stayed home with ILI symptoms; however, these children were captured within the category of absenteeism. Because the model captured ILI symptoms without laboratory confirmation, the type of influenza strain implicated in an epidemic must be inferred from community data. Lastly, the ILI definition used by ESSENCE is not identical to that used by SILISS, although it is very similar; therefore, direct comparison poses limitations.

CONCLUSIONS

Future considerations and challenges include determining when to initiate and terminate school-based influenza surveillance (year-long vs. at the start/end of flu season). Although feasibility is uncertain, the surveillance system could also be extended to include identification of additional symptom groups (e.g., nausea/vomiting/diarrhea in the setting of gastroenteritis). AACoDOH continues to conduct school-based influenza surveillance, and ongoing use of SILISS will enhance our interpretation and application of influenza school surveillance methods and data. No other Maryland counties are currently participating in SILISS, although collaboration would likely contribute to the richness of surveillance data. Given the relatively modest requirement of additional resources, including health room staff and county epidemiologist participation and subscription to SurveyMonkey, long-term sustainability of SILISS does not appear problematic.

The use of school-based influenza surveillance is an important tool for monitoring illness trends in the community. In addition to vaccination, influenza surveillance represents an important public health intervention that may potentially decrease influenza morbidity and mortality and positively impact lives. The AACoDOH ILI school surveillance initiative demonstrates ease of administration and accuracy as a tool for detecting community influenza activity, and may be a practical model for a school-based influenza surveillance system. Based on minimal resources and high-impact measures, the surveillance system is readily applicable to adaptation in other counties and public health departments.