Outbreak of influenza A in a boarding school in South Africa, 2016

Introduction We investigated an outbreak of influenza-like illness (ILI) at a boarding school in Eastern Cape Province, South Africa. We aimed to confirm the etiological agent, estimate attack rates and identify risk factors for illness. Methods We conducted a retrospective cohort study including senior school boarders (n=308). Students with ILI (cough and fever) were identified through school medical records. We also conducted a questionnaire-based cross-sectional study among senior students including boarders (n=107) and day students (n=45). We collected respiratory specimens for respiratory pathogen testing by real-time polymerase chain reaction from a subset of symptomatic students. We calculated attack rates of medically attended ILI (medILI) and identified factors associated with medILI using logistic regression. We calculated seasonal influenza vaccine effectiveness (VE) against medILI. Results Influenza A (H3N2) virus was detected in 61% (23/38) of specimens. Attack rate for medILI was 13% among boarders (39/308) in the cohort study and 20% in both day students (9/45) and boarders (21/107) in the cross-sectional study. Playing squash was associated with medILI (aOR 5.35, 95% confidence interval [95% CI]: 1.68-17.07). Of the boarders, 19% (57/308) were vaccinated before the outbreak. The adjusted VE against medILI was 18% (aOR 0.82, 95% CI 0.38-1.78). The outbreak led to cancellation of several events and the need for academic remedial sessions. Conclusion We confirmed an influenza A (H3N2) virus outbreak with a high attack rate. The outbreak affected academic and sports activities. Participation in sports and social gatherings while experiencing ILI should be discouraged to reduce viral transmission and impact on school activities.


Introduction
Influenza-like illness (ILI) and influenza outbreaks commonly occur in schools [1,2], with areas like United Kingdom reporting as many as 658 school respiratory illness outbreaks in one season, with attack rates ranging from 14-45% [3]. There are few data describing outbreaks of influenza or ILI in South Africa. To our knowledge, only three outbreaks have been described in peer-reviewed journals. This includes an outbreak of influenza A (H3N2) virus in a police residential college in Pretoria in 2003 with an attack rate of 20-47% [4], an outbreak affecting students from several schools in Bophuthatswana in 1978 where the strain was not characterized [5], and an outbreak of influenza A (H2N2) virus in gold miners near Johannesburg in 1958 with an attack rate of 15% [6].
On the 14 th of July 2016, the resident doctor at a boarding school in the Eastern Cape Province, South Africa, contacted the National Institute for Communicable Diseases (NICD) of the National Health Laboratory Service (NHLS) and reported a large number of students presenting with ILI since the 9 th of July, more cases than he had experienced for any influenza season in his time working at the school. Symptoms among students presenting to the sanatorium included fever, cough and generalised body pains. Due to the high number of cases, the disruption of school activities and the unique opportunity to describe an influenza outbreak in a closed population, we conducted an investigation during the 26-29 th of July 2016, aimed to confirm the etiological agent, estimate attack rates among students and identify risk factors for illness.

Methods
The investigation took place at a co-educational private school in the Eastern Cape province. Besides South Africans, the nationalities of the students varied and included more than 20 countries.
Interviews were held with the school sanatorium doctor and nurse, the senior school headmaster, head of academics and head of sport to gather information on the setting and the impact of the outbreak on the school. The outbreak was confined mainly to senior school students and therefore the focus of this investigation was on the senior school students (grades 8-12 and bridge year students). Cases were identified from medical record review and questionnaires administered to staff and students.
indicate the baseline level of medILI before the outbreak and the medILI that occurred until the end of the outbreak investigation (25 June -27 July 2016). Attack rates for self-reported medILI (selfreported cough and fever and seeking medical care from a nurse or clinician) were calculated from the start of the school term (3 July 2016) to when the investigation was performed (27 July 2016) by dividing the number of individuals reporting each condition by the specified population. A multivariable logistic regression analysis compared the proportion reporting medILI in each of the subgroups, assessing all variables with a p-value <0.2 in the univariate analysis.
Variables with a p-value <0.05 were included in the final analysis.
Variables considered in the model were age, grade, gender, nationality, boarding/day student, boarding house, having cubicles in the boarding room, participation in sports and other activities and having underling medical conditions.

Data processing and analysis
The web-based questionnaire used in the cross-sectional study was done on the Google forms platform and the output exported to Microsoft Excel 2010 (Microsoft Corporation, Redmond, WA, USA).
Data from medical records and the student register collected for the cohort study were captured on the same database. Data were analysed in Stata version 14.0 (StataCorp LLC, College Station, TX, USA).

Laboratory investigations
Respiratory specimens were collected by the resident doctor and nurse from a convenient subset of boarding and day senior students clinically diagnosed with influenza from 13 to 29 July 2016. The swabs were transported in universal transport medium to the Centre for Respiratory Diseases and Meningitis, NICD, Johannesburg. To determine confidence for support of the branches, 500 bootstrap replicates were employed.

Ethical considerations
Prior to the investigation, guidance on managing parental consent for this outbreak investigation was sought from the chairperson of the Informed consent and assent were obtained from parents and students to complete the questionnaires in the cross-sectional study.
All information collected was linked to an anonymised study identification number. The raw data were only accessed by the researcher doing the analysis, which was done on a passwordprotected computer.

Setting
The school employed a nurse and offered primary health care services for boarders and day students. A school doctor consulted when required. The school annually offered the Southern Hemisphere influenza vaccine to consenting students as soon as the vaccine was available in South Africa, which was billed to the student's medical aid.
The senior school hosted grade 8 to 12 as well as bridge year students. The bridging year offered students that finished grade 12 an additional year of school before starting tertiary education. There were 427 senior students of which 308 (72%) were boarders, aged It was recommended that symptomatic students not participate in any sporting events, and from 11-24 July there were multiple students that did not attend sports practice due to illness. Several sports events and practices were cancelled. The sports program recommenced on 25 July. Except for sports practices and matches, many other meetings and cultural events had to be rescheduled or cancelled.
Although academic classes were not cancelled during the outbreak, remedial days had to be held after the outbreak for students to catch up on their work.

Cohort study
We identified 39 medILI cases occurring between 5-27 July ( Figure 1). The attack rate of medILI (Table 1) amongst senior boarders was 13% (39/308). There were no significant differences in attack rates by age group, gender or residence. Of 308 boarders, 19% (n=57) were vaccinated prior to the outbreak. Additional vaccinations were not provided during the outbreak. After adjusting for gender, the vaccine effectiveness against medILI was 18% (aOR 0.82, 95%CI 0.38-1.78). Only two cases resulted in complications. There was one hospitalization reported when a student fainted at an event and one case of suspected pneumonia; both students recovered.
Of 152 students who completed the questionnaire, 116 days of school were missed due to absenteeism caused by ILI. Among students, 33% (n=50) missed school, 29% (n=44) missed study time, 45% (n=68) missed sport and 14% (n=21) missed other events or time with their friends. The mean time missed from school was 3.4 days per ill student (range 1-11 days). Thirty percent (n=45) of students that completed the questionnaire reportedly received the 2016 vaccine before the outbreak. The reasons provided for not receiving the vaccine are shown in Table 3.

Laboratory investigations
Of the first four nasal specimens collected before the outbreak

Discussion
The outbreak occurred from 9-28 July based on the epidemiological curve and the last samples that tested positive for influenza A (H3N2) virus. Laboratory testing identified influenza A (H3N2) virus as the etiological agent with an overall high attack rate of medILI (13% in cohort and 20% in cross-sectional study), with participating in the sport squash being the strongest risk factor associated with medILI.
Multiple sporting and cultural events were cancelled during the outbreak and academic remedial sessions were rescheduled afterwards. We had a low response rate (36%) in the cross-sectional study. This was influenced by lack of interest among some students and by lack of parental consent for others. This may have resulted in participation bias where students that were not affected by ILI might have been less likely to participate, inflating the attack rate of ILI in the crosssectional study. The attack rate for medILI in the cohort study may therefore be a more accurate representation of the outbreak.

Conclusion
We confirmed an outbreak of influenza A (H3N2) virus in a boarding school possibly initiated by a sentinel case after the return of students from a half-term break. The outbreak impacted academic, cultural and sports activities at the school. Playing squash was associated with medically attended influenza-like illness. Participation in sports activities and attending social gatherings in confined spaces should be discouraged when showing ILI symptoms to prevent infection.

Seasonal influenza vaccination should be encouraged in boarding schools to prevent influenza and reduce academic and sports program disruption.
What is known about this topic  Influenza-like illness (ILI) and influenza outbreaks commonly occur in schools and can disrupt educational activities;  There are few data describing outbreaks of influenza or ILI in South Africa.

What this study adds
 Outbreaks of influenza in schools may have high attack rates and disrupt educational and extracurricular activities;  Participation in sport activities while infected with influenza A viruses may lead to further spread;  The most common reason students were not vaccinated was the perception that the vaccine doesn't work.

Competing interests
The authors declare no competing interests.

Acknowledgments
We express our appreciation to the boarding school, their medical and academic staff that assisted in the collection of samples, collating informed consent forms, facilitating questionnaire completion and for their hospitality. We thank the parents and students who participated in the outbreak investigation, the NICD, the National Department of Health and the Eastern Cape Department of Health. We acknowledge the authors, originating and submitting laboratories (WHO CCs Australia, Atlanta and London) of the sequences from the GISAID's EpiFlu TM Database used in this study. Table 1: description of boarders, medically attended influenza-like illness cases and analysis of potential risk factors for medILI