As described previously, the VWP is a national sentinel influenza surveillance programme consisting of approximately 180 outpatient practitioners, of whom 90% are general practitioners of all race groups in all socioeconomic areas; however, only 11 (6%) were based at clinics in the public sector.7 Of the nine provinces of South Africa, the largest percentage of participating VWP practitioners during 2010–2013 were in Gauteng (40–47% of practitioners), which has 23% of the population of South Africa, and the Western Cape (23–28% of practitioners), which has 11% of the population. Influenza-like illness (ILI) was defined as an acute respiratory illness with a measured temperature of ≥38°C or a history of fever, and cough, with onset within the past 7 days. Clinical, demographic and influenza vaccination data were collected from each patient at the time of nasopharyngeal swab collection.7 Data on race and socioeconomic group were not collected. Practitioners were supplied with VW specimen forms to record underlying medical conditions using a tickbox format. Details were completed from medical records where available, or by self-report. Practitioners were limited to submitting no more than five specimens per week throughout the year to avoid overburdening the laboratory, and the choice of patient was left to the individual practitioner's discretion. Specimens were tested using multiplex reverse transcription real-time polymerase chain reaction (rRT-PCR) assays for influenza A and B as well as other respiratory viruses. Influenza-A-positive specimens were further subtyped by rRT-PCR.8 Specimens from seven of the nine provinces were tested at the National Institute for Communicable Diseases (NICD), whereas specimens from the other two provinces were tested at central laboratories in those provinces using the same test and assay. All positive specimens were sent to the NICD for sequencing.

A test-negative case–control study was conducted among patients enrolled as part of the VWP to estimate the VE of the Southern Hemisphere TIV. Patients from all nine provinces presenting with ILI to participating centres of the VWP from 1 January 2010 to 31 December 2013 were eligible for inclusion. Patients who met the ILI case definition, had a known influenza vaccine history and were 6 months or older were included in the VE analysis. Vaccine history was recorded from medical records where available, and if not was self-reported. It was not recorded whether children under 9 years had received two doses. Patients were considered to be vaccinated if they had received influenza vaccine >14 days prior to the onset of illness. Patients who had received vaccine ≤14 days prior to onset of symptoms were excluded from the study. Patients in whom influenza was detected were considered cases and those who tested negative for influenza were unmatched controls.9–12 Patients who presented at a clinic at the international arrivals of the major airport into South Africa were excluded as it was assumed that these patients contracted influenza outside of South Africa.

The start of the influenza season was defined as the week when at least two consecutive weekly influenza detection rates of ≥10% in specimens received from the VWP, and the season was considered to have ended when the detection rate dropped below 10% for two consecutive weeks, or when the number of specimens received were <10 per week. Only specimens collected during the season were included in the VE analysis. Each influenza season was divided into three equal parts, that is early, mid and late season, to assess the possibility of waning immunity during the season.12,13 As there were too few patients in the extremes of age, three age groups were selected for analysis, that is children and teenagers (<20 years), young adults (20–44 years) and older adults (≥45 years).

The Southern Hemisphere influenza vaccine for 2010–2012 was unchanged and contained an A/California/7/2009 (H1N1)-like virus, an A/Perth/16/2009 (H3N2)-like virus, and a B/Brisbane/60/2008-like virus. For 2013, the vaccine contained an A/California/7/2009 (H1N1)-like virus, an A/Victoria/361/2011 (H3N2)-like virus, and a B/Wisconsin/1/2010-like virus.

Categorical variables were compared using the chi-squared test, and multivariable logistic regression was used for the VE estimates to adjust for age, underlying medical conditions and season period where possible as we were not powered to obtain stratum-specific estimates. VE was calculated as 1 odds ratio (OR) for laboratory-confirmed influenza in vaccinated and unvaccinated patients.

Over the study period, there were approximately 1300 patients eligible (according to the above-mentioned criteria) for VE calculations on average every year, with a mean annual influenza detection rate of about 50% and a mean average vaccine coverage of about 5% among controls. Given the available data, we estimated that we would be able to obtain statistically significant VE estimates of approximately 40% and above assuming 95% confidence intervals and 80% power.

To assess reasons for low vaccine uptake, a written survey on the knowledge, attitudes and practices regarding influenza vaccination was conducted among VWP practitioners during an annual Influenza Symposium in March 2013. In May 2013, the survey was sent out electronically requesting those who were unable to attend to complete the online survey. The survey consisted of 20 multichoice questions and one open-ended question. There were eight questions that assessed the practitioners' knowledge of influenza and influenza vaccine. The survey was anonymous with no demographic data required about the respondents.

Ethical approval for this programme and survey was obtained from the University of the Witwatersrand Human Research Ethics Committee (Medical). Individual informed consent was not required as the test was performed for diagnostic purposes as part of surveillance. The Centers for Disease Control and Prevention's Institutional Review Board relied on the ethical approval from the University of the Witwatersrand.

In total, 9420 patients with ILI were enrolled over the 4-year study period (Figure1). Of these, 5344 (56·7%) were included in the VE analysis. Of the 4076 excluded specimens, 1402 (34·4%) were collected out of season, 1340 (32·9%) had an unknown date of illness onset, 750 (18·4%) did not meet the ILI case definition, 276 (6·8%) were collected at the point of entry into South Africa, 162 (4·0%) were less than 6 month of age or age unknown, and 146 (3·6%) had unknown influenza vaccine status. Among included individuals, 2666 (49·9%) were classified as cases (influenza test-positive) and 2678 (50·1%) as controls (influenza test-negative).

During the 4-year period, the annual influenza season occurred between May and September. The dominant types or subtypes differed by year with influenza B accounting for 310/585 (53·0%) of total influenza-positive cases in 2010, influenza A(H1N1)pdm09 for 561/698 (80·4%) in 2011, influenza A(H3N2) for 360/586 (61·4%) in 2012 and influenza A(H1N1)pdm09 for 539/797 (68·0%) in 2013 (Table1).

Nearly half (2569, 48·1%) of specimens came from patients aged 20–44 years, 2564 (48·0%) of patients were male, and 2951 (55·2%) of patients attended VWP sites in the central plateau region with a temperate climate (Free State, Gauteng, Northern Cape and North West provinces). A total of 4573 (85·6%) specimens were collected during the early (first third) and mid (second third)part of the influenza season. The majority of specimens 4913 (91·9%) were collected within 3 days of onset of symptoms. Six hundred and seventy-seven (12·7%) patients had an underlying medical condition, such as chronic pulmonary and cardiac disease (320; 47·3%), HIV infection (96; 14·3%), other immunosuppression (82; 12·0%), metabolic disorders (121; 17·9%), pregnancy and post-parturition (55; 8·1%), and morbid obesity defined as a body mass index of ≥40 (3; <1%). Depending on year, there were differences in age, seasonality, region and interval between onset and sampling between cases and controls (Table2).

Annual influenza vaccine coverage in patients included in the VE analysis was low (a mean of 1·4% and 4·5% for cases and controls, respectively; Table3). Coverage was highest in those ≥45 years of age (2·9% and 7·0% for cases and controls, respectively) and lowest in those <20 years of age (1·2% and 2·1% for cases and controls, respectively; Table3). Over the 4-year period, 116/158 (73·4%) of patients received their vaccine before the onset of the season and an additional 36 (22·8%) received it before the peak of the season. Vaccinated cases had received influenza vaccine between 3 weeks and 6 months prior to onset of symptoms. Influenza vaccine was found to be effective in three of the 4 years (Table3). VE estimates adjusted for age, underlying medical conditions and seasonality for 2010–2013 were 54·2% (95% CI: 2·4–78·6%) for 2010, 57·1% (95% CI: 15·5–78·2%) for 2011, 38·4% (95% CI: −71·7–78·1%) for 2012 and 87·2% (95% CI: 67·2–95·0%) for 2013.

Of the 167 registered VWP practitioners in 2013, 14 could not be reached. The survey was completed by 96/153 (62·7%) of practitioners (11 at the Influenza Symposium and the remainder electronically). The majority of VWP members to whom the KAP survey was administered were general practitioners 154/167 (92·2%), nine (5·4%) were occupational health nurses or doctors, and 4 (2·4%) were paediatricians.

The vast majority (94/96, 97·9%) of respondents agreed that it was possible for patients to be hospitalised or die from influenza. More than 85% of the 96 respondents correctly identified pregnant women, persons older than 65 years, young children, persons with human immunodeficiency virus (HIV)/acquired immunodeficiency syndrome (AIDS), persons with asthma or other chronic lung conditions, persons with tuberculosis and persons with chronic medical conditions as groups more likely to have influenza-related complications such as pneumonia or bronchitis. However, 43 (44·7%) of respondents felt that healthcare workers were also at higher risk of complications as well. Sixty-four (68·1%) of 94 respondents thought the vaccine to be highly effective, 29/94 (30·1%) thought it to be of average effectiveness, and one respondent did not think that influenza vaccine was effective.

Of the 96 participants, 85 (88·5%) responded that South Africa has a national recommendation regarding seasonal influenza vaccine and 86/92 (93·5%) knew that vaccine needed to be given annually. Of those who knew the vaccine needed to be given annually, 53/86 (61·6%) specified a time. Of these respondents, 32/53 (60·4%) answered that the vaccine should be given between March and July; 21 (39·6%) correctly responded more specifically that the vaccine should be given during autumn months (March–May) or before the influenza season. In addition, 82/96 (85·4%) agreed that vaccinating healthcare workers protected patients from becoming infected with influenza. Of the 96 respondents, 87 (90·6%) had previously been vaccinated against influenza and had received the vaccine in 2012. Seventy-seven (80·2%) had ready access to the vaccine, and 15 (15·6%) were able to get it free of charge. The three main reasons given for being vaccinated were to protect themselves and their patients against influenza (95, 99·0%), to protect family members against influenza (95, 99·0%) and to avoid missing work due to illness (88, 91·7%).

Although 83 (86·5%) practitioners mostly or always recommended influenza vaccine to their patients, 94 (97·9%) always recommended it to patients at risk of complications. The obstacles reported by practitioners to vaccinating patients were mainly patient refusal (59, 61·5%) and patients not coming to the practice when vaccinations were due (54, 56·3%). VWP practitioners also reported that there were several misconceptions regarding influenza vaccine influencing patient's decisions: that the vaccine caused influenza, that the vaccine had severe side effects and that influenza was always a mild illness.