Established in 1984, the Viral Watch (VW) programme is an active, prospective surveillance programme run by volunteer medical practitioners to monitor outpatient influenza. ILI surveillance is performed at 205 sentinel sites in private and public clinics in all 9 of South Africa’s provinces (Fig. 1)[7]. A case of ILI is defined as: an acute respiratory infection with a measured fever (≥38°C), AND cough, AND onset within the past 7 days. Sites are requested to enrol patients with onset of symptoms within 72 hours of presentation, and to enrol no more than 5 patients per week. Nose and throat swabs are collected from consenting patients and most samples (>80%) are sent directly to the National Institute for Communicable Diseases (NICD) for influenza testing by real-time reverse transcription polymerase chain reaction (rRT-PCR). Specimens from selected provinces are first tested for influenza by shell viral culture at local laboratories, after which positive samples are shipped to NICD for influenza subtyping by rRT-PCR. NICD is a current WHO National Influenza Centre[13] and VW is fully funded by NICD.

SARI surveillance is performed by teams of dedicated personnel at 6 sentinel sites in public hospitals in 4 provinces (Fig. 1)[9]. Established in February 2009, this active, prospective, hospital-based surveillance programme collects detailed epidemiological and virological influenza surveillance data. In persons ≥5 years of age, a case of SARI is defined as: an acute respiratory infection with a history of fever or measured fever (≥38°C), AND cough, AND onset within the past 7 days, AND requiring hospitalization. In children 3–59 months of age, a case of SARI is defined as physician-diagnosed lower respiratory tract infection (LRTI) with onset in the past 7 days and requiring hospitalization, while for infants aged 2 days to <3 months the definition includes suspected sepsis or LRTI irrespective of signs and symptoms with onset in the past 7 days. All patients admitted on week days are eligible for enrolment, except adult patients at Chris Hani Baragwanath Hospital, where systematic sampling is used on 2 of every 5 week days (selected days rotate) due to high patient load and resource constraints. Specimens collected include nasopharyngeal aspirates from patients <5 years of age and nasopharyngeal and throat swabs from patients ≥5 years of age. Respiratory specimens are transported on ice to NICD within 72 hours of collection and tested for 10 respiratory viruses, including influenza A and B viruses, by multiplex rRT-PCR[14]. Influenza virus subtypes are also identified by rRT-PCR. The SARI programme is coordinated by NICD with financial and technical assistance provided by the U.S. Centers for Disease Control and Prevention (CDC).

For each system we describe the weekly proportion of samples testing positive for influenza, the week of peak detection, and season start and end dates (seasonal threshold defined as influenza detection rates ≥10% and <10% for two consecutive weeks, respectively) from 2009 to 2012. The number of respiratory specimens tested by week and the proportion testing positive for influenza and influenza types and subtypes was also assessed. Only samples tested by rRT-PCR were included in our analysis.

CDC guidelines[15] suggest the usefulness of a surveillance system is dependent on the actions that can be taken as a result of data collection and analysis; specifically, whether the system is able to: (1) guide disease prevention and control activities through the timely detection of adverse health-events, (2) estimate the magnitude of morbidity and mortality and associated risk factors, (3) detect trends that signal changes in incidence, including epidemics, (4) permit assessment of prevention and control measures, (5) lead to improved health and social policy or clinical practice, and (6) stimulate research to inform prevention and control measures. Using criteria established by the CDC[15], we assessed nine surveillance system attributes that can affect usefulness, including quantitative analyses of data quality, timeliness, sensitivity, and positive predictive value, and qualitative descriptive analyses of representativeness, simplicity, flexibility, acceptability, and stability.

Data quality was assessed by measuring the completeness of patient interview forms, form transmission, respiratory specimen collection and testing, and by checking patient sign and symptom records and primary diagnosis to determine whether surveillance case definitions had been adhered to properly. Data collected by each system was also compared against minimum data collection standards for ILI and SARI surveillance (S1 Table)[12].

Timeliness was assessed by measuring the duration of time for the collection, transfer, and processing of forms and specimens, and for the availability of laboratory results.

Sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) were obtained for one SARI and two ILI case definitions for predicting influenza infection (Table 2). The clinical case definitions selected for this comparison included those used by each system for screening and enrolment, and an older ILI case definition used by VW prior to March 2012.

The SARI system collects detailed epidemiological patient data; when compared against minimum data collection standards for ILI and SARI surveillance (S1 Table), we found patient interview forms contained all data elements recommended by WHO[12]. By contrast, VW collects minimal epidemiological patient data. When compared against minimum data collection standards for ILI surveillance, we found VW does not collect data on the use of antivirals for current illness, or pre-existing conditions such as chronic neurological or neuromuscular disease and haematological disorders.

Data quality was a major strength of the SARI system, with 40/43 key data elements on patient interview forms having completeness measures above 90% (n = 15,189 forms). Completeness of other SARI forms (i.e., lab slips, Hospital Results Forms, Final Outcome Forms) were similarly high, as were measures of form transmission (98% of enrolled patients had all surveillance forms transferred from sentinel sites to NICD), data capturing (>99.9% of all forms received at NICD were captured by data entry clerks), and specimen collection and testing (98% of SARI patients had respiratory specimens collected and tested for influenza). Using available data on primary diagnosis, adherence to the SARI case definition (S3 Table) in persons aged 2 days to <3 months was at least 89% (725/812) and adherence in persons aged 3–59 months was at least 88% (2,192/2,486). We were not able to evaluate adherence to the SARI case definition in persons aged 5 years or older because the appropriate data were not routinely collected during the study timeframe.

Data quality was a relative weakness of VW. Over the four years analysed, 62% (n = 8,556) of patient interview forms had complete data available for all data elements, though completeness improved with new forms introduced in March 2012 (83% between April–June 2012, n = 698). Measures of specimen collection and testing were very high; 100% of ILI patients had respiratory specimens collected and tested for influenza, and 99% (n = 2,667) of influenza A results were subtyped. Adherence to the ILI case definition in use from 2009 to March 2012 was 82% (n = 5,447), but decreased to 65% (n = 592) with the introduction of a new ILI case definition in March 2012.

Timeliness was a relative strength of the SARI programme; 93% (12,835/13,813) of patients had respiratory specimens collected within 7 days of symptom onset (median: 3 days, IQR: 2–5 days), and 95% (13,291/14,020) of patients had respiratory specimens collected within 1 day of hospital admission. The time interval from specimen collection to capturing of results into analytic datasets at NICD was a median of 8 days (IQR: 7–11, n = 7,559). There were weaknesses, however; only 51% (7,770/15,243) of case investigation forms were received at NICD within the programme target of 14 days after patient interviews (median: 14 days, IQR: 7–28).

Timeliness was also a relative strength of VW; 98% (n = 5,659) of ILI cases had respiratory specimens collected within 7 days of symptom onset (median: 1 day), and the median time from specimen collection to receipt at NICD is 1 day (n = 7,744). Once at NICD, however, the median time to availability of laboratory results was 4 days (n = 612), which exceeded the programme target by 1 day. Although case investigation forms for both systems are currently paper-based, electronic data collection is being considered and may help to improve the timeliness of future data transfer and capturing.

The SARI programme’s SARI case definition in persons aged 2 days to <3 months was introduced in February 2009 (S3 Table); using incorrectly enrolled patients who did not meet the case definition as a comparator, we found the sensitivity of identifying laboratory-proven influenza was high (89.6%; 95% CI: 82.4–94.1) and specificity (10.8%; 95% CI: 8.7–13.3) and PPV were low (13.1%; 95% CI: 10.8–15.8). Similarly, in children aged 3–59 months, using enrolled patients who did not meet the case definition as a comparator, we found sensitivity was high (94.3%; 95% CI: 91.4–96.2) and specificity (12.9%; 95% CI: 11.5–14.4) and PPV were low (15.8%; 95% CI: 14.4–17.4). We were not able to evaluate sensitivity and PPV of SARI case definitions in persons aged 5 years or older because the appropriate data was not routinely collected during the study timeframe.

The VW’s current ILI case definition was introduced in March 2012 (S3 Table). Using incorrectly enrolled patients who did not meet the case definition as a comparator, this slightly improved sensitivity (89.7%; 95% CI: 88.3–90.9) and specificity (28.7%; 95% CI: 27.0–30.4) of identifying laboratory-proven influenza compared with the ILI case definition in use prior to March 2012 [sensitivity 88.7% (95% CI: 87.3–90.0), specificity 22.2% (95% CI: 20.7–23.8)] (Table 2). PPV estimates were somewhat higher for ILI case definitions than for SARI, ranging from 45.9% (95% CI: 44.4–47.4) for the ILI case definition in use before March 2012 to 48.3% (95% CI: 46.8–49.9) for the definition VW now uses. Although they were not included in this analysis, 18% (1,998/11,102) of VW specimens were first tested at local laboratories using less sensitive viral culture techniques. Since only positive samples are shipped to NICD for subtyping by rRT-PCR, it is likely that true positives are being missed, though the system reports case detections rates using samples tested at NICD only.

With public sentinel sites in just 4 out of 9 provinces, national representativeness was a relative weakness of the SARI programme. Though its six public sector sentinel sites were selected to represent a variety of geographic and demographic areas, the Cape provinces (i.e. Western Cape, Eastern Cape, and Northern Cape) are not represented. By comparison, representativeness was a relative strength of VW. Although VW sample collection diminished rapidly outside the regular influenza season, VW has excellent geographic and population coverage, with 205 sentinel sites servicing in all 9 provinces, including private sector general practitioners, primary care clinics, paediatric outpatient departments, and occupational health clinics.

Simplicity was relative weakness of the SARI programme; with an annual operating cost of approximately $500,000–800,000 USD, the system requires substantial investments in human resources and the training of dedicated personnel. Though the system collects extensive clinical and epidemiological data, it displays good flexibility, and has adapted to changing information needs by enhancing surveillance at selected sites to include patients with suspected or confirmed tuberculosis (TB) and patients with severe respiratory infection (respiratory symptoms >7 days).

Simplicity is a major strength of VW; participation of sentinel sites is voluntary, data and specimen collection materials are few, and minimal epidemiological data is collected. Laboratory tests account for the majority of the programme’s operating costs (approximately $100,000 USD per annum), while coordination and overhead costs are low (approximately $20,000 USD per annum) and can be scaled up or down as funds permit. VW also displays good flexibility, and has expanded from fewer than 20 sentinel sites in 2005 to 205 sites in 2012 with little growth in the programme’s overhead costs.

Indirect measures of acceptability include completeness and timeliness of data reporting; on each account, the SARI programme performs well. Specimens and forms are couriered to NICD to ensure transportation is timely and successful. Participation in the SARI programme is controlled through dedicated personnel and the high level of resources supporting the programme ensure stability. The willingness of VW sentinel sites to participate voluntarily suggests acceptability is also a relative strength of the programme, though many do not send samples regularly. In 2011, 21% (42/202) of sites sent fewer than 5 samples in the entire year. The number of participating sites also supports programme stability, though with few resources to support the system, VW specimens and forms are sent to NICD via private laboratories used by participating clinicians.