From February 2009, active, prospective, hospital-based surveillance (the Severe Acute Respiratory Illness (SARI) programme) was implemented in three of the nine provinces of South Africa (Chris Hani-Baragwanath Academic Hospital (CHBAH) in an urban area of Gauteng Province, Edendale Hospital in a peri-urban area of KwaZulu-Natal Province and Matikwana and Mapulaneng Hospitals in a rural area of Mpumalanga Province). In June 2010, an additional surveillance site was introduced at Klerksdorp and Tshepong Hospitals in a peri-urban area of the Northwest Province.

A case of SARI was defined as a hospitalised individual with symptom onset less than seven days prior to admission meeting an adapted World Health Organisation (WHO) case definition for SARI: (1) sudden onset of fever (>38°C) or reported fever, (2) cough or sore throat, and (3) shortness of breath, or difficulty breathing.[7]

All patients admitted during Monday through Friday were eligible, except for adult patients at CHBAH where enrolment occurred for two of every five working days (enrolment days varied systematically according to the intake days of the two participating wards) per week due to large patient numbers and limited resources. Daily numbers of patients admitted, numbers screened, numbers meeting study case definitions and numbers enrolled were collected in study logs. Study staff completed case report forms until discharge and collected nasopharyngeal (NP) and throat swabs as well as blood specimens for pneumococcal testing from consenting patients. Hospital and ICU admission and collection of specimens for CD4+ T-cell counts was performed at the discretion of the attending-physician. Underlying medical conditions were defined as documented presence of asthma, other chronic lung disease, chronic heart disease, liver disease, renal disease, diabetes mellitus, immunocompromising conditions (excluding HIV infection) or neurological disease.

NP and throat swabs were transported in a single viral transport medium tube at 4–8°C to the National Institute for Communicable Diseases (NICD) within 72 hours of collection. Respiratory specimens were tested by a multiplex real-time reverse-transcription polymerase chain reaction (PCR) assay for influenza A and B viruses, parainfluenza virus 1–3, respiratory syncytial virus (RSV), enterovirus, human metapneumovirus (hMPV), adenovirus and human rhinovirus.[8] Influenza positive specimens were subtyped using the US Centers for Disease Control and Prevention (CDC) real-time reverse-transcription PCR protocol for characterisation of influenza virus. Streptococcus pneumoniae was identified by quantitative real-time PCR detecting the lytA gene from whole blood specimens.[9] The focus of the surveillance programme was viral pathogens and pneumococcus, therefore patients were not systematically tested for tuberculosis or other respiratory pathogens.

HIV-infection status data was obtained based on testing undertaken as part of standard-of-care,[10] or through anonymised linked dried blood spot specimen testing by enzyme-linked immunosorbent assay (ELISA) in patients providing written informed consent. Results from anonymised testing were used preferentially if both standard-of-care and anonymised results were available. CD4+ T-cell counts were determined by flow cytometry.[11] Patients were categorised into two immunosuppression categories: (1) moderate immunosupression (CD4+ T-lymphocytes ≥200/mm³), or (2) severe immunosuppression (CD4+ T-lymphocytes <200/mm³).[12] Patients diagnosed by clinicians as HIV-infected on the current admission were referred for HIV management as part of routine care.

Calculation of incidence was conducted at one surveillance site (CHBAH) where population denominator data were available. This hospital is the only public hospital serving a community of about 1.8 million persons aged ≥5 years in 2012 amongst whom ~10% have private medical insurance.[13] The vast majority (>80%) of uninsured individuals and approximately 10% of medically-insured individuals seek care at public hospitals, consequently the majority of individuals requiring hospitalisation from this community are admitted to CHBAH. We estimated the total number of SARI hospitalisations from the number of enrolled individuals adjusting for non-enrollment in three of five adult wards and during weekends and refusal to participate using information from study logs. The total number of SARI hospitalizations at CHBAH was obtained using the following formula:

Where SARI_Totalij is the estimated total number of SARI hospitalization in year i (2009–2012) and age group j (5–14, 15–24, 25–44, 45–64 and ≥65 years of age); SARI_Enrolledij is the number of SARI cases enrolled in year i and age group j; 5/2 is the coefficient used to adjust for enrolment of patients in 2/5 adult wards; 7/5 is the coefficient used to adjust for non-enrolment over weekends; and X_ij is the proportion of all eligible cases that were enrolled in year i and age group j. The adjustment factor varied from 2.2 to 7.9 depending on the age-group and year of enrolment. We estimated incidence of SARI hospitalisations per 100,000 individuals by age groups and HIV status using the adjusted number of SARI hospitalisations divided by the mid-year total population estimates for each year, multiplied by 100,000.[14] HIV prevalence in the study population was estimated from the projections of the Actuarial Society of South Africa AIDS and Demographic model.[4] For estimation of incidence, we assumed that the HIV prevalence by age group amongst patients not tested for HIV was the same as that amongst those tested.

Confidence intervals for incidence estimates were calculated using the Poisson distribution. Age-specific and overall age-adjusted relative risk of SARI hospitalisation in HIV-infected compared to -uninfected persons was determined using log-binomial regression. To explore the possible effect of missing data on estimates of hospitalisation incidence by HIV status, we conducted a sensitivity analysis in which all cases not tested for HIV were assumed to be HIV uninfected.

Univariable and multivariable analyses were performed with Stata version 12 (StataCorp Limited, College Station, United States). To identify factors associated with HIV-infection status and death among SARI patients we implemented multivariable logistic regression models, starting with all variables that were significant at p<0.1 on univariable analysis and dropping non-significant factors with stepwise backward selection. All pairwise interactions of factors significant at the final multivariable additive model were evaluated. Two-sided p-values <0.05 were considered significant. For each univariable analysis, we used all available case information. In the multivariable model, patients with missing data for included variables were dropped from the model. Age group, duration of hospitalisation and year of admission were defined as categorical variables in multiple levels. All other variables were defined as the presence or absence of the attribute excluding missing data. To explore possible bias, individuals tested for HIV were compared to those not tested.

The protocol was approved by the Research Ethics Committees of the Universities of the Witwatersrand and KwaZulu-Natal. This surveillance was deemed non-research by the U.S. CDC and did not need human subjects review by that institution. Written informed consent was obtained from all participants.

From February 2009 through December 2012, 7977 individuals ≥5 years of age who fulfilled the SARI case definition were screened for study enrolment, of whom 7193 (90%) were enrolled (Fig. 1). The most common reasons for non-enrolment were being confused or too ill to consent (55%) and study refusal (11%). Of the 7193 enrollees, 8% were 5–14 years of age, 8% 15–24 years, 53% 25–44 years, 25% 45–64 years and 6% ≥65 years (Table 1). The majority of subjects were enrolled at CHBAH (76%), and 61% were female. Among patients with available information, the overall case-fatality ratio was 7%.

HIV-infection status was available for 6334 (88%) of enrolled individuals. Age-specific HIV prevalence findings were not significantly different when only patients tested through anonymised linked testing were included (data not shown). When comparing patients tested for HIV to those not tested for HIV, controlling for year of test, surveillance site and age group there were no differences in patient epidemiologic characteristics or case-fatality ratios (data not shown). The overall HIV prevalence among persons ≥5 years with available data was 74% (4663/6334) and was highest in the 25–44 year age group (88%, 3016/3421) (Table 1). Twelve percent of individuals had an underlying medical condition, excluding HIV. 53 women were pregnant. Only 14 individuals reported having been vaccinated against influenza in the current year and no subject had received pneumococcal vaccines.

Enrolment occurred throughout the year and peaked in the winter months (May-August) (Fig. 2). Overall, among those tested for respiratory viruses, 18% were positive for rhinovirus, 10% for adenovirus and 9% for influenza (Table 2). Other respiratory viruses tested positive in less than 5% of individuals. Adenovirus, rhinovirus and enterovirus were more commonly identified in individuals 5–14 years old than other age groups. Also, 9% of subjects tested positive for pneumococcus on PCR of whole blood specimens. The detection of influenza virus-associated SARI peaked during the winter months (Fig. 2). Although pneumococcus (on lytA PCR or culture) was detected perennially, detection increased during winter-months of at least two years (2009 and 2010).

The annual incidence of hospitalisation (per 100,000) for SARI at CHBAH ranged between 325 (95% CI 315–335) in 2012 and 617 (95% CI 603–632) in 2010 and was highest in the 45–64 year age-group; annual range 501 to 1284 (Table 3). HIV-infected individuals experienced an age-adjusted increased relative risk of 13 to 19 times for SARI hospitalisation compared to HIV-uninfected individuals. On sensitivity analysis, assuming that all patients not tested for HIV were HIV-uninfected, the trend towards a higher incidence of SARI hospitalisations in HIV-infected individuals remained in all age groups and years.

Compared to HIV-uninfected cases, using multivariable analysis, in addition to other factors, HIV-infected subjects were more likely to be receiving tuberculosis treatment at admission (OR 1.7; 95%CI: 1.1–2.7), have pneumococcal infection (OR 2.4; 95%CI: 1.7–3.3), be hospitalised for >7 days (OR 1.7; 95%CI: 1.2–2.3 as compared to <2 days), and had a higher case-fatality ratio (OR1.7; 95%CI: 1.2–2.3; Table 1). In contrast, HIV-infected subjects were less likely to have an underlying medical condition (OR 0.3; 95%CI: 0.2–0.3), or be infected with influenza (OR 0.6; 95%CI: 0.5–0.8).

Only 1455 (31%) of 4663 HIV-infected patients had available CD4+ T cell count data, of whom 68% (987) had CD4+ T-lymphocyte cell counts <200/mm3. The case-fatality ratio was significantly higher in HIV-infected subjects with severe immunosuppression (12%, 117/983) than those with CD4+ T-lymphocyte count of >200/mm3 (5%, 22/462, p<0.001). Of those with available data, 41% (1083/2629) reported currently receiving ART and 34% (1566/4569) reported receiving prophylaxis with trimethoprim-sulfamethoxazole. The case-fatality ratio was similar in individuals receiving (7%, 80/1075) and not receiving ART (vs. 8%, 121/1536, p = 0.681). The proportion of patients with CD4+ T-lymphocyte cell counts <200/mm3 was higher in patients not receiving ART (388/571, 68%) as compared to patients receiving ART (221/392, 56%).

The overall case fatality ratio was 7% (514/7154), with a median age of 42 years (interquartile range 23–74) in those who died. The case-fatality ratio was 1.5 times greater amongst HIV-infected (8%) as compared to HIV-uninfected (5%) individuals with SARI (Table 4). On multivariable analysis, independent risk indicators associated with death included increasing age group, HIV infection (OR 1.8 95%CI: 1.3–2.4), receipt of mechanical ventilation (OR 6.5; 95%CI: 1.3–32.0) and receiving supplementary-oxygen therapy (OR 2.6; 95%CI: 2.1–3.2) (Table 4).