We included data collected from January 2010 through December 2013 from active, prospective, hospital-based surveillance (the Severe Acute Respiratory Illness [SARI] Program) in 3 of the 9 provinces of South Africa (Chris Hani-Baragwanath Academic Hospital [CHBAH] in an urban area of Gauteng Province, Edendale Hospital in a periurban 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 Hospital in a periurban area of the North West Province.¹⁹

A case of hospitalized LRTI was defined as a hospitalized infant with ≤7 days’ symptom duration meeting age-appropriate clinical case definitions (aged 2 days through <3 months of age with physician-diagnosed sepsis or LRTI or aged 3 to <6 months with physician-diagnosis of bronchitis, bronchiolitis, pneumonia, and/or pleural effusion). We included infants aged <3 months with physician-diagnosed sepsis because LRTI could present nonspecifically in this age group.

All patients admitted from Monday through Friday were eligible for enrollment. The total number of infants admitted meeting study case definitions and numbers enrolled were documented for the entire study period. In 2013, at CHBAH, enrollment of patients was limited to systematically varying 2 of every 5 working days because of high patient numbers and limited resources. Study staff completed case report forms until discharge and collected nasopharyngeal aspirates and blood specimens. All decisions on medical care, including mechanical ventilation and testing for Mycobacterium tuberculosis, were undertaken at the discretion of the attending physician. Data in case report forms were reviewed regularly to identify inconsistencies, and regular site visits were conducted to ensure adherence to study procedures.

The determination of HIV infection and exposure status used information from testing undertaken as part of standard of care during the current admission²⁰ and through anonymized linked dried blood spot specimen testing by HIV polymerase chain reaction (PCR) as part of surveillance. HUU infants were defined as infants with a negative HIV PCR result on admission and a documented or reported negative maternal HIV status. HEU infants were defined as infants with a negative HIV PCR result on admission and a documented or reported positive maternal HIV status. HIV-infected infants were defined as infants with a positive HIV PCR result and CD4+ T-cell counts were determined by flow cytometry as part of standard of care.²¹ Maternal breastfeeding was not considered in assessment of HIV infection status because PCR results at the time of admission in infants aged <6 months are likely to reflect current infant HIV status irrespective of whether infection was acquired at birth or through breastfeeding. Definitions for other underlying conditions are provided in the Methods section of the online Supplemental Information.

For additional details, please see the online Supplemental Information.

Nasopharyngeal aspirates were immersed in universal transport medium at 4° to 8°C and transported to the National Institute for Communicable Diseases within 72 hours of collection. Specimens were tested by multiplex real-time reverse-transcription PCR for 10 respiratory viruses including influenza A and B viruses, parainfluenza virus types 1, 2, and 3; respiratory syncytial virus (RSV); enterovirus; human metapneumovirus (hMPV); adenovirus; and human rhinovirus.²² Streptococcus pneumoniae was identified by quantitative real-time PCR detecting the lytA gene from whole blood specimens.²³

Incidence estimates were calculated for Soweto (CHBAH) because population denominators were available for this site. This hospital is the only public hospital serving a community of ~12 000 infants aged <6 months in 2012 among whom <10% have private medical insurance.²⁴ Consequently, the majority of individuals requiring hospitalization from this community are admitted to CHBAH. We estimated the incidence of hospitalization stratified by HIV infection and exposure status for the years 2010 and 2011 because data on HIV prevalence in pregnant women and rates of MTCT were only available for these years.^{17, 25} We assumed that the HIV exposure and infection status among infants with unknown status was similar to those with known status. We calculated the annual incidence of LRTI hospitalization for infants <6 months of age per 100 00 individuals, by HIV exposure and infection status, by dividing the number of enrolled SARI cases each year (adjusted for nonenrollment on weekends and through refusal) in each category (HUU, HEU, and HIV-infected) by midyear population estimates for each group, multiplied by 100 000. We obtained estimates of the numbers of infants in the population in each HIV exposure group using population denominators for infants aged <1 year from the 2011 census for Region D (Soweto) and dividing these by 2. We then applied published data on HIV prevalence in pregnant women and rates of MTCT from the 2011 National South African PMTCT survey for Gauteng Province, where Soweto is located.^{25, 26} Etiology-specific numbers were estimated by multiplying the adjusted numbers of SARI cases by the age- and HIV exposure status-specific detection rates for each pathogen. We calculated the relative risk of SARI and etiology-specific hospitalization comparing HUU to HEU and HIV-infected infants. Confidence intervals (CIs) were calculated by using the Poisson distribution for incidence rates (IRs) and log binomial model for incidence rate ratios (IRRs).

For the analysis of factors associated with HIV-exposure status and death we included infants from any surveillance site from 2010 through 2013 with known HIV exposure status and in-hospital outcome. We implemented 2 multivariable models to identify factors associated with 2 outcome variables: (1) HIV infection/exposure status and (2) mortality. Multinomial regression was used for the comparison of factors associated with HIV infection/exposure. Multinomial regression allows modeling of outcome variables with >2 categories and relates the probability of being in category j to the probability of being in a baseline category. A complete set of coefficients was estimated for each of the j levels being compared with the baseline, and the effect of each predictor in the model was measured as relative risk ratio (RRR). For this analysis, HUU infants were used as the referent group and compared with HEU and HIV-infected infants. The model to assess factors associated with mortality was implemented using logistic regression. For both models, we assessed all variables that were significant at P < .2 on univariate analysis and dropped nonsignificant factors (P ≥ .05) with stepwise backward selection. Patients with missing data for included variables were dropped from the model. In addition, for the mortality analysis, we excluded variables considered to be on the causal pathway for mortality (eg, length of hospital stay, ICU admission). The statistical analysis was implemented using Stata version 12 (StataCorp Inc, College Station, TX). We conducted a sensitivity analysis of the association between HIV exposure and infection status and mortality restricted to individuals with documented maternal HIV status and a separate analysis excluding infants with suspected sepsis. Results of these analyses are reported in the online Supplemental Information.

From January 2010 through December 2013, 3537 infants aged <6 months fulfilling the LRTI case definition were enrolled (Fig 1). HIV infection and exposure status were determined in 2507 (71%) enrolled infants. Maternal HIV status was obtained from medical records or laboratory report (vs maternal verbal report) in 27% (229/850) of HEU infants and 19% (268/1446) of HUU infants. The proportion of infants with available HIV exposure status information was 66% (578/878) in 2010, 73% (805/1096) in 2011, 72% (687/949) in 2012, and 71% (437/614) in 2013 (P = .001). A higher proportion of infants had available HIV exposure status information at Edendale (81%, 454/562) and Matikwana/Mapulaneng Hospitals (82%, 367/450) than CHBAH (67%, 1511/2263) or Klerksdorp Hospital (67%, 175/262) (P < .001). Infants with known HIV exposure status did not differ from those with unknown HIV exposure status with regard to other epidemiologic or clinical characteristics such as age group, symptom duration, or in-hospital outcome (data not shown). The proportion of enrolled infants who were HIV infected was 6% (87/1511) at CHBAH, 15% (54/367) at Matikwana/Mapulaneng, 9% (41/454) at Edendale, and 17% (29/175) at Klerksdorp (P < .001). The proportion of infants who were HEU was 31% (468/1511) at CHBAH, 30% (110/367) at Matikwana/Mapulaneng, 44% (201/454) at Edendale, and 41% (71/175) at Klerksdorp (P < .001).

Only 7% (196/2391) of infants had an underlying medical condition other than HIV infection, the commonest of which was premature birth (n = 164). Among infants with available data, 22% (170/773) were malnourished, the overall case fatality ratio was 3% (71/2488), and 2% (55/2493) were admitted to the ICU. Tuberculosis testing was performed on 372 (15%) children, from whom M tuberculosis was identified in 37 (10%) cases (18/195 [9%] in HUU, 13/134 [10%] in HEU, and 6/43 [14%] in HIV infected, P = .640). An additional 43 children were started on tuberculosis treatment based on clinical suspicion without laboratory confirmation (19/1434 [1%] in HUU, 12/845 [1%] in HEU, 12/207 [6%] in HIV-infected, P < .001). Supplemental oxygen was administered to 47% (1160/2493) of enrolled infants. The percentage of infants aged >6 weeks who had received age-appropriate doses of pneumococcal conjugate vaccine was 76% (1113/1455). Information on receipt of antiretroviral treatment (ART) was only available for 48 of 211 (23%) HIV-infected infants, and 16 (33%) had reported receipt of ART. CD4+ T-cell count was only available for 56 (27%) of HIV-infected infants, and 82% (n = 46) had CD4+ T lymphocytes <25%.²⁷

The annual incidence (per 100 000 population) of LRTI hospitalization among infants <6 months of age in Soweto was lowest in HUU infants (10 404; 95% CI 9947–10 877), and 1.4-fold higher in HEU (14 288; 95% CI 13 437–15 178; IRR 1.4; 95% CI 1.3–1.5) and 3.7-fold greater among HIV-infected infants (38 445; 95% CI 33 080–44 438; IRR 3.7; 95% CI 3.2–4; Table 1). Among viruses with high likelihood of etiologic attribution, RSV-associated (IRR 1.4; 95% CI 1.3–1.5), hMPV-associated (IRR 1.4; 95% CI 1.1–2.0), and influenza-associated (IRR 1.2; 95% CI 0.8–1.8) LRTI point estimates of incidence were higher in HEU relative to HUU infants, although this was not statistically significant for influenza (for which there was the lowest number of cases, n = 56). For other viruses, estimates were generally similar, although numbers were small in some groups (Table 1).

Compared with HUU on multivariable analysis, and controlling for hospital, race, and year of admission, HEU infants were more likely to be hospitalized for 2 to 7 days (adjusted relative risk ratio [aRRR] 1.5; 95% CI 1.3–1.9) or >7 days (aRRR 1.8; 95% CI 1.4–2.4) compared with <2 days; to require mechanical ventilation (aRRR 2.3; 95% CI 1.1–5.0); and to die in-hospital (aRRR 2.1; 95% CI 1.1–3.9; Supplemental Table 4).

Compared with HUU on multivariable analysis controlling for hospital, age, and year, HIV-infected infants were less likely to test positive for RSV (aRRR 0.3; 95% CI 0.2–0.04); however, HIV-infected infants were more likely to be hospitalized for >7 days (aRRR 5.6; 95% CI 3.1–10.0) compared with <2 days, to require oxygen (aRRR 2.0; 95% CI 1.3–2.9) or ventilation (aRRR 2.9; 95% CI 1.1–7.7), and to die in-hospital (aRRR 6.3; 95% CI 3.1–13.1; Supplementa Table 4).

On multivariable analysis, HEU had 2 times (adjusted odds ratio [aOR] 2.1; 95% CI 1.1–4.5), and HIV-infected 5 times (aOR 5.3; 95% CI 2.4–11.8), greater odds of in-hospital mortality compared with HUU infants. PCR detection of pneumococcus in blood was associated with increased mortality (aOR 4.8; 95% CI 1.8–12.7), and detection of any virus (vs no virus detected) was associated with decreased mortality (aOR 0.5; 95% CI 0.2–0.9; Table 2). Mortality was substantially higher at Matikwana/Mapulaneng Hospitals (aOR 17.2; 95% CI 7.0–45.0) compared with CHBAH. When stratified by etiology, for most pathogens, case fatality ratios were elevated in HEU and HIV-infected individuals compared with HIV-uninfected individuals, although numbers of cases and deaths were low in some subgroups (Table 3). Case fatality ratios were significantly elevated for HEU infants with RSV compared with HUU infants.