The aetiology of severe community-acquired pneumonia requiring intensive care unit admission in the Western Cape Province, South Africa

Background Community-acquired pneumonia (CAP) is a common condition, with mortality increasing in patients who require intensive care unit (ICU) admission. A better understanding of the current aetiology of severe CAP will aid clinicians in requesting appropriate diagnostic tests and initiating appropriate empiric antimicrobials. Objectives To assess the comorbidities, aetiology and mortality associated with severe CAP in a tertiary ICU in Cape Town, South Africa. Methods We retrospectively analysed a prospective registry of all adults admitted to the medical intensive care unit at Tygerberg Hospital with severe CAP over a 1-year period. Results We identified 74 patients (mean (SD) age 40.0 (15.5) years; 44 females). The patients had a mean (SD) APACHE II score of 21.4 (7.9), and the mean ICU stay was 6.6 days. Of the 74 patients, 16 (21.6%) died in ICU. Non-survivors had a higher mean (SD) APACHE II score than survivors (28.3 (6.8) v. 19.4 (7.1); p<0.001). Mycobacterium tuberculosis (n=16; 21.6%) was the single most common agent identified, followed by Pseudomonas aeruginosa (n=9; 12.2%). All P. aeruginosa isolates were sensitive to first-line treatment. No organism was identified in 32 patients (43.2%). Conclusion M. tuberculosis was the single most common agent identified in patients presenting with CAP. The mortality of CAP requiring invasive ventilation was relatively low, with a strong association between mortality and a higher APACHE II score.

If HIV is left untreated, it increases the incidence of pneumococcal pneumonia by 17 -35-fold. [10] A surveillance study conducted in rural Kenya with an HIV incidence of 50 -75% showed an incidence of pneumococcal pneumonia in 5 and 67 per 1 000 participants in the HIV-negative and HIV-positive populations, respectively. [7] Poor socioeconomic circumstances such as overcrowding, malnutrition and poor indoor ventilation contribute to the burden of tuberculosis in SA, and may also contribute to increased incidence of CAP. [7] Furthermore, it has been shown that 18 to 40% of patients with CAP may test positive for TB, either as a co-existing infection or a primary pathogen. [11] S. pneumoniae is the most common identified bacterial pathogen leading to CAP worldwide. Mortality rates as a result of S. pneumoniae have been shown to be three times higher than those for other pathogens. [1] Atypical pathogens (Mycoplasma pneumoniae, Legionella spp., Chlamydia pneumoniae) account for <2% of CAP in SA. [3] Rational antimicrobial prescription is an essential task, as emerging antimicrobial resistance is a global concern. [3] Guidelines for the management of CAP are well documented by internationally reputable sources, but mainly targeted towards the developed world.
Patients with severe CAP treated with combination therapy, including a macrolide, have improved survival compared with those who receive monotherapy. [5] Population-based surveillance programmes must identify the emergence of resistant or atypical organisms in order to guide clinicians on empirical antimicrobial choices for CAP.
The findings of international studies may not always be relevant to SA. We therefore aimed to assess the comorbidities, aetiology and mortality associated with severe CAP in a tertiary ICU in Cape Town, SA.

Study population
All patients admitted to the ICU of Tygerberg Hospital from 1 June 2016 to 31 May 2017 with CAP requiring invasive ventilation were identified from an existing prospectively collected registry. This institution is a 1 380-bed academic hospital in Cape Town, SA. It is one of two academic referral centres in the city, and renders a tertiary service to a population of ~1.5 million, with a local incidence of TB in the order of almost 1 000 per 100 000. [12] The study was approved by the Stellenbosch University Health Research Ethics Committee (ref. no. S18/10/243).
Pneumonia was defined as any patient exhibiting a triad of infection, signs or symptoms localised to the lower respiratory tract and a new radiological infiltrate. [13] We included all patients with complete medical, microbiological and radiological records, and excluded patients hospitalised for >48 hours in our or any other medical care centre.

Data collection
The demographic data, acute physiology and chronic health evaluation (APACHE) II and comorbidities of all patients were documented, as well as all positive microbiological culture results. Blood cultures and tracheal aspirates were routinely collected from all patients on admission to ICU at the time of the study, and as part of ICU protocol, Xpert MTB/RIF was performed on tracheal aspirates of all patients admitted to the ICU with severe CAP, irrespective of clinical and radiological suspicion for TB. Moreover, urinary testing was performed to detect the presence of Legionella antigen, and nasopharyngeal swabs for influenza were performed at the discretion of the treating physician. Final microbiological diagnoses were made in conjunction with the divisions of infectious diseases and microbiology in each case.

Statistical aspects
Descriptive statistics and χ 2 or Fisher's exact tests (where indicated) were performed on dichotomous categorical variables, and t-tests on continuous data. Unless stated otherwise, data are displayed as means and standard deviations (SD).

Results
During the study period, there were 423 admissions to the ICU, with all patients requiring invasive ventilation on admission. We identified 74 patients who met the study criteria (mean (SD) age 40.0 (15.5 years); 44 females). The patients had a mean (SD) APACHE II score of 21.4 (7.9; range 6 -39), and the mean ICU stay was 6.6 days (range 1 -41 ; p=0.72). Therefore, no comorbidity was protective against the development of severe CAP. There was no mortality observed in pregnant patients. M. tuberculosis (n=16; 21.6%) was the single most common organism identified, followed by Pseudomonas aeruginosa (n=9; 12.2%). All P. aeruginosa isolates were sensitive to first-line treatment (e.g. ceftriaxone). Of the patients with P. aeruginosa pneumonia, 22% were HIV-positive and 11% had underlying chronic lung disease. No organism was identified in 32 patients (43.2%). The diagnosis of a single case of Varicella zoster pneumonia was based on the clinical picture (including classical skin rash) and chest radiograph findings and there was one case of Pneumocystis jirovecii.

Discussion
We found that Mycobacterium tuberculosis was the single most common agent identified, followed by P. aeruginosa and S. pneumoniae. The ICU mortality was 22%, and the only predictor of mortality was a higher APACHE II score.
There were five confirmed cases of S. pneumoniae pneumonia in this study. This may be explained by preceding use of a beta-lactam antibiotic that is routinely commenced at primary and secondary healthcare level, often prior to taking appropriate samples for culture. The low diagnostic yield was also illustrated in international studies in the USA and the Netherlands. [1,4] Despite sophisticated methods used to identify the aetiology of CAP, there was an inability to identify a cause for pneumonia in more than 50% of patients. [14] Another explanation is the indirect reduction in colonisation with S. pneumoniae in adults as a result of pneumococcal conjugate vaccination in infants. This herd immunity has been illustrated in international studies, [15] as well as a study conducted in a part of rural SA with a high HIV burden. [16] The present study highlights the significance of TB presenting as lobar consolidation. It is in keeping with previous studies from other high TB-burden countries. [8,11,17] Of the study patients diagnosed with TB, only 15% had HIV co-infection, highlighting the high prevalence of pulmonary TB irrespective of HIV status. Nyamande et al. [11] identified M. tuberculosis as the most common organism in a study conducted in KwaZulu-Natal. Among the patients diagnosed with pulmonary TB in the present study, there was a mortality rate of 31% during ICU stay. It should be emphasised that, unlike most blood and tracheal culture techniques, microbiological confirmation of TB remains possible even weeks after empirical therapy. This may therefore have influenced the high yield of M. tuberculosis in this study. Surprisingly, P. aeruginosa was identified as an aetiological pathogen in 12% of patients. It is generally considered an uncommon cause of CAP, and mostly associated with nosocomial infections, with associated high mortality and typically extended antibiotic resistance patterns. [7,18] In the context of CAP, it is more likely to be seen in patients who have an identifiable risk factor for developing P. aeruginosa pneumonia. [19] P. aeuruginosa has been recognised as the second-most common bacterial organism for CAP in patients with HIV co-infection, [20] but this was not reflected in the present study.
The study showed an in-ICU mortality of 22%, even though the average APACHE score predicted a mortality of ~39%. Previous studies conducted in Europe had case-fatality rates ranging from 11% to 48% in ICU. [1] A higher APACHE II score at ICU admission, male sex, chronic heart failure and dialysis were identified as independent risk factors for in-hospital mortality in a study by Li et al. [21] An association between mortality and underlying chronic lung diseases is well described in the literature. Vidal et al. [22] identified a 23% mortality in patients with CAP and comorbid chronic respiratory disease in a study conducted in Portugal. A retrospective cohort study in the USA showed a statistically significant increase in 30-day mortality, rate of ICU admission and length of hospital stay among patients with co-existing COPD who were hospitalised for CAP. [23] In our population, however, mortality was not associated with underlying lung diseases.

Study limitations
Our study had some limitations. Owing to the small sample size, we were not able to identify statistically significant associations between certain risk factors and mortality. No organism was identified in just over 40% of cases, many of whom in all probability may have had Streptococcus pneumoniae or other common causes of bacterial pneumonia. Testing for viral infections and atypical pathogens (including Legionellae) was only performed when requested and not routinely, which may have led to a significant proportion of cases not being accurately diagnosed.

Conclusion
M. tuberculosis was the single-most common agent identified in patients presenting with severe CAP requiring ventilation. The relatively high percentage of confirmed community-acquired P. aeruginosa could be related to the concomitant severe drought experienced at the time of the study. The mortality of CAP requiring invasive ventilation was relatively low, and the only predictor of mortality was a higher APACHE II score.
Declaration. The research for this study was done in partial fulfillment of the requirements for AM's MMed (Internal Medicine) at Stellenbosch University. Acknowledgements. None. Author contributions. AM and CFNK conceptualised the study. TJJ, KGJ and JS collected data. CFNK, UL and JTT were responsible for critical revision of the manuscript. Funding. None. Conflicts of interest. None.