This retrospective observational study provided some important insights into the characteristics and outcomes of HIV-positive patients admitted with severe pneumonia and highlights the fact that, despite HIV now being a manageable, chronic disease, there is much work to be done in attaining the WHO goal of having 90% of patients diagnosed and 90% of those on appropriate treatment . In this study only 35.9% were on HAART at the time of admission to the hospital and 58.1% were only diagnosed with HIV during the presenting admission. Late diagnosis of HIV has previously been shown to be a risk factor for admission to the ICU .
Taking into account the young age (mean 38.3 years) and high mortality (40.1%) in the group studied, it is clear that earlier diagnosis and appropriate treatment is necessary to improve outcomes. The high mortality (40.1%; 47/117) in this study cohort should be contextualised, taking into account that this was a referral centre and only those with severe pneumonia who required admission to high care or ICU. The criteria for admission to high care is the requirement of at least one organ system support (oxygen > 60%, high flow oxygen, non-invasive ventilation, inotropic support or dialysis). Patients managed in the general wards were not included. As more than half of patients (52%) required admission to the intensive care unit, and the mean length of hospitalisation was 15.8 days, it is clear that management is resource-intensive and costly.
Evidence in the literature regarding predictors of outcome in patients with HIV and pneumonia has been heterogenous [4,5,6,7,8,9,10,11,12] and criteria for admission to critical care units of HIV-positive patients is controversial [4,5,6,7,8,9,10,11,12]. The CD4 counts and severity-of-illness scores have previously been associated with outcome , but in this study no level of CD4 was predictive of mortality, even at counts of < 100 cells/µL. The CD4 of the study population was low with a mean value of 120.2 cells/µL. Furthermore, the need for ICU admission and ventilation increased mortality risk.
In the era of widespread HAART availability, the spectrum of disease seen in people living with HIV and admitted to ICU in developed countries, has shifted to non-AIDS-related illnesses, similar to those without HIV . However, in more resource-limited areas, late-stage diagnosis of HIV with associated opportunistic infections, remains a common reason for ICU admission in developing countries [4, 17]. Chiang and colleagues (2011) did not find the presence of HAART to be associated with better outcomes in patients with HIV admitted to the ICU , whilst Morquin and colleagues (2012) found that the initiation of HAART may improve survival in patients with HIV admitted to the ICU . The patients on HAART in this study cohort were more likely to survive than those who were not. In the cohort, 38.6% (27/70) of survivors were on HAART compared to 31.0% (15/47) of non-survivors, but this was not statistically significant (p 0.46). When the subset of 61 patients who required admission to ICU were considered, however, there was a statistically significant survival benefit for those on HAART: 100% (21/21) of survivors were on HAART, while only 25% (10/40) of non-survivors were on HAART (p 0.03). The initiation of HAART in ICU has been associated with improved survival in previous studies .
The presence of CMV viremia (viral load > 1000 copies/mL) was associated with a higher mortality (p 0.028). The reactivation of CMV in HIV-positive patients with immune suppression has previously been found to be associated with worse clinical outcomes .
Patients with TB admitted to the ICU had a high mortality (33% to 67%) in the published literature [5,6,7,8,9,10,11,12,13]. A South-African study in patients with TB admitted to the ICU, showed that 53% of participants were co-infected with HIV and the mortality was 59% . In the current study, 35% (41/117) of the patients were found to be culture-positive for TB. There was no statistically significant difference in mortality between those with TB (46.4%) and the cohort as a whole (40.1%).
Imaging by chest X-ray is the first-line imaging for patients presenting with respiratory symptoms. Chest X-ray findings have a wide differential diagnosis, and pneumonia often has overlapping features . The chest X-ray remains a good tool to confirm the diagnosis and look for complications of pneumonia, but it is inferior to computed tomography (CT) in providing diagnostic certainty and establishing a potential causative organism . In this study chest X-ray imaging did not correlate with a specific etiological diagnosis.
Lung ultrasound has become a useful bedside tool in evaluating patients with dyspnea  and in those with pneumonia . Although lung ultrasound is comparatively novel when one considers chest X-ray and CT scan as imaging modalities, it has the advantage of being a bedside tool that can be repeated at regular intervals without having to transport patients. In the diagnosis of pneumonia, the pooled sensitivity and specificity were 94% (CI 92% to 96%) and 96% (CI 94% to 97%), respectively. In a small case series Limonta and colleagues (2019) found that all patients with P. jirovecii infection had multiple B-lines on ultrasound of the lung; the so-called B-pattern or interstitial pattern . Another, series of 14 patients found B-lines to be 100% sensitive for the presence of P. jirovecii and also found subpleural consolidation and cystic changes in patients with P. jirovecii.  In the current study the presence of B-lines on lung ultrasound were strongly associated with P. jirovecii as an aetiology for pneumonia in patients with HIV (positive predictive value 80%, 95% CI 65.7–94.3). Notably the use of lung ultrasound picked up a small pneumothorax, which was not visible on conventional chest radiography.
Kahn and colleagues (2020) considered a combination of ultrasound features to predict TB in patients with HIV and pneumonia. The presence of a pleural effusion argued strongly for TB as the aetiology of pneumonia in this cohort of patients . In this study, 15/117 patients had pleural effusions on lung ultrasound. |Seventy-three percent (11/15) of these proved to have TB as the aetiology for their pneumonia. Three patients were found to have pleural effusion with visible fibrin strands and all three also tested positive on TB culture. This number was too small to make inferences on, but the signal warrants further study in a larger cohort. Larger studies in the future are likely to delineate the role of ultrasound in the diagnosis of TB in the context of HIV to a greater extent. The current study was limited by the small number of patients studied. There was selection bias as in that the study was performed at a referral centre, which means that patients are generally more severely ill.