To our knowledge, this single-center study is the first report of hospitalized adult patients with SARS-CoV-2 pneumonia in Andean subregion in a high-altitude population (Bogotá is situated at an altitude of 2,640 meters [8,660 feet] above sea level). We observed that COVID-19 hospitalized patients were more likely men over 50 years of age. Demographic characteristics and symptoms of COVID-19 were similar to previous reported data from patients admitted to hospitalization in China, United States, and the UK (11–14). In our study, in-hospital mortality was 22.6%; age, ICU admission and arterial pH were factors associated with this outcome.
Even though mortality in the present study was consistent with what has been reported, severity of respiratory failure seemed to be worse considering the high proportion of patients admitted to ICU (41.9%) and use of invasive mechanical ventilation (32.6%) in comparison to what was reported in China (26% and 17%, respectively), New York (14.2% and 12.2%, respectively), and the UK (17% and 10%, respectively) (10, 13, 14). This could be partially explained because one third (34.2%) of our patients didn’t receive corticosteroid therapy for COVID-19, due to part of our population was enrolled before the release of the RECOVERY trial report; although in the dexamethasone group in the RECOVERY trial the use of invasive mechanical ventilation was way lower (5.7%) than in the present study mortality was similar (22.9%) (15).
In Latin America, several reports have found a case fatality rate and mechanical ventilation use around 24% in hospitalized patients in Brazil (16, 17). However, in the COALITION II trial, that assessed efficacy and safety of adding azithromycin to COVID-19 treatment in Brazilian patients, mortality rate and use of mechanical ventilation was even higher to what we showed (40% and 52% in the control group, respectively) (18).
It has been suggested that some local factors in Latin America could influence clinical presentation of COVID-19 in comparison to Europe, such as the younger age of populations, tropical climate, and the immune regulation induced by helminthic infections or extensive BCG vaccination (19, 20). Colombia has a lower proportion of population over 60 years (13%) in comparison to Italy (29%) or Spain (25%), but at the same time, a lower hospital bed to population ratio and a fragmented healthcare system (21). These environmental and physiological characteristics may affect the course of COVID-19.
Moreover, PaO2/FiO2 ratio is lower at higher altitudes. Observational studies have been suggested that high-altitude is associated with infectivity and case fatality rate of COVID-19, due to factors such as adaptation to chronic hypobaric hypoxia, angiotensin-converting enzyme 2 expression, ultraviolet radiation and vitamin D production (22). However, results are conflicting and may be explained by differences in population density, underreporting of cases and barriers of access to healthcare among populations (23–25). Although altitude does not affect the mortality rate in general patients undergoing invasive mechanical ventilation, specific features of subgroups of patients with acute respiratory distress syndrome in COVID-19 may influence the need of ventilatory support at high-altitude (26). We theorize that high-altitude hypoxemia could have impacted in severity and course of acute respiratory failure in our COVID-19 population.
On the other hand, this study was conducted in a tertiary care center with one of the largest ICU in Bogotá, so presumably we admitted more severe patients prone to invasive mechanical ventilation from the area. The median duration of symptoms before admission (7 days [IQR 4–9]) was a little bit higher to what was reported in New York and the UK (13, 14); factors not yet assessed and involved in late admission of COVID-19 patients could have affected our results.
In our study, most patients had a mild pneumonia on admission, according to CURB-65 and qSOFA scores. Zhou et al. (10) described in a cohort of 191 patients in Wuhan a CURB-65 score 0 to 1 in most of them (75%) as well. It is possible that clinical prediction rules traditionally used to evaluate severity of community-acquired pneumonia may underestimate risk of mortality or ICU admission in SARS-CoV-2 pneumonia, since they were not developed to predict outcomes in viral pneumonia. Clinical deterioration in COVID-19 occurs later in comparison to bacterial pneumonia (in the present study 9 days from illness onset to ICU admission), so prediction rules at admission might be inaccurate. Data published is conflicting about the performance of these prediction rules in COVID-19 (27–30). Scores developed for viral pneumonia, such as MuLBSTA, 4C or CALL scores, may better predict the severity in this subset of patients, although they haven’t been validated in high-altitude populations (30–33).
Pulmonary embolism in COVID-19 has been described in one out of three patients admitted to ICU, even higher in histopathological studies, suggesting a main role of this complication in adverse patient outcomes (34, 35). In our study, pulmonary embolism occurred in just 1.6% of patients, with no associated deaths. Probably, there was underdiagnosis because CT pulmonary angiography was performed only in 66 patients, due to limitations for its use in ICU patients with acute renal failure, hemodynamic or ventilatory instability.
In the logistic models developed in this study, age and male sex were associated with COVID-19 severity; these results are consistent with the risk factors for poor prognosis previously reported (14, 36). Comorbidities such as arterial hypertension, diabetes mellitus, coronary heart disease, and obesity have been described as factors associated with mortality as well (14, 36, 37). In the present study, most patients had at least one underlying disease. The prevalence of obesity (30.2%) was considerably higher than the overall prevalence in Colombian adults (18.7%), suggesting that obesity increases the risk for COVID-19 requiring hospitalization (38).
Inflammatory biomarkers, such as C-reactive protein, ferritin and procalcitonin, have been associated with mortality among COVID-19 patients (36, 37). Likely, since some our patients did not have these markers measured on admission, we could not validate them as independent risk factors. On the other hand, biological variations on biomarkers due to different ethnic backgrounds might modify their prognostic ability in populations like ours. Regarding laboratory findings, in our model for mortality pH in arterial blood gas test on admission was validated as an important biomarker, this factor had not been associated with severe disease before. Most studies in COVID-19 like this have assessed prognostic markers on admission, further studies should address the diagnostic accuracy of markers follow-up.
There are some limitations to our study. First, clinical data collected relied on medical records which might lead to misclassification or recall biases. Nevertheless, we verified thoroughly the collected data; significant underreporting was unlikely because report of clinical characteristics and underlying comorbidities was consistent with existing literature. Second, there were missing data of symptoms and laboratory findings in some cases. This limitation is common in observational studies and might contribute to the underestimation of the true strength of any association. Third, power of statistical analyses may have been affected by the sample size and categorization of variables. Fourth, this study was conducted with hospitalized patients in a single tertiary care center in a high-altitude city, so it is possible that the sickest patients with highest degree hypoxemia were admitted. Patients were included by convenience sampling during the first months of the pandemic to describe the characteristics in our center; thus, our population is not representative of the general population through the whole pandemic. Caution should be exercised about generalizing these data to different settings. Finally, due to the study design we cannot establish a causal connection between risk factors and outcomes; our results and the model developed need a prospective validation.