Risk factors for intensive care unit admission or mortality in adults hospitalized for COVID-19 at high altitude

ABSTRACT Objective. To determine the risk factors for admission to the intensive care unit (ICU) or mortality in patients hospitalized for COVID-19 in a hospital in Puno, Peru. Materials and methods. Retrospective cohort study in adults hospitalized between April and December 2020. We evaluated Sociodemographic characteristics, vital functions, comorbidities, treatment received and its association with admission to ICU or mortality (adverse outcome). Poisson regression with robust variance was used to calculate crude and adjusted relative risks (RR) with their 95% confidence intervals (95%CI). Results. A total of 348 medical records were analyzed. The median age in years was 42.5 (IQR: 30.0; 58.0); 38.2% were male, and 35.3% died or were admitted to the ICU. Those admitted with an oxygen saturation ≤ 75% were 2.79 times more likely to have the adverse outcome (p < 0.001), compared to those admitted with a saturation ≥ 85%; those admitted with a value between 75-79% were 2.92 times more likely to have the adverse outcome (p < 0.001); likewise, those admitted with saturation between 80-84% were 1.70 times more likely to have the adverse outcome; however, the difference was not statistically significant (p=0.066). In addition, male patients, RR= 1.75 (p<0.001); those aged > 40 years, RR 3.5 (p=0.001); those with tachypnea, RR=1.66 (p=0.010); or with diabetes, RR = 1.53 (p=0.011) had higher risk of presenting the adverse outcome. Conclusions. The risk factors for ICU admission or mortality due to COVID-19 were male sex, age over 40 years, low saturation, diabetes and tachypnea.


INTRODUCTION
A significant number of severe pneumonia cases were reported in Wuhan, China in December 2019 (1) . Subsequently, this new disease would be identified as COVID-19, caused by severe acute respiratory syndrome type 2 virus (SARS-CoV-2) (1) . On March 11, 2020, The World Health Organization (WHO) proclaimed COVID-19 as a pandemic due to its extensive worldwide spread, suggesting that countries follow a comprehensive strategy aimed at preventing infections and deaths (2) . As of October 26, 2021, more than 4.9 million deaths caused by this disease had been reported worldwide. In response, the Peruvian Ministry of Health (MINSA) decreed a state of national health emergency (3) . Confirmed cases of coronavirus in the country totaled 2,197,052, with 200,118 deaths and a case fatality rate of 9.11% (4) . By August 2020, Peru became the country with the highest COVID-19 mortality per number of inhabitants (4,5) .
There are studies that suggest that the prevalence and impact of COVID-19 could be diminished by a potential protective factor such as altitude (6,7) ; either due to the drastic changes that usually exist in these environments, high ultraviolet radiation and dryness of the air (6) . Likewise, the decreased expression of angiotensin-converting enzyme (ACE) in the pulmo-Motivation for the study: Risk factors for adverse outcome due to COVID-19 are not well defined in populations in altitudes greater than 3000 m.
Main findings: Factors associated with death due to COVID-19 were age > 40 years, male sex, diabetes, admission with oxygen saturation less than 80% and respiratory rate > 22 bpm. Implications: Identifying these risk factors in a population located at 3827 m altitude will allow us to recognize patients with worse prognosis in order to implement adequate measures.

KEY MESSAGES
nary endothelium of high-altitude inhabitants could generate a physiological protective factor as it is the binding site of SARS-CoV-1 and SARS-CoV-2 (6,7) . However, these are ecological studies, with inconclusive statements, and more research is needed.
Oxygen saturation (SatO 2 ) is an important prognostic factor, since having a low percentage at hospital admission is associated with higher mortality (3,15,16) . A study conducted in Lima, a city at an altitude of 161 m, found that having saturation < 90% on admission was associated with a higher risk of death compared to those who arrived with > 90% (3) . We have not found studies that determine what percentage saturation indicates a poor prognosis in high-altitude residents. However, it is known that high-altitude residents have different SatO 2 values than residents at sea level. A study conducted in healthy residents of Huánuco (1894 m altitude) and Cerro de Pasco (3399 m altitude) found that the mean saturation was 96.2 and 87.0%, respectively, concluding that the higher the altitude of residence, the more noticeable the changes in the normal values of SatO 2 (17) , it should be noted that this study was conducted outside the context of a pandemic.
Puno is a region located at an altitude of 3827 m, so the results of this study may be useful to understand the dynamics of the disease in high-altitude populations and to propose future studies.
This research aimed to determine the risk factors for intensive care unit (ICU) admission or mortality in a high-altitude population of adults hospitalized for COVID-19 in a hospital in Puno, Peru.

Design and context
A retrospective cohort study was designed at the Manuel Núñez Butrón Regional Hospital (HRMNB) in Puno, Peru, that included adult patients hospitalized in the COVID area between April and December 2020. The HRMNB is a ca-tegory II hospital, administered by the Ministry of Health (MINSA) located at an altitude of 3827 m. The first CO-VID-19 cases were recorded in early April, with the first wave occurring between July and August.

Population
We included adult patients (≥18 years), hospitalized in CO-VID areas of the HRMNB who previously resided in Puno for at least two months. Patients who voluntarily withdrew and those without confirmed diagnosis by a rapid test were excluded.

Sample
We carried out a census of patients hospitalized in the COVID area of the HRMNB of Puno from April to December 2020, obtaining 542 medical records. Epidat 4.2 was used to define a minimum sample size. Based on a previous study (3) , we constructed a contingency table and calculated the probability of dying between exposed and unexposed; we considered a ratio of 0.33 with a confidence level of 99.9% and power of 99.9%. p1 represents the probability of dying in patients with low saturation (<85%) being 61.6%; while; p2, is the probability of dying in patients without low saturation (≥ 85%) which was 14%. The minimum required sample size was 192 and considering 10% of incorrectly filled histories, it was 214.

Variables
The dependent variable was the adverse outcome, which consisted of admission to the ICU or in-hospital mortality. This variable was categorical dichotomous (died or was admitted to the ICU / did not die or was not admitted to the ICU).

Data Collection
Data collection was carried out after approval from the Research Ethics Committee of the Universidad Peruana de Ciencias Aplicadas and authorization from the Regional Health Directorate (DIRESA) of Puno. We obtained access to the medical records of patients hospitalized by COVID-19 at the HRMNB of Puno between April and December 2020. Those that met the selection criteria were selected. Between June and July 2021, the data collection form was filled out with the variables chosen by literature review, then the data were organized in Excel.

Data analysis
We used the STATA v16 statistical analysis program. For the univariate analysis we calculated percentages for categorical variables, measures of central tendency and dispersion for quantitative variables, the mean and standard deviation (SD) were calculated if there was normality, otherwise, median and interquartile range (IQR). For the bivariate analysis, we used Student's t test / Mann Whitney U test and chi-square / Fisher's test, according to compliance with assumptions. Finally, for the multivariate analysis, we used Poisson regression with robust variance to calculate the crude and adjusted relative risk (RR) with 95% confidence intervals (95%CI). An alpha of 0.05 was considered. Variables were entered into the model according to epidemiological/theoretical criteria. Collinearity with VIF (variance inflation factor) was assessed using a cutoff point of 2.

Ethical aspects
The protocol was approved by the Ethics Committee of the Universidad Peruana de Ciencias Aplicadas (SCEI 024-01-  and authorized by the DIRESA of Puno. Since the data were already in the medical records, the research team had no direct contact with the research subjects and confidentiality was maintained. PRISA registration code: EI00001585.

RESULTS
A total of 542 medical records were obtained from patients hospitalized due to COVID-19 in the hospital of Puno, from April to December 2020. Of these, 194 were excluded because they did not meet inclusion criteria or because they were not available. Finally, 348 were included as part of the analyzed sample ( Figure 1).

Descriptive analysis
With respect to the baseline characteristics (Table 1), 48.2% were under 40 years of age, 29.1% were between 40 and 59 years of age and 22.7% were 60 years of age or older; 61.8% were female. Regarding the outcome, 18.4% were admitted to the ICU, while 5.5% used invasive mechanical ventilation, and in-hospital mortality was 29.9%. In terms of comorbidities, 13.8% had diabetes, 7.5% had AHT, 4.6% had obesity and, finally, 21.8% had one comorbidity and 6.4% had two or more of them.
Regarding oxygen saturation, 69.9% of patients were admitted with a saturation greater than 85%, 5.5% were admitted with a saturation ranging from 80 to 84%, 6.1% with a saturation ranging from 76 to 79% and 18.5% were admitted with less than 75% saturation.
Regarding diagnosis, 48.3% had clinical diagnosis. In addition, 66.9% were positive for the IgM rapid test and 81.9% were positive for the IgG rapid test. As for laboratory tests, the mean hemoglobin was 15.1 g/dL with a standard deviation (SD) of 3.0. In addition, 46.7% had lymphopenia and 33.9% had leukocytosis.

Bivariate analysis
During the bivariate analysis, we observed a statistically significant association between age and adverse outcome (p <  0.001) as well as with male sex (p < 0.001) and reason for admission (p < 0.001) ( Table 2).
Regarding comorbidities, diabetes (p<0.001) and obesity (p<0.001) had a significant association with adverse outcome; no significant association was observed with AHT (p=0.104).
As for vital signs, we found a significant association between oxygen saturation and adverse outcome (p < 0.001), as well as with respiratory frequency (p < 0.001), lower systolic blood pressure (p = 0.017) and lower diastolic blood pressure (p = 0.025).
With respect to laboratory tests, a significant association was found between adverse outcome and higher hemoglobin level (p < 0.001), as well as with higher glucose level (p < 0.001) and lymphopenia (p < 0.001). However, no significant association was found between adverse outcome and blood group (p = 0.200).

Analysis of multiple variables
During the multivariate analysis we found a statistically significant association between age and the adverse outcome, with those between 40 and 59 years of age having 3.7 times the probability of developing an adverse outcome (p < 0.001) and those over 60 years of age having 3.5 times the risk compared to those under 40 years of age (p = 0.001) (Table 3). Likewise, men had 1.8 times the risk of adverse outcome (p < 0.001) after adjusting for the rest of the variables in the equation. Regarding comorbidities, patients with diabetes had 1.3 times the risk of developing the adverse outcome (p = 0.011) and obese patients had 1.2 times the risk of developing it; however, the association was not significant (p = 0.260). On the other hand, there was no statistically significant association with AHT (p = 0.851). Patients with oxygen saturation ranging between 80 and 84% had 1.7 times the risk of developing an adverse outcome compared to those admitted with more than 85%, without a significant association (p = 0.066); on the other hand, those with values between 75 and 79% had 2.9 times the probability of developing an adverse outcome (p < 0.001), the same for those admitted with less than 75%, who had a risk of 2.8 (p < 0.001).
Patients admitted with tachypnea (> 22 bpm) had 1.7 times the risk of an adverse outcome, with a statistically significant difference (p = 0.010). There was no association between the use of anticoagulants and the risk of having an adverse outcome (p > 0.05). Similarly, there was no association between adverse outcome and dexamethasone (p = 0.769), azithromycin (p = 0.705) and hydroxychloroquine (p = 0.386). The variables "chronic kidney disease" and "congestive heart failure" were eliminated from the adjusted model due to a significant percentage of missing values and the variable "ivermectin" was eliminated due to collinearity with hydroxychloroquine.

DISCUSSION
The risk factors identified with admission to the ICU or death were age over 40 years, male sex, admission with oxygen saturation less than 80%, admission with respiratory rate > 22, or diabetes mellitus. No significant association was found between adverse outcome and obesity or hypertension. As secondary findings, 18.4% of patients were admitted to the ICU and 29.9% died; 30.1% were admitted with saturation < 85%. The mortality we found was lower (29.9%) than that reported for other hospitals, 49% (3) and 46.4% (18) ; however, these studies would not be comparable as they are from different contexts, characteristics and times of the pandemic in the country.
Nonetheless, our study reports characteristics and risk factors similar to those described in other cohorts (3,13,15,(19)(20)(21) . We observed an association between age over 40 years and the adverse outcome, which is also evident in other studies (9,10,13,23) , possibly due to the increased prevalence of chronic diseases in this age group. It could also be due to vulnerability to infections, explained by immunosenescence (9) .
In this study we observed an association between male sex and outcome. This association has been described in studies carried out in cities at sea level (3,8,10) . It has been described that biological mechanisms, such as a decreased immune response in men versus women may be involved (23) . It has been suggested that sex hormones, such as androgens, could play an important role in regulating proteins such as transmembrane serine protease 2 (TMPRSS2), which is involved in viral entry and spread, resulting in more severe forms of COVID-19 (24) .
We identified that having an admission oxygen saturation < 80% was a risk factor, this value differs from the cut-off point proposed in other similar studies performed in populations at sea level (3,15,18) and in populations with a lower altitude than Puno (19) . This could be explained by the low oxygen pressure, since the saturation described in patients at high altitude without any pathology is lower, compared to that of residents at sea level (17) , so it would be expected that in patients with COVID-19, the saturation would be even lower since it is a pathology that adds hypoxemia. It is difficult to establish a cut-off point in high altitude populations. An inversely proportional relationship between oxygen saturation and altitude level has been previously described (17) . More studies are needed to quantify this relationship and determine cut-off points.
Likewise, we found a significant association between tachypnea (>22 bpm) and the adverse outcome. This relationship is understandable because an increasing respiratory rate is the physiological response to hypoxemia and acidosis (25) . Again, it should be noted that more studies evaluating tachypnea in high-altitude conditions are needed to determine the cutoff points for poor prognosis in this population.
Regarding comorbidities, we found an association between the adverse outcome and diabetes mellitus (DM),

REFERENCES
which is similar in other studies (11)(12)(13)(14) , since DM involves immunosuppression due to alterations in the immune response that include increased production of proinflammatory cytokines, which add inflammation to the patient with CO-VID-19 (26) . However, no association was found between the adverse outcome and obesity, as has been found in other studies (20,22) where the outcome was mortality and, like DM, obesity involves a proinflammatory state due to cytokines (TNF-alpha, MCP-1 and IL-6) produced by adipose tissue (12) . On the other hand, although studies report an association between mortality and AHT (3,11,15) , no association was found between AHT and adverse outcome in the adjusted analysis. It has been proposed that there is a higher expression of ACE in hypertensive patients due to treatment with ACE inhibitors, which could increase the risk of severe forms of COVID-19 (14) .
We did not find statistically significant associations in the laboratory results; however, there is literature describing that severe patients in high altitude areas show a decrease of the platelet count.
Finally, no significant association was found between adverse outcome and receiving anticoagulants (enoxaparin) either in prophylactic dose (40 mg) or full dose (60 mg), nor with dexamethasone, azithromycin or hydroxychloroquine. Regarding hydroxychloroquine, there are studies that show that there is no benefit in reducing mortality compared to those who do not receive it (27,28) . On the contrary, there is literature that supports the association between the use of low-dose dexamethasone and a reduction in mortality in patients with severe COVID-19 (29) .
The main limitation of this study is the use of the rapid test, due to its low diagnostic sensitivity. However, it proved to be the most accessible resource to define the cases admitted to the COVID area. This screening method was used to confirm or rule out cases according to the Ministerial Resolution 193-2020; because diagnosis was based on epidemiological history and clinical characteristics of the patient. On the other hand, the laboratory for molecular tests was available since August 2020, before this date the tests were neral-s-opening-remarks-at-the-media-briefing-on-covid-19---11march-2020. Despite the limitations, this is one of the few studies that explores the relationship between saturation and adverse outcomes in high-altitude populations. In addition, our study identifies different factors that possibly influence the prognosis of the high-altitude resident with COVID-19.
In conclusion, in the Manuel Núñez Butron Regional Hospital, the risk factors for an adverse outcome (admission to the ICU or mortality) due to COVID-19 were male sex, age over 40 years, low saturation (< 80%) on admission, tachypnea (> 22 bpm) on admission, and diabetes.
The definition of admission for patients should be improved for future studies. On the other hand, it would be relevant to define temporal contexts because epidemiological surveillance in Peru determined the existence of two waves of COVID-19 cases; in addition, there are variants of the SARS-CoV-2 virus that appeared over time, as well as vaccines from different laboratories (Pfizer, Astrazeneca, Sinopharm). These variables could be included when replicating this study.
COVID-19 represents a huge public health burden globally and nationally, and identification of risk factors at hospital admission in high-altitude populations could help to take early measures that contribute to clinical improvement of the patients. These findings should be confirmed in future studies.
Authors contribution: NA, GAM and GA conceived the research idea, conceptualized, wrote the article, and carried out the data analysis and interpretation. NA, GAM and FM were responsible for data collection. All authors approved the final version of the article, as well as the critical review of the article and assume responsibility for the article.

Funding:
The study was self-funded.

Conflict of interest:
The authors deny having any conflict of interest.