Mortality risk of COVID-19 in elderly males with comorbidities: a multi-country study

The COVID-19 pandemic causes severe morbidity and mortality. This multi-country study aimed to explore risk factors that drive mortality in COVID-19 patients who received neither dexamethasone nor remdesivir. We analyzed a cohort of 568 survivors and 507 non-survivors from China, European regions, and North America. Elderly males ≥70 years accounted for only 25% of survivors, but this rate was significantly higher in non-survivors from China (55%), European regions (63%), and North America (47%). Compared with survivors, non-survivors had more incidences of comorbidities such as cerebrovascular disease and chronic obstructive pulmonary disease (COPD, p-values<0.05). Survival analyses revealed age, male gender, shortness of breath, cerebrovascular disease, and COPD as mortality-associated factors. Survival time from symptom onset was significantly shorter in elderly versus young patients (median: 29 versus 62 days), males versus females (median: 46 versus 59 days), and patients with versus without comorbidities (mean: 41 versus 61 days). Mortality risk was higher in elderly males with comorbidities than in young females without comorbidities (p-value<0.01). Elderly male survivors with comorbidities also had longer hospital stays than other survivors (25 versus 18.5 days, p-value<0.01). Overall, the high mortality risk in elderly males with COVID-19-associated comorbidities supports early prevention and critical care for elderly populations.


INTRODUCTION
Coronavirus Disease 2019 (COVID-19) is a global pandemic that is causing significant, widespread increases in morbidity and mortality, while effective antiviral treatments are still under development [1]. As of November 26 th , 2020, there were more than 60.4 million cases of COVID-19, and the overall casefatality risk was approximately 2.4% according to an update by the World Health Organization (WHO). With a strong potential of sustained human-to-human transmission, severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) can quickly spread to vulnerable populations such as elderly individuals with aging-related immune disorders.
Early studies of COVID-19-associated mortality and risk factors have mostly involved small cohorts of non-survivors in Wuhan [2][3][4][5][6][7]. For instance, COVID-19-associated mortality was observed in 32 (62%) of 52 critically ill patients in Wuhan who received medical care in the intensive care unit [2]. A singlecenter study reported an increased risk of mortality in COVID-19 patients who were severely or critically ill at hospitalization [3]. Additionally, mortality was observed in both non-severe (1.1%) and severe (32.5%) patients during the 32-day follow-up [4]. A retrospective study of 54 non-survivors showed the increasing odds of death in association with older age, a high sequential organ failure assessment score, and high d-dimer levels at baseline [5]. Another study in Wuhan reported strong associations of older ages with a high risk of acute respiratory distress syndrome and death [6]. Moreover, CD3 + CD8 + T cells and cardiac troponin I were found to be important risk factors based on a small cohort of 21 non-survivors and 158 survivors [7]. Despite the above findings, it remains unclear why some patients are more susceptible to a fatal outcome of COVID-19 and whether elderly patients outside Wuhan harbor unique features with regard to mortality.
To reveal mortality-associated factors, our study collected a large cohort of survivors who recovered from COVID-19 and non-survivors who died of the disease before May 1, 2020. Due to the authorization of remdesivir on May 1 and dexamethasone on June 17 that may affect our results of mortality-related factors, our data collection initially retrieved clinical records of COVID-19 patients who received neither remdesivir nor dexamethasone. By using epidemiological and clinical records of COVID-19 patients from different countries, this multi-country study revealed a high-risk population of elderly patients with specific comorbidities. Given the rapid spread of COVID-19 which is causing high morbidity and mortality worldwide, a special focus on elderly patients with specific comorbidities in current COVID-19 guidelines should be implemented.

Epidemiological features
This study retrieved a multi-country dataset of 1075 COVID-19 patients, including 232 survivors and 183 non-survivors in China, 208 survivors and 258 nonsurvivors in European regions, and 128 survivors and 66 non-survivors in North America ( Table 1). None of the COVID-19 patients received remdesivir or dexamethasonetwo approved drugs that may potentially reduce the mortality rate of COVID-19 [8,9]. All COVID-19 patients were reported between January and April 2020. Our workflow and data summary are provided in Figure 1.

Clinical features of survivors and non-survivors
We analyzed the proportions of those with pre-existing comorbidities in a subset of 359 survivors and 239 nonsurvivors whose comorbidities were recorded ( Figure  1). As expected, proportions of comorbidities increased with age in males and females, though comorbidity proportions were slightly higher (but insignificant) in males than in females (Figure 3A and Supplementary  Table 1). Moreover, non-survivors had higher proportions of comorbidities than survivors aged from AGING  Figure 3C).
We next analyzed baseline findings of computed tomography (CT) scans. Incidences of ground-glass opacity, pleural adhesions, and pleural effusion were significantly higher in non-survivors than in survivors (p-values<0.05). Figure 4 highlights the CT results of a non-survivor who had abnormal bilateral lungs with ground-glass opacities and pleural adhesions.

Risk factors associated with COVID-19 mortality
To reveal mortality-associated factors, we examined a subset of 242 survivors and 157 non-survivors whose records contained complete information regarding age, gender, comorbidities, symptoms, clinical outcomes, and survival time from symptom onset to clinical outcomes ( Figure 1). As shown in Table 2, the most common symptoms at baseline were fever, cough, shortness of breath, and fatigue (proportions ≥25%). In the comparison of survivors and non-survivors, significant differences in patient age, male gender, shortness of breath, any comorbidity, cerebrovascular disease, cardiovascular disease, chronic liver disease, and COPD were detected (  Table 3). These five factors with the effect size of Cohen's d ≥0.4 remained significant in the propensity score-matched samples that potentially reduced the confounding effect (Table 4).
A tree model was built in Figure 5 to visualize proportions of survivors and non-survivors based on 8 possible conditions of three pre-existing factors at baseline: age (<70 or ≥70 years), gender (male or female), and any comorbidity (yes or no). Among the 8 possible conditions, the highest proportion of mortality was observed in elderly males with comorbidities (28%, 44/157), whereas the highest proportion of survivors was found in young females without comorbidities (18%, 44/242). A higher proportion of non-survivors was often found among males, those ≥70 years, and/or those with comorbidities than among females, those <70 years, and/or those without comorbidity ( Figure 5). Similar patterns were also observed in the tree models of age, gender plus either cerebrovascular disease or COPD (Supplementary Figure 1).

Impact of risk factors on survival time and length of hospital stay
Kaplan-Meier curves were evaluated to understand whether risk factors such as age, gender, and comorbidities affect survival time from symptom onset to death -the primary endpoint of clinical outcomes. As shown in Figure 6A, the median survival time of elderly patients ≥70 years was significantly shorter than that of www.aging-us.com   Figure 6B). The survival time of patients with comorbidities was shorter than those without any comorbidity (mean: 41 versus 61 days, p-value=0.001, Figure 6C). Furthermore, the survival time of patients with ≥3 comorbidities (mean: 27, IQR: 20 to 33 days) was shorter than those with 1 or 2 comorbidities (mean: 43, IQR: 40 to 46.5 days) and those without any comorbidity (mean: 60, IQR: 55 to 66 days, p-values<0.01, Supplementary Figure 2A). Patients with AGING cerebrovascular disease (median: 14 versus 59 days, p-value<0.001, Figure 6D) or COPD (median: 11 versus 59 days, p-value<0.001, Supplementary Figure 2B) also had a shorter survival time than patients without these comorbidities.
We next evaluated the length of hospital stay under preexisting conditions of age, gender, and comorbidities ( Figure 6E). The length of hospital stay was shorter among non-survivors (median: 11, interquartile range (IQR): 7 to 21 days) than among survivors (median: 20,

DISCUSSION
To understand why some COVID-19 patients are more susceptible to fatal outcomes, we explored epidemiological and clinical records of survivors and non-survivors from China, European regions, and North America. Our study revealed three major findings. (i) High mortality risk of COVID-19 was consistently found among elderly males from China, Europe, and North America. (ii) Risk factor and survival analyses revealed mortality-associated risk factors of age, male gender, shortness of breath, and chronic comorbidities such as cerebrovascular disease and COPD. (iii) Mortality-associated risk factors exert an impact on the survival time and hospital stays of elderly patients. To reduce mortality rates and hospital burden, we carried out a multi-country study on COVID-19-induced mortality, and the results support the early prevention and critical care of elderly males with specific comorbidities.

Elderly adults with an increased risk of COVID-19
In agreement with previous findings [2,4,10,11], our survival analyses suggested that the mortality risk of COVID-19 increased with patient age (HR: 1.04 increase per year), and most deaths were observed in elderly adults, especially those ≥70 years (>50% of nonsurvivors). Of interest, the high mortality risk of elderly adults was consistently observed in China, European regions, and North America ( Figure 2). Indeed, similar findings have been consistently reported by singlecountry studies of COVID-19 patients in many countries such as America [12], Belgium [13], Canada [14], France [15], Germany [16], Italy [17], Mexico [18], Poland [19], Romania [20], Russia [21], Spain [22], and the United Kingdom [23]. Of note, age, gender, and chronic comorbidities are key factors that determine hospitalization rather than outpatient care [10]. For instance, elderly survivors required a longer length of hospital stay than non-survivors ( Figure 6). The elderly required longer hospital stays than young adults, probably because close monitoring of their recovery requires more time in the hospital. Similar to our findings, a small-cohort study of 17 non-survivors reported a short time (median: 11.5 days) from symptom onset to death in patients ≥70 years [24]. Overall, our study and literature findings support the early prevention of COVID-19 in elderly adults due to their high risk.
Why are elderly adults more vulnerable to COVID-19? Although the exact mechanisms remain unclear, several hypotheses have been proposed [25]. First, compared with young patients, elderly adults with compromised immunity might have slower, less coordinated, and less efficient immune responses to COVID-19 [26,27]. Additionally, the innate and adaptive systems are possibly suboptimal in elderly patients. For instance, T and B lymphocytes are critical components of adaptive immune responses to emerging infections, but elderly patients may experience age-related dysfunction with a decreased production of T and B lymphocytes [28]. Second, elderly adults may have an increased risk of exposure to COVID-19, problems with access to health services, and less opportunity to receive respiratory support in resource-limited settings [29]. Third, elderly adults usually have comorbid conditions such as acute respiratory distress syndrome, which may diminish their recovery from COVID-19 [30].
Why does COVID-19 kill more males than females? Although the exact mechanisms remain unknown, several hypotheses have been proposed to explain gender disparities. (i) The X chromosome encodes the human receptor angiotensin-converting enzyme 2 (ACE2, gene ID: 59272), which is the human receptor mediating SARS-CoV-2 entry into human cells. Expression of ACE2 might vary in different populations due to genetic differences [34]. (ii) Males and females have different immune responses to COVID-19 due to the sex chromosomes (XY in males, XX in females), sex hormones (androgens, estrogens), and regulatory genes associated with the immune system [35]. Since the X chromosome encodes the largest number of immunerelated genes in the human genome, females carrying two copies of the X chromosome may have immunological advantages regarding resistance to viral pathogens, but they also develop more autoimmune diseases [36,37].
(iii) Different socioeconomic and lifestyle behaviors such as smoking, alcohol consumption, and personal hygiene may also play a role in gender differences in susceptibility to COVID-19 [38].

Comorbidities increase the risk of COVID-19
In agreement with previous studies [39][40][41], we observed a higher proportion of comorbidities (e.g. cerebrovascular disease, cardiovascular disease, COPD) in non-survivors than in survivors (Table 2). Furthermore, pre-existing comorbidities may reduce the survival time in non-survivors and increase the length of hospital stay of survivors ( Figure 6). Similar to our results in Table 2, severely ill patients had more comorbidities, including hypertension (56%), heart disease (21%), diabetes (18%), cerebrovascular disease (12%), and cancer (7%) [42]. Comorbidities were also observed in 94% of 5700 COVID-19 patients hospitalized in New York [43]. Our study showed slightly higher (but insignificant) proportions of comorbidities such as diabetes mellitus, cardiovascular disease, cerebrovascular disease, chronic bronchitis, and COPD in males than females (Supplementary Table 1). Multisystem organ dysfunction in males might be associated with a high rate of mortality [32].
What are the impacts of comorbidities on the morbidity and mortality of COVID-19? Although the possible mechanisms are still under investigation, answers likely depend on the type of comorbidity. First, our analyses revealed that the comorbidities of cerebrovascular disease and COPD are time-dependent factors associated with a fatal outcome of COVID-19 (Tables 3, 4). Of interest, cerebrovascular disease was also recognized as a risk factor in a retrospective study of 50 non-survivors and 1540 survivors [40]. Furthermore, a meta-analysis reported an approximately 2.5-fold increase in the odds of severe disease in the presence of cerebrovascular disease [44]. Pre-existing COPD was previously confirmed as a key factor that contributes to the worse progression and outcome of COVID-19 [45]. Due to SARS-CoV-2 invasion of the central nervous system, severely ill patients are more likely to develop neurologic manifestations, especially acute cerebrovascular disease [46]. Second, possible crosstalk between diabetes and COVID-19 might be involved with ACE2 and dipeptidyl peptidase-4, two important human proteins in the biological pathways of both diseases [47,48]. Third, cardiovascular disease involves an increased risk of inhospital death in COVID-19, which might be mediated by ACE2-dependent myocardial infection [49] or the high inflammatory burden that can induce vascular inflammation, myocarditis, and cardiac arrhythmias [50]. Fourth, patients with lung cancer, hematological cancer, or cancers in metastatic stages have higher rates of severe outcomes than patients without cancer [51]. Overall, comorbidities are likely associated with the severity of COVID-19 [39,41].

Limitations
First, we initially intended to collect all available online records of COVID-19 cases worldwide, but detailed medical records of individual patients were mostly lacking probably because the privacy of patient data is protected and detailed records are often not published in an emergent outbreak. Nevertheless, our major findings of elderly males at high risk for COVID-19 have been consistently confirmed by single-country studies [12][13][14][15][16][17][18][19][20][21][22][23]. Future multi-country studies are yet needed to retrieve complete records worldwide and characterize COVID-19 mortality from a global perspective. Second, our retrospective study describes potential associations of risk factors with mortality, but crosstalk between human and viral proteins in the lifecycle of SARS-CoV-2 needs to be clarified by future studies. Third, the role of treatments in fatal outcomes of COVID-19 was not analyzed because multiple treatments (e.g. antiviral/antibiotic/antifungal therapies, ventilation, oxygen therapy) are often administered. The possible role of prevention strategies in reducing mortality was not analyzed due to limited data. Moreover, controlled cohorts were not available for our study. Randomized controlled cohorts are thus needed to assess the use of potential treatments in elderly patients with COVID-19.

CONCLUSIONS
Given the rapid spread of COVID-19 worldwide, a better understanding of high-risk populations is important for decision making in the treatment and prevention of the disease. The clinical features of nonsurvivors inside and outside Wuhan are similar [28], supporting universal strategies to handle COVID-19 across different cities and countries. As of today, there is no cure to eliminate COVID-19 so that prevention strategies are critical for reducing COVID-19 infections. In addition to prevention measures (e.g. face masks, good hygiene, social distancing, contract tracing, travel restrictions, lockdowns, and quarantine) that limit the spread of COVID-19, intensive surveillance and hospital/outpatient caregiving should be adapted for elderly patients with COVID-19. Current guidelines should also pay more attention to high-risk populations of elderly patients with COVID-19-associated comorbidities, complications, and polypharmacy.

Data collection of COVID-19 patients
Our workflow, as shown in Figure 1, was conducted to collect clinical records for COVID-19 patients in China, North America, and European regions. First, we collected epidemiological and clinical records of COVID-19 patients who were hospitalized before May 1 st , 2020, in the Sino-French New-City Tongji Hospital in Wuhan. At hospital admission, all patients were diagnosed with COVID-19. Nasal and pharyngeal swab specimens were collected to assess the presence of SARS-CoV-2 using real-time RT-PCR tests according to WHO interim guidelines. All survivors, who fulfilled discharge criteria, were discharged for 14-day home quarantine based on the New Coronavirus Diagnosis and Treatment Guidelines in China (Supplementary Method 1). A total of 232 survivors and 60 non-survivors were included in our dataset. Second, to increase the sample size, we extracted clinical records for all COVID-19 non-survivors from The First Hospital of Changsha in China. Third, we attempted to extract online records for COVID-19 individuals in China, European regions, and North America. Online collections were conducted using a four-fold procedure. (i) We searched online records of COVID-19 patients in China, 15 countries in North America, and 43 countries in Europe (Supplementary  Table 4) where a large number of COVID-19 patients were reported between January and April 2020. (ii) We searched records of COVID-19 individuals in the scientific literature, official media/conferences, hospital websites, government (state, provincial, national) websites, and Centers for Disease Control and Prevention. Only individual cases with epidemiological and clinical information were retrieved, while clinical studies or meta-analyses that shared no individual records were not considered. (iii) We used keywords such as: "COVID", "SARS-CoV-2", "2019-nCoV", "new coronavirus", "mortality", "death" and "survivor" to screen online data. (v) Duplicates were removed and all records were cross-checked by at least two investigators.
The data extraction (as described above) retrieved a total of 1267 patients, including 666 survivors and 601 nonsurvivors. Because remdesivir and dexamethasone significantly reduce COVID-19 mortality [8,9], our analyses excluded patients who received remdesivir (N=8) and/or dexamethasone (N=73). Moreover, 42 patients were excluded because they had no treatment records and were hospitalized after May 1, 2020. Of 59 countries in our search list, 35 had fewer than five extracted records in our dataset, which is unlikely to be representative; therefore, 73 cases from these countries were excluded. Supplementary Table 4 summarizes the country list and case numbers in 59 countries. For data transparency, the compiled online resources of 981 COVID-19 patients are shared in Supplementary Data 1.
A total of 568 survivors and 507 non-survivors with basic information regarding age, gender, country, and clinical outcomes were included in our age-gender analysis (Table 1 and Figure 2). Additional information on baseline symptoms and comorbidities was also recorded in a subset of 359 survivors and 239 nonsurvivors, which was used for comorbidity analyses (Figure 3). Survival time, as defined by the days from symptom onset of clinical outcomes, was identified in a subset of 242 survivors and 157 non-survivors and, subsequently used for risk-factor analyses and survival analyses (Tables 2-4 and Figures 5, 6). Figure 1 summarizes our workflow and data structures.

Data management
Our datasets included demographic features (age, gender, cities), signs and symptoms, pre-existing comorbidities, complications before death, and other available data (e.g. symptom onset date, admission date, computerized tomography data) for COVID-19 survivors and nonsurvivors. Symptoms, comorbidities, and complications were treated as categorical variables (yes or no). Clinical outcomes were defined by the status of death (nonsurvivors) or hospital discharge after the clearance of COVID-19 (survivors). For non-survivors, the survival time was defined by days from symptom onset to the date of death. Since death was not observed before hospital discharge, survivors were treated as censored data in which the observation time was defined by days from symptom onset to hospital discharge.

Statistical analysis
In comparisons of COVID-19 survivors and nonsurvivors, continuous variables were analyzed using Mann-Whitney U tests, while categorical variables were analyzed using χ 2 tests or Fisher's exact tests as appropriate. Survival analyses were performed using Cox proportional hazards models. Hazard ratios with the 95% confidence interval were assessed to reveal the effects of risk factors on survival-time outcomes: either death for non-survivors or hospital discharge for survivors. Kaplan-Meier curves were built to show survival time among patient groups (e.g. elderly versus young patients), and significant differences were examined by log-rank tests. A common approach called pairwise deletion was applied to handle missing data. Since no random sampling was conducted, all statistical analyses were descriptive. To reduce the confounding effect in this nonrandomized study, the propensity-score matching method in SPSS V26.0 was applied with default settings (e.g. one-to-one matching, matching tolerance: 0.02) to assess whether any observed confounding was adequately reduced. The effect size of risk factors was estimated by Cohen's d [52]. In all significance tests, two-tailed tests were performed and P-values <0.05 were considered statistically significant.

Ethics declarations
This retrospective study was performed in accordance with the Helsinki Declaration and was approved by the Ethics Committees of The Second Xiangya Hospital, Central South University (ID: LYF2020060), and The First Hospital of Changsha (ID: KX2020002). Written informed consent was waived for analyzing archived records in this retrospective study.

Availability of clinical data
Supplementary Data 1 shares online records of COVID-19 patients.

AUTHOR CONTRIBUTIONS
GL, YL, and YW performed statistical analyses and drafted the manuscript; YL, XJ, MM, and YW performed data acquisition of non-survivors; MZ, JL, and PJ performed data acquisition of survivors; ZZ, YX, YH, JL, and GG conducted clinical practice and data interpretation; YTH, NRF, and EDC contributed with discussions and writing of the manuscript; GL and MZ supervised the study, obtained funding, and revised the manuscript. All authors contributed to the final article.