Prevalence and predictors of death and severe disease in patients hospitalized due to COVID-19: A comprehensive systematic review and meta-analysis of 77 studies and 38,000 patients

Introduction Progression of COVID-19 to severe disease and death is insufficiently understood. Objective Summarize the prevalence of risk factors and adverse outcomes and determine their associations in COVID-19 patients who were hospitalized. Methods We searched Medline, Embase and Web of Science for case-series and observational studies of hospitalized COVID-19 patients through August 31, 2020. Data were analyzed by fixed-effects meta-analysis using Shore’s adjusted confidence intervals to address heterogeneity. Results Seventy-seven studies comprising 38906 hospitalized patients met inclusion criteria; 21468 from the US-Europe and 9740 from China. Overall prevalence of death [% (95% CI)] from COVID-19 was 20% (18–23%); 23% (19–27%) in the US and Europe and 11% (7–16%) for China. Of those that died, 85% were aged≥60 years, 66% were males, and 66%, 44%, 39%, 37%, and 27% had hypertension, smoking history, diabetes, heart disease, and chronic kidney disease (CKD), respectively. The case fatality risk [%(95% CI)] were 52% (46–60) for heart disease, 51% (43–59) for COPD, 48% (37–63) for chronic kidney disease (CKD), 39% for chronic liver disease (CLD), 28% (23–36%) for hypertension, and 24% (17–33%) for diabetes. Summary relative risk (sRR) of death were higher for age≥60 years [sRR = 3.6; 95% CI: 3.0–4.4], males [1.3; 1.2–1.4], smoking history [1.3; 1.1–1.6], COPD [1.7; 1.4–2.0], hypertension [1.8; 1.6–2.0], diabetes [1.5; 1.4–1.7], heart disease [2.1; 1.8–2.4], CKD [2.5; 2.1–3.0]. The prevalence of hypertension (55%), diabetes (33%), smoking history (23%) and heart disease (17%) among the COVID-19 hospitalized patients in the US were substantially higher than that of the general US population, suggesting increased susceptibility to infection or disease progression for the individuals with comorbidities. Conclusions Public health screening for COVID-19 can be prioritized based on risk-groups. Appropriately addressing the modifiable risk factors such as smoking, hypertension, and diabetes could reduce morbidity and mortality due to COVID-19; public messaging can be accordingly adapted.


Introduction
Coronavirus disease-19  caused by severe acute respiratory syndrome-coronavirus-2 (SARS-CoV-2) that first emerged in Wuhan, China in late December 2019 has spread with such rapidity and efficiency that in less than 10 months, it has caused more than 36 million cases and million deaths globally [1]. Driven by an urgency to solve the crisis, studies are being published at an unprecedented pace. However, across the publications, prevalence of death, severe disease and their association with epidemiological risk factors have greatly varied [2,3], with studies showing conflicting results for association of key risk factors such as sex [4][5][6][7][8], smoking [9][10][11][12], hypertension [4,7,8,13,14] and diabetes [4,7,8,13,14] with COVID-19 disease severity and death. Whether or how cardiovascular risk factors, especially prior hypertension, diabetes and heart disease are associated with acquisition of SARS-CoV-2 and progression to severe disease or death is not understood well [15][16][17][18]. Meta-analyses conducted so far on prevalence of epidemiological risk factors and association with disease progression were mostly based on studies from China [9,11,[18][19][20] and many of the analyses on prevalence estimates included studies focused on critically ill patients [9,19], which can overestimate the prevalence and affect generalizability of results. To our knowledge, none of the analyses were restricted to hospitalized COVID-19 patients. Restricting our analysis to hospitalized patients provides an efficient sampling frame to investigate disease progression in relation to risk factors. Therefore, we undertook a comprehensive systematic review and meta-analysis to investigate the association between key epidemiological factors-age, gender, smoking, hypertension, diabetes, heart disease, chronic obstructive pulmonary disease (COPD), chronic kidney disease (CKD) and chronic liver disease (CLD)-and progression to death and severe disease in patients hospitalized due to COVID-19. We additionally compared the 1) the prevalence of risk factors and death in the US-Europe with that of China; 2) the prevalence of co-morbidities at baseline with the general population prevalence, and 3) prevalence of cardiovascular disease, COPD and CKD at baseline with corresponding organ injuries (acute cardiac injury, acute lung injury, and acute kidney injury) during hospital admission.

Literature search, study selection and data abstraction
We searched Medline, Embase, Web of Science and the WHO COVID-19 database to identify studies published through August 31, 2020 that investigated the risk of severe disease or death in hospitalized patients with confirmed COVID-19 disease. We used search terms, 'coronavirus disease 19', 'COVID-19', 'severe acute respiratory syndrome coronavirus 2' and 'SARS-CoV-2' for COVID-19 and the string ((characteristics) OR (risk factors) OR (epidemiology) OR (prevalence) OR (intensive care) OR (ventilator) OR (mechanical ventilator) OR (mortality) OR (survivor � ) OR (smoking) OR (smoker � )) AND ((COVID-19) OR (COVID) OR (coronavirus)) for studies published between December 15, 2019 and August 31, 2020. We started the search on March 18, 2020 with biweekly search thereafter and final search on August 31, 2020. We included case series and observational studies that described the prevalence of death or severe disease in adult population stratified by risk factors: age, sex, hypertension, diabetes, heart disease, COPD, CKD and CLD. We excluded studies that included nonconsecutive patients or exclusively focused on pregnant women, children, and elderly patients. We excluded studies that exclusively studied critically ill patients from calculation of prevalence of death but included them for calculating the association of risk factors with death. Screening of abstracts and full-text reviews were conducted using Covidence (Melbourne, Australia).

Risk factors and outcomes
Primary outcomes were prevalence of death and association of risk factors with death. We extracted data on death as recorded in the publications. We measured prevalence of severe disease and association with risk factors as secondary outcomes. We defined outcome as severe disease for any of 1) the study classified COVID-19 disease as severe or critical, 2) intensive care unit (ICU) admission, 3) acute respiratory distress syndrome, or 4) mechanical ventilation. Severe disease was defined by studies as respiratory rate�30 per minute, oxygen saturation�93%, and PaO 2 /FiO 2 <300 and/or lung infiltrates>50% within 24-48 hours [3]. Critical illness was defined as respiratory failure, shock and/or multiple organ dysfunction or failure [3]. Heart disease as a pre-existing condition was broadly defined by most studies as 'cardiovascular disease' (CVD). Additional outcomes were acute cardiac and kidney injury in the hospitalized patients that were defined as such by the studies.

Statistical analysis
We calculated and reported summary estimates from fixed-effects models [21]. We assessed heterogeneity across studies using Cochran's Q-test (χ 2 p value <0.10) [22] and I 2 statistics (I 2 >30%) [23]. In the presence of heterogeneity, we adjusted the confidence intervals for between-study heterogeneity using the method described by Shore et al. [24]. We presented the results from random effects meta-analysis as well. The meta-analysis was performed in Microsoft 1 Excel 2020 (Microsoft Corporation, Redmond, WA). We analyzed publication bias using funnel plots and Egger's tests. Quality of each study was assessed using the Newcastle-Ottawa assessment scales using the PRISMA guidelines. We calculated the following as a part of our analyses: 1) prevalence of severe disease or death, 2) prevalence of risk factors, and 3) relative risk for the association of age, sex, and comorbidities with outcome. When not reported or when unadjusted odds ratio was presented, we calculated the relative risk (95% CI) using the frequencies provided. Adjusted estimates were used where available. Case fatality risk (and case severity risk) for a specific risk factor was calculated as number of deaths (or severe disease) in patients with a risk factor out of all patients possessing the risk factor.

Prevalence of death and severe disease
We calculated an overall prevalence of death of 20% [95% CI: 18-23%; I 2 = 96%; n = 60], ranging from 1% to 38% across the studies, and of severe disease of 28% [95% CI: 24-33%; I 2 = 98%; n = 60] for all patients hospitalized due to COVID-19 (Tables 2 and 3). Data on prevalence of death, severe disease, and risk factors (S1 Table), and association of the risk factors with death (S2 Table) and severe disease (S3 Table) for the individual studies are presented in the supplemental tables.
Predictors of death and severe disease (Tables 2 and 3 (Fig 2A).

COVID-19 related organ system injury
To understand how pre-existing health conditions may be correlated with the risk of specific organ injury, we calculated the prevalence of acute injury to lung, heart and kidney for studies that reported prevalence of both the pre-existing condition(s) and corresponding organ injury (Fig 4A). Pooled across 12 studies [14,25,32,45,48,49,52,54,60,62,79]

Regional difference in prevalence of death and risk factors
Upon sub-group analysis, we noted significantly higher prevalence of death and risk factors among COVID-19 patients in the US and Europe than in China (Fig 4B).

Comorbidities in COVID-19 patients and the general populations in the US and China
In order to gain some understanding of whether patients with comorbidities are at higher risk of COVID-19 infection or hospitalization, we compared the prevalence of comorbidities between COVID-19 patients hospitalized in the US and the prevalence of comorbidities in the general US population. We observed that the prevalence of hypertension (55%), diabetes (33%), CVD (17%), and smoking history (23%) were substantially higher in the COVID-19 patients than in the general US population (Fig 4C). For the Chinese population, the overall prevalence of hypertension (23%) and diabetes (12%) in the COVID-19 patients were similar to that of the general Chinese population. However, the prevalence of smoking history (11%), COPD (4%), CKD (2%), and heart disease (16%) in the COVID-19 patients hospitalized in China were unexpectedly lower as compared to their corresponding prevalence in the general Chinese population (Fig 4D).

Sensitivity analyses
The positive associations of age�65 years, male sex, smoking history, COPD, hypertension and diabetes with the risk of death in the COVID-19 patients were relatively homogenous (I 2 <70%). However, we carried out sensitivity analyses to assess the effects of outliers. For the risk of death for hypertension and smoking history, we removed the study by Yao et al. [86] which showed significantly higher risk compared to other studies; the results for both hypertension [sRR = 1.74; 95% CI: 1.58-1.94] and smoking [sRR:1.24; 95% CI: 1.08-1.42] remained significant. Guan et al. [13] had published a second study with additional patients and reported adjusted estimates for COPD, diabetes and hypertension. We used the adjusted risk estimates for the analyses. For the risk of death with other risk factors (CVD, CKD, and CLD) for Guan et al. [45], we conducted sensitivity analyses by using the counts only from the original study.

Small study effects and quality assessment
We observed asymmetry in the funnel plot for studies that reported prevalence of death in COVID-19 patients (Egger's test p = 0.001) (S1 Fig). On further analysis, the plot remained asymmetrical when restricted to studies from China (Egger's p = 0.003) but was symmetrical for studies from US-Europe (Egger's p = 0.160). We observed symmetrical funnel plots with no bias for pooled prevalence severe disease (Egger's p = 0.128). On average, prospective or retrospective studies scored a score of 6 out of 9 and cross-sectional studies scored 6 out of 10. Many studies did not get a full score because they did not adjust for confounders (age, sex, or other risk factors) or patients remained hospitalized even after the follow-up ended, suggesting inadequate follow-up period (S4 Table).

Discussion
We carried out a comprehensive systematic review and meta-analysis of 77 studies that included 38906 hospitalized patients to investigate the prevalence and risk factors for death and severe disease in COVID-19 patients. We calculated an overall prevalence of death of 20% and severe disease of 28%. Nearly 50% of the patients admitted to hospitals due to COVID-19 were �60 years of age and 59% were males. We observed high prevalence of hypertension and diabetes of 50% and 28%, respectively, for the patients. The risk factors were more prevalent in patients who died and were distributed as: age �60 years: 85%; males: 66%; hypertension: 66%; diabetes: 39%; heart disease: 37%; CKD: 27%; smoking history: 44%; COPD: 12%, and CLD: 9%. In comparison with the overall prevalence of death of 20% for all COVID-19 hospitalized patients, the CFR was higher for male patients (26%) and for patients having the following risk factors: age�60 years (35%), heart disease (52%), COPD (51%), CKD (48%), CLD (39%), hypertension (28%), diabetes (24%), and smoking history (27%). The elevation in the risk of death was statistically significant for age �60 (sRR = 3.6; 95% CI: 3.0-4.4), male sex 1.3 (95% CI: 1.2-1.4), smoking history (sRR = 1.3; 95% CI: 1.1-1.6), COPD (sRR = 1.7; 95% CI: 1.4-2.0), heart disease (sRR = 2.1; 95% CI: 1.8-2.4), CKD (sRR = 2.5; 95% CI: 2.1-3.0), hypertension (sRR = 1.8; 95% CI: 1.7-2.1), and diabetes (sRR = 1.5; 95% CI: 1.4-1.7). All of the risk factors we analyzed were positively associated with progression to severe disease as well. The results suggest that older age, male sex and the co-morbidities increase the risk of progression to severe disease and death in COVID-19 patients. We observed significant difference in the prevalence of death between US-Europe (23%) and China (11%). This lower risk of death from COVID-19 for the hospitalized patients in China may be explained by the lower median age as well as lower prevalence of co-morbidities for COVID-19 patients in China. However, this >200% lower prevalence of death in China is incommensurate with our finding of a higher prevalence of severe disease observed for patients in China (39%) as compared to patients in the US-Europe (20%). Notably, we observed asymmetry in the funnel plot and a statistically significant tests for publication bias or small study effects for the prevalence of death for studies from China that could suggest selective outcome reporting. As such, while the lower median age and prevalence of co-morbidities for COVID-19 patients in China may explain the lower prevalence of death, it is also possible that a selective under-reporting of death had occurred for studies from China. The death toll in China was initially under-reported and later updated on April 17, 2020 [95].
Whether or not cigarette smoking has been associated with SARS-CoV-2 acquisition or progression to severe disease has been strongly debated with studies showing both positive, null, and inverse association between smoking and COVID-19 [10,11,[96][97][98]. We found that patients with any history of smoking have both a higher risk of death (RR: 1.28; 95% CI: 1.06-1.55) and severe disease (1.29; 95% CI: 1.18-1.42). The case fatality risk for those with smoking history (27%) was also higher than the overall CFR of 20%. Whereas a higher COVID-19 mortality and morbidity among smokers may be due its causal association with COPD and CVD, Cai et al. [99] has also observed upregulation of pulmonary Angiotensin Converting Enzyme 2 (ACE2) gene expression and hence, pulmonary ACE2 receptors in smokers suggesting a direct effect of smoking on COVID-19 susceptibility and disease progression. ACE2 receptors are used by SARS-CoV-2 to translocate intracellularly [15,[100][101][102][103][104].
Our results of higher risk of death and severe disease associated with hypertension, diabetes and CVD in COVID-19 patients concurred with most studies conducted to date including studies that specifically investigated these associations [14,65,105,106]. However, it is unclear if cardiovascular risk factors including smoking, hypertension, diabetes, heart disease and CKD increases the susceptibility toward SARS-CoV-2 infection in the population [15,100,101,107]. On one hand, angiotensin-converting enzyme 2 (ACE2)-by blocking the renin angiotensin aldosterone system (RAAS) and decreasing or countering the vasoconstrictive, proinflammatory and profibrotic properties of angiotensin-II through catalysis of angiotensin-II to angiotensin-(1-7)-have been shown to exert cardiovascular protective effect and prevent acute lung injury from SARS-CoV-2 [15,100,101]. However, on the other hand, a possible greater expression of ACE2, the functional receptor mediating cellular entry of SARS-CoV-2 in humans, in patients with cardiovascular disease and other comorbidities can lead to increased susceptibility towards infection with SARS-CoV-2 [108,109]. In this context, it would be reasonable to posit that a substantially higher prevalence of cardiovascular comorbidities in the hospitalized patients compared to the prevalence in the general population may suggest elevated risk of acquisition of SARS-CoV-2 for patients with cardiovascular risk factors. To this end, we found that the prevalence of smoking history (23%), hypertension (55%), diabetes (33%) and heart disease (17%) in the hospitalized COVID-19 patients in the US were substantially higher than the corresponding prevalence of smoking (14%) [110], hypertension (29%) [111], diabetes (13%) [112] and heart disease (9%) [113] in the general US population that could suggest an association between these comorbidities and risk of SARS-CoV-2 infection or disease progression. However, we note that if the prevalence of these comorbidities in the asymptomatic individuals with COVID-19 in the general population is similar to that of their prevalence in the non-COVID-19 general population, then this difference-the higher prevalence of comorbidities in the hospitalized patients compared to the general populationcould simply imply a higher risk of symptomatic infection or hospitalization for individuals having SARS-CoV-2 infection. The prevalence of other risk factors i.e. COPD (9%) and CKD (15%) in the COVID-19 patients in the US was similar to the overall prevalence of COPD (7%) [114] and CKD (15%) [115] in the country. Generally, we noted a lower prevalence of comorbidities for patients in China. The prevalence of hypertension (23%) and diabetes (12%) in the hospitalized patients in China, which were lower than that of the US, approximate the respective prevalence of hypertension (23%) [116] and diabetes (15%) [117] in the general population of China. A previous meta-analysis also noted this observation [19]. Surprisingly, the prevalence of smoking (11%) in the COVID-19 patients hospitalized in China are inexplicably lower than the corresponding prevalence of smoking (23%) among COVID-19 patients in the US despite a higher prevalence of smoking (47% in Chinese males) [118] in the general Chinese population is significantly higher than that of the US. The prevalence of CVD (16%), COPD (4%) and CKD (2%) among COVID-19 patients in China are substantially lower than the corresponding prevalence of CVD (21%) [119], COPD (14%) [120], and CKD (11%) [121] in the general Chinese population. Given these discrepancies, we are unsure whether the lower prevalence of comorbidities noted for the COVID-19 patients in China are representative of the true prevalence. There was a great sense of urgency and a race to publish data in the early phase of the outbreak. As such, there exists the possibility of substantial under-recording of data on covariables. Had there been under-reporting, the implication would be a higher true prevalence estimate. We do not see reason for any systematic difference in reporting of risk factors based on outcome, or vice-versa, and hence, our summary relative risk estimates for association of risk factors with death or severe disease should not have been affected.
We assessed if patients with specific co-morbidities at baseline had higher risk of specific organ injury from SARS-CoV-2 during hospitalization. While the available data did not allow direct assessment of this relation, we compared the prevalence of comorbidities with the prevalence of corresponding organ system injury for studies that reported both baseline comorbidity and corresponding organ injury. We observed that the risk of acute lung injury/ARDS (48%), ACI (21%), and AKI (27%) were substantially higher than the baseline prevalence of COPD (6%), heart disease (11%) and CKD (14%), respectively. The higher prevalence of acute organ injury than the prevalence of baseline comorbidity simply indicates that ARDS, ACI and AKI were also occurring in patients who did not have a corresponding comorbidity at baseline in addition to people having the comorbidities.
Most studies reported only frequencies of risk factors and did not present adjusted measures for disease severity or death. Given this limitation, the risk ratio we calculated from the frequencies are largely unadjusted estimates. Future studies could additionally present, at the least, age-and sex-adjusted measures for association of risk of comorbidities with death or severe disease. Many studies reported odds ratio for the measure of association between preexisting conditions and risk of severe disease or death. Odds ratio poorly approximates risk ratio when the disease prevalence is high at baseline. For example, Zhou et al. [14] calculated an odds ratio of 5.4 (95% CI: 0.96-30.4) for risk of death from COPD in COVID-19 patients whereas the risk ratio we calculated from the frequencies presented is RR = 2.47 (95% CI: 1. 34-4.55). Prevalence of severe disease or death in COVID-19 patients was high in several studies. Similarly, several meta-analyses calculated odds ratios instead of risk ratios to summarize the risk of disease severity or death in association with risk factors such as smoking, diabetes, hypertension and cardiovascular disease [10,11,18], often to be interpreted by media and even by researchers as a measure of relative risk. Lack of rigor in research design, analysis and interpretation could generate inconsistent and ungeneralizable results across studies leading to controversy and confusion around serious public health issues such as that existing for association (or not) of smoking with COVID-19 disease acquisition, severity or death. As publications evolve at a pace that could be overwhelming for researchers and practitioners, we attempted to present a meaningful summary and inference for association of risk factors with death or severe disease from literatures published globally. Additionally, we provide an epidemiological framework for the risk of infection by SARS-CoV-2 based on presence of cardiovascular risk factors. This analysis can inform public health measures for COVID-19 screening and prevention, risk stratification and management of patients in clinical practice, analysis and presentation strategies for research data and inspire etiological investigations.

Conclusion
Epidemiological risk factors for progression of COVID-19 to severe disease and death and for acquisition of SARS-CoV-2, the causal agent for COVID-19, based on presence of pre-existing conditions have been insufficiently understood. Meta-analysis of 77 studies including 39023 COVID-19 patients hospitalized globally revealed case fatality risk of 52% for those having heart disease, 51% for COPD, 48% for CKD, 39% for CLD, 28% for hypertension, 27% for smoking history, 24% for diabetes, 35% for age�60 years, and 26% for males. Of all the patients who died, an overwhelming majority (85%) were in people aged�60 years. Also, of the people who died, 66% were males, 66% had hypertension, 44% had history of smoking, 39% had diabetes, 37% had CVD, 27% had CKD, and 6% had CLD. All of the above risk factors were significantly associated with death and severe disease in the patients hospitalized for COVID-19. The prevalence of ARDS was 48%, ACI 21%, and AKI 28% in the hospitalized patients. A higher prevalence of hypertension, diabetes, smoking and heart disease in the COVID-19 inpatients as compared to that of the general population could imply a higher risk of SARS-CoV-2 infection or disease progression for patients having these risk factors. These findings could inform public health strategies for targeted screening and appropriate control of modifiable risk factors such as smoking, hypertension, and diabetes to reduce morbidity and mortality. Finally, based on the published literature, there were vast differences in the prevalence of death and risk factors for the populations in China and in US-Europe that should be further investigated.  Table. Newcastle-Ottawa quality assessment (modified) for studies # . # Award of Points: Selection: points were awarded based on representativeness of the exposed group and unexposed group (2 points), ascertainment of exposures (1 point), and demonstration that outcome of interest was not present at the start of the study (1 point). Comparability (2 points): points were awarded based on whether the analyses were adjusted for age, sex, and other risk factors (2 points for adjustment to age and sex). Outcome (3points): points were awarded based on ascertainment of outcome through record linkage or independent blind assessment (1 points); duration of follow-up (1 point) (hospitalization till discharge); and adequacy of follow up for study population (complete follow up for the patients (vs whether patients were currently under treatment at the time of study report) (1 point), or if the patients currently under admission are excluded from outcome assessment (1 point).