The initial search of articles from the databases using the keywords resulted in 4275 articles (PubMed=1986, Embase= 2006, CINAHL =224, and Cochrane Library = 59) out of which 1527 were duplicates (Fig 1). Of the non-duplicate 2748 articles, 2444 were ineligible after screening the titles and abstracts, and 233 articles were excluded after full-text review. This resulted in 71 articles appropriate for inclusion in the systematic review [5-10, 24-89], and 60 of them were also included in the meta-analysis. The majority of the studies were conducted in China (n=53), and the others from South Korea (n=2), Iran (n=1), Europe (n=6) and USA (n=9) (S1 Table 3). Study designs included retrospective (n=42), cross-sectional (n=25), and prospective (n=4). Clinical outcomes reported were death in 34 studies, severe illness in 45 studies, and admission to ICU in 11 studies. Exposure variables examined in the studies included hypertension, cardiovascular disease, diabetes, chronic respiratory disease, cancer, chronic kidney disease, chronic liver disease, cerebrovascular disease, smoking, age and sex.
Hypertension
Out of 71 studies approved for inclusion in this review, 38 examined the correlation of hypertension with death, severe illness, and admission to ICU among COVID-19 patients. Out of the 38 studies, 36 included in the meta-analyses. Figure 2 shows summary estimates of the odds ratio of death (vs survival: n=13 studies), severe illness (vs moderate or mild: n=21 studies), and admission to ICU (vs non-ICU: n=4 studies) among hypertensive vs non-hypertensive patients. While some studies reported increased odds of death (n=9 studies) and severe illness (n=10 studies) among hypertensive patients, others reported a lack of correlation between hypertension and death (n=4 studies) or severe illness (n=11 studies). A summary analysis of pooled data from these studies showed moderate to high heterogeneity. Though, there were greater odds of death (OR 2.60, 95% CI 1.95–3.25, I2 = 52.6%, number of studies (n)= 13, number of participants (N)=53,222 ) and severe illness (OR 1.70, 95% CI 1.30 –2.10, I2 = 47.8%, n=21, N=6,172) among hypertensive as compared to non-hypertensive patients. The odds of admission to ICU were comparable between those who were hypertensive vs those who were not hypertensive (OR 1.20, 95% CI 0.30 –2.18, I2 = 79.2%, n=5, N=1,832).
Cardiovascular disease
Twenty studies examined the nature of the relationship of cardiovascular disease with the odds of developing death (n=7), severe illness (n=11) and admission to ICU (n=5) among COVID-19 patients. Out of the twenty studies, the majority showed higher odds of death (n=6) and severe illness (n=8) among COVID-19 patients with cardiovascular disease (Fig 3). The remaining 1 out of 7 studies showed a lack of association between cardiovascular disease and death, and 3 out of 11 showed a lack of association between cardiovascular disease and the odds of developing severe illness among COVID-19 patients. A meta-analysis of the studies showed higher odds of death (OR 5.16, 95% CI 4.10–6.22, I2 = 0.0%, n=6, N=47,134), severe illness (OR 2.04, 95% CI 1.01–3.08, I2 = 30.6%, n=11, N=48,535) and admission to ICU (OR 1.36, 95% CI 1.04–1.69, I2 = 0.0%, n=5, N=8,346) among COVID-19 patients who had cardiovascular disease as compared to those without this health problem (Fig 3).
Diabetes
Diabetes also is posited to be linked with the risk of developing severe illness and death among COVID-19 patients. Twelve studies examined if having diabetes is correlated with the odds of death among COVID-19 patients, half of which reported increased odds of death in COVID-19 patients with diabetes. Correlation of having diabetes with the odds of developing severe illness and the odds of admission to ICU was assessed in 18 (eight reported increased odds) and five (two reported increased odds) studies, respectively. A summary analysis of these studies showed greater odds of death (OR 2.11, 95% CI 1.35–2.87, I2 = 67.4%, n=12, N=59281), severe illness (OR 1.65, 95% CI 1.23–2.08, I2 = 24.9%, n=18, N=5811) and admission to ICU (OR 1.55, 95% CI 1.20–1.90, I2 = 0.0%, n=5, N=8469) among patients with diabetes than those who had no diabetes (Fig 4).
Chronic respiratory disease
Of 71 included studies, 19 compared the odds of death vs survival (n=9), severe vs moderate or mild illness (n=9), and admission to ICU vs not (n=4) among COVID-19 patients who had chronic respiratory disease vs those without this problem. Of the 9 studies which compared the odds of death vs survival, six reported significantly greater odds of death among COVID-19 patients with chronic respiratory disease, but three documented lack of association between chronic respiratory disease and the odds of death. On the other hand, out of the 9 studies which compared the odds of severe vs mild or moderate illness among COVID-19 patients, only three reported significantly greater odds of severe illness, but five showed a lack of association between chronic respiratory disease and the odds of developing severe illness. One study reported lower odds of severe illness among COVID-19 patients with chronic respiratory disease as compared to those without chronic respiratory disease. Meta-analysis of the 19 studies showed association of chronic respiratory disease with increased odds of death (OR 2.83, 95% CI 2.14–3.51, I2= 0.0%, n=9, N=59,624) and admission to ICU (OR 1.52, 95% CI 1.09–1.94, I2 = 0.0%, n=5, N=8,346), but there was lack of correlation of the disease with severe illness (OR 1.39, 95% CI 0.15–2.64, I2 = 27.0%, n=9, N=48,381) among COVID-19 patients (Fig 5).
Cancer
Thirteen studies examined the nature of the relationship of cancer with the odds of death (n=5), severe illness (n=6) and admission to ICU (n=3) among COVID-19 patients. Increased odds of death, severe illness and admission to ICU among COVID-19 patients who had cancer was reported in 2 (out of 5), 3 (out of 6) and 1 (out of 3) studies, respectively. The remaining studies showed a lack of association between having cancer and the odds of death (n=3), severe illness (n=3) and admission to ICU (n=2) among COVID-19 patients. Meta-analysis of the 13 studies showed lack of association of cancer with the odds of death (OR 1.72, 95% CI 0.39 – 3.05, I2 = 0.0%, n=5, N=46,793), severe illness (OR 2.32, 95% CI 0.45 – 4.19, I2 = 0.0%, n=6, N=3,169) and admission to ICU (OR 1.79, 95% CI -0.74 – 4.31, I2 = 77.1%, n=3, N=1,804) among COVID-19 patients (S1 Fig).
Chronic kidney disease
Twelve studies examined the nature of the relationship of chronic kidney disease with the odds of death (n=3), severe illness (n=7) and admission to ICU (n=2) among COVID-19 patients. Of the 12 studies, two showed increased odds of death (n=1) or severe illness (n=1) among COVID-19 patients who had vs didn’t have chronic kidney disease. The remaining 10 studies showed a lack of association of chronic kidney disease with the odds of death (n=3), severe illness (n=3) and admission to ICU (n=2). Meta-analysis of the 12 studies showed lack of association of chronic kidney disease with the odds of death (OR 2.32, 95% CI 0.06 – 4.58, I2 = 0.0%, n=3, N=4951), severe illness (OR 1.42, 95% CI 0.46 –2.38, I2 = 0.0%, n=7, N=7788) and admission to ICU (OR 1.37, 95% CI 0.88–1.86, I2 = 0.0%, n=2, N=6653) among COVID-19 patients (S2 Fig).
Chronic liver disease
Out of the 71 studies included in this review, only seven tested the association of chronic liver disease with death or severe illness in COVID-19 patients. Of the seven studies, the only one reported a correlation of chronic liver disease with COVID-19 severity. A summary analysis of the seven studies showed lack of correlation of chronic liver disease with death (OR 1.41, 95% CI -2.20 – 5.02, I2 = 0.0%, n=2, N=352), severe illness (OR 1.35, 95% CI 0.67–2.03, I2 = 0.0%, n=4, N=6855) and admission to ICU (OR 1.77 95% CI 0.65 – 4.80, n=1, N=6637) (S3 Fig).
Cerebrovascular diseases
Six studies reported findings on the odds of death (n=2), severe illness (n=2) and admission to ICU (n=2) among COVID-19 patients with cerebrovascular diseases vs those without having this comorbidity. Two studies by Guan et al. 2020 [6] and Wang et al. 2020 [72] reported increased odds of death among COVID-19 patients with cerebrovascular diseases. Another two studies also showed association of cerebrovascular diseases with increased odds of developing severe illness [37] and admission to ICU [8].
A meta-analysis of the two studies showed increased odds of death among COVID-19 patients with cerebrovascular diseases (OR 5.14, 95% CI 1.08 – 9.19, I2 = 0.0%, n=2, N=1,983), but the odds of severe illness (OR 2.18, 95% CI -0.07 – 4.44, I2 = 0.0%, n=2, N=928) and admission to ICU (OR 4.09, 95% CI -0.20 – 8.38, I2 = 0.0%, n=2, N=1,606) were similar between patients with cerebrovascular diseases and those without this comorbidity (S4 Fig).
Smoking
A total of 14 studies included in the current meta-analyses that examined the impact of tobacco smoking on the clinical outcomes of COVID-19. Of these 14 studies, most showed a lack of association of smoking with death (2 out of 2), severe illness (9 out of 11), or admission to ICU (1 out of 1) among COVID-19 patients. Only two studies showed increased odds of severe illness among COVID-19 patients who were former or current smokers than non-smokers. A meta-analysis of the studies showed similar odds of death (OR 0.90, 95% CI 0.70 – 1.10, I2 = 0.0%, n=2, N=9155) or severe illness (OR 1.31, 95% CI 0.96 – 1.67, I2= 0.0%, n=11, N=4509) among COVID-19 patients who were former or current smokers than non-smokers (S5 Fig).
Age
Out of the 71 studies, 27 examined the effect of age on the odds of death, severe illness and admission to ICU. Out of 27 studies, 15 treated age as >65 vs ≤ 65 years and 10 as ≥ 60 vs <60 years. Of the 15 studies which treated age as > 65 vs ≤ 65 years, 13 reported increased odds of death (n=5) or severe illness (n=9) among patients of ages > 65 years as compared to those of ≤ 65 years. Similarly, nine studies that treated age as ≥ 60 vs < 60 years reported increased odds of death (n=6) or severe illness (n=3) among patients of ages ≥ 60 years as compared to those < 60 years. Chen et al. [89] also showed an increased hazard ratio of severe illness among patients of Age ≥ 65 years as compared to younger individuals (Hazard ratio 3.43; 95% CI, 1.24-9.5). Dong et al. [88] also reported an increased prevalence of severe COVID-19 illness in children than those of adult patients. However, a study showed similar odds of severe illness among individuals of ages ≥ 60 vs <60 years [80]. Two studies also reported similar odds of admission to ICU among individuals of ages > 65 years vs those of ≤ 65 years [10,25]. A meta-analysis of these studies showed a greater odds of death (OR 3.56, 95% CI 1.21 – 5.90, I2 = 18.2%, n=6, N=13,964) and severe illness (OR 2.55, 95% CI 1.94 – 3.17, I2 = 24.5%, n=9, N=3374) among patients of ages > 65 years than those of ≤ 65 years. The summary odds ratio estimates of death (OR 6.09, 95% CI 3.53 – 8.66, I2 = 95.5%, n=6, N=100,733) and severe illness (OR 4.91, 95% CI 1.35 – 8.47, I2 = 0.0%, n=4, N=333) among older-aged COVID-19 patients than younger ones were even much greater when age was treated as ≥ 60 vs < 60 (Fig 6).
Gender
Fifty-two studies examined the effect of gender on the odds of death (n=16), severe illness (n=32) and admission to ICU among COVID-19 patients (n=6). Out of 16 studies which compared the odds of death between males and females, 10 reported increased odds in males and six documented similar odds between males vs female. A meta-analysis of the 16 studies showed increased odds of death among males than females (OR 1.34, 95% CI 1.18 – 1.50, I2 = 38.7%, n=16, N=68,609). Similarly, of the 32 studies which compared the odds of severe illness between males vs females, 11 reported increased odds and 21 studies document similar odds between males vs female. A summary analysis of the 32 studies showed increased odds of severe illness among males than females (OR 1.35, 95% CI 1.23 – 1.47, I2 =0.0%, n=32, N=13,426); however, this could not be further assessed by tandem risk factors (e.g., underlying cardiovascular disease) in the available data.
There were six studies that examined the relationship of gender with the odds of admission to ICU. Out of these six studies, two showed increased odds of admission to ICU in males than females, but four studies showed a lack of association between gender and the odds of admission to ICU. A summary of the 6 studies showed a lack of correlation between gender and admission to ICU in COVID-19 patients (OR 1.52, 95% CI 0.69 – 2.36, I2 =2.9%, n=6, N=666) (Fig 7).
Heterogeneity assessment
There was no heterogeneity (I2= 0.0%) observed among the studies included in the meta-analyses that estimated the summary odds ratio of death, severe illness or admission to ICU among COVID-19 patients with cerebrovascular diseases, chronic liver disease, chronic kidney disease, cancer or smoker vs those without these comorbidities. There was also no or low heterogeneity among the studies included in the meta-analyses that examined correlation of clinical outcomes in COVID-19 patients with cardiovascular disease (I2= 0.0% for death and admission to ICU, I2=30. 6% for severe illness), diabetes (I2= 0.0% for admission to ICU, I2=24.9% for severe illness), chronic respiratory disease (I2= 0.0% for death and admission to ICU, I2=27.0% for severe illness ), age in years > 65 vs ≤65 (I2= 18.2% for death, I2=24.5% for severe illness), and gender (I2= 0.0% for severe illness, I2=2.9% for admission to ICU, I2=38.7% for death). The heterogeneity level in meta-analyses performed to examine associations of hypertension with the odds of death (I2= 52.6%) and severe illness (I2= 47.8%) among COVID-19 patients was moderate.
However, heterogeneity was high among the studies combined in the meta-analyses to examine the association of death with diabetes (I2 = 67.4%) and age, ≥ 60 vs < 60 years (I2 = 95.5%) in COVID-19 patients. Studies that were included in the meta-analyses to assess the nature of the relationship of admission to ICU with hypertension (I2 = 79.2%) also had high heterogeneity. Subgroup analyses by study design dropped heterogeneity: diabetes (I2 = 0.0% in retrospective), and age ( I2 = 0.0% in retrospective; I2 =0.0% in cross-sectional). Due to the few numbers of studies that reported results on the effect of hypertension on the risk of admission to ICU, we didn’t perform subgroup analyses by study region/country or study design. Removal of a study by Guan et al. 2020a [6], however, significantly minimized the heterogeneity of the studies that examined the effect of hypertension on the risk of admission to ICU (I2 = 1.0%).
On the other hand, the meta regression showed lack of effect of study area or country (meta regression coefficient (β)= -0.50, P= 0.073), study design (β = -0.23, P= 0.400) and sample size (1.25e-06, P= 0.912) on the log ORs of death among COVID-19 patients with diabetes vs those without this problem. Similarly, study area or country (β= -0.30, P= 0.380), study design (β = β = -0.04, P= 0.863) and sample size (β = 8.74e-06, P= 0.631) didn’t significantly predict the log ORs of admission to ICU among COVID-19 patients of age ≥60 years vs <60 years in the meta regression model.
Publication bias assessment
Odds ratio distributions against their standard error estimates and Egger tests for asymmetry did not indicate significant publication bias: hypertension (bias 6.22, P=0.11), diabetes (bias 1.43, P=0.634), cardiovascular disease (bias 1.64, P=0.71), chronic respiratory diseases (bias 6.47, P=0.13), chronic kidney disease (bias 88.99, P=0.11), and cancer (Egger’s regression test (bias) -0.26, P=0.953) (Fig 8 and S6 Fig). Funnel plots of odds ratios of the likelihood of developing severe vs moderate or mild illness among COVID-19 patients and their Egger tests also were not significant: hypertension (bias 5.05, P=0.21); diabetes (bias 3.29, P=0.353), chronic respiratory diseases (bias -8.36, P=0.12); chronic kidney disease (bias -0.57, P=0.807); and, cancer (bias 3.34, P=0.598) (Fig 8 and S6 Fig). Studies comparing the odds ratio of admission to ICU vs not admission to ICU among COVID-19 patients with cancer (bias -6.27, P=0.51), hypertension (bias -5.36, P=0.12), diabetes (bias 0.53, P=0.763), cardiovascular (bias 1.37, P=0.40), chronic respiratory diseases (bias 1.98, P=0.67) were also spread evenly on both sides of the average OR estimates, creating an approximately symmetrical funnel-shaped distribution (Fig 8 and S6 Fig). The odds ratio estimates of the studies which examined the relationship of chronic liver disease (bias 0.47, P=0.752), cerebrovascular diseases (bias -5.83, P=0.395) and smoking behavior (bias 0.98, P=0.217) with severe illness or death in COVID-19 patients also had approximately symmetrical funnel-shaped distribution and the Egger’s test for the asymmetry of the plots were not significant (S6 Fig). However, there was publication bias among the studies comparing the odds of i) death vs survival in patients with age > 65 vs ≤65 years (bias 11.29, P=0.062) and males vs females (bias 2.90, P=0.046); ii) severe vs mild or moderate illness in patients with age > 65 vs ≤65 years (bias 14.63, P=0.032), males vs females (bias 2.18, P=0.039) and those with cardiovascular disease vs without having this problem (bias 5.62, P=0.0.034), the odds ratio estimates scattered asymmetrically in the funnel plot (Fig 8 and S6 Fig).
Risk of bias and quality of the studies
The study's risk of bias and quality included in this study is summarized in the supplementary table (S4 table). Out of the 71 studies, 54 were moderate quality, and 18 were good quality in recruiting the study participants. The majority of the studies also used good (n=54) or moderate quality (n=17) reliable and valid tools to determine COVID-19 severity status and related deaths among the study participants. In many studies (n=44), the researchers or data collectors were not aware of the group (exposed vs. unexposed) to which the study participants belong in the data collection process and/or the participants were blinded about the research question. About 2/3 of the included studies were of moderate quality in terms of the study design (retrospective case-control or cohort) and controlling confounders that may affect the relationship between comorbidities and COVID-19 related outcomes or death. On the other hand, several studies rated as low quality based on study design (n=15), data collection (n=8), confounders (n=26), blinding (n=28) and dropouts and withdrawals (n=29). None of the studies were of low quality in terms of blinding of the outcome assessor and the study participants. Total rating using the six criteria showed that 14 studies were of strong quality, 23 studies were of moderate quality, and 34 studies were rated as low quality. Studies were included in this review, regardless of their qualities.