Prognostic factors for severity and mortality in patients infected with COVID-19: A systematic review.

BACKGROUND AND
PURPOSE: The objective of our systematic review is to identify prognostic factors that may be used in decision-making related to the care of patients infected with COVID-19. DATA SOURCES: We conducted highly sensitive searches in PubMed/MEDLINE, the Cochrane Central Register of Controlled Trials (CENTRAL) and Embase. The searches covered the period from the inception date of each database until April 28, 2020. No study design, publication status or language restriction were applied. STUDY SELECTION AND DATA EXTRACTION: We included studies that assessed patients with confirmed or suspected SARS-CoV-2 infectious disease and examined one or more prognostic factors for mortality or disease severity. Reviewers working in pairs independently screened studies for eligibility, extracted data and assessed the risk of bias. We performed meta-analyses and used GRADE to assess the certainty of the evidence for each prognostic factor and outcome.
RESULTS: We included 207 studies and found high or moderate certainty that the following 49 variables provide valuable prognostic information on mortality and/or severe disease in patients with COVID-19 infectious disease: Demographic factors (age, male sex, smoking), patient history factors (comorbidities, cerebrovascular disease, chronic obstructive pulmonary disease, chronic kidney disease, cardiovascular disease, cardiac arrhythmia, arterial hypertension, diabetes, dementia, cancer and dyslipidemia), physical examination factors (respiratory failure, low blood pressure, hypoxemia, tachycardia, dyspnea, anorexia, tachypnea, haemoptysis, abdominal pain, fatigue, fever and myalgia or arthralgia), laboratory factors (high blood procalcitonin, myocardial injury markers, high blood White Blood Cell count (WBC), high blood lactate, low blood platelet count, plasma creatinine increase, high blood D-dimer, high blood lactate dehydrogenase (LDH), high blood C-reactive protein (CRP), decrease in lymphocyte count, high blood aspartate aminotransferase (AST), decrease in blood albumin, high blood interleukin-6 (IL-6), high blood neutrophil count, high blood B-type natriuretic peptide (BNP), high blood urea nitrogen (BUN), high blood creatine kinase (CK), high blood bilirubin and high erythrocyte sedimentation rate (ESR)), radiological factors (consolidative infiltrate and pleural effusion) and high SOFA score (sequential organ failure assessment score).
CONCLUSION: Identified prognostic factors can help clinicians and policy makers in tailoring management strategies for patients with COVID-19 infectious disease while researchers can utilise our findings to develop multivariable prognostic models that could eventually facilitate decision-making and improve patient important outcomes. SYSTEMATIC REVIEW REGISTRATION: Prospero registration number: CRD42020178802. Protocol available at: https://www.medrxiv.org/content/10.1101/2020.04.08.20056598v1.

Introduction

COVID-19 is an infectious disease caused by the SARS-CoV-2 coronavirus [1]. It was first identified in Wuhan, China, on December 31, 2019 [2]; five months later, more than six million cases had been identified across 215 countries [3]. On March 11, 2020, WHO characterised the COVID-19 outbreak as a pandemic [1]. While the majority of cases present with mild symptoms, a minority progress to acute respiratory illness and hypoxia requiring hospitalization, and a subset develop acute respiratory distress syndrome, multi-organ failure or have fatal outcomes [4]. The case fatality rate reported across countries, settings and age groups is highly variable, ranging from about 0.5% to 10% [5]. In hospitalised patients it has been reported to be higher than 20% [6].

Prognostic factors (stand-alone or combined in risk assessment models) may guide the stratification of patients with SARS-CoV-2 infectious disease based on their risk of severe disease or death. This risk stratification may subsequently guide optimised management and resource utilisation strategies in the care of these patients.

Although multiple prognostic factors have been proposed and some have been accepted as “established” by the scientific community (i.e age), the predictive value of most of these potential prognostic factors has not been robustly evaluated and remains uncertain. As pointed out by Wynants et.al: “unreliable predictors could cause more harm than benefit in guiding clinical decisions” [7]. For example, aggressive and risky interventions might be attempted if the risk of poor outcomes are inaccurately defined as high based on unreliable predictors.

Using innovative and agile processes supported by advanced evidence synthesis tools and collaborative efforts across several international research groups, this systematic review aims to provide a rigorous summary of the evidence available on prognostic factors that may be used in decision-making related to the care of patients infected with COVID-19.

Methods

Protocol registration

We published [8] and registered the protocol for this systematic review with PROSPERO (CRD42020178802).

Search strategy

We conducted highly sensitive searches in PubMed/MEDLINE, the Cochrane Central Register of Controlled Trials (CENTRAL) and Embase. The searches covered the period from the inception date of each database until April 28, 2020. No study design, publication status or language restriction were applied.

Detailed strategies for each database are reported in the S1 Text.

In order to identify articles that might have been missed in the electronic searches, we reviewed the reference list of each included study and performed cross citation in Google Scholar using each included study as the index reference.

Study selection

Four reviewers working independently and in duplicate, performed study selection, including screening of titles and abstracts and of potentially eligible full-text articles. Reviewers resolved disagreements by discussion.

We included studies examining individual prognostic factors or risk assessment models based on the typologies of prognosis proposed by Iorio and colleagues [9] and the PROGnosis RESearch Strategy (PROGRESS) Group framework [10] without applying any restrictions based on analytical methods (i.e performing multivariable analysis).

Specifically, we included studies that evaluated patients with confirmed SARS-CoV-2 infectious disease regardless of the healthcare setting (i.e ambulatory or inpatients) and of baseline disease severity. We investigated all candidate prognostic factors reported by individual studies and compared patients exposed (the factor was present) with patients unexposed (the factor was absent) to each one of those factors.We considered studies that assessed mortality or severe COVID-19 disease as outcomes and accepted the author's definitions of the latter. When severe COVID-19 disease was reported as a multi-categorical scale, we used the most severe category. Additionally, when severe COVID-19 disease was not reported as an outcome, we considered ICU requirement, invasive mechanical ventilation (IVM) and acute respiratory distress syndrome (ARDS) as surrogate outcomes.

Data extraction

For each eligible study, five pairs of reviewers, independently, abstracted the following information on study characteristics (year of publication, country, medical center and time period in which the study was conducted); population characteristics (sample size, context in which the study was conducted and other population characteristics); description of prognostic factors and outcomes and their definitions; and study results (measures of association or crude event rates for every candidate prognostic factor and outcome reported).

Risk of bias assessment

Two reviewers assessed the risk of bias of individual included studies independently and in duplicate. Discrepancies were resolved by consensus. We used the Quality in Prognosis Studies (QUIPS) tool for prognostic factor studies [11] which considers population characteristics, attrition, prognostic factor and outcome measurement and potential residual confounding. For “study confounding summary” and “statistical analysis and presentation domains”, in order to assess adequacy of the multivariable models, we considered appropriate model adjustment as based on inclusion of age, one comorbidity (e.g diabetes) and one parameter of disease severity (e.g. respiratory rate) at minimum.

Data synthesis and analysis

We presented the results of individual prognostic factors in both tabular and narrative formats. We standardized the units of measurement for each prognostic factor, unifying the direction of the predictors and adjusting the weights of the studies, and calculated crude effect estimates when not provided [12]. When possible, we meta-analysed all prognostic factors whose association with the selected outcomes of interest was explored and were reported by more than one study. We used the generic inverse variance-based method to produce an overall measure of association and random-effect models based on the DerSimonian-Laird method provided by the metafor package for R software [13].

For every candidate prognostic factor, we presented the measure of association as odds ratios (ORs) and their corresponding 95% confidence intervals (CI). In studies that reported the measure of association as a hazard ratio (HR) or risk ratio (RR), we converted them to ORs using the baseline risk (death rate or incidence of severe COVID-19 out of the total sample) reported in the studies [14,15]. For dichotomous variables, we used the crude event rate to calculate ORs when no measures of association were provided. We excluded information on continuous variables for which no measures of association were available. We also calculated absolute risk differences (RDs) that can be attributed to every individual candidate prognostic factor by applying the ORs to estimated baseline risks (see below “Baseline risks”). When the same candidate prognostic factor was assessed in alternative ways (e.g. dichotomic and continuous) we used the one for which we found better certainty of evidence.

For every explored candidate variable, we performed sensitivity analysis excluding high risk of bias studies and studies that did not report adjusted estimates. In cases where the effect estimates provided by the primary analysis and the sensitivity analysis significantly differed, we either presented the moderate/low risk of bias–adjusted estimates or the primary analysis estimates but rated down certainty of the evidence because for risk of bias (see below). In addition, when we observed inconsistent results for disease severity outcome, we performed subgroup analyses accounting for outcome definition (i.e severity scale vs. IVM vs. ARDS) as a potential source of heterogeneity.

Assessment of certainty of the evidence

We assessed certainty of the evidence for each candidate prognostic factor, by outcome, based on the GRADE approach [16]. The approach considers the following domains: risk of bias, indirectness, inconsistency, imprecision, and publication bias. We produced summary of findings tables and rated the certainty of the evidence as high, moderate, low or very low depending on the grading of the individual domains [16]. See S1 Text for a detailed description of the certainty of the evidence assessment.

Result interpretation

To define which candidate variables provide valuable prognostic information we adopted a minimally contextualized approach [17]. To this end, we arbitrarily set thresholds to define important incremental increase in the risk of our outcomes. In setting those thresholds we aimed to define the minimal incremental increase in the risk of mortality or severe COVID-19 disease that could be interpreted as valuable prognostic information without considering the potential consequences of using that information in healthcare decision-making. These thresholds represent the line that separates a risk increase that is trivial from a small but important risk increase. We set those thresholds in 0.5% increase in mortality and 1% increase insevere COVID-19 disease. We performed a sensitivity analysis in which we adopted a purely non-contextualized approach [17] to assess mortality outcome. In doing so we only considered the relative measures of association and used an OR of 1 as the threshold for minimal important risk increment.

Baseline risks

To define baseline risks we selected clinical scenarios based in the severity categories proposed by WHO [18]. To assess the prognostic value on mortality, we used the clinical scenario of a patient infected with COVID-19 with severe but not critical disease (i.e patients with respiratory failure but not invasive mechanical ventilation and/or hemodynamic support requirement). We identified one study informing prognosis in this specific subgroup with a mortality risk of 9% [19]. However, as we identified significant variability in mortality risks reported for similar clinical scenarios (i.e in the RECOVERY trial [20] mortality risk in hospitalized patients assigned to the control arm, with no baseline oxygen requirement was 14%), we performed a sensitivity analysis using a baseline risk of 26% as reported by a large cohort of non-ICU inpatients treated in 255 sites across 36 countries [6].

To assess the prognostic value on severe COVID-19 disease, we used the clinical scenario of a patient infected with COVID-19 with non-severe disease. We identified 7 studies informing prognosis in this specific subgroup with a median risk of progression to severe or critical state of 13% [21-27]. We calculated baseline risks (risks in patients not exposed to the prognostic factor) by also considering the prevalence of every prognostic factor and the estimates of association [28]. When prevalence of prognostic factors was not available we used described baseline risks (9% for mortality and 13% for severe COVID-19 disease).

Update of this systematic review

An artificial intelligence algorithm deployed in the Coronavirus/COVID-19 topic of the L.OVE platform (https://app.iloveevidence.com/loves/5e6fdb9669c00e4ac072701d) will provide instant notification of articles with a high likelihood of eligibility. These will be screened by paired reviewers iteratively and will conduct data extraction and iterative updates of estimates for selected prognostic factors accordingly. We will consider resubmission to a journal if there is a substantial modification on the measure of association or the certainty of the evidence for a given prognostic factor such that it is clinically significant, at the discretion of the reviewer team. This review is part of a larger project established to produce multiple parallel systematic reviews relevant to COVID-19 [29].

Results

(Fig 1) illustrates the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) flow diagram. Our search identified 7631 citations of which we included 569 studies for full text assessment. 207 studies fulfilled the inclusion criteria [21,23-25,27,30-231]. These 207 studies, with sample sizes from 10 to 8910, enrolled a total of 75607 patients and were conducted in 12 different countries (China, USA, Canada, Spain, France, Turkey, Korea, Japan, Italy, Germany, India and Singapore).

Fig 1

PRISMA (preferred reporting items for systematic reviews and meta-analyses) flowchart of study inclusions and exclusions.

Description of included studies

S1 Table describes the characteristics of the included studies reporting on mortality and/or severe COVID-19 disease. Regarding candidate prognostic factors, of the 207 included studies, 184 (88.9%) reported socio-demographic variables, 180 (86.9%) comorbidities, 178 (86%) clinical findings, 176 (85%) laboratory findings and 106 (51.3%) imaging findings. The outcomes reported were mortality in 116 (56%) and progression to severe/critical status in 131 (63.3%). In 78 (37.7%) of the included studies a multivariable analysis was performed. In the 150 studies in which the severity of included patients was described the mean proportion of patients in each category was: non-severe disease 63.8%, severe disease 22.6%, critical disease 13.6%.

Risk of bias

Risk of bias was high across most identified studies. Among the 207 included studies only 7 were judged as low risk of bias [21,24,56,57,59,130,142] as the remaining presented important limitations in at least one domain or item. Most frequent were retrospective design, which may have introduced classification bias, and lack or inappropriate adjusted analysis. S2 Table provides detailed judgements for each of the risk of bias domains criteria.

Prognostic factors for mortality

We investigated 96 candidate prognostic factors for mortality from 116 studies including 57044 patients. S3 Table provides a summary of findings for all the candidate prognostic factors and S1 Appendix includes the corresponding forest plots.

We found high or moderate certainty that the following 35 variables provide valuable prognostic information on mortality outcome (Table 1).

Table 1

Prognostic factors for mortality and/or severe COVID-19 disease.

Prognostic factor	Mortality					Severe COVID-19 disease
	Number of patients (studies)	Odds ratio (95%CI)	Risk without prognostic factor	Risk with prognostic factor	Certainty of the evidence	Number of patients (studies)	Odds ratio (95%CI)	Risk without prognostic factor	Risk with prognostic factor	Certainty of the evidence
Socio-demographic characteristics
Age Definition: 10 years increase	11962 (19)	1.80 (1.54–2.10)	9%	15.1%	⨁⨁⨁⨁ HIGH	14456 (53)	1.63 (1.47–1.80)	13%	19.6%	⨁⨁⨁⨁ HIGH
			6.1% increase in mortality. Between 4.2% more and 8.2% more					6.6% increase in severe COVID-19 disease. Between 5% more and 8.2% more
Sex Definition: Male	31948 (58)	1.72 (1.5–1.98)	8%	13%	⨁⨁⨁◯ MODERATEd	25032 (122)	1.53 (1.4–1.67)	10.8%	15.5%	⨁⨁⨁⨁ HIGH
			5% increase in mortality. Between 4% more and 7% more					4.7% increase in severe COVID-19 disease. Between 3.7% more and 5.6% more
Smoking Definition: Active, present smoker	12025 (16)	1.57 (1.19–2.07)	8.7%	13%	⨁⨁⨁⨁ HIGH	9147 (45)	1.65 (1.25–2.17)	12.1%	18.4%	⨁⨁⨁◯ MODERATE d
			4.3% increase in mortality. Between 1.5% more and 7.5% more					6.3% increase in severe COVID-19 disease. Between 2.7% more and 10.2% more
Medical illness and patient history
Any chronic condition or comorbidities	4406 (16)	3.3 (2.18–5)	5.9%	16.2%	⨁⨁⨁◯ MODERATE a	6640 (40)	3.16 (2.71–3.68)	8.2%	20.1%	⨁⨁⨁⨁ HIGH
			10.3% increase in mortality. Between 6.8% more and 13.4% more					12% increase in severe COVID-19 disease. Between 10.6% more and 13.2% more
Cerebrovascular disease Definition: History of stroke or CNS disease	15294 (26)	2.85 (2.02–4.01)	8.7%	21.3%	⨁⨁⨁⨁ HIGH	11050 (42)	2.67 (1.84–3.87)	12.7%	27.8%	⨁⨁⨁◯ MODERATE d
			12.6% increase in mortality. Between 7.5% more and 18.5% more					15.1% increase in severe COVID-19 disease. Between 8.4% more and 22.8% more
COPD	34759 (41)	2.43 (1.88–3.14)	8.5%	18.4%	⨁⨁⨁⨁ HIGH	15468 (65)	2.7 (2.14–3.4)	12.6%	27.9%	⨁⨁⨁⨁ HIGH
			9.8% increase in mortality. Between 6.4% more and 13.6% more					15.3% increase in severe COVID-19 disease. Between 11% more and 20% more
Chronic kidney disease Definition: KDIGO definition of CKD	23448 (28)	2.27 (1.69–3.05)	8.5%	17.2%	⨁⨁⨁⨁ HIGH	12056 (42)	2.21 (1.51–3.24)	12.8%	24.5%	⨁⨁◯◯ LOW a,d
			8.8% increase in mortality. Between 5.1% more and 12.9% more					11.7% increase in severe COVID-19 disease. Between 5.4% more and 19.2% more
Cardiovascular disease Definition: Coronary heart disease or congestive heart failure	37156 (51)	2.12 (1.77–2.56)	8.1%	15.5%	⨁⨁⨁◯ MODERATE d	16679 (73)	3.34 (2.71–4.1)	12.2%	31.3%	⨁⨁⨁◯ MODERATE d
			7.5% increase in mortality. Between 5.4% more and 9.7% more					19.1% increase in severe COVID-19 disease. Between 15.1% more and 23.1% more
Cardiac arrhythmia	12729 (6)	2.13 (1.72–2.65)	7%	13.6%	⨁⨁⨁⨁ HIGH	747 (4)	16.51 (6.69–40.77)	6.5%	35.5%	⨁⨁◯◯ LOW a,c,e
			6.5% increase in mortality. Between 4.7% more and 8.4% more					29% increase in severe COVID-19 disease.Between 22.6% more and 32.3% more
Arterial hypertension	31341 (52)	2.02 (1.71–2.38)	7%	13%	⨁⨁⨁⨁ HIGH	20817 (94)	2.5 (2.21–2.92)	11.1%	23.3%	⨁⨁⨁◯ MODERATEd
			6% increase in mortality. Between 4.5% more and 7.3% more					12.1% increase in severe COVID-19 disease. Between 10.4% more and 14.4% more
Diabetes	30303 (52)	1.84 (1.61–2.1)	7.9%	13.6%	⨁⨁⨁⨁ HIGH	21381 (97)	2.51 (2.2–2.87)	12%	25.2%	⨁⨁⨁⨁ HIGH
			5.6% increase in mortality. Between 4.3% more and 7% more					13.2% increase in severe COVID-19 disease. Between 11% more and 15.5% more
Dementia	8922 (3)	1.54 (1.31–1.81)	9%	13.2%	⨁⨁⨁⨁ HIGH	0 (0)	NA	NA	NA	NA
			4.2% increase in mortality. Between 2.5% more and 6.2% more					NA
Obesity: BMI > 25–30	9127 (3)	1.41 (1.15–1.74)	8.5%	11.5%	⨁⨁⨁⨁ HIGH	1140 (8)	3.74 (2.37–5.89)	10.2%	35%	⨁⨁⨁⨁ HIGH
			3.1% increase in mortality. Between 1.2% more and 5.1% more					16.7% increase in severe COVID-19 disease. Between 9.6% more and 24.7% more
Cancer Definition: Solid or active haematologic cancer	22734 (25)	1.35 (1.17–1.55)	8.9%	11.6%	⨁⨁⨁⨁ HIGH	15156 (58)	2.06 (1.64–2.58)	12.8%	23.2%	⨁⨁⨁◯ MODERATEd
			2.7% increase in mortality. Between 1.4% more and 4.2% more					10.4% increase in severe COVID-19 disease. Between 6.6% more and 14.5% more
Dyslipidemia	11273 (4)	1.26 (1.06–1.5)	8.9%	11%	⨁⨁⨁◯ MODERATEb	559 (4)	0.63 (0.22–1.83)	13.1%	8.7%	⨁⨁◯◯ LOW a,b
			2.1% increase in mortality. Between 0.5% more and 3.9% more					4.4% decrease in severe COVID-19 disease. Between 10% less and 8.3% more
Symptoms, vital signs and physical examination
Respiratory failure Definition: increased respiratory rate, abnormal blood gases (hypoxemia, hypercapnia, or both), and evidence of increased work of breathing	1887 (8)	21.17 (4.9–91.3)	3.1%	23.4%	⨁⨁⨁◯ MODERATEa	1156 (7)	23.21 (12.07–44.62)	NA	NA	NA
			20.3% increase in mortality. Between 13.4% more and 22.4% more					NA
Low blood pressure Definition SBP less than 90–100 mmHg	1269 (2)	6.7 (3.14–14.33)	9%	39.9%	⨁⨁⨁◯ MODERATEa	480 (2)	1.29 (0.72–2.29)	NA	NA	NA
			30.9% increase in mortality. Between 14.7% more and 49.6% more					NA
Hypoxemia Definition: Low digital saturation (below 90–93%)	1047 (5)	5.46 (2.05–14.53)	2.3%	9.1%	⨁⨁⨁◯ MODERATEa	1331 (5)	4.69 (1.56–14.09)	NA	NA	NA
			6.7% increase in mortality. Between 4.2% more and 7.7% more					NA
Tachycardia Definition: More than 90–100 bpm	1269 (2)	2.61 (1.62–4.22)	9%	20.5%	⨁⨁⨁◯ MODERATEa	78 (1)	1.54 (0.31–7.58)	13%	18.7%	⨁◯◯◯ VERY LOWa,f
			11.5% increase in mortality. Between 4.8% more and 20.4% more					5.7% increase in severe COVID-19 disease. Between 8.6% less and 40.1% more
Dyspnea Definition: Dyspnea or shortness of breath	6613 (28)	3.45 (2.72–4.38)	4.9%	13.8%	⨁⨁⨁⨁ HIGH	16803 (78)	4.23 (3.32–5.38)	9.3%	27.8%	⨁⨁◯◯ LOWa,d
			8.9% increase in mortality. Between 7.5% more and 10.2% more					18.5% increase in severe COVID-19 disease. Between 15.4% more and 21.3% more
Anorexia	1483 (8)	2.16 (1.14–4.12)	7.3%	14.4%	⨁⨁⨁◯ MODERATEc	5495 (26)	2.86 (2.16–3.84)	10.4%	24%	⨁⨁⨁◯ MODERATEa
			7% increase in mortality. Between 1.1% more and 13.1% more					13.6% increase in severe COVID-19 disease. Between 9.8% more and 17.5% more
Tachypnea Definition: More than 20–24 bpm	202 (1)	1.21 (1.12–1.31)	7.6%	9%	⨁⨁⨁◯ MODERATEa	518 (7)	7.51 (1.66–33.91)	13%	52.9%	⨁⨁⨁◯ MODERATEd
			1.4% increase in mortality. Between 0.9% more and 1.9% more					39.9% increase in severe COVID-19 disease. Between 6.9% more and 70.5% more
Haemoptysis	781 (5)	2.91 (0.74–11.4)	8.3%	20.6%	⨁⨁◯◯ LOWa,b	3317 (14)	4.39 (2.18–8.81)	12.7%	38.6%	⨁⨁⨁◯ MODERATEa
			12.3% increase in mortality. Between 2.2% less and 34.6% more					25.9% increase in severe COVID-19 disease. Between 11.4% more and 42.1% more
Abdominal pain	1127 (5)	1.06 (0.53–2.11)	9%	9.5%	⨁⨁◯◯ LOWa,b	4896 (22)	1.95 (1.36–2.79)	12.7%	22%	⨁⨁⨁◯ MODERATEa
			5% increase in mortality. Between 4% less and 8% more					9.4% increase in severe COVID-19 disease. Between 4% more and 15.8% more
Fatigue	3725 (21)	1.66 (1.27–2.17)	7.1%	11.1%	⨁⨁◯◯ LOWa,d	13262 (71)	1.41 (1.19–1.68)	11.6%	15.5%	⨁⨁⨁◯ MODERATEd
			4% increase in mortality. Between 1.9% more and 5.9% more.					3.9% increase in severe COVID-19 disease. Between 2% more and 5.9% more.
Fever. Definition: More than 37.5°C	6154 (31)	1.04 (0.77–1.4)	8.7%	9.1%	⨁⨁◯◯ LOWa,b	20026 (102)	1.84 (1.54–2.21)	9.3%	15.4%	⨁⨁⨁◯ MODERATEd
			0.3% increase in mortality. Between 2.3% less and 2.5% more					6.1% increase in severe COVID-19 disease. Between 4.5% more and 7.6% more
Myalgia/arthralgia Definition: myalgia or arthralgias	3436 (18)	0.96 (0.77–1.23)	9.1%	8.7%	⨁⨁◯◯ LOWa,b	13814 (61)	1.29 (1.03–1.61)	12.5%	15.6%	⨁⨁⨁◯ MODERATEd
			0.3% decrease in mortality. Between 2% less and 1.8% more					3% increase in severe COVID-19 disease. Between 0.3% more and 5.9% more
Laboratory measures (blood or plasma)
High procalcitonin Definition: More than 0.01–05 ng/ml	4735 (10)	12.42 (7.18–21.5)	6.3%	38.5%	⨁⨁⨁◯ MODERATEc	7923 (28)	5.13 (3.16–8.35)	10.6%	35.4%	⨁⨁◯◯ LOWa,d
			32.3% increase in mortality. Between 25% more and 38.1% more					24.8% increase in severe COVID-19 disease. Between 16.7% more and 32.3% more
Myocardial injury Definition: Reported as myocardial injury or as increase in blood troponins	3855 (21)	10.89 (5.39–22.04)	3.5%	20.4%	⨁⨁⨁◯ MODERATEd	3627 (20)	10 (6.84–14.62)	11.1%	51.3%	⨁⨁⨁⨁ HIGH
			16.9% increase in mortality. Between 13.4% more and 19% more					40.2% increase in severe COVID-19 disease. Between 33.1% more and 46.4% more
High WBC Definition: greater than 10.0 x 109/L	2870 (10)	4.06 (2.7–6.12)	7.8%	24.7%	⨁⨁⨁◯ MODERATEd	9331 (32)	4.67 (3.17–6.88)	11.2%	35.6%	⨁⨁⨁⨁ HIGH
			16.9% increase in mortality. Between 11% more and 23.3% more					24.3% increase in severe COVID-19 disease. Between 17.3% more and 31.2% more
High lactate: Definition More than 1.5–2.2 mmol/L	1078 (1)	3.66 (2.26–5.94)	7.3%	21.7%	⨁⨁⨁◯ MODERATEa	812 (3)	3.74 (0.69–20.16)	12.1%	33.3%	⨁◯◯◯ VERY LOW a,b,d
			14.3% increase in mortality. Between 8.3% more and 20.6% more					21.2% increase in severe COVID-19 disease. Between 3.7% less and 51.4% more
Low platelet count Definition: Less than 100–150 x 109/L	3676 (10)	5.43 (2.55–11.56)	5%	19.3%	⨁⨁⨁⨁ HIGH	8081 (32)	1.93 (1.52–2.46)	11.1%	19.2%	⨁⨁◯◯ LOWa,d
			14.3% increase in mortality. Between 8.3% more and 18.6% more					8% increase in severe COVID-19 disease. Between 5% more and 11.1% more
High D-dimer Definition: More than 500–1000 ng/ml	4361 (17)	4.81 (3.15–7.34)	4.3%	15.6%	⨁⨁⨁◯ MODERATEd	6356 (24)	3.27 (2.46–4.36)	8.2%	20.7%	⨁⨁⨁◯ MODERATEd
			11.2% increase in mortality. Between 8.8% more and 13.1% more					12.5% increase in severe COVID-19 disease. Between 9.8% more and 14.8% more
High LDH Definition: More than 240–250 U/L	1440 (6)	4.09 (1.18–14.17)	4.7%	15.2%	⨁⨁⨁◯ MODERATEd	7955 (26)	4.48 (3.21–6.25)	7.8%	23.9%	⨁⨁⨁◯ MODERATEd
			10.4% increase in mortality. Between 1.4% more and 15.3% more					16.2% increase in severe COVID-19 disease. Between 13.1% more and 18.8% more
High CRP Definition: More than 1–100 mg/l	2107 (8)	6.6 (3.36–12.99)	2.3%	10.3%	⨁⨁⨁◯ MODERATEd	9094 (37)	4.5 (3.1–6.23)	6.3%	19.5%	⨁⨁⨁⨁ HIGH
			7.9% increase in mortality. Between 6.4% more and 8.7% more					13.2% increase in severe COVID-19 disease. Between 10.8% more and 14.9% more
Decrease in Lymphocyte count Definition: per 1 x 109 U/L decrease	544(3)	3.57 (2–6.67)	9%	26.1%	⨁⨁⨁◯ MODERATEd	1909 (7)	2.28 (1.21–4.30)	13%	25.4%	⨁⨁⨁◯ MODERATEd
			17.1% increase in mortality. Between 7.5% more and 30.7% more					12.4% increase in severe COVID-19 disease. Between 2.3% more and 26.1% more
High AST level Definition: More than 32–40 U/l	2969 (7)	3.5 (1.59–7.71)	6%	17.1%	⨁⨁⨁◯ MODERATEd	9179 (32)	3.41 (2.7–4.3)	9.9%	25.8%	⨁⨁⨁◯ MODERATEa
			11.1% increase in mortality. Between 4% more and 16.8% more					15.8% increase in severe COVID-19 disease. Between 12.7% more and 18.8% more
Decrease in albumin: Definition: 20 g/L decrease	336 (3)	1.53 (1.32–1.78)	9%	13.2%	⨁⨁⨁◯ MODERATEc	1266 (5)	1.11 (1.01–1.21)	13%	14.2%	⨁⨁⨁◯ MODERATEb
			4.2% increase in mortality. Between 2.5% more and 6% more					1.2% increase in severe COVID-19 disease. Between 0.1% more and 2.3% more
Increase in creatinine Definition: per 1 mg/dL increase	1508 (9)	1.14 (1.02–1.28)	9%	10.1%	⨁⨁⨁◯ MODERATEb	1116 (4)	1.89 (0.87–4.10)	13%	22%	⨁⨁⨁◯ MODERATEb
			1.1% increase in mortality. Between 0.2% more and 2.3% more					9% increase in severe COVID-19 disease. Between 1.5% less and 25% more
High Neutrophil count Definition: greater than 6.3 x 109/L	727 (2)	6.78 (3.07–14.97)	5.2%	23%	⨁⨁◯◯ LOWa,c	4945 (16)	5.66 (3.71–8.63)	9%	31%	⨁⨁⨁◯ MODERATEa
			17.8% increase in mortality. Between 10% more and 23% more					22% increase in severe COVID-19 disease. Between 17% more and 27% more
High BNP: More than 500–900 pg/mL	1283 (6)	3.27 (1.24–8.63)	7%	19%	⨁⨁◯◯ LOWa,d	1086 (1)	4.99 (3.2–7.77)	9.4%	30.9%	⨁⨁⨁◯ MODERATEa
			12% increase in mortality. Between 1.9% more and 21.9% more					21.5% increase in severe COVID-19 disease. Between 15.5% more and 26.7% more
High BUN Definition: mmol/L, > 5.2–9.5	1258 (2)	10.56 (6.76–16.48)	5.2%	29.6%	⨁⨁◯◯ LOWa,c,e	3890 (10)	3.66 (2.82–4.74)	11.1%	30.2%	⨁⨁⨁◯ MODERATEa
			24.4% increase in mortality. Between 20.2% more and 27.7% more					19.1% increase in severe COVID-19 disease. Between 14.8% more and 23.4% more
High CPK Definition: More than 185–200 U/L	407 (3)	1.35 (0.58–3.14)	8.8%	11.5%	⨁⨁◯◯ LOWa,b	3292 (13)	3.1 (2.32–4.16)	11.5%	28.1%	⨁⨁⨁◯ MODERATEa
			2.7% increase in mortality. Between 3.7% less and 13% more					16.5% increase in severe COVID-19 disease. Between 11.7% more and 21.6% more
High total bilirubin Definition: More than 17–21 pg/ml	2715 (3)	3.03 (1.87–4.92)	8.1%	20.7%	⨁⨁◯◯ LOWa,c	5098 (14)	2.94 (2.18–3.97)	12.5%	29.3%	⨁⨁⨁◯ MODERATEa
			12.6% increase in mortality. Between 6.3% more and 19.9% more					16.8% increase in severe COVID-19 disease. Between 11.3% more and 22.9% more
High interleukin-6 Definition: More than 5–20 pg/ml	436 (4)	1.31 (0.14–12.27)	8.1%	10.3%	⨁⨁◯◯ LOWa,b	1211 (7)	7.36 (2.97–18.27)	6.5%	26.2%	⨁⨁⨁◯ MODERATEa
			2.2% increase in mortality. Between 11.6% less and 15% more					19.7% increase severe COVID-19 disease. Between 12.2% more and 23.9% more
High Definition More than 10–20 mm/H	628 (3)	0.89 (0.54–1.45)	9.7%	8.7%	⨁⨁◯◯ LOWa,b	2557 (12)	3.08 (2.04–4.65)	6.6%	15.6%	⨁⨁⨁◯ MODERATEa
			1% decrease in mortality. Between 5.6% less and 2.8% more					9.4% increase in severe COVID-19 disease. Between 6.7% more and 11.3% more
Radiological signs
Pleural effusion Definition: X ray or CT assessment	820 (5)	1.38 (0.63–3.06)	8.8%	11.7%	⨁⨁◯◯ LOWa,b	5289 (23)	3.31 (2.03–5.38)	12.5%	32%	⨁⨁⨁◯ MODERATEd
			3% increase in mortality. Between 3% less and 13% more					19% increase in severe COVID-19 disease. Between 10% more and 30% more
Consolidation pattern Definition: X ray or CT assessment	795 (4)	1.93 (1.31–2.84)	7.5%	13%	⨁⨁◯◯ LOWa,c	6133 (27)	2.46 (1.54–3.93)	11.2%	23.2%	⨁⨁⨁◯ MODERATEd
			5.8% increase in mortality. Between 2.3% more and 9.4% more					12% increase in severe COVID-19 disease. Between 5.4% more and 18.8% more
Others
High SOFA score Definition: More than 2	585 (3)	1.97 (1.22–3.2)	9%	16.3%	⨁⨁⨁◯ MODERATEd	92 (2)	21.31 (6.26–72.6)	13%	76.1%	⨁⨁◯◯ LOWa,c
			7.3% increase in mortality. Between 1.8% more and 15% more					63% increase in severe COVID-19 disease. Between 35.3% more and 78.6% more

Glossary

FDP: Fibrin Degradation Product.

PT: Prothrombin time.

APTT: Activated partial thromboplastin time.

APACHE: Acute Physiology And Chronic Health Evaluation II.

SOFA: The sequential organ failure assessment score.

qSOFA: Quick sepsis related organ failure assessment.

AST: Aspartate aminotransferase.

ALT: Alanine aminotransferase.

BUN: Blood urea nitrogen.

NA: Not applicable, either because there is no information or because the addressed variable does not represent a potential prognostic factor in that clinical scenario.

Explanations

a. Risk of bias due to study limitations (unadjusted estimates, inappropriate prognostic factor or outcome assessment, inappropriate population inclusion criteria, study attrition).

b. Imprecision: Confidence interval includes significant and non-significant risk increase.

c. Imprecision due to fragility: Less than 200 events.

d. Inconsistency: Unexplained visual heterogeneity.

e. Risk of selective reporting: Most of the pooled estimate weight from studies that performed multivariable analysis but did not report adjusted estimates.

f. Very serious imprecision: Very wide confidence interval.

Demographic factors

Age per 10 years increase (OR 1.8, 95% CI 1.54 to 2.1; RD 6.1%, 95% CI 4.2 to 8.2%), male sex (OR 1.72, 95% CI 1.5 to 1.98; RD 5%, 95% CI 4 to 7%) and active smoker (OR 1.57, 95% CI 1.19 to 2.07; RD 4.3%, 95%CI 1.5 to 7.5%).

Medical illness and patient history factors

Any chronic condition or comorbidity (OR 3.3, 95% CI 2.18 to 5; RD 10.3%, 95% CI 6.8 to 13.4%), cerebrovascular disease (OR 2.85, 95% CI 2.02 to 4.01; RD 12.6%, 95% CI 7.5 to 18.5%), chronic obstructive pulmonary disease (COPD) (OR 2.43, 95% CI 1.88 to 3.14; RD 9.8%, 95% CI 6.4 to 13.6%), chronic kidney disease (CKD), (OR 2.27, 95% CI 1.69 to 3.05; RD 8.8%, 95% CI 5.1 to 12.9%), cardiovascular disease (defined as coronary heart disease and/or cardiac failure) (OR 2.12, 95% CI 1.77 to 2.56; RD 7.5%, 95% CI 5.4 to 9.7%), cardiac arrhythmia (OR 2.13, 95% CI 1.72 to 2.65; RD 6.5%, 95% CI 4.7 to 8.4%), arterial hypertension (OR, 2.02, 95% CI 1.71 to 2.38; RD 6%, 95% CI 4.5 to 7.33), diabetes (OR 1.84, 95% CI 1.61 to 2.1; RD 5.6%, 95% CI 4.3 to 7%), dementia (OR 1.54, 95% CI 1.31 to 1.81; RD 4.2%, 95% CI 2.5 to 6.2%), obesity (OR 1.41, 95% CI 1.15 to 1.74; RD 3.1%, 95% CI 1.2 to 5.1%), cancer (OR 1.35, 95% CI 1.17 to 1.55; RD 2.7%, 95% CI 1.4 to 4.2) and dyslipidemia (OR 1.26, 95%CI 1.06–1.5; RD 2.1%, 95%CI 0.5% to 3.9%).

Symptoms, vital signs and physical examination factors

Respiratory failure (OR 21.17, 95% CI 4.9 to 91.3; RD 20.3%, 95% CI 13.4% to 22.4%), low blood pressure (OR 6.7, 95% CI 3.14 to 14.33; RD 30.9%, 95% CI 14.7% to 49.6%), hypoxemia (OR 5.46, 95% CI 2.05 to 14.53; RD 6.7%, 95% CI 4.2% to 7.7%), tachycardia (OR 2.61, 95% CI 1.62 to 4.22; RD 11.5%, 95% CI 4.8% to 20.4%), dyspnea (OR 3.45, 95% CI 2.72 to 4.38; RD 8.9%, 95% CI 7.5% to 10.2%), anorexia (OR 2.16, 95% CI 1.14 to 4.12; RD 7%, 95% CI 1.1% to 13.1%) and tachypnea (OR 1.21, 95%CI 1.12 to 1.31; RD 1.4%, 95% CI 0.9% to 1.9%).

Laboratory factors (measured in blood or plasma)

High procalcitonin (OR 12.42, 95% CI 7.18 to 21.5; RD 32.3%, 95% CI 25% to 38.1%), myocardial injury markers (OR 10.89, 95% CI 5.39 to 22.04; RD 16.9%, 95% CI 13.4% to 19%), high white cell count (WBC) (OR 4.06, 95% CI 2.7 to 6.12; RD 16.9%, 95% CI 11% to 23.3%), high lactate (OR 3.66, 95% CI 2.26 to 5.94; RD 14.3%, 95% CI 8.3% to 20.6%), low platelet count (OR 5.43, 95% CI 2.55 to 11.56; RD 14.3%, 95% CI 8.3% to 18.6%), high D-dimer (OR 4.81, 95% CI 3.15 to 7.34; RD 11.2%, 95% CI 8.8% to 13.1%), high lactate dehydrogenase (LDH) (OR 4.09, 95% CI 1.18 to 14.17; RD 10.4%, 95% CI 1.4% to 15.3%), high c-reactive protein (CRP) (OR 6.6, 95% CI 3.36 to 12.99; RD 7.9%, 95% CI 6.4% to 8.7%), decrease in lymphocyte count (OR 3.57, 95% CI 2 to 6.67; RD 17.1%, 95% CI 7.5% to 30.7%), high aspartate aminotransferase (AST) (OR 3.5, 95% CI 1.59–7.71; RD 11.1%, 95% CI 4% to 16.8%), albumin increase (OR 1.53, 95% CI 1.32 to 1.78; RD 4.2%, 95% CI 2.5% to 6%) and creatinine increase (OR 1.14, 95%CI 1.02 to 1.28; RD 1.1%, 95% CI 0.2% to 2.3%).

Others

SOFA score> 2 (OR 1.97, 95% CI 1.22 to 3.2; RD 7.3%, 95% CI 1.8% to 15%).

Prognostic factors for severe COVID-19 disease

We investigated 96 candidate prognostic factors for severe COVID-19 disease from 131 studies including 28538 patients. S3 Table provides a summary of findings for all the candidate prognostic factors and S2 Appendix includes the corresponding forest plots.

In addition to identified prognostic factors for mortality, we found high or moderate certainty that the following 14 variables provide valuable prognostic information on severe COVID-19 disease outcome (Table 1).

Haemoptysis (OR 4.39, 95% CI 2.18 to 8.81; RD 25.9%, 95% CI 11.4% to 42.1%), abdominal pain (OR 1.95, 95% CI 1.36 to 1.79; RD 9.4%, 95% CI 4% to 15.8%), fatigue (OR 1.41, 95% CI 1.19 to 1.68; RD 3.9%, 95% CI 2% to 5.9%), fever (OR 1.84, 95% CI 1.54 to 2.21; RD 6.1%, 95% CI 4.5% to 7.6%) and myalgia or arthralgia (OR 1.29, 95% CI 1.03 to 1.61; RD 3%, 95%CI 0.3% to 5.9%).

High neutrophil count (OR 5.66, 95% CI 3.71 to 8.63; RD 22%, 95% CI 17% to 27%), high B-type natriuretic peptide (BNP) (OR 4.99, 95% CI 3.2 to 7.77; RD 21.5%, 95% CI 15.5% to 26.7%), High urea nitrogen (BUN) (OR 3.66, 95% CI 2.82 to 4.74; RD 19.1%, 95% CI 14.8% to 23.4%), high creatine kinase (CK) (OR 3.1, 95% CI 2.32 to 4.16; RD 16.5%, 95% CI 11.7% to 21.6%), high bilirubin (OR 2.94, 95% CI 2.18 to 3.97; RD 16.8%, 95% CI 11.3% to 22.9%), high interleukin-6 (IL-6) (OR 7.36, 95% CI 2.97 to 18.27; RD 13.3%, 95% CI 8.5% to 15.9%), high erythrocyte sedimentation rate (ESR) (OR 3.08, 95% CI 2.04 to 4.65; RD 9.4%, 95% CI 6.7% to 11.3%).

Radiological factors

Consolidative infiltrate (OR 2.46, 95% CI 1.54 to 3.93; RD 12%, 95% CI 5.4% to 18.8%) and pleural effusion (OR 3.31, 95% CI 2.03 to 5.38; RD 19%, 95% CI 10% to 30%).

Other analysed variables

The remaining variables analysed were: asthma, tuberculosis, HIV infection, immunocompromise, autoimmune disease, malnutrition, chronic liver disease, thyroid disease, chronic gastric disease, chest pain, high fever, cough, rhinorrhea, odynophagia, conjunctivitis, sputum production, enlarged lymph nodes, rash, headache, vomits, diarrhea, anemia, low WBC, low neutrophil count, glomerular filtration rate, blood urea, cystatin C, prothrombin time, APTT time, ferritin, cholinesterase, alanine aminotransferase (ALT), fibrinogen degradation products, globulin, prealbumin, blood glucose, alfa-HBDH, low density lipoprotein (LDL), triglycerides, any abnormal radiologic finding, radiological interstitial pattern, ground glass opacity, crazy paving pattern, radiological evidence of enlarged lymph nodes, bilateral radiological compromise, APACHE (Acute Physiology And Chronic Health Evaluation II), qSOFA (quick sepsis related organ failure assessment).

For all of these variables we found low or very low certainty evidence both for mortality and severe COVID-19 disease. Hence, it is uncertain if these variables provide prognostic value in the context of COVID-19 infected patients.

Additional analysis

We performed a sensitivity analysis on mortality outcome using a non-contextualized approach and assuming adjusted estimates as at low risk of being biased (less demanding risk of bias approximation in comparison with the primary analysis) (see methods, risk of bias assessment in S1 Text). The results were similar to the primary analysis. However our certainty increased to moderate or high for some prognostic factors for which we had very low or low certainty: Chest pain, cough, sputum production, anemia, high ferritin, high ALT, increase in blood glucose and high APACHE score.

A second sensitivity analysis in which we set a significantly higher baseline mortality risk (26%) for patients with severe but non-critical COVID-19 disease did not show differences with the primary analysis.

When we found inconsistent results for severe COVID-19 disease outcome (n = 43), we performed subgroup analyses accounting on outcome definition. Observed heterogeneity could not be explained by this analysis for any of the candidate prognostic factors explored (see S3 Appendix).

Discussion

In this systematic review we evaluated prognostic factors for poor outcome in patients with covid-19 infectious disease. We found 49 variables that provide valuable prognostic information for mortality and/or severe COVID-19 disease. Identified prognostic factors include socio-demographic characteristics (age, male sex and smoking) medical illness and patients history information (comorbidities including chronic respiratory, cardiac and endocrinologic conditions), physical examination findings (respiratory failure related symptoms as well as general clinical condition deterioration), laboratory (multiple biomarkers and alterations in basic laboratory tests) and radiological findings (consolidation pattern and pleural effusion) (Table 1). Overall the risk of severe COVID-19 disease or death resulted higher in older patients and those with previous medical conditions including COPD and cardiovascular disease as some of the most relevant predictors. Additionally, those patients that presented with clinical signs and symptoms suggesting respiratory failure or laboratory biomarkers showing inflammation or organ damage were also at increased risk of severe COVID-19 disease or death. Radiological features did not show good predictive value.

Strengths and limitations of the study

Our systematic review has a number of strengths. First, it provides the most comprehensive and trustworthy body of evidence up to date as it includes a significant number of studies not included in prior reviews. Secondly, we followed the GRADE approach to summarize and rate the certainty on the evidence. And thirdly, we presented our results both as relative estimates of association as well as absolute risk differences and used the latter to interpret and analyse our results. We consider that the absolute risk modification that can be attributed to a prognostic factor is a critical piece of information for those aiming to make decisions using prognostic information.

Regarding limitations, most of the studies included in this review were not published in peer review journals (only as preprint) at the time we performed the search. We identified most of those studies by cross reference search in google scholar, but it is possible that some may have not been detected by our search strategy. Additionally, given the high publication speed of COVID-19 studies it is probable that new relevant information not included in our review is available at the time our review is published. We aim to address this issue by updating our results in the short term. Although we made efforts to identify data duplication, in many instances it was not clear if studies reported, totally or partially, on the same cohorts of patients hence we assume there is a considerable chance of some degree of data overlap between included publications.

Significant variability in study design, study type, patient eligibility criteria, prognostic factor definition was observed, however, given the huge amount of information analysed, it was not feasible to explore subgroup effects accounting for those differences.

In analysing our results we implemented a minimally contextualized approach for which we arbitrarily set thresholds to define the minimal important risk difference necessary to assume valuable prognostic information. As the degree of contextualization was minimal, we set very low thresholds (near the point of no effect). We acknowledge that readers might find those thresholds inappropriate, hence we also provided relative estimates of association which can be used with alternative thresholds or analytical approximations (e.g partially contextualized approach). In addition, for some candidate prognostic factors, baseline risk could not be adjusted for prevalence which might have resulted in an overestimation or underestimation of the risk difference estimates. Finally, for risk of bias assessment, we defined a set of requirements for dealing with potential confounders that were not previously validated.

Relation to prior work

We identified multiple systematic reviews addressing prognostic factors in patients with COVID-19 infectious disease [232-251]. All analysed certain prognostic factors or groups or prognostic factors that we included in the present review, and measured mortality and/or disease severity as outcomes. Most of the reported results are in consonance with our findings with only a few exceptions. Kumar et.al [246] reported diarrhea, productive cough and high ALT as prognostic factors however we found low certainty evidence on those variables in our primary analysis. Wang et.al [249] reported no association between chronic kidney disease and malignancy with poor outcomes in COVID-19 patients however we found that both conditions are associated with an increased risk of mortality and severe COVID-19 disease. Other significant differences of our review in relation to these prior reviews include multiple characteristics that were previously not identified as prognostic factors. In contrast to previous reviews, here we provide both relative and absolute estimates of risk and provide our certainty in those estimates.

Significant information has been published since our search was finalized. An update of the ISARIC registry [252] including 15194 hospitalised patients discharged or dead, the openSAFELY registry [253] included 17425445 adults potentially exposed to COVID-19 infection and a Chinese registry [254] that included 44672 patients with COVID-19.

These studies identified the following variables as prognostic factors for COVID-19 related mortality: Age, sex (male), obesity, cardiovascular disease, diabetes, arterial hypertension, dyslipidemia, COPD, smoking, malignancy, cerebrovascular disease, dementia and chronic kidney disease. All these variables were captured by our analysis as predictors of COVID-19 related mortality or severe COVID-19 disease for which moderate or high certainty evidence exists. In addition, other prognostic factors were identified by these studies: race (not white) [253] and deprivation [253], two variables we did not explore, and immunocompromise [253], asthma [253], autoimmune diseases [253], and chronic liver disease [252,253], four variables for which we found low certainty evidence.

Implications of study

Our approach considered two clinical scenarios in which we assumed that the prognostic information of each predictor for each outcome could potentially impact decision-making. In patients presenting with mild disease, predicting the risk of progression to severe status could support decisions on the level of healthcare required and more extensive follow up strategies. In the same way, in patients presenting with severe disease, predicting mortality risk could support the use of certain, more aggressive, therapeutic interventions. Clinicians or decision-makers can use our results to tailor management strategies for patients with COVID-19. For example, they could select a set of prognostic factors for which there is high certainty in a significant risk incremental increase (e.g Age, gender, comorbidities, respiratory failure and myocardial injury) and use them to define hospitalization rules for patients consulting to the emergency department. However, to what extent accounting for these prognostic factors will improve clinically important outcomes is a question that cannot be addressed with our results. Furthermore using information on multiple individual prognostic factors for outcome prediction is challenging. Multivariable models provide a solution to this limitation, however, considering the high demand for accurate risk prediction models for patients with COVID-19 [7], our work can also provide solid grounds for development of these prognostic tools.

Conclusions

We have identified a set of variables that provide valuable prognostic information in patients with COVID-19 infectious disease. Clinicians and policy makers can use our results to tailor management strategies for patients with this condition while researchers can utilise our findings to develop multivariable prognostic models that could eventually facilitate decision-making and improve patient important outcomes.

Included studies characteristics.

This table presents detailed information on individual included studies.

(PDF)

Click here for additional data file.

Risk of Bias (RoB) assessment.

This table contains a detailed RoB assessment of included studies.

(PDF)

Click here for additional data file.

Summary of findings of all candidate variables.

This table presents the complete results including all assessed candidate variables.

(DOCX)

Click here for additional data file.

Supplementary methods.

This file contains additional details on methods.

(DOCX)

Click here for additional data file.

PRISMA document.

This file contains the references of the pages according to PRISMA Statement.

(DOCX)

Click here for additional data file.

Mortality forest plots.

This file contains the Forest plots for all assessed candidate variables.

(PDF)

Click here for additional data file.

COVID-19 severe disease forest plots.

This file contains the Forest plots for all assessed candidate variables.

(PDF)

Click here for additional data file.

Subgroup analyses forest plots.

This file contains the Forest plots for all performed subgroup analyses.

(PDF)

Click here for additional data file.

18 Sep 2020

PONE-D-20-20042

Prognostic factors for severity and mortality in patients infected with COVID-19: A systematic review

PLOS ONE

Dear Dr. Izcovich,

Thank you for submitting your manuscript to PLOS ONE. After careful consideration, we feel that it has merit but does not fully meet PLOS ONE’s publication criteria as it currently stands. Therefore, we invite you to submit a revised version of the manuscript that addresses the points raised during the review process.

Please submit your revised manuscript by Nov 02 2020 11:59PM. If you will need more time than this to complete your revisions, please reply to this message or contact the journal office at plosone@plos.org. When you're ready to submit your revision, log on to https://www.editorialmanager.com/pone/ and select the 'Submissions Needing Revision' folder to locate your manuscript file.

Please include the following items when submitting your revised manuscript:

A rebuttal letter that responds to each point raised by the academic editor and reviewer(s). You should upload this letter as a separate file labeled 'Response to Reviewers'.

A marked-up copy of your manuscript that highlights changes made to the original version. You should upload this as a separate file labeled 'Revised Manuscript with Track Changes'.

An unmarked version of your revised paper without tracked changes. You should upload this as a separate file labeled 'Manuscript'.

If you would like to make changes to your financial disclosure, please include your updated statement in your cover letter. Guidelines for resubmitting your figure files are available below the reviewer comments at the end of this letter.

If applicable, we recommend that you deposit your laboratory protocols in protocols.io to enhance the reproducibility of your results. Protocols.io assigns your protocol its own identifier (DOI) so that it can be cited independently in the future. For instructions see: http://journals.plos.org/plosone/s/submission-guidelines#loc-laboratory-protocols

We look forward to receiving your revised manuscript.

Kind regards,

Chiara Lazzeri

Academic Editor

PLOS ONE

Journal Requirements:

When submitting your revision, we need you to address these additional requirements.

1. Please ensure that your manuscript meets PLOS ONE's style requirements, including those for file naming. The PLOS ONE style templates can be found at

https://journals.plos.org/plosone/s/file?id=wjVg/PLOSOne_formatting_sample_main_body.pdf and

https://journals.plos.org/plosone/s/file?id=ba62/PLOSOne_formatting_sample_title_authors_affiliations.pdf

2. We note that you have indicated that data from this study are available upon request. PLOS only allows data to be available upon request if there are legal or ethical restrictions on sharing data publicly. For information on unacceptable data access restrictions, please see http://journals.plos.org/plosone/s/data-availability#loc-unacceptable-data-access-restrictions.

In your revised cover letter, please address the following prompts:

a) If there are ethical or legal restrictions on sharing a de-identified data set, please explain them in detail (e.g., data contain potentially identifying or sensitive patient information) and who has imposed them (e.g., an ethics committee). Please also provide contact information for a data access committee, ethics committee, or other institutional body to which data requests may be sent.

b) If there are no restrictions, please upload the minimal anonymized data set necessary to replicate your study findings as either Supporting Information files or to a stable, public repository and provide us with the relevant URLs, DOIs, or accession numbers. Please see http://www.bmj.com/content/340/bmj.c181.long for guidelines on how to de-identify and prepare clinical data for publication. For a list of acceptable repositories, please see http://journals.plos.org/plosone/s/data-availability#loc-recommended-repositories.

We will update your Data Availability statement on your behalf to reflect the information you provide.

3. Please include captions for your Supporting Information files at the end of your manuscript, and update any in-text citations to match accordingly. Please see our Supporting Information guidelines for more information: http://journals.plos.org/plosone/s/supporting-information.

[Note: HTML markup is below. Please do not edit.]

Reviewers' comments:

Reviewer's Responses to Questions

Comments to the Author

1. Is the manuscript technically sound, and do the data support the conclusions?

The manuscript must describe a technically sound piece of scientific research with data that supports the conclusions. Experiments must have been conducted rigorously, with appropriate controls, replication, and sample sizes. The conclusions must be drawn appropriately based on the data presented.

Reviewer #1: Yes

Reviewer #2: Yes

**********

2. Has the statistical analysis been performed appropriately and rigorously?

Reviewer #1: Yes

Reviewer #2: Yes

**********

3. Have the authors made all data underlying the findings in their manuscript fully available?

The PLOS Data policy requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified.

Reviewer #1: Yes

Reviewer #2: Yes

**********

4. Is the manuscript presented in an intelligible fashion and written in standard English?

PLOS ONE does not copyedit accepted manuscripts, so the language in submitted articles must be clear, correct, and unambiguous. Any typographical or grammatical errors should be corrected at revision, so please note any specific errors here.

Reviewer #1: Yes

Reviewer #2: Yes

**********

5. Review Comments to the Author

Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters)

Reviewer #1: First of all one has to applaud the authors for the sheer amount of data, they have provided in this systematic review! Generaly the manuscript is well writen, but still contains a few typing errors. Some formulations formulations are a matter of taste. Nevertheless, I would recommend to change them (detailed information can be found below).

Abstract:

Since you are describing/systematically reviewing prognostic factors, I would go so far and claim that it influences decision making! Prognostic factors are importnat as well, but they are not predictive factors. Therefore, developing a predictive model (would call it multivariate model not multivariable model) to estimate disease course, severity and mortality is, what the data support.

Since arterial hypertension, and cardiac arrythmias are mentioned seperately, it should be defined, what cardiovascular disease includes. Basically all laboratory test you mention are tested in the blood/serum or plasma. Please omit using blood ... (e.g. WBC instead of blood WBC). It is heavily affecting the fluidity of the text and adds no information. Propably it makes sense to define it at one point. Ultimately, it is also not consistant through the text/tables.

Covid-19 is defined as an infectious disease (does not include asymptomatic SARS-CoV-2 infections). Please change infection to infectious disease.

Introduction:

Please cite the RECOVERY trial (contains important and prospective reference information regarding mortality rate and the rate of severe Covid-19 disease). "Can may be used" (page 8) can be formulated a bit more eloquent (e.g. have to potential to be used/applied or can be potentially used). What is mentioned above about decision making applies also here.

Methods:

The sentence on page 17, chapter 1, line 3 is difficult to understand. "Exposed" is not the correct wording here. The biomarkers/clinical factors/symptoms/etc. are absent or present. It is also not clear if the biomarkers are measured/determined at baseline/SARS-CoV-2 diagnosis or crossed/never crossed the treshold during the complete disease until recovery. Please be more specific in your description. This is essential information!

On page 17, chapter 2, line 5 please change studies results to study results.

That you promote an update (AI based) is a huge plus point and might already be necessary!

The mortality has currently dropped to 5%. Your reference value is no longer up to date!!!

On page 13, line 2 please change "inseverty" to "in severity".

Although the cut-offs used are arbitrary, they seem to be weel chosen (e.g. 0.5% in overall mortality is one tenth of the current overall mortality).

Results:

Please change "Physical examination factors" to a more appropriate term. Fatique is rather subjective and a symptom. Personally I am also not aware of a test for myalgia. Rewording is required for example to "symptoms, vital signs & abnormal physical examination results".

Laboratory factors: as mentioned above.

Some terms (e.g. immuncompromise) need a better definition (lymphopenia, immunosupressive medication, etc. ???).

Please mention why you didn't look at co-medication! This might be the only factor, which is relevant for decision making!!!

Overall it is an important review, which summarizes the current biomarker landscape for Covid-19. The cut-off date (End of April 2020) is very early and new discoveries have been made since. NLR and lymphopenia have been even published earlier. They should actually have met the inclusion criteria, but are not included (Lancet, Lancet Resp. Med. Feb 2020). Data have been homgenised and adjusted. The tables, figures and supplementary figures are very informative and have been nicely prepared!

Results:

The mortality risks

Reviewer #2: This study sets out to report a very important issue: prognostic factors that may be used in decision-making related to the care of patients infected with COVID-19. Although, this topic is not novel, the current research investigated it in a much more detail way than previous studies. In addition, authors also raised the possible questions that must be tackled in near future in order to develop multivariable prognostic models that could eventually facilitate decision-making and improve patients’ outcomes.

Reviewer’s comments:

1. The definitions of severe COVID-19 disease are adopted from multiple studies and therefore are inconsistent. How do the authors eliminate the effect of inconsistent definitions on the association analysis?

2. When severe COVID-19 disease was not reported as an outcome, the authors considered ICU requirement, invasive mechanical ventilation (IVM) and acute respiratory distress syndrome (ARDS) as surrogate outcomes. Did authors analyze the correlation between primary outcome and surrogate outcomes in this study?

3. The ICU requirement, invasive mechanical ventilation (IVM) and acute respiratory distress syndrome (ARDS) were used as surrogate outcomes. As a results, the outcomes in this study include mortality, severe COVID-19 disease, ICU requirement, invasive mechanical ventilation (IVM) and acute respiratory distress syndrome (ARDS). Multiple outcomes make the results difficult to be interpretated. If those outcomes could be analyzed separately, the results can be interpretated more precisely.

4. In the analysis of multivariable models, the variables used for model adjustment include age, one comorbidity (e.g diabetes) and one parameter of disease severity (e.g. respiratory rate) at minimum. How do the authors validate that this setting is statistically appropriate?

5. In this study, the authors arbitrarily set thresholds to define important incremental increase in the risk of their outcomes, including mortality or severe COVID-19. How do the authors validate that those thresholds can select proper prognostic information?

6. The thresholds were set in 0.5% increase in mortality and 1% increase in severe COVID-19 disease. Do the authors analyze the correlation between 0.5% increase in mortality and 1% increase in severe COVID-19 disease?

7. The authors used the clinical scenario of a patient infected with COVID-19 with severe but not critical disease to assess the prognostic value on mortality. Please clarify the definition of critical disease.

8. When prevalence of prognostic factors was not available, the authors used described baseline risks (9% for mortality and 13 % for severe COVID-19 disease). This setting will inevitably lead to bias in the results.

9. This study enrolled 207 studies and only 7 were judged as low risk of bias as the remaining presented important limitations in at least one domain or item. Hence, there are multiple inevitable bias in the current study, limiting the reliability of the results.

10. There is too much information in Table 1. Is there any way to condense the information or split it into different tables?

11. The authors described that clinicians can use their results to tailor management strategies for patients with COVID-19. It would be of great help to the reader if the authors could formulate a summarized management strategy based on the results of this study.

**********

6. PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy.

Reviewer #1: No

Reviewer #2: No

[NOTE: If reviewer comments were submitted as an attachment file, they will be attached to this email and accessible via the submission site. Please log into your account, locate the manuscript record, and check for the action link "View Attachments". If this link does not appear, there are no attachment files.]

While revising your submission, please upload your figure files to the Preflight Analysis and Conversion Engine (PACE) digital diagnostic tool, https://pacev2.apexcovantage.com/. PACE helps ensure that figures meet PLOS requirements. To use PACE, you must first register as a user. Registration is free. Then, login and navigate to the UPLOAD tab, where you will find detailed instructions on how to use the tool. If you encounter any issues or have any questions when using PACE, please email PLOS at figures@plos.org. Please note that Supporting Information files do not need this step.

21 Oct 2020

1. Please ensure that your manuscript meets PLOS ONE's style requirements, including those for file naming. The PLOS ONE style templates can be found at

https://journals.plos.org/plosone/s/file?id=wjVg/PLOSOne_formatting_sample_main_body.pdf and

https://journals.plos.org/plosone/s/file?id=ba62/PLOSOne_formatting_sample_title_authors_affiliations.pdf

We made several modifications to meet style requirements.

2. We note that you have indicated that data from this study are available upon request. PLOS only allows data to be available upon request if there are legal or ethical restrictions on sharing data publicly. For information on unacceptable data access restrictions, please see http://journals.plos.org/plosone/s/data-availability#loc-unacceptable-data-access-restrictions.

We now included all the forest plots in the appendices S5, S6 and S7

3. Please include captions for your Supporting Information files at the end of your manuscript, and update any in-text citations to match accordingly. Please see our Supporting Information guidelines for more information: http://journals.plos.org/plosone/s/supporting-information.

We included captions for supporting files as suggested.

4. Reviewer #1: First of all one has to applaud the authors for the sheer amount of data, they have provided in this systematic review! Generally the manuscript is well writen, but still contains a few typing errors. Some formulations formulations are a matter of taste. Nevertheless, I would recommend to change them (detailed information can be found below).

We want to thank the reviewers for taking the time and performing such a thorough review which will certainty result in significant improvements of our manuscript.

Abstract:

5. Since you are describing/systematically reviewing prognostic factors, I would go so far and claim that it influences decision making! Prognostic factors are importnat as well, but they are not predictive factors. Therefore, developing a predictive model (would call it multivariate model not multivariable model) to estimate disease course, severity and mortality is, what the data support.

We agree, but in the absence of well developed multivariate models clinicians and policymakers will need to rely on the best available evidence on individual prognostic factors, which we are providing in this review. We discuss this in page 23: “Implications of study”

6. Since arterial hypertension, and cardiac arrythmias are mentioned seperately, it should be defined, what cardiovascular disease includes. Basically all laboratory test you mention are tested in the blood/serum or plasma. Please omit using blood ... (e.g. WBC instead of blood WBC). It is heavily affecting the fluidity of the text and adds no information. Propably it makes sense to define it at one point. Ultimately, it is also not consistant through the text/tables.

In page 25 paragraph 2, line 6 we added: “(defined as coronary heart disease and/or cardiac failure)

On pages 26, paragraph 2, line 1 and 27 , paragraph 3, line 1 we added:”(measured in blood or plasma)”. We deleted the word blood or plasma in the rest of the manuscript accordingly.

7. Covid-19 is defined as an infectious disease (does not include asymptomatic SARS-CoV-2 infections). Please change infection to infectious disease.

We changed “infection” for “infectious disease” as suggested.

Introduction:

8. Please cite the RECOVERY trial (contains important and prospective reference information regarding mortality rate and the rate of severe Covid-19 disease). "Can may be used" (page 8) can be formulated a bit more eloquent (e.g. have to potential to be used/applied or can be potentially used). What is mentioned above about decision making applies also here.

In page 8, paragraph 1, line 9 we changed: “...provide a rigorous summary of the evidence available on prognostic factors that can be used in decision-making …” for “...provide a rigorous summary of the evidence available on prognostic factors that may be used in decision-making…”

In page 13, paragraph 1, line 6, we changed: “However, as we identified significant variability in mortality risks reported for similar clinical scenarios we performed a sensitivity analysis using a baseline risk of 26% as reported by a large cohort of non-ICU inpatients treated in 255 sites across 36 countries [6].” for “However, as we identified significant variability in mortality risks reported for similar clinical scenarios (i.e in the RECOVERY trial [20] mortality risk in hospitalized patients assigned to the control arm, with no baseline oxygen requirement was 14%), we performed a sensitivity analysis using a baseline risk of 26% as reported by a large cohort of non-ICU inpatients treated in 255 sites across 36 countries [6].

Methods:

9. The sentence on page 17, chapter 1, line 3 is difficult to understand. "Exposed" is not the correct wording here. The biomarkers/clinical factors/symptoms/etc. are absent or present. It is also not clear if the biomarkers are measured/determined at baseline/SARS-CoV-2 diagnosis or crossed/never crossed the treshold during the complete disease until recovery. Please be more specific in your description. This is essential information!

In page 9, paragraph 3, line 3 we changed: ”We investigated all prognostic factors reported by individual studies and compared patients exposed with patients unexposed to each one of those factors” for “We investigated all candidate prognostic factors reported by individual studies and compared patients exposed (the factor was present) with patients unexposed (the factor was absent) to each one of those factors”

We used the candidate prognostic factors determinations as reported by primary studies authors. How and when those measurements were performed, was a critical part of the risk of bias assessment. Studies in which candidate prognostic factors were not measured at baseline were judged as moderate or high risk of bias.

In page 10 paragraph 2, line 3, we changed: “ We used the Quality in Prognosis Studies (QUIPS) tool for prognostic factor studies [11].” for “ We used the Quality in Prognosis Studies (QUIPS) tool for prognostic factor studies [11] which considers population characteristics, attrition, prognostic factor and outcome measurement and potential residual confounding.”

In S1 Text, page 3, “Risk of bias”, we changed: “We used the Quality in Prognosis Studies tool (QUIPS) for prognostic factor studies.11 To be rated as low risk of bias studies needed to be prospective, have appropriately assessed prognostic factors and outcomes…” for “We used the Quality in Prognosis Studies tool (QUIPS) for prognostic factor studies.11 To be rated as low risk of bias studies needed to be prospective, have appropriately assessed prognostic factors (measured at baseline) and outcomes…”

10. On page 17, chapter 2, line 5 please change studies results to study results.

We made the suggested modification

11. That you promote an update (AI based) is a huge plus point and might already be necessary!

We acknowledge that an update will provide additional valuable information. However our review includes a significant body of evidence (207 studies including 75607 patients) and we found moderate/high certainty of the evidence for 49 prognostic factors. This means that it is implausible that additional information will modify our conclusions on those factors. We are planning an update which will be focused on the remaining 47 candidate variables for which we found low or very low certainty of the evidence.

12. The mortality has currently dropped to 5%. Your reference value is no longer up to date!!!

As we mention in the paper we found significant variability in mortality estimates. In order to avoid overestimating the predictive value of the analyzed variables, we decided to use a conservative one which also focused on our population of interest (severe but non-critical patients). We are not aware of a universal significant drop in mortality, for example ISARIC registry reported 26% mortality for non-ICU inpatients until May and 27% mortality for the same population until August (https://isaric.tghn.org/). We think that 9% mortality for severe but non-critical inpatients is reasonable, even if some lower estimates have been reported since our initial analysis.

13. On page 13, line 2 please change "inseverty" to "in severity".

The term “inseverity” is not included in the manuscript anymore

Although the cut-offs used are arbitrary, they seem to be weel chosen (e.g. 0.5% in overall mortality is one tenth of the current overall mortality).

Results:

14. Please change "Physical examination factors" to a more appropriate term. Fatique is rather subjective and a symptom. Personally I am also not aware of a test for myalgia. Rewording is required for example to "symptoms, vital signs & abnormal physical examination results".

All along the manuscript we changed: “Physical examination” for “Symptoms, vital signs and physical examination”

15. Laboratory factors: as mentioned above.

Some terms (e.g. immuncompromise) need a better definition (lymphopenia, immunosupressive medication, etc. ???).

In S3 Table, immunosuppression line, we added:”Definition: As defined by the authors including patients on immunosuppressive medications and/or with immunosuppressive medical conditions”

16. Please mention why you didn't look at co-medication! This might be the only factor, which is relevant for decision making!!!

Our systematic review focuses on prognostic/predictive factors for which we used a analytical framework specifically designed to summarize this type of evidence.1 Addressing the effects of interventions (previous or new) requires a different analytical approximation (https://gdt.gradepro.org/app/handbook/handbook.html). These two approximations have substantial differences for example in the process of assessing the certainty of the evidence. A central point is that issues of co-intervention are pertinent to causal associations while for prognostic factors we are merely interested in associations. We then decided not to include interventions in our review.

17. Overall it is an important review, which summarizes the current biomarker landscape for Covid-19. The cut-off date (End of April 2020) is very early and new discoveries have been made since. NLR and lymphopenia have been even published earlier. They should actually have met the inclusion criteria, but are not included (Lancet, Lancet Resp. Med. Feb 2020). Data have been homgenised and adjusted. The tables, figures and supplementary figures are very informative and have been nicely prepared!

Please see reply to point 10.

Regarding NLR and lymphopenia, we found moderate certainty that lymphopenia is probably a prognostic factor for mortality and severe disease (page 26, paragraph 2, line 10) and high neutrophil count is probably a prognostic factor for severe disease (page 28, paragraph 1, line 1). Aiming for simplicity as we addressed a significant number of variables, we decided to include candidate prognostic factors only as standalone measurements and not in combination (i.e NLR).

Reviewer #2: This study sets out to report a very important issue: prognostic factors that may be used in decision-making related to the care of patients infected with COVID-19. Although, this topic is not novel, the current research investigated it in a much more detail way than previous studies. In addition, authors also raised the possible questions that must be tackled in near future in order to develop multivariable prognostic models that could eventually facilitate decision-making and improve patients’ outcomes.

Reviewer’s comments:

18. The definitions of severe COVID-19 disease are adopted from multiple studies and therefore are inconsistent. How do the authors eliminate the effect of inconsistent definitions on the association analysis?

We acknowledge that using different definitions of severe COVID-19 disease may have introduced some degree heterogeneity to our results. In response to the reviewers comment we performed subgroup analyses accounting for COVID-19 severe disease definition for 43 of the candidate prognostic factors in which primary analysis had shown significant inconsistency. Observed heterogeneity could not be explained by this analysis for any of the candidate prognostic factors.

In page 11, paragraph 3, line 6 we added: ”In addition, when we observed inconsistent results for disease severity outcome, we performed subgroup analyses accounting for outcome definition (i.e severity scale vs. IVM vs. ARDS) as a potential source of heterogeneity.”

In page 29, paragraph 4, line 1 we added: “When we found inconsistent results for severe COVID-19 disease outcome (n=43), we performed subgroup analyses accounting on outcome definition. Observed heterogeneity could not be explained by this analysis for any of the candidate prognostic factors explored (see S7 appendix).”

19. When severe COVID-19 disease was not reported as an outcome, the authors considered ICU requirement, invasive mechanical ventilation (IVM) and acute respiratory distress syndrome (ARDS) as surrogate outcomes. Did authors analyze the correlation between primary outcome and surrogate outcomes in this study?

Unfortunately we did not address the correlation between disease severity scales and ICU requirement, IVM or ARDS. However most studies (almost all) used only one of these variables to assess disease severity.

20. The ICU requirement, invasive mechanical ventilation (IVM) and acute respiratory distress syndrome (ARDS) were used as surrogate outcomes. As a results, the outcomes in this study include mortality, severe COVID-19 disease, ICU requirement, invasive mechanical ventilation (IVM) and acute respiratory distress syndrome (ARDS). Multiple outcomes make the results difficult to be interpretated. If those outcomes could be analyzed separately, the results can be interpretated more precisely.

We decided to condense those definitions in one outcome (“Disease severity”) mainly for two reasons: 1) We think that providing separate estimates according to the severity definition used would have resulted in a less clear and possibly more confusing message; 2) All utilized definitions are highly related to each other and reflect severity to a greater or lesser extent, hence a significant effect modification depending on which severity definition was used seemed implausible to us. The latter is supported by the analysis performed in response to comment 17.

21. In the analysis of multivariable models, the variables used for model adjustment include age, one comorbidity (e.g diabetes) and one parameter of disease severity (e.g. respiratory rate) at minimum. How do the authors validate that this setting is statistically appropriate?

In the absence of appropriately validated multivariable prognostic models to rely on,2 we used our clinical judgment and background knowledge to define the thresholds for risk of bias assessment. We acknowledge this as a limitation of our work. In page 31, paragraph 3, we added: “Finally, for risk of bias assessment, we defined a set of requirements for dealing with potential confounders that were not previously validated.”

22. In this study, the authors arbitrarily set thresholds to define important incremental increase in the risk of their outcomes, including mortality or severe COVID-19. How do the authors validate that those thresholds can select proper prognostic information?

We did not validate the thresholds used for result interpretation. We acknowledge that those thresholds would vary depending on multiple factors (e.g. the context and the availability of useful treatments). Although we mention this as a limitation in page 31 we think that being explicit about how the result interpretation process was carried out provides transparency to the process and gives the reader the chance of using alternative interpretation approaches.

23. The thresholds were set in 0.5% increase in mortality and 1% increase in severe COVID-19 disease. Do the authors analyze the correlation between 0.5% increase in mortality and 1% increase in severe COVID-19 disease?

We did not analyze the correlation between mortality and severe COVID-19 disease in this review.

24. The authors used the clinical scenario of a patient infected with COVID-19 with severe but not critical disease to assess the prognostic value on mortality. Please clarify the definition of critical disease.

In page 13, paragraph 2, line 3 we added: ”(i.e patients with respiratory failure but not invasive mechanical ventilation and/or hemodynamic support requirement)”

25. When prevalence of prognostic factors was not available, the authors used described baseline risks (9% for mortality and 13 % for severe COVID-19 disease). This setting will inevitably lead to bias in the results.

In page 31, paragraph 3, line 7, we added: ”In addition for some candidate prognostic factors, baseline risk could not be adjusted for prevalence which might have resulted in an overestimation or underestimation of the risk difference estimates.”

26. This study enrolled 207 studies and only 7 were judged as low risk of bias as the remaining presented important limitations in at least one domain or item. Hence, there are multiple inevitable bias in the current study, limiting the reliability of the results.

We agree that the body of evidence has important limitations. Nevertheless exploring how and to what extent those limitations impact the certainty of the observed results is something that requires an extensive analytical process. We have performed that process following state of the art guidance and tools.1 As a result we provide certainty estimates for every candidate prognostic factor and outcome.

27. There is too much information in Table 1. Is there any way to condense the information or split it into different tables?

We agree. We think that all the information included in table 1. is relevant. We are open to suggestions if the editors consider a better way to present it. For example we could split the table according to the different groups of candidate variables.

28. The authors described that clinicians can use their results to tailor management strategies for patients with COVID-19. It would be of great help to the reader if the authors could formulate a summarized management strategy based on the results of this study.

In page 33, paragraph 2, line 9 we added: ”For example, they could select a set of prognostic factors for which there is high certainty in a significant risk incremental increase (e.g Age, gender, comorbidities, respiratory failure and myocardial injury) and use them to define hospitalization rules for patients consulting to the emergency department.”. We have not discussed specific management strategies, given this is the beyond the scope of our review.

References

1. Foroutan, Farid, Gordon Guyatt, Victoria Zuk, Per Olav Vandvik, Ana Carolina Alba, Reem Mustafa, Robin Vernooij, et al. 2020. “GRADE Guidelines 28: Use of GRADE for the Assessment of Evidence about Prognostic Factors: Rating Certainty in Identification of Groups of Patients with Different Absolute Risks.” Journal of Clinical Epidemiology 121 (May): 62–70. https://doi.org/10.1016/j.jclinepi.2019.12.023.

2. Wynants, Laure, Ben Van Calster, Gary S Collins, Richard D Riley, Georg Heinze, Ewoud Schuit, Marc M J Bonten, et al. 2020. “Prediction Models for Diagnosis and Prognosis of Covid-19: Systematic Review and Critical Appraisal.” BMJ, April, m1328. https://doi.org/10.1136/bmj.m1328.

Submitted filename: Response to reviewers.docx

Click here for additional data file.

26 Oct 2020

Prognostic factors for severity and mortality in patients infected with COVID-19: A systematic review

PONE-D-20-20042R1

Dear Dr. Izcovich,

We’re pleased to inform you that your manuscript has been judged scientifically suitable for publication and will be formally accepted for publication once it meets all outstanding technical requirements.

Within one week, you’ll receive an e-mail detailing the required amendments. When these have been addressed, you’ll receive a formal acceptance letter and your manuscript will be scheduled for publication.

An invoice for payment will follow shortly after the formal acceptance. To ensure an efficient process, please log into Editorial Manager at http://www.editorialmanager.com/pone/, click the 'Update My Information' link at the top of the page, and double check that your user information is up-to-date. If you have any billing related questions, please contact our Author Billing department directly at authorbilling@plos.org.

If your institution or institutions have a press office, please notify them about your upcoming paper to help maximize its impact. If they’ll be preparing press materials, please inform our press team as soon as possible -- no later than 48 hours after receiving the formal acceptance. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information, please contact onepress@plos.org.

Kind regards,

Chiara Lazzeri

Academic Editor

PLOS ONE

Additional Editor Comments (optional):

Reviewers' comments:

30 Oct 2020

PONE-D-20-20042R1

Prognostic factors for severity and mortality in patients infected with COVID-19: A systematic review

Dear Dr. Izcovich:

I'm pleased to inform you that your manuscript has been deemed suitable for publication in PLOS ONE. Congratulations! Your manuscript is now with our production department.

If your institution or institutions have a press office, please let them know about your upcoming paper now to help maximize its impact. If they'll be preparing press materials, please inform our press team within the next 48 hours. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information please contact onepress@plos.org.

If we can help with anything else, please email us at plosone@plos.org.

Thank you for submitting your work to PLOS ONE and supporting open access.

Kind regards,

PLOS ONE Editorial Office Staff

on behalf of

Dr. Chiara Lazzeri

Academic Editor

PLOS ONE