Table of Contents
Study cohort
This report focuses on patients who were admitted to the “ASST Bergamo Ovest” hospital wards between February 21 and May 31, 2020, due to COVID-19-related symptoms. The included patients had a laboratory-confirmed SARS-CoV-2 infection and were either discharged or died by the end of the follow-up period, which concluded late on the night of May 31, 2020, totaling 723 patients. The study population represents 25.8% of total hospital admissions during the study period at “ASST Bergamo Ovest”, diagnosed as suspected COVID-19 cases based on WHO interim guidance [7–8]. Please refer to the Cohort Flow Chart in “Figure S2” in “Supplementary material” section for more details.
Patients’ characteristics, respiratory healthcare and drugs administration
Of the patients, 84.8% reside in the geographical area falling within the jurisdiction of “ASST Bergamo Ovest”. Among them, 65.6% are males, resulting in a male-to-female ratio of 1.9:1. A majority of patients (90.5%) exhibit specific respiratory symptoms associated with COVID-19, and their predominant mode of hospital arrival is by ambulances (77.7%), whit 10% arriving within three days of the onset of clinical manifestation. On average, patients arrive 8.3 days (SD 5.0) after the onset of symptoms, with no significant differences by sex but statistically significant differences by age, particularly between patients under 50s and those over 80s (9.0 and 7.8 days respectively). Under a logistic regression, adjusting for patient’s sex and age, a statistically significant 5% decrease in the probability of death is observed for every day earlier arrival (p = 0.015*). The time waited between symptom onset and hospital arrival does not seem to be associate with oxygen saturation levels. Analyzing data by exitus, for patients hospitalized within 14 days of symptom onset (91.04%), a statistically significant correlation is found between the time waited and oxygen saturation levels at hospital arrival in patients who survived (Spearman’s rho= -0.59, p = 0.034*, “Figure 1”). The remaning 9.5% of patients contracted the infection during a planned hospital stay or arrived at hospitals for other symptoms.
Correlation between the time (in days) waited from symptom onset to hospital admission and oxygen saturation level (% SpO2) while breathing ambient air at hospital admission in survivor (Panel A) and non-survivor patients (Panel B). The green line represents the average oxygen saturation levels of patients (95% CI, shown in the grey area) recorded each day (d) up to 14 days of waiting from symptom onset to hospitalization. The correlation is statistically significant in survivor patients, as detailed in the text
Overall, the median age of the study population is 75 years (IQR 64–82) in femalesand 71 years (IQR 62–78) in malesrespectively (p < 0.001*). A total of 77.9% of patients have at least one concomitant disease considered in data collection: 35.0% had one comorbidity, 29.2% had two, and 13.7% had three or more, with no differences by sex. The most prevalent comorbidity is hypertension (61.8% and 63.1% in male and female, respectively, p = 0.744), followed by diabetes (28.9 and 21.7%, p = 0.037*), cardiovascular disease, including ischemic heart disease, coronary heart disease and myocardial infarction (21.5 and 14.9%, p = 0.031*), obesity (12.5 and 14.9%, p = 0.363), active tumour (11.4 and 10.4%, p = 0.699) and kidney failure (4.4 and 4.8%, p = 0.812). The distribution of the AB0 blood type system among patients is not different from the distribution among the Italian population, even adjusting for patient sex, age or exitus. Overall, the median oxygen saturation recorded by first aid operators before any supporting oxygen flow administration is 88% (83–94). Of the inpatients, 16% recorded a normal resting oxygen saturation above 95%, 30% recorded levels between 95 − 90%, and the remaning 54% had values lower than 90%. A scatter distribution by exitus is reported in “Figure 2”.
Distribution of oxygen saturation levels recorded by first aid operators before any supporting oxygen flow administration, categorized by outcome
The average length of ICU stay was 24.2 days (20.0 SD) with statistically significant differences based on exitus. When adjusting for age, number of comorbidities, and oxygen saturation capacity, female patients appear to have a lower probability of ICU hospitalization (-56.5%, IC95% -79.1, -9.3; p = 0.03*), as well as patients of both sexes over 70 years old, showing a probability reduction of -87,6% compared to the 60–69 years old age group (IC95% -95.2,-68.0, p < 0.001*; -90.8% if over 80s, IC95% -97.1,-70.2, p < 0.001*). However, with the inclusion of laboratory tests covariatesin the model, gender loses its statistical significance.
The administration of antiviral therapy tipically used to treat HIV infection is associated with the sex of the patient (more frequent in males, p = 0.015*), age group (decreasing with increasing age, p < 0.05*) and oxygen saturation level (increasing with decreasing SpO2, p < 0.001*).
The primary class of anticoagulant employed is low-molecular-weight heparin (LMWH) administered in 68.0% of patients. Other administered anticoagulant include antiplatelets (3.8%, 24 patients), a combination of LMWH and antiplatelets (22.0%), oral anticoagulants (2.1%, 13 patients), a combination of LMWH and oral anticoagulants (3.2%, 20 patients) and direct oral anticoagulant molecules in less than 1%. Overall, anticoagulant therapies were administered in 87.4% with an association by sex (χ², p = 0.003*, aORmale=1.92, IC95% 1.17–3.13; p = 0.009*).
Aside of analgesic therapy administered during ICU stay, continuous opioid administration from hospital admission is statistically significant associated with sex (28.9% female and 36.3% male, p = 0.046*) and age (positive correlation, p < 0.001*). If one out of two patients over 70s received the therapy, less than one out of ten under 50s did. The association is confirmed when controlling for sex, age, oxygen saturation, and comorbidities: the probability of receiving continuous opioid therapy outside the ICU is lower in female (OR = 0.66, IC95% 0.45–0.96, p = 0.029*) and increases with age (p < 0.001*) or the number of comorbidities (OR = 1.9 IC95% 1.03–3.5, p = 0.040*).
Forty-nine patients were enrolled in the TOCIVID-19 clinical trial (EudraCT Number: 2020-001110-38) without differences by sex. Due to the enrollment inclusion criteria, we did not evaluate administration differences between groups for tocilizumab (RoActemra®).
In “Table S1” and “Table S2” reported in “Supplementary material” section, patient characteristics, respiratory healthcare, and drugs administration are summarize by exitus as well, using Cox PH regression models and Kaplan-Meier curves.
Laboratory tests
Laboratory tests data are collected on the day of hospital admission, at the day of discharge or death, and at the day of ICU admission. Due to the low number of available data, tests collected on the day of ICU admission are not reported. It follows a description of biomarkers recorded on the day of hospital admission and their association with patients’ characteristics. Associations with the outcome are listed in “Table S1” and “Table S2” in “Supplementary material” section, Kernel density distributions by outcome are illustrated in “Figure 3”, and ROC analysis is reported in “Table S3” in the “Supplementary material” section.
Kernel density function of laboratory tests performed on the day of hospital admission. Legend: the unit of measurements is the same as in the text, continuous line represents discharged live patients, while dotted line is for died patients. In the PT, PTT, CK, Hs Tn-I and AST graphs, outlier values (< 1%) are not included to better highlight differences in the probability distribution of biomarkers by outcome
Hypoxia
In 70% of patients, the levels of deoxyhemoglobin at hospital admission exceed the upper limit, with an overall median value equal to 7.5% (IQR 4.0-13.7) and a statistically significant association with age (p = 0.024*).
Coagulation defects
Median platelet count value is 205 10^3/µl (IQR 157–277, range 29-1137 10^3/µl) with an association between sexes: male patients more often have values below the lower limit, otherwise females have value above the upper limit (p < 0.001*); and with class of age (decreasing values with increasing age, p < 0.001*). The lowest PT value recorded is 11.1 s, indicating that no one has a PT value below the normal range. Overall, the PT median value is 13.9 s (IQR 13.1–15.2), with longer PT times in older patients (p < 0.001*). Median PTT value is 33.3 s (IQR 30.5–36.9), with differences by sex (p = 0.021*) and class of age (p = 0.002*). D-dimer levels were within normal range in 8.6% of patients. Overall, the median D-dimer value is 1398.5 (IQR 881-2712.5) ng/mL aside from sex, and with association by age (p < 0.001*).
Immunity system disorders
Data indicate changes in the proportion of WBCs, showing an increase in neutrophils and a decrease in lymphocytes, with a statistically significant association with sex. Evaluating the neutrophils and lymphocytes ratio, values greater than 9.8 [21] can double the probability of death (OR = 1.98 IC95% 1.4–2.7, p < 0.001*). Monocytes, eosinophils and basophils data are also reported in “Table 1”.
Non-specific biomarkers of tissue damage
LDH is above the upper limit in 83.3% of patients, with a median value at hospital admission of 368.1 (277.4-513.7) U/L, without statistically significant differences between groups. Investigation on CRP at hospital admission was performed in a too few patients to draw any meaningful conclusion, data not shown.
Heart failure markers
CK median value is 124.7 (65.2-286.8) U/L, with higher values in male patients (p < 0.0001*). Hs-Tn I median value is equal to 0.02 (0.01–0.05) µg/L, with statistically significant differences by class of age (positive association, p = 0.0001*).
Liver failure markers
AST median value is equal to 47.9 U/L (IQR 32.7–73.7), with an association by sex (p < 0.001*) and age (p = 0.017*), with values over the threshold mainly reported in male patients (58.7% vs. 41.7%, p < 0.001*). ALT median value is equal to 32.8 U/L (21.9–57.4) with statistically significant differences between sexes (35.3 in male and 27.3 in female, p < 0.0001*) and by class of age (p = 0.0001*). Categorizing the variable as “in” or “out” of range, groups that more often have values over the upper limit are male (35% vs. 24.5%, p = 0.004*) and patients between 50 and 70 years old (p = 0.003*, 4 DF). The AST/ALT ratio confirms the correlation with death (OR = 2.31 IC95% 1.77–3.02, p < 0.001*). APRI index has a median value equal to 0.50 (0.29–0.87) with a statistically significant difference by sex (0.57 and 0.42 as median value in male and female respectively, p < 0.0001*). FIB-4 index collected at hospital admission has a median value equal to 2.81 (1.73–4.78), with statistically significant differences by sex (2.9 and 2.6 as median value in male and female respectively, p = 0.005*) and by class of age (positive linear correlation, p = 0.0001*). Having an elevated APRI or FIB-4 index statistically significantly increases the probability of death (ORAPRI=2.10 IC95% 1.57–2.81, p < 0.001* and ORFIB−4=1.36 IC95% 1.25–1.47, p < 0.001*).
Trends
When laboratory biomarkers are available at both hospital admission and discharge for the same patient, changes over time are analyzed by sex and outcome and illustrated with boxplots in “Figure S3” in Supplementary Material. According to our data, AST is the only biomarker that decreases to the normal range in a statistically significant way only in survivors. A similar but opposite trend is recorded for the platelet count that increases in a statistically significant way only in survivors, regardless of sex. Furthermore, survivors are characterized by a statistically significant increase in the proportion of lymphocytes and a reduction in neutrophils. In died patients, WBCs alteration are worse, aside from sex, with a more evident increase in neutrophils and decrease in lymphocytes. PTT values decrease in survivors and increase in died, with statistically significance differences only in male patients. A similar trend is appreciable in LDH values, even if available data are too few to lead to conclusion. Overall, the values of D-dimer remain above the range at discharge; moreover, we noticed a reduction in survivors and further increases in died patients, statistically significant in dead male and surviving female. The PT, Hs-Tn I and ALT are the biomarkers that change less between time points. CK values seem to decrease in both survivors and died patients, but data on CK are inconsistent to lead to conclusion, as previously clarified.
Survival analysis
The first day of FU was set on February 21, 2020, and the May 31, 2020, marks the last day of FU. The total duration of FU spans 101 days. Individually, hospital stays ranged from one day to over three months, with the longest stay recorded at 94 days. The median length of hospital stay was ten days (mean 14.6 days, SD 12.7 days), aside from sex (10, IQR 6–20 in male and 10, IQR 6–23 in female; mean: 14.5 SD 12.8 and 14.7 SD 12.5), but depending on exitus: 6 days (IQR 3–10; mean: 9.0 days, SD 9.3) between who died and more than a double length of stay, LoS (14 days, IQR 8–25; mean: 18.1 days, SD 13.3) between survivors (p < 0.001*), with no differences by sexes (“Figure 4). At the end of the FU, thatwere 281deaths: 186 male and 95 female patients (p = 0.776), disclosing a lethality rate of 38.9% among frail patients during the first pandemic wave. The 75.8% of deaths occurred within ten days of hospital admission (73.7% in male and 80% in female, p = 0.665), with 18.5% occurring within two days (17.7% in male and 20.0% in female, p = 0.703). Overall, the median survival time was about 34 days (26–50): 30 days (23–43) in male and 37 days (28-nr) in female, withno statistically significant differences between sexes. Kaplan Meyer survival analyses are summarize in “Table S2” in “Supplementary material” section. Cox PH regression analysis on survival is reported in “Table S1” in “Supplementary material” section as a single model on covariate, adjusted for sex and age. Multivariable Cox PH model, characterized by Harrell’s C index of 0.92, is represent with a Forest Plot in “Figure S4” in the “Supplementary material” section. To avoid redundancy of information we only considered the FIB-4 index to disclose liver failure, categorizing the values of the marker according to the literature [22]. Due to the low rate of laboratory tests, we could not include LDH in the analysis, despite its association with the risk of death, as demonstrate in the univariate analysis.
Kaplan-Meier survival estimates by sex. Legend “d” is for day and “p” is for probability, very short dash line is for female, long dash line is for male. See text for details. Note: time of follow-up is the overall LoS, and events are deaths occurred during the hospitalization
Thromboembolic events
Overall, thromboembolic events were reported in 124 SDOs (17.2%), with no association by sex and exitus, but with a statistically significant association with the age of the patient, indicating a lower risk of events in younger patients (p = 0.008*). Under sex and age equality, the risk of thromboembolic events during hospital stays increased by 74% in obese subjects (p = 0.043*) and decreased by 37% in patients who received oxygen supplementation (p = 0.025*). Otherwise, still under sex and age equality, preventive therapy for deep venous thrombosis did not appear to decrease the likelihood of thromboembolic events, even after controlling for other administered drugs. Of the specified events (N = 107), the majority affected the heart (75.4%), followed by the lung (12.3%), the central nervous system (4.9%) and an arm or leg (3.3%). There was a statistically significant association between the site of onset and the exitus (p = 0.010*). Specifically, events predominantly affected the heart tissue in patients who died (92% versus 65%), while the lung or central nervous system were more common in survivors (19% vs. 2% and 6.8% vs. 2.1% respectively).
