Epidemiological Study of Patients Admitted in Intensive Care Unit with Severe Acute Respiratory Illness with a Possible Diagnosis of COVID19 (EPIC19), a Multicentre Study.


 BackgroundGlobal pandemic of COVID 19 has affected many countries. The initial epicenter was in China with gradual spread to various countries including India.For a developing country like India with limited resources and high population, it is worthwhile to know how these patients requiring intensive care admission were managed and the outcome of these patients. To address these issues, a prospective observational study was planned.Methods A multicenter study was conducted from June 2020 to December 2020 including 4 centers across India. Patients > 18 years of age admitted in the intensive care unit (ICU), with the diagnosis of COVID 19 pneumonia confirmed by reverse transcriptase –polymerase chain reaction (RT-PCR) or rapid antigen test (RAT) as applicable were included. Factors associated with ICU mortality were examined using multivariable logistic regression analysis and Cox proportional hazard model. ResultsOf 667 patients were included in the study. ICU mortality was 60 %. In multivariable analysis, history of cerebral vascular accident (CVA), day 1 acute physiology and chronic health evaluation (APACHE II) score, need for invasive ventilator support, minimum PO2, fluid balance and complications such as pneumothorax and arrhythmia during ICU admission were associated with mortality. Among these parameters, day 1 need for invasive ventilator support (odds ratio OR: 3.01(1.81, 5.00) and development of arrhythmia (OR 3.85 [1.56, 8.06]) had higher odds of mortality. Cox proportional hazard analysis showed, history of ischemic heart disease (IHD) (Hazard Ratio, HR 1.64, 95% CI:1.13, 2.38), day1 APACHE II (HR 1.03, 95% CI:1.00, 1.07), arterial blood gas (ABG) pH (HR 0.14, 95% CI:0.03, 0.56) and use of therapeutic anticoagulation (HR 0.42,95% CI:0.29, 0.61) as a predictor of 7 days ICU mortality. Daywise trend of ventilator parameters showed dynamic compliance was higher on day3 and 4 in survivors.ConclusionIn this cohort of ICU patients, ICU mortality was 60%. The reason for higher mortality could be the severity of illness as suggested by the day 1 PF ratio (109.31 [77.79-187.26]).Trial Registration-(IEC131/2020, CTRI/2020/06/025858).


Abstract
Background Global pandemic of COVID 19 has affected many countries. The initial epicenter was in China with gradual spread to various countries including India.
For a developing country like India with limited resources and high population, it is worthwhile to know how these patients requiring intensive care admission were managed and the outcome of these patients.
To address these issues, a prospective observational study was planned.

Methods
A multicenter study was conducted from June 2020 to December 2020 including 4 centers across India.
Patients > 18 years of age admitted in the intensive care unit (ICU), with the diagnosis of COVID 19 pneumonia con rmed by reverse transcriptase -polymerase chain reaction (RT-PCR) or rapid antigen test (RAT) as applicable were included. Factors associated with ICU mortality were examined using multivariable logistic regression analysis and Cox proportional hazard model.

Results
Of 667 patients were included in the study. ICU mortality was 60 %. In multivariable analysis, history of cerebral vascular accident (CVA), day 1 acute physiology and chronic health evaluation (APACHE II) score, need for invasive ventilator support, minimum PO 2 , uid balance and complications such as pneumothorax and arrhythmia during ICU admission were associated with mortality. Among these parameters, day 1 need for invasive ventilator support (odds ratio OR: 3.01(1.81, 5.00) and development of arrhythmia (OR 3.85 [1.56, 8.06]) had higher odds of mortality. Cox proportional hazard analysis showed, history of ischemic heart disease (IHD) (Hazard Ratio, HR 1.64, 95% CI:1.13, 2.38), day1 APACHE II (HR 1.03, 95% CI:1.00, 1.07), arterial blood gas (ABG) pH (HR 0.14, 95% CI:0.03, 0.56) and use of therapeutic anticoagulation (HR 0.42,95% CI:0.29, 0.61) as a predictor of 7 days ICU mortality. Daywise trend of ventilator parameters showed dynamic compliance was higher on day3 and 4 in survivors.

Conclusion
In this cohort of ICU patients, ICU mortality was 60%. The reason for higher mortality could be the severity of illness as suggested by the day 1 PF ratio (109.31 [77.79-187.26]).
1 parameters such as APACHE II score, need for invasive ventilator support, minimum PO2, uid balance and complications observed such as pneumothorax and arrhythmia during ICU admission.

Background
The global pandemic of COVID 19 has affected more than 200 countries. The initial outbreak was in China, December 2019, where Wuhan was the epicenter for this outbreak [1]. In India rst case of this novel illness was seen in January 2020. The disease which was initially limited to international travelers, within few months spread to the community. The illness caused by the novel Coronavirus was similar to viral pneumonia, characterized by Type I respiratory failure with acute respiratory distress syndrome (ARDS), which was of mild, moderate, or severe category [2,3].
Evidence was constantly evolving amidst the pandemic with emerging new strategies and management protocols. Response to disease, population demographics and case fatalities were variable in different countries [4,5]. The successful response to any pandemic depends upon how we use available health resources, infrastructure and implement therapeutic strategies [6]. Effective management and treatment of these patients was a major challenge for countries such as India, with high population and limited resources.
The preliminary study published from India with 235 patients showed 24hrs mortality as 8.5% [7]. The study on the timing of intubation including 147 patients showed higher mortality of around 60% [8]. These two studies were single center studies. However, there is limited data that has been published from India that describes the pro le of severe COVID cases requiring admission in Intensive Care Units (ICU) and the management of these cases. Therefore this study evaluates the epidemiology and clinical characteristics of the patients with the diagnosis of COVID pneumonia, requiring ICU admission in large tertiary care hospitals across India.

Methods
The study was approved by the Institutional Ethics Committee (IEC) of St. John's Medical College (131/2020) (CTRI/2020/06/025858). Critical care units across the country were invited by email to participate in the study. Nine centers responded and were interested in participating in the study. However, only 4 centers were recruited as others had logistic issues due to a rapid surge in the caseload and delay in getting the IEC approval. Each recruited center obtained approval from the local IEC's of the respective hospitals. As it was an observational study, a waiver of consent was given for three centers, except for one center.
The data were collected prospectively from June 2020 to December 2020, from four tertiary care hospitals. Online training was provided to each center's investigators about data collection and data entry. Data were entered in a standardized excel spreadsheet which was shared with all the centers.
Regular checking for data validation and correctness was done throughout the data collection process.
Demographic details, epidemiological details, ventilator parameters and treatment details were collected. As data for 7 days of ICU stay was required, there was an initial hurdle in prospective data collection, this was overcome by retrieving the prospectively collected data from the health records. Data related to the type of oxygen therapy, invasive or noninvasive ventilation therapy, (PO 2 /FiO 2 ) PF ratio, use of rescue therapy such as prone ventilation, or Extracorporeal Membrane Oxygenation (ECMO) were collected. Acute physiology and chronic health evaluation (APACHE II) and Sequential organ failure assessment (SOFA) scores were calculated in the rst 24 hours of ICU admission. The sample size for estimating mortality of 50% with the margin of error as 5% and 95% con dence interval was 385. Considering the possibility of missing data in 10% of subjects, the required sample size was decided to be 400. The primary outcome of interest was ICU mortality and the secondary outcomes included length of ICU stay, duration of mechanical ventilation and hospital mortality.
A total of 1273 patients were screened, based on the inclusion criteria, reverse transcriptase-polymerase chain reaction (RT-PCR) or rapid antigen test (RAT) con rmed cases were included in the study. Five hundred and sixty-three patients were RT-PCR negative. Among 710 RT-PCR positive patients, 40 patients were excluded due to incomplete records. Total of 667 patients were included in the study across the 4 centers with maximum enrolment from one center as shown in Figure 1.
Statistical analysis-All statistical analyses were performed using STATA v19 (StataCorp. 2019. Stata Statistical Software: Release 16. College Station, TX: StataCorp LLC). The clinical and epidemiological characteristics are reported as mean (standard deviation SD) or median (Interquartile range IQR) as applicable for continuous variables and number (%) for categorical variables. Factors associated with ICU mortality were identi ed by multivariable logistic regression. Several variables which were initially considered as predictors of mortality were collinear and hence a careful selection of variables was made based on perceived importance in the clinical and statistical association. Adjusted odds ratio (95% con dence interval) was reported from the multivariable logistic regression. For ventilated patients, the change in ventilator settings over 7 days of ventilator usage were analysed using the mixed logistic model and are represented as trend lines with error bars. The Kaplan Meir survival curve was used to estimate the seven-day survival probabilities in ICU. Seven-day mortality was considered because mortality within this period can be attributed to COVID infection and not affected by known confounders such as hospital-acquired infections. The factors associated with 7day mortality in the ICU were identi ed using the Cox-proportional Hazards model and hazard ratios (HR) with 95% con dence interval (CI) are reported. All p values less than 5% were considered statistically signi cant.

Results
The patients were predominantly male (70.46%), and the average age was 57 years (SD=15). The majority of the patients were admitted directly to the respective hospitals (65%) and the remaining (35%) were transferred from either local clinics or tertiary care hospitals ( Table 1). The predominant comorbidities were hypertension followed by Diabetes Mellitus (DM). Among the comorbidities, the presence of DM, hypertension, Ischemic heart disease (IHD) and cerebral vascular accident (CVA) were signi cantly associated with mortality when not adjusted for any confounders. There was no effect of recent travel or attending a mass gathering or any contact with COVID suspects or COVID positive patients on mortality as shown in Table 1.
The mortality in ICU was 60% and mortality in the hospital was 63%. There was no difference in mortality between males and females (59% vs 61% respectively, p=0.504). Length of ICU stay and number of days on ventilator support were similar in survivors and non-survivors (Table2).
Echocardiography and ultrasonography on admission were routinely done in one of the centers. Among echocardiographic parameters, the presence of systolic dysfunction as evaluated by the visual gestalt method and the presence of regional wall motion abnormality were associated with worse outcomes. E/A ratio for evaluating diastolic dysfunction was lower in non-survivors than survivors (0.91 vs 1.13 p= <0.001) ( Supplementary Table S1) The E/e` ratio indicative of LV (left ventricle) lling pressure was lower 8.06(2.63) and was not signi cant among survivors 8.03(2.59) and non-survivors 8.09(2.66), p=0.87 (Supplementary Table S1). There was no difference in outcome based on lung score by USG (ultrasonography) and the number of quadrants involved based on CXR.
Among the medications, the steroid was used in 84.6% survivors vs 92.5% non-survivors (p=0.001) and plasma was used in 10.8% survivors vs 5.3% non-survivors (p=0.007). (Supplementary Table S1) Various complications were observed in these patients. Predominant complication being the development of pneumothorax 66(10%) followed by subcutaneous emphysema 28(4%), both were associated with the higher mortality. Incidence of Pneumothorax was seen in 32 (10.70%) patients requiring invasive ventilator support on day 1 (Supplementary Table S1).
The most common mode of ventilation used was pressure regulated volume control (PRVC) (43%).
Among the total of 667 patients, invasive ventilation was used in 445 (66.7%) patients during their hospitalization and the mortality was 59.8% among the intubated patients. The average time taken for intubation was 2days (1-5) from the date of ICU admission. Prone positioning was used in 67 (22.41%) mechanically ventilated patients and awake proning was used in 44(11.96%) spontaneously breathing patients on the day1 of ICU admission. Severe ARDS was seen in 132 (43.4%) intubated (299) patients on day1. Awake proning was used in 134(23.1%) patients at least once during their ICU stay.
Among the ventilator parameters, tidal volume was around 6.24(1.36) ml/kg of predicted body weight. Other ventilator parameters like minimum PO 2 , PCO 2 , Peak inspiratory pressure (P Peak) and positive endexpiratory pressure ( PEEP) were found to be signi cant as shown in Table 3.

Factors associated with mortality
In multivariable analysis, comorbidities such as CVA, day 1 parameters such as APACHE II score, need for invasive ventilator support, minimum PO 2 , uid balance and development of complications such as pneumothorax and arrhythmia during ICU stay were associated with mortality (Table 4). Among these parameters, the requirement of invasive ventilator support on day 1 and the development of arrhythmia during ICU admission were statistically signi cant with OR= 3.01 (1.81-5.00) and 3.85 (1.56-8.06) respectively.
Among the 667 patients, 208 (31.18%) died within the rst seven days of ICU admission ( Figure 2). The estimated mean survival time in the rst seven days of ICU admission was 5.9 days (95% CI: 5.8-6.1). Cox proportional hazard regression analysis was done to identify the predictors of 7day ICU mortality. Patients with IHD as comorbidity and higher APACHE II score at ICU admission had signi cantly higher mortality as shown by hazard ratios, HR=1.64(95% CI:1.13,2.38) and HR=1.03(95% CI:1.00,1.07) respectively (Table 5). Patients who received therapeutic anticoagulation were 58% less likely to be nonsurvivors at Day 7 of ICU stay compared to the patients who did not receive this treatment. Day 7 mortality was negatively associated with higher ABG (Arterial blood gas) pH done on Day 1 of ICU admission, HR=0.14(95% CI: 0.03, 0.56) ( Table 5). Figure 3 shows the trend of ventilator parameters over 7 days. Mean PCO 2 was signi cantly higher in the non-survivors throughout the 7 days. The mean driving pressure was higher in the survivors on days 5 and 6. Although static compliance was comparable between the time groups at all time points, mean dynamic compliance was higher in the survivor group on days 3 and 4. However, there was no difference in the change in any of these ventilator parameters over time (day 1 to day 7) between the survivor and non-survivor groups (Figure 3).

Discussion
This is one of the largest multicenter studies from India which showed 7 days trend of patients with COVID pneumonia. The parameters which were statistically signi cant among survivors and nonsurvivors were CVA, APACHE II score, day1 requirement of invasive ventilator support, minimum PO 2 , uid balance and complications such as pneumothorax and arrhythmia as shown by multivariable analysis ( Table 4). The ICU mortality observed was higher (60%) as compared to other studies.
The (Pro vent COVID) study which included 553 patients who were mechanically ventilated showed median tidal volume was 6·3 ml/kg of predicted body weight (IQR 5·7-7·1), PEEP was 14  with the higher mortality as shown by previous studies of ARDS [9].
The study by Ferrando, et al, of 742 patients requiring invasive ventilation showed all-cause ICU mortality was 32% and mortality seen in severe ARDS patients was 39% [10]. The multicenter study from the US with 2215 patients showed a mortality of 39.5% [11].
One of the largest multicenter studies of COVID patients, from 138 hospitals, including > 4000 patients showed that 90 days mortality was around 31%. Mortality was higher among older, obese, diabetics and severe ARDS patients. In this study, the PF ratio on day1 was 154 (106-223) which was higher than that observed in our study 109.31(77.79-187.26) [12].
In our study among the ventilator parameters, routine parameters such as tidal volume and plateau pressure (P Plat) were not found signi cant. Probably due to the practice of low tidal volume and monitoring of P Plat was the established standard in the majority of the ICUs. P Peak and PCO 2 were found to be signi cant parameters, along with minimum PO 2 . This suggests the possibility of impaired oxygenation as well as ventilation, indicating higher PCO 2 levels in non-survivors as compared to survivors (46.15mm of Hg (18.30) vs 39.64mm of Hg (11.36). This also suggests the possibility of intrapulmonary shunt due to micro-thrombi or dead space.
Another peculiar nding in our study was the 7days trend of ventilator parameters. Among the ventilator parameters, change in any of the ventilator parameters over 7 days was not statistically signi cant, but dynamic compliance was higher on day3 and 4 in survivors (Figure 3). This highlights that pathophysiology of COVID is not only due to impaired static compliance alone but dynamic compliance also plays a role which is a measure of compliance and airway resistance [13]. It also indicates possibility of different phenotypes of ARDS in COVID patients [14].
A retrospective study by Xie, et al, of 733 patients showed a median age of 65 years, and mortality was around 53.8 %. The higher mortality was due to associated organ failure like respiratory failure, shock and acute kidney injury. These ndings were similar to our study as indicated by higher APACHE II score and SOFA score in non-survivors [15].
There were two small studies by Bhataraju, et al and Arentz, et al, that showed higher mortality of 50% and 67% respectively [16, 17].
The study by Yang, et al, of 52 critically ill patients showed 28-day mortality of 61.5%, with a lower PF ratio, similar to our study. Surprisingly these patients had a lower incidence of barotrauma of 2 % as against 9.9% in our study, possibly due to timely use of rescue therapy such as ECMO in 17% of patients. In our study, only 1 patient received ECMO [18].
A study looking at the timing of intubation from India showed baseline mortality of 60 % which was similar to our study [8].
The strengths of our study include, the patients who were admitted to the ICU were having severe disease, which is a true representation of the critically ill population. Also as shown in Table S1, echocardiographic parameters suggest, these were the patients, with low LV lling pressure as shown by E/e' ratio, as a true representation of the ARDS population. This study also describes characteristics of the patients requiring invasive as well as non-invasive ventilator support. The trend of parameters for 7 days for each patient helped in understanding the progress of the disease. Collection of data was a big challenge when healthcare workers were already overburdened by the caseload which was overcome by a volunteered team of research enthusiasts.
There were certain limitations to our study. Although the study was designed as a multicenter project, the actual number of ICUs which could participate was only 4, hence the data cannot represent the practices across various ICUs. The second limitation was depending on the type of hospital and case surge, criteria for admission for each ICU may be variable hence mortality among various ICUs may differ. The third limitation was the enrolment of different centers that happened at variable times during the study, which might have affected the number of patients recruited from each center. In addition to this, there was a difference in the timing of the surge of cases in various states. So during the study period, it was di cult to gure out the approximate number of patients who could be enrolled in the study, which resulted in a total enrolment of 667patients that was higher than the calculated sample size.
This study helped in evaluating the management of patients with COVID across the centers and also helped to check the practice patterns of managing ARDS, especially the use of low tidal volume strategy and prone positioning for which evidence is already established [19]. It also highlights the challenges for healthcare workers while taking care of these sick patients, such as the use of awake proning, prone positioning and managing the complications as mentioned in one of the largest meta-analyses [20].

Conclusion
In this cohort of ICU population of 667 patients, the primary outcome of ICU mortality was 60% and risk factors associated with poor prognosis as shown in multivariable analysis were history of CVA, day 1 parameters such as APACHE II score, need for invasive ventilator support, minimum PO 2 , uid balance and development of complications such as pneumothorax and arrhythmia during ICU stay. Predictors of 7 days mortality were IHD, APACHE II score, ABG pH and use of therapeutic anticoagulation. Our study also showed improving trend of dynamic compliance in survivors. Future studies with classifying ARDS based on the phenotypes such as compliance, resistance and elastance, will be helpful to guide in the management.

Consent for publication-Not applicable
Availability of data and material-Upon request deidenti ed data will be available after approval from the ethics committee.
Competing/ Con icts of Interest-On behalf of all the authors as a PI,I declare that there are no con icts of interest or competing interest.
Funding-This study did not receive any funding.    Tidal volume (ml), Fluid balance (ml) Table 3 Oxygen therapy and ventilator parameters   The trend of ventilator parameters

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download. SupplementaryMaterial.docx