key: cord-0995972-91zf31ds
authors: Monforte, Antonella d’Arminio; Tavelli, Alessandro; Bai, Francesca; Tomasoni, Daniele; Falcinella, Camilla; Castoldi, Roberto; Barbanotti, Diletta; Mulè, Giovanni; Allegrini, Marina; Tesoro, Daniele; Tagliaferri, Gianmarco; Mondatore, Debora; Augello, Matteo; Cona, Andrea; Ancona, Giuseppe; Gazzola, Lidia; Iannotti, Nathalie; Tincati, Camilla; Viganò, Ottavia; De Bona, Anna; Bini, Teresa; Cozzi-Lepri, Alessandro; Marchetti, Giulia
title: The importance of patients’ case-mix for the correct interpretation of the hospital fatality rate in COVID-19 disease
date: 2020-09-17
journal: Int J Infect Dis
DOI: 10.1016/j.ijid.2020.09.037
sha: ef32304b4d2ded7ccd77836fa331ff9018cc095d
doc_id: 995972
cord_uid: 91zf31ds

Objective We aimed to document data on the epidemiology and factors associated with clinical course leading to death of patients hospitalised with COVID-19. Methods Prospective observational cohort study on patients hospitalised with COVID-19 disease in February-24th/May-17th 2020 in Milan, Italy. Uni-multivariable Cox regression analyses were performed. Death’s percentage by two-weeks’ intervals according to age and disease severity was analysed. Results A total of 174/539 (32.3%) patients died in hospital over 8,228 person-day follow-up; the 14-day Kaplan-Meier probability of death was 29.5% (95%CI: 25.5-34.0). Older age, burden of comorbidities, COVID-19 disease severity, inflammatory markers at admission were independent predictors of increased risk, while several drug-combinations were predictors of reduced risk of in-hospital death. The highest fatality rate, 36.5%, occurred during the 2nd-3rd week of March, when 55.4% of patients presented with severe disease, while a second peak, by the end of April, was related to the admission of older patients (55% > = 80 years) with less severe disease, 30% coming from long-term care facilities. Conclusions The unusual fatality rate in our setting is likely to be related to age and the clinical conditions of our patients. These findings may be useful to better allocate resources of the national healthcare system, in case of re-intensification of COVID-19 epidemics.

SARS CoV-2 epidemic is one of the most devastating worldwide epidemics in the last century. From the initial outbreak in China, it reached Europe by the end of February, 2020.

Lombardy was the first region to be affected, thus representing an unexpected challenge for region governors and causing a dramatic overload of hospitals and intensive care units (ICU). In a short time span, all the hospitals were overwhelmed by dozens of citizens suffering from acute respiratory distress and the different departments were rapidly shifted to areas for patients. On March 11, WHO declared the state of pandemic (World Health Organization 2020).

We faced a number of deaths that seemed to be higher than what could be expected from the data of the Chinese epidemic, where a fatality rate on hospitalised patients ranging from 1.4-2.3% J o u r n a l P r e -p r o o f (Guan et al. 2020 , Wu C et al. 2020 , Wu Z et al. 2020 ) to 28% was documented in an early report from Wuhan by Zhou et al. (2020) .

When reporting fatality rates, several factors should be taken in consideration, that include the reference population, disease severity, but also presence of comorbidities and the availability of ICU beds; all these factors might contribute to disentangle differences on COVID-19 fatality rates in different settings.

Bearing this in mind, we aimed to identify factors associated with the risk of in-hospital death in a cohort of hospitalised patients with COVID-19 disease in a single hospital in Milan.

San Paolo hospital is a University hospital with 426 beds of all specialities, including ICU, infectious diseases, and pneumology. Since end of February, increasing number of ICU and non-ICU beds were saved for COVID-19 patients (Supplemental Fig 1) . Doctors and nurses converged in multidisciplinary teams leaded by infectious diseases, pneumology and intensive care physicians.

Prospective observational cohort study including all patients admitted to the San Paolo Hospital in Milan with symptomatic SARS CoV-2 infection between February 24 and May 17, 2020.

Inclusion criteria were: -confirmed diagnosis of symptomatic SARS CoV-2 infection by RT-PCR on naso-pharyngeal or oro-pharyngeal or broncho-alveolar swab specimens; -age>=18 years;hospitalisation in February 24-May 17. Patients who died in the emergency room within 24 hours and patients not hospitalised were not included.

Data were entered into an electronic database, including: age; sex; ethnicity; risk factors for SARS CoV-2; ongoing or previous comorbidities; age-unadjusted Charlson comorbidity index (Charlson et al. 1987 ); symptoms; obesity; respiratory rate (RR), oxygen saturation percent (SO2); computerised tomography (CT); laboratory examinations.

CT scan was evaluated as: no pathological findings; interstitial pneumonia; consolidation; pleural effusion. Mono-or bilateral extension was collected.

Disease severity at admission was classified as mild (no pneumonia); moderate (radiological demonstration of pneumonia; RR>26/min; SO2>96% in room air; PaO2/FiO2 >300 mmHg); severe (RR < 24/min; SO2<92%; PaO2/FiO2 100-300 mmHg); critical disease (PaO2/FiO2 <100 mmHg).

The highest intensity of ventilation was recorded as: no need; low/high flow supplemental oxygen by nasal cannula/face mask; continuous positive airway pressure device (cPAP); mechanical non-invasive or invasive ventilation. J o u r n a l P r e -p r o o f Criteria for invasive mechanical ventilation were acute respiratory distress (PaO2/FiO2 <100 mmHg) and no major conditions determining short life expectancy.

Antivirals; low molecular weight heparin; hydroxychloroquine+/-azithromycin; immunomodulatory agents; high-dose corticosteroids were collected and grouped according to the combinations used.

Primary end-point was time to in-hospital death. Factors associated were evaluated in the whole cohort and in patients undergoing mechanical ventilation, with a special focus on the possible role of comorbidities. We also evaluated the dynamics of the disease in terms of severity at presentation, age of patients, availability of ICU beds, and fatality-rate according to two-week time frames.

The study was approved by Ethic Committee Area 1, Milan (2020/ST/049 and 2020/ST/049_BIS, 11/03/2020). Informed consent was obtained whenever possible.

Follow-up was censored at June 17, so that each patient had at least 30 days' observation.

Statistics included: Chi-square and Kruskal-Wallis test, to compare characteristics of in-hospital survivors vs non survivors and characteristics of population according to 2-week time-span of admission.

We calculated the in-hospital mortality by age strata according to the number of comorbidities and formally tested for interaction between age and number of comorbidities using Wald test.

Kaplan-Meier curves were used to estimate the probability of in-hospital death. The time-to event was calculated from the date of hospital admission to the date of death or last day of hospitalisation or to June 17th, whichever occurred first. We evaluated the possible association between burden of comorbidities and time to in-hospital death using 2 Cox-proportional hazard regression models with 2 different definitions of the exposure: Age-unadjusted Charlson index (Model 1) or individual comorbidities (Model 2). We also evaluated the possible association of other variables at admission including demographics, period of admission, disease severity; CRP, D-dimer; use of anti-Covid-19 drug combinations. A sensitivity analysis in the patients undergoing mechanical ventilation was performed.

All analyses were done using Stata v.14.

Of the 687 patients entered the emergency room of the San Paolo hospital in the period February 24-May 17, 2020, 43 (6.2%) died within 24 hours and 105 (15.3%) did not require J o u r n a l P r e -p r o o f hospitalisation and were excluded. A total of 539 (78.5%) patients were hospitalised for SARS During hospitalisation, most of the patients received hydroxychloroquine+/-azithromycin (427, 79.2%) and low weight heparin at prophylactic doses (355, 65.8%), 134 (24.8%) received antivirals (lopinavir/r or darunavir/c in 126, remdesivir n 8 cases), 122 (22.6%) corticosteroids and 43 (8.0%) immunomodulatory drugs. More than half of the patients were given combinations including hydroxychloroquine +heparin +/-lopinavir/r or darunavir/c or azithromycin (301, 55.8%).

A total of 117 patients (21.9%) required mechanical invasive (N=68) or non-invasive (N=49) ventilation, 156 (29.0%) required cPAP, 202 (37.5%) only high or low flow oxygen support and 64 (11.9%) no oxygen at all.

A number of factors were differently distributed among survivors and non survivors: non survivors were older, more frequently Caucasian, more frequently affected by comorbidities, suffered by a more severe disease at presentation, and showed more frequently higher serum inflammatory parameters than survivors. During hospitalisation, oxygen support was more Intensive in those patients who subsequently died (Table 1) .

We aso also observed that the highest percentage of deaths occurred in the oldest patients, being 58.5% in those aged 80 or above ( Fig 1A) . Percentages of deaths also increased with number of comorbidities, being 19.3% in patients with no comorbidity and 58% in those with at least 4 J o u r n a l P r e -p r o o f comorbidities ( Fig 1B) . By analyzing the percentage of deaths in relation to the age groups and comorbidities together, we verified that no one under the age of 40 died independently of comorbidities, while comorbidities weighed on the number of deaths in other age groups, apart from the oldest one, aged 80 years and over, in which the percentage of deaths is very high independently from comorbidities ( Fig 1C) . We confirmed this different effect of comorbidities on in-hospital death according to age, by identifying an interaction between age and number of comorbidities (interaction p-value < 0.001).

Over 8,228 person-day follow-up (PDFU), 174 patients died in hospital. The Kaplan Meir probability of in-hospital death by 14 days was of 29.5% (IQR:25.5-34.0) (Fig 2A) . The 14-day probability of death was associated with age, being 0% in patients below 40, and highest in those above 80 (52.0%, 95%CI: 43.1-61.6) ( Fig 2B) .

In the unadjusted analysis, a number of factors were associated with time to in-hospital death: age, individual comorbidities and Charlson index, inflammatory markers and d-dimer, severity of disease and therapy combinations.

A severe burden of comorbidity as by age unadjusted Charlson Index (Model 1) and not individual comorbidities (Model 2) was independently associated with the risk of in-hospital death (Table 2) . A number of other variables were independently associated with risk of time to inhospital death: age, with every 10 years older showing 53% higher risk (AHR 1.53, 95% CI: 1.32-1.78); CRP >60 mg/dL (AHR 2.14,95% CI:1.49-3.08); D-dimer >1000 ng/mL (AHR 1.67; 95% CI:1.12-2.46), severe and critical COVID-19 disease at presentation (AHR: 1.77, 95%CI: 1.24-2.53 and AHR 5.27, 95% CI: 2.82-9.85); finally, the use of most of the drugs combinations was associated with reduced risk of in-hospital death ( A total of 55/117 (47.0%) patients who underwent mechanical ventilation died in hospital; again age, Charlson index, inflammatory markers were associated with increased risk of death. In this setting obesity was independently associated with a 2-fold higher risk of in-hospital death (AHR 2.45; 95%CI: 1.11-5.42) (Supplemental Table 1 ).

Looking at the 174 died patients, 32 received invasive mechanical ventilation and 147 not.

Among these last ones, 67 patients showed P/F <100 and were not admitted in ICU and died. Main reasons for not admission were age, presence of severe comorbidities and short life expectancy.

We then studied the dynamics of COVID-19 disease, by investigating patients admitted and disease severity in the 2-week time frames. We observed a first peak of deaths in patients hospitalised in the 2 nd -3 rd week of March (36.5%), and a second one in the last two weeks of April (32.5%). While during the first peak most of the patients presented with severe disease (55.4%) but only 13% were older than 79, during the second peak the disease was less frequently severe (35%) J o u r n a l P r e -p r o o f but more than half of the patients (55%) were aged >=80 and 30% acquired the infection in longterm facilities residency (Table 3 ).

In our study population of 539 patients hospitalised for COVID-19 disease we found an inhospital mortality of 32%, reaching 44% in patients undergoing mechanical ventilation. Patients' age, disease severity at presentation, level of inflammation and concomitant comorbidities appeared to be the main drivers of fatality events. We also observed different waves of patients' admissions Looking at our results in detail, the association of age and comorbidities is particularly evident in the age stratum 60 to 79, showing a fatality rate ranging from 27% in case of no comorbidity, to 62% when 4 or more comorbidities are present.

Main drivers of high in-hospital mortality rate in our cohort are age, disease severity at admission and weight of comorbidities, as represented by Charlson index, as well as inflammatory and pro-coagulatory markers, as shown by others (Ciceri et al. 2020 , Wu C et al. 2020 J o u r n a l P r e -p r o o f 2020). We didn't find an association between individual comorbidities and risk of death, differently from other reports (Ciceri et al. 2020 , Cummings et al. 2020 .

We also observed a reduced risk of in-hospital death according to the treatment received. Interesting is the dynamics of the epidemics in our hospital. Looking at data on the characteristics of patients admitted in the different 2-weeks periods, from the beginning of outbreak, we can observe two different waves, the first one represented by severely ill patients, and the second one, one month later, by very old patients, with less severe disease, in 20% coming from long-term care facilities. The scandal of elderly people getting infected and dying in long-term care facilities occurred in Lombardy, other Italian regions and European countries, determining an epidemic inside the epidemic, affecting fragile people (Surveillance of COVID-19 at long-term care J o u r n a l P r e -p r o o f facilities in the EU/EEA 2020, Survey nazionale sul contagio COVID-19 nelle strutture residenziali e sociosanitarie 2020) Finally, we cannot exclude that the high fatality rate found in our setting could be related to the limited number of beds in ICU, even if these were increased every day while facing the wave of admissions. Actually, to date there are no standardised criteria for invasive mechanical ventilation in COVID-19 patients (Wunsch 2020). A multidisciplinary equip of intensive care, infectious diseases and pneumology specialists, evaluated all the 67 patients with P/F<100 taking into account level of respiratory distress, age, comorbidities and life expectancy.

It is reasonable that the weight of each of these reasons was largely dependent of ICU beds availability, given that theoretically all the patients should have the opportunity of being ventilated even if their life expectancy is short. The balance between health care system offer and individual patient request applies for all health care systems and all diseases, but high income countries give better opportunities to their citizens as compared to low income ones.

Our study has several limitations. First, a number of variables have not been adequately collected, in particularly obesity (Simonnet et al. 2020) , 50% of patients with unknown BMI.

Second, we are not aware whether patients died for COVID-19 disease or with COVID-19 disease:

actually, the association of risk of death with number of comorbidities might suggest either that COVID-19 was more aggressive in fragile patients, or that the cause of death was the pre-existing condition, and COVID-19 acted as a trigger on a precarious condition.

In conclusion, we showed a high rate of in-hospital death in patients with COVID-19 disease in a University hospital in Milan, the first European city to be overwhelmed by the epidemic of SARS CoV-2 infection in Europe. The severity of disease at presentation, the advanced age of patients, the level of inflammation, and the presence of comorbidities, together with the small number of ICU beds are the most likely explanations for the outcome observed. These findings may be useful to better allocate resources of the national healthcare systems, in case of re-intensification of COVID-19 epidemics.

The authors declare that there are no conflicts of interest regarding the publication of this study

☒ All the authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper. 

Leukocytes count, 10^3/uL, median (IQR)

Lymphocyte count, 10^3/uL

Platelets,10^3/uL, median (IQR)

Creatinin, mg/dL, median (IQR)

Pharmacological support, N (%)