Comorbidities and Risk Score in COVID-19 Patients

Background

A virus causing clusters of severe pneumonia was first detected in the city of Wuhan, China,
in December 2019.

This pathogen was designated as SARS-CoV-2. Although of probable zoonotic origin,
human-to-human transmission has rapidly fuelled the spread of SARS-CoV-2 infection globally.

On February 20th, the first case of locally acquired SARS-CoV-2 infection was diagnosed in
Northern Italy in a critically ill, hospitalized young man with no travel history to known
areas of viral circulation or link to a probable or confirmed coronavirus infectious disease
2019 (COVID-19) case. Prior to this date, only three cases of COVID-19 had been reported in
Central Italy, all with a travel history to Wuhan. Following this unexpected finding, case
counts, and death tolls has increased rapidly in the country with a total of 192,994
confirmed cases and 25,969 deaths as of 24 April 2020.

Study rationale

Multiple variables have been described as possible risk factors for SARS-CoV-2
susceptibility, severity and prognosis, among which age, sex and comorbidities play an
important role.

Centers for Disease Control and Prevention (CDC) listed the underlying medical conditions
that have shown to increase the risk of severe illness from SARS-CoV-2. While some
comorbidities, such as serious heart conditions and chronic kidney disease have a consistent
and strong evidence as bad prognostic factors in SARS-CoV-2 infection, others as HIV have a
limited evidence and heterogeneous results.

Further, despite it is well-known that the burden of co-existing diseases may be additive or
even multiplicative, the effect of specific disease cluster on COVID-19 adverse outcomes has
never been evaluated. Finally, the proposed models and risk scores currently available to
predict disease severity and mortality are poorly reported and at high risk of bias, raising
concern that their predictions could be unreliable when applied in daily practice. A reliable
risk/prognostic score developed by a multidimensional and cross-validated approach will pave
the way for future research on frail sectors of the population and on the use of health
system resources. At the clinical level, a prognostic score will allow to predict severity
and mortality risk in patients requiring hospitalization and to stratify patients according
to clinical severity helping clinicians in their therapeutic decision-making.

Objectives

The primary objective of the study is to evaluate the role of patient's comorbidities on
clinical outcome in patients hospitalized for SARS-CoV-2. The investigators will confirm risk
predictors already known and provide evidence for the uncertain ones. The investigators will
also develop a prognostic score able to predict negative clinical outcomes (primarily
short-term mortality), that will be useful to stratify patients at hospital admission
according to their different risk profiles, and therefore to "tailor" the individuals' level
of care.

A secondary objective could be that to extend this approach at the susceptible population
level, especially the elders, to stratify according to the higher risk of being infected by
SARS-CoV-2, hospitalized and to have a dismal outcome (not developed here but related to a
possible amendment).

Sample size

The investigators expect the total number of patients with complete data to be approximately
2500, based on the expected recruitment of each center.

Analysis Plan

Data will be summarized by counts and percentage and quartiles for categorical and continuous
variables, respectively. Multi-state models will be used to describe patient's hospital
mortality and discharge. In-hospital mortality will be estimate accounting for discharge as
competing event. Kaplan-Meier estimator will be used to estimate mortality up to 3-months
from admission. The role of patient's comorbidities on clinical outcome will be evaluated by
the Cox model adjusting for relevant confounders. A clinically-based prognostic score will be
developed including comorbidities and other risk factors. The score will be constructed by a
multidimensional approach and Lasso approach will be used to select relevant risk factors.
The Area Under the Receiving Operating Characteristics curve (AUC) and Brier score will be
used to evaluate model performance and the final score will be cross-validated. A sensitivity
analysis will be performed using a training and test validation approach. The use of
regression trees for a practical definition of risk subgroups and latent variable models will
also be considered.

Multiple imputation will be performed if missing would exceed 10%.

Data collection

Consecutive patients hospitalized for SARS-CoV-2 up to October 2020 will be included. Given
the difficulty in systematically obtaining written informed consent and given the great
public interest of the project, the research will be conducted in the context of the
authorizations guaranteed by Article 89 of the General Data Protection Regulation (GDPR) EU
Regulation 2016/679, which guarantees processing for purposes of public interest, for
scientific or historical research or for statistical purposes of health data.