Prognostication of Oxygen Requirement in Non-severe SARS-CoV-2 Infection

Background

At the time of writing the novel betacoronavirus SARS-CoV-2 has infected more than 11 million
individuals and killed more than 500,000. The most recent epidemiological surveillance curve
shows that the number of new infections continues to rise, with over 230,000 new cases
reported in the last 24 hours.

Whilst the absolute number of deaths attributable to COVID-19 is substantial, effectively
managing high patient volumes is also a major challenge facing health systems, particularly
in under-resourced settings. With an estimated case-fatality rate amongst symptomatic
patients of between 1-2%, it is vitally important that health workers are able to accurately
identify the majority of patients at low risk for progression to severe disease. These
patients can be safely discharged away from the health facility ensuring the available
resources are allocated to patients most likely to benefit. If low-risk patients cannot be
readily identified, there is a real risk that health facilities in these regions will be
overrun, with consequent substantial avoidable mortality.

COVID-19 has already started to hit extremely vulnerable populations in refugee camps and
conflict areas. In Bangladesh, cases within the Rohingya refugee megacamps have already been
reported, and local transmission is ongoing. Such contexts already suffer from a lack of
medical facilities and even a moderate number of COVID-19 cases will overburden existing, and
proposed, capacity. Having the ability to prognosticate the need for supplemental oxygen (the
main treatment available in these settings) through measurement of parameters available at
the time of arrival at a healthcare facility would strengthen the capability to identify
those patients presenting with moderate symptoms that can be safely discharged away from the
facility. Based on recent estimates, only 20% of all symptomatic patients with COVID-19
develop a requirement for supplemental oxygen.

Although existing prognostic scores have yielded disappointing results in patients with
SARS-CoV-2 infection a number of demographic, clinical and laboratory parameters are
associated with a more severe disease course and worse patient outcomes. However, to our
knowledge, few studies have examined the performance of prognostic markers in patients who do
not require supplemental oxygen at presentation. Only one of these studies included
outpatient or ambulatory care settings and none were conducted in resource-limited settings.
Hence, whether measurement of these parameters can inform the decision to safely discharge a
patient away from a health facility is as yet unclear.

Main research question

In adults presenting to care with non-severe COVID-19, can subsequent need for oxygen be
predicted from parameters measured at the time of arrival at a healthcare facility?

This study will evaluate several clinical and biochemical biomarkers that have been
identified as possible predictors of deterioration in patients with COVID-19. The primary
objective is to develop a prognostic tool combining up to four markers (including a maximum
of two biochemical markers) with a high negative predictive value (NPV) for subsequent
supplemental oxygen requirement (WHO Grade ≥ 5). Clinical biomarkers will be limited to
simple clinico-demographic variables (for example, age, sex and duration of symptoms) in
order to ensure the tool remains as simple as possible. Biochemical biomarkers for which
point-of-care (POC) and/or near-patient tests are either already commercially available or in
late-stage development have been prioritised, in order to maximise the chance of translation
on to the field within a time period that is useful for the current global pandemic response.

The requirement for supplemental oxygen has been selected as the primary outcome: it is a
relatively objective endpoint (based either on the measurement of peripheral oxygen
saturation [sO2 ≤ 93%] or a respiratory rate > 30 breaths per minute or a clinical decision
to give supplemental oxygen) and from a practical perspective is the main evidence-based
therapeutic intervention available in such settings, but in very limited supply in the
majority of low- and middle-income countries (LMICs).

We have identified near-patient tests for one of the priority biochemical biomarkers that
have not yet been field-evaluated in LMICs (suPAR). We will evaluate the performance of this
test under field conditions. We have selected this test for evaluation because unlike the
other potential markers for which near-patient tests also exist, there is a relative paucity
of field experience in tropical settings. As all biochemical markers will be measured using a
validated multi-analyte quantitative immunoassay (using the Ella platform), this study will
provide an opportunity to evaluate the field performance of these tests. If either of the
markers is found to be useful prognosticators for patients with non-severe SARS-CoV-2
infection, this may facilitate a more rapid translation of the results of this study into
practice.

Assumptions, limitations, and generalisability

The main assumption that underlies this work is that a validated, rapid, low-cost
antigen-based test for SARS-CoV-2 will become available in the near future. There is reason
to believe that this will be the case.Availability of such a test would allow the prognostic
tool developed in this study to be applied to the appropriate patient population at the point
of arrival at a health facility. Whilst the absence of such a test would undoubtedly limit
the utility of the tool, it is likely that when local incidence of patients presenting with
confirmed SARS-CoV-2 infection exceeds a threshold, use of the tool could be justified based
upon a clinical case definition alone. Furthermore, in order to mitigate this risk, we plan
to use the data collected in this study (specifically the baseline and outcome data collected
from participants who are subsequently confirmed to be SARS-CoV-2 negative by RT-PCR) to
inform exploratory simulation analyses, with the aim of customizing of the tool for use in
different prevalence settings, thereby extending utility to different phases of the pandemic.

An additional limitation of this study is that due to its single-site nature we will not be
able to explore whether the performance of the prognostic markers is impacted by different
host populations, or perform an external geographic validation of the prognostic model. To
mitigate this risk we have designed a pragmatic study that could be implemented in other
sites. We are actively exploring options for conducting the study at other sites if
sufficient additional funding can be identified. We will inform and seek approval from all
relevant regulatory bodies (international and local) prior to the inclusion of any additional
study sites.