Improving the early identification of COVID-19 pneumonia: a narrative review | BMJ Open Respiratory Research

Sung and colleagues9 provide a useful insight into the transition from mild to severe cases in their retrospective analysis of 3060 cases from South Korea. In the mild group (n=2585), symptoms included cough (40%), fever (30%) and dyspnoea (5%). Cough, history of fever and dyspnoea all increased in prevalence between groups, reaching 70.4%, 57.1% and 70.3% in the critical group, respectively.

While Sung et al provides a useful insight into early symptoms in mild COVID-19, other epidemiological studies provide larger datasets for mixed severity and severe COVID-19 cohorts. Two of these studies together with the South Korean mild cohort have been collated into a table for comparison of symptom trajectory (table 1).

Comparing the three different groups (mild, mixed and severe/inpatient), symptoms that may suggest COVID-19 progression include: shortness of breath, dry cough, fever, fatigue, chest tightness, confusion, diarrhoea, vomiting and abdominal pain. Of these, shortness of breath remains the most predictive of disease progression (summary box 1).

There is an association between time from symptom onset to receiving clinical care and disease severity. Within a specified cohort, the longer the delay between symptom onset and admission, the greater the duration of stay and the higher the mortality.10 11 Some patients however, can deteriorate rapidly, requiring intubation within 1 day of symptom onset.12–14

This was highlighted well in a study from Mexico. Data were collected from 65 000 patients who had suspected COVID-19 between January and April 2020, with an overall Case Fatality Rate of 3.32%. The time from initial symptoms to actual clinical suspicion of COVID-19 was recorded and then analysed against medical disposition decision and mortality. Fourteen per cent of patients presented to healthcare within 24 hours of first symptoms and of those nearly half were admitted directly to hospital, with 2.8% being admitted to intensive care unit (ICU). The overall mortality for those presenting within 24 hours of symptom onset was 5.2%. Mortality fell to 2.5% if the patient presented to medical care when the symptom onset was between 1 and 3 days and rose significantly to 3.6% in the 4–7 days group. Where patients were admitted after 7 days of symptoms, mortality rose even further (4.1%). The delayed presentation group had a 59% higher chance of ICU admission if presenting after 7 days versus presentation between 1 and 3 days.15

While there have been reports of a ‘second hit’ deterioration in patients occurring between 5 and 10 days of symptom onset, the evidence for such a trajectory is limited. It remains feasible that such a ‘second hit’ merely relates to the delayed presentation of COVID-19 pneumonia and natural evolution of the disease. Regardless of what future research will show, there is clear evidence of sudden and rapid deterioration in a proportion of patients with COVID-19,12–15 and as such, vigilance throughout the illness seems warranted.

Identifying disease progression in COVID-19 has been challenging in part due to the variable symptoms and in part due to the minimal changes that occur in basic observations (blood pressure (BP), temperature (Temp), oxygen saturations (SpO2), heart rate (HR) and RR) despite severely progressive disease.

Even within a relatively severe cohort, the observation changes are modest. Data from the multicentre retrospective inpatient cohort study, International Severe Acute Respiratory and emerging Infection Consortium cohort (ISARIC) (n=122 361), yield an overall mortality rate of 31% (where outcomes are recorded) and mean CRP of ~90 mg/L. Despite this level of severity, the mean observations in the age bracket of 60–69 years at presentation were RR: 22, HR: 91, SpO2: 95%, temp 37.3° and systolic BP: 130 mm Hg (table 2).16

Oxygen saturations are the most consistent predictor for disease severity.18 23 24 Oxygen saturations below 95% have been reported to be associated with twice the risk of death in comparison with normoxaemia at presentation.25

While SpO2 seems to have a fairly clear relationship with disease progression, shortness of breath is more complicated. Silent hypoxia is the popular term used to describe patients who have limited sensation of feeling short of breath and/or no increased RR and yet when examined are found to be hypoxic.26 This has been a feature of COVID-19 since the first cases were reported.27 Imaging studies also demonstrate the disconnect between severity of lung pathology and the sensation of breathlessness. In a study from Marseilles, France, 757 (68%) of patients who did not complain of breathlessness had pneumonia on CT.28

The same study investigated the prevalence and consequences of hypoxia without the sensation of dyspnoea and reported 28.1% of patients who did not complain of dyspnoea were in fact hypoxic on blood gas analysis (n=96). The investigators also reported a dramatically increased rate of ICU admission in such patients of 42.6%, compared with 5.7% in dyspnoeic hypoxic patients, the mortality rate being 20.4% versus 5.7%, respectively. Such ‘silent hypoxic’ patients presented later than others in the group (half after day 5 of symptoms).28

Another useful study examined first responder observations of patients with COVID-19 throughout the first wave of infections in March 2020 in Paris (n=1201). There was a marked disconnect between level of hypoxia (mean SpO2 of 90%) and RR (mean 20), comparing with the previous year where the mean SpO2 was 96%, and RR was 22.29

The inability to rely on the self-reported symptom of breathlessness to identify progressive disease and hypoxia presents a substantive challenge.

Very little has been published on clinical examination findings in COVID-19 pneumonia, beyond the value of basic observations as discussed previously. Reports have suggested certain characteristic features on auscultation (such as ‘velcro’ crackles, atypical bronchial breathing and basolateral distribution) that may denote underlying COVID-19 pneumonia and correlate with radiological changes. Such findings are subject to observer variation however, and the absence of such signs does not denote absence of disease.35

Inferentially, there is likely to be a demonstrable value in the ‘eyeball’ clinical assessment of patients with COVID-19. A Danish study compared an established triage system (based on observations and presenting complaint) versus an ‘eyeball’ assessment by a phlebotomist. The eyeball assessment was significantly superior, particularly in detecting those who may be incorrectly triaged low (green or yellow categories).36 A further study examined physician and nurse predictions of mortality in ED at first assessment. Observations were to hand, but no test results. Both groups performed well, and performance improved with years of experience. When combined (ie, both physician and nurse were in agreement), the predictive value was excellent.37

While formal clinical examination is of unknown value, there is likely significant value in assessment by an experienced healthcare professional.

CRP has been reported as a reliable marker for disease severity and is routinely available including through point-of-care testing.38–40 The optimum cut-off value to indicate significant risk of disease progression remains unknown. It has been previously established, however, that the majority of viral infections we are likely to encounter will rarely raise CRP >30 mg/L, and a CRP of >30 mg/L during a viral illness such as influenza would typically indicate progression of disease and risk of viral pneumonia.41 Further elevations raise the possibility of severe inflammation, a bacterial or other invasive infection.42 Additionally, imaging studies have consistently demonstrated that even a modest CRP rise is associated with infiltrative changes on CT prior to respiratory symptoms.43 In this regard, CRP can be of use in identifying patients with disease progression.

CRP also has some reported use in the hyperacute COVID-19 phenotype. Manson et al14 identified a significant subgroup of patients presenting to two tertiary hospitals in the UK in March 2020, where CRP >150 mg/L or doubling from 50 mg/L within 24 hours was strongly predictive of death or the need for intubation within the following 24 hours. A further analysis from South Korea reported that an admission CRP >80 mg/L had a higher sensitivity for predicting adverse outcome in COVID-19 than a NEWS score of 2 or more.25

There are several limitations with using CRP as a monitor for disease severity, or to detect those patients that require closer follow-up. Primarily, CRP has a lag time before rising. CRP results indicate the severity of inflammation from the previous day (with levels peaking 6–72 hours following an insult).44 This makes the utility of an isolated CRP measurement taken at the point of symptom onset fairly limited. Furthermore, CRP rises in the elderly or those with multiple comorbidities are often blunted. In a multicentre study (although an inpatient high severity cohort), initial CRP measurement correlated with COVID-19 disease severity in all age groups except those in the >75-year-old age group.45 Early changes in CRP in the elderly may still be predictive, although less so.

There are other biochemical parameters that have been associated with deterioration in the patient with COVID-19. Lymphocyte count, lactate dehydrogenase and D-dimer have been quite reliably reported as prognostic markers in the patient with COVID-19, although predominantly in inpatient cohorts. Whether there is clinical utility for monitoring these markers in the community setting remains unknown. Lymphocyte count may provide better prognostication in the elderly and offset the shortcomings of the CRP in this patient group.38 39

In those cases where there are some concerns of progression but equivocal objective signs, there is a question as to the utility of CXR within the community setting.

Borakati et al produced a useful study exploring this, although in a high-severity cohort (n=763, mortality=24%). More than half of the patients had ‘classic’ CXR signs of COVID-19, and as such would have confirmed COVID-19 pneumonia (sensitivity 0.56 (95% CI 0.51 to 0.60)). Of considerable note, nearly one-third of PCR negative referrals with ‘possible’ COVID-19 also had ‘classic’ COVID-19 features on CXR.46

An urgent care cohort that was likely less severe (n=636) reported around 40% of confirmed cases had a positive CXR, although this was based on a consensus view of 12 radiologists and in the absence of objective disease severity markers.47 A further milder cohort from Hong Kong (n=64, dyspnoea rate=6%) with a mean age of 56 years (SD: 19) reported 69% of chest X-rays were abnormal at baseline (but this also included all consolidations).48