Our observational analysis demonstrates that performing CT scan to diagnose clinically occult pulmonary involvement of COVID-19 in patients who need admission to the hospital for alternative diseases cannot be supported. This is particularly valid in a moment of decrease of the prevalence of the infection during the COVID-19 pandemic.
Our geographic area was first hit by the COVID-19 pandemic outbreak during spring 2020. At that time, our institution identified emergent priorities for the surge of admissions. The priorities included, among others, the need for protecting patients and personnel from the risk of cross-infections inside the hospital. For this reason, it was decided to secure adequate supply of personnel protective equipment and institute separated areas inside the hospital for patients with and without COVID-19 active infection. These areas included beds dedicated both to patients needing intensive or emergency care and to patients needing low-intensity care. Due to the known sub-optimal sensitivity of the first RT-PCR testing and delay in obtaining the laboratory result, our group experienced a strategy based on a combination of lung ultrasound and clinical evaluation to guide a safe initial allocation of patients suspected of COVID-19 [8, 9]. The protocol was then validated by an international multicenter study . Based on that experience, we reiterate here that chest imaging needs to be integrated with the clinical condition and presenting symptoms to allow an accurate diagnostic process. After the first allocation in the ED, in cases that remain doubtful, serial repetition of RT-PCR together with extension of chest imaging to CT scan allow completion of the diagnostic process to guide admission of the patients to dedicated COVID or non-COVID areas . During the outbreak waves, the number of alternative pathologies needing hospital admission dramatically fell to a minimum that was never experienced in the past, and almost all the available beds became dedicated to COVID-19 patients.
However, during these last 2 years, it became evident that the pandemic may manifest with highly variable prevalence . After the first dramatic wave, our institution experienced a fall in the prevalence of COVID-19 and an increase in the number of alternative pathologies needing admissions. Progressively, the COVID-19-dedicated beds were converted to general use, with a growing need to limit the risk of subclinical COVID-19 cases admitted in COVID-19-free areas. To this aim, in our institution, it was deliberated the implementation of a diagnostic algorithm based on the evaluation by CT scan to diagnose clinically occult COVID-19 pneumonia . This algorithm was applied to any patient over 50 years old who needed admission from the ED for any diagnosis alternative to COVID-19 and showing negative RT-PCR tests. The protocol was approved and applied during the period of our analysis; after 3 months of application, our analysis demonstrates its failure.
One of the main objective reasons of failure is the variability in the prevalence of COVID-19. This variability strongly affects the diagnostic accuracy of RT-PCR and chest imaging, the two main diagnostic tools in the management of COVID-19 [4, 11]. For example, during the peak of the surge, the probability that a positive RT-PCR assay is effectively linked to the SARS-CoV-2 infection is extremely high and negativity of the test will have not enough power to exclude, while the opposite happens when the prevalence of the infection drops down. Thus, a negative RT-PCR during a fall in the prevalence is far more sensitive in excluding than during the peak of the surge. The same can be said for chest imaging. In a moment of high prevalence, signs of interstitial pneumonia become highly specific, while they fail in ruling-in COVID-19 pneumonia when the prevalence approaches 0 cases and differential diagnoses become more probable. Moreover, chest imaging can only diagnose the pulmonary involvement during the time course of the disease but cannot be of any help in diagnosing the infection not complicated by pneumonia. Thus, the usefulness of CT scan when used to diagnose COVID-19 is strongly limited [11, 12]. Finally, following the Bayesian reasoning, the accuracy of any diagnostic test fails when applied to patients without clinical suspicion. Thus, like for any other test, CT scan simply loses specificity in diagnosing COVID-19 pneumonia when applied to patients without symptoms of the disease .
In the opinion of these authors, a strategy based on systematic RT-PCR and CT imaging for the identification of COVID-19-infected patients in the ED to guide allocation in watertight compartments inside the hospital may hide pitfalls. Based on doubtful imaging obtained during our study, some patients were allocated in COVID-19 areas, even if sometimes in protected beds, but then revealed not infected. Moreover, our observation is limited by the fact that the algorithm was applied in a moment of fall in the prevalence when the probability of false negative RT-PCR became extremely low. It is not said that, during a rise in the prevalence, the percentage of false negative swabs might become significantly higher. In these cases, CT imaging may be of help in finding clinically occult COVID-19 pneumonia but cannot help in diagnosing those cases without pulmonary involvement. For all these reasons, more than relying on a strict selection in the ED of COVID-19 and non-COVID-19 patients for admissions in separated areas, it is safer considering any patient and personnel a possible source of infection. We speculate that a general strategy based on reinforcement of bed distancing, personnel equipping, and strict monitoring of patients with negative RT-PCR tests inside the areas dedicated to alternative COVID-19 pathologies remains a more efficient and safer alternative.