COVID-19 is a disease that has challenged health systems all over the world1.

The majority of patients progress favorably. However, 20% of patients require hospitalization and their disease can progress rapidly to acute respiratory distress syndrome (ARDS), multiple organ dysfunction syndrome, or even death2. In addition, the symptoms are characterized by a significant clinical-radiological dissociation, such that clinical deterioration occurs suddenly3. Therefore, the identification of risk factors is a priority for the proper management of these patients.

The SEMI-COVID-19 Registry has revealed that mortality in Spain reached 21%, according to data from 20204. In addition, the percentage of hospitalized patients who required mechanical ventilation was as high as 65%.

The aim of this study was to identify factors of poor progress in SARS-CoV-2 infection in patients hospitalized due to COVID-19. Poor progress was defined as onset of severe ARDS during hospitalization (P/F ratio ≤100 mmHg), admission to the intensive care unit (ICU), or in-hospital death.

This is a nested case-control study within the COVID-19 Osakidetza cohort (NCT04463706). In this study, cases were patients admitted to the hospitalization ward due to COVID-19 who presented with clinical deterioration (presence during the hospitalization of severe ARDS (PaO2/FiO2 ≤ 100 mmHg), admission to the ICU, or in-hospital death). The controls were patients hospitalized due to COVID-19 who did not present with severe ARDS. Cases and controls were matched according to age groups. The exclusion criteria were those who died in the emergency department, those who were admitted directly to the ICU, or those who presented with a P/F ratio ≤ 100 mmHg in the first 24 h of hospitalization. Two controls were identified per case. The inclusion period was from March 3, 2020 to April 10, 2020.

The project was approved by the Basque Country Ethics Committee (PI2020059). Simple random sampling was used.

Variables were obtained from the Osakidetza-Basque Health System electronic medical record (EMR) data processing system by means of a review of the EMRs by trained reviewers supervised by collaborating clinicians.

The following variables were gathered: sociodemographic variables (age and sex); personal medical history (comorbidities included in the Charlson Comorbidity Index); disease history (symptoms and date of onset, date of first contact with the healthcare system, persistence of symptoms, treatments indicated, contact with primary care in the previous month); current hospitalization episode (symptoms upon arrival at the emergency department); vital signs (oxygen saturation as measured by a pulse oximeter (SaO2p); clinical signs and physical examination; laboratory tests upon admission (biochemistry: renal function, albumin, liver panel; acute-phase reactants: LDH, CRP; complete blood count and blood differential; coagulation, prothrombin time, and d-dimer); pathological X-ray upon admission, defined as an X-ray which showed previously unknown infiltrates or condensations or data indicative of ARDS (presence of bilateral opacities not fully explained by effusion, lobar collapse, a collapsed lung, or lung nodules).

The lowest P/F ratio (PaO2/FiO2) was calculated at various points in time: upon arrival at the emergency department, the day the patient was transferred to the hospital ward, and through day three of the hospital stay. In addition, the lowest P/F ratio reached during hospitalization was recorded for both the cases and the controls.

A case was defined as a patient who presented with clinical deterioration during hospitalization, defined as presence of a P/F ratio ≤100 mmHg, admission to the ICU, or in-hospital death (date).

A total of 366 patients were randomly selected from the 4447 patients who were hospitalized during the study period in participating centers. Once ineligible patients were excluded, our sample comprised 99 cases and 205 controls, of which 198 eligible patients were selected in order to have a case/control ratio of 1:2 (Fig. 1).

Figure 1.

Flowchart.

(0.44MB).

Males were more numerous among the cases than the controls (79% vs. 65%, p = .014) as were patients who had had an acute myocardial infarction (AMI) (14% vs. 4%, p = .03), diabetes with organ damage (7% vs. 1.5%, p = .025), and hypertension (63% vs. 49.4%, p = .031). A lower frequency was observed among cases versus controls for the symptoms of dry cough (51.5% vs. 65%, p = .021) and sore throat (5% vs. 18.7%, p = .003). More cases than controls had consulted with their primary care physicians (48.5% vs. 34%, p = .016). The plain X-ray was more often pathological among the cases (91% vs. 77%, p = .004). There was an association between levels of glucose, LDH, CRP, and platelets and clinical deterioration; patients who deteriorated presented with higher figures of the first three parameters and lower platelet figures (Table 1). All the variables described in the methods section were considered on the univariate analysis (Table 2).

When these variables were combined on the multivariate model, patients who presented with an O2 saturation in the emergency department of ≤90% had an increased risk of clinical deterioration (OR = 16.57, 95% CI 4–68, p < .001) as well as those who had had AMI (OR = 15.72, 95% CI 3.3–75, p = .0006), hypertension (OR = 2.21, 95% CI 1.1–4.4, p = .024), and COPD (OR = 3.05, 95% CI 1.4–6.5, p = .004). In addition, a pathological X-ray entailed a risk that was 5.6 times greater. CRP levels >100 were positively related to possibility of deterioration, as were blood glucose levels >200. Patients with platelet levels less than 15 × 103/µL presented with a risk of deterioration that was four times higher than those who had platelet levels within normal or elevated ranges (Table 3).

The variability explained by the model measured by R2 was 25% and the predictive power as measured by AUC was 0.86.

The P/F ratio was lower among the cases at the various moments it was analyzed (p < .001 for all). The difference in the P/F ratio trend between cases and controls was also statistically significant overall (p < .001), but can be divided into two intervals: the difference is not significant in measurements upon arrival at the emergency department and hospitalization day one; furthermore, the trends are practically parallel during this period (p = .064). However, after hospitalization day one, the cases continued on a downward trend whereas controls stabilized more (p < .001). On the second day of hospitalization, 75% of the cases had values lower than 253.6 whereas 75% of the controls had values higher than 271.4. The decline in the third day of hospitalization until the day a patient presented with deterioration or the critical day was more marked among the cases (p < .001) (Fig. 2).

Figure 2.

Evolution of P/F ratio during the hospitalization.

(0.13MB).

This study sheds light on the main clinical characteristics of patients hospitalized due to COVID-19 who present with poor progress, defined as PaO2/FiO2 ≤100, admission to the ICU, or death. In this work, it was possible to determine that presence of SaO2 ≤90%, hyperglycemia, elevated CRP levels, thrombocytopenia, a pathological X-ray upon admission, and a previous associated disease (AMI, COPD, or hypertension) were associated with clinical deterioration.

Numerous studies coincide on highlighting age, male sex, and presence of dyspnea as predictive factors of mortality or ICU admission in COVID-19 patients5,6. As this study paired cases and controls by age, we were not able to evaluate this variable’s effect on outcomes. In regard to sex, 78.79% of patients who developed clinical deterioration in this study were men.

A meta-analysis by Jain and Yuan7 evaluated symptoms and comorbidities that were predictors of severity and ICU admission in patients with COVID-19 and found that dyspnea was the only predictive symptom of both outcomes. Unlike other published works, dyspnea was not significantly more frequent among the cases in our study. This is probably related to the exclusion of patients who required mechanical ventilation in the first 24 h, given that patients who were admitted to the ICU in that time period were transferred due to presence of severe ARDS.

It is important to note that our main objective was to identify patients who, hospitalized in the pulmonary phase of the disease as per the phases proposed by Siddiqi and Mehra8, presented with clinical deterioration during hospitalization. Thus, the cases more frequently had an SaO2p that was lower than the controls upon arrival to the emergency department. Despite the fact that the oxygen values were practically within normal ranges, there was an objective clinical repercussion given that in addition, the cases also presented with tachypnea more often, which is undoubtedly a determining factor of poor progress in pneumonia and is included on COVID-19 severity scales9.

Fu et al.’s10 study on 200 patients observed that both age and comorbidities such as low oxygenation were associated with greater mortality. Therefore, it seems that lower oxygenation values upon admission may predict later deterioration. Beyond this, as Fig. 2 shows, there is a point of inflection in oxygenation, which has been termed the “critical day” for patients. Therefore, proper identification of the clinical characteristics of patients at high risk of worsening is crucial. Bilateral lung involvement and ground-glass opacities have been demonstrated to be the most frequent radiological form of presentation in the literature11,12. However, in our study, we could not determine the type of radiological involvement; it was only known that patients who developed clinical worsening during the course of the disease presented with some type of initial radiological abnormality versus the controls, who had normal X-rays.

In regard to associated diseases, it seems logical that patients with chronic diseases such as COPD who presented with chronic lung parenchyma inflammation, limitation of expiratory flow, and greater susceptibility to infections would be more vulnerable to the disease12. Likewise, in line with other Spanish cohorts13, a greater association between COPD and clinical deterioration has been observed, which is likely related to presence of a greater number of comorbidities than COVID-19 patients usually have14.

Hypertension and AMI are significantly related to clinical deterioration15. Zhao et al.16 identified that age >60 years, cardiovascular disease, hypertension, and diabetes were independent factors of mortality due to COVID-19. Another two recent studies indicated that these patients more often had myocardial damage and arrhythmias during the hospitalization, which are associated with greater mortality17,18. Direct viral involvement as well as inflammation secondary to the cytokine storm in SARS-CoV-2 infection could be behind this myocardial damage19. The potential impact this damage may have in the long-term is still to be determined.

Lastly, elevated glucose and CRP levels were predictors of clinical deterioration. Both parameters have been widely studied in pneumonia. Hyperglycemia that is either secondary to the stress of the infection or due to diabetes or an underlying metabolic syndrome could increase mortality in COVID-19 patients20–22.

Diabetes is one of the most frequent comorbidities in COVID-19 and can worsen its prognosis23. An inappropriate immune response along with exacerbation of the proinflammatory storm could explain this association24. Undoubtedly, proper management and control of glucose levels must be a key part of COVID-19 treatment. In regard to CRP, its elevation has been associated with disease severity and, in addition, it seems to have surpassed other biomarkers more commonly used in sepsis25.

On the other hand, the cases presented with lower platelet levels than the controls. A recent retrospective study which included 1476 patients observed that 20.7% of the sample presented with thrombocytopenia, which was significantly associated with in-hospital mortality26. It has been observed that disseminated intravascular coagulation is frequent among those who have died due to SARS-CoV-2 infection, which could explain the increase in platelet consumption27.

This study has certain limitations inherent to its case-control design. In addition, it was not possible to identify the specific radiological involvement, there were some missing analytical variables, and the results are not able to be extrapolated to patients excluded from the study due to deterioration in the first 24 h. In regard to the study period, in the initial phases of the pandemic, it was less likely that patients with mild-moderate disease would have been admitted to the hospital; the majority were severe patients. In addition, variables that were later determined to be predictors of poor progress in COVID-19 patients were not included because at the time of recruitment, they were not routinely requested (for example, ferritin).

New studies are needed that help identify patients at risk of clinical deterioration so that healthcare professionals can make appropriate decisions about their destination and treatment.

This project has been funded by the Carlos III Institute of Health in the extraordinary call for research projects on SARS-CoV-2 and COVID-19 disease, file number COV20/0459, and the Network for Research in Chronic Disease Healthcare Departments (REDISSEC, for its initials in Spanish).

The authors declare that they do not have any conflicts of interest.