Population, study design and data collection

A retrospective cohort study was carried out. We included all patients over 18 years of age treated between April 1 and 30, 2020 in the emergency department of HNERM in Lima, Peru, who were confirmed cases of COVID-19 and who had chest CT scans requested during their initial evaluation.

We used the operational definition by the Ministry of Health (MINSA) ^{21 )} and, therefore, the suspected cases who had positive molecular or serological test for SARS-CoV-2 infection were considered as confirmed cases. Patients who had incomplete data in their clinical record or a CT scan with important artifacts that limited their evaluation were excluded. Thus, out of 309 patients with a chest CT scan and confirmed diagnosis, 53 with incomplete data and two with significant artifacts were excluded, resulting in a total of 254 patients (Annex 1).

Demographic and clinical variables were obtained from the electronic medical record, such as age in years (>60 years, ≤60 years), sex (female, male), clinical severity (mild, moderate, severe), time of illness in days, symptoms (dyspnea, fever ≥38 °C, cough, malaise, sore throat, diarrhea), comorbidities (arterial hypertension, diabetes mellitus (DM), obesity, chronic kidney disease, asthma) and time of hospitalization in days.

Laboratory data were collected, such as leukocytes (x10³/uL), lymphocytes (x10³/uL), aspartate aminotransferase (AST) (U/L), alanine aminotransferase (ALT) (U/L), lactate dehydrogenase (LDH) (U/L), prothrombin time (PT) (s), activated partial thromboplastin time (APTT) (s), creatinine (mg/dl), C-reactive protein (mg/dl), serum ferritin (ng/ml) and D-dimer (ug/ml).

Clinical severity was determined according to the guidelines established by MINSA ^{( 21}: in mild cases there were at least two signs or symptoms of COVID-19; in moderate cases there were at least one hospitalization criterion (dyspnea, respiratory frequency (RF) > 22 breaths per minute (bpm), oxygen saturation <95%, altered level of consciousness, hypotension, signs of pneumonia, lymphocyte count <1000 cells/μL) and in severe cases there were signs of sepsis and/or respiratory failure (at least two of RF >22 bpm, PaCO₂ <32 mmHg, altered level of consciousness, hypotension, PaO₂ <60 mmHg or P/F ratio <300, clinical signs of muscle fatigue, serum lactate >2 mOsm/L).

Image collection and analysis

Chest CT scans were carried out in the hospital with the patient in the supine position, contrast material was not used, in a Toshiba Aquilion Prime tomograph with 80 detector rows and multislice system.

Image analysis was carried out by two radiologists with 3 and 21 years of experience (JCG and MM), who used the tomographic terms listed in the Fleischner Society glossary ²². The following tomographic findings were registered: ground-glass opacities, consolidations, crazypaving pattern, interlobular interstitial thickening (IIT), “reverse halo” sign, parenchymal bands, air bronchogram, pleural effusion, adenopathy (mediastinal lymph node with short axis >10 mm) and the “subpleural preservation” sign (1-2 mm region of non-opacified subpleural lung parenchyma). The distribution of the lesions (peripheral, central or mixed), their pattern (patchy or diffuse), as well as the number of lung lobes involved were also registered. The TSS was calculated by assigning a value between 0 and 4 to each of the five lung lobes according to their degree of involvement: 0 (0%), 1 (1-25%), 2 (26-50%), 3 (5175%) and 4 (76-100%), and the sum of these values determined the final score, with values between 0 and 20 ¹⁶.

Statistical analysis

Quantitative variables were expressed as mean and standard deviation (SD) or as median and interquartile range (IQR), while categorical variables were expressed as percentages.

For qualitative variables, Fisher’s exact test or the Chi-square test were used to determine differences between survivors and deceased. The normality of quantitative variables was assessed by using histograms and the Jarque-Bera test. If the distribution was normal, the Student’s t-test was used; otherwise, the Wilcoxon rank sum test was used.

We used an intraclass correlation coefficient (ICC) to assess the consistency between both observers in the assessment of TSS. Likewise, ROC curves were constructed to determine the cut-off points with the highest combination of sensitivity and specificity to identify moderate and severe cases.

The tomographic findings associated with clinical severity were evaluated using a multivariate logistic regression model, grouping the mild and moderate categories and comparing them with the severe cases category. The model was adjusted for variables such as age, sex, time of illness, and comorbidities, such as arterial hypertension (AHT), obesity and diabetes mellitus, since they are recognized risk factors for severe and mortal cases in patients with COVID19 ^{23 - 26}.

To determine the tomographic findings associated with mortality in patients with severe disease, we carried out a multivariate Cox regression analysis adjusted for the same variables of the logistic model, obtaining hazard ratios (HR) for mortality. The proportionality of the HR of each parameter and of the overall model was verified using the Schoenfeld residuals plot and the nonzero slope test. Survival time in days was counted from the day of patient admission to the eventual death, hospital discharge or end of the follow-up period (86 days). Statistical analysis was carried out using Stata version 16.1.

Ethical aspects

This study was approved by the Specific Research Ethics Committee for COVID-19 of EsSALUD and the Institutional Research Ethics Committee (IREC) of the Universidad Peruana Cayetano Heredia (certificate E086-16-20) under the category of “exempt from review”. The application of an informed consent form was not required since the study only included the review of clinical records and tomographic results, preserving patient confidentiality.

Demographic, clinical and laboratory characteristics

Of the 254 patients included in the study, 137 (53.9%) had positive RT-PCR and 117 (46.1%) had positive serology; 173 (68.1%) were male and 81 (31.9%) female; median age was 59.1 (SD: 14.7 years). Regarding clinical severity, 13 (5.1%) cases were mild, 89 (35.0%) moderate and 152 (59.9%) severe. At the end of the follow-up period, 158 (62.2%) patients were discharged, 95 (37.4%) died and 1 (0.4%) remained in the intensive care unit (ICU). The remaining demographic and clinical variables are shown in Table 1.

The variables that showed significant differences between survivors and deceased were age, dyspnea, fever, diarrhea, AHT and length of hospitalization (Table 1). The laboratory tests evaluated at admission that showed significant differences between survivors and deceased were leukocytes, lymphocytes, LDH, PT, APTT, creatinine, C-reactive protein, serum ferritin and D-dimer (Annex 2).

Tomographic findings

Of the 254 chest CT scans, 243 (95.7%) showed ground-glass opacities; 200 (78.7%), consolidations; 184 (72.4%), “subpleural preservation”; 182 (71.7%), IIT; 120 (47.2%), air bronchograms; 90 (35.4%), “crazy-paving” pattern, and 47 (18.5%), parenchymal bands. Adenopathies were observed in 15 (5.9%) patients, pleural effusion in 7 (2.8%) and the “reverse halo” sign in 5 (2.0%). The most frequent lung lesions had mixed distribution (78.3%), followed by those with a diffuse pattern (74%) and those with patchy pattern (26%).

The “subpleural preservation” sign, IIT, “crazy-paving” pattern, adenopathy, mixed distribution and diffuse pattern of lesions were more frequent in the deceased; while parenchymal bands, peripheral distribution of lesions and patchy pattern were more frequent in the survivors. All these findings showed significant difference between both groups (Table 2).

Regarding the extent of pulmonary involvement, we observed bilateral involvement in 250 (98.4%) cases; the five pulmonary lobes were more frequently involved together (94.5%), and of these, the right lower lobe (RLL) was the most frequently involved (61.4%) (Annex 3).

Tomographic severity score (TSS)

The application of the TSS by both evaluators showed an adequate repeatability analysis with an ICC of 0.93 (95% CI: 0.85-0.97). The median TSS value in the 254 cases was 13 (IQR: 11-15); for the survivors it was 12 (IQR: 9-14) and for the deceased it was 16 (IQR: 14-17), a significant difference was found between both groups, p<0.001 (Table 2). Regarding the degree of clinical severity, the median TSS value for moderate cases was 11 (IQR: 9-13) and for severe cases it was 15 (IQR: 13-16) (Table 3).

The ROC curve analysis for moderate cases showed an area under the curve (AUC) of 0.99 (95% CI: 0.98-1.00) and a cutoff point of 7 showed sensitivity of 94.4% and specificity of 100%. The ROC curve for severe cases showed an AUC of 0.86 (95% CI: 0.82-0.90) and a cutoff point of 13 showed sensitivity of 84.9% and specificity of 70.6% (Annex 4).

Figure 1. presents tomographic images of three patients included in the study, and the description of the main findings identified and the assigned TSS value.

Figure 1 Axial non-contrast chest tomography (a, b, c). a) 31-year-old woman with fever, malaise and diarrhea. There are ground-glass opacities (white arrow) with peripheral distribution, and the sign of “subpleural preservation” (black arrow); tomographic severity score (TSS): 2. b) 45-year-old male with fever, cough, dyspnea, diarrhea and chest pain. There are consolidations (black arrows) with central and peripheral distribution; TSS: 9. c) 45-year-old male with fever and dyspnea. There are extensive areas of “crazy-paving” pattern (white arrows) and with diffuse pattern in both lungs, as well as the sign of subpleural preservation (black arrow) and consolidation (black arrow), TSS: 19. 

Multivariate analysis, Cox regression and Kaplan-Meier survival curve

The tomographic findings associated with clinical severity were the diffuse pattern (OR: 3.23; 95% CI: 1.46-7.14), “crazy-paving” pattern (OR: 2.48; 95% CI: 1.08-5.68) and a higher value in TSS (OR: 1.73; 95% CI: 1.49-2.02) (Table 3).

The tomographic findings associated with higher mortality in severe patients according to the Cox regression analysis of survival were the presence of “crazy-paving” pattern (HR: 1.78; 95% CI: 1.03-3.06) and a higher TSS value (HR: 1.33; 95% CI: 1.20-1.48) (Table 4). In addition, we identified that survival in severe patients, after 86 days of follow-up, was lower in those with TSS values ≥14 compared to those with TSS values <14. This result is shown in the Kaplan Meier curve presented in Figure 2.

Figure 2 Kaplan-Meier survival curve. Comparison of estimated survival rates in severe cases with tomographic severity score (TSS) <14 and severe cases with TSS ≥14.