Radiographic patterns and severity scoring of COVID-19 pneumonia in children: a retrospective study

The World Health Organization has officially declared coronavirus disease (COVID-19) outbreak a global pandemic. According to the largest epidemiologic survey, the radiological features of COVID-19 have been mostly described in adults. Typical chest radiographic features are peripheral or subpleural ground-glass opacities (GGOs) or consolidation. The distribution is usually bilateral with more prominence in the lower lung than in the upper lung zones [1, 2]. In adults, the radiographic features vary from negative chest radiography (CXR) [3] to diffuse disease between the 1st and 3rd week and may be associated with severe acute respiratory distress syndrome (ARDS); however, cases of children are usually mild and less severe and critical (5.9%) than cases of adults (18.5%) [4‐6]. The incidence of pediatric COVID-19 pneumonia with mild clinical symptoms was low, and one-fourth to one-half of those cases had abnormal CXR [7, 8]. Moreover, children (50%) have less involvement on computed tomography (CT) than adults (91.5%) [9]. To prevent unnecessary radiation exposure, CXR is the preferred imaging procedure in clinical decisions and management of children with COVID-19 pneumonia. Additional CT findings do not change management in childhood [10, 11].

Although the heterogeneous image quality of pediatric CXR affects the interpretation of radiographic findings, CXR highlights the spectrum of findings in COVID-19 [12]. A previous study showed that peribronchial thickening is a major radiographic feature in children, followed by GGOs [13]. The severity scores of CXR focuses on the assessment of COVID-19 pneumonia in adults [14], but the lung scoring system for pediatric CXR has not been widely reported [15, 16]. Therefore, this study aimed to assess the radiographic patterns and severity scores on CXR in children with confirmed COVID-19 pneumonia. We also evaluated the association between the CXR findings, lung scoring system, and clinical patterns (e.g., symptoms, medical treatment, and cycle threshold value).

In Thailand, the third outbreak of COVID-19 occurred in April 2021, coinciding with an increasing proportion of the Delta variant, which is more transmissible and might cause more severe illness than other virus strains during that period. We assumed that most of the patients with COVID-19 in our study were infected with the Delta variant. The proportion of pediatric patients with COVID-19 in our study, who had abnormal radiographic features (106/976, 10.9%), was lower than that in a previously published study (27/59, 46%) [7]. It could be attributed to the different COVID-19 variant between Palabiyik et al.’s study and our study. Our outcomes suggest that an initial CXR can have a role in the initial screening if there is high clinical suspicion of COVID-19 pneumonia. Therefore, our institutions did not routinely perform follow-up chest radiographs in all pediatric patients with COVID-19 who had initially normal CXR findings at presentation. Abnormal CXR features often progress over time, as correlated with clinical information [21]. Overall, 3.4% of adult patients with COVID-19 pneumonia developed ARDS and had a slow resolution of abnormalities over weeks [22]. However, children seem to have milder symptoms and lower mortality and 2.1% of the cases in the pediatric group (younger than 18 years of age) showed progression to severe forms of disease [23, 24]. Herein, only five of 106 pediatric cases (4.7%) had normal initial CXR findings that later became abnormal within 5 days on average. Accordingly, follow-up CXR is necessary for pediatric patients with clinical deterioration or complicated pneumonia after treatment.

Pediatric patients, particularly those aged > 9 years, were symptomatic. However, the children aged < 5 years were asymptomatic. A few studies hypothesized that the younger children might have protective immune system and healthier respiratory endothelial function than older children and adult [25, 26]. We hypothesized that older children and adolescents can communicate their illness more than infants and those in early childhood.

Prior research has used combined CXR and chest CT (sample size = 80) and chest CT alone (varying sample sizes from 1 to 171 cases) to identify the imaging characteristics of pediatric COVID-19 pneumonia [27‐30]. Although chest CT provides more details in mediastinal or hilar adenopathies, more accuracy in lobar distribution, and more sensitivity to detect GGO than CXR, there are still inconsistent CT descriptors in pediatric COVID-19 pneumonia [30]. Moreover, extra information from the initial chest CT scan did not alter the course of treatment for the patient, and the scan should only be used for severely symptomatic patients and/or pre-existing comorbidities [10, 11]. The necessity of chest CT and the radiation risk need to be balanced. The previous study used CT for imaging CT features in COVID-19 pneumonia with severe-to-critical cases (57.5%) and comorbid conditions (50%), whereas our study had only 4.7% (5/106) with severe pneumonia and appropriately focused on CXR [27].

The main finding of peribronchial thickening on CXR in pediatric patients with COVID-19 pneumonia was consistent with that of a previously published article [13]. Peribronchial thickening was the most common finding in our study followed by GGO. The distribution of these lesions was neither peripheral nor perihilar. The combination of peripheral and perihilar zones was found most frequently in all age groups, which is in concordance with the result of another study [7]. Peripheral distribution in children may not be as common as that in adults [31].

In our study, 31.1% and 21.7% of the lesions occurred in the right lower lobe and both lower lobes, respectively. Similar to another study, unilateral lung involvement was more frequent in children than in adults (30% versus 11.7%) [32]. Research conducted on children with COVID-19 pneumonia using chest CT findings assessment also revealed that unilateral lung involvement was greater than bilateral lung involvement (29.4% versus 24.7% in Qi et al.’s study, 54.5% versus 45.5% in Zhang et al.’s study, and 55% versus 45% in Katal et al.’s study) [28‐30]. Distribution was slightly predominant in the right lower lung in our study, which was different from that in a previous study, in which 59–73% of the lesions were in the left lower lung zone [7, 29]. Subtle lesions (especially peribronchial thickening) on CXR may be obscured by cardiac shadow, which mostly occupies the left lung.

Only a few studies have categorized the imaging features into age groups [33, 34]. Our study found the significant differences between age groups and density categories. GGO was predominant in school-age children and those aged > 9 years (73.1%). Whereas peribronchial thickening was the most common in children aged < 5 years (46.3%). These findings are similar to those of a recent study in which GGO/consolidations were more prevalent in older children and perihilar markings were more prevalent in younger children, although there was some controversy regarding age [33]. Nino et al.’s study found that GGO/consolidation was common in aged 3–10 years (36.4%) and 11–19 years (42.4%), whereas peribronchial markings were common in aged 0–2 years (37.5%) and 3–10 years (40.6%) [33]. However, there was no significant difference regarding GGO and peribronchial thickening by age groups in Bayramoglu et al.’s study [34]. The possible explanation could be the small sample size of the presence of GGO (2/20 patients aged 0– < 6 years and 3/26 patients aged 12–18 years) and peribronchial thickening (6/20 patients aged 0– < 6 years and 1/26 patients aged 12–18 years) on the chest radiograph in that study [34].

The study's small sample size of GGO (5/69 patients) and the presence of peribronchial thickening (7/69 patients) on the chest radiograph could be the cause.

The distribution of radiographic severity scores across various pediatric age groups was initially determined by our study. We discovered that 76.4% of patients across all age categories had a mild severity score. There was no statistical significance, despite the fact that the patients with severe radiographic score were limited to the age group < 5 years (Table 3). Our study also analyzed the radiographic severity scoring (in initial and follow-up CXR), correlation with clinical context and medical treatments (Tables 1 and 4). Most pediatric patients with COVID-19 pneumonia in our study had mild lung abnormalities. More extensive lung involvement was found in 21 (19.8%) and 4 (3.8%) pediatric patients with moderate and severe lung abnormalities, respectively. No significant differences were observed between total lung scores and symptoms in the initial CXR. The correlation between initial radiological severity scoring and clinical outcomes has been studied in earlier research [27, 33]. The high severity score was associated with hospitalization, the need for ICU admission, oxygen supplementation, and COVID-19 complications [27, 33]. Similarly, our study found significant differences between severity of lung score and oxygen requirement and additional details in medical treatments. The severe lung scoring group received favipiravir, remdesivir, an antibiotic, and a steroid more than the mild and moderate lung scoring groups.

In our subgroup with radiographic recovery after treatment (25/40, 62.5%), there was no statistically significant difference between their median LOHS and recovery of approximately 10 days. In a subgroup analysis of 40 cases, six patients (15%) had progressive lung involvement at the time of follow-up. However, there was no significant difference between the radiographic findings in follow-up CXR and radiographic severity scores. Unfavorable outcomes were more frequent when initial CXR showed bilateral involvement and a combination of patchy opacities and peribronchial thickenings [12]. We still believe that chest imaging should be used as an adjunct tool in the monitoring of the severe course of COVID-19 pneumonia until further evidence is obtained.

Our study has a few limitations. First, all pediatric patients with COVID-19 were not followed until complete clearance of the radiographic abnormalities. Second, some of the clinical data collected were limited because of incompleteness of the electronic medical records. Third, the time intervals between symptom onset and a sequential chest radiograph were not uniform, affecting the accuracy of our analysis. Finally, the subtle radiographic abnormalities may limit reliability in suboptimal assessment and are highly variable depending on the grade of inspiration on the chest radiographs. Future studies are needed to develop a clinical practice guideline for follow-up CXR to improve the monitoring of the disease course and the time of resolution.

Our study demonstrated significant differences between age groups and density categories in the pediatric patients with COVID-19 pneumonia. The most common radiographic feature in infants and those in early childhood aged < 5 years was peribronchial thickening; however, GGO was more frequent in children aged > 9 years. Both lower lobes were predominantly affected, but the right lower lung was slightly more affected than left lower lung. Chest radiographs are usually indicated to establish a baseline study and to assess for an alternative diagnosis. The role of total severity lung scores in the initial CXR of COVID-19 pneumonia in pediatric patients might predict the clinical outcome and medical treatments.