Abdominal Imaging Findings on Computed Tomography as a Tool for COVID-19 Mortality Risk Assessment: Comparison With Chest Radiograph Severity Scores : Journal of Computer Assisted Tomography

Secondary Logo

Journal Logo

Abdominopelvic Imaging: Gastrointestinal

Abdominal Imaging Findings on Computed Tomography as a Tool for COVID-19 Mortality Risk Assessment: Comparison With Chest Radiograph Severity Scores

Balthazar, Patricia MD; Mercaldo, Nathaniel PhD; Pisuchpen, Nisanard MD†,‡; Mendoza, Dexter P. MD; Little, Brent P. MD; Flores, Efren J. MD; Kambadakone, Avinash MD

Author Information
Journal of Computer Assisted Tomography 47(1):p 3-8, 1/2 2023. | DOI: 10.1097/RCT.0000000000001393
  • Free

Abstract

The global COVID-19 pandemic, caused by the severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), has had a devastating impact on the United States, affecting more than 34 million people and resulting in more than 610,000 deaths as of June 1, 2021.1 Most commonly manifesting with fever and respiratory symptoms, the lung imaging findings of COVID-19 are well described in the literature.2–4 As the disease became more widespread, gastrointestinal symptoms and abdominal imaging findings also became evident.4–7 The proposed mechanism is the expression of angiotensin-converting enzyme 2 not only in the lungs but also in the gastrointestinal, hepatobiliary and vascular cell membranes.8

Multiple disease severity models have been studied for the prognostication of COVID-19, some of which include well-validated chest imaging severity scores.9–14 However, the added value of abdominal computed tomography (CT) findings in the prediction of COVID-19 mortality has yet to be explored. We aimed to quantify the association between CT abdomen and pelvis with contrast (CTAP) findings and CXR severity score, and the incremental effect of incorporating CTAP findings into predictive models of COVID-19 mortality.

MATERIALS AND METHODS

Study Design and Setting

Institutional review board approval was obtained for this Health Insurance Portability and Accountability Act–compliant retrospective cross-sectional study performed at a large quaternary care medical center. The institutional review board waived the requirement for informed patient consent.

Study Cohort

We included all adult patients (≥18 years) who presented to our institution between March 15, 2020, and June 15, 2020 (first wave of the pandemic) with the diagnosis of COVID-19 by PCR test and had a chest radiograph (CXR) the same day or up to 48 hours before a CT abdomen and pelvis with contrast (CTAP) examination. Two patients were excluded due to primary admission diagnosis of trauma with intra-abdominal hemorrhage. For patients with multiple CTAPs with prior CXR, only the first study was included. For patients with multiple CXR within 48 hours before CTAP, only the closest one to CTAP was included. A flowchart of our patient population is presented in Figure 1.

F1
FIGURE 1:
Patient population flowchart.

Outcomes and Measures

Our primary outcomes were the severity of lung disease before CTAP, measured by using the modified Radiographic Assessment of Lung Edema score on CXR, and mortality within 14 and 30 days.

Data Collection

We queried our institutional research patient database repository to identify all patients with COVID-19 in the study period and their radiology examinations. Subsequently, we identified those with CTAP and at least 1 prior CXR up to 48 hours.

Chest radiographs were independently reviewed and scored by 2 fellowship-trained cardiothoracic radiologists blinded to both clinical information and the abdominal findings (B.P.L. with 12 years of subspecialty experience and D.P.M. with 2 years of subspecialty experience). Each lung received a score from 0 to 4 based on the percentage involvement of the opacified lung (0 = no involvement; 1, <25%; 2, 25%–50%; 3, 50%–75%; 4, >75%). This score was then multiplied by a second score representing the overall density of the opacity (1 hazy; 2 moderate; 3, dense). The sum of the scores from each lung is the total CXR severity score.13,15 If a score differed by less than 2 points, then the average score was used as the final CXR score. If the total CXR score differed by more than 2 points, then the CXR was reviewed by a third fellowship-trained cardiothoracic radiologist (E.J.F. with 10 years of subspecialty experience) and the average of the 3 scores was used as the final score. These final scores were then categorized as mild (0–4), moderate (>4–10), and severe (>10).15

The CTAPs were independently and blindly reviewed by 1 fellowship-trained abdominal radiologist (N.P. with 3 years of subspecialty experience) and 1 abdominal imaging fellow (P.B.) to extract the following variables: pyelonephritis, acute tubular necrosis (Fig. 2), distended fluid-filled bowel (Fig. 3), bowel wall thickening, pneumatosis, portal venous gas, bowel perforations, gallbladder sludge, heterogeneous liver (Fig. 4), solid organ infarct, peritonitis, major vessel thrombosis, hematoma. For the purpose of this study, CTAPs with none of these findings were classified as negative, whereas positive CTAPs had at least 1 of the findings. Because there is no current consensus for some of the nonspecific findings, they were subjectively assessed by the reviewers and if discordant, a third reviewer (A.K. with 10 years of subspecialty experience) made the final determination.

F2
FIGURE 2:
Acute tubular necrosis example. Axial (A) and coronal (B) abdomen and pelvis CT shows relative hypoenhancement and loss of corticomedullary differentiation of both kidneys.
F3
FIGURE 3:
Distended fluid-filled bowel example. Axial (A), coronal (B), and sagittal (C) abdomen and pelvis CT shows distended fluid-filled colon with air-fluid levels.
F4
FIGURE 4:
Heterogeneous liver example. Axial (A) and coronal (B) abdomen and pelvis CT shows heterogeneous liver enhancement pattern with geographic areas of hypoenhancement in the right hepatic lobe.

Statistical Analysis

Descriptive summaries were computed for continuous and categorical variables; continuous variables were summarized as the median and interquartile range (IQR) and categorical variables as frequencies (proportions). Differences in the distribution of CXR scores by CTAP findings were assessed either using the χ2 test (categorical variables) or the Kruskal-Wallis test (continuous variables).

Both univariable and multivariable proportional odds logistic regression models were constructed to quantify the association between CXR score and CTAP findings. Odds ratios (ORs), and their 95% confidence intervals (95% CIs) and P values, were estimated to summarize odds of having a higher CXR score when a CTAP finding was observed. Variables that were not identified or very rarely identified (less than 3 cases) were not included on the multivariable analysis.

A penalized (LASSO) logistic regression model was constructed to identify key predictors (demographics, CTAP findings, and CXR score) and their functional relationship with death within 14 days. A separate model was developed for death within 30 days.

For each outcome, we first constructed these main effects-only models (linear no interactions) using only the subset of data with complete data (10% [n = 20] of subjects were missing race/ethnicity). Categorical variables were modeled as a series of indicator variables, whereas continuous predictors were modeled using either linear or nonlinear (restricted cubic splines) terms. All variables were standardized before inclusion of the model building process. 10-fold cross-validation was used to identify the LASSO tuning parameter (lambda = 1 standard error) and associated parameter estimates, as well as model predictions. These predictions, and the true outcome status, were then used to estimate receiver operating characteristic (ROC) curves and area under the curve (AUC) values and CIs. Area under the curve estimates based on random forests were used as an informal comparator to the penalized logistic regression modeling approach.

Multiple imputation, using chained equations, was performed to account for the missing demographic information. Death data were not included in the imputation modeling to reduce data leakage, which would lead to overly optimistic predictive performance. Parameter estimates and predictions were stored for each analysis performed on each of the 100 imputed data sets. We define variable importance as both the absolute value of the average penalized regression coefficient and the percentage of times these estimates were nonzero. Final predictions were generated by averaging model predictions across all imputed data sets. Like above, these final predictions were used to estimate ROC curves and their corresponding AUC values.

To assess the incremental effect of CTAP findings, all analyses were repeated without CTAP findings in the eligible feature set for the penalized logistic regression models and the random forests. P values less than 0.05 indicated a significant difference. All statistical analyses were performed using R (version 4.0.2; R Foundation for Statistical Computing, Vienna, Austria) and the mice, glmnet, and pROC libraries.

RESULTS

Characteristics of 195 included patients with COVID-19 who had a CXR and a CTAP within 48 hours during the study period are summarized in Table 1. Most patients were White (55%) and English-speaking (56%). The median (IQR) patient age was 63 years (50–75 years). Mild, moderate, and severe CXR severity scores were observed in 55.4%, 26.7%, and 17.9% of patients, respectively. Approximately half of the patients (50.3%) had a normal CTAP. The most common findings on CTAP were distended fluid-filled bowel (15.9%), gallbladder sludge (15.9%), and bowel wall thickening (13.8%).

TABLE 1 - Descriptive Summaries of 195 Patients With COVID-19 Who Had a CTAP and a CXR Within 48 h
Variables Summary
Sex
 Male 119/195 (61.0)
Age 63 (50, 75)
Race/ethnicity
 White 97/175 (55.4)
 Black 23/175 (13.1)
 Hispanic 12/175 (6.9)
 Asian 7/175 (4.0)
 Other 36/175 (20.6)
CXR score 4 (0.75, 8)
CXR severity category
 Mild: [0–4] 108/195 (55.4)
 Moderate [>4–10] 52/195 (26.7)
 Severe [>10–24] 35/195 (17.9)
CTAP
 Negative 98/195 (50.3)
 Positive (any finding) 97/195 (49.7)
CTAP findings*
 Gallbladder sludge 31/195 (15.9)
 Distended fluid-filled bowel 31/195 (15.9)
 Bowel wall thickening 27/195 (13.8)
 Ileus 12/195 (6.2)
 Pyelonephritis 11/195 (5.6)
 Solid organ infarct 8/195 (4.1)
 Heterogeneous liver 7/195 (3.6)
 Peritonitis 7/195 (3.6)
 Hematoma 6/195 (3.1)
 Major vessel thrombosis 3/195 (1.5)
 Bowel perforation 2/195 (1.0)
 Acute tubular necrosis 1/195 (0.5)
Categorical variables are summarized as frequencies (proportions) and continuous variables are summarized as median (IQR).
* Pneumatosis and portal venous gas were not observed within this cohort.

Association Between CTAP and CXR Severity

Positive CTAP status was more frequently seen in patients with severe CXR severity category (74% versus 26% compared with normal CTAPs). The odds of having a positive CTAP status was 3.89 (95% CI, 1.67–9.10) times greater when a CXR score was classified as severe compared with mild (P = 0.002; Table 2). Normal CTAPs had a median CXR severity score of 3.0 (IQR, 0.0–5.9), whereas positive CTAPs had a median CXR severity score of 6.0 (IQR, 1.0–11.0). For each unit increase in CXR severity score, the odds of having a positive CTAP was 1.11 (1.04, 1.17; P = 0.001).

TABLE 2 - Association Between CXR Severity Category and Severity Score, and Negative/Positive Status of CTAP
CXR Severity CTAP Status OR (95% CI) P
Negative Positive
Total score
3.0 (0.0–5.9) 6.0 (1.0–11.0) 1.11 (1.04–1.17) 0.001
Total score categories
 Mild: [0–4] 62/108 (57.4) 46/108 (42.6) Reference
 Moderate [>4–10] 27/52 (51.9) 25/52 (48.1) 1.25 (0.64–2.43) 0.513
 Severe [>10–24] 9/35 (25.7) 26/35 (74.3) 3.89 (1.67, 9.10) 0.002
Categorical variables are summarized as frequencies (proportions) and continuous variables are summarized as median [IQR]. ORs summarize the association between severity score and CTAP status (negative, 0; positive, 1).

Table 3 summarizes the associations between CXR severity category and CTAP findings. Significant univariate associations included: distended fluid-filled bowel (OR, 3.79; 2.80–7.98; P < 0.001), gallbladder sludge (2.35, 1.12–4.92; 0.024), and major vessel thrombosis (11.4; 1.10–118; 0.041). When simultaneously adjusting for all CTAP findings, all estimates were attenuated (likely due to the relatively rare frequency of some of the CTAP findings) yet remained positively associated with higher CXR severity scores (eg, distended fluid-filled bowel [2.18 {0.92–5.20}], gallbladder sludge [2.00 {0.89–4.50}], and major vessel thrombosis [9.74 {0.84–113}]).

TABLE 3 - Association Between CTAP Findings and CXR Severity Category
CTAP Findings* CXR Severity Category (%) Univariate Summaries Multivariable Summaries
Mild n = 108 Moderate n = 52 Severe n = 35 OR (95% CI) P OR (95% CI) P
Pyelonephritis 3.7 7.7 8.6 2.05 (0.68–6.21) 0.203 1.86 (0.54–6.41) 0.327
Distended fluid-filled bowel 9.3 15.4 37.1 3.79 (2.80–7.98) <0.001 2.18 (0.92–5.20) 0.078
Ileus 6.5 3.8 8.6 1.03 (0.32–3.32) 0.958 0.78 (0.21–2.88) 0.714
Wall thickening 13.9 15.4 11.4 0.94 (0.43–2.04) 0.869 0.96 (0.42–2.21) 0.932
Perforation 0.0 1.9 2.9 6.56 (0.54–80.3) 0.141 5.52 (0.30–102.8) 0.252
Gallbladder sludge 12.0 13.5 31.4 2.35 (1.12–4.92) 0.024 2.00 (0.89–4.50) 0.095
Heterogeneous liver 2.8 5.8 2.9 1.34 (0.35–5.19) 0.669 1.14 (0.27–4.74) 0.858
Solid organ infarct 0.9 11.5 2.9 2.52 (0.80–7.91) 0.113 2.43 (0.71–8.32) 0.156
Peritonitis 3.7 3.8 2.9 0.89 (0.21–3.83) 0.874 0.46 (0.08–2.79) 0.399
Major vessel thrombosis 0.0 1.9 5.7 11.38 (1.10–117.6) 0.041 9.74 (0.84–112.6) 0.068
Hematoma 2.8 3.8 2.9 1.15 (0.25–5.24) 0.856 1.07 (0.21–5.43) 0.931
ORs summarize the odds of having a higher CXR severity categorization (mild < moderate < severe) in the presence of each CTAP finding (negative vs positive).15
*Pneumatosis and portal venous gas were not observed and only 1 subject had acute tubular necrosis and omitted from the above analysis.

Mortality Prediction

The cross-validated AUC estimates associated with 14-day mortality were 0.67 (95% CI, 0.56–0.78) and 0.71 (0.62–0.81) when using penalized logistic regression and random forests, respectively. Similar values for 30-day mortality were 0.76 (0.66–0.85) and 0.75 (0.66–0.84).

Independent predictors, sorted by importance (absolute value of the standardized coefficient), of mortality within 14 days included: age (0.62), hematoma (0.276), CXR severity score (0.275), bowel wall thickening (0.164), heterogeneous liver (0.147), bowel perforation (0.143), and distended fluid-filled bowel (0.112). Similar values for 30-day mortality were: age (0.456), CXR severity score (0.400), bowel wall thickening (0.142), and hematoma (0.016).

When CTAP findings were omitted from the feature set, the cross-validated AUC estimates associated with 14-day mortality were 0.62 (95% CI, 0.51–0.73) and 0.70 (95% CI, 0.60–0.80) when using penalized logistic regression and random forests, respectively. Similar values for 30-day mortality were 0.74 (95% CI, 0.65–0.83) and 0.74 (95% CI, 0.64–0.83).

DISCUSSION

To our knowledge, this is the first study exploring the association and discriminatory ability of abdominal imaging findings, chest radiographs, and patient demographics in the prognostication of patients with COVID-19. We found a significant incremental rate of positive CTAP across the low, moderate, and—most notably—severe CXR severity score categories (42.6%, 48.1%, and 74.3%, respectively). These findings confirm those of previous studies showing that abdominal imaging findings are associated with worse outcomes in COVID-19 patients5,6; however, previous studies did not evaluate the added value of CTAP to the validated chest radiograph severity score as a COVID-19 outcome predictor. Further, we found that multiple abdominal imaging findings may serve as independent predictors of COVID-19 mortality within 14 and 30 days, but CXR severity score and patient age remained important predictors of mortality.

Chest radiograph severity score is associated with worse outcomes in COVID-19.10–13 Our findings suggest that positive CTAP is also associated with worse CXR severity score, most notably with those with severe chest imaging findings. Based on our multivariable analysis, distended fluid-filled bowel, gallbladder sludge, and major vessel thrombosis were marginally associated with CXR severity. Given this association, it is not surprising that the same CTAP findings were not identified as predictors of mortality—especially after adjusting for CXR severity score. Schalekamp et al. found that a risk model using CXR and laboratory findings predicted critical illness (death and/or intensive care unit admission requiring mechanical ventilation) with an AUC of 0.77.11 Similarly, our imaging-focused risk model including CXR severity score, and patient age predicted mortality within 30 days with AUC of 0.74 (0.65–0.83) and 0.74 (0.64–0.83) when using penalized logistic regression and random forests, respectively. With the inclusion of CTAP findings, the AUCs associated with 30-day mortality were 0.76 (0.66–0.85) and 0.75 (0.66–0.84).

Our results suggest that certain abdominal imaging findings may serve as independent predictors of mortality in patients with COVID-19. Potential explanations rely on SARS-CoV-2 affinity for angiotensin-converting enzyme 2 receptors in gastrointestinal, hepatobiliary, and vascular cell membranes.8 Possible pathophysiology mechanisms for the abdominal findings in COVID-19 include the prothrombotic state with embolization and ischemia, abnormal systemic inflammatory response, cytokine storm, or angiopathy.16–19 An interconnected multifactorial pathophysiology is possible, where more severe chest manifestations are associated with worse system-wide disease, including abdominal findings.

Bowel wall thickening, pneumatosis, and portal venous gas have been previously associated with ICU admission.5 In our study, pneumatosis and portal venous gas were rare events, which limited the evaluation of these variables as mortality predictors. However, we similarly found that bowel wall thickening was an independent predictor of mortality within both 14 and 30 days, along with the presence of hematoma, worse CXR severity score, and more advanced patient age. Therefore, certain CTAP findings, especially those not directly associated with CXR severity, may be important predictors of COVID-19 when used in combination with CXR severity score and patient age.

This study has several limitations. First, this is a single-institution retrospective study, which limits generalizability. Second, the relatively small sample size considering the multiple possible abdominal imaging findings limits evaluation of rare abdominal findings as mortality predictors and limits the exploration of interactions between abdominal findings. Third, abdominal imaging findings in the patient population might not necessarily be related to COVID-19, such as preexisting conditions or stress related to hospitalization, and should be interpreted as an exploratory analysis. Fourth, the evaluation of some of the abdominal imaging findings was subjective. However, this reflects real-life assessment and discordant cases between the 2 readers were assessed by a third fellowship-trained abdominal radiologist with 10 years of subspecialty experience. Lastly, we did not evaluate laboratory findings as predictors of mortality in this imaging-focused study.

In conclusion, findings on CT abdomen and pelvis were associated with more severe CXR score and may serve as predictors of COVID-19 mortality.

REFERENCES

1. Worldometer. United States COVID: 34,136,738 Cases and 610,432 Deaths. https://www.worldometers.info/coronavirus/country/us/. Accessed June 1, 2021.
2. Wong HYF, Lam HYS, Fong AH, et al. Frequency and distribution of chest radiographic findings in patients positive for COVID-19. Radiology. 2020;296:E72–E78.
3. Bernheim A, Mei X, Huang M, et al. Chest CT findings in coronavirus disease-19 (COVID-19): relationship to duration of infection. Radiology. 2020;295:200463.
4. Song F, Shi N, Shan F, et al. Emerging 2019 novel coronavirus (2019-nCoV) pneumonia. Radiology. 2020;295:210–217.
5. Bhayana R, Som A, Li MD, et al. Abdominal imaging findings in COVID-19: preliminary observations. Radiology. 2020;297:E207–E215.
6. Horvat N, Pinto PVA, Araujo-Filho JAB, et al. Abdominal gastrointestinal imaging findings on computed tomography in patients with COVID-19 and correlation with clinical outcomes. Eur J Radiol Open. 2021;8:100326.
7. Tirumani SH, Rahnemai-Azar AA, Pierce JD, et al. Are asymptomatic gastrointestinal findings on imaging more common in COVID-19 infection? Study to determine frequency of abdominal findings of COVID-19 infection in patients with and without abdominal symptoms and in patients with chest-only CT scans. Abdom Radiol (NY). 2021;46:2407–2414.
8. Jothimani D, Venugopal R, Abedin MF, et al. COVID-19 and the liver. J Hepatol. 2020;73:1231–1240.
9. Colombi D, Bodini FC, Petrini M, et al. Well-aerated lung on admitting chest CT to predict adverse outcome in COVID-19 pneumonia. Radiology. 2020;296:E86–E96.
10. Balbi M, Caroli A, Corsi A, et al. Chest X-ray for predicting mortality and the need for ventilatory support in COVID-19 patients presenting to the emergency department. Eur Radiol. 2020;31:1999–2012.
11. Schalekamp S, Huisman M, van Dijk RA, et al. Model-based prediction of critical illness in hospitalized patients with COVID-19. Radiology. 2021;298:E46–E54.
12. Al-Smadi AS, Bhatnagar A, Ali R, et al. Correlation of chest radiography findings with the severity and progression of COVID-19 pneumonia. Clin Imaging. 2021;71:17–23.
13. Li MD, Arun NT, Gidwani M, et al. Automated Assessment and Tracking of COVID-19 Pulmonary Disease Severity on Chest Radiographs using Convolutional Siamese Neural Networks. Radiol Artif Intell. 2020;2:e200079.
14. Yuan M, Yin W, Tao Z, et al. Association of radiologic findings with mortality of patients infected with 2019 novel coronavirus in Wuhan, China. PLoS One. 2020;15:e0230548.
15. Joseph NP, Reid NJ, Som A, et al. Racial and ethnic disparities in disease severity on admission chest radiographs among patients admitted with confirmed coronavirus disease 2019: a retrospective cohort study. Radiology. 2020;297:E303–E312.
16. Mao R, Qiu Y, He J-S, et al. Manifestations and prognosis of gastrointestinal and liver involvement in patients with COVID-19: a systematic review and meta-analysis. Lancet Gastroenterol Hepatol. 2020;5:667–678.
17. Sultan S, Altayar O, Siddique SM, et al. AGA institute rapid review of the gastrointestinal and liver manifestations of COVID-19, meta-analysis of international data, and recommendations for the consultative management of patients with COVID-19. Gastroenterology. 2020;159:320–334.e27.
18. Chan NC, Weitz JI. COVID-19 coagulopathy, thrombosis, and bleeding. Blood. 2020;136:381–383.
19. McElvaney OJ, McEvoy NL, McElvaney OF, et al. Characterization of the inflammatory response to severe COVID-19 illness. Am J Respir Crit Care Med. 2020;202:812–821.
Keywords:

COVID-19; computed tomography; abdominal imaging; chest radiograph; severity score

Copyright © 2022 Wolters Kluwer Health, Inc. All rights reserved.