Can the baseline chest X-ray brixia score and other factors predict outcomes of individuals admitted with COVID-19 pneumonia in resource-limited settings? A retrospective cohort study at a tertiary hospital in Northwestern Tanzania

¹Department of Internal Medicine, Bugando Medical Centre, Mwanza, Tanzania, ²Department of Radiology, Bugando Medical Centre, Mwanza, Tanzania, ³Department of Research, Bugando Medical Centre, Mwanza, Tanzania, ⁴Department of Emergency Medicine, Bugando Medical Centre, Mwanza, Tanzania, ⁵Department of Microbiology and Immunology, Weill Bugando School of Medicine, Catholic University of Health and Allied Sciences, Mwanza, Tanzania, ⁶Department of Radiology, Weill Bugando School of Medicine, Catholic University of Health and Allied Sciences, Mwanza, Tanzania

^&Corresponding author
Patrick Ngoya, Department of Radiology, Weill Bugando School of Medicine, Catholic University of Health and Allied Sciences, Mwanza, Tanzania

Introduction    

COVID-19 pneumonia presents in various degrees of severity and respiratory impairment, ranging from mild cases with flu-like symptoms to severe forms with acute respiratory distress syndrome (ARDS), requiring intensive care support [1]. Despite potential multi-systemic involvement, the lung is by far the organ most commonly affected. Imaging of the chest is important in evaluating the extent, complications, and follow-up of disease. The chest x-ray (CXR) has relatively high portability, speed of acquisition, accessibility, and low cost, especially in low-resource settings and where serial imaging is required for follow-up [2,3]. Computed tomography is indicated in specific clinical scenarios where it is expected to significantly contribute to clinical management [4]. To improve risk stratification, CXR scoring systems for quantifying the severity and progression of lung abnormalities in COVID-19 pneumonia were introduced [5]. The Brixia score has the ability to provide relevant information for clinicians as well as identify high-risk individuals and those who require specific treatment strategies [6,7]. The Brixia score has been shown to correlate positively with biomarkers of inflammation and organ injury known to be associated with COVID-19 pneumonia [8]. The COVID-19 pneumonia burden was alarmingly high, particularly among individuals with severe and critical disease who had been documented to have up to a 34% mortality rate in Tanzania [9,10]. Therefore, this study was designed to determine whether the baseline CXR Brixia score and other factors can predict outcomes of individuals admitted with COVID-19 pneumonia in resource-limited settings.

Methods     

Study design, duration, and setting: this was a retrospective cohort hospital-based study conducted from 1 June 2023 to 31 May 2024 at Bugando Medical Centre (BMC), Mwanza, Tanzania.

Study population: all individuals diagnosed with COVID-19 pneumonia who were admitted between 1 September 2021 and 31 May 2022 during the third and fourth waves of the COVID-19 pandemic.

Sample size estimation: the sample size was estimated using the Taro Yamane formula [11]:

Where n was the sample size estimate, N was the population size of diagnosed COVID-19 individuals during the study period (N=298), and e was the level of precision or sampling error (5%). Thus, the minimum sample size was estimated at 170.

Sampling method: a consecutive sampling technique was applied until an adequate sample size was reached.

Selection criteria: adults aged 18 years and older diagnosed with COVID-19 pneumonia were included. Individuals with incomplete data were excluded.

Study variables: exposure variables included age, sex, presence of comorbidities, respiratory rate, pulse rate, systolic and diastolic blood pressure, temperature, oxygen saturation, and the total Brixia score. Outcome variables included disease severity, length of hospital stay, and in-hospital mortality.

Data collection and procedures: data were retrieved from the BMC emergency preparedness and response team database, which included socio-demographic and clinical data and baseline CXR. Nasopharyngeal swabs were collected at the time of admission and sent to the BMC laboratory for processing using viral transport media. The Abbott m2000 RT-PCR assay (Abbott Laboratories, USA) was used to detect RNA in respiratory specimens collected from individuals suspected of COVID-19. Respiratory rate was measured in cycles per minute (cpm) during physical examination. Pulse rate in beats per minute (bpm) and systolic and diastolic blood pressure in millimeters of mercury (mmHg) were acquired using a digital blood pressure monitor. Temperature was acquired using an axillary thermometer and recorded in degrees celsius (°C). Oxygen saturation in ambient room air was acquired using a pulse oximeter and recorded as a percentage (%). Baseline CXRs were acquired using a GE XR6000 digital radiography system (GE Healthcare, USA) with an exposure range of 70-120 kilovoltage peak (kVp) and 1-5 milliampere-seconds (mAs), following the as low as reasonably achievable (ALARA) principle. The CXRs were reviewed by an experienced radiologist blinded to the clinical outcomes and assigned a total Brixia score. The Brixia score involved dividing the bilateral lung fields into six zones (three on the right and three on the left) and assigning a score ranging from 0 to 3 to each zone: i) 0 indicated no abnormalities; ii) 1 indicated interstitial opacities only; iii) 2 indicated both interstitial and alveolar opacities, with interstitial opacities predominating; iv) 3 indicated both, with alveolar opacities predominating.

A total Brixia score was derived from the sum of the Brixia scores in the six zones of the bilateral lung fields. Mild lung involvement if the total Brixia score is 0 to 6, moderate lung involvement if the total Brixia score is 7 to 12, and severe lung involvement if the total Brixia score is 13 to 18 [6,7]. Disease severity was categorized clinically as follows [12]: i) mild disease: individuals with various symptoms (e.g. fever, cough, fatigue) but without shortness of breath, dyspnea, or abnormal chest imaging; ii) moderate disease: Individuals with evidence of lower respiratory disease during clinical assessment or imaging and an oxygen saturation of 94% on room air; iii) severe disease: Individuals with an oxygen saturation of < 94% on room air, a respiratory rate > 30 breaths/min, or lung infiltrates > 50%; iv) critical disease: Individuals with respiratory failure, septic shock, and/or multiple organ dysfunction. Length of hospital stay was categorized as < 2 weeks or ≥2 weeks. In-hospital mortality was determined on the day of discharge and categorized as alive or deceased.

Data analysis: data were entered in Microsoft Excel (Microsoft, USA), then exported to STATA version 17 (Statacorp LLC, USA) for cleaning and analysis. Continuous variables were summarized as means with standard deviation (SD) or median with interquartile range (IQR) depending on their distribution. Continuous variables were categorized. Categorical variables were summarized as frequencies and percentages. Predictors of outcomes were determined using Firth logistic regression, with all exposure variables included as potential confounders in the multivariate analysis. A p-value of less than 0.05 was considered significant.

Ethical consideration: ethical clearance was obtained from the CUHAS/BMC Research and Ethical Review Committee (CREC/680/2023). Informed consent was waived due to the retrospective nature of the study. Data confidentiality was strictly maintained throughout the study through anonymization and restricted access to patient records.

Results     

Characteristics of adults diagnosed with COVID-19 pneumonia: a total of 298 individuals were diagnosed with COVID-19 pneumonia. However, 78 (26%) individuals were excluded due to incomplete data (Figure 1). Enrolled participants were aged between 22 and 98 years, with a mean age (SD) of 59 (13) years and negligible sex predominance. Comorbidities were present in 52% of the participants and included hypertension (40%), diabetes mellitus (19%), chronic kidney disease (9%), cardiac disease (4%), and human immunodeficiency virus (HIV) infection (4%). Oxygen saturation in ambient room air ranged from 34% to 97%, with the majority (76%) having less than 94% oxygen saturation (Table 1). Based on the Brixia score, mild lung involvement was seen in 12 individuals (6%), moderate in 57 individuals (26%), and severe in 151 individuals (69%) (Table 1), as illustrated on CXR in Figure 2.

Outcomes of adults diagnosed with COVID-19 pneumonia: most individuals (70%) had a severe or critical disease, with 32 (14%) requiring intensive care unit (ICU) admission. The median (IQR) of length of hospital stay was 9 (6 -13) days, with a range of 2 to 71 days. In-hospital mortality was recorded in 37.7% of participants (Table 1).

Predictors of outcomes in adults diagnosed with COVID-19 pneumonia: oxygen saturation (OR = 14.9, 95% CI: 3.2-69.2), moderate Brixia score (OR = 15.9, 95% CI: 4.2-59.5), and severe Brixia score (OR = 85.2, 95% CI: 4.2-1742.9) were significant independent predictors of disease severity (Table 2). The presence of comorbidities (OR = 2.4, 95%CI: 1.2-4.8), oxygen saturation (OR = 12.6, 95%CI: 2.4-66.2), the moderate Brixia score (OR = 6.4, 95% CI: 2.3-18.2), and severe Brixia score (OR = 5.3, 95% CI: 1.7-17.0) were dependent predictors of length of hospital stay (Table 3). Age (OR = 3.3, 95%CI: 1.5-7.0), moderate Brixia score (OR = 19.5, 95% CI: 1.1-350.7), and severe Brixia score (OR = 425.6, 95% CI: 19.3-9381.4) were significant independent predictors of in-hospital mortality (Table 4).

Discussion     

This study retrospectively reviewed data and imaging of individuals diagnosed with COVID-19 pneumonia during the third and fourth waves of the COVID-19 pandemic. Most participants with severe lung involvement (69%) on baseline CXR had a Brixia score clinically, which positively correlated with those that had severe or critical disease (70%). Oxygen saturation and the Brixia score were significant independent predictors of disease severity, each having at least fifteenfold increased odds of disease severity in cases of low oxygen saturation or a high Brixia score. These findings are similar to previous studies, which showed that a low oxygen saturation and higher Brixia score at the time of admission correlated with disease severity [13,14]. Prolonged hospital stay was significantly associated with comorbidities, low oxygen saturation, and high Brixia score. The baseline CXR Brixia score has also been reported to predict the length of hospital stay in a study done in Sri Lanka [15]. This is contrary to findings from Indonesia [16] which showed that the CXR Brixia score cannot predict the length of stay of hospitalized, COVID-19 confirmed patients. The contradictory findings may be related to differences in methodology, as the Indonesian study was a case-control study.

We report an in-hospital mortality rate of more than one-third (38%) among the individuals admitted with COVID-19 pneumonia. This is slightly higher in comparison to previous studies from Tanzania [9,10]. Death was significantly more common in individuals with advanced age, comorbidities, low oxygen saturation, and a high Brixia score. These findings are similar to a retrospective study conducted in Italy, which demonstrated that patients with a high Brixia score and at least one other predictive factor had the highest risk of in-hospital death [17]. Another retrospective study in the Philippines also suggested that Brixia scores obtained from baseline CXR have a significant association with in-hospital mortality [18]. More than half of the participants (52%) had comorbidities, with hypertension followed by diabetes mellitus being the most predominant. Comorbidities have been widely reported to be associated with adverse clinical outcomes [19].

The exclusion of 26% (78 out of 298) of individuals due to missing data represents a substantial attrition rate. This may have underpowered the study and introduced selection bias, as evidenced by small cell sizes, extremely high odds ratios, and wide confidence intervals in some cross-tabulations. While the predictive trends remain robust and statistically significant, the precise magnitude of these odds ratios is likely mathematically unstable due to these small cell counts. Future research should consider multiple imputation techniques rather than listwise deletion to handle missingness and maximize statistical power as described in a recent study [20]. The primary reasons for missing data were the unprecedented strain on health resources, where clinical care took precedence over research database maintenance, and the subsequent loss of patients to follow-up during the COVID-19 pandemic.

Study limitations: as a single-center study, the findings may not be generalized to other populations. Furthermore, the exclusion of approximately one-quarter of the potential participants due to missing data may have introduced selection bias and reduced the study's statistical power. While baseline CXR may have been subject to observer variability, this was counteracted by having an experienced thoracic radiologist review all CXR. We recommend larger, multi-center studies with robust methodology in the future to address these limitations.

Conclusion     

The baseline CXR Brixia score can independently predict outcomes such as disease severity and in-hospital mortality of individuals admitted with COVID-19 pneumonia in resource-limited settings. Age, comorbidities, and oxygen saturation were also key predictors of outcomes.

Competing interests     

All authors declare no competing interests.

Authors' contributions     

Conceived, designed and executed the study: Dominic Mbwilo, Ziad Byekwaso, Ernest Elisenguo, Ally Munir Akrabi, Jeremiah Seni, Bahati Wajanga and Patrick Ngoya). Supervised the study, and coordinating the COVID-19 database at Bugando Ziad Byekwaso, Ernest Elisenguo, Ally Munir Akrabi, Jeremiah Seni, Bahati Wajanga and Patrick Ngoya. Data and sample collections: Dominic Mbwilo, Ziad Byekwaso, Ernest Elisenguo, Fredy Hyera and Evarist Msaki. Patients’ management: Dominic Mbwilo, Ziad Byekwaso, Ernest Elisenguo, Ally Munir Akrabi, Jeremiah Seni, Bahati Wajanga and Patrick Ngoya. Data analysis, literature search and interpretation of data: Dominic Mbwilo, Ziad Byekwaso, Ernest Elisenguo, Fredy Hyera, Evarist Msaki, Ally Munir Akrabi, Jeremiah Seni, Bahati Wajanga and Patrick Ngoya. Preparation of the first draft of the manuscript: Dominic Mbwilo, and Patrick Ngoya. Critical review of the manuscript: Dominic Mbwilo, Ziad Byekwaso, Ernest Elisenguo, Fredy Hyera, Evarist Msaki, Ally Munir Akrabi, Jeremiah Seni, Bahati Wajanga and Patrick Ngoya. All authors have read and approved the final version of this manuscript.

Acknowledgments     

The authors would like to thank the BMC Emergency Preparedness and Response Team for providing data and support to this study.

Tables and figures     

Table 1: characteristics of adults diagnosed with COVID-19 pneumonia in Northwestern Tanzania between 1 September 2021 and 31 May 2022 by exposure (n=220)

Table 2: predictors of disease severity in adults diagnosed with COVID-19 pneumonia in Northwestern Tanzania between 1 September 2021 and 31 May 2022 (n=220)

Table 3: predictors of length of hospital stay in adults diagnosed with COVID-19 pneumonia in Northwestern Tanzania between 1 September 2021 and 31 May 2022 (n=220)

Table 4: predictors of mortality in adults diagnosed with COVID-19 pneumonia in Northwestern Tanzania between 1 September 2021 and 31 May 2022 (n=220)

Figure 1: recruitment flowchart of adults diagnosed with COVID-19 pneumonia in Northwestern Tanzania between 1 September 2021 and 31 May 2022

Figure 2: chest X-ray showing (a) mild, (b) moderate and, (c) severe lung involvement in adults diagnosed with COVID-19 pneumonia in Northwestern Tanzania based on the total Brixia score from the bilateral lung fields’ six zones

References