Lung ultrasound (LUS) has proven high diagnostic accuracy for community-acquired pneumonia (CAP) in developed countries. However, its diagnostic performance in resource-limited settings with high pulmonary tuberculosis (TB) incidence is less established. Additionally, the role of LUS in monitoring CAP progression remains underexplored.

To validate the diagnostic performance, monitoring and prognostic utility of LUS for CAP in a high pulmonary TB incidence setting.

Prospective single-centre cohort study.

Pulmonary department of a tertiary hospital in Vietnam.

A total of 158 patients suspected of having CAP were enrolled, with 136 (mean age 62 years, 72.8% male) included in the final analysis.

Patients underwent LUS and chest X-ray (CXR) within 24 hours of admission, with a follow-up LUS on days 5–8.

The primary outcome was the diagnostic accuracy of LUS and CXR compared with discharge diagnosis. Secondary outcomes included the accuracy compared with CT scan results, changes in LUS parameters—consolidation size, number and Lung Ultrasound Score (LUSS)—and their association with in-hospital mortality.

LUS demonstrated higher sensitivity than CXR (96.0% (95% CI 90.0% to 99.0%) vs 82.8% (95% CI 73.9% to 89.7%)). LUS specificity was 64.9% (95% CI 47.5% to 80.0%), compared with 54.1% (95% CI 36.9% to 70.5%) for CXR. The moderate specificity for LUS was due to sonographic-similar conditions, notably TB in 5.1% of patients. Consolidation size and numbers showed marginal resolution, while LUSS showed more pronounced decreases over time. The baseline LUSS showed limited discriminative ability for predicting mortality (area under the curve, AUC 0.65, 95% CI 0.55 to 0.75), while follow-up LUSS and changes in LUSS (ΔLUSS) demonstrated higher levels of discrimination (AUC 0.81 (95% CI 0.71 to 0.89) and 0.89 (95% CI 0.80 to 0.95), respectively). For each one-point increase in ΔLUSS, the odds of in-hospital mortality went up by 70% (p=0.002). An improved LUSS effectively ruled out mortality (negative predictive value 97.4%).

Although LUS is highly sensitive for diagnosing CAP, its specificity in TB-endemic regions warrants further caution. Serial LUS assessments, particularly monitoring LUSS changes, are valuable for tracking disease progression and prognostication, with increasing LUSS indicating potential clinical deterioration.

Data are available upon reasonable request.

http://creativecommons.org/licenses/by-nc/4.0/

This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See: http://creativecommons.org/licenses/by-nc/4.0/.

Community-acquired pneumonia (CAP) is the leading global infectious disease, presenting significant challenges to public health due to its high hospitalisation and mortality rates.1–3 Effective diagnosis and monitoring are crucial to improve patient outcomes and reduce the healthcare burden. Despite being frequently encountered in both outpatient and inpatient settings, pneumonia diagnosis remains complex. CAP is rarely confirmed through the gold standard of pathology. Instead, the diagnosis relies on concordant evidence of clinical symptoms, microbiological detection and compatible imaging findings, typically new infiltrates on chest radiographs (CXR).4 Despite being a staple for diagnosing CAP for years, CXR may fail to detect or correctly identify pneumonic lesions.5–7

In recent years, alternative diagnostic tools such as lung ultrasound (LUS) have emerged.8 Besides the advantages of being radiation-free, bedside-available and repeatable, studies have shown that LUS offers substantial diagnostic accuracy.9 Multiple meta-analyses revealed that the LUS sensitivity for diagnosing CAP ranges from 85% to 97%, with specificity between 80% and 96%.10–18 However, most of the evidence on LUS diagnostic accuracy for pneumonia was derived from developed countries. There is less emphasis on low-resource settings, where diseases such as tuberculosis (TB) and bronchiectasis can mimic pneumonia sonographically, potentially affecting diagnostic properties.17 Furthermore, the potential of LUS in monitoring and stratifying CAP patients at risk of clinical deterioration is not well understood. In this study, we aim to investigate the diagnostic performance of LUS in a developing country. Additionally, we seek to identify which LUS parameters can effectively monitor and prognosticate CAP.

This study confirmed that LUS has a higher sensitivity than CXR for diagnosing CAP. However, its moderate specificity may be influenced by the difficulty in differentiating pneumonia from other respiratory conditions, particularly TB. To the best of our knowledge, this is one of the first studies to incorporate LUSS for monitoring CAP. Our findings indicate that LUSS changes over time may offer preliminary prognostic insights, potentially aiding in the identification of disease progression and mortality risk stratification.

Previous studies have shown that LUS has a high sensitivity for detecting pneumonia.14–17 Our study aligns with these findings, demonstrating greater sensitivity than CXR. These results reaffirm LUS as a reliable tool for ruling out pneumonia. However, if ultrasound is negative but other pneumonia signs persist, further investigation and close monitoring after antibiotic treatment are necessary for a definitive diagnosis. Despite showing great sensitivity, LUS specificity was lower than in prior reports14–17 and varied depending on the reference standard used. The lower specificity observed with CT as the standard, compared with a clinical panel combining clinical features and CXR, reflects CT’s superior ability to detect detailed pulmonary changes. Studies have also shown that clinical features and CXR frequently lead to misdiagnosis of CAP compared with CT.5 While LUS is more sensitive than CXR in detecting interstitial abnormalities and consolidations, it shares similar limitations, such as difficulty in distinguishing acute from chronic changes and less detailed lung pattern analysis compared with CT. For example, B-lines on LUS may indicate acute infections or chronic fibrotic processes, and hypoechoic lesions may also signify various pathologies, including pneumonia, atelectasis, lung cancer, pulmonary embolism, or nodular scarring.

Our findings indicate that LUS has difficulty in differentiating pneumonia from other respiratory diseases, with TB being the most frequently misdiagnosed. In this study, we classified pulmonary TB as false positive rather than a type of CAP. This decision was based on the rationale that the diagnosis determines subsequent antibiotic strategies, which differ between the two conditions. While some sonographic findings (eg, subpleural nodules, pleural effusion and consolidation with fluid collections) may suggest TB, the modality is inherently limited in detecting cavity lesions, which are a consistent radiological feature in our TB patients, due to air within cavities preventing ultrasound penetration. This is particularly relevant in our setting, which reported the highest number of TB cases among LUS studies on CAP. In contrast, previous research, primarily conducted in low-TB-prevalence settings, found no TB cases, while studies in endemic areas such as Liu26 in China and Amatya27 in Nepal reported zero and one case, respectively. Our study’s TB prevalence of 6.7% (7/105) notably surpasses the global average of 0.86% reported by a multicentre CAP study, which included non-endemic regions such as Europe (0.97%) and North America (1.02%),28 while aligning more closely with figures from other high-burden settings, including Hong Kong29 (8.1%) and the Philippines30 (9.8%). These findings highlight the diagnostic challenges of LUS for pneumonia in TB-endemic regions, where sonographic presentations of TB and pneumonia often overlap. Clinically, when consolidations (with or without complementary features such as pleural effusion or subpleural nodules) appear alongside a clinical suspicion of TB, further evaluation with CT scans and TB-specific workup is indispensable. Larger, targeted studies are needed to better characterise ultrasound findings in TB.

Besides evaluating the diagnostic properties, our study aimed to observe sonographic changes in CAP over time and assess whether these changes could aid in monitoring and predicting clinical outcomes. We focused on three ultrasound parameters: consolidation size, number of consolidations and LUSS. Previous studies in both paediatric28–30 and adult populations21 suggest that disease remission can be observed through the resolution of lesion sizes and numbers. However, our findings indicate that changes in the size and overall number of consolidations during follow-up assessments were relatively small. These marginal changes may not be readily apparent to clinicians, making it less ideal to utilise these parameters for monitoring purposes. The difference in pneumonic lesion resolution between our study and that reported in the adult population by Reissig21 may stem from variations in measurement methods and sample selection. We used a one-dimensional measure for the largest consolidation, whereas Reissig et al employed a two-dimensional measure in square centimetres. Additionally, for comparisons of lesion size at two time points, our initial assessment only included subjects available for a follow-up ultrasound, in contrast to Reissig’s approach, which involved measuring pneumonic size in all patients, regardless of follow-up availability.21

The LUSS has recently emerged as a useful tool for assessing severity, and the baseline score is closely related to adverse outcomes in COVID-19 patients.31 Our analysis showed that the baseline score has limited predictive value for in-hospital mortality. Instead, the dynamic changes in the LUSS during follow-up may offer a more reliable indication of mortality. Hypothetically, since the LUSS incorporates both consolidation and interstitial components, and considering that changes in consolidation measurements were small, it is possible that changes in the interstitial pattern occur earlier and are more predictive of the clinical course of CAP than consolidative changes. From the clinical practice perspective, LUSS progression should alert physicians about a deteriorating clinical course. A ΔLUSS cut-off of 0 is clinically applicable as it allows for the simple categorisation of patients into groups with improved or unimproved LUSS over time. Patients with no improvement in monitoring LUS (ΔLUSS≥0) demonstrated a sensitivity of 88.9%, specificity of 57.6%, NPV of 97.4% and PPV of 22.2% in predicting mortality. The high sensitivity and NPV suggest that a ΔLUSS≥0 is effective in identifying patients at risk of mortality, the moderate specificity and PPV indicate that ΔLUSS should be used in conjunction with other clinical indicators. Utilising LUSS for stratification may lead to a more efficient allocation of medical resources, ensuring that attention and care are prioritised for patients with a higher risk of mortality. Timely interventions, such as escalating antibiotics, advanced imaging and microbiology workup may prevent progression to critical illness and ultimately reduce mortality. However, as our observations are based on a limited sample size, further studies on sonographic pneumonic lesion evolution and their impact on clinical outcomes are needed to validate these findings.

This study has several limitations. First, due to ethical reasons, a CT scan was not performed on all patients, leaving the possibility of missing or misidentifying pneumonic lesions. However, in those who did receive a CT scan, the performance of LUS was found to be comparable to both discharge diagnosis and CT imaging, indicating the former’s reliability. Second, the consolidation size was recorded in a single dimension, which does not fully capture the lesion’s three-dimensional volume. However, measurements in one dimension have been shown to effectively represent overall lesion volume.29 Third, as the study focused on inpatients, its findings may not extend to outpatients, who often have smaller, more rapidly resolving lesions. Additionally, some patients were discharged before the second ultrasound, potentially skewing follow-up data away from those with milder disease. However, similar to CXR, follow-up ultrasounds may be unnecessary for patients showing early recovery, as their clinical symptoms suggest resolution without additional imaging. For patients who died early before the second ultrasound, it is plausible they had more progressive lesions, potentially amplifying our findings. Fourth, while some clinical data (eg, mechanical ventilation and RICU admission) were recorded at the time of the second LUS, other dynamic parameters such as trends in vital signs, oxygen therapy escalation, lactate levels or renal function were not captured. Incorporating these variables could provide a more comprehensive prognostic assessment, as the absence of such data renders the prognostic value of LUSS less certain. For example, in patients with clear signs of deterioration, conducting an intensive LUS protocol may offer limited benefit, whereas follow-up LUS could be more valuable for those with uncertain trajectories. Future research should explore integrating LUSS with dynamic clinical data to improve risk stratification. Finally, ultrasound is an operator-dependent tool, and its interpretation is subjective to sonographer’s experience. Nevertheless, our study demonstrated high reliability between performers.

Data are available upon reasonable request.

The authors thank Thong Dang-Vu, Dung Lam-Quoc and the staff at the Pulmonary Department, Cho Ray Hospital, for their assistance in data collection and all the patients for participating in this study.