Screening strategies for tuberculosis prevalence surveys: the value of chest radiography and symptoms

Cost-Effectiveness of Community-Based Active Case Finding Strategy for Tuberculosis: Evidence From Shenzhen, China

BACKGROUND

Active case finding (ACF) is a potentially promising approach for the early identification and treatment of tuberculosis patients. However, evidence on its cost-effectiveness, particularly in low- and middle-income countries, remains limited. This study evaluates the cost-effectiveness of a community-based ACF practice in Shenzhen, China.

METHODS

We employed a Markov model-based decision analytic method to assess the costs and effectiveness of 3 tuberculosis detection strategies: passive case finding (PCF), basic ACF, and advanced ACF. The analysis was conducted from a societal perspective on a dynamic cohort over a 20-year horizon, focusing on active tuberculosis (ATB) prevalence and the incremental cost-effectiveness ratio (ICER).

RESULTS

Compared to the PCF strategy, the basic and advanced ACF strategies effectively reduced ATB cases by 6.8 and 10.2 per 100 000 population, respectively, by the final year of this 20-year period. The ICER for the basic and advanced ACF strategies were ¥14 757 and ¥8217 per quality-adjusted life-year, respectively. Both values fell below the cost-effectiveness threshold.

CONCLUSIONS

Our findings indicate that the community-based ACF screening strategy, which targets individuals exhibiting tuberculosis symptoms, is cost-effective. This underscores the potential benefits of adopting similar community-based ACF strategies for symptomatic populations in tuberculosis-endemic areas.

Clinical and imaging factors that can predict contagiousness of pulmonary tuberculosis

BACKGROUND

Knowledge on predicting pulmonary tuberculosis (PTB) contagiosity in the hospital admission setting is limited. The objective was to assess clinical and radiological criteria to predict PTB contagiosity.

METHODS

Retrospective analysis of 7 clinical, 4 chest X-ray (CXR) and 5 computed tomography (CT) signs in 299 PTB patients admitted to an urban tertiary hospital from 2008 to 2016. If the acid fact bacilli stain was positive (AFB+) on admission, the case was considered high contagiosity.

RESULTS

Best predictors for high PTB contagiosity (AFB+) were haemoptysis (OR 4.33), cough (3.00), weight loss (2.96), cavitation in CT (2.75), cavitation in CXR (2.55), tree-in-bud-sign in CT (2.12), German residency of the patient (1.89), and abnormal auscultation findings (1.83). A previous TB infection reduced the risk of contagiosity statistically (0.40). Radiographic infiltrates, miliary picture, and pleural effusion were not helpful in predicting high or low contagiosity. 34% of all patients were clinically asymptomatic (20% of the highly contagious group, 50% of the low contagious group).

CONCLUSION

Haemoptysis, cough and weight loss as well as cavitation and tree-in-bud sign in CXR/CT can be helpful to predict PTB contagiosity and to improve PTB management.

Tuberculosis Diagnosis Using Deep Transferred EfficientNet

Tuberculosis is a very deadly disease, with more than half of all tuberculosis cases dead in countries and regions with relatively poor health care resources. Fortunately, the disease is curable, and early diagnosis and medication can go a long way toward curing TB patients. Unfortunately, traditional methods of TB diagnosis rely on specialist doctors, which is lacking in areas with high TB mortality rates. Diagnostic methods based on artificial intelligence technology are one of the solutions to this problem. We propose a Deep Transferred EfficientNet with SVM (DTE-SVM), which replaces the pre-trained EfficientNet classification layer with an SVM classifier and achieves auspicious performance on a small dataset. After ten runs of 10-fold Cross-Validation, the DTE-SVM has a sensitivity of 93.89±1.96, a specificity of 95.35±1.31, a precision of 95.30±1.24, an accuracy of 94.62±1.00, and an F1-score of 94.62±1.00. In addition, our study conducted ablation studies on the effect of the SVM classifier on model performance and briefly discussed the results.

Weak Localization of Radiographic Manifestations in Pulmonary Tuberculosis from Chest X-ray: A Systematic Review

Pulmonary tuberculosis (PTB) is a bacterial infection that affects the lung. PTB remains one of the infectious diseases with the highest global mortalities. Chest radiography is a technique that is often employed in the diagnosis of PTB. Radiologists identify the severity and stage of PTB by inspecting radiographic features in the patient's chest X-ray (CXR). The most common radiographic features seen on CXRs include cavitation, consolidation, masses, pleural effusion, calcification, and nodules. Identifying these CXR features will help physicians in diagnosing a patient. However, identifying these radiographic features for intricate disorders is challenging, and the accuracy depends on the radiologist's experience and level of expertise. So, researchers have proposed deep learning (DL) techniques to detect and mark areas of tuberculosis infection in CXRs. DL models have been proposed in the literature because of their inherent capacity to detect diseases and segment the manifestation regions from medical images. However, fully supervised semantic segmentation requires several pixel-by-pixel labeled images. The annotation of such a large amount of data by trained physicians has some challenges. First, the annotation requires a significant amount of time. Second, the cost of hiring trained physicians is expensive. In addition, the subjectivity of medical data poses a difficulty in having standardized annotation. As a result, there is increasing interest in weak localization techniques. Therefore, in this review, we identify methods employed in the weakly supervised segmentation and localization of radiographic manifestations of pulmonary tuberculosis from chest X-rays. First, we identify the most commonly used public chest X-ray datasets for tuberculosis identification. Following that, we discuss the approaches for weakly localizing tuberculosis radiographic manifestations in chest X-rays. The weakly supervised localization of PTB can highlight the region of the chest X-ray image that contributed the most to the DL model's classification output and help pinpoint the diseased area. Finally, we discuss the limitations and challenges of weakly supervised techniques in localizing TB manifestations regions in chest X-ray images.

Sensitivity of Various Case Detection Algorithms for Community-based Tuberculosis Screening

Using data from 388 people diagnosed with tuberculosis through a community-based screening program in Lima, Peru, we estimated that cough screening followed by sputum smear microscopy would have detected only 23% of cases found using an algorithm of radiographic screening followed by rapid nucleic acid amplification testing and clinical evaluation.

Diagnostic Accuracy of the Artificial Intelligence Methods in Medical Imaging for Pulmonary Tuberculosis: A Systematic Review and Meta-Analysis

Tuberculosis (TB) remains one of the leading causes of death among infectious diseases worldwide. Early screening and diagnosis of pulmonary tuberculosis (PTB) is crucial in TB control, and tend to benefit from artificial intelligence. Here, we aimed to evaluate the diagnostic efficacy of a variety of artificial intelligence methods in medical imaging for PTB. We searched MEDLINE and Embase with the OVID platform to identify trials published update to November 2022 that evaluated the effectiveness of artificial-intelligence-based software in medical imaging of patients with PTB. After data extraction, the quality of studies was assessed using quality assessment of diagnostic accuracy studies 2 (QUADAS-2). Pooled sensitivity and specificity were estimated using a bivariate random-effects model. In total, 3987 references were initially identified and 61 studies were finally included, covering a wide range of 124,959 individuals. The pooled sensitivity and the specificity were 91% (95% confidence interval (CI), 89-93%) and 65% (54-75%), respectively, in clinical trials, and 94% (89-96%) and 95% (91-97%), respectively, in model-development studies. These findings have demonstrated that artificial-intelligence-based software could serve as an accurate tool to diagnose PTB in medical imaging. However, standardized reporting guidance regarding AI-specific trials and multicenter clinical trials is urgently needed to truly transform this cutting-edge technology into clinical practice.

Economic analysis of different throughput scenarios and implementation strategies of computer-aided detection software as a screening and triage test for pulmonary TB

BACKGROUND

Artificial Intelligence (AI) systems have demonstrated potential in detecting tuberculosis (TB) associated abnormalities from chest X-ray (CXR) images. Thus, they might provide a solution to radiologist shortages in high TB burden countries. However, the cost of implementing computer-aided detection (CAD) software has thus far been understudied. In this study, we performed a costing analysis of CAD software when used as a screening or triage test for pulmonary TB, estimated the incremental cost compared to a radiologist reading of different throughput scenarios, and predicted the cost for the national scale-up plan in Pakistan.

METHODS

For the study, we focused on CAD software reviewed by the World Health Organization (CAD4TB, Lunit INSIGHT CXR, qXR) or listed in the Global Drug Facility diagnostics catalogue (CAD4TB, InferRead). Costing information was obtained from the CAD software developers. CAD4TB and InferRead use a perpetual license pricing model, while Lunit and qXR are priced per license for restricted number of scans. A major implementer in Pakistan provided costing information for human resource and software training. The per-screen cost was estimated for each CAD software and for radiologist for 1) active case finding, and 2) facility based CXR testing scenarios with throughputs ranging from 50,000-100,000 scans. Moreover, we estimated the scale-up cost for CAD or radiologist CXR reading in Pakistan based on the National Strategic Plan, considering that to reach 80% diagnostic coverage, 50% of TB patients would need to be found through facility-based triage and 30% through active case finding (ACF).

RESULTS

The per-screen cost for CAD4TB (0.25 USD- 2.33 USD) and InferRead (0.19 USD- 2.78 USD) was lower than that of a radiologist (0.70 USD- 0.93 USD) for high throughput scenarios studied. In comparison, the per-screen cost for Lunit (0.94 USD- 1.69 USD) and qXR (0.95 USD-1.9 USD) were only comparable with that of the radiologists in the highest throughput scenario in ACF. To achieve 80 percent diagnostic coverage at scale in Pakistan, the projected additional cost of deploying CAD software to complement the current infrastructure over a four-year period were estimated at 2.65-19.23 million USD, whereas Human readers, would cost an additional 23.97 million USD.

CONCLUSIONS

Our findings suggest that using CAD software could enable large-scale screening programs in high TB-burden countries and be less costly than radiologist. To achieve minimum cost, the target number of screens in a specific screening strategy should be carefully considered when selecting CAD software, along with the offered pricing structure and other aspects such as performance and operational features. Integrating CAD software in implementation strategies for case finding could be an economical way to attain the intended programmatic goals.

Explanatory classification of CXR images into COVID-19, Pneumonia and Tuberculosis using deep learning and XAI

Chest X-ray (CXR) images are considered useful to monitor and investigate a variety of pulmonary disorders such as COVID-19, Pneumonia, and Tuberculosis (TB). With recent technological advancements, such diseases may now be recognized more precisely using computer-assisted diagnostics. Without compromising the classification accuracy and better feature extraction, deep learning (DL) model to predict four different categories is proposed in this study. The proposed model is validated with publicly available datasets of 7132 chest x-ray (CXR) images. Furthermore, results are interpreted and explained using Gradient-weighted Class Activation Mapping (Grad-CAM), Local Interpretable Modelagnostic Explanation (LIME), and SHapley Additive exPlanation (SHAP) for better understandably. Initially, convolution features are extracted to collect high-level object-based information. Next, shapely values from SHAP, predictability results from LIME, and heatmap from Grad-CAM are used to explore the black-box approach of the DL model, achieving average test accuracy of 94.31 ± 1.01% and validation accuracy of 94.54 ± 1.33 for 10-fold cross validation. Finally, in order to validate the model and qualify medical risk, medical sensations of classification are taken to consolidate the explanations generated from the eXplainable Artificial Intelligence (XAI) framework. The results suggest that XAI and DL models give clinicians/medical professionals persuasive and coherent conclusions related to the detection and categorization of COVID-19, Pneumonia, and TB.

The Contribution of Chest Radiography to the Clinical Management of Children Exposed to Tuberculosis

Rationale: Although World Health Organization guidelines emphasize contact investigation for tuberculosis (TB)-exposed children, data that support chest radiography as a useful tool are lacking. Objectives: We evaluated the diagnostic and prognostic information of chest radiography in children exposed to TB and measured the efficacy of isoniazid preventive therapy (IPT) in those with relevant radiographic abnormalities. Methods: Between September 2009 and August 2012, we enrolled 4,468 TB-exposed children who were screened by tuberculin skin testing, symptom assessment, and chest radiography. Those negative for TB disease were followed for 1 year for the occurrence of new TB diagnoses. We assessed the protective efficacy of IPT in children with and without abnormal chest radiographs. Measurements and Main Results: Compared with asymptomatic children with normal chest films, asymptomatic children with abnormal radiographs were 25.1-fold more likely to have coprevalent TB (95% confidence interval [CI], 1.02-613.76) and 26.7-fold more likely to be diagnosed with incident TB disease during follow-up (95% CI, 10.44-68.30). Among the 29 symptom-negative and CXR-abnormal child contacts, 20% (3/15) of the isoniazid recipients developed incident TB, compared with 57% (8/14) of those who did not receive IPT (82% IPT efficacy). Conclusions: Our results strongly support the use of chest radiography as a routine screening tool for the evaluation of child TB contacts, which is readily available. Radiographic abnormalities not usually considered suggestive of TB may indicate incipient or subclinical disease, although TB preventive treatment is adequate in most cases.

Role of chest radiography in the diagnosis of pulmonary tuberculosis during nCovid19 pandemic

Pulmonary tuberculosis and nCovid 19 share many common risk factors. nCovid19 may increase the risk to develop pulmonary tuberculosis. Pulmonary tuberculosis may precede, co-exist or follow nCovid19. Careful evaluation of chest radiography is useful to differentiate tuberculosis from nCovid19 bronchopneumonia. Symptoms of tuberculosis may be mistaken for long covid. A normal chest x ray in the absence of sputum production may help to rule out tuberculosis in such cases. All patients with nCovid19 bronchopneumonia should undergo a careful chest x ray evaluation for any lesions suggestive of tuberculosis. All patients with chest radiological abnormality should undergo sputum examination to rule tuberculosis as atypical radiological manifestations may be more common in patients with nCovid19. Symptoms, signs, clinical features and chest radiographic features of Pulmonary tuberculosis and nCovid19 bronchopneumonia may overlap in some cases. Correlation of chest radiographic findings with epidemiologic history, clinical presentation, and RT-PCR test results or in later stages antibody titres will help in confirming or excluding the diagnosis in suspected cases of nCovid19. In pulmonary tuberculosis definitive diagnosis should be established by bacteriological confirmation. Molecular diagnostic tools should be used to confirm or exclude tuberculosis in suspect cases as the results are rapid, accurate and reliable.

Estimation of tuberculosis incidence at subnational level using three methods to monitor progress towards ending TB in India, 2015-2020

OBJECTIVES

We verified subnational (state/union territory (UT)/district) claims of achievements in reducing tuberculosis (TB) incidence in 2020 compared with 2015, in India.

DESIGN

A community-based survey, analysis of programme data and anti-TB drug sales and utilisation data.

SETTING

National TB Elimination Program and private TB treatment settings in 73 districts that had filed a claim to the Central TB Division of India for progress towards TB-free status.

PARTICIPANTS

Each district was divided into survey units (SU) and one village/ward was randomly selected from each SU. All household members in the selected village were interviewed. Sputum from participants with a history of anti-TB therapy (ATT), those currently experiencing chest symptoms or on ATT were tested using Xpert/Rif/TrueNat. The survey continued until 30 Mycobacterium tuberculosis cases were identified in a district.

OUTCOME MEASURES

We calculated a direct estimate of TB incidence based on incident cases identified in the survey. We calculated an under-reporting factor by matching these cases within the TB notification system. The TB notification adjusted for this factor was the estimate by the indirect method. We also calculated TB incidence from drug sale data in the private sector and drug utilisation data in the public sector. We compared the three estimates of TB incidence in 2020 with TB incidence in 2015.

RESULTS

The estimated direct incidence ranged from 19 (Purba Medinipur, West Bengal) to 1457 (Jaintia Hills, Meghalaya) per 100 000 population. Indirect estimates of incidence ranged between 19 (Diu, Dadra and Nagar Haveli) and 788 (Dumka, Jharkhand) per 100 000 population. The incidence using drug sale data ranged from 19 per 100 000 population in Diu, Dadra and Nagar Haveli to 651 per 100 000 population in Centenary, Maharashtra.

CONCLUSION

TB incidence in 1 state, 2 UTs and 35 districts had declined by at least 20% since 2015. Two districts in India were declared TB free in 2020.

Simple tuberculosis screening tool using signs, symptoms, and risk factors to reduce the missed opportunity in the older population

Background

There is a growing concern on how to increase tuberculosis (TB) case detection in resource-poor settings. The healthcare facilities routinely providing services to the elderly for chronic diseases often failed to detect TB cases, causing a missed opportunity. This study aimed to develop a simple and sensitive screening tool using signs, symptoms, and risk factors for TB case detection in the elderly.

Methods

This was a cross-sectional study conducted from August to December 2020. A random sample of 302 subjects was taken from the elderly aged ≥ 60 years attending the outpatient polyclinic at Mangunharjo hospital, Madiun, East Java, Indonesia, for a chronic disease problem. The test was developed using 16 TB signs, symptoms, and risk factors. Test performance was assessed by comparing it against the GeneXpert MTB/RIF.

Results

Marginal analysis resulted in the optimal cut-point of ≥ 7 for the test, which gave an area under the curve (AUC) of 0.62, with the maximum marginal AUC of 0.04 (p < 0.001). The sensitivity and specificity were 60.26% and 64.29%, meaning 60 TB cases for every 100 elderly with TB would be otherwise left undetected if this screening test did not take place.

Conclusion

A simple TB screening tool with moderate sensitivity and AUC has been developed using TB signs, symptoms, and risk factors. It can be used as an initial step of the systematic TB screening in the elderly visiting healthcare facilities for routine chronic disease examination, with the additional utility of reducing the missed opportunity.

TB-Net: A Tailored, Self-Attention Deep Convolutional Neural Network Design for Detection of Tuberculosis Cases From Chest X-Ray Images

Tuberculosis (TB) remains a global health problem, and is the leading cause of death from an infectious disease. A crucial step in the treatment of tuberculosis is screening high risk populations and the early detection of the disease, with chest x-ray (CXR) imaging being the most widely-used imaging modality. As such, there has been significant recent interest in artificial intelligence-based TB screening solutions for use in resource-limited scenarios where there is a lack of trained healthcare workers with expertise in CXR interpretation. Motivated by this pressing need and the recent recommendation by the World Health Organization (WHO) for the use of computer-aided diagnosis of TB in place of a human reader, we introduce TB-Net, a self-attention deep convolutional neural network tailored for TB case screening. We used CXR data from a multi-national patient cohort to train and test our models. A machine-driven design exploration approach leveraging generative synthesis was used to build a highly customized deep neural network architecture with attention condensers. We conducted an explainability-driven performance validation process to validate TB-Net's decision-making behavior. Experiments on CXR data from a multi-national patient cohort showed that the proposed TB-Net is able to achieve accuracy/sensitivity/specificity of 99.86/100.0/99.71%. Radiologist validation was conducted on select cases by two board-certified radiologists with over 10 and 19 years of experience, respectively, and showed consistency between radiologist interpretation and critical factors leveraged by TB-Net for TB case detection for the case where radiologists identified anomalies. The proposed TB-Net not only achieves high tuberculosis case detection performance in terms of sensitivity and specificity, but also leverages clinically relevant critical factors in its decision making process. While not a production-ready solution, we hope that the open-source release of TB-Net as part of the COVID-Net initiative will support researchers, clinicians, and citizen data scientists in advancing this field in the fight against this global public health crisis.

Symptom- and chest-radiography screening for active pulmonary tuberculosis in HIV-negative adults and adults with unknown HIV status

BACKGROUND

Systematic screening in high-burden settings is recommended as a strategy for early detection of pulmonary tuberculosis disease, reducing mortality, morbidity and transmission, and improving equity in access to care. Questioning for symptoms and chest radiography (CXR) have historically been the most widely available tools to screen for tuberculosis disease. Their accuracy is important for the design of tuberculosis screening programmes and determines, in combination with the accuracy of confirmatory diagnostic tests, the yield of a screening programme and the burden on individuals and the health service.

OBJECTIVES

To assess the sensitivity and specificity of questioning for the presence of one or more tuberculosis symptoms or symptom combinations, CXR, and combinations of these as screening tools for detecting bacteriologically confirmed pulmonary tuberculosis disease in HIV-negative adults and adults with unknown HIV status who are considered eligible for systematic screening for tuberculosis disease. Second, to investigate sources of heterogeneity, especially in relation to regional, epidemiological, and demographic characteristics of the study populations.

SEARCH METHODS

We searched the MEDLINE, Embase, LILACS, and HTA (Health Technology Assessment) databases using pre-specified search terms and consulted experts for unpublished reports, for the period 1992 to 2018. The search date was 10 December 2018. This search was repeated on 2 July 2021.

SELECTION CRITERIA

Studies were eligible if participants were screened for tuberculosis disease using symptom questions, or abnormalities on CXR, or both, and were offered confirmatory testing with a reference standard. We included studies if diagnostic two-by-two tables could be generated for one or more index tests, even if not all participants were subjected to a microbacteriological reference standard. We excluded studies evaluating self-reporting of symptoms.

DATA COLLECTION AND ANALYSIS

We categorized symptom and CXR index tests according to commonly used definitions. We assessed the methodological quality of included studies using the QUADAS-2 instrument. We examined the forest plots and receiver operating characteristic plots visually for heterogeneity. We estimated summary sensitivities and specificities (and 95% confidence intervals (CI)) for each index test using bivariate random-effects methods. We analyzed potential sources of heterogeneity in a hierarchical mixed-model.

MAIN RESULTS

The electronic database search identified 9473 titles and abstracts. Through expert consultation, we identified 31 reports on national tuberculosis prevalence surveys as eligible (of which eight were already captured in the search of the electronic databases), and we identified 957 potentially relevant articles through reference checking. After removal of duplicates, we assessed 10,415 titles and abstracts, of which we identified 430 (4%) for full text review, whereafter we excluded 364 articles. In total, 66 articles provided data on 59 studies. We assessed the 2 July 2021 search results; seven studies were potentially eligible but would make no material difference to the review findings or grading of the evidence, and were not added in this edition of the review. We judged most studies at high risk of bias in one or more domains, most commonly because of incorporation bias and verification bias. We judged applicability concerns low in more than 80% of studies in all three domains. The three most common symptom index tests, cough for two or more weeks (41 studies), any cough (21 studies), and any tuberculosis symptom (29 studies), showed a summary sensitivity of 42.1% (95% CI 36.6% to 47.7%), 51.3% (95% CI 42.8% to 59.7%), and 70.6% (95% CI 61.7% to 78.2%, all very low-certainty evidence), and a specificity of 94.4% (95% CI 92.6% to 95.8%, high-certainty evidence), 87.6% (95% CI 81.6% to 91.8%, low-certainty evidence), and 65.1% (95% CI 53.3% to 75.4%, low-certainty evidence), respectively. The data on symptom index tests were more heterogenous than those for CXR. The studies on any tuberculosis symptom were the most heterogeneous, but had the lowest number of variables explaining this variation. Symptom index tests also showed regional variation. The summary sensitivity of any CXR abnormality (23 studies) was 94.7% (95% CI 92.2% to 96.4%, very low-certainty evidence) and 84.8% (95% CI 76.7% to 90.4%, low-certainty evidence) for CXR abnormalities suggestive of tuberculosis (19 studies), and specificity was 89.1% (95% CI 85.6% to 91.8%, low-certainty evidence) and 95.6% (95% CI 92.6% to 97.4%, high-certainty evidence), respectively. Sensitivity was more heterogenous than specificity, and could be explained by regional variation. The addition of cough for two or more weeks, whether to any (pulmonary) CXR abnormality or to CXR abnormalities suggestive of tuberculosis, resulted in a summary sensitivity and specificity of 99.2% (95% CI 96.8% to 99.8%) and 84.9% (95% CI 81.2% to 88.1%) (15 studies; certainty of evidence not assessed).

AUTHORS' CONCLUSIONS

The summary estimates of the symptom and CXR index tests may inform the choice of screening and diagnostic algorithms in any given setting or country where screening for tuberculosis is being implemented. The high sensitivity of CXR index tests, with or without symptom questions in parallel, suggests a high yield of persons with tuberculosis disease. However, additional considerations will determine the design of screening and diagnostic algorithms, such as the availability and accessibility of CXR facilities or the resources to fund them, and the need for more or fewer diagnostic tests to confirm the diagnosis (depending on screening test specificity), which also has resource implications. These review findings should be interpreted with caution due to methodological limitations in the included studies and regional variation in sensitivity and specificity. The sensitivity and specificity of an index test in a specific setting cannot be predicted with great precision due to heterogeneity. This should be borne in mind when planning for and implementing tuberculosis screening programmes.

Applications of artificial intelligence in the thorax: a narrative review focusing on thoracic radiology

Objective

This review will focus on how AI—and, specifically, deep learning—can be applied to complement aspects of the current healthcare system. We describe how AI-based tools can augment existing clinical workflows by discussing the applications of AI to worklist prioritization and patient triage, the performance-boosting effects of AI as a second reader, and the use of AI to facilitate complex quantifications. We also introduce prominent examples of recent AI applications, such as tuberculosis screening in resource-constrained environments, the detection of lung cancer with screening CT, and the diagnosis of COVID-19. We also provide examples of prognostic predictions and new discoveries beyond existing clinical practices.

Background

Artificial intelligence (AI) has shown promising performance for thoracic diseases, particularly in the field of thoracic radiology. However, it has not yet been established how AI-based image analysis systems can help physicians in clinical practice.

Methods

This review included peer-reviewed research articles on AI in the thorax published in English between 2015 and 2021.

Conclusions

With advances in technology and appropriate preparation of physicians, AI could address various clinical problems that have not been solved due to a lack of clinical resources or technological limitations.

Keywords

Artificial intelligence (AI); deep learning (DL); computer aided diagnosis (CAD); thoracic radiology; pulmonary medicine

The definition of tuberculosis infection based on the spectrum of tuberculosis disease

Latent tuberculosis infection was the term traditionally used to indicate tuberculosis (TB) infection. This term was used to define “a state of persistent immune response to stimulation by Mycobacterium tuberculosis antigens through tests such as the tuberculin skin test (TST) or an interferon-γ release assay (IGRA) without clinically active TB”. Recent evidence indicates that the spectrum from TB infection to TB disease is much more complex, including a “continuum” of situations didactically reported as uninfected individual, TB infection, incipient TB, subclinical TB without signs/symptoms, subclinical TB with unrecognised signs/symptoms, and TB disease with signs/symptoms. Recent evidence suggests that subclinical TB is responsible for important M. tuberculosis transmission. This review describes the different stages described above and their relationships. It also summarises the new developments in prevention, diagnosis and treatment of TB infection as well as their public health and policy implications.

The evolution from TB infection to disease is now described as a “continuum process”. Understanding of this is important to appreciate what is new on prevention, diagnosis and treatment of TB infection.https://bit.ly/3jauRKA

On the Influence of the Quality of Databases of X-Ray Images of Patients with Tuberculosis on the Diagnostics of Deceases

The reliability of the training samples used for training convolutional neural networks applied for diagnostics of lung diseases has been investigated. By the example of the Kaggle database, it has been shown that the sample containing 3500 roentgenograms of healthy people and 3500 roentgenograms of patients with tuberculosis is very inhomogeneous. When training by using different parts of the sample and recognizing its different parts, the results obtained are significantly different.

Triage of Persons With Tuberculosis Symptoms Using Artificial Intelligence-Based Chest Radiograph Interpretation: A Cost-Effectiveness Analysis

BACKGROUND

In settings without access to rapid expert radiographic interpretation, artificial intelligence (AI)-based chest radiograph (CXR) analysis can triage persons presenting with possible tuberculosis (TB) symptoms, to identify those who require additional microbiological testing. However, there is limited evidence of the cost-effectiveness of this technology as a triage tool.

METHODS

A decision analysis model was developed to evaluate the cost-effectiveness of triage strategies with AI-based CXR analysis for patients presenting with symptoms suggestive of pulmonary TB in Karachi, Pakistan. These strategies were compared to the current standard of care using microbiological testing with smear microscopy or GeneXpert, without prior triage. Positive triage CXRs were considered to improve referral success for microbiologic testing, from 91% to 100% for eligible persons. Software diagnostic accuracy was based on a prospective field study in Karachi. Other inputs were obtained from the Pakistan TB Program. The analysis was conducted from the healthcare provider perspective, and costs were expressed in 2020 US dollars.

RESULTS

Compared to upfront smear microscopy for all persons with presumptive TB, triage strategies with AI-based CXR analysis were projected to lower costs by 19%, from $23233 per 1000 persons, and avert 3%-4% disability-adjusted life-years (DALYs), from 372 DALYs. Compared to upfront GeneXpert, AI-based triage strategies lowered projected costs by 37%, from $34346 and averted 4% additional DALYs, from 369 DALYs. Reinforced follow-up for persons with positive triage CXRs but negative microbiologic tests was particularly cost-effective.

CONCLUSIONS

In lower-resource settings, the addition of AI-based CXR triage before microbiologic testing for persons with possible TB symptoms can reduce costs, avert additional DALYs, and improve TB detection.

Symptom-based Screening Versus Chest Radiography for TB Child Contacts: A Systematic Review and Meta-analysis

BACKGROUND

Accessibility to chest radiography remains a major challenge in high burden and low-income countries. The World Health Organization (WHO) guidelines acknowledge that for child contacts under 5 years, a negative symptom-based screening is sufficient to exclude active tuberculosis (TB), but in child contacts older than 5 years, a chest radiograph should be considered. We performed a systematic review and meta-analysis to assess the performance of symptom-based screening compared with chest radiography in household contacts under 15 years in low-income and middle-income countries.

METHODS

Screening articles published prior 1 October 2020 and data extraction were performed by 2 independent reviewers. The primary outcome was the concordance between symptom screening and chest radiography using the prevalence adjusted bias adjusted kappa coefficient (PABAK) and the proportion of asymptomatic children with negative chest radiography. The analysis was stratified by age group.

RESULTS

Of 639 identified articles, 10 were included. PABAK varied between 0.09 and 0.97 and between 0.22 and 0.98, in children less than 5 years and 5-14 years, respectively. The pooled proportion of children with both non-TB suggestive symptoms and chest radiography findings was 98.7% (96.9-99.8) in children less than 5 years and 98.1% (93.8-100) in children of age 5-14 years.

CONCLUSIONS

Despite low concordance between symptom-based screening and chest radiography, most children without TB suggestive symptoms did not have chest radiography findings suggestive of TB. These results suggest that a negative symptom screening is sufficient to rule out active TB, supporting the WHO recommendation to use symptom-based screening alone when chest radiography is not available.

Comparison of yield and relative costs of different screening algorithms for tuberculosis in active case-finding: a cross-section study

Background

Part of tuberculosis (TB) patients were missed if symptomatic screening was based on the main TB likely symptoms. This study conducted to compare the yield and relative costs of different TB screening algorithms in active case-finding in the whole population in China.

Methods

The study population was screened based on the TB likely symptoms through a face-to-face interview in selected 27 communities from 10 counties of 10 provinces in China. If the individuals had any of the enhanced TB likely symptoms, both chest X-ray and sputum tests were carried out for them furtherly. We used the McNemar test to analyze the difference in TB detection among four algorithms in active case-finding. Of four algorithms, two were from WHO recommendations including 1a/1c, one from China National Tuberculosis Program, and one from this study with the enhanced TB likely symptoms. Furthermore, a two-way ANOVA analysis was performed to analyze the cost difference in the performance of active case-finding adjusted by different demographic and health characteristics among different algorithms.

Results

Algorithm with the enhanced TB likely symptoms defined in this study could increase the yield of TB detection in active case-finding, compared with algorithms recommended by WHO (p < 0.01, Kappa 95% CI: 0. 93–0.99) and China NTP (p = 0.03, Kappa 95% CI: 0.96–1.00). There was a significant difference in the total costs among different three algorithms WHO 1c/2/3 (F = 59.13, p < 0.01). No significant difference in the average costs for one active TB case screened and diagnosed through the process among Algorithms 1c/2/3 was evident (F = 2.78, p = 0.07). The average costs for one bacteriological positive case through algorithm WHO 1a was about two times as much as the costs for one active TB case through algorithms WHO 1c/2/3.

Conclusions

Active case-finding based on the enhanced symptom screening is meaningful for TB case-finding and it could identify more active TB cases in time. The findings indicated that this enhanced screening approach cost more compared to algorithms recommend by WHO and China NTP, but the increased yield resulted in comparative costs per patient. And it cost much more that only smear/bacteriological-positive TB cases are screened in active case-finding.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12879-021-06486-w.

Identifying barriers and facilitators to implementation of community-based tuberculosis active case finding with mobile X-ray units in Lima, Peru: a RE-AIM evaluation

OBJECTIVES

Identify barriers and facilitators to integrating community tuberculosis screening with mobile X-ray units into a health system.

METHODS

Reach, effectiveness, adoption, implementation and maintenance evaluation.

SETTING

3-district region of Lima, Peru.

PARTICIPANTS

63 899 people attended the mobile units from 7 February 2019 to 6 February 2020.

INTERVENTIONS

Participants were screened by chest radiography, which was scored for abnormality by computer-aided detection. People with abnormal X-rays were evaluated clinically and by GeneXpert MTB/RIF (Xpert) sputum testing. People diagnosed with tuberculosis at the mobile unit were accompanied to health facilities for treatment initiation.

PRIMARY AND SECONDARY OUTCOME MEASURES

Reach was defined as the percentage of the population of the three-district region that attended the mobile units. Effectiveness was defined as the change in tuberculosis case notifications over a historical baseline. Key implementation fidelity indicators were the percentages of people who had chest radiography performed, were evaluated clinically, had sputum samples collected, had valid Xpert results and initiated treatment.

RESULTS

The intervention reached 6% of the target population and was associated with an 11% (95% CI 6 to 16) increase in quarterly case notifications, adjusting for the increasing trend in notifications over the previous 3 years. Implementation indicators for screening, sputum collection and Xpert testing procedures all exceeded 85%. Only 82% of people diagnosed with tuberculosis at the mobile units received treatment; people with negative or trace Xpert results were less likely to receive treatment. Suboptimal treatment initiation was driven by health facility doctors' lack of familiarity with Xpert and lack of confidence in diagnoses made at the mobile unit.

CONCLUSION

Mobile X-ray units were a feasible and effective strategy to extend tuberculosis diagnostic services into communities and improve early case detection. Effective deployment however requires advance coordination among stakeholders and targeted provider training to ensure that people diagnosed with tuberculosis by new modalities receive prompt treatment.

Computer-aided interpretation of chest radiography reveals the spectrum of tuberculosis in rural South Africa

Computer-aided digital chest radiograph interpretation (CAD) can facilitate high-throughput screening for tuberculosis (TB), but its use in population-based active case-finding programs has been limited. In an HIV-endemic area in rural South Africa, we used a CAD algorithm (CAD4TBv5) to interpret digital chest x-rays (CXR) as part of a mobile health screening effort. Participants with TB symptoms or CAD4TBv5 score above the triaging threshold were referred for microbiological sputum assessment. During an initial pilot phase, a low CAD4TBv5 triaging threshold of 25 was selected to maximize TB case finding. We report the performance of CAD4TBv5 in screening 9,914 participants, 99 (1.0%) of whom were found to have microbiologically proven TB. CAD4TBv5 was able to identify TB cases at the same sensitivity but lower specificity as a blinded radiologist, whereas the next generation of the algorithm (CAD4TBv6) achieved comparable sensitivity and specificity to the radiologist. The CXRs of people with microbiologically confirmed TB spanned a range of lung field abnormality, including 19 (19.2%) cases deemed normal by the radiologist. HIV serostatus did not impact CAD4TB's performance. Notably, 78.8% of the TB cases identified during this population-based survey were asymptomatic and therefore triaged for sputum collection on the basis of CAD4TBv5 score alone. While CAD4TBv6 has the potential to replace radiologists for triaging CXRs in TB prevalence surveys, population-specific piloting is necessary to set the appropriate triaging thresholds. Further work on image analysis strategies is needed to identify radiologically subtle active TB.

'If not TB, what could it be?' Chest X-ray findings from the 2016 Kenya Tuberculosis Prevalence Survey

BACKGROUND

The prevalence of diseases other than TB detected during chest X-ray (CXR) screening is unknown in sub-Saharan Africa. This represents a missed opportunity for identification and treatment of potentially significant disease. Our aim was to describe and quantify non-TB abnormalities identified by TB-focused CXR screening during the 2016 Kenya National TB Prevalence Survey.

METHODS

We reviewed a random sample of 1140 adult (≥15 years) CXRs classified as 'abnormal, suggestive of TB' or 'abnormal other' during field interpretation from the TB prevalence survey. Each image was read (blinded to field classification and study radiologist read) by two expert radiologists, with images classified into one of four major anatomical categories and primary radiological findings. A third reader resolved discrepancies. Prevalence and 95% CIs of abnormalities diagnosis were estimated.

FINDINGS

Cardiomegaly was the most common non-TB abnormality at 259 out of 1123 (23.1%, 95% CI 20.6% to 25.6%), while cardiomegaly with features of cardiac failure occurred in 17 out of 1123 (1.5%, 95% CI 0.9% to 2.4%). We also identified chronic pulmonary pathology including suspected COPD in 3.2% (95% CI 2.3% to 4.4%) and non-specific patterns in 4.6% (95% CI 3.5% to 6.0%). Prevalence of active-TB and severe post-TB lung changes was 3.6% (95% CI 2.6% to 4.8%) and 1.4% (95% CI 0.8% to 2.3%), respectively.

INTERPRETATION

Based on radiological findings, we identified a wide variety of non-TB abnormalities during population-based TB screening. TB prevalence surveys and active case finding activities using mass CXR offer an opportunity to integrate disease screening efforts.

FUNDING

National Institute for Health Research (IMPALA-grant reference 16/136/35).

Active case-finding of tuberculosis in general populations and at-risk groups: a systematic review and meta-analysis

The World Health Organization (WHO) recommends active case-finding (ACF) of Tuberculosis (TB) in certain high-risk groups; however, more evidence is needed to elucidate the scope of ACF beyond the current recommendations. In this study we aimed to systematically review yields (the prevalence of active TB) of studies on ACF in general populations and at-risk groups.The review protocol was registered with PROSPERO (registration no.: CRD42020206856). A literature search in PubMed, Embase, and CENTRAL was performed for studies concluded after 31/12/1999 and published before 01/09/2020. Screening yields were estimated and yield/prevalence ratios (ratio between yield of study and WHO estimated prevalence of TB) were calculated to assess which groups might especially benefit from ACF. Finally, risk of bias was assessed, and heterogeneity was investigated using meta-regression and sensitivity analyses.We included 197 studies, with a total of 12 372 530 screened and 53 158 cases found. Yields were high among drug users, close contacts, the poor and marginalised, people living with HIV (PLHIV), and prison inmates across incidence strata and estimated yield/prevalence ratios in screenings of general populations tended to be >1 with an overall ratio of 1.4 and ranging between 1.0 and 1.5. Sensitivity analyses suggested that inclusion of studies at high risk of bias contributed to underestimation of yields.Despite many studies using insensitive screening methods, these results suggest that more at-risk groups should be considered for inclusion in future screening recommendations and that screening of general populations may outperform current case-finding practices, providing evidence for extending ACF beyond the current recommendations.

Case finding of tuberculosis among mining communities in Ghana

Background

Data on active TB case finding activities among artisanal gold mining communities (AMC) is limited. The study assessed the yield of TB cases from the TB screening activities among AMC in Ghana, the factors associated with TB in these communities and the correlation between the screening methods and a diagnosis of TB.

Methods

We conducted secondary data analyses of NTP program data collected from TB case finding activities using symptom screening and mobile X-ray implemented in hard to reach AMC. Yield of TB cases, number needed to screen (NNS) and the number needed to test (NNT) to detect a TB case were assessed and logistic regression were conducted to assess factors associated with TB. The performance of screening methods chest X-ray and symptoms in the detection of TB cases was also evaluated.

Results

In total 10,441 people from 78 communities in 24 districts were screened, 55% were female and 60% (6,296) were in the aged 25 to 54 years. Ninety-five TB cases were identified, 910 TB cases per 100,000 population screened; 5.6% of the TB cases were rifampicin resistant. Being male (aOR 5.96, 95% CI 3.25–10.92, P < 0.001), a miner (aOR 2.70, 95% CI 1.47–4.96, P = 0.001) and age group 35 to 54 years (aOR 2.27, 95% CI 1.35–3.84, P = 0.002) were risk factors for TB. NNS and NNT were 110 and 24 respectively.; Cough of any duration had the strongest association with X-ray suggestive of TB with a correlation coefficient of 0.48. Cough was most sensitive for a diagnosis of TB; sensitivity of 86.3% (95% CI 79.4–93.2) followed by X-ray, sensitivity 81.1% (95% CI 71.7–88.4). The specificities of the symptoms and X-rays ranged from 80.2% (cough) to 97.3% (sputum).

Conclusion

The high risk of TB in the artisanal mining communities and in miners in this study reinforces the need to target these populations with outreach programs particularly in hard to reach areas. The diagnostic value of cough highlights the usefulness of symptom screening in this population that may be harnessed even in the absence of X-ray to identify those suspected to have TB for further evaluation.

"Diagnosis on the Dock" project: A proactive screening program for diagnosing pulmonary tuberculosis in disembarking refugees and new SEI model

OBJECTIVE

From 2011 to 2017, the total number of refugees arriving in Europe, particularly in Italy, climbed dramatically. Our aim was to diagnose pulmonary TB in migrants coming from the African coast using a clinical-based port of arrival (PoA) screening program.

METHODS

From 2016 to 2018, migrants coming via the Mediterranean Route were screened for body temperature and the presence of cough directly on the dock: if they were feverish with productive cough, their sputum was examined with NAAT; with a dry cough, they underwent Chest-X-ray (CXR). Those migrants with positive NAAT or CXR suggestive for TB were admitted to our ward. In addition, we plotted an SEI simulation of our project to evaluate the epidemiological impact of our screening.

RESULTS

Out of 33.676 disembarking migrants, 314 (0.9%) had fever and cough: 80 (25.47%) with productive cough underwent NAAT in sputum, and 16 were positive for TB; 234 (74.52%) with dry cough had a CXR examination, and 39 were suggestive of TB, later confirmed by mycobacterial culture. The SEI-new model analysis demonstrated that our screening program significantly reduced TB spreading all over the country.

CONCLUSIONS

For possible future high migrant flows, PoA screening for TB has to be considered feasible and effective in decreasing TB spreading.

Application of artificial intelligence in digital chest radiography reading for pulmonary tuberculosis screening

Currently, the diagnosis of tuberculosis (TB) is mainly based on the comprehensive consideration of the patient's symptoms and signs, laboratory examinations and chest radiography (CXR). CXR plays a pivotal role to support the early diagnosis of TB, especially when used for TB screening and differential diagnosis. However, high cost of CXR hardware and shortage of certified radiologists poses a major challenge for CXR application in TB screening in resource limited settings. The latest development of artificial intelligence (AI) combined with the accumulation of a large number of medical images provides new opportunities for the establishment of computer-aided detection (CAD) systems in the medical applications, especially in the era of deep learning (DL) technology. Several CAD solutions are now commercially available and there is growing evidence demonstrate their value in imaging diagnosis. Recently, WHO published a rapid communication which stated that CAD may be used as an alternative to human reader interpretation of plain digital CXRs for screening and triage of TB.

Evaluation of underweight status may improve identification of the highest-risk patients during outpatient evaluation for pulmonary tuberculosis

BACKGROUND

When evaluating symptomatic patients for tuberculosis (TB) without access to same-day diagnostic test results, clinicians often make empiric decisions about starting treatment. The number of TB symptoms and/or underweight status could help identify patients at highest risk for a positive result. We sought to evaluate the usefulness of BMI assessment and a count of characteristic TB symptoms for identifying patients at highest risk for TB.

METHODS

We enrolled adult patients receiving pulmonary TB diagnoses and a representative sample with negative TB evaluations at four outpatient health facilities in Kampala, Uganda. We asked patients about symptoms of chronic cough, night sweats, chest pain, fever, hemoptysis, or weight loss; measured height and weight; and collected sputum for mycobacterial culture. We evaluated the diagnostic accuracy (for culture-positive TB) of two simple scoring systems: (a) number of TB symptoms, and (b) number of TB symptoms plus one or more additional points for underweight status (body mass index [BMI] ≤ 18.5 kg/m2).

RESULTS

We included 121 patients with culture-positive TB and 370 patients with negative culture results (44 of whom had been recommended for TB treatment by evaluating clinicians). Of the six symptoms assessed, the median number of symptoms that patients reported was two (interquartile range [IQR]: 1, 3). The median BMI was 20.9 kg/m2 (IQR: 18.6, 24.0), and 118 (24%) patients were underweight. Counting the number of symptoms provided an area under the Receiver Operating Characteristic curve (c-statistic) of 0.77 (95% confidence interval, CI: 0.72, 0.81) for identifying culture-positive TB; adding two points for underweight status increased the c-statistic to 0.81 (95%CI: 0.76, 0.85). A cutoff of ≥3 symptoms had sensitivity and specificity of 65% and 74%, whereas a score of ≥4 on the combined score (≥2 symptoms if underweight, ≥4 symptoms if not underweight) gave higher sensitivity and specificity of 69% and 81% respectively. A sensitivity analysis defining TB by Xpert MTB/RIF status produced similar results.

CONCLUSION

A count of patients' TB symptoms may be useful in clinical decision-making about TB diagnosis. Consideration of underweight status adds additional diagnostic value.

Trade-offs between cost and accuracy in active case finding for tuberculosis: A dynamic modelling analysis

BACKGROUND

Active case finding (ACF) may be valuable in tuberculosis (TB) control, but questions remain about its optimum implementation in different settings. For example, smear microscopy misses up to half of TB cases, yet is cheap and detects the most infectious TB cases. What, then, is the incremental value of using more sensitive and specific, yet more costly, tests such as Xpert MTB/RIF in ACF in a high-burden setting?

METHODS AND FINDINGS

We constructed a dynamic transmission model of TB, calibrated to be consistent with an urban slum population in India. We applied this model to compare the potential cost and impact of 2 hypothetical approaches following initial symptom screening: (i) 'moderate accuracy' testing employing a microscopy-like test (i.e., lower cost but also lower accuracy) for bacteriological confirmation and (ii) 'high accuracy' testing employing an Xpert-like test (higher cost but also higher accuracy, while also detecting rifampicin resistance). Results suggest that ACF using a moderate-accuracy test could in fact cost more overall than using a high-accuracy test. Under an illustrative budget of US$20 million in a slum population of 2 million, high-accuracy testing would avert 1.14 (95% credible interval 0.75-1.99, with p = 0.28) cases relative to each case averted by moderate-accuracy testing. Test specificity is a key driver: High-accuracy testing would be significantly more impactful at the 5% significance level, as long as the high-accuracy test has specificity at least 3 percentage points greater than the moderate-accuracy test. Additional factors promoting the impact of high-accuracy testing are that (i) its ability to detect rifampicin resistance can lead to long-term cost savings in second-line treatment and (ii) its higher sensitivity contributes to the overall cases averted by ACF. Amongst the limitations of this study, our cost model has a narrow focus on the commodity costs of testing and treatment; our estimates should not be taken as indicative of the overall cost of ACF. There remains uncertainty about the true specificity of tests such as smear and Xpert-like tests in ACF, relating to the accuracy of the reference standard under such conditions.

CONCLUSIONS

Our results suggest that cheaper diagnostics do not necessarily translate to less costly ACF, as any savings from the test cost can be strongly outweighed by factors including false-positive TB treatment, reduced sensitivity, and foregone savings in second-line treatment. In resource-limited settings, it is therefore important to take all of these factors into account when designing cost-effective strategies for ACF.

The impact of chest radiography and Xpert MTB/RIF testing among household contacts in Chennai, India

Tuberculosis prevalence surveys have demonstrated the benefit of screening with chest x-ray (CXR) and sensitive diagnostic tests compared to symptoms and smear microscopy. However, in programmatic practice there is little evidence on the yield of different algorithms. We implemented contact tracing in Chennai, India for adult sputum-positive TB patients registered from January 2015 to March 2016. Patients with symptoms or abnormal X-ray findings further underwent testing using Xpert MTB/RIF (Xpert) and smear microscopy. A retrospective cohort study was done to summarize the key findings. We verbally screened 5553 contacts for symptoms, CXR through private sector collaboration, Xpert, and smear microscopy. Overall, 1312 (23.6%) contacts screened positive. CXR alone identified 531 (40.5%) of them, 679 (51.8%) were symptom-positive only, while 102 (7.8%) were positive on both the symptom and CXR screen. Overall, 35 bacteriologically positive cases were identified (0.7%). A standard approach of symptoms screening followed by microscopy identified only 9 (25.7%) of the total number of bacteriologically positive cases, whereas the combination of a CRX screening followed by microscopy identified 13 (37.1%) of the cases. The algorithm of symptoms screening followed by Xpert testing, detected 20 cases, whereas the combination of symptoms and CXR followed by Xpert increased this number to 35 (75% increase compared to symptoms and Xpert). Optimal use of more sensitive screening tests, better diagnostic tests, and novel private sector engagement can improve diagnostic yield in a programmatic setting.

Development and Validation of a Deep Learning-based Automatic Detection Algorithm for Active Pulmonary Tuberculosis on Chest Radiographs

Background

Detection of active pulmonary tuberculosis on chest radiographs (CRs) is critical for the diagnosis and screening of tuberculosis. An automated system may help streamline the tuberculosis screening process and improve diagnostic performance.

Methods

We developed a deep learning–based automatic detection (DLAD) algorithm using 54c221 normal CRs and 6768 CRs with active pulmonary tuberculosis that were labeled and annotated by 13 board-certified radiologists. The performance of DLAD was validated using 6 external multicenter, multinational datasets. To compare the performances of DLAD with physicians, an observer performance test was conducted by 15 physicians including nonradiology physicians, board-certified radiologists, and thoracic radiologists. Image-wise classification and lesion-wise localization performances were measured using area under the receiver operating characteristic (ROC) curves and area under the alternative free-response ROC curves, respectively. Sensitivities and specificities of DLAD were calculated using 2 cutoffs (high sensitivity [98%] and high specificity [98%]) obtained through in-house validation.

Results

DLAD demonstrated classification performance of 0.977–1.000 and localization performance of 0.973–1.000. Sensitivities and specificities for classification were 94.3%–100% and 91.1%–100% using the high-sensitivity cutoff and 84.1%–99.0% and 99.1%–100% using the high-specificity cutoff. DLAD showed significantly higher performance in both classification (0.993 vs 0.746–0.971) and localization (0.993 vs 0.664–0.925) compared to all groups of physicians.

Conclusions

Our DLAD demonstrated excellent and consistent performance in the detection of active pulmonary tuberculosis on CR, outperforming physicians, including thoracic radiologists.

Providing TB and HIV outreach services to internally displaced populations in Northeast Nigeria: Results of a controlled intervention study

BACKGROUND

A decade of Boko Haram insurgency brought conflict, mass displacement, and the destruction of basic infrastructure to Northeast Nigeria. Over 2 million internally displaced persons (IDPs) suffering from lack of basic hygienic conditions, malnutrition, and disease live in camps or are hosted by communities in the region, where the conflict has contributed to a massive destruction of health facilities. Infectious diseases like tuberculosis (TB) and HIV are especially difficult to address under such conditions, and IDPs are vulnerable to both. Although international investment supports some health interventions among IDPs, locally sourced solutions are lacking.

METHODS AND FINDINGS

We evaluated the impact of an active case finding (ACF) intervention for TB and testing for HIV in IDP communities and provided linkages to treatment in 3 states in Northeast Nigeria: Adamawa, Gombe, and Yobe. The ACF was a component of a multistakeholder collaboration between government, civil society, and IDP community partners, which also included mapping of IDP populations and health services, supporting existing health facilities, developing a sample transport network, and organizing community outreach to support ACF. Between July 1, 2017, and June 30, 2018, ACF was conducted in 26 IDP camps and 963 host communities in 12 local government areas (LGAs) with another 12 LGAs serving as a control population. Outreach efforts resulted in 283,556 screening encounters. We screened 13,316 children and 270,239 adults including 150,303 (55.6%) adult women and 119,936 (44.4%) men. We tested 17,134 people for TB and 58,976 for HIV. We detected 1,423 people with TB and 874 people living with HIV. We linked 1,419 people to anti-TB treatment and 874 people with HIV to antiretroviral treatment sites. We evaluated additional TB cases notified and conducted comparative interrupted time series (ITS) analyses to assess the impact of ACF on TB case notifications. Through our efforts, bacteriologically confirmed TB notifications increased by 847 (45.1%) during the intervention period, with IDPs accounting for 46% of these notifications. The ITS analyses detected significant positive postintervention trend differences in TB notification rates between the intervention and control areas in all forms TB (incidence rate ratio [IRR] = 1.136 [1.072, 1.204]; p ≤ 0.001) and bacteriologically positive TB (IRR = 1.141 [1.058, 1.229]; p = 0.001). The TB prevalence (502 cases per 100,000 screening encounters) was 10 times the national notification rates and 2.3 times the estimated national incidence. Rates of HIV infection (1.8%) were higher than HIV prevalence estimates in the 3 states. Our study was limited by the nonrandom selection of LGAs. Furthermore, we did not use sensitive screening tools like chest X-ray and likely missed people with TB.

CONCLUSIONS

In this study, we observed a burden of TB in IDP populations of Northeast Nigeria many times higher than national rates and HIV rates higher than state level estimates. The impact of the intervention showed that ACF can greatly increase TB case notifications. Engaging IDP communities, local governments, and civil society organizations is essential to ensuring the success of interventions targeting TB and HIV, and such approaches can provide sustained solutions to these and other health crises among vulnerable populations.

Prompt recognition of infectious pulmonary tuberculosis is critical to achieving elimination goals: a retrospective cohort study

INTRODUCTION

All pulmonary tuberculosis (PTB) cases are presumed to be infectious to some degree. This spectrum of infectiousness is independently described by both the acid-fast bacilli smear and radiographic findings. Smear-positive patients with chest radiographic findings that are typical for adult-type PTB are believed to be most infectious.

HYPOTHESIS

Characterisation of the presumed most infectious PTB case is possible by reference to readily available clinical features and laboratory results.

METHODS

Retrospective cohort study of adult, culture-positive PTB cases (151 smear-positive; 162 smear-negative) diagnosed between 1 January 2013 and 30 April 2017 in Canada. We describe cases according to demographic, clinical and laboratory features. We use multivariable multinomial logistic regression to estimate the relative risk ratio (RRR) with 95% CI of features associated with an outcome of smear-positive PTB, characterised by 'typical' chest radiograph findings.

RESULTS

Being Canadian-born, symptomatic, having a subacute duration of symptoms and broad-spectrum antibiotic prescriptions were all more commonly associated with smear-positive than smear-negative disease (36% vs 20%; 95% vs 63%; 88% vs 54%; and 59% vs 28%, respectively). After combining smear status and radiographic features, we show that smear-positive patients with typical chest radiographs were younger, had a longer duration of symptoms (RRR 2.41; 95% CI 1.01 to 5.74 and 2.93; 95% CI 1.20 to 7.11, respectively) and were less likely to be foreign-born, or have a moderate to high-risk factor for reactivation (RRR 0.40; 95% CI 0.17 to 0.92 and 0.18; 95% CI 0.04 to 0.71, respectively) compared with smear-negative patients with atypical chest radiograph findings.

CONCLUSION

A clear picture of the presumed most infectious PTB case emerges from available historical and laboratory information; vigilance for this presentation by front-line providers will support elimination strategies aimed at reducing transmission.

Clinical Implementation of Deep Learning in Thoracic Radiology: Potential Applications and Challenges

Chest X-ray radiography and computed tomography, the two mainstay modalities in thoracic radiology, are under active investigation with deep learning technology, which has shown promising performance in various tasks, including detection, classification, segmentation, and image synthesis, outperforming conventional methods and suggesting its potential for clinical implementation. However, the implementation of deep learning in daily clinical practice is in its infancy and facing several challenges, such as its limited ability to explain the output results, uncertain benefits regarding patient outcomes, and incomplete integration in daily workflow. In this review article, we will introduce the potential clinical applications of deep learning technology in thoracic radiology and discuss several challenges for its implementation in daily clinical practice.

Tuberculosis disease and infection among household contacts of bacteriologically confirmed and non-confirmed tuberculosis patients

OBJECTIVE

To compare the prevalence of tuberculosis infection and disease in household contacts of patients with bacteriologically confirmed tuberculosis disease and contacts of non-bacteriologically confirmed disease in western Kenya.

METHODS

We enrolled newly diagnosed index patients and their household contacts from March 2014 to June 2016. All contacts were evaluated with a symptom questionnaire, tuberculin skin test (TST) and HIV test. Clinical evaluation and sputum testing were performed for those with symptoms, positive TST result or HIV infection.

RESULTS

We enrolled 1155 contacts of 330 index patients with bacteriologically confirmed tuberculosis and 192 contacts of 55 index patients with non-bacteriologically confirmed tuberculosis. 3.5% of contacts of patients with bacteriologically confirmed tuberculosis were diagnosed with tuberculosis, whereas no contacts of index patients with non-bacteriologically confirmed tuberculosis were. Of those diagnosed with tuberculosis disease, 58.5% reported symptoms, 34.1% reported no symptoms but had positive TST results, and 7.3% had neither symptoms nor positive TST but were HIV-positive. Among 872 contacts with a TST result, 50.9% of contacts of index patients with bacteriologically confirmed tuberculosis and 41.0% of contacts of index patients with non-bacteriologically confirmed tuberculosis had a positive result (prevalence ratio = 1.16, 95% confidence interval 0.92-1.48).

CONCLUSION

In a high-burden setting, tuberculosis disease was more prevalent among contacts of patients with bacteriologically confirmed tuberculosis than contacts of patients with non-bacteriologically confirmed disease. TST was feasible to perform and helped to detect cases that would have been missed had only symptomatic contacts been evaluated.

Realizing the "40 by 2022" Commitment from the United Nations High-Level Meeting on the Fight to End Tuberculosis: What Will It Take to Meet Rapid Diagnostic Testing Needs?

Existing rapid diagnostics offer faster and more sensitive diagnosis of tuberculosis (TB) and simultaneous detection of multidrug-resistant TB. A 5-fold increase in investment in these tools is needed to meet the needs of the TB community and the United Nations’ ambitious 40 million by 2022 diagnosis and treatment target.

The potential gains from full adoption of World Health Organization (WHO)-recommended rapid diagnostics (WRDs) for tuberculosis (TB) are significant, but there is no current analysis of the additional investment needed to reach this goal. We sought to estimate the necessary investment in instruments, tests, and money, using Xpert MTB/RIF (Xpert), which detects Mycobacterium tuberculosis (MTB) and tests for resistance to rifampicin (RIF), as an example. An existing calculator for TB diagnostic needs was adapted to estimate the Xpert needs for a group of 24 countries with high TB burdens. This analysis assumed that countries will achieve the case-finding commitments agreed to at the recent United Nations High-Level Meeting on the Fight to End Tuberculosis, and that countries would adopt the WHO-recommended algorithm in which all people with signs and symptoms of TB receive an Xpert test. When compared to the current investments in these countries, this baseline model revealed that countries would require a 4-fold increase in the number of Xpert modules and a 6-fold increase in the number of Xpert test cartridges per year to meet their full testing needs. The incremental cost of the additional instruments for these countries would total approximately US$474 million, plus an incremental cost each year of cartridges of approximately $586 million, or a 5-fold increase over current investments. A sensitivity analysis revealed a variety of possible changes under alternative scenarios, but most of these changes either do not meet the global goals, are unrealistic, or would result in even greater investment needs. These findings suggest that a major investment is needed in WRD capacity to implement the recommended diagnostic algorithm for TB and reach the case-finding commitments by 2022.

Using artificial intelligence to read chest radiographs for tuberculosis detection: A multi-site evaluation of the diagnostic accuracy of three deep learning systems

Deep learning (DL) neural networks have only recently been employed to interpret chest radiography (CXR) to screen and triage people for pulmonary tuberculosis (TB). No published studies have compared multiple DL systems and populations. We conducted a retrospective evaluation of three DL systems (CAD4TB, Lunit INSIGHT, and qXR) for detecting TB-associated abnormalities in chest radiographs from outpatients in Nepal and Cameroon. All 1196 individuals received a Xpert MTB/RIF assay and a CXR read by two groups of radiologists and the DL systems. Xpert was used as the reference standard. The area under the curve of the three systems was similar: Lunit (0.94, 95% CI: 0.93-0.96), qXR (0.94, 95% CI: 0.92-0.97) and CAD4TB (0.92, 95% CI: 0.90-0.95). When matching the sensitivity of the radiologists, the specificities of the DL systems were significantly higher except for one. Using DL systems to read CXRs could reduce the number of Xpert MTB/RIF tests needed by 66% while maintaining sensitivity at 95% or better. Using a universal cutoff score resulted different performance in each site, highlighting the need to select scores based on the population screened. These DL systems should be considered by TB programs where human resources are constrained, and automated technology is available.

Accuracy of an automated system for tuberculosis detection on chest radiographs in high-risk screening

SETTING:

Tuberculosis (TB) screening programmes can be optimised by reducing the number of chest radiographs (CXRs) requiring interpretation by human experts.

OBJECTIVE:

To evaluate the performance of computerised detection software in triaging CXRs in a high-throughput digital mobile TB screening programme.

DESIGN:

A retrospective evaluation of the software was performed on a database of 38 961 postero-anterior CXRs from unique individuals seen between 2005 and 2010, 87 of whom were diagnosed with TB. The software generated a TB likelihood score for each CXR. This score was compared with a reference standard for notified active pulmonary TB using receiver operating characteristic (ROC) curve and localisation ROC (LROC) curve analyses.

RESULTS:

On ROC curve analysis, software specificity was 55.71% (95%CI 55.21–56.20) and negative predictive value was 99.98% (95%CI 99.95–99.99), at a sensitivity of 95%. The area under the ROC curve was 0.90 (95%CI 0.86–0.93). Results of the LROC curve analysis were similar.

CONCLUSION:

The software could identify more than half of the normal images in a TB screening setting while maintaining high sensitivity, and may therefore be used for triage.

A comparison of the yield and relative cost of active tuberculosis case-finding algorithms in Zimbabwe

Setting

Ten districts and three cities in Zimbabwe.

Objective

To compare the yield and relative cost of identifying a case of tuberculosis (TB) using the three WHO-recommended algorithms (WHO2b, symptom inquiry only; WHO2d, chest X-ray [CXR] after a positive symptom inquiry; WHO3b, CXR only) and the Zimbabwe active case finding (ZimACF) algorithm (symptom inquiry plus CXR) to everyone.

Design

Cross-sectional study using data from the ZimACF project.

Results

A total of 38 574 people were screened from April to December 2017; 488 (1.3%) were diagnosed with TB using the ZimACF algorithm. Fewer TB cases would have been diagnosed with the WHO-recommended algorithms. This ranged from 7% fewer (34 cases) with WHO3b, 18% fewer (88 cases) with WHO2b and 25% fewer (122 cases) with WHO2d. The need for CXR ranged from 36% (WHO2d) to 100% (WHO3b). The need for bacteriological confirmation ranged from 7% (WHO2d) to 40% (ZimACF). The relative cost per case of TB diagnosed ranged from US$180 with WHO3b to US$565 for the ZimACF algorithm.

Conclusion

The ZimACF algorithm had the highest case yield, but at a much higher cost per case than the WHO algorithms. It is possible to switch to algorithm WHO3b, but the trade-off between cost and yield needs to be reviewed by the Zimbabwean National TB Programme.

Utility of a clinical scoring system in prioritizing TB investigations - a systematic review

Introduction: Tuberculosis (TB) is among the 10 most common causes of death worldwide and it is the leading cause of mortality in people with human immunodeficiency virus (HIV). Clinical scoring systems have the potential to improve case finding and to prioritize patients for TB testing. Areas covered: This systematic review investigated the utility of prediction models to improve pulmonary tuberculosis (pTB) case finding. Studies were searched through PubMed until 15th of August 2018 and 20 studies were eligible according to the inclusion criteria. Data on study population, outcome measurements, predictors, and performance were extracted. Many studies showed promising results but lacked external validation. Furthermore, head-to-head studies are needed to compare the different prediction models. Sensitivities of the prediction models ranged from 26% to 96% and specificities from 18% to 92%, negative likelihood ratios (LR-) from 0.22 to 0.8 and positive likelihood ratios(LR+) 1.07 to 7.32. Composite scores including paraclinical measures added to sensitivity. Expert opinion: TB case finding is of utmost importance to advance the quest for global TB elimination, and simple measures to identify high-risk populations or persons to undergo further diagnostic evaluation are highly needed. A number of clinical scores are available and could be implemented in practice to improve case finding.

Tuberculosis household accompaniment to improve the contact management cascade: A prospective cohort study

Background

Appropriate management of people exposed in the home to tuberculosis is essential to prevent morbidity. These household contacts, particularly children, should receive preventive therapy to prevent them from falling ill. However, few people receive preventive therapy worldwide. We sought to determine whether a community-based accompaniment intervention could improve tuberculosis contact management.

Methods

We conducted a prospective cohort study of household contacts of tuberculosis patients who initiated treatment during September 2015-June 2016 in Lima, Peru. Enrolled households received an intervention comprising home visits, transport vouchers, assistance coordinating evaluation procedures, and adherence support during preventive therapy. To evaluate the impact of the intervention, we conducted retrospective chart reviews of all patients initiating treatment during 6-month baseline and intervention periods.

Results

We enrolled 314 household contacts of 109 index patients. Of these, 283 (90%) completed evaluation, and 4 (1%) were diagnosed with tuberculosis. Preventive therapy was prescribed for 35/38 (92%) contacts 0–19 years old who were eligible under Peruvian guidelines. Preventive therapy was also prescribed for 6/26 (23%) contacts with unknown eligibility due to lack of a tuberculin skin test (TST), and 20/69 (29%) who were ineligible either because of a negative TST result or exposure to a drug-resistant or extrapulmonary case. Of the 61 contacts who were prescribed preventive therapy, 57 (93%) initiated treatment, and 51 (91%) completed treatment. The proportion of contacts who completed evaluation increased from 42% during the baseline period to 71% during the evaluation period (risk ratio [RR] = 1.73, 95% confidence interval [95% CI]: 1.41–2.13). The proportion of contacts who initiated preventive therapy increased from 15% to 40% (RR = 2.45, 95% CI: 1.42–4.22).

Conclusion

Accompaniment of TB patient households greatly improved the evaluation of household contacts for TB and increased the use of preventive therapy.

Sub-national TB prevalence surveys in India, 2006-2012: Results of uniformly conducted data analysis

Setting

Community based tuberculosis (TB) prevalence surveys in ten sites across India during 2006–2012

Objective

To re-analyze data of recent sub-national surveys using uniform statistical methods and obtain a pooled national level estimate of prevalence of TB.

Methods

Individuals ≥15 years old were screened by interview for symptoms suggestive of Pulmonary TB (PTB) and history of anti-TB treatment; additional screening by chest radiography was undertaken in five sites. Two sputum specimens were examined by smear and culture among Screen-positives. Prevalence in each site was estimated after imputing missing values to correct for bias introduced by incompleteness of data. In five sites, prevalence was corrected for non-screening by radiography. Pooled prevalence of bacteriologically positive PTB was estimated using Random Effects Model after excluding data from one site. Overall prevalence of TB (all ages, all types) was estimated by adjusting for extra-pulmonary TB and Pediatric TB.

Results

Of 769290 individuals registered, 715989 were screened by interview and 294532 also by radiography. Sputum specimen were examined from 50 852 individuals. Estimated prevalence of smear positive, culture positive and bacteriologically positive PTB varied between 108.4–428.1, 147.9–429.8 and 170.8–528.4 per 100000 populations in different sites. Pooled estimate of prevalence of bacteriologically positive PTB was 350.0 (260.7, 439.0). Overall prevalence of TB was estimated at 300.7 (223.7–377.5) in 2009, the mid-year of surveys. Prevalence was significantly higher in rural compared to urban areas.

Conclusion

TB burden continues to be high in India suggesting further strengthening of TB control activities.

Estimating the annual risk of tuberculosis infection in Japan from interferon-gamma release assay data

To assess tuberculosis transmission frequency at a population level, the age-dependent Mantoux test has been used widely to estimate the annual risk of infection (ARI) with Mycobacterium tuberculosis. However, the widespread Bacille Calmette-Guerin (BCG) immunization program implemented in Japan in the 20th century has made natural infections with M. tuberculosis difficult to distinguish from immune responses against this vaccine. Consequently, a recognized alternative method for measuring the frequency of primary infections, the interferon-gamma release assay (IGRA), which partially decays as a function of time after infection, is used. We aimed to estimate the ARI in Japan from IGRA data along with its response decay information using mathematical modeling. Devising a partial differential equation system, we computed the probability of IGRA positivity as a function of time and age, accounting for the time-varying force of infection and decay function of the IGRA response. Jointly estimating the force of infection and the parameters governing the decay function of the IGRA response, we found that the age-dependent increasing pattern of the IGRA response was captured by the proposed simple model, yielding estimates of the time-dependent force of infection. ARI decreased as a function of time in the study subjects for all geographic locations. By the year 2030, our model showed that the median age of infection is predicted to be delayed by 40-50 years compared with that in 1940. The geographic variations in the ARI were striking, ranging from under 0.1% to 0.6% in 2018, which echoes the longstanding notion of highly heterogeneous geographical tuberculosis transmission in Japan.

Screening for active pulmonary tuberculosis: Development and applicability of artificial neural network models

Tuberculosis (TB) remains a significant public health challenge, motivated by the diversity of healthcare epidemiological settings, as other factors. Cost-effective screening has substantial importance for TB control, demanding new diagnostic tools. This paper proposes a decision support tool (DST) for screening pulmonary TB (PTB) patients at a secondary clinic. The DST is composed of an adaptive resonance model (iART) for risk group identification (low, medium and high) and a multilayer perceptron (MLP) neural network for classifying patients as active or inactive PTB. Our tool attains an overall sensitivity (SE) and specificity (SP) of 92% (95% CI; 79-97) and 58% (95% CI; 47-68), respectively. SE values for smear-positive and smear-negative patients are 96% (95% CI; 80-99) and 82% (95% CI; 52-95), as well as higher than 83% (95% CI; 43-97) in low and high-risk cases. Even in scenarios with prevalence up to 20%, negative predictive values superior to 95% are obtained. The proposed DST provides a quick and low-cost pretest for presumptive PTB patients, which is useful to guide confirmatory testing and patient management, especially in settings with limited resources in low and middle-incoming countries.

Smear Microscopy for Diagnosis of Pulmonary Tuberculosis in Eastern Sudan

Background

In Sudan, tuberculosis diagnosis largely relies on clinical symptoms and smear microscopy as in many other low- and middle-income countries. The aim of this study was to investigate the positive predictive value of a positive sputum smear in patients investigated for pulmonary tuberculosis in Eastern Sudan.

Methods

Two sputum samples from patients presenting with symptoms suggestive of tuberculosis were investigated using direct Ziehl-Neelsen (ZN) staining and light microscopy between June to October 2014 and January to July 2016. If one of the samples was smear positive, both samples were pooled, stored at −20°C, and sent to the National Reference Laboratory (NRL), Germany. Following decontamination, samples underwent repeat microscopy and culture. Culture negative/contaminated samples were investigated using polymerase chain reaction (PCR).

Results

A total of 383 samples were investigated. Repeat microscopy categorized 123 (32.1%) as negative, among which 31 were culture positive. This increased to 80 when PCR and culture results were considered together. A total of 196 samples were culture positive, of which 171 (87.3%), 14 (7.1%), and 11 (5.6%) were M. tuberculosis, M. intracellulare, and mixed species. Overall, 15.6% (57/365) of the samples had no evidence of M. tuberculosis, resulting in a positive predictive value of 84.4%.

Conclusions

There was a discordance between the results of smear microscopy performed at local laboratories in the Sudan and at the NRL, Germany; besides, a considerable number of samples had no evidence of M. tuberculosis. Improved quality control for smear microscopy and more specific diagnostics are crucial to avoid possible overtreatment.

Tuberculosis case finding using population-based disease surveillance platforms in urban and rural Kenya

Background

Tuberculosis (TB) case finding is an important component of TB control because it can reduce transmission of Mycobacterium tuberculosis (MTB) through prompt detection and treatment of infectious patients.

Methods

Using population-based infectious disease surveillance (PBIDS) platforms with links to health facilities in Kenya we implemented intensified TB case finding in the community and at the health facilities, as an adjunct to routine passive case finding conducted by the national TB program. From 2011 to 2014, PBIDS participants ≥15 years were screened either at home or health facilities for possible TB symptoms which included cough, fever, night sweats or weight loss in the preceding 2 weeks. At home, participants with possible TB symptoms had expectorated sputum collected. At the clinic, HIV-infected participants with possible TB symptoms were invited to produce sputum. Those without HIV but with symptoms lasting 7 days including the visit day had chest radiographs performed, and had sputum collected if the radiographs were abnormal. Sputum samples were tested for the presence of MTB using the Xpert MTB/RIF assay. TB detection rates were calculated per 100,000 persons screened.

Results

Of 11,191 participants aged ≥15 years screened at home at both sites, 2695 (23.9%) reported possible TB symptoms, of whom 2258 (83.8%) produced sputum specimens. MTB was detected in 32 (1.4%) of the specimens resulting in a detection rate of 286/100,000 persons screened. At the health facilities, a total of 11,762 person were screened, 7500 (63.8%) had possible TB symptoms of whom 1282 (17.1%) produced sputum samples. MTB was detected in 69 (5.4%) of the samples, resulting in an overall detection rate of 587/100,000 persons screened. The TB detection rate was higher in persons with HIV compared to those without at both home (HIV-infected - 769/100,000, HIV-uninfected 141/100,000, rate ratio (RR) – 5.45, 95% CI 3.25–22.37), and health facilities (HIV-infected 3399/100,000, HIV-uninfected 294/100,000, RR 11.56, 95% CI 6.18–18.44).

Conclusion

Facility-based intensified TB case finding detected more TB cases per the number of specimens tested and the number of persons screened, including those with HIV, than home-based TB screening and should be further evaluated to determine its potential programmatic impact.

Population-based cohort study examining the association between weight loss and pulmonary tuberculosis in adults

BACKGROUND/PURPOSE

Little research is currently available on the relationship between weight loss and pulmonary tuberculosis in Taiwan. This study aimed to evaluate whether weight loss is an early clinical feature of pulmonary tuberculosis in Taiwan.

METHOD

This population-based retrospective cohort study was conducted using the Taiwan National Health Insurance Program database. There were 6051 subjects aged 20 to 84 years with newly diagnosed weight loss from 2000 to 2012 as the weight loss group, and 24081 randomly selected subjects without weight loss from the same period as the non-weight loss group. The weight loss and the non-weight loss groups were matched by sex, age, and comorbidities. The incidence of pulmonary tuberculosis at the end of 2013 was evaluated in both groups. A multivariable Cox proportional hazards regression model was used to evaluate the hazard ratio (HR) and 95% confidence interval (CI) for association of pulmonary tuberculosis with weight loss.

RESULTS

The incidence of pulmonary tuberculosis was 15.2-fold higher in the weight loss group than in the non-weight loss group during the first 3 months of follow-up (22.8 vs. 1.50 per 1000 person-years, 95% CI 13.7, 16.9). After adjusting for covariables, the subsequent HR of pulmonary tuberculosis was 2.36 for the weight loss group (95 % CI 1.88, 2.97), compared with the non-weight loss group.

CONCLUSION

Although our finding is not novel, it does support the notion that weight loss is significantly associated with increased hazard of pulmonary tuberculosis in Taiwan. The risk was found to be particularly high during the first 3 months of follow-up.