In Sub-Saharan Africa, lower respiratory tract infections (LRTIs) and tuberculosis (TB) jointly are the leading cause of overall mortality. There is a need to integrate sustainable triage and management strategies into standard care. The TrUST study investigates the utility of point-of-care ultrasound (POCUS) for diagnosis and prognosis of LRTIs in TB endemic regions in the outpatient triage setting. Automated interpretation of POCUS by artificial intelligence (AI) may further standardize and improve its predictive utility as well as facilitate its implementation into usual practice.

Design International multicentre prospective cohort study

Setting This study aims to explore the performance of POCUS in a pre-hospital setting in countries with different prevalences of HIV: South- Africa (high prevalence), Benin (low prevalence) and Mali (low prevalence).

Study procedures Pre-recruitment training Before initiating the study, participating clinicians will benefit from lung ultrasound (LUS) and POCUS protocols for extra pulmonary TB (FASH+ exam) training with a field expert. The participating clinicians are required to pass a final expert test before start of patient recruitment and to have a minimum of 20 supervised US.

Inclusion (Day 0) Patients inclusion will be performed over a period of 18 to maximum 24 months. Recruitment will stop before the anticipated end if the inclusion of 1000 patients is reached.

Clinical data Data collected include demographic characteristics (age, sex, occupation etc.), past medical history (previous TB, Coronavirus Disease (COVID), vaccination status, documented past episodes of COVID-19, HIV status,…), symptoms (respiratory symptoms, specific COVID-19 symptoms (from the Centers for Disease Control COVID checklist), specific pulmonary TB symptoms (from the CDC TB checklist) and routine clinical examination findings (vital parameters, oxygen saturation, lung auscultation,…).

Laboratory analyses

Analyses performed on site:

• HIV screening with a serial two rapid test algorithm as per World Health Organization (WHO) guidelines : Screening with Alere Determine combo®; if positive, confirmation by a second rapid test First Response® HIV 1-2-0, CD4 count and/or Viral load as per national guidelines.
• Sputum for GeneXpert Mycobacterium Tuberculosis (MTB)/rifampicine (RIF) assay
• Induced sputum for a new GeneXpert MTB assay if a routine spot and second morning GeneXpert MTB is negative and sputum quality was low (salivary) as per WHO consolidated guidelines
• Sputum for Gram stain and standard bacteriological culture
• Nasopharyngeal swab for Severe Acute Respiratory Syndrome Coronovirus (SARS-CoV-2) real-time polymerase chain reaction (RT-PCR) (Cepheid Xpert Xpress SARS-CoV-2) Since the additional diagnostic yield of mycobacterial culture is shown to be very low in adult patients with cough in Benin, it is reserved for GenXpert MTB/RIF resistant cases, specimens other than sputum (e.g. pleural effusions, where sensitivity of GeneXpert MTB/RIF is low) and patients with treatment failure. Mycobacterial culture will be done on solid medium (Löwenstein-Jensen).

Analyses performed retrospectively on stored samples:

• Nasopharyngeal swab: PCR for influenza
• Sputum: syndromic molecular panel targeting influenza and atypical bacteria. Radiological examinations
• Chest X-Ray (CXR) - face: digital format will be stored
• LUS exam using a standard point-of-care ultrasound device (ultrasound-on-a-chip), LUS will be recorded according to a standard 12-point acquisition protocol with 4 additional apical sites (2 subclavicular and 2 axillary) for improved TB detection.
• This involves scanning both apexes, the anterior and posterior apexes, anterior superior, anterior inferior, posterior superior, posterior inferior and lateral thorax regions
• FASH plus exam using a standard point-of-care ultrasound device. The FASH plus protocol will be performed in a standardized manner as previously described.

All US images captured will be digitally recorded and transferred via a secured internet connection along with relevant metadata to a secure server for storage. For study quality control purposes, the quality of the image and the interpretation of a random sample of images will be evaluated retrospectively by an experienced radiologist.

Clinician-evaluated LUS and FASH All LUS and FASH exams will be read by 2 independent readers blinded to patients diagnosis and outcome. A third reader will solve discrepancies between the two readers. They will fill in a standardized LUS and FASH report.

AI-evaluated LUS and FASH The US images will be further used for secondary studies developing deep learning algorithms for LUS and FASH diagnosis.

Follow-up (D7 and D28) Patients will receive a follow-up phone call by the study nurse on day 7 and day 28. Data on the clinical evolution are collected: hospitalization, non-invasive ventilation, intubation, death.

Follow-up of TB treatment failures and deaths until D180 TB patients will be followed in the national TB register and treatment failures will be recorded at 6 months of treatment (according to national guidelines).

Deaths will be recorded until D180 as well.

Inclusion Criteria:

• Age ≥18 years
• Suspicion of lower respiratory tract infection (defined as presence of cough with at least one of the following: fever, dyspnea/tachypnea, sputum production, chest pain, hemoptysis or cachexia)

Exclusion Criteria:

• Cough/dyspnea from a definite non-infectious origin (cardiac, asthmatic...)
• Inability to cooperate with ultrasound procedure
• Inability to sign informed consent (third witness procedure available for illiterate patients)

• Benin: 6'500 cumulative total detected and declared COVID-19 cases since march 2020 to date (NB true COVID-19 incidence unknown due to undertesting), TB incidence 55/100'000/year, low HIV (prevalence < 2%)
• South Africa: 1,522'697 cumulative total detected and declared COVID-19 cases since march 2020 to date TB incidence 615/100.000/year, high HIV (prevalence approximately 20%)
• Mali: 30'420 cumulative total detected and declared COVID-19 cases since march 2020 to date (NB true COVID-19 incidence unknown due to undertesting), TB incidence 52/100'000/year, low HIV (prevalence < 2%)