Introduction: Pneumonia is a leading cause of mortality and a common indication for antibiotic in elderly patients. However, its diagnosis is often inaccurate. The investigators aim to compare the diagnostic accuracy, the clinical and cost outcomes and the use of antibiotics associated with three imaging strategies in patients >65 years old with suspected pneumonia in the emergency room (ER): Chest-X ray (CXR, standard of care), low-dose CT scan (LDCT) or lung US (LUS).

Methods and analysis: This is a multicenter randomized superiority clinical trial with three parallel arms. Patients will be allocated in the ER to a diagnostic strategy based on either CXR, LDCT, or LUS. All three imaging modalities will be performed but the results of two of them will be masked during 5 days to the patients, the physicians in charge of the patients and the investigators according to random allocation. The primary objective is to compare the accuracy of LDCT vs CXR- based strategies. As secondary objectives, antibiotics prescription, clinical and cost outcomes will be compared, and the same analyses repeated to compare the LUS and CXR strategies. The reference diagnosis will be established a posteriori by a panel of experts. Based on a previous study, the investigatory expect an improvement of 16% of the accuracy of pneumonia diagnosis using LDCT instead of CXR. Under this assumption, and accounting for 10% of drop out, the enrolment of 495 patients is needed to prove the superiority of LDCT over CRX (alpha error =0.05, beta error=0.10).

Impact of the study: Superiority of the LDCT or LUS strategy over CXR would affect recommendations for the diagnosis of pneumonia in elderly patients. A higher accuracy of one of the strategies may decrease antibiotics overuse and lead to better outcomes and reduced costs.

This study will be conducted in three academic hospitals and one tertiary care hospital in Switzerland: Geneva University Hospitals, Geneva; Inselspital, Bern; Regional Hospital Lugano, Lugano; and Riviera Chablais Hospital, Rennaz.

Study design This is a multicenter randomized superiority clinical trial with three parallel arms aiming to compare the accuracy of imaging-based strategies for diagnosis of pneumonia in elderly patients admitted to the ER.

Each patient will be randomly allocated in the ER to one of the three imaging examination (CXR, LDCT or LUS), which will be immediately performed, interpreted by one of two independent radiologists (one for CXR and one for LDCT) or by an independent emergency physician trained in ultrasonography (LUS) and reported in a standardized form. The physician in charge of the patient will have access to the imaging examination and the corresponding report, in addition to usual clinical and biological data obtained in the diagnostic work-up of suspected pneumonia; he/she will be asked to assess the probability of pneumonia before the patient is discharged from the ER.

For each patient, the two other imaging examinations will also be performed and interpreted as described above, but the physician in charge of the patient will be blinded to these results. The results of all three imaging examinations and the interpretation will however be available to the panel of experts, whose final diagnosis of pneumonia will be the reference diagnosis for the study.

All enrolled patients will be followed up by study staff during hospitalization and by phone at month 1 and 3.

The sample size calculation was based on the accuracy of diagnostic strategies for pneumonia assessed in the PneumO-LD-CT cohort (unpublished results, table 1 bis in appendix). In this study, the accuracy of the clinician's diagnosis was 68% when based on CXR and 84% when based on LDCT.

With an expected improvement of 16% of the accuracy using LDCT instead of CXR, 150 patients will be required in each arm to demonstrate the superiority of LDCT over CXR with a two-sided alpha error of 0.05 and a power of 90%. Allowing for a 10% dropout after randomization, the final recruitment objective is 165 patients in each arm, for a total of 495 patients.

Patients aged >65 years consulting in the ER with suspected community acquired pneumonia (CAP) or nursing-home acquired pneumonia. Any patients admitted to the ER will be included in the study if eligible according pre-specified criteria. They may be referred by a doctor, an ambulance, a relative, or come on their own initiative.

At the ER of each recruiting site, dedicated research staff will screen admissions and ask suitable patients for participation in the study. In addition, triage nurses and physicians working at the ER will be asked to call research staff when identifying any potential participant.

Due to the high complexity of the inclusion process and of the simultaneous realisation of three imaging modalities, the patients will only be included during working hours.

The inclusions began on June 2021 and in the study is planned to last until August 2023. However, recruitment may be extended if necessary, in particular in view of the COVID-19 pandemic interfering with inclusions in times of high demands on emergency departments.

Patients will be randomised using Research Electronic Data Capture (REDCap®) tool, a secured web-based application designed to support data capture and randomisation for research studies. Randomization will be done immediately after inclusion, stratified by centre and using permutated block sizes.

Interventions The 3 images can be performed in random order and order will depend on the availability of the clinician practicing LUS. However, images will be obtained in the shortest time possible to avoid any significant impact on patient's care.

Standard of care: CXR will be done preferentially standing and with 2 incidences, which is the recommended and most commonly used diagnostic imaging modality for pneumonia in guidelines.5 6 LDCT scan will be obtained without administration of intra-venous contrast. Its performance lasts 10 minutes. Mean radiation exposure is 1.5 +/- 0.47 mSv, to be compared with a mean exposure of 0.05 +/- 0.03 mSv for conventional CXR, 7 mSv for a full-dose CT scan, and to Switzerland's natural background radiation level of 4 mSv/year.35 LDCT and CXR will be interpreted by two independent radiologists who will not be allowed to communicate.22 LUS will be performed at bedside by a trained physician not involved in management of the patient, using the device available in the corresponding ER (models and commercial brand may differ between different ER). All physicians performing the LUS (therafter: ultrasonographists) will be board-certified in the realization of Point-Of-Care Ultrasonography (POCUS). To enhance homogeneity of LUS reporting, physicians performing LUS will be trained to use the standardized report form before the beginning of inclusions, using a common protocol agreed upon by al site investigators. All examiners will have participated in a joint POCUS workshop to standardize the use of the study protocol and practice and their years of practice will be recorded.

Blinding/unblinding procedure The two masked radiological examinations will be concealed to the physicians caring for the patients during 5 days.

As soon as the randomization has been conducted, research staff will know which tests to blind. Those will be sent to a research PACS instead of the electronic health records with the help of the local IT team. Furthermore, the radiologists or sonographist who perform/interpret the tests will be asked not to communicate with each other. Research staff will be present throughout the process to ensure a smooth conduct of the study.

Emergency unblinding will be allowed in case of identification of an immediately life-threatening finding (table 2). In this occurence, the radiologist or ultrasonographist will make an emergent call to the investigator before communicating the results to the clinician.

The clinician in charge of the patient will be allowed to prescribe any new imaging deemed necessary in case of later clinical deterioration (eg. full dose chest CT scan with intravenous contrast for suspected pulmonary embolism).

At day 5, research staff will unmask all radiological examinations along with the standardized reports in the patient's medical record.

Reasons for emergency unblinding Pneumothorax Haemothorax Indirect signs of aortic dissection Indirect signs of aneurysmal rupture (haemomediastinum) Massive pericardial effusion Tracheal foreign body Pneumoperitoneum Pneumomediastinum Malignant airway obstruction Suspected acute tuberculosis

Reference diagnosis A panel of experts composed of senior clinicians and board-certified specialists, including internists, geriatricians, infectious diseases specialists and radiologists, blinded to the allocation arm and the probability of pneumonia estimated by the clinician in charge, will rate prospectively and a posteriori the probability of pneumonia. They will be trained before the adjudication process and asked to follow international guidelines for the diagnosis of pneumonia. They will have access to all available but de-identified patient data present in the medical records, including clinical data, biological, microbiological data -as results of PCR viral detection on naso-pharyngeal swabs- and images of CXR, LDCT, LUS and corresponding reports, hospital notes, and the final medical report. Each patient's diagnosis of pneumonia will be analysed using a Delphi method as follows: each expert will give an individual opinion on the probability of pneumonia on a 3-point Likert scale (low, intermediate, high). Next, each expert will re-examine the cases where there was a disagreement between expert ratings, in full knowledge of the other experts' first decisions. Finally, the adjudication committee will make consensus decisions in a plenary session and in the presence of a radiologist. The adjudication committee's final decision will be considered as the reference diagnosis.

Data safety monitoring board A safety analysis will be performed after 200 patients have completed the 1-month follow-up. Safety outcomes will be: unplanned transfer to the ICU and 1-month mortality. Based on this analysis, the data safety monitoring board can recommend to discontinue one arm or all arms of the trial. No interim analysis for futility or superiority is planned.

OBJECTIVES

Comparison between CXR- and LDCT-based diagnostic strategies for:

• Diagnostic performance, including accuracy (primary outcome), sensitivity, specificity, positive and negative predictive value and likelihood ratio
• Treatment and management (antibiotic prescription and additional imaging)
• Clinical outcomes (including length of stay, mortality, quality of life)
• Cost per patient:

Cost outcomes are defined as costs within the hospital calculated using a Swiss standard called REKOLE (https://rekole.hplus.ch/fr/produkt/rekole-comptabilite-analytique-a-lhopital/).

The main costs components are: nursing care, physician, imaging, laboratory, treatment (including antibiotic therapy) and others per patient during hospitalization, health related quality of life at 3 months; unit of work consumption per hospital (number of minutes of care, physician, laboratory and imaging points) up to 3 months.

Comparison between CXR- and LUS-based diagnostic strategies (secondary objectives) for the same outcomes as above Comparison between LDCT- and LUS based diagnostic strategies (secondary objectives) for the same outcomes as above

Other secondary objectives:

• Association between biomarkers and imaging-based diagnosis
• Factors of the physician (such as experience or gender), of the patient (such as urgency or presenting complaint) and of the context such as daytime affecting physician confidence in their diagnosis (CIRCUS substudy for Calibration of reasoning confidence in uncertain situations).
• Calibration of physician confidence with their actual diagnostic accuracy (CIRCUS)

STATISTICAL METHODS The investigators hypothesize that the diagnostic accuracy will be higher with the LDCT-based strategy than with CXR-based strategy. The null hypothesis that will be tested in the primary analysis is the equality of accuracy with these two strategies.

Population analysis The analysis will follow the intention to treat principle. A sensitivity analysis will be conducted on the per protocol population (i.e. excluding patients for whom an imaging or its standardized report which should have been masked has been available to the clinician who assesses the probability of pneumonia at the ER).

Primary analysis The proportions of correctly classified patients in the LDCT and CXR arms will be calculated with 95% Clopper-Pearson confidence intervals (CIs), and will be compared with a logistic regression model adjusted for sites to account for the stratified randomization. The statistical test will be two-sided and the significance threshold will be 0.05.

Handling of missing data and drop-outs Some patients will be discharged from the ED, but they won't be considered as dropout patients as the diagnosis of pneumonia will be made before the discharge, and follow-up will be done by phone. All available data from all included patients will be included in the intention-to-treat analysis whereas patients with missing data will be excluded from the complete case analysis. In addition, multiple imputation will be performed if more than 10% of data of the outcome are missing.

Inclusion criteria:

• Aged >65 years
• Suspected community-acquired or nursing-home acquired pneumonia consulting to the emergency room with at least one respiratory symptom (new or increasing among: cough, purulent sputum, pleuritic chest pain, dyspnea, respiratory rate >20/min, focal auscultatory findings or oxygen saturation <90% on room air) AND at least one symptom or laboratory finding compatible with an infection (temperature >37.8°C or <36.0°C, C reactive protein (CRP) >10 mg/L, PCT >0.25 µg/L, leukocyte count >10 G/L with >85% neutrophils or band forms)
• Signed informed consent
• In the oldest old (patients aged >80 years), the presence of acute delirium or unexplained acute fall can substitute for the presence of either the respiratory or the infectious symptom

Exclusion Criteria:

• Immediate admission to the intensive care unit (ICU)
• Pneumonia in the past 3 months
• PCR or antigenic test positive for SARS-CoV-2 in the 3 past weeks
• Transfer from another hospital with a diagnosis of pneumonia
• CXR or thoracic CT scan or US already done during the present episode
• Immediate contrast-enhanced CT scan needed
• Advanced care planning limiting therapy to comfort care only
• Prisoners
• Known uncontrolled psychiatric disorders
• Previous enrollment into the current study.

As described in the data management plan, data will be shared according to FAIR data principles and thus in a FAIR-compliant data repository.

After the end of the protocol

As described in the data management plan, data will be shared according to FAIR data principles and thus in a FAIR-compliant data repository

The care provider (clinician in charge) will be also be the outcome assessor and will know the allocation arm.