The effect of focused lung ultrasound on antibiotic prescribing in patients with acute lower respiratory tract infections in Danish general practice (PLUS-FLUS): a pragmatic, randomised controlled, post-market trial

Summary

Background

Antimicrobial resistance is a growing global concern. Antibiotic use is the main driver and rational use is essential. General practitioners (GPs) prescribe most antibiotics for patients with acute lower respiratory tract infections (LRTI). This trial assessed whether focused lung ultrasound (FLUS) added to usual care in general practice could reduce antibiotic prescribing at the index consultation.

Methods

In this pragmatic, randomised, post-market, superiority trial, adults presenting with acute cough and at least one other symptom of acute LRTI, in whom the GP suspected pneumonia, were invited to participate. 34 GPs across 28 Danish office-based general practice clinics recruited participants. Patients were randomly assigned 1:1, stratified by GP, to receive either usual care (controls), reflecting current standard examination and care, or usual care supplemented with FLUS. The primary outcome was the proportion and relative risk (RR) of patients having antibiotics prescribed at the index consultation. Outcome assessors were blinded to group allocation. All patients were included in the intention-to-treat analyses. The trial was registered at ClinicalTrials.gov (NCT06210282).

Findings

Between November 1, 2023, and July 10, 2024, 390 patients were enrolled and analysed; 192 were assigned to the FLUS group. In the usual care group (n = 198), 81 patients (41%) were prescribed antibiotics at the index consultation compared with 69 (36%) in the FLUS group (n = 192; RR 0·88, 95% CI: 0·68–1·14). No evidence of harm was identified with the addition of FLUS to usual care.

Interpretation

Addition of FLUS did not lead to a 15% or larger reduction in antibiotic prescribing at the index consultation. Routine implementation of FLUS in this patient population is not supported by the findings. However, no evidence of harm was observed with the addition of FLUS.

Funding

The Health Foundation, The Foundation for General Practice, The Lung Association's Research Fund, and regional committees for Quality and Education.

Research in context.

Evidence before this study

We searched PubMed for articles published, in any language, between database inception and August 27, 2025, using the terms ((lung ultrasound) AND (general practice OR primary care)) AND ((community acquired pneumonia) OR (antibiotic)), yielding a total of 143 articles. Compared with chest x-ray, pathological focused lung ultrasound (FLUS) findings of a consolidation measuring >1 cm or a focal/asymmetrical B-lines pattern have shown a sensitivity of 87·8% and a specificity of 58·5% in one study in a general practice population. A practical algorithm incorporating chest x-ray to improve specificity has been proposed but may be unfeasible in many primary care settings. The effect of FLUS on antibiotic prescribing has only been evaluated once but in combination with procalcitonin point-of-care (POC) testing, in which it did not provide additional reduction in prescribing rates.

Added value of this study

The clinical impact of adding FLUS as a standalone POC examination on antibiotic prescribing in general practice was unexplored prior to this study. We show that our approach of adding FLUS, guided by clinical integration principles rather than a fixed diagnostic algorithm, to usual care management containing C-reactive protein POC testing, does not lead to a 15% or larger reduction in antibiotic prescribing. As this trial was not powered to detect a smaller reduction, more or larger studies are needed to detect a potential more modest effect of FLUS on antibiotics prescribing. However, FLUS was associated with fewer reconsultations for the same lower respiratory tract infection (LRTI) episode in general practice or out-of-hours primary care, possibly reflecting a reassurance effect for both GPs and patients regarding the initial diagnosis. This trial was conducted with a pragmatic design across a broad sample of general practices in Denmark, enhancing the generalisability of the findings. The near-complete follow-up and high diary completion rate further strengthen the validity of the results. Although the primary outcome did not reach statistical significance, the observed reduction in reconsultations suggests that FLUS may contribute to improved confidence in clinical decision-making.

Implications of all the available evidence

Current evidence does not support routine use of FLUS in all adults presenting in general practice with symptoms of acute lower respiratory tract infection where pneumonia is suspected. However, there is no evidence to suggest that FLUS poses a safety concern in this patient population. More evidence is needed to inform clinical integration strategies or diagnostic algorithms, that combine clinical assessment and FLUS findings to enhance specificity without increasing the risk of undertreatment.

Introduction

The threat of antimicrobial resistance is among the most pressing concerns for global public health. Antibiotic use is recognised as the main driver,¹ and prescribing in primary healthcare, particularly in general practice, is a major contributor.² Acute lower respiratory tract infections (LRTIs) are among the most common indications for antibiotic use in general practice,³ despite most being of viral origin and unlikely to benefit from antibiotic treatment.⁴ The diagnosis of pneumonia in general practice relies heavily on clinical assessment,⁵ and differentiating self-limiting viral infections from bacterial pneumonia is often challenging.^6,7 It is well known that this diagnostic uncertainty results in overuse of antibiotics.⁷

Point-of-care (POC) test strategies have been introduced to support general practitioners (GPs) in differentiating bacterial from viral infections. The C-reactive protein (CRP) POC test is the most widely used and reduces the likelihood of unnecessary antibiotic prescribing.⁸ Danish national guidelines recommend the CRP ‘speedometer’ model, in which the recommendation to prescribe antibiotics increases from CRP 20 mg/L towards values above 100 mg/L.^5,9 Within this CRP interval, the decision to prescribe antibiotics should primarily depend on clinical assessment. However, multiple CRP algorithms exist, and studies report variation in adherence.¹⁰ Chest x-ray is used in fewer than 10% of patients as a diagnostic aid for pneumonia in Danish general practice.¹¹ Thus, imaging is not routinely used in the diagnostic work-up of patients presenting with acute LRTI in Danish general practice. Consequently, strategies to reduce diagnostic uncertainty are needed to support GPs in adopting more restrictive antibiotic prescribing, without increasing the risk of delayed antibiotic prescribing. Focused lung ultrasound (FLUS) has gained increasing attention as a POC diagnostic tool in recent years. Numerous reviews and meta-analyses support FLUS as a valuable tool for diagnosing pneumonia with high diagnostic accuracy.12, 13, 14 However, most evidence stems from secondary care settings and hospitalised populations. Ultrasound is now increasingly available and used in general practices for various purposes.¹⁵ To date, general practice research on FLUS has focused on training GPs,16, 17, 18, 19 the diagnostic accuracy and feasibility of FLUS^20,21 and its effect on antibiotic prescribing when combined with procalcitonin POC testing.²² The clinical impact of adding FLUS as a standalone POC modality on antibiotic prescribing in general practice remains unexplored.²³

The aim of this randomised controlled trial was to evaluate the effect of adding FLUS to usual care on antibiotic prescribing in patients with acute LRTI in whom the GP suspected pneumonia. A potential reduction in antibiotic prescribing was not intended to result in undertreatment of bacterial infections; consequently, patients’ clinical course was closely monitored as a secondary objective.

Methods

Study design and participants

The PLUS-FLUS study was conducted as a pragmatic, randomised post-market (phase 5) superiority trial with a parallel two-group design, involving 34 GPs from 28 office-based general practice clinics across Denmark. GPs were involved as an advisory panel during the study design phase to ensure clinical relevance and feasibility. This trial is registered and statistical analysis plan available at ClinicalTrials.gov, NCT06210282.

Eligible patients were consecutively identified by the GP or assisting staff upon presentation to participating clinics. Adults (≥18 years) with acute cough (<21 days) and at least one additional symptom of acute LRTI (e.g. fever, sputum, dyspnoea, wheezing, chest discomfort), in whom the GP suspected bacterial pneumonia, were invited by the GP to participate. Patients were excluded if they: (1) had received antibiotic treatment for the current episode of acute LRTI, (2) were not listed with the GP (making medical record follow-up impossible), (3) were unable to understand or provide informed consent, or (4) declined to participate.

Specialists in family medicine, working as GPs and using ultrasound weekly in practice or out-of-hours primary care, were eligible to enrol patients. Completion of a FLUS training programme with a validated assessment was required; those already trained attended a 1-h brush-up session. Accordingly, a FLUS training programme was offered before patient enrolment. Details of GP recruitment strategy, timeline of enrolment, and description of the FLUS training programme offered, are found in the supplementary material.

All participants gave written informed consent, patients including consent for data extraction from medical and Shared Medication Records. Patients’ sex was recorded using the unique personal identification number assigned to all Danish citizens while GPs self-reported this as free-text.

Ethics

The study protocol²⁴ was approved by The Research Unit for General Practice in Aalborg (194-2) and the Scientific Ethics Committee of North Denmark Region (reference number: N-20230038). The study was conducted in accordance with the protocol without any adaptations. Written informed consent was obtained from all GPs and patients ahead of enrolment.

Randomisation and masking

Patients were randomly assigned in a 1:1 ratio to either usual care (control) or FLUS (intervention), where GPs performed and interpreted FLUS as an additional procedure during the index consultation (day 0). Randomisation was performed using a web-based system with permuted block sizes (2, 4, 6), stratified by GP for balanced enrolment. Allocation sequences and sealed envelopes were prepared by two affiliated researchers not otherwise involved in the trial. At randomisation, the GP opened the top envelope and group allocation was revealed to both the GP and the patient. Outcome assessors and data analysts were blinded to group allocation. until database lock and primary analysis completion.

Procedures

All patients received usual care, reflecting current standard diagnostic assessments and clinical management in Danish general practices. Usual care was recommended to follow national guidelines, which define pneumonia as a clinical diagnosis based on symptoms (acute cough and at least one other LRTI symptom) and clinical signs such as poor general condition and/or fever.^9,25 POC CRP testing was recommended. No clear recommendations were provided regarding referral for chest x-ray. FLUS was not considered part of usual care. Randomisation took place after usual care completion and before any treatment was commenced.

Patients randomised to the FLUS group received a FLUS examination in addition to usual care. GPs were instructed to follow a 14-zone FLUS scanning protocol previously applied across various settings,^26,27 including general practice in a similar patient population.¹⁹ The positioning of the patient during the examination was decided by the GP, either an upright or supine position or both. GPs used the ultrasound device available to them in their clinic. Types, brands, models and transducers are provided in the supplementary material.

GPs were trained to identify and classify FLUS pathological findings according to the European Federation of Societies for Ultrasound in Medicine and Biology (EFSUMB) coursebook,²⁸ the feasibility of which had previously been tested in general practice.¹⁹ See the supplementary material for the classification of pathological findings.

GPs were guided to interpret FLUS findings according to the following principles: Small subpleural consolidations (2–20 mm), multiple and bilateral, with pleural irregularities and bilateral B-lines typically suggest a viral infection. In contrast, larger unilateral consolidations with air bronchogram, and in some cases pleural effusion, often indicate bacterial infection. Depending on infection severity, focal multiple B-lines might be the only detectable sign of bacterial infection, potentially indicating early or mild disease; however, their diagnostic specificity remains limited.^29,30 Nevertheless, FLUS cannot detect infections confined to the deeper lung parenchyma that do not extend to the tissue directly beneath the visceral pleura. Based on this knowledge, the clinical integration of FLUS and its implications for patient management were left to the discretion of the GPs. The FLUS training programme and 1-h brush-up session included guidance on the interpretation of pathological findings and the principles for clinical integration as described above.

For each enrolled patient, the GP completed an electronic case report form in REDCap©³¹ (developed by Vanderbilt University and hosted by Aalborg University, Denmark) during or immediately after the index consultation. Information included patient baseline characteristics, symptoms, clinical findings, and any POC tests performed (e.g. CRP). The GP also registered in the case report form whether antibiotics were prescribed at the index consultation (yes/no) (primary outcome data), specifying if the prescription was immediate or wait-and-see. For patients allocated to the FLUS group, the GP documented FLUS findings.

During follow-up, GPs were permitted to reassess patients at any time if deemed necessary, and no concomitant care or diagnostic procedures were restricted. All follow-up data were likewise collected using REDCap©.

Patients received a validated LRTI Symptom Diary via short message service (SMS), covering the period from day 0 to day 21.³² The diary included items covering six dimensions: cough, dyspnoea, sputum production, general well-being, sleep disturbance and activity limitations. Patients were asked to score each of the six dimensions daily using a seven-point Likert scale (0 = no problem to 6 = as bad as it could be). The diary also contained daily questions regarding cancellation of work-related and leisure activities. On the day of the index consultation, patients were also asked to rate their satisfaction with the consultation, assessed on a five-point Likert scale (1 = very dissatisfied to 5 = very satisfied). A paper version of the diary was offered to patients who preferred not to receive the electronic format.

Data on changes in medication or new prescriptions across all healthcare sectors are automatically uploaded to a Danish national digital system, the Shared Medication Record. The principal investigator (PI) collected data on all antibiotics prescribed for systemic use (Anatomical Therapeutic Chemical Classification System [ATC] code J01) from day 0–28, regardless of indication, using the patients’ Shared Medication Records.

The PI collected patients' medical records on-site at general practice clinics, covering all documentation from general practice and correspondence from both primary and secondary care from day 0–60. An affiliated researcher, not otherwise involved in the trial, pseudonymised the records and concealed group allocation. The PI and a second affiliated researcher independently extracted relevant data using a pilot-tested data extraction tool. The datasets were compared, and discrepancies resolved through reassessment of the respective patient's medical record.

Outcomes

The primary outcome was the proportion of patients prescribed antibiotics at the index consultation (day 0), reported by the GP. The relative risk (RR) of antibiotic prescribing at the index consultation was also calculated to compare groups.

Secondary outcomes and prespecified exploratory outcomes (∗) from the LRTI symptom diary were: (1) The daily total LRTI symptom score, calculated as the sum of scores for six symptoms, with values ranging from 0 to 36. The scores were presented as mean or median from day 0 to day 21. (2) The number of days with symptoms rated as a 'moderate problem' or worse (at least one symptom received a score of 3 or higher), presented as mean or median days in the period from day 0 to day 21. (3) The number of days patients signed in sick or cancelled work-related activities, presented as mean or median days in the period from day 0 to day 21. (4) The number of days patients cancelled leisure activities, presented as mean or median days in the period from day 0 to day 21. (5) ∗Satisfaction with the index consultation on day 0 was measured by the proportion of patients who reported being satisfied or very satisfied (scores of 4 or 5) with the index consultation.

Secondary outcomes and prespecified exploratory outcomes (∗) from patients’ Shared Medication Records and regarding type of antibiotic prescription were: (1) The proportion of patients prescribed antibiotics within 7 days following the index consultation. (2) The proportion of patients prescribed antibiotics within 28 days following the index consultation. (3) ∗The proportion of prescriptions prescribed as delayed antibiotic prescription at the index consultation.

Secondary outcomes and prespecified exploratory outcomes (∗) from patients’ Electronic Medical Records were: (1) ∗The proportion of patients who had any primary care contact (defined as any contact with general practice or out-of-hours services) within 28 days after the index consultation. (2) ∗The proportion of patients who had any primary care contact (defined as any contact with general practice or out-of-hours services) for the LRTI episode within 28 days after the index consultation. (3) The proportion of patients admitted to the hospital within 28 days following the index consultation. (4) The proportion of patients with complications during hospitalisation (pleural infection (complicated parapneumonic effusion or empyema), lung abscess, or sepsis)) within 28 days after the index consultation. (5) ∗The proportion of patients who underwent any imaging other than FLUS in secondary health care services within 28 days after the index consultation. (6) ∗The types of other imaging modalities performed within 28 days after the index consultation. (7) ∗The proportion of patients referred for suspected cancer within 60 days after the index consultation. (8) ∗The proportion of patients diagnosed with cancer within 60 days after the index consultation. (9) The number of spontaneously reported unintended events by GPs up to 60 days after the index consultation. (10) The proportion of patients who died from any cause by day 28 and day 60.

Statistical analysis

A sample size of 340 patients, with 170 allocated to each arm, was calculated to provide 80% power at a 5% significance level to detect an absolute reduction in antibiotic prescribing from 50% in the usual care group to 35% in the  FLUS group. These assumptions were informed by previous studies on patients with LRTI in general practice.^7,11,33 We assumed a maximum dropout or discontinuation rate of 10% and increased the sample size by an additional 5% to adjust for covariates in the analyses. Consequently, we aimed to include a total of 390 patients (195 in each arm).

Baseline characteristics were tested for normality. Normally distributed data were presented as means with standard deviation. Non-normally distributed data were presented as medians with interquartile ranges. All patients were analysed according to the intention-to-treat principle. Mixed-effects Poisson regressions with robust variance estimation were used to estimate RRs with 95% confidence intervals (CIs). Chi-square test was used for comparison of proportions. A mixed-effects repeated-measures model was used for comparing daily symptom diary scores. Generalised linear models were used to estimate outcome means. Formal interaction tests were performed on predefined subgroups. All analyses accounted for clustering by the 34 recruiting GPs to adjust for variations in usual care. To address non-normality in outcome data, bootstrap resampling with 1000 iterations was applied. No data monitoring committee was established and no interim analyses performed. All analyses were conducted using STATA 17·0 (StataCorp. 2021. Stata: Release 17. Statistical Software. College Station, TX: StataCorp LLC.)

Role of funding source

The funders of the study had no role in study design, data collection, data analyses, interpretation, or writing of the report.

Results

Between November 1, 2023, and July 10, 2024, 523 patients were assessed for eligibility and 392 were enrolled, meeting the planned sample size. Of these, 199 patients were randomly assigned to the usual care group and 193 patients to FLUS group. One patient in each group was excluded after enrolment but before analysis due to being under 18 years of age. One patient in the usual care group received FLUS at the index consultation but was included in the intention-to-treat analysis, resulting in 390 patients analysed for the primary outcome (Fig. 1).

Fig. 1.

Trial profile. LRTI, Lower respiratory tract infection; GP, General practitioner; FLUS, Focused lung ultrasound.

Baseline characteristics were generally similar between groups (Table 1), apart from a higher prevalence of asthma in the FLUS group and more non-pulmonary comorbidities in the usual care group that could influence the indication for antibiotics. The median age of patients was 58 years (IQR:44–70), and 211 of 390 (54%) were female. The most common symptoms at the index consultation were sputum production in 78% (303/390), fever (temperature ≥38·0 °C) in 54% (209/390) and dyspnoea in 52% of patients (202/390). The most frequent clinical finding reported by GPs was abnormal lung auscultation in 41% of patients (158/390), and almost all patients (99% (388/390)) underwent CRP POC testing. The median CRP value was 16 mg/L (IQR:5·8–51·0).

Table 1.

Baseline characteristics of patients.

	FLUS (N = 192)		Usual care (N = 198)
Female, sexg	116	(60%)	95	(48%)
Age at randomisation (years)g	57	(42–69)	59	(46–71)
Comorbidities
None	109	(57%)	109	(55%)
Heart failure	3	(2%)	6	(3%)
COPD	15	(8%)	23	(12%)
Asthma	31	(16%)	20	(10%)
Other lung disease comorbiditiesa	3	(2%)	8	(4%)
Other relevant to FLUS findingsb	2	(1%)	7	(4%)
Other relevant to antibioticsc	7	(4%)	19	(10%)
Smoking status
Current smoker	33	(17%)	34	(17%)
Former smoker	65	(34%)	80	(41%)
Patient-reported symptoms
Sputum production	146	(76%)	157	(79%)
Dyspnoea	107	(56%)	95	(48%)
Fever (≥38.0 °C)	103	(54%)	106	(54%)
Temperature (°C)d	38·4	(38–39)	38·4	(38-38·9)
Chest discomfort	91	(47%)	82	(41%)
Poor general well-being	49	(26%)	39	(20%)
Other	11	(6%)	9	(5%)
Clinical findings
Abnormal lung auscultation	73	(38%)	85	(43%)
Sputum	55	(29%)	61	(31%)
Poor general appearance	34	(18%)	25	(13%)
None	7	(4%)	8	(4%)
Vital signs
Increased respiratory rate measurede	25	(13%)	16	(8%)
Respiratory rate (breaths/min.)d	23	(22–26)	22	(20–25)
O2 saturation performed	70	(37%)	61	(31%)
O2 saturation (%)df	97	(95–98)	97	(95–98)
Point-of-care testing
CRP poct performed	190	(99%)	198	(100%)
CRP level (mg/L)d	15	(5–48)	16	(7–53)
Leucocytes poct performed	20	(10%)	19	(10%)
Leucocytes (109/L)d	8·4	(7·4–9·6)	8·9	(7·5–11·0)

Values presented as n (%) or median (IQR).

FLUS, Focused lung ultrasound; COPD, Chronic obstructive pulmonary disease; poct, point-of-care test; CRP, C-reactive protein; IQR, Interquartile range.

^aInterstitial lung disease, lung cancer, lung tuberculosis and bronchiectasis.

^bComorbidities potentially affecting FLUS findings, as judged by the GP.

^cComorbidities potentially affecting antibiotic prescribing decisions, as judged by the GP.

^dValues are based on patients in whom the test was performed and result reported: Temperature (FLUS: 72 patients/usual care: 75 patients), Respiratory rate (19/14), O₂ saturation (70/61), CRP level (189/198), Leucocytes (19/18).

^eIncreased respiratory rate defined as ≥ 20 breaths/min.

^fWithout O₂ supplementation.

^gMedian age of the 131 excluded patients was 60 years and 56·5% were female.

For the primary outcome, 69 of 192 patients (36%) in the FLUS group were prescribed antibiotics at index consultation. In the usual care group, 81 of 198 patients (41%) were prescribed antibiotics, corresponding to a non-significant reduction in risk in the FLUS group, with a RR of 0·88 (95% CI: 0·68–1·14) (Table 2). Detailed subgroup analyses of the primary outcome for patients with CRP ≥50 mg/L, age ≥80 years, COPD or general lung disease comorbidities are presented in the supplementary material.

Table 2.

Primary and secondary outcomes: antibiotics prescribing, reconsultations, complications and imaging.

	FLUS (N = 192)		Usual care (N = 198)		RR (95% CI)
Antibiotics at index consultation	69/192	(36%)	81/198	(41%)	0·88 (0·68–1·14)
Wait-and-see prescriptions	4/69	(6%)	7/81	(9%)	0·65 (0·32–1·33)
Antibiotics within 7 days	81/192	(42%)	99/195	(50%)	0·83 (0·66–1·05)
Antibiotics within 28 days	88/192	(46%)	105/195	(54%)	0·85 (0·67–1·08)
Reconsultationsac	118/192	(62%)	133/195	(68%)	0·90 (0·73–1·11)
Reconsultations for LRTIac	79/192	(41%)	106/195	(54%)	0·76 (0·58–0·98)f
Admission to hospitala	11/192	(6%)	11/195	(6%)	1·02 (0·44–2·35)
Complicationsad	1/192	(0·5%)	1/195	(0·5%)	1·02 (0·06–16·8)
Other imagingae	26/192	(14%)	38/195	(20%)	0·69 (0·46–1·04)
Referred with suspicion of cancerb	5/191	(3%)	13/195	(6%)	0·39 (0·11–1·42)
Diagnosed with cancerb	0/191	(0%)	0/195	(0%)	1·00 (NA)
All-cause mortalityb	0/191	(0%)	0/195	(0%)	1·00 (NA)

Values presented as n/N (%).

FLUS, Focused lung ultrasound; RR, Relative risk; CI, Confidence interval.

^aWithin 28 days after the index consultation.

^bWithin 60 days after the index consultation.

^cReconsultations defined as any primary care contact (general practice or out-of-hour services).

^dComplications defined as pleural infection (complicated parapneumonic effusion or empyema), lung abscess or sepsis during hospital admission.

^eAny other imaging than FLUS performed in the secondary healthcare sector.

^fp = 0·038.

For secondary outcomes, follow-up data from Shared Medication Records and medical records were available for 387 patients. Secondary outcome results are presented in Table 2. In both groups, antibiotic prescriptions accumulated through day seven, with a higher proportion in the usual care group (99 of 195 patients, 50%) than in the FLUS group (81 of 192 patients, 42%). Still, the RR of having antibiotics prescribed up until day seven was not significantly lower in the FLUS group than in the usual care group (Table 2). A total of 18 patients were referred for suspected cancer within 60 days. Of these, five patients in the usual care group and one in the FLUS group were specifically referred for suspected lung cancer. The RR of patient reconsultation for the current LRTI episode in general practice or out-of-hour primary care within 28 days was significantly lower in the FLUS group (RR 0·76, CI: 0·58–0·98, p = 0·038), but no significant difference was found for reconsulting for any reason (Table 2). Chest x-ray was the most frequently used secondary imaging modality, performed in 19 of 192 patients (10%) in the FLUS group and 22 of 195 patients (11%) in the usual care group. Chest CT was performed in four of 192 patients (2%) in the FLUS group and six of 195 patients (3%) in the usual care group. Imaging unrelated to the lungs (e.g., other ultrasonography modalities, magnetic resonance imaging (MRI), CT or x-ray (not of the chest)) was performed in five patients in the FLUS group and 13 patients in the usual care group.

Of the 390 patients enrolled, 353 (91%) completed at least one day of the LRTI symptom diary: 183 in the FLUS group and 170 in the usual care group. The overall response rate was 81% with a mean of 18 (SD 5·7) days completed in the FLUS group and 17 (SD 6·5) days in the usual care group. No significant differences were observed between the two groups in LRTI symptom diary outcomes (Figs. 2 and 3). Patient satisfaction with the index consultation was high as 91% of responding patients in both groups reported being satisfied or very satisfied (137/149 (usual care) and 145/159 (FLUS)). No unintended events within 60 days were reported.

Fig. 2.

Lower respiratory tract infection symptom diary outcomes. Red shows the FLUS group. Blue shows the Usual care group. CI, Confidence Interval; FLUS, Focused lung ultrasound.

Fig. 3.

Daily total lower respiratory tract infection symptom score (0–36). Red shows the FLUS group. Blue shows the Usual care group. CI, Confidence Interval; FLUS, Focused lung ultrasound.

Significant pathological findings on FLUS were reported by GPs in 85 of the 192 (44%) patients randomised to the FLUS group. Among patients with pathological FLUS findings, 55 of 85 (65%) were prescribed antibiotics at the index consultation, compared with only 14 of 107 (13%) among those without pathological findings (unadjusted for confounders). The most common pathological findings in the 85 patients included focal ≥3 B-lines in 61 patients (72%), subpleural consolidation(s) (2–20 mm) in 17 (20%), pleural effusion in 11 (13%), larger consolidations in 9 (11%), interstitial syndrome in 5 (6%) and isolated visceral pleural pathology in 4 (5%).

Characteristics of the 34 GPs who enrolled study participants are shown in Table 3. Fourteen GPs had no or less than one year of experience using FLUS when recruited. Twenty-nine of the GPs completed the FLUS training programme offered before patient enrolment. Results from the FLUS training programme and comparisons between participating GPs and the national population of Danish GPs are presented in the supplementary material.

Table 3.

Characteristics of general practitioners enrolling patients in the trial.

	General practitioners (N = 34)
Age, years	48·4	(5·9)
Sex
Male	26	(76%)
Female	8	(24%)
Experience as a GP
< 5 years	5	(15%)
5–15 years	19	(56%)
>15 years	10	(29%)
Experience using POCUS
< 2 years	6	(18%)
2–5 years	19	(56%)
>5 years	9	(26%)
Experience using FLUS
0 years	14	(41%)
1–3 years	8	(24%)
>3 years	12	(35%)
Location in Denmarka
North Denmark Region	4	(12%)
Central Denmark Region	4	(12%)
Southern Denmark Region	22	(64%)
Region Zealand	1	(3%)
Capital Region	3	(9%)
Type of general practice clinic
Partnership Practice	27	(79%)
Solo Practice	7	(21%)
Baseline mean AB prescriptionsb	249	(83)

Values presented as n (%) or mean (SD).

GP, General Practitioner; POCUS:,Point of care ultrasound; FLUS, Focused lung ultrasound; AB, Antibiotics; SD, Standard deviation.

^aThe Capital Region is an urban area. The other regions are mixed rural and urban areas.

^bMean (SD) number of antibiotic prescriptions per 1000 patients listed in participating general practice clinics per year (one year before trial commencement, Nov 1, 2022–Oct 30, 2023.).

Discussion

Adding FLUS to usual care did not lead to the expected 15% reduction in antibiotic prescriptions at the index consultation. However, reconsultations related to the current LRTI episode in general practice or out-of-hours primary care were significantly reduced in the FLUS group. FLUS was not associated with any adverse effects on patient safety, symptom duration or severity or patients’ satisfaction with the consultation. However, the wide CIs observed for some secondary outcomes introduce uncertainty about the true RRs.

The reduction in antibiotic prescribing found in our study with the addition of FLUS to usual care was smaller than hypothesised and modest compared with other POC testing interventions evaluated for similar purposes in general practice settings.^8,22,33 Our study was powered to detect a 15% absolute reduction; therefore, it was underpowered to identify the 5% absolute reduction at the index consultation. Considering antibiotics prescribed during follow-up, the accumulated absolute reduction within the first week was 8% in the FLUS group, but this difference was also not statistically significant. At the index consultation, the antibiotic prescribing rate in the usual care group was 41%, which was lower than anticipated. This expectation had been based, among other sources, on a 2007 Danish study that reported a 56% prescribing rate in a comparable patient population.⁷ However, our findings did not indicate a stronger effect of FLUS on antibiotic prescribing within subgroups where prescribing rates were naturally higher (see the supplementary material).

The effect of adding FLUS into routine diagnostic procedures on antibiotic prescribing has previously been evaluated in a prospective observational study with 148 patients in French emergency departments.³⁴ Use of a 12-zone FLUS protocol led to changes in antibiotic management in 47 of 148 patients (32%), with 21 treatments initiated and 26 discontinued.³ Consistent with our study, this study showed no significant overall reduction in antibiotic prescribing. However, it did suggest a shift in the subset of patients to whom antibiotics were prescribed. Whether a similar pattern occurred in our study remains unclear. Nevertheless, any potential redistribution of antibiotic prescribing associated with FLUS did not translate into a clinically meaningful difference in patient outcomes, as the clinical course was comparable between groups.

As previously stated, this is the first randomised trial evaluating the effect of adding FLUS alone to usual care. To date, only one previous randomised trial has assessed the combined effect of the addition of sequential procalcitonin POC testing and FLUS to usual care on antibiotic prescribing in Swiss general practices.²² In that three-armed trial did the addition of FLUS at a predefined procalcitonin cut-off in a small patient sample not significantly reduce antibiotic prescribing compared with procalcitonin POC testing alone added to usual care. Usual care in the Swiss study did not include CRP testing, whereas nearly all patients (99%) in our study received a CRP POC test as part of usual care. Consequently, our study did not assess the impact of FLUS in isolation, but rather its effect when added to a standard care protocol that included CRP testing. While incorporating a study arm evaluating FLUS in the absence of CRP testing might have been informative and broadened the generalizability of our findings, such an approach would not have pragmatically reflected standard care in Danish general practice.

A key strength of this study is its pragmatic design, which closely mirrors routine clinical practice and the patient population typically encountered by GPs. This strengthens the study's external validity. The study was designed to minimize recall bias in the registrations by both GPs and patients. Furthermore, there was no indication of information bias in the GPs' registrations of antibiotic prescribing at the index consultation, as these data were cross-validated with entries in the Shared Medication Records by the PI for quality assurance. Comprehensive follow-up of the clinical course was achieved, with minimal loss to follow-up and a high response rate for the symptom diary. These features contributed to robust evidence regarding the impact of adding FLUS to usual care for the patient population studied.

However, several limitations should be acknowledged. First, participating GPs were self-selected and represented a subgroup (12%) of Danish GPs using POC ultrasound,³⁵ potentially limiting generalisability. Therefore, our results may not fully reflect average prescribing practices among Danish GPs. Baseline prescribing rates reported in Table 3 reflect practice-level data rather than individual prescribing habits, which were not available. The low baseline antibiotic prescribing and also the high uptake of CRP POC testing may potentially limit international applicability. Furthermore, participation in this trial may itself alter clinical and prescribing behaviour (Hawthorne effect), especially when GPs are unblinded to both study outcomes and allocation.³⁶ Another limitation is the absence of data on symptom duration before the index consultation. Symptom duration could have affected both clinical assessment and FLUS findings, as well as the GP's subsequent management decision.

This study reflects the impact of adding FLUS to usual care when GPs are guided by principles of clinical integration rather than a fixed diagnostic algorithm. Such an algorithm has, however, been proposed by Rodriguez-Contreras and colleagues.²⁰ However, their algorithm would require a chest x-ray in all cases of focal or asymmetrical B-lines pattern without consolidations. Based on previous research, we anticipated that such patterns would predominate in most patients with pathological findings.^19,20 Introducing this algorithm could lead to an increased use of chest x-ray, with associated increases in costs and patient radiation exposure in a Danish clinical context.¹¹ Consequently, we chose a pragmatic design in which the FLUS training programme and the brush-up session incorporated the best available knowledge on pathological findings and clinical interpretation.¹⁹ The results of our study probably reflect persistent uncertainties regarding the relationship between pathological FLUS findings, pathogen likelihood and disease severity progression in general practice. Whether a more conservative definition of pneumonia on FLUS (requiring consolidation) can be used in general practice, without compromising patient safety and clinical course, is unknown. As these relationships become clearer, the clinical utility of FLUS may improve. Notably, only 14 of the 107 patients in the FLUS group with no pathological FLUS findings were prescribed antibiotics. The underlying mechanism remains unclear. Possible explanations include that pathological FLUS findings aligned with pronounced symptoms supporting antibiotic use, or that a normal FLUS had a strong deterrent effect on prescribing decisions. These hypotheses fall outside the scope of this study but warrant further exploration.

Patients in the FLUS group had a lower likelihood of reconsulting general practice or out-of-hours primary care for the same LRTI episode within 28 days than patients receiving usual care. Any contact initiated by either the GP or the patient that related in any way to the current LRTI episode was classified as a reconsultation. Despite the uncertainties described above, this finding suggests a possible reassurance mechanism affecting both the GP and the patient regarding the initial diagnosis. Previous research supports this interpretation^19,37 as GPs reported increased diagnostic confidence in 69 of 91 cases (76%) after performing FLUS in a similar patient population.¹⁹

Future research should investigate if a more modest effect on antibiotic prescribing can be detected or if there is a larger potential of adding FLUS to usual care in settings with higher antibiotic prescribing rates. Human antibiotic consumption in Denmark is relatively low compared with that in other European countries,³⁸ and antibiotic prescribing rates in the usual care group were lower than anticipated. Moreover, longer-term studies are warranted to assess the safety of FLUS over time and to inform clinical referral guidelines for suspected pulmonary cancer based on FLUS findings in general practice. In our study, none of the patients referred with suspected cancer were diagnosed with cancer within the 60-day follow-up period; however, in a few cases, diagnostic clarification was still pending.

FLUS is not recommended as a routine tool for all adults with suspected pneumonia based on this trial's findings. However, the study found no evidence of patient harm associated with the addition of FLUS to usual care. Future studies should carefully evaluate whether FLUS may contribute to missed diagnoses of pulmonary cancer; however, our study indicates no safety concerns related to its use in this patient population.

Contributors

The study was conceptualised by JJS, MPH, CAA, MBJ, and JLT, with substantial scientific contributions from CBL, SHS and HHS. All authors agreed on the study design and practical considerations. Data collection, statistical analyses, and drafting of the initial manuscript were carried out by JJS. All authors contributed to and approved the final manuscript. Furthermore, all authors had full access to all study data and shared final responsibility for the decision to submit the manuscript for publication.

Data sharing statement

Deidentified individual participant data underlying the results reported in this article, including study registrations and informed consent forms, are available upon reasonable request. Upon approval of a data sharing request by the regional legal body, information necessary to address the proposed research question will be shared under the terms of a signed data sharing agreement. Requests should be directed to jrjepsen@dcm.aau.dk. The study protocol²⁴ is publicly available and the statistical analysis plan is accessible at clinicaltrials.gov (NCT06210282).

Declaration of interests

CBL coauthored the e-learning material used for the FLUS training programme, Basal Klinisk Ultralydsdiagnostik [Basic Clinical Ultrasound Diagnostics], published by Munksgaard (see the supplementary material) and receives royalties for the use of this material. CBL, HHLS and SHS received honoraria for lectures and education activities by Boehringer Ingelheim. CBL also received honoraria for education activities from AstraZeneca, GSK, Chiesi Pharma AB, and HHLS from Takeda. Congress travel was sponsored by Boehringer Ingelheim for HHLS. SHS is Head of the European Respiratory Society Group 14.3, Ultrasound. All other authors declare no competing interests.