Introduction

Community-acquired pneumonia (CAP) is a leading cause of hospitalisation and mortality [1,2]. Antibiotic treatment should be initiated timely [3] to avoid serious complications such as bacteremia, sepsis, organ failure, and death [4]. Initial antimicrobial treatment is often empiric, and an uncertain or delayed diagnosis often leads to use of broad-spectrum antibiotics [5]. This, in turn, contributes to adverse effects and complications, such as Clostridioides difficile infection, super-infections with resistant bacteria, poor patient outcomes, and general development of antibiotic resistance [6–9]. In Denmark antimicrobial resistance is low, and almost all Streptococcus pneumoniae are susceptible to benzylpenicillin and 93% to erythromycin, and 75% of Haemophilus influenzae are susceptible to benzylpenicillin [10]. Danish guidelines recommend narrow-spectrum penicillin for empirical treatment of CAP with CURB-65 <3 and broad-spectrum antibiotics for severe CAP with CURB-65 ≥3 [11,12]. The CAP diagnosis is based on clinical symptoms such as cough, dyspnea, fever, and sputum production, combined with unspecific diagnostic tools such as auscultation of the lungs, chest radiography, blood tests, and microbiological analysis of sputum samples [13–15].

Sputum samples can be cultivated to determine bacterial agents; however, samples are often of poor quality, many patients cannot deliver a sample and laboratory turnaround time is typically 2 days [16,17]. The lack of precise, timely microbiological results may delay or hinder targeted antimicrobial treatment.

In addition, CAP is often caused by viral infections that can be treated without antibiotics but usually are indistinguishable from bacterial infections without specific microbiological tests [18–20]. Consequently, molecular diagnostic methods, including rapid polymerase chain reaction (PCR) panels for viruses and bacteria, have been developed and tested in clinical settings [21–23]. These panels are simple to use, sensitive, generate rapid results, and significantly contribute to the management of CAP [21,23,24].

By identifying pathogenic organisms earlier, studies have reported faster de-escalation of antibiotic treatment, reduced duration of broad-spectrum empirical antibiotic therapy, reduced length of stay (LOS), and reduced hospital costs [25,26]. However, evidence of clinical impact of point-of-care (POC)-PCR testing of sputum samples in EDs is limited and a recent feasibility study advocates the need for randomised controlled trials (RCTs) to test POC-PCR panels in acute settings [27].

In this multicentre, randomised study, we aimed to investigate the effect of adding POC-PCR to standard care in an emergency department (ED) setting. Our hypothesis was that POC-PCR testing of sputum samples from suspected CAP patients would increase the proportion of patients treated with no or narrow-spectrum antibiotics. The objectives were (i) to investigate the effect of POC-PCR testing of sputum from suspected CAP patients on the prescriptions of antibiotic treatment compared to usual care; and (ii) to investigate if the addition of POC-PCR testing to the diagnostic setup affects LOS, intensive care unit (ICU) admission, 30-days mortality, in-hospital mortality, or readmissions within 30 days.

Methods

Results

Patients admitted with suspected CAP were enrolled from March 1, 2021 to February 28, 2022. The last follow-up for mortality and readmission was on April 1, 2022. We screened 379 patients for eligibility and collected 294 (77.6%) LRT samples (78.4% tracheal secretions and 21.6% expectorated sputa) from patients who underwent randomisation. The 294 patients were allocated to either the POC-PCR group (148 patients (50.4%)) or the SCO group (146 patients (49.6%)), and those patients were included in the intention-to-treat analysis. Per protocol analyses for the primary outcome included 291 (99.0%) patients with no or narrow antibiotic treatment registered within 4 h (POC-PCR 145 (49.8%) and SCO 146 (50.2%)) (Fig 1).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Trial profile.

LRT, lower respiratory tract; POC-PCR, point-of-care polymerase chain reaction; SCO, standard care only.

https://doi.org/10.1371/journal.pmed.1004314.g001

Baseline data

Demographic and clinical characteristics are shown in Table 1.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Baseline characteristics of patients included in the analysis.

https://doi.org/10.1371/journal.pmed.1004314.t001

Number of patients prescribed “no or narrow,” targeted, and adequate antibiotic at 4 h, 48 h, and 5 days is presented in Table 2. Because of the observed difference in triage between the intervention and control, unadjusted and adjusted results are presented in Tables 3 and 4.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Absolute values for “no or narrow (no and narrow), targeted and adequate treatments” at 4 h, 48 h, and day 5.

Analyses of targeted and adequate treatment were based on 55 positive culture results from 290 patients.

https://doi.org/10.1371/journal.pmed.1004314.t002

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Unadjusted and adjusted per protocol analyses for the primary and secondary outcomes: Prescriptions of no or narrow, targeted, and adequate antibiotic treatment at 4 h, 48 h, and day 5.

The control group (SCO) is the reference. Analyses of targeted and adequate treatment were based on 55 positive culture results and routine PCR from 290 patients.

https://doi.org/10.1371/journal.pmed.1004314.t003

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 4. Adverse events and LOS for 291 patients.

https://doi.org/10.1371/journal.pmed.1004314.t004

Prescription of targeted and adequate antibiotics

Prespecified analysis of targeted antibiotic treatment and exploratory analyses of adequate antibiotics were based on positive culture results from 290 specimens after exclusion of one sample missing from the culture analysis. We identified 68 (23%) bacterial agents from 55 (19%) patients. Targeted treatment was used significantly more often in the POC-PCR compared with the SCO group at both 4 h and 48 h but not at day 5 (Table 3). Analysis of adequate treatment did not show a statistically significant difference between the groups at 4 h but more patients were treated with adequate antibiotics at 48 h and on day 5 in the POC-PCR compared to the SCO group (Table 3). A graphical presentation of changes in (2A) no or narrow, (2B) targeted, and (2C) adequate treatment for both groups is presented in Fig 2.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Changes in (2A) no or narrow-, (2B) targeted-, and (2C) adequate treatment prescription at 4 h, 48 h, and day 5.

Targeted and adequate treatments are based on culture results and routine PCR and include a sample of 55 patients. Results were adjusted for triage. POC-PCR, point-of-care polymerase chain reaction; SCO, standard care only.

https://doi.org/10.1371/journal.pmed.1004314.g002

Discussion

In this randomised study, adding sputum-POC-PCR to our diagnostic setup did not affect prescriptions of no or narrow-spectrum antibiotics during the first 2 days of admission, but less patients in the POC-PCR-group were treated with no or narrow-spectrum antibiotics after 5 days. Interestingly, patients in the POC-PCR-group were more likely to receive early targeted and adequate treatment. Number of readmissions, ICU admissions, and mortality were unchanged but we found a nonsignificant one-day reduction in LOS. Several prospective studies have reported sputum-POC-PCR as a method to support clinical decisions by fast and accurate detection of CAP pathogens [21–23,35]. Studies have shown a reduction in both use of intravenous antibiotics and number of days treated with antibiotics. In contrast to our study, most previous studies in ED settings only used panels for detecting upper respiratory pathogens [25,26,36,37].

Our outcomes were different focusing on type of antibiotic instead of length of treatment and route of administration, but nevertheless, the failure of POC-PCR to increase the use of no or narrow-spectrum antibiotics may seem to contrast these previous results.

There are some likely explanations. In an international context, the level of antimicrobial resistance is very low in Denmark and most pneumococci and H. influenzae are susceptible to penicillins [10]. Consequently, Danish treatment guidelines recommend relatively narrow-spectrum penicillins for CAP and reserve broad-spectrum antibiotics for severe pneumonia and/or sepsis [11,12]. This may have affected the study. For example, a patient with severe CAP may have been treated with penicillin instead of broad-spectrum antibiotics if POC-PCR detected pneumococci and another patient with mild CAP may have been treated with broad-spectrum antibiotics instead of penicillin if POC-PCR detected M. catarrhalis. Both actions were in agreement with the provided action card and both actions would result in a more targeted treatment—but also blur the effect of POC-PCR. This explanation is in line with the observation that patients in the POC-PCR group were more likely to receive early targeted and adequate treatment. In addition, the detection of Enterobacterales and Pseudomonas aeruginosa with POC-PCR may result in broad antimicrobial therapy even though they rarely cause CAP in a medical ED [38]. We excluded Enterobacterales from the analysis of targeted and adequate treatment due to the low incidence (1.3%) and because they usually represent colonisation [38].

Another possible explanation is the very low prevalence of common respiratory viruses in the study period related to the SARS-CoV-2 pandemic [39]. In other studies, virus accounted for 20% to 40% of CAP cases [19,20,37]. Some patients with CAP and a detected viral cause may be treated without antibiotics, and it is therefore possible that POC-PCR would have reduced the use of antibiotics in a period with a higher transmission of respiratory viruses.

The increased prescription of targeted and adequate antibiotics in the POC-PCR-group within the first 2 days is an interesting observation. It is based on analysis of a small subset of culture-positive samples; therefore, it is unknown if the result completely or in part can be extrapolated to the rest of the study population. Nevertheless, it highlights the question if POC-PCR improves patient outcome. We did not find any difference in mortality or transferal to ICU—but the number of events was very low. There was no difference in the number of readmissions but we did find a nonsignificant reduction in LOS from 4.3 to 3.6 days (p = 0.164) when adjusted for triage. It was not significant, but it might on the other hand reflect improved patient treatment and a possible reduction in LOS of almost 20% for one of the most common infections in the ED is very interesting from a hospital management and economic perspective.

At day 5, more patients in the SCO group were treated with no or narrow-spectrum antibiotics and there was no difference in the use of targeted antibiotics. This observation may be explained by routine microbiological results being available between day 2 and 5—allowing adjustment of treatment. Even though, we detected statistically significant differences they might be without clinical significance as they were quite small and day 5 is at the end of our recommended treatment duration. The strength of our study is the pragmatic multicentre, RCT design. The randomised design ensured that severity of illness, CAP diagnosis, and other patient characteristics were distributed equally between intervention and control group, and therefore causal inference is likely as the assumption of positivity is fulfilled. The POC-PCR analysis was integrated in the usual workflow in our ED suggesting that the test is technically feasible and easy to implement in other EDs. Project assistants were trained in collecting LRT-specimens and in using the POC-PCR platform and the primary investigator monitored the project closely to ensure a high level of internal validity. Almost 80% of the collected samples were tracheal secretions and this may have increased the reliability of the microbiological results by reducing upper airway contamination [40,41]. To ensure a uniform and correct clinical interpretation, we provided all POC-PCR results with a clear guideline-based action card.

There are also a number of limitations. Only few patients with CURB-65 scores ≥3 (14.4%) were included in the study. The inability to consent is likely linked to severe disease and acute cognitive impairment. In addition, restriction to weekdays and daytime may have reduced the number of severe cases as admission on weekends and at night are known to be associated with increased mortality and risk of referral to ICU [42]. Therefore, results can only be generalised to patients admitted on weekdays during daytime. In the secondary analysis of targeted and adequate treatment, only few culture-positive samples were included. The sensitivity of culture may be very low, and a high number of patients were treated with antibiotics before admission [43]. We could have circumvented this challenge by also analysing samples in the SCO group with FilmArray with a random disclosure design where results only are available in the intervention group. It would also allow subgroup analysis to investigate the effect of POC-PCR separately in test-positive and test-negative patients. It would straighten the results, leading to evidence-practice recommendations for implementing the test in clinical practice. However, it would be more expensive and may introduce ethical issues [44,45].

Both culture and POC-PCR may detect commensals, which was stated clearly in the provided action card. It is therefore possible that the clinicians in some situations chose to ignore the result—e.g., based on severity of illness, response to current treatment, fear of prescribing inadequate treatment, likelihood of commensal pathogen, and expected virulence of the pathogen [46]. We did not measure to what extent the action card recommendations were followed.

A possible interpretation of the overall results is that the current restrictive prescribing strategy in Denmark may be unable to provide targeted and adequate treatment for some patients. This may be overcome by introducing broad-spectrum empirical regimes—but that would fuel a further rise in resistance, may introduce side effects, and go against our general antimicrobial stewardship interventions. However, as indicated in this study, we might get around this problem by introducing fast and sensitive diagnostic methods. Future studies should focus on (i) the impact of POC-PCR on clinical outcome in a larger scale—e.g., LOS, length of treatment, and patient quality of life; (ii) hospitalisation costs; and (iii) the use of adequate and target treatment in a blinded setup where sensitive molecular methods are applied in both intervention and control groups. In conclusion, in this randomised trial introduction of POC-PCR did not increase the proportion of patients prescribed no or narrow-spectrum antibiotics but it might increase early treatment with adequate and targeted antibiotics and may be associated with a reduced LOS. The results apply to a setting with restrictive use of antibiotics and a very low level of antimicrobial resistance and may be quite different in other settings. Fast and accurate diagnostic tools may aid to maintain a restrictive use of antibiotics in the future.

Supporting information

Acknowledgments

The authors appreciated the text editing from Caroline Moos, research consultant at the University Hospital of Southern Denmark.