Ventilator-associated pneumonia in neonates: the role of point of care lung ultrasound

Ventilator-associated pneumonia (VAP) is the second most common cause of antibiotic use on the neonatal intensive care unit (NICU) [1]. VAP is associated with higher incidence of bronchopulmonary dysplasia, prolonged mechanical ventilation and hospital stay [2].

Definition of VAP is challenging in neonates with no international consensus [3]. In adults, VAP is generally defined as a nosocomial lower airway infection in intubated patients with onset after more than 48 h of invasive mechanical ventilation [4]. In neonates, frequently applied definitions from Centres for Disease Control and Prevention (CDC) and European Centre for Disease Prevention and Control (ECDC) are difficult to apply mainly due to the absence of specific clinical and laboratory findings. Several modified criteria have been developed. However, comparison of agreement between four diagnostic criteria showed only moderate agreement [5].

CDC and ECDC criteria require chest X-ray changes as one of the main criteria for VAP diagnosis. However, interpreting radiographic changes in patients with underlying lung pathology and on mechanical ventilation can be challenging and comorbidities such as chronic lung disease (CLD) can obscure radiographic evidence of VAP [6, 7]. Reliability of chest X-ray interpretation has been shown to be low for neonatal VAP [5].

Lung ultrasound (LU) is increasingly used in neonates to diagnose lung pathologies [8‐11]. LU has been shown to be equivalent to chest X-ray for diagnosis of respiratory distress syndrome in infants [8, 9] and is highly accurate for the diagnosis of pneumonia in children [10, 11]. Multiparameter systematic approach for diagnosis and monitoring of VAP with clinical, laboratory, microbiological and LU data has recently been proposed in adults [12]. However, it has never been evaluated for diagnosis of VAP in neonates, especially in preterm infants with underlying CLD.

Our aim was to define clinical, laboratory, microbiological, chest X-ray and LU characteristics of VAP in our population and propose a multiparameter score for VAP diagnosis in neonates.

This quality improvement project was part of the pre-implementation phase of a bundle of care aiming at reducing the number of VAP episodes and antibiotics use in our level III NICU at the Evelina London Children’s Hospital, London (UK). Infants with suspected VAP were prospectively identified between March 25, 2018, and May 25, 2019, by the attending medical team led by the NICU consultant using the a modified version of the previously published “NICU Lucerne” definition [5] of ventilation for more than 48 h and new start or change of antibiotic therapy due to worsening ventilation requirements (increased FiO2 by > 20%, increased pCO₂ or ventilation demand) and at least one of the following: (i) clinical deterioration (temperature instability defined as axilla temperature of > 37.5 °C or < 36.5 °C that was deemed not to be due to environmental factors by the attending medical team, tachycardia, hypotension, increased frequency of episodes with bradycardias and/or desaturations), (ii) new or worsening opacity, consolidation or pleural effusion on chest X-ray, (iii) changes of tracheal secretions, (iv) abnormal laboratory parameters (C-reactive protein (CRP) > 10 mg/l, leucocytosis (white cell count > 20 × 10⁹/l) or leucopenia (white cell count < 5 × 10⁹/l)). Confirmed VAP was defined according to the above criteria plus isolation of a pathogenic microorganism in the airway aspirate.

Demographic, clinical, laboratory and microbiology data were collected from electronic patient records (BadgerNet®, Electronic Patient Records). Demographic data were collected on gestational age at birth and corrected age at VAP diagnosis, birth weight and actual weight. Clinical data included type of and reason for respiratory support, presence of respiratory deterioration, temperature instability within 24 h of VAP diagnosis and change in the amount and/or characteristics of airway secretions as recorded in the medical notes. Clinical data on risk factors known to be associated with VAP were also collected including number of intubation episodes, length of respiratory support with positive pressure ventilation and number of antibiotic courses. Laboratory data included CRP and white cell count at the time of VAP diagnosis and within the next 48 h. We collected semi-quantitative microbiology results and data on culture and sensitivity profile of pathogens that grew in the airway secretions.

We also applied the definition of Paediatric Ventilator-Associated Events (Ped-VAE), including Ventilator-Associated Condition (Ped-VAC), Infectious VAC (Ped-IVAC) and Possible Ventilator-Associated Pneumonia (Ped-PVAP) [13] to all cases to assess whether our definition for neonatal VAP were consistent with Ped-VAE, Ped-VAC, Ped-IVAC and/or Ped-PVAP.

Chest X-rays acquired as part of routine clinical care and stored on the Patient Archiving and Communication System were reported by the attending consultant radiologist. Reports were retrospectively reviewed by a neonatal consultant (VMP) to determine whether based on the CDC, ECDC, Lucerne and Dutch NICU definitions [14] of VAP, the following criteria were met: new emergence or worsening of consolidation, infiltration, pleural effusion or presence of cavitation or pneumatocele. Using the diagnosis on the radiological report, episodes were dichotomized based on whether the chest X-ray report was suggestive of VAP.

Additionally, LU scans were done within 24 h of VAP diagnosis as part of routine clinical care. LU scans were performed by trained members of the attending neonatal medical team using a GE Logiqe ultrasound machine (GE Healthcare®) with a L10–22 linear probe. The probe was held perpendicular to the ribs. For each hemithorax, three areas were defined by the anterior and posterior axillary lines (anterior, axillary and posterior). Anterior and posterior areas were divided into superior and inferior regions. At least nine standard views were acquired including right anterior apex and base, right axilla, right posterior apex and base, left anterior apex, left axilla, left posterior apex and base. A short clip was recorded in each view, downloaded anonymised and reviewed retrospectively offline by two neonatal consultants, NT and VMP with 1.5 and 6 years of experience respectively in performing and interpreting LU in neonates. In each view, the following parameters were assessed: aspect of the pleural line (presence of pleural sliding, thickness and regularity of the pleural line), presence of A-lines and B-lines, presence of consolidations and their characteristics (presence of static and/or dynamic air bronchograms, size and location in terms of translobar/non-translobar) and presence of pleural effusion. Translobar consolidation was defined as tissue-like image of the lung reflecting a loss of aeration with a regular inner limit (opposed visceral pleura). Non-translobar consolidation was defined as a tissue-like image limited by irregular “shredded” border as the non-aerated lung is in continuity with the aerated lung. When a consolidation was found, in some patients, it was assessed whether the consolidated area could be re-opened if indicated and possible by patient positioning and/or by increasing the mean airway pressure. For positioning, the consolidated area was positioned upward. For example, if a large consolidation was found in the posterior views, the patient was positioned prone for at least 1 h and the LU was repeated subsequently to assess re-opening of the consolidated area. Increasing mean airway pressure was achieved by stepwise increments of the positive end expiratory pressure and the peak inspiratory pressure, with close pressure-volume curve monitoring on the ventilator to detect over-distension [15].

Based on pathophysiological significance clinical, radiology and microbiology parameters [12] were identified and a scoring system for VAP diagnosis (multiparameter VAP score) was built post-hoc (Table 1). Clinical signs included respiratory deterioration, temperature instability and changes to airway secretions. In terms of radiological signs, for the purposes of scoring, on LU consolidations (> 0.5 cm depth), areas with linear/arborescent dynamic bronchograms and pleural effusion were included. The VAP score was calculated on Day 1 when antibiotics were started or changed and on Day 3 when microbiology result was available and antibiotic therapy was reviewed. To assess the clinical value of LU, the multiparameter VAP score was computed with and without LU findings on Day 1 and Day 3.

A group of infants who received positive pressure respiratory support for at least 48 h and underwent LU scanning because of respiratory deterioration but did not have a VAP diagnosis was identified. Data were similarly collected and VAP scores calculated.

Data was also collected on presence of chronic pulmonary insufficiency of prematurity (CPIP), defined as respiratory morbidity after preterm birth during the birth hospitalization and through infancy and childhood [16], and CLD, defined according to National Institute of Health Consensus definition [17] and death.

We demonstrated that adding LU to clinical information could be an alternative to chest X-Ray to diagnose VAP in neonates. As there is no widely accepted consensus regarding definition of neonatal VAP, we have also introduced a multiparameter score comprising clinical, laboratory, microbiological and LU data that showed high predictive values for VAP diagnosis in neonates with underlying lung pathology.

Chest X-Ray has been the gold standard imaging method for diagnosing chest infection in paediatric patients, and it is a mandatory criterion in the most often applied VAP definitions, such as the CDC guidelines for infants ≤ 1 year and the ECDC criteria [5]. However, challenges in interpreting radiographic changes in patients with underlying lung pathology, like infants with CLD have been previously reported [5‐7] and inter-rater reliability of chest X-ray interpretation was shown to be poor in this context [5]. In our series, only 64.5% of episodes met radiological criteria for VAP diagnosis based on chest X-ray. On the other hand, we found that LU had a high inter-rater reliability with only one disagreement between the two experts. Subpleural consolidation, consolidations > 1 cm, dynamic bronchogram, focal B-lines, pleural line abnormalities and pleural effusion have been shown to be characteristics of community-acquired pneumonia in hospitalized children with high specificity in suspected pneumonia [10, 11, 21, 22]. In adult VAP, subpleural consolidation and dynamic arborescent/linear air-bronchogram had high positive predictive values [23]. Of all the LU findings described in paediatric patients and adults, sub-pleural consolidations were also present in all our patients who did not have VAP and were thus not considered to be suggestive of VAP diagnosis. All patients in both VAP and no-VAP groups had diffuse lung rockets, related to the underlying lung pathology. We therefore excluded focal B-lines from our score. The LU findings that were suggestive of VAP in our patient population were consolidations > 0.5 cm accompanied by linear/arborescent dynamic bronchograms and/or associated with pleural effusion.

All LU in the VAP group showed large consolidations, which could either be atelectasis, collapse in dependant areas or VAP. Atelectasis could easily be recognized as a well limited area of consolidation with absent or very rare static bronchograms. Most challenging part of the assessment was to differentiate between areas of collapse and infection. Linear/arborescent dynamic bronchograms have been shown to be highly specific for pneumonia but identifying such changes in neonates with respiratory rates as high as 60/min and small tidal volumes could be challenging. Of note, identifying dynamic bronchograms was the only disagreement between the two experts in our cohort. Surprisingly, two infants in the no-VAP group had consolidations with linear dynamic bronchograms. Thus, specificity of linear/arborescent dynamic bronchograms in the collapsed dependant areas of infants with CLD needs further study. Punctiform dynamic bronchograms were identified in consolidations > 0.5 cm in both groups. However, these are not specific to pneumonia [23], rather a sign of de-recruitment. Recruitment of consolidated areas was another criterion to differentiate between collapsed areas and VAP. We hypothesised that when a consolidation cannot be completely recruited either by positioning of the infant or by recruitment airway manoeuvres, this would be suggestive of VAP. Most patients in the VAP group had consolidations that partially re-opened, changing from a consolidated area to white lungs following recruitment manoeuvres. Partial re-opening may be related to the presence of collapsed areas around areas of infection. All consolidations in the control group were possible to recruit.

LU to monitor recovery following VAP and efficacy of antibiotics was previously studied [24]. In our cohort, all patients (6/12) who had persisting consolidation suggestive of VAP at the end of antibiotic treatment (without any other clinical symptoms) had a relapse within the next month, whereas only 1/6 patients without LU findings suggestive of VAP at end of treatment relapsed. Performing LU at end of antibiotic treatment was not standard of care during this pre-implementation phase, explaining the small number of episodes where such assessment was performed. A LU score to quantify the global loss of aeration has been proposed in adult VAP patients. Its modification in days showed a significant correlation with re-aeration as assessed by quantitative CT scan (rho = 0.85) when computed after 7 days of antibiotics [25]. LU to monitor response to antibiotics treatment in neonates should be investigated further as optimal duration for treatment is currently unknown.

Despite LU being a promising tool to aid VAP diagnosis, similar to other neonatal respiratory pathologies [26], there is no single highly specific marker of VAP in neonates. Clinical and laboratory findings have been shown to be nonspecific [3, 25]. In our series, only 26% of infants had abnormal white cell count and only 41.8% had increased CRP. Other markers, such as procalcitonin have no positive impact on the accuracy of clinical and ultrasound-based scores for VAP diagnosis [27]. Microbiological samples cannot guide early decisions on starting antibiotics. A multiparameter systematic approach for diagnosis and monitoring of VAP with clinical, laboratory, microbiological and LU data is essential. A multiparameter score has recently shown to have greater specificity and sensitivity than clinical signs alone in adults [12]. To our knowledge, this is the first time that a multiparameter score combining clinical, microbiological and LU characteristics has been studied in neonates. Furthermore, sensitivity, specificity and AUC showed that predictive values increased when LU was included in the multiparameter VAP score.

Incidence of VAP in our NICU was 10.47/1000 ventilator days, which is in line with published incidence of 2.7–11.8/1000 ventilator days in developed countries [5]. Recent studies have reported lower incidence of VAP after implementation of VAP prevention bundles that combined antimicrobial stewardship and infection control measures, such as hand hygiene, isolation guidelines, handling and disinfection of patient care equipment, instruments and devices and use of personnel protective equipment [5, 28, 29]. This reduction had a significant impact on antibiotic use.

The main limitation of our cohort is its small sample size. As a result, it was not possible to undertake multivariate analysis adjusting for cofounders. Also, there is a mismatch between the size of the group of patients with and without VAP. We prospectively identified patients over a 14 months period during the pre-implementation phase of a VAP prevention care bundle. However, some potentially eligible patients might have been inevitably missed in context of non-trial settings. Furthermore, although LU is routinely performed on our NICU since 2018, experience and confidence of physicians, especially of those in training, performing LU varies; hence, LU was not available in every episode. Recruitment manoeuvre, although a dynamic and important element to explore further, was attempted only in a small number of episodes.

In summary, LU is a potential diagnostic tool in neonatal VAP. Compared with CXR, we found LU to be more useful to differentiate areas of atelectasis from areas of de-recruitment or areas of pneumonia, to assess the effect of positioning and recruitment manoeuvres on large consolidations and potentially to monitor evolution of consolidations following antibiotic treatment. However, this needs to be explored further in larger studies. While the proposed multiparameter VAP score, incorporating LU findings, showed promising predictive values for diagnosis of VAP in neonates with underlying CLD, it needs further testing in larger patient populations under prospective, multi-centre clinical trial settings.