Bacteriological analysis based on disease severity and clinical characteristics in patients with deep neck space abscess

Deep neck space abscess (DNSA) is defined as a collection of pus in the fascial planes and potential spaces of the neck, such as parapharyngeal, retropharyngeal, prevertebral, submandibular, carotid, visceral, and submental spaces, except for the peritonsillar space. Because of its acute onset, hidden location, and rapid progress, delayed treatment can result in life-threatening complications such as airway obstruction, descending mediastinitis, and sepsis [1]. Thus, DNSA has been recognized globally as a serious public health disease.

The administration of antibiotics is an essential part of the treatment strategy for patients with DNSA. Generally, antibiotics should be selected based on bacterial culture and susceptibility test results. However, according to previous reports, the positivity rate of bacterial culture is approximately 30%–50%, and at least 48 h are required to attain the results, indicating that most patients must rely on empiric antibiotic treatment within 48 h or even throughout the treatment course [2‐4]. Therefore, the selection of effective empiric antibiotics is a conundrum that has always plagued clinicians.

Considering that DNSA is typically a polymicrobial infection (e.g., coinfection with aerobic and anaerobic bacteria), most clinicians recommend the use of broad-spectrum antibiotics as empiric treatment [2, 5‐7]. However, using a narrow-spectrum antibiotic would reduce changes to the normal flora and adverse events, thereby limiting the overgrowth of resistant nosocomial organisms, avoiding the potential toxicity of multiple-drug regimens, and reducing costs, which is in line with the goals of the antimicrobial stewardship program (ASP) in China [8, 9]. But so far there are no guidelines in clinical settings on the use of empiric antibiotics before bacterial culture and drug susceptibility results are available. Bakir et al. suggested that empiric antibiotic treatments should be selected based on disease severity in patients with DNSA [10].

Therefore, the rational use of empiric antibiotics in DNSA patients must be considered in combination with the potential types of bacteria and disease severity. In this study, we established a DNSA clinical severity prediction model and further predicted the types of potential infecting bacteria (Gram-staining positive/negative) based on relevant clinical characteristics, aiming to provide clinical knowledge with certain guidelines for the rational selection of empiric antibiotics in clinical practice.

To the best of our knowledge, this is the first bacteriological analysis based on disease severity and clinical characteristics in patients with DNSA. We developed a new clinical prediction model based on disease severity in patients with DNSA, which demonstrated good discrimination and generalizability. This prediction model can help clinicians to early identify DNSA patients who are more likely to develop a severe illness. Moreover, we analyzed the distribution and predictive factors of the type of Gram-straining strains based on disease severity and clinical characteristics, and identified the existence of anaerobic infection through metagenomic sequencing analysis. These results can provide a new clinical basis for clinicians on how to reasonably select empiric antibiotics for patients with DNSA.

Prediction models for disease severity in patients with DNSA are lacking. In this study, a novel prediction model was constructed to assist clinicians in the early identification of patients who are likely to develop life-threatening complications. The model composed seven independent predictors that covered clinical symptoms and signs, imaging findings, and laboratory tests that can be used to systematically assess the severity of DNSA patients. Although previous studies have found that some inflammatory indicators are related to life-threatening complications in patients with DNSA, PLR is a new inflammatory marker that we first applied to the model for indicating disease severity [15‐20]. In addition to inflammatory indicators, we also found that the primary regions of infection were strongly associated with disease severity. The primary region of infection originating from the suprahyoid region can cause Ludwig’s angina, and abscesses originating from the retropharyngeal space can easily spread downward and cause complications with mediastinal abscesses, while congenital branchial cleft fistulae mostly originate from the infrahyoid region, which are more limited in deep neck space and involve fewer life-threatening complications. Although the cut-off point of the model is relatively low, we still need to be vigilant about patients at high risk for developing critical complications. We divided the patients into low- (mild) and high-risk (severe) groups based on the cut-off point. Thus, this prediction model can provide a novel evaluation reference model for clinicians to distinguish the disease severity in patients with DNSA.

Previous studies have explored the distribution of bacteria in patients with DNSA, but none of them have performed a stratified analysis based on the disease severity and clinical characteristics. In our study, the proportion of Gram-positive and Gram-negative strains in patients with mild disease was approximately the same. Among them, the most common Gram-positive strain was Streptococcus, and the most common Gram-negative strain was K. pneumoniae. Interestingly, only DM was associated with the type of Gram-staining strain, indicating that patients with mild DNSA and DM were mainly infected by K. pneumoniae, which is consistent with the findings of a previous study [21]. Therefore, third-generation cephalosporins, piperacillin, quinolones can be recommended as empiric antibiotics to treat patients with mild DNSA and DM (Gram-negative strains), whereas penicillin and first/second-generation cephalosporins can be recommended as an empiric antibiotic in patients with mild DNSA without DM (Gram-positive strains). The low resistance rates to these antibiotics also support these findings.

Notably, more than half (62.1%) of patients with severe disease were infected by Gram-negative strains, and the overall multi-drug resistance rate reached 12.1%, which indicated that broad-spectrum antibiotics in combination with beta-lactamase inhibitors would be more in line with the bacteriological characteristics in patients with severe disease. Additionally, it is interesting that trismus and gas formation was significantly correlated with the type of Gram-positive strain in patients with severe disease. The most common cause of trismus is tonsillogenic or odontogenic infection in patients with DNSA, and the main pathogens of both are Gram-positive strains (e.g., Streptococcus) [22‐24]. Gas formation is a sign of anaerobic pathogens (e.g., Clostridium perfringens, which is also a Gram-positive strain) and a higher complication rate [25, 26]. Poeschl et al. [27] reported that β-lactamase inhibitor combinations (amoxicillin-clavulanate) are more effective than penicillin for treating life-threatening complications such as mediastinitis, thoracic empyema, and septicemia. The isolates obtained from the severe disease group also showed a low resistance rate to amoxicillin-clavulanate, vancomycin, piperacillin-tazobactam, cefoperazone-sulbactam, and carbapenem. Therefore, according to the Gram reaction of strains mentioned above in patients with severe disease, amoxicillin-clavulanate or vancomycin can be selected as an empiric antibiotic in patients with severe disease exhibiting trismus and gas formation (Gram-positive strains), whereas piperacillin-tazobactam or cefoperazone-sulbactam can be selected as an empiric antibiotic for patients with severe disease who do not present the above symptoms (Gram-negative strains). According to the International Guidelines for the Management of Sepsis (2016), if the infection is poorly controlled and develops into sepsis, and the pathogen remains unidentified, the initial empiric antibiotics can be replaced with carbapenems [28]. In addition, it is worth noting that empiric antibiotic treatment should be deescalated to the narrowest effective agent once pathogen identification and sensitivities are established and/or adequate clinical improvement is noted.

Although bacterial culture is still the first-line detection method for identifying pathogens, the detection time required for metagenomic sequencing in our study was shorter (< 24 h) than that of bacterial culture, and the positive and accuracy rates were sufficient to outweigh the shortcomings of traditional methods. This advanced method can provide a reference for rapid pathogen diagnosis and allow the use of narrower-spectrum antibiotics in patients with DNSA. Moreover, metagenomic sequencing can also be used for subtyping the bacterial species and monitoring hospital outbreaks to support infection control and public health surveillance efforts [29]. Besides, the positivity rate of anaerobic culture was extremely low (9.5%) in both patients with mild and severe cases. This may be related to the harsh culture conditions used to grow anaerobic strains. However, the results of metagenomic sequencing showed that anaerobic strains were mixed in the pus, with a positivity rate as high as 83.3%. Many studies have confirmed that DNSA most commonly involves mixed infections of aerobic and anaerobic strains [2, 30‐32]. Therefore, treatment of anaerobic infections should be considered in patients with DNSA regardless of the disease severity.

This study had several limitations. First, DNSA is a relatively rare infectious disease; we observed a low positivity rate in bacterial culture, resulting in a relatively small sample size. Second, different regions across the country showed different distributions of strains and drug resistance, and the results may be biased for cases outside of the Pearl River Delta region in Guangdong Province. Finally, we did not analyze prognosis, as this study was focused on bacteriology. With the support of this bacteriological evidence, it can provide a scientific basis for us to carry out the clinical trial of antibiotic treatment in patients with DNSA in the next research phase.

In this study, we developed a prediction model and web-based calculator to estimate the risk of severe illness development in patients with DNSA. Estimating disease severity in patients with DNSA is a prerequisite for the rational selection of empiric antibiotics. The Gram reaction of the strain depends on the presence of DM in patients with mild DNSA, whereas in patients with severe DNSA, it mainly depends on the presence of trismus and gas formation. Multi-drug resistant strains were more common in patients with severe disease. Metagenomic sequencing can improve the positive detection rate of pathogenic bacteria and accurately detect the existence of anaerobic bacteria. Regardless of disease severity, it is necessary to consider the treatment of anaerobic infections. Empiric antibiotics must be modified after the results of bacterial culture and susceptibility tests are available.