Study design/sample

After obtaining institutional review board approval, the investigators implemented a retrospective cohort study and enrolled a sample derived from the population of subjects who presented to a medical center for the evaluation and management of odontogenic infections between January 1, 2001 and December 31, 2013. To be included in the study sample, the subjects had to have one of the following discharge diagnoses, according to the International Classification of Diseases, Ninth Revision (ICD-9 codes): cellulitis (528.3), cellulitis and abscess of face/neck (682, 682.1), mediastinitis (519.2), neck swelling/mass (784.2), dental caries (521.09), and necrotizing fasciitis (728.86). Alternatively they could have a Current Procedural Terminology (CPT) code for debridement (11,040–11,044, 41,000–41,008, 41,015–41,018, 42,725, 97,597), intraoral/extraoral incision and drainage (21,501), sternal debridement (21,627), open treatment of sternum fracture with or without skeletal fixation (21,825), thoracoscopy (32,651–32,652), creation of pericardial window (33,025), mediastinotomy with exploration (39,000, 39,010), removal of devitalized tissue from wound(s) (97,602), and negative pressure therapy (97,605). A computer search of the electronic medical records was used to identify potential subjects for inclusion. Exclusion criteria were patient age <18 years, any non-odontogenic infection, and patient not pursuing treatment or for whom data were not complete.

Variables

The primary predictor variable was the LRINEC score. The LRINEC score was computed using the parameters described by Wong et al. using admission laboratory values for CRP, WBC count, Hb, sodium, creatinine, and blood glucose. For each patient, the LRINEC score was calculated as the sum of these six laboratory values as per Table 1 .

The primary outcome variable was whether or not the patient developed a NSTI and/or descending necrotizing mediastinitis. This was determined by searching for ICD codes 728.86 (necrotizing fasciitis) and 519.2 (mediastinitis), respectively. These diagnoses were confirmed clinically, operatively, and by biopsy of necrotic tissue. The secondary outcome variable was the total dollar amount billed to each subject derived from billing data. The data were collected in a Microsoft Excel spreadsheet (Microsoft, Redmond, WA, USA).

The other study variables were grouped into the following sets: demographic data, patient history, values on admission, presentation, evaluation, treatment, and outcome. The demographic variables were age (years), race (Caucasian, African American, or other), and sex (male or female). Weight (kg), height (cm), and body mass index (BMI, kg/m ² ) were also recorded. The patient history variables were history of diabetes (insulin-dependent and non-insulin-dependent), history of dental disease, and other medical/social history (cardiac disease, liver disease, hypertension, chronic obstructive pulmonary disease, asthma, HIV/AIDS, cancer, psychiatric disorder, tobacco use, intravenous drug use, alcohol abuse, and homelessness).

Age was determined by subtracting the date of birth in the demographic information from the date of admission. Sex and race were gathered from the patient demographic information. Weight, height, and BMI were collected from the nursing notes recorded on the day of admission. Medical and social histories were determined from the emergency department notes. Any history of dental disease specifically was determined from the emergency department notes or preoperative notes indicating decay of one or more teeth on admission.

The values recorded on admission were temperature (°C), CRP (mg/l), total WBC count (×10 ⁹ /l), Hb (g/dl), sodium (mmol/l), creatinine (mg/dl), and blood glucose (mg/dl). Temperature on admission was that recorded in the emergency department notes. Admission laboratory data for CRP, WBC, Hb, sodium, creatinine, and blood glucose were obtained from the laboratory results section of the patient’s chart.

The presentation variables were the presence of a draining wound (neck or oral), indication of gas on initial imaging, and whether there was airway compromise. Information on the use of imaging was obtained from the radiology section of the patient’s chart, and the assessment of gas on imaging was obtained from the radiology report and a review of the imaging. Airway compromise was determined by ‘airway involvement’ or ‘airway deviation’ on the radiology report. The evaluation variables were the number and type of imaging modalities used: computed tomography (CT) and magnetic resonance imaging (MRI).

The treatment variables were the treatment rendered, use of steroid treatment, type and route of antibiotics given, total number of operating room procedures within 48 h following imaging (including all washouts, which were done exclusively in the operating room), and whether surgery involved the mediastinum. Steroid and antibiotic use was determined from the inpatient medication list. The number and type of surgeries, including mediastinal involvement, was determined from the operative notes and discharge summaries. Comparison of imaging dates and procedure notes allowed the determination of which surgeries occurred within 48 h of imaging as per the imaging recommendations of Freeman et al.

All of the NSTI cases were treated with the same NSTI algorithm, as per the medical center protocol. This included quad antibiotic treatment (penicillin G, vancomycin, clindamycin, and levofloxacin or gentamicin ) until culture results were obtained, emergent debridement, serial washouts, and 48-h CT imaging.

The outcome variables were total days spent in the intensive care unit (ICU), total length of hospital stay, whether imaging was obtained within 48 h of discharge, any complications from the infection or treatment, and whether the patient died during the 90-day postoperative period. ICU use was determined from the discharge summary and inpatient notes. The discharge summary also indicated the total length of hospital stay. If the patient was not admitted overnight, the length of stay was recorded as 0. Imaging dates were compared to the discharge summary to determine whether or not imaging was obtained within 48 h of discharge. Additional information received from the billing department included type of medical coverage, actual payment, and remaining balance.

Data analyses

All data were abstracted from the subjects’ medical records and recorded using a spreadsheet (Microsoft Excel, Redmond, WA, USA). Descriptive statistics including the mean, frequency, range, and standard deviations were computed for each study variable. A Mann–Whitney U -test was used to compare the NSTI and odontogenic infection groups with respect to continuous variables. Fisher’s exact test was used to evaluate differences between groups with respect to categorical variables such as sex and race. To assess the prognostic value of the LRINEC score, diagnostic statistics were computed: sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV); this was done for the subset of patients for whom complete LRINEC laboratory parameters were obtained. To account for the small sample size, binomial proportion confidence intervals (CI) were constructed using Clopper–Pearson intervals rather than relying on the normal approximation to the binomial distribution. Statistical calculations were carried out in R version 3.2.2 (R Development Core Team, Vienna, Austria; www.R-project.org ). Statistical significance was set at P = 0.05.