Accepted Manuscript

Evidence-Based Clinical Criteria for Computed Tomography Imaging in Odontogenic

Infections

Brian J. Christensen, DDS, MD, Earl Peter Park, DMD, MD, Salvador Suau, MD,
David Beran, DO, MPH, Brett J. King, DDS

PII: S0278-2391(18)31092-9
DOI: 10.1016/j.joms.2018.09.022
Reference: YJOMS 58467

To appear in: Journal of Oral and Maxillofacial Surgery

Received Date: 3 August 2018

Revised Date: 13 September 2018
Accepted Date: 15 September 2018

Please cite this article as: Christensen BJ, Park EP, Suau S, Beran D, King BJ, Evidence-Based Clinical
Criteria for Computed Tomography Imaging in Odontogenic Infections, Journal of Oral and Maxillofacial
Surgery (2018), doi: https://doi.org/10.1016/j.joms.2018.09.022.

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to
our customers we are providing this early version of the manuscript. The manuscript will undergo
copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please
note that during the production process errors may be discovered which could affect the content, and all
legal disclaimers that apply to the journal pertain.
 ACCEPTED MANUSCRIPT

PT
RI
Evidence-Based Clinical Criteria for Computed
Tomography Imaging in Odontogenic Infections

SC
Brian J Christensen, DDS, MD1, Earl Peter Park, DMD, MD2, Salvador Suau, MD3, David

U
Beran, DO, MPH4, and Brett J King, DDS5*
1
AN
Chief Resident, Department of Oral & Maxillofacial Surgery, Louisiana State University Health
Sciences Center. 2Maxillofacial Oncologic and Reconstructive Surgery Fellow, Department of
Oral & Maxillofacial Surgery, University of Alabama at Birmingham. 3Assistant Professor and
Residency Program Director, Section of Emergency Medicine, Louisiana State University Health
M

Sciences Center. 4Assistant Professor, Section of Emergency Medicine, Louisiana State

University Health Sciences Center, and Medical Director of University Medical Center New
Orleans Department of Emergency Medicine. 5Assistant Professor, Department of Oral &
D

Maxillofacial Surgery, Louisiana State University Health Sciences Center.

Brian J Christensen – bchri2@lsuhsc.edu, E Peter Park – epark1@lsuhsc.edu, David Beran –
TE

dbera1@lsuhsc.edu, Salvador Suau – ssuau@lsuhsc.edu, Brett J King – bking6@lsuhsc.edu

* Corresponding Author: Dr. Brett King, Department of Oral and Maxillofacial Surgery,
Louisiana State University, 1100 Florida Ave Box 220 Room 5303, New Orleans, LA 70119,
EP

USA. Email bking6@lsuhsc.edu, Tel: (504)941-8212, Fax: (504)941-8215

This material has never been published and is not currently under evaluation in any other peer-
reviewed publications
C
AC
 ACCEPTED MANUSCRIPT

PURPOSE

Odontogenic infections are a common problem in emergency departments and impose a burden

on hospital budgets and provider time. Compounding this, is the lack of evidence guiding the

patient’s initial evaluation. The purpose of this study is to derive evidence-based guidelines for

PT
the use of computed tomography (CT) imaging in the management of odontogenic infections.

RI
METHODS

A prospective cohort study was designed. Patients with an odontogenic infection presenting to

SC
the emergency department from 11/1/2016 to 11/30/2017 were eligible for inclusion. The

outcome variable was need for CT imaging, which was based on the location of the abscess. The

U
potential predictor variables were demographics, history items, exam findings and laboratory
AN
values. The association between the outcome and predictor variables was determined using

classification and regression tree analysis as well as a standard logistic regression analysis.
M

RESULTS
D

There were 129 patients that met the inclusion criteria and consented for participation. The
TE

patients were 53.5% male and the mean age was 42.5 years. The most common fascial spaces

involved were vestibular (58.2%), submandibular (18.6%), pterygomandibular (6.2%), buccal

EP

(5.4%), and lateral pharyngeal (5.4%). The classification and regression tree analysis identified

mandibular inferior border blunting at the body as the best predictor for necessitating a CT scan
C

and identified a mouth opening of <25mm as the second best predictor. These two predictors had
AC

an accuracy of 96.9% for needing a CT scan. The logistic regression analysis identified these two

variables, as well as odynophagia, floor of mouth induration and white blood cell count as

significant predictors for needing CT imaging.

CONCLUSION
 ACCEPTED MANUSCRIPT

The two physical exam findings of mandibular inferior border blunting at the body and restricted

mouth opening were found to be highly associated with the need for CT imaging. Further studies

should be directed at validating these criteria in larger multi-center studies.

PT
RI
U SC
AN
M
D
TE
C EP
AC
 ACCEPTED MANUSCRIPT

INTRODUCTION

Odontogenic infections are a common problem in emergency departments across the country1-3.

These patients present with acute pain and swelling which can lead to airway compromise if not

treated promptly. However, they impose a burden on hospital budgets and provider time as both

PT
hospitals and clinicians are pressured to provide more cost-efficient care4, 5. Compounding this is

RI
the lack of evidence guiding the initial evaluation of these patients, which makes the job of the

emergency medicine (EM) physicians who triage these patients even more challenging. Current

SC
recommendations vary widely and are based on anecdotal experience, rather than evidence. As a

result, a recent study showed that EM physicians misjudge the clinical needs of a large portion of

U
patients presenting with odontogenic infections. Computed tomography (CT) imaging studies
AN
were ordered for over 60% of patients that did not require CT imaging and oral and maxillofacial

surgery (OMS) was consulted urgently for almost 80% of patient who did not require specialist
M

consultation6. The authors suggested that EM physicians were not able to determine the need for
D

CT imaging in patients with odontogenic infections; the study highlights the need for evidence-
TE

based guidelines to direct the initial triage of these patients.

The purpose of this study is to define evidence-based guidelines for the use of CT imaging in the
EP

management of odontogenic infections that cause facial swelling. These guidelines would ideally

allow for physicians and dentists in the acute care setting to quickly and accurately assess the
C

need for a CT scan with intravenous contrast. The investigators hypothesize there are a group of
AC

discrete symptoms, replicable physical exam findings, and basic laboratory values that predict

the involvement of fascial spaces associated with the need for a CT scan. The aims of the study

are to use a classification and regression tree analysis (CART) to identify the most important
 ACCEPTED MANUSCRIPT

predictor variables for the involvement of certain fascial spaces that conventionally define the

need for CT imaging in odontogenic infections.

METHODS

After obtaining institutional board review approval (IRB #9340), the authors began a prospective

PT
observational cohort study. The eligible patients included all patients presenting to the

RI
emergency department with facial swelling of suspected odontogenic origin over the dates

11/1/2016 to 11/30/2017 at University Medical Center in New Orleans, LA, an urban tertiary

SC
teaching hospital. The inclusion criteria were age greater than 18 years, presence of visible facial

swelling and confirmed odontogenic origin. Odontogenic origin was confirmed through presence

U
of carious lesion on physical exam at site chief complaint, presence of an antecedent tooth pain
AN
in the region of chief complaint and imaging (CT or panoramic) confirmation of the anatomic

relationship between a tooth and the swelling or periapical radiolucency. The exclusion criteria
M

were a recent surgical procedure or trauma near or at the site of the chief complaint, refusal to
D

consent to participate in the study or incomplete records as defined by the study protocol.
TE

The outcome variable analyzed was the need for CT scan. The involvement of particular fascial

spaces in the head and neck was used as a guideline for the need for CT scan. The fascial spaces
EP

considered to not need a CT scan were periodontal, palatal, vestibular, buccal and canine.

Therefore, the fascial space involvement that was considered to need CT imaging was the
C

submental, sublingual, submandibular, submassteric, pterygomandibular, superficial temporal,

AC

deep temporal, infratemporal, lateral pharyngeal and retropharyngeal spaces. The existing

practice patterns of CT imaging ordering by the EM physicians was not changed for this study. A

previous study at the author’s institution found that CT imaging was ordered for the vast

majority of patients and that only the most superficial locations (e.g. vestibular space) and
 ACCEPTED MANUSCRIPT

unequivocal diagnostic findings did not receive CT imaging6. The fascial space involvement was

determined by a combination of physical exam findings, imaging (if ordered) and findings at the

time of the incision and drainage. In order for a space to be considered involved, at least two of

the following must have been observed: at the time of drainage, a space contained a collection of

PT
fluid or gentle finger dissection was able to open the space, physical exam findings could clearly

RI
localize the abscess to a space, the space was involved on CT imaging.

Potential predictor variables analyzed are listed in Table 1 and included demographics, patient

SC
history items, focused physical exam findings and relevant laboratory values. The study protocol

was applied for patients meeting the inclusion criteria: The EM physicians examined the patients

U
and used an electronic data sheet template to record their history and physical exam. They then
AN
ordered a standard set of laboratory studies, if indicated, and consulted OMS. The OMS resident

on call interviewed and examined the patient with the same technique utilized by the emergency
M

medicine physicians. The OMS resident ensured the patient met the inclusion criteria. The
D

resident physician consented the patient for inclusion in the study after explaining the risks and
TE

benefits to participation. Because no interventions were a part of the study design, further care

was carried out in routine fashion.

EP

To organize data collection, a Microsoft Excel spreadsheet was created and used to record

information. Data were compiled and imported into JMP Statistical Discovery for analysis (JMP,
C

Version 13.0. SAS Institute Inc., Cary, NC, 1989-2017). Descriptive statistics were presented for
AC

predictor and outcome variables. Classification and regression tree (CART), a type of recursive

partition analysis, was used to identify the strongest predictors of the outcome variable. A

receiver operator characteristic curve was used to calculate the area under the curve for accuracy

evaluation of the model. Additionally, K-fold cross-validation with K=10 was used for model
 ACCEPTED MANUSCRIPT

validation. The CART method was used because it is well-suited for creation of clinical decision

rules and has been used in multiple medical specialties to create reliable scoring criteria and

decision rules7-11, including models for complications and mortality in aneurysmal subarachnoid

hemorrhage and acute decompensated heart failure. In addition to CART analysis, standard

PT
bivariate and multivariable analyses were performed. The relationships of the predictor variables

RI
to the outcome variable were analyzed using t-test, chi-square test, and ANOVA, as appropriate

for the type of data. A p-value of less than 0.05 was considered statistically significant. A

SC
preliminary multivariable logistic regression model was created by including all variables with

significant associations to the outcome variable. The final multivariable logistic regression model

U
was created using a backward elimination technique until all variables had p-values of less than
AN
0.05.

RESULTS
M

A total of 129 patients met the inclusion criteria and were enrolled in the study. The patients
D

were 53.5% male and the mean age was 42.5 ± 13.3 years. The most severe space involved in
TE

each patient was vestibular in 58.2%, submandibular in 18.6%, pterygomandibular in 6.2%,

buccal in 5.4%, lateral pharyngeal in 5.4%, canine in 2.3%, submasseteric in 2.3%, sublingual in
EP

0.8% and infratemporal in 0.8%. Descriptive statistics for the variables collected were reported

in Table 2. CT imaging was ordered for 82.1% of the study patients (106/129). CT imaging was
C

needed for 34.1% of patients (44/129) in the study cohort. The CART analysis identified
AC

blunting of the inferior border of the mandible at the body as the single best predictor for needing

a CT scan and identified a mouth opening of less than 25mm as the second best predictor. These

two predictors had an accuracy of 96.9% for needing a CT scan in this cohort. Of the 129

patients, 33 had inferior border blunting and 31 of these required a CT scan based on the location
 ACCEPTED MANUSCRIPT

of their abscess. Of the 96 patients that did not have inferior border blunting, 10 had a mouth

opening of less than 25mm and 9 of these required a CT scan based on the location of their

abscess. Four patients who did not have inferior border blunting at the body or a restricted mouth

opening were considered to have needed a CT scan based on the location of their abscess (Figure

PT
1). The four patients who were considered to need a CT scan by the location of their abscess but

RI
did not meet the criteria derived from the CART analysis shared few features except mandibular

tooth as a source and the size of the abscess and surrounding tissue effect was minimal. The K-

SC
fold cross validation average R2 value was 0.600 and the overall R2 was 0.672. Based on the

CART analysis, a suggested clinical guideline is depicted in Figure 2. Using the location of the

U
abscess as described above as the “gold standard,” sensitivity, specificity, positive predictive
AN
value and negative predictive value were calculated for the protocol guidelines. The sensitivity

was 90.9%, specificity was 92.9%, positive predictive value was 87.0% and negative predictive
M

value was 95.2%.

D

The predictor variables were also individually compared to the outcome variable as shown in
TE

Table 3. The predictor variables that were statistically significantly associated with the outcome

variables were location of swelling, antibiotic use over the preceding 2 weeks, dysphagia,
EP

odynophagia, lack of periorbital swelling, tongue range of motion, blunting of the inferior border

of the mandible at the angle, blunting of the inferior border of the mandible at the body, floor of
C

mouth elevation, floor of mouth induration, reduced maximum incisal opening, elevated white
AC

blood cell count, and elevated C-reactive protein. These variables were entered into a

preliminary multivariable logistic regression model. The final multivariable logistic regression

model was created as described above and is shown in Table 4. Blunting of the inferior border at

the body, maximum incisal opening, floor of mouth induration, odynophagia and white blood
 ACCEPTED MANUSCRIPT

cell count were the statistically significant predictors of the need for CT imaging in the final

model. Age and gender were not included in the final model because the unadjusted and adjusted

models did not significantly alter the results. For the CART analysis, the area under the receiver

operating characteristic curve was 0.939. For the final multivariable logistic regression analysis,

PT
the area under the receiver operating characteristic curve was 0.982.

RI
DISCUSSION

The purpose of this study was to define evidence-based guidelines for obtaining CT imaging for

SC
odontogenic infections that cause facial swelling. The aim of the study was to use a classification

and regression tree analysis to identify the most important predictor variables for the

U
involvement of certain fascial spaces that conventionally necessitate CT imaging. The CART
AN
method identified two variables that appear to be extremely important in predicting the outcome

variable: the need for CT imaging. A maximum incisal opening of less than 25mm and blunting
M

of the inferior border of the mandible at the body have both been long-recognized as important
D

hallmarks of more serious odontogenic infections12. It comes as no surprise that these predictors
TE

are the most closely associated with the group of diagnoses commonly designated as needing CT

imaging.
EP

In addition to the CART analysis, a standard bivariate analysis and logistic regression analysis

was performed. The multivariable logistic regression analysis identified the same two variables
C

as the CART analysis did for the best predictors of the need for CT imaging. However, it did
AC

identify three additional variables associated with the need for CT imaging: odynophagia, floor

of mouth induration and leukocytosis, though their effect on predicting the outcome variable was

much smaller. Overall, the multivariable regression analysis was mildly more accurate than the

CART model based on the area under the receiving operating characteristic curves (CART
 ACCEPTED MANUSCRIPT

model 93.9% and logistic regression model 98.2%). However, the CART model identifies the

variables in a way that is much easier to use to develop a set of clinical criteria. Based on these

models, the clinical algorithm in Figure 2 is suggested as a method for determining the need for

CT imaging of patients with odontogenic infections that present to the emergency department

PT
with facial swelling. However, it should be noted that in this study population, 3.1% of the

RI
patients who needed a CT scan according to the definition used in this paper would not have

received one if the suggested criteria were strictly applied. When re-examining these four

SC
patients, it is clear they had small abscesses with minimal surrounding inflammation. However, it

is unknown what their outcome would be if no CT imaging was obtained but instead a panoramic

U
film taken and a tooth extraction without an extraoral incision and drainage. The potential
AN
disadvantages of obtaining unnecessary CT images include an increased dose of ionizing

radiation, delay in care and exposure to nephrotoxic intravenous contrast agents as well as
M

increased diagnostic costs. Likewise, the potential disadvantage of not obtaining necessary CT
D

images include a delay in diagnosis, a return visit to the emergency department or even airway
TE

compromise outside of immediate access to medical personnel. Therefore, it is extremely

important that guidelines for CT images be associated with small amounts of error.
EP

This study is the first to examine the relationship of clinical predictor variables to a diagnostic

outcome variable in a prospective fashion in an attempt to define clinical guidelines for

C

odontogenic infections. The need for this clinical criteria was made clear by a recent
AC

retrospective study which notes that EM physicians were unable to accurately judge the need for

CT imaging and specialist consultation6. However, it is far from the first to use CART analyses

to define clinical guidelines. The method has been used to create scoring criteria for S. aureus
 ACCEPTED MANUSCRIPT

bacteremia, congestive heart failure, acute pancreatitis and aneurysmal subarachnoid hemorrhage

as well as for guidelines in many other areas of medicine7, 9-11.

The study has several limitations. While it is the largest prospective study on odontogenic

infections performed to the authors’ knowledge, the sample size of 129 is still limiting. There are

PT
a number of rare infection patterns that are not well-represented in our study, such as orbital

RI
involvement, temporal space involvement, mediastinal involvement and other rare but serious

conditions. There were also a relatively small number of lateral pharyngeal abscesses. Although

SC
it is not possible to develop a lateral pharyngeal space abscess from a tooth without first passing

through the submandibular or pterygomandibular space, it is possible that the major component

U
of an infection could be in the lateral pharyngeal space with minimal swelling in the
AN
submandibular or pterygomandibular spaces. In this scenario, it is possible the criteria developed

in this study would not correctly place this type of infection in the “requiring CT” category.
M

Additionally, if the sample size was larger, it would permit the use of a derivation and validation
D

cohort instead of the K-fold cross-validation technique. An additional limitation is the selection
TE

bias inherent in asking a busy OMS resident to enroll a patient in an observational study,

sometimes in the early morning hours. While attempts such as frequent reminders were made to
EP

minimize these missed data points, it was not possible to eliminate them completely and it is not

possible to determine their exact frequency. It is possible that these patients have a different set
C

of predictor variables that would alter the interpretation of the data, although it is unlikely.
AC

Another limitation is the definition we chose for the infections requiring CT imaging. While we

chose this outcome variable for ease of application of the study findings, the outcome variable is

based on the grouping of fascial involvement into “serious” and “less serious” groups based on

their potential to spread and compromise the airway. Most clinicians would agree with this
 ACCEPTED MANUSCRIPT

definition of the need for CT imaging, however, not all would. The application of these clinical

criteria for CT imaging does have a potential for error, as discussed above. An additional

problem is the repeatability of these physical exam findings. The instructions given to the patient

on how to open for an maximum incisal opening measurement can make a drastic difference in

PT
the measurement. Likewise, blunting of the inferior border of the mandible is not a physical

RI
exam maneuver that is self-explanatory or commonly taught in medical schools. For providers

that have never seen a submandibular abscess, it may be difficult to understand. However, these

SC
discrepancies should be able to be overcome by educational opportunities, in the form of lectures

or one-on-one explanations.

CONCLUSION
U
AN
The present study demonstrated that through the use of a CART analysis and a multivariable

logistic regression analysis, two physical exam findings were found to be highly associated with
M

the need for CT imaging: mandibular inferior border blunting at the body and maximum incisal
D

opening. A simple clinical criteria was developed based on these results; if a patient with facial
TE

swelling from an odontogenic infection presents to the emergency department, a physical exam

finding of mandibular inferior border blunting at the body or trismus should be considered an
EP

indication for CT imaging. Further studies should be directed at the validation of these criteria in

larger multi-center studies.

C
AC
 ACCEPTED MANUSCRIPT

PT
RI
U SC
Variables
AN
Age Gender Days since beginning of Location of swelling
swelling
Location of tooth pain Rate of progression of Prior antibiotic treatment Use of analgesics
M

swelling
Dysphagia Odynophagia Dyspnea Medical comorbidities
Substance use disorders Tobacco use disorders Immunocompromising Triage heart rate
conditions
D

Triage systolic blood Triage diastolic blood Triage temperature Triage pain score
pressure pressure
TE

Tongue range of motion Maximum incisal opening Blunting of inferior border of Tooth etiology
mandible
Number of carious teeth Floor of mouth elevation Floor of mouth induration Periorbital swelling
EP

Orbital proptosis Uvular deviation White blood cell count C-Reactive protein level
Creatinine
Table 1. Potential Predictor Variables Included.
C
AC
 ACCEPTED MANUSCRIPT

Selected Study Variables N = 129

Demographic
Age 42.5 ± 13.3
Gender (Male) 53.5%
Patient History

PT
Location of Swelling
Upper 31.0%
Lower 69.0%

RI
Progression of Swelling
Increasing 88.4%
Stable 7.0%

SC
Decreasing 4.6%
Antibiotic Use in 2 Weeks Prior 47.2%
Diabetic 13.2%

U
Immunocompromising Diseases 16.3%
Smokers 51.9%
AN
Smoking (packs per day) 0.39 ± 0.46
Physical Exam Findings
Systolic Blood Pressure (mm Hg) 136.6 ± 19.6
Diastolic Blood Pressure (mm Hg) 82.2 ± 12.6
M

Heart Rate (beats per minute) 87.9 ± 17.6

Respiratory Rate (breaths per minute) 17.2 ± 1.7
Temperature (degrees Fahrenheit) 98.8 ± 1.1
D

Pain (0-10 scale) 7.7 ± 2.6

Periorbital Swelling 17.1%
TE

Uvula Deviation 2.4%

Blunting of Inferior Border of Mandible 25.6%
at Body
EP

Blunting of Inferior Border of Mandible 10.1%

at Angle
Number of Carious Teeth 2.0 ± 2.1
Floor of Mouth Elevation 7.2%
C

Floor of Mouth Induration 6.4%

Maximum Incisal Opening (mm) 34.1 ± 12.5
AC

Laboratory Values
White Blood Cell Count (1000 cells/µL) 11.4 ± 4.3
C-Reactive Protein (mg/dL) 9.6 ± 5.8
Creatinine (mg/dL) 0.98 ± 1.3
Diagnosis
Tooth Etiology
Maxillary Incisor 6.2%
Maxillary Canine 11.6%
Maxillary Premolar 8.5%
 ACCEPTED MANUSCRIPT

Maxillary First Molar 2.3%

Maxillary Second Molar 1.6%
Maxillary Third Molar 0.8%
Mandibular Incisor 0.0%
Mandibular Canine 2.3%
Mandibular Premolar 10.1%
Mandibular First Molar 21.7%

PT
Mandibular Second Molar 16.3%
Mandibular Third Molar 18.6%
Number of Spaces 1.3 ± 0.7

RI
Table 2. Descriptive Statistics for Study Population.

U SC
AN
M
D
TE
C EP
AC
 ACCEPTED MANUSCRIPT

Study Variable Needed CT Did Not Need Mean Likelihood Odds 95% CI of p Value
Imaging CT Imaging Estimate and Ratio the Odds
N = 44 N=85 Standard Error Ratio
(34.1%) (65.9%)
Age 39.9 43.9 -0.024 ± 0.015 0.97 0.95-1.00 0.097

PT
Gender -0.245 ± 0.187 0.61 0.29-1.27 0.188
Male 29.0% 71.0%
Female 40.0% 60.0%

RI
Duration of Swelling (days) 4.1 3.2 0.130 ± 0.071 1.14 0.99-1.31 0.060
Location of Swelling n/a n/a n/a <0.0001
Upper 0.0% 100.0%

SC
Lower 49.4% 50.6%
Antibiotic Use in 2 Weeks 0.439 ± 0.191 2.41 1.14-5.09 0.020
Prior
Yes 45.0% 55.0%

U
No 25.4% 74.6%
AN
Dyspnea 0.341 ± 0.509 1.98 0.27-14.53 0.507
Yes 50.0% 50.0%
No 33.6% 66.4%
Dysphagia 0.901 ± 0.242 6.06 2.35-15.64 <0.0001
M

Yes 68.0% 32.0%

No 26.0% 74.0%
Odynophagia 1.513 ± 0.246 20.62 7.85-54.17 <0.0001
D

Yes 79.0% 21.0%

No 15.4% 84.6%
TE

Diabetes Mellitus -0.124 ± 0.284 0.78 0.26-2.37 0.658

Yes 29.4% 70.6%
No 34.8% 65.2%
Systolic Blood Pressure 133.3 138.3 -0.013 ± 0.010 0.99 0.97-1.01 0.161
EP

(mm Hg)
Heart Rate (beats per 89.2 87.2 0.006 ± 0.010 1.01 0.99-1.03 0.553
minute)
C

Respiratory Rate (breaths 16.9 17.4 -0.203 ± 0.118 0.82 0.65-1.03 0.077
per minute)
AC

Temperature (degrees 99.0 98.7 0.236 ± 0.180 1.27 0.89-1.80 0.184

Fahrenheit)
Periorbital swelling -1.323 ± 0.521 0.07 0.01-0.55 0.0003
Yes 4.6% 95.4%
No 40.2% 59.8%
Tongue Range of Motion 1.725 ± 0.531 31.5 3.93-252.20 <0.0001
Abnormal 92.3% 7.7%
Normal 27.6% 72.4%
Maximum Incisal Opening 24.0 39.4 -0.144 ± 0.027 0.87 0.821-0.913 <0.0001
 ACCEPTED MANUSCRIPT

(mm)
Blunting of Inferior Border 1.725 ± 0.531 31.5 3.93-252.20 <0.0001
of Mandible at Angle
Yes 92.3% 7.7%
No 27.6% 72.4%
Blunting of Inferior Border 2.297 ± 0.394 99.0 21.11-463.84 <0.0001
of Mandible at Body

PT
Yes 93.9% 6.1%
No 13.5% 86.5%
Floor of Mouth Elevation 1.088 ± 0.414 8.80 1.74-44.59 0.003

RI
Yes 77.8% 22.2%
No 28.5% 71.5%
Floor of Mouth Induration 1.00 ± 0.421 7.32 1.41-38.11 0.010

SC
Yes 75.0% 25.0%
No 29.1% 70.9%
White Blood Cell Count 13.6 10.2 0.200 ± 0.053 1.22 1.10-1.35 <0.0001

U
(1000 cells/µL)
C-Reactive Protein (mg/dL) 11.2 7.5 0.121 ± 0.064 1.13 1.00-1.28 0.046
AN
Table 3. Relationship of the Primary Predictor Variables to the Outcome Variables.
M

Study Variable Maximum Standard Adjusted Odds 95% CI of p Value

Likelihood Error Ratio Odds Ratio
D

Estimate
Blunting of Inferior Border of 243.25 18.97-3120.8
TE

Mandible at Body 2.747 0.6509 <0.0001

Maximum Incisal Opening -0.245 0.0718 0.78 0.68-0.90 <0.0001
Floor of Mouth Induration 1.634 0.8383 26.27 0.98-702.51 0.0301
Odynophagia 1.002 0.4833 7.41 1.11-49.29 0.0307
EP

White Blood Cell Count 0.246 0.1237 1.28 1.00-1.63 0.0351

Table 4. Final Multivariable Logistic Regression Analysis.
C
AC
 ACCEPTED MANUSCRIPT

REFERENCES

1. Kim MK, Allareddy V, Nalliah RP, Kim JE, Allareddy V: Burden of facial cellulitis:

estimates from the Nationwide Emergency Department Sample. Oral Surg Oral Med Oral Pathol

PT
Oral Radiol 114:312, 2012

RI
2. Nalliah RP, Allareddy V, Elangovan S, Karimbux N, Lee MK, Gajendrareddy P,

Allareddy V: Hospital emergency department visits attributed to pulpal and periapical disease in

SC
the United States in 2006. J Endod 37:6, 2011

3. Sun BC, Chi DL, Schwarz E, Milgrom P, Yagapen A, Malveau S, Chen Z, Chan B,

U
Danner S, Owen E, Morton V, Lowe RA: Emergency department visits for nontraumatic dental
AN
problems: a mixed-methods study. Am J Public Health 105:947, 2015

4. Christensen B, Han M, Dillon JK: The cause of cost in the management of odontogenic
M

infections 1: a demographic survey and multivariate analysis. J Oral Maxillofac Surg 71:2058,
D

2013
TE

5. Jundt JS, Gutta R: Characteristics and cost impact of severe odontogenic infections. Oral

Surg Oral Med Oral Pathol Oral Radiol 114:558, 2012

EP

6. Christensen BJ, Park EP, Nelson S, King BJ: Are Emergency Medicine Physicians Able

to Determine the Need for Computed Tomography and Specialist Consultation in Odontogenic
C

Maxillofacial Infections? Journal of Oral and Maxillofacial Surgery, 2018

AC

7. Lo BW, Fukuda H, Angle M, Teitelbaum J, Macdonald RL, Farrokhyar F, Thabane L,

Levine MA: Aneurysmal subarachnoid hemorrhage prognostic decision-making algorithm using

classification and regression tree analysis. Surg Neurol Int 7:73, 2016
 ACCEPTED MANUSCRIPT

8. Henrard S, Speybroeck N, Hermans C: Classification and regression tree analysis vs.

multivariable linear and logistic regression methods as statistical tools for studying haemophilia.

Haemophilia 21:715, 2015

9. Kobayashi D, Yokota K, Takahashi O, Arioka H, Fukui T: A predictive rule for mortality

PT
of inpatients with Staphylococcus aureus bacteraemia: A classification and regression tree

RI
analysis. Eur J Intern Med 25:914, 2014

10. Fonarow GC, Adams KF, Jr., Abraham WT, Yancy CW, Boscardin WJ, Adhere

SC
Scientific Advisory Committee SG, Investigators: Risk stratification for in-hospital mortality in

acutely decompensated heart failure: classification and regression tree analysis. JAMA 293:572,

2005
U
AN
11. Wu BU, Johannes RS, Sun X, Tabak Y, Conwell DL, Banks PA: The early prediction of

mortality in acute pancreatitis: a large population-based study. Gut 57:1698, 2008

M

12. Miloro M, Peterson LJ: Peterson's principles of oral and maxillofacial surgery. (ed. 3rd).
D

Shelton, CT, People's Medical Pub. House-USA,, 2012, p p841

TE
C EP
AC
 ACCEPTED MANUSCRIPT

Figure 1. Classification and Regression Tree Analysis for Predictors of Need for CT Imaging in
the Study Cohort. The diagram is a depiction of the branch points used in the classification and
regression tree analysis performed on the study population. Each node represents the best branch
point identified by the algorithm. Also included in the diagram is the portion of the study patients
that would fall into each of the proposed categories.

Figure 2. Proposed Criteria for CT Imaging in Odontogenic Infections. This diagram represents a

PT
suggested algorithm for determining the need for CT imaging in patients with odontogenic
infections. Note that regardless of the criteria, CT imaging should be considered for patients with
unusual findings such as orbital proptosis, pain with extraocular movements, visual changes and

RI
fluctuant swelling in the temporal hairline as orbital and temporal space infections were not well-
represented in this study population.

U SC
AN
M
D
TE
C EP
AC
 ACCEPTED MANUSCRIPT

PT
RI
U SC
AN
M
D
TE
EP
C
AC
 ACCEPTED MANUSCRIPT

PT
RI
U SC
AN
M
D
TE
EP
C
AC