Ogwo et al.

BMC Oral Health (2024) 24:313 BMC Oral Health

https://doi.org/10.1186/s12903-024-04012-3

RESEARCH Open Access

Predicting dental anxiety in young adults:

classical statistical modelling approach versus
machine learning approach
Chukwuebuka Ogwo1*, Wisdom Osisioma2, David Ifeanyi Okoye1 and Jay Patel1

Abstract
Objectives To predict and identify the key demographic and clinical exposure factors associated with dental anxiety
among young adults, and to compare if the traditional statistical modelling approach provides similar results to the
machine learning (ML) approach in predicting factors for dental anxiety.
Methods A cross-sectional study of Western Illinois University students. Three survey instruments (sociodemographic
questionnaire, modified dental anxiety scale (MDAS), and dental concerns assessment tool (DCA)) were distributed
via email to the students using survey monkey. The dependent variable was the mean MDAS scores, while the
independent variables were the sociodemographic and dental concern assessment variables. Multivariable analysis
was done by comparing the classical statistical model and the machine learning model. The classical statistical
modelling technique was conducted using the multiple linear regression analysis and the final model was selected
based on Akaike information Criteria (AIC) using the backward stepwise technique while the machine learining
modelling was performed by comparing two ML models: LASSO regression and extreme gradient boosting machine
(XGBOOST) under 5-fold cross-validation using the resampling technique. All statistical analyses were performed
using R version 4.1.3.
Results The mean MDAS was 13.73 ± 5.51. After careful consideration of all possible fitted models and their
interaction terms the classical statistical approach yielded a parsimonious model with 13 predictor variables with
Akaike Information Criteria (AIC) of 2376.4. For the ML approach, the Lasso regression model was the best-performing
model with a mean RMSE of 0.617, R2 of 0.615, and MAE of 0.483. Comparing the variable selection of ML versus the
classical statistical model, both model types identified 12 similar variables (out of 13) as the most important predictors
of dental anxiety in this study population.
Conclusion There is a high burden of dental anxiety within this study population. This study contributes to reducing
the knowledge gap about the impact of clinical exposure variables on dental anxiety and the role of machine
learningin the prediction of dental anxiety. The predictor variables identified can be used to inform public health
interventions that are geared towards eliminating the individual clinical exposure triggers of dental anxiety are
recommended.

*Correspondence:
Chukwuebuka Ogwo
[email protected]
1
Department of Oral Health Sciences, Maurice H Kornberg School of
Dentistry, Temple University, Philadelphia, PA 19140, US
2
School of Public Health, Texas A&M University, College Station, TX, USA

© The Author(s) 2024. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use,
sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and
the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this
article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included
in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will
need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/. The
Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available
in this article, unless otherwise stated in a credit line to the data.

Content courtesy of Springer Nature, terms of use apply. Rights reserved.

Ogwo et al. BMC Oral Health (2024) 24:313 Page 2 of 9

Introduction comparing classical statistical modeling and ML model

Dental fear or anxiety (DA) can be described as a subjec- in predicting oral health outcomes and more specifically
tive negative reaction to dental treatment resulting from dental anxiety.
a learned negative behavior and often attributed to the The objectives of this study were to predict and identify
aggressive conditioning process which occurred during the key demographic and clinical exposure factors asso-
childhood [1]. Fear of pain has been found to be the main ciated with dental anxiety among young adults, and to
cause of anxiety and a major barrier to seeking dental compare if the traditional statistical modelling approach
care [2, 3]. Dental avoidance has been linked with den- provides similar results to the machine learning (ML)
tal fear and anxiety in many patients and thus has led to approach in predicting factors for dental anxiety.
the deterioration of their oral health state [4, 5]. In severe
cases of dental anxiety, the dentist-patient relation may Methods
be hampered and sometimes lead to misdiagnoses of This is a cross-sectional study of university students
anxiety for pain which might result in wrong treatment recruited from Western Illinois University (WIU). Eli-
administration [4, 6]. The standardized and validated tool gible participants were graduate and undergraduate stu-
for the measurement of dental anxiety is known as the dents registered for the 2017 Spring semester at WIU
Modified Dental Anxiety Scale (MDAS) [7]. at both campuses (Macomb and the Quad Cities cam-
Globally, several reports have shown the prevalence puses). Using the pwr package [18] in R for generalized
of dental anxiety to be between 2.5 and 20% depending linear regression, we estimated the sample size of 1062
on population and methods of assessment [8–10]. Stud- students based on the following assumptions – small
ies have also reported dental anxiety to occur more in effect size of 2%, 21 variables, and 80% power, and 95%
females than males [7, 11]. Recent studies have shown confidence level.
that 51% of subjects reported dental anxiety onset in
childhood, 22% in adolescence, and 27% in adulthood [8, Data collection
12]. Studies have shown gender and age differences in the After obtaining approval from the IRB, we obtained the
prevalence of dental anxiety but more importantly, the list of emails of the students registered for the Spring
socioeconomic differences which bother mostly on fear semester of 2017 and emailed the survey instruments to
of treatment cost [7, 11, 13]. the participants electronically via Survey Monkey. The
There are limited studies on the burden of dental anxi- survey instrument consisted of a self-administered struc-
ety among university students in the US, which constitute tured questionnaire, a validated modified dental anxiety
a reasonable population of adolescents and young adults scale (MDAS) [7], and a Dental Concern/Fear Assess-
in the country [14]. Also, very few studies have explored ment tool (DCA) [19]. The inclusion criteria were all stu-
the clinical exposure variables that may be associated dents (undergraduate, graduate, and doctoral) registered
with dental anxiety among young adults [15]. A study of for the spring 2017 semester. The e-mails were sent two
Washington University students showed that about one times, the first one in March and the second one in April
out of five students reported high levels of dental anxi- 2017. The termination date for the survey return was set
ety (mean DAS = 9.2; SD = 3.4) and most of the students at two weeks after the first survey was sent. The survey
reported that their dental anxiety was due to dental emails were sent in the evening, assuming that students
injection [16]. Previous studies in the U.S. have focused would be more relaxed in the evenings and have more
on dental anxiety among American children and older time to complete the survey. Informed consent was
adults with only a few studies on young adults or ado- obtained from all subjects and their legal guardian(s)
lescents [14, 15]. Thus, there is a need for more studies before responding to the questionnaires.
among young adults to understand the clinical exposure The dependent variable was the mean MDAS scores.
variables contributing to dental anxiety. The independent variables (see Appendix I for details)
Supervised machine learning (ML) is a subset of artifi- include age group, sex, socioeconomic status, dental visit,
cial intelligence used in the prediction of outcome mea- frequency of dental visit, level of education, and the den-
sures based on several input measures. The goal of the tal concern variables (Sound or vibration of the drill, dis-
ML model is to optimize the bias-variance trade-off and like of the numb feeling, injection in the mouth, sound
prevent model underfitting or overfitting [17]. Machine or feel of scraping during teeth cleaning, cold air on the
Learningprovides a robust approach for the identification teeth, root canal treatment, tooth removal, fear of being
and selection of the most important predictors, without injured, panic attacks, fear of feeling pain during treat-
running into issues of numerical convergence and the ment, the concern of being embarrassed, smells in the
“curse of dimensionality” a phenomenon common in dental office, worried about need a lot of dental treat-
classical statistical modeling with a lot of predictor vari- ment, cost of the dental treatment). Studies have identi-
ables. Most importantly, there is a dearth of literature fied the cognitive conditioning pathway as a framework

Content courtesy of Springer Nature, terms of use apply. Rights reserved.

Ogwo et al. BMC Oral Health (2024) 24:313 Page 3 of 9

elucidating the aetiology of dental anxiety. Within this Results

framework, individuals with negative experiences dur- A total of 454 students (45% response rate) completed the
ing dental visits may establish a conditioned association dental anxiety questionnaire (i.e. completely answered
linking the dentist with anxiety [20, 21]. The DCA tool at least the Sociodemographic and MDAS sections of
was modified and adapted for our study to consist of 15 the questionnaire) by the end of the survey period and
short dental anxiety-specific questions with three rank- thus were included in the study. No data were inputted.
order response options and was used to measure dental The pilot study of 30 students showed a response rate
anxiety concerning individual triggers or aggravating of 100%. All the 30 participants accepted the question-
factors from dental procedures. Before the primary data naires and had no worries or questions about the con-
collection, a pilot study was conducted using a randomly tent of measurement tools. Thus, the questionnaires were
selected thirty (30) participants to check the response validated for the measurement of dental anxiety within
rate, acceptability and the validity of the assessment tools this population and their responses were included in the
that were utilized for the study. study.
As shown in Table 1, about two-thirds of the partici-
Statistical analysis pants were female (69.40%). Most of the respondents
Only participants who have complete data for all the vari- were within the age range of 15 to 24 (68.10%). Most of
ables were included in the final analysis. No imputation the respondents (98.20%) have visited a dentist. About
of data was done. The univariate analyses were conducted 35.20% of the respondents had not visited the dentist in
for both the dependent and independent variables. The the last 12 months. More than half of the respondents
one-way ANOVA test was used to assess the bivari- reported a household income of $74,999 or less.
ate relationship between each categorical independent Table 1 also showed the different levels of dental con-
variable and the continuous dependent variable (MDAS cerns/fear about various triggers of dental anxiety among
score). Only the statistically significant variables (p < 0.05) the study participants. The study participants had a fairly
were included in the final model for both the classical equal level of dental concern/fear about the sound or
statistical approach and the ML approach. vibration of the drill. Up to 57% expressed a low level of
Multivariable analysis was done to ascertain the rela- concern about the numb feeling from dental treatment
tionship between the independent variables and the while 45.2% said they have a high level of concern about
MDAS score and identify the predictors of dental anxiety injection in the mouth. About 41% and 44% had a low
by comparing the results from two modelling approaches: concern about the sound or feel of scraping during teeth
The classical statistical approach and the machine learn- cleaning and cold air in the mouth respectively. The study
ing approach. The classical statistical modelling tech- participants had high levels of concerns (67%) for both
nique was conducted using the multiple linear regression root canal treatment and extractions. About 62% had
analysis and the final model was selected based on high dental concerns for panic attacks during treatment.
Akaike information Criteria (AIC) using the backward The mean MDAS was 13.73 ± 5.51 (Table 1). The
stepwise technique. The root mean square error (RMSE), prevalence of dental anxiety among the respondents
and coefficient of determination (R2) for the classical was 63.90%. The prevalence of extreme anxiety was
statistical modelling approach was calculated. For com- 19.50% while high anxiety and moderate anxieties were
parison, the machine learning technique was performed 21.50% and 22.90% respectively (See details in Table 1
by comparing two ML models: LASSO regression [22] on how the MDAS scores were categorized). As shown
and extreme gradient boosting machine (XGBOOST) in Table 2, all predictor variables were statistically sig-
[23] under 5-fold cross-validation using the resampling nificant (p < 0.001) except age, dental visit, and education
technique. The data pre-processing (Standardization and (see Table 2 for details).
normalization) for the machine learning model was done
using the recipes package [24]. The model performance Multivariable analyses
was based on RMSE, R2, and mean absolute error (MAE). Classical statistical model (multiple Linear regression)
The RMSE value was the main metric for performance As shown in Table 3, after careful consideration of all
assessment and comparison of both model types. The possible fitted models and their interaction terms for
RMSE value indicates measures difference between the multiple linear regression analysis, the parsimonious
predicted values and the observed values. Therefore, the model had 9 variables with Akaike Information Criteria
lower the RMSE, the better the model performance. All (AIC) of 2376.4. The RMSE and R2 were 3.16 and 0.67,
statistical analyses were performed using R version 4.1.3. respectively.
Holding all other variables in the model constant,
high and moderate fear of panic attacks during treat-
ment were significantly associated with higher mean DA

Content courtesy of Springer Nature, terms of use apply. Rights reserved.

Ogwo et al. BMC Oral Health (2024) 24:313 Page 4 of 9

Table 1 Descriptive statistics of the dental anxiety scores, demographic and dental concern assessment variables
N Frequency (%)
Sex Female 135 69.40
Male 139 30.60
Age Less than 24 309 68.10
25 to 34 84 18.50
35 to 44 29 6.40
45 to 54 24 5.30
55 or older 8 1.80
Dental visit Yes 446 98.20
No 8 1.80
Frequency of dental visits* Less than 3 months 115 25.30
3 months to < 6 months 89 19.60
6 months to < 12 months 90 19.80
More than 12 months 160 35.20
Household Income* $0 - $24,999 126 27.80
$25,000 - $74,999 172 37.90
$75,000 - $124,999 119 26.20
$125,000 - $149,999 19 4.20
$150,000 and up 18 4.00
Level of Education Doctoral Student 9 2.00
Graduate Student 142 31.30
Undergraduate Student 303 66.70
Sound or vibration of the drill Low 142 31.30
Moderate 156 34.40
High 156 34.40
Dislike the numb feeling Low 262 57.70
Moderate 152 26.70
High 71 15.60
Injection in the mouth Low 97 21.40
Moderate 152 33.50
High 205 45.20
The sound or feel of scraping during teeth cleaning Low 185 40.70
Moderate 140 30.80
High 129 28.40
Cold air on the teeth Low 200 44.10
Moderate 154 33.90
High 100 22.00
Root canal treatment Low 49 10.80
Moderate 97 21.40
High 308 67.80
Tooth removal Low 53 11.70
Moderate 95 20.90
High 306 67.40
Fear of being injured Low 167 36.80
Moderate 131 28.90
High 156 34.40
Panic attacks Low 281 61.90
Moderate 100 22.00
High 73 16.10
Fear of feeling pain during treatment Low 62 13.70
Moderate 147 32.40
High 245 54.00

Content courtesy of Springer Nature, terms of use apply. Rights reserved.

Ogwo et al. BMC Oral Health (2024) 24:313 Page 5 of 9

Table 1 (continued)
N Frequency (%)
Concern of being embarrassed Low 269 59.30
Moderate 102 22.50
High 83 18.30
Smells in the dental office Low 327 72.00
Moderate 89 19.60
High 38 8.40
Worried about needing a lot of dental treatment Low 250 55.10
Moderate 102 22.50
High 102 22.50
Cost of the dental treatment Low 127 28.00
Moderate 140 30.80
High 187 41.20
Dental anxiety Level MDAS Scale N Prevalence (%)
No Anxiety 5 to 10 240 36.10
Moderate Anxiety 11 to 14 152 22.90
High Anxiety 15 to 18 143 21.50
Extreme Anxiety 19 to 25 130 19.50
Mean Median Variance Standard Deviation
MDAS Score 13.73 13.00 30.35 5.51

Table 2 One-way ANOVA: independent variables versus mean

MDAS score
associated with higher mean DA compared to low fear of
F-value p-value
root canal treatment (β = 2.69, p < 0.01; β = 1.62, p = 0.01).
Sex 10.02 0.002 High fear of injection was associated with higher mean
Age group 0.38 0.822 DA compared to low fear of injection (β = 1.91, p < 0.01).
Dental visit 2.36 0.166 High fear of needing a lot of dental treatment was signifi-
Frequency of dental visit 7.22 1.198e-4 cantly associated with higher mean DA compared to low
Household Income 2.62 0.042 fear (β = 1.60, p < 0.01). High fear of being embarrassed
Level of Education 0.96 0.399 was significantly associated with higher mean DA com-
Fear of sound or vibration of the drill 156.20 < 2.2e-16 pared to low fear of being embarrassed (β = 1.52, p = 0.01).
Dislike the numb feeling 40.45 2.99e-15 The frequency of dental visits of less than 3 months and
Fear of injection in the mouth 92.74 < 2.2e-16 3 months to less than 6 months were associated with a
Fear of sound or feel of scraping during 62.69 < 2.2e-16 lower mean DA score compared to dental visit frequency
teeth cleaning of 12 months and above (β = -0.94, p = 0.02; β = -1.17,
Fear of cold air on the teeth 34.98 4.316e-14 p = 0.01). Moderate fear of the sound or feel of scraping
Fear of root canal treatment 114.26 < 2.2e-16 during teeth cleaning was associated with a higher mean
Tooth removal 158.04 < 2.2e-16 DA compared to low fear (β = 0.82, p = 0.04).
Fear of being injured 112.39 < 2.2e-16
Fear of panic attacks 112.20 < 2.2e-16 Machine learning model
Fear of feeling pain during treatment 136.78 < 2.2e-16 After a comparison of the Lasso regression and
Concern of being embarrassed 43.79 3.41e-16 XGBOOST model, the Lasso regression model was
Dislike Smells in the dental office 72.48 < 2.2e-16 found to be the best-performing model with a mean
Fear of needing a lot of dental treatment 73.36 < 2.2e-16
RMSE of 0.617, R2 of 0.615, and MAE of 0.483 (Table 4).
Fear of cost of the dental treatment 6.53 1.69 e-3
The details of the performance of the XGBOOST model
can be found in Appendix VII. The calibration plot of
compared to low fear of panic attacks (β = 3.14, p < 0.01; the Lasso regression model showed a close calibration
β = 1.28, p < 0.01, respectively). High fear of feeling pain between the observed and the predicted means (Appen-
during treatment was associated with a higher mean DA dix III). The variable importance assessment found iden-
compared to low fear (β = 2.70, p < 0.01). High and mod- tified 28 predictors of dental anxiety in this population
erate fear of the sound or vibration of drills was associ- and these variables were ranked based on their permuted
ated with a higher mean DA score compared to having mean RMSE score (Table 5).
low fear (β = 2.70, p < 0.01; β = 0.98, p = 0.02, respectively).
High and moderate fear of root canal treatment was

Content courtesy of Springer Nature, terms of use apply. Rights reserved.

Ogwo et al. BMC Oral Health (2024) 24:313 Page 6 of 9

Table 3 Summary output of the final (reduced) model from Table 5 Showing variable importance based on permitted
generalized linear model (Arranged in the order of variable mean RMSE.
importance based on the magnitude of their beta coefficient) Per-
Final model: Akaike Information Criteria (AIC) = 2376.4 muted
Categories Estimate P-value mean
RMSE*
Intercept 6.55 < 0.001
High fear of feeling pain during treatment 0.093
Fear of panic attacks High 3.14 < 0.001
High fear of sound or vibration of the drill 0.062
Moderate 1.28 0.003
High fear of root canal treatment 0.061
Low Ref
High fear of panic attacks 0.052
Fear of feeling pain High 2.76 < 0.001
during treatment High fear of injection in the mouth 0.043
Moderate 0.88 0.09
High fear of needing a lot of dental treatment 0.026
Low Ref
High fear of being embarrassed 0.020
Fear of Sound or High 2.70 < 0.001
vibration of the drill Moderate Moderate fear of root canal treatment 0.014
0.98 0.02
Moderate fear of panic attacks 0.012
Low Ref
High fear of being injured 0.012
Fear of root canal High 2.69 < 0.001
treatment Moderate fear of the sound or vibration of the drill 0.011
Moderate 1.62 0.01
Frequency of dental visits (3 months to less than 6 months) 0.010
Low Ref
Frequency of dental visits less than 3 months 0.006
Fear of injection in High 1.91 < 0.001
the mouth Moderate fear of sound or feel of scraping during teeth 0.005
Moderate 0.42 0.37
cleaning
Low Ref
High fear of cost of the dental treatment 0.003
Fear of needing High 1.60 0.003
High fear of cold air on the teeth 0.003
a lot of dental Moderate 0.70 0.10
treatment High dislike of smells in the dental office 0.003
Low Ref
Moderate fear of needing a lot of dental treatment 0.003
Fear of being High 1.52 0.01
High fear of sound or feel of scraping during teeth cleaning 0.003
embarrassed Moderate 0.35 0.41
High fear of tooth removal 0.002
Low Ref
Moderate fear of concern of being embarrassed 0.002
Frequency of dental Less than 3 months -0.94 0.02
Moderate fear for the numb feeling 0.002
visit 3 months to less than 6 -1.17 0.01
Moderate fear of tooth removal 0.001
months
Frequency of dental visits (6 months but less than 12 months) 0.001
6 months to less than 12 -0.74 0.09
months Moderate fear of injection in the mouth 0.001
12 months and above Ref Moderate dislike of smells in the dental office 0.001
Fear of the sound or High 0.70 0.12 Moderate fear of feeling pain during treatment 0.000
feel of scraping dur- Moderate 0.82 0.04 Sex (Male) 0.000
ing teeth cleaning Low Ref Moderate fear of cost of the dental treatment 0.000
Dispersion parameter for gaussian family taken = 10.48 High dislike the numb feeling 0.000
Null deviance: 13749.2 on 453 degrees of freedom Moderate fear of being injured -0.0004
Residual deviance: 4546.4 on 434 degrees of freedom Moderate fear of cold air on the teeth -0.0005
*Higher permuted mean RMSE score means higher variable importance

Table 4 Showing the performance of the LASSO regression

model similar variables (out of 13) as the most important pre-
Metrics Mean Median SD Min Max dictors of dental anxiety in this study population. The
RMSE 0.617 0.597 0.040 0.585 0.672 ranking of the variable importance across the differ-
R2 0.615 0.638 0.056 0.525 0.660 ent model types varied however the top 5 predictors of
MAE 0.481 0.480 0.020 0.457 0.512 DA were high fear of pain, panic attack, sound or vibra-
RMSE = Root mean square error; R2 = Coefficient of determination; MAE = Mean tion of drill, root canal, and injection in the mouth. The
absolute error RMSE and R2 of the classical statistical model were 3.16
and 0.67, respectively versus the RMSE and R2 of the ML
Comparing variable importance of ML versus classical model which were 0.617 and 0.615, respectively (Table 6).
statistical model
As shown in Table 3, the classical statistical model had Discussion
only 13 statistically significant variables and as such we Dental anxiety is a huge concern for dental profession-
selected the top 13 predictors of DA from the ML model als, public health specialists, and patients because of its
variable importance estimation for comparison (Table 5). association with poor oral health outcomes. Recent stud-
As shown in Table 6, both model types identified 12 ies have shown a surge in the prevalence of dental anxiety

Content courtesy of Springer Nature, terms of use apply. Rights reserved.

Ogwo et al. BMC Oral Health (2024) 24:313 Page 7 of 9

Table 6 Comparison of the 10 most important predictors of DA 25 and come from lower-income households. Almost all
identified using ML model versus classical statistical model the participants had visited a dentist before and therefore
ML model (Lasso regression) * Classical statistical model have had some prior exposure to the clinical triggers of
(Multiple Linear regres-
sion) **
dental anxiety assessed in this study.
High fear of feeling pain during High fear of panic attack When compared to the ML model approach, the clas-
treatment sical statistical model approach showed a much higher
High fear of the sound or vibration of High of feeling pain during RMSE and slightly higher R2. This implies that our ML
the drill treatment model performed better than the classical statistical
High fear of root canal treatment High fear of the sound or model in predicting dental anxiety due to the higher
vibration of the drill error rate in the classical statistical model and bearing
High fear of panic attacks High fear of root canal in mind that R2 is sensitive to the number of variables in
treatment
the model and therefore not a very accurate measure of
High fear of injection in the mouth High fear of injection in the
mouth
model performance. Also, our classical statistical model
High fear of needing a lot of dental Moderate fear of root canal
(multiple linear regression) revealed only 13 predictors
treatment treatment of DA in this study population based on the beta coef-
High fear of being embarrassed High fear of needing a lot of ficient and p-value. In contrast, our ML model (Lasso
dental treatment regression) identified 28 predictors of DA based on the
Moderate fear of root canal treatment High fear of being permuted mean RMSE. This highlights the ability of
embarrassed machine learning to model complex interactions between
Moderate fear of panic attacks Moderate fear of panic attacks variables and identify a wider range of predictors beyond
High fear of being injured Frequency of dental visits (3 the classical model.
months to less than 6 months)
Interestingly, our study showed a very comparable
Moderate fear of the sound or vibration Moderate fear of the sound or
of the drill feel of scraping during teeth
performance between the classical statistical modelling
cleaning approach and the machine learning approach in terms
Frequency of dental visits (3 months to Moderate fear of the sound or of variable selection. When we compared the 13 pre-
less than 6 months) vibration of the drill dictors of DA from the classical statistical model to the
Frequency of dental visits less than 3 Frequency of dental visits less top 13 predictor variables from the machine learning
months than 3 months model, both models identified 12 similar predictors of
Comparison of performance of the ML model versus classical dental anxiety. The predictors include high fear of feeling
statistical model
pain during treatment, high fear of the sound or vibra-
RMSE = 0.617 RMSE = 3.16
tion of the drill, high fear of root canal treatment, high
R2 = 0.615 R2 = 0.67
fear of panic attacks, high fear of injection in the mouth,
* Ranking was based on the permuted mean RMSE.
high fear of needing a lot of dental treatment, high fear
** Ranking was based on the beta coefficient and statistical significance
(p < 0.05) of being embarrassed, moderate fear of root canal treat-
Note: Lower RMSE equals better model performance. ment moderate fear of panic attacks, moderate fear of
sound or vibration of the drill, frequency of dental visit
[25, 26]. However, there are very few recent studies on (3 months to less than 6 months) and frequency of dental
young adults, especially in the United States of America visit (less than 3 months).
[27, 28]. There are no existing studies that compared clas- High fear of pain during treatment and fear of injection
sical statistical models versus machine learning models in the mouth were associated with higher dental anxiety.
in predicting and identifying the predictors of dental More than half of the participants rated their level of fear
anxiety. of pain during treatment as high while more than three-
The high mean MDAS score found in this study is quarters of the participants indicated moderate to high
higher than the average MDAS score of 12.34 found in levels of fear of injection. Our findings align with the
the Saatchi et al. study [29]. Comparable studies in the studies from Georgelin-Gurgel et al. [30] that found an
U.S. by Locker and Liddle (mean DAS = 7.8) and Kaako association between higher levels of DA and fear of intra-
et al. study among university students (mean DAS = 9.2) oral injection. Individuals who had high fear of the sound
have also shown lower scores even though they used or vibration of drills had a 2.70 higher mean DA score
DAS for measurement [13, 16]. It is crucial to highlight compared to those who had low low fear of the sound
that the DAS (Corah’s Dental Anxiety Scale) comprises or vibration of drills. This agrees with the Cohen et al.28
four items, yielding scores from 4 to 20. In contrast, the study that found a significant relationship between DA
MDAS (Modified Dental Anxiety Scale) utilized in our and the sound/vibration of drills.
study is a five-item measure, with score range of 5 to 25. High fear of root canal treatment was associated with
Most of the study participants were females below age a 2.6 higher mean DA score compared to low fear of root

Content courtesy of Springer Nature, terms of use apply. Rights reserved.

Ogwo et al. BMC Oral Health (2024) 24:313 Page 8 of 9

canal treatment with more than about two-thirds of the using the modified DCA questionnaires outside this
participants indicating high level of fear about root canal study population. Similarly, it was not feasible to deter-
treatment which aligns with similar findings by Algho- mine the reasons behind non-responses from certain
faily et al. [31]. study participants or elucidate factors implicated in the
Individuals with high fear of panic attacks during low response rates. This limitation may have significant
treatment had a 3.22 higher mean DA score compared implications for the strength of our study’s conclusions.
to those who had low panic attacks during treatment. However, a smaller sample size may affect the statisti-
Also, participants with high fear of being embarrassed cal power of our analysis, which may underestimate the
had a 1.5 higher mean DA compared to those who had actual effect or relationships present in this population.
low fear of being embarrassed. A high level of concern Also, due to the sampling method and sensitivity of the
about needing a lot of dental treatment was significantly topic, the true cases might have been missed out. Other
associated with 1.84 higher mean DA compared to a low than comparing the variable selection, there are no exist-
level of concern about needing a lot of dental treatment. ing objective metrics for comparing classical statistical
The public health relevance of these findings is that if an versus machine learning models.
individual feels they are going to be embarrassed or get In future studies, a more diverse and larger sample size
diagnosed with more dental issues, they become more will be considered to enhance the strength, reliability and
anxious and avoid routine dental care visits altogether. applicability of our study.
Individuals who visited the dentist more frequently
were significantly less likely to have dental anxiety. Indi- Conclusion
viduals who have visited a dentist in the past 6 months There is a high burden of dental anxiety within this study
had at least a 1.18 lower average DA score compared population and continues to constitute a serious dental
to those who have not visited the dentist in the past 12 public health issue because those impacted are known to
months. This finding conforms with the study by Doerr avoid dental visits. More frequent exposure to the clinic
et al. study where those who did not go for a checkup at environment through routine visits plays a huge role in
least once a year were found to be more dentally anx- reducing the burden of dental anxiety, especially in young
ious than subjects receiving more frequent dental care adults. This study contributes to reduce the knowledge
[27]. This implies that frequent visits to the dentist could gap about the impact of clinical exposure variables on
help decrease dental anxiety due to continuous exposure dental anxiety and the role of machine learningin the
to dental anxiety stimuli thereby improving the patient’s prediction of dental anxiety. Behavioral theory (such as
self-efficacy. Inversely, high dental anxiety can be said to motivational interviewing) based public health inter-
have caused the low frequency of visits within this study ventions that are geared towards eliminating the indi-
population possibly due to previous personal experi- vidual clinical exposure triggers of dental anxiety are
ence or experiences of a family member, friends, or col- recommended.
leagues. Our study found no association between dental
anxiety and the average household income, level of edu- Supplementary Information
cation, and sex. Both Our Machine Learning (ML) and The online version contains supplementary material available at https://doi.
org/10.1186/s12903-024-04012-3.
the classical statistical approach identified set of variables
as predictors for dental anxiety. These variables hold the Supplementary Material 1
potential to form the basis for developing a web applica-
tion tailored to aid the diagnosis of dental anxiety and Acknowledgements
the customization of patient-specific interventions. It I would like to acknowledge the contributions of Chukwuebuka Ogwo (the
is important to highlight that our findings were derived corresponding authors) led and contributed to the conception, design, data
acquisition, and interpretation, performed all statistical analyses, and drafted
from a sample of 454 students, which differs greatly from and critically revised the manuscript. Ifeanyi David Okoye and Wisdom
the initially calculated sample size of 1062 students and Osisioma contributed to the conception, design, data analysis, writing, and
may have limited external validity. preparing the tables and figures for the manuscript. Jay Patel contributed
to the data analysis, interpretation, drafting, and revision of the manuscript.
The limitations of this study include the lack of gen- Thank you, Professor Marisol Tellez Merchan for her expert guidance and
eralizability to other populations due to the differences comments during the final revision of the manuscript.
in population characteristics. The data obtained from
Author contributions
WIU might not be representative of other universities Chukwuebuka Ogwo (the corresponding authors) led and contributed to
in America. The study participants might not have given the conception, design, data acquisition, and interpretation, performed all
adequate and accurate information regarding the level of statistical analyses, and drafted and critically revised the manuscript. Ifeanyi
David Okoye and Wisdom Osisioma contributed to the conception, design,
dental anxiety since its measurement is subjective. Reli- data analysis, writing, and preparing the tables and figures for the manuscript.
ability testing of the modified DCA questionnaire was Jay Patel contributed to the data analysis, interpretation, drafting, and revision
not conducted, therefore caution should be applied when

Content courtesy of Springer Nature, terms of use apply. Rights reserved.

Ogwo et al. BMC Oral Health (2024) 24:313 Page 9 of 9

of the manuscript. All authors gave their final approval and agreed to be 11. Frazer M, Hampson S. Some personality factors related to dental anxiety and
accountable for all aspects of the work. fear of pain. Br Dent J. 1988;165:436–9.
12. Ost I. Age of onset of different phobias. J Abnorm Psychol. 1987;96:223–9.
Funding 13. Locker D, Liddell A. Correlates of Dental anxiety among older adults. J Dent
This research was supported by Temple University Kornberg School of Res. 1991;70(3):198–203. https://doi.org/10.1177/00220345910700030801.
Dentistry and the Cary Kleinman Oral Health Sciences Research Fund. 14. Raadal M, Milgrom P, Weinstein P, Mancl L, Cauce AM. The prevalence of
dental anxiety in children from low-income families and its relationship to
Data availability personality traits. J Dent Res. 1995;74(8):1439–43.
The data that support the findings of this study are available on request from 15. Elter JR, Strauss RP, Beck JD. Assessing dental anxiety, dental care use, and oral
the corresponding author. health status in older adults. JADA. 1997;128(5):591–7.
16. Kaakko T, Milgrom P, Coldwell SE, Getz T, Weinstein P, Ramsay DS. Dental
fear among university students: implications for pharmacological research.
Declarations Anesth Prog. 1998;45(2):62–7.
17. James G, Witten D, Hastie T, Tibshirani R. An introduction to statistical learn-
Competing interests ing. Volume 112. New York: springer; 2013 Jun. p. 18.
The authors declare no competing interests. 18. Champely S et al. pwr: Basic Functions for Power Analysis_. R package version
1.3-0. 2020. https://CRAN.R-project.org/package=pwr.
Ethics approval and consent to participate 19. Clarke JH, Rustvold S. Corah’s dental anxiety scale, revised (DAS-R). Oregon
Approval for the conduct and collection of the data for this study was Health Sciences University School of Dentistry, 1993 [revised 1998]. Retrieved
obtained from Western Illinois University Institutional Review Board. All from http://www.dentalfearcentral.org/media/dental_anxiety_scale.pdf.
components and procedures of the study were conducted in accordance with 20. Klingberg G, Broberg AG. Dental fear/anxiety and dental behaviour manage-
the guidelines and regulations of the Western Illinois University Institutional ment problems in children and adolescents: a review of prevalence and
Review Board. Informed consent was obtained from all subjects and their legal concomitant psychological factors. Int J Paediatr Dent. 2007;17:391–406.
guardian(s) before responding to the questionnaires. 21. Carter AE, Carter G, Boschen M, AlShwaimi E, George R. Pathways of fear and
anxiety in dentistry: a review. World J Clin Cases. 2014;2(11):642–53. https://
Consent for publication doi.org/10.12998/wjcc.v2.i11.642.
Not applicable. 22. Tibshirani R. Regression shrinkage and selection via the lasso. J Roy Stat Soc:
Ser B (Methodol). 1996;58(1):267–88.
Received: 1 July 2023 / Accepted: 11 February 2024 23. Chen T, Guestrin C, Xgboost. A scalable tree boosting system. in Proceedings
of the 22nd acm sigkdd international conference on knowledge discovery
and data mining. 2016 Aug 13 (pp. 785–794).
24. Kuhn M, Wickham H. recipes: Preprocessing tools to create design matrices R
package version (2020) 1.8. 0. https://github.com/tidymodels/recipes.
References 25. Humphris G, Crawford JR, Hill K, Gilbert A, Freeman R. UK Population norms
1. Rubin JG, Slovin HI, Krochak M. The psychodynamics of dental anxiety and for the modified dental anxiety scale with percentile calculator: adult
dental phobia. Dent Clin North Am. 1988;32:647–56. dental health survey 2009 results. BMC Oral Health. 2013;13:29. https://doi.
2. Klepac RI, Doding J, Hauge G. Characteristics of clients seeking therapy for org/10.1186/1472-6831-13-29.
the reduction of dental avoidance: reactions to pain. J Bebar Exp Psychiat. 26. Geer JH. The development of a scale to measure fear. Behav Res Ther.
1982;13:293–300. 1965;3:45–53. https://doi.org/10.1016/0005-7967(65)90040-9.
3. Woolgrave J, Cumberbatch G. Dental anxiety and regularity of dental atten- 27. Doerr PA, Lang WP, Nyquist LV, Ronis DL. Factors Associated with Dental
dance. J Dent. 1986;14:209–13. anxiety. J Am Dent Assoc. 1998;129(8):1111–9.
4. Eli I. Dental anxiety: a cause for possible misdiagnosis of tooth vitality. Int 28. Bare LC, Dundes L. Strategies for combating dental anxiety. J Dent Educ.
Endod J. 1993;26:251–3. 2004;68(11):1172–7. doi:68/11/1172 [pii].
5. Freeman R. A psychodynamic theory for dental phobia. Br Dent J. 29. Saatchi M, Abtahi M, Mohammadi G, Mirdamadi M, Binandeh ES. The preva-
1998;184:170–2. lence of dental anxiety and fear in patients referred to Isfahan Dental School,
6. Johansson P, Berggren U, Hakeberg M, Hirsch JM. Measures of dental beliefs Iran. Dent Res J. 2015;12(3):248–53.
and attitudes: their relationships with measures of fear. Community Dent 30. Georgelin-Gurgel M, Diemer F, Nicolas E, Hennequin M. Surgical and nonsur-
Health. 1993;10(1):31–9. gical endodontic treatment-induced stress. J Endod. 2009;35(1):19–22.
7. Humphris GM, Dyer TA, Robinson PG. The modified dental anxiety 31. Alghofaily M, Alsalleeh F. Levels of anxiety and fear related to non-surgical
scale: UK general public population norms in 2008 with further psy- root canal treatment performed by endodontic residents and endodontists.
chometrics and effects of age. BMC Oral Health. 2009;9:20. https://doi. Front Dent Med:20.
org/10.1186/1472-6831-9-20.
8. Milgrom P, Fiset L, hlelnick S, et al. The prevalence and practice manage-
ment consequences of dental fear in a major US city. J Am Dent Assoc. Publisher’s Note
1988;116:641–7. Springer Nature remains neutral with regard to jurisdictional claims in
9. Locker D, Liddel A, Burman D. Dental fear and anxiety in an older adult popu- published maps and institutional affiliations.
lation. Community Dent Oral Epidemiol. 1991;19:120–4.
10. Hakeberg M, Berggren U, Carlsson SG. Prevalence of dental anxiety in an
adult population in a major urban area in Sweden. Community Dent Oral
Epidemiol. 1992;20:97–101.

Content courtesy of Springer Nature, terms of use apply. Rights reserved.

Terms and Conditions
Springer Nature journal content, brought to you courtesy of Springer Nature Customer Service Center GmbH (“Springer Nature”).
Springer Nature supports a reasonable amount of sharing of research papers by authors, subscribers and authorised users (“Users”), for small-
scale personal, non-commercial use provided that all copyright, trade and service marks and other proprietary notices are maintained. By
accessing, sharing, receiving or otherwise using the Springer Nature journal content you agree to these terms of use (“Terms”). For these
purposes, Springer Nature considers academic use (by researchers and students) to be non-commercial.
These Terms are supplementary and will apply in addition to any applicable website terms and conditions, a relevant site licence or a personal
subscription. These Terms will prevail over any conflict or ambiguity with regards to the relevant terms, a site licence or a personal subscription
(to the extent of the conflict or ambiguity only). For Creative Commons-licensed articles, the terms of the Creative Commons license used will
apply.
We collect and use personal data to provide access to the Springer Nature journal content. We may also use these personal data internally within
ResearchGate and Springer Nature and as agreed share it, in an anonymised way, for purposes of tracking, analysis and reporting. We will not
otherwise disclose your personal data outside the ResearchGate or the Springer Nature group of companies unless we have your permission as
detailed in the Privacy Policy.
While Users may use the Springer Nature journal content for small scale, personal non-commercial use, it is important to note that Users may
not:

1. use such content for the purpose of providing other users with access on a regular or large scale basis or as a means to circumvent access
control;
2. use such content where to do so would be considered a criminal or statutory offence in any jurisdiction, or gives rise to civil liability, or is
otherwise unlawful;
3. falsely or misleadingly imply or suggest endorsement, approval , sponsorship, or association unless explicitly agreed to by Springer Nature in
writing;
4. use bots or other automated methods to access the content or redirect messages
5. override any security feature or exclusionary protocol; or
6. share the content in order to create substitute for Springer Nature products or services or a systematic database of Springer Nature journal
content.
In line with the restriction against commercial use, Springer Nature does not permit the creation of a product or service that creates revenue,
royalties, rent or income from our content or its inclusion as part of a paid for service or for other commercial gain. Springer Nature journal
content cannot be used for inter-library loans and librarians may not upload Springer Nature journal content on a large scale into their, or any
other, institutional repository.
These terms of use are reviewed regularly and may be amended at any time. Springer Nature is not obligated to publish any information or
content on this website and may remove it or features or functionality at our sole discretion, at any time with or without notice. Springer Nature
may revoke this licence to you at any time and remove access to any copies of the Springer Nature journal content which have been saved.
To the fullest extent permitted by law, Springer Nature makes no warranties, representations or guarantees to Users, either express or implied
with respect to the Springer nature journal content and all parties disclaim and waive any implied warranties or warranties imposed by law,
including merchantability or fitness for any particular purpose.
Please note that these rights do not automatically extend to content, data or other material published by Springer Nature that may be licensed
from third parties.
If you would like to use or distribute our Springer Nature journal content to a wider audience or on a regular basis or in any other manner not
expressly permitted by these Terms, please contact Springer Nature at

[email protected]