The Risk Factors of Laryngeal Pathology in Korean

Adults Using a Decision Tree Model

*,Haewon Byeon, *yGwangju, Republic of Korea
Summary: Objective. The purpose of this study was to identify risk factors affecting laryngeal pathology in the
Korean population and to evaluate the derived prediction model.
Study Design. Cross-sectional study.
Methods. Data were drawn from the 2008 Korea National Health and Nutritional Examination Survey. The subjects
were 3135 persons (1508 male and 2114 female) aged 19 years and older living in the community. The independent
variables were age, sex, occupation, smoking, alcohol drinking, and self-reported voice problems. A decision tree
analysis was done to identify risk factors for predicting a model of laryngeal pathology.
Results. The significant risk factors of laryngeal pathology were age, gender, occupation, smoking, and self-reported
voice problem in decision tree model. Four significant paths were identified in the decision tree model for the
prediction of laryngeal pathology. Those identified as high risk groups for laryngeal pathology included those who
self-reported a voice problem, those who were males in their 50s who did not recognize a voice problem, those
who were not economically active males in their 40s, and male workers aged 19 and over and under 50 or 60 and
over who currently smoked.
Conclusions. The results of this study suggest that individual risk factors, such as age, sex, occupation, health
behavior, and self-reported voice problem, affect the onset of laryngeal pathology in a complex manner. Based on
the results of this study, early management of the high-risk groups is needed for the prevention of laryngeal pathology.
Key Words: Decision treeVoice disorderRisk factorLaryngeal pathologyDysphonia.
INTRODUCTION
Laryngeal pathology is very frequent in the general population.
In Korea, the prevalence of laryngeal pathology has been
reported to be 7%.1 Also, considering that the lifetime
prevalence of laryngeal pathology has been reported to be
approximately 29% in a recent epidemiologic study,2
approximately 12 million Koreans experience a laryngeal
pathology at least once in their lifetime. Laryngeal pathology
had a higher prevalence than stroke in the same period in
2008.3 However, the importance of laryngeal pathology has
been overlooked in terms of public health because it does not
cause death directly. However, laryngeal pathology limits daily
life including labor activities due to functional problems in
communication, resulting in economic loss. As an example,
social costs in teachers caused by dysphonia are estimated at
$2.5 billion/y in the United States of America.4 Accordingly,
it is important to clarify risk factors clearly in laryngeal
pathology as taking precautions is more efficient than giving
treatment in reducing economic loss due to laryngeal pathology
and in decreasing the prevalence rate of laryngeal pathology.
According to the recent epidemiological studies, age,
gender, self-reported voice problem, occupation, alcohol drinking, smoking, high education level, xerostomia, and asthma
were reported as independent risk factors for dysphonia.2,58
However, the preceding studies have differences in the

Accepted for publication April 4, 2014.

From the *Department of Speech Language Pathology & Audiology, Nambu University,
Gwangju, Republic of Korea; and the ySpeech-Language Pathology Center, Nambu University, Gwangju, Republic of Korea.
Address correspondence and reprint requests to Haewon Byeon, Department of Speech
Language Pathology & Audiology, Nambu University, Cheomdan Jungangro 23,
Gwangsan-gu, Gwangju 506-706, Republic of Korea. E-mail: byeon@nambu.ac.kr
Journal of Voice, Vol. 29, No. 1, pp. 59-64
0892-1997/$36.00
2015 The Voice Foundation
http://dx.doi.org/10.1016/j.jvoice.2014.04.004

applied confounding variables for each study and they

applied regression models as a method to estimate risk
factors. Hence, they are effective in exploring independent
risk factors but have difficulties in showing the priority of
risk factors. In addition, it is likely that the assumptions that
are made in a regression model may be violated in the case
when the data of diseases or disorders are used in the model,
because linear regression models need assumptions to be
made, including assumptions about the linearity, normality,
and homoscedasticity of the data among others.
Recently, data mining, which is used in a decision tree and
neural network analysis among others, has been applied to
methods in classifying or estimating the objects of interest in
various areas including health studies, medicine, and so on. In
particular, the decision tree analysis not only can analyze
nonparametric data but also is expressed by a tree structure
for its analysis process, and the basis for prediction can be
easily understood compared with the neural network or cluster
analysis.9 In addition, decision tree model is effective to use in
clinical settings because the primary related factors, among
other several related factors, can be identified.
The aim of this study was to investigate risk factors of
laryngeal pathology and develop a prediction model for
laryngeal pathology.
METHODS
Subjects
This research targeted the general population aged 19 years or
older who had completed the 2008 Korea National Health and
Nutrition Examination Survey (KNHANES), a nationwide
survey of noninstitutionalized residents in South Korea,
and who then participated in an otolaryngology examination.
The multi-stage cluster sampling design and administration of
KNHANES are described in detail elsewhere.3 Briefly, from

60
each of the 200 sampling units, 2023 households were selected
by systematic sampling, yielding 12 528 persons in 4600
households. The field survey was conducted by specially
trained interviewers at mobile centers and in the participants
households. Among questionnaire surveys, survey on level of
education and economic activity was conducted by individual
interviews, and survey on the self-reported voice problem and
behaviors regarding health such as smoking was executed
with self-administered questionnaires from January to
December in 2008. These surveys were completed by 9308
participants. This research targeted 3626 men and women
who completed all of the health surveys, otolaryngology
surveys and examinations, and laryngoscope examinations.
Among them, 15 nonrespondents and 476 persons whose
laryngoscopic findings could not be determined were excluded
from the research, and 3135 persons (1310 men and 1825
women) were analyzed.
Measurement
Laryngeal pathology. Laryngeal pathology in this study
were included as benign vocal fold lesions (eg, vocal nodules,
vocal polyp, and vocal fold cyst), Reinke edema, laryngeal
granuloma, laryngeal keratosis, laryngeal papilloma, sulcus
vocalis, laryngitis, and suspected malignant neoplasm of the
larynx. Although a variety of definitions of the term vocal
fold lesions have been suggested, this study used the definition
suggested by Rosen et al.10 Laryngoscopic examination of
adults above the age of 19 was carried out by otolaryngologists
using 70 endoscopes, and abnormalities of the larynx were
studied. The index of coincidence evaluation was executed
twice, and the quality improvement committee reevaluated
the pictures and videos examined by the otolaryngologists
and computed the results. The index of coincidence for the
laryngoscopic examination was 75%. Voice ailment data
that were classified from the laryngoscopic examination were
reclassified as laryngeal pathology and no laryngeal
pathology by the researcher.
Demographic factors. The age, gender, education level, and
occupation were examined. Education levels were classified
into below elementary school graduation, below , below high
school graduation, and above college graduation. The occupation was classified as economically inactive (unemployed
persons, housewives, and students), nonmanual (managers,
clerical workers, and service and sales workers), and manual
(skilled agricultural, forestry and fishery workers, craft and
related trades workers, and elementary occupations).11
Healthy behavior factor. Health behavioral factors
included smoking and drinking status. Current smokers were
defined as those who had smoked 100 cigarettes or more during
their lifetime and were currently smoking. Nonsmokers were
those who had never smoked or had smoked less than 100
cigarettes in their lifetime but currently did not smoke. Alcohol
consumption was defined as less than once a week, twice or three
times a week, or above four times a week during the last 1 year.
Self-reported voice problem. Self-reported voice problems were surveyed based on the question Do you think

Journal of Voice, Vol. 29, No. 1, 2015

that you have any voice problem now? in the otolaryngology

survey. Those who currently reported having problems with
their voice were classified as having self-reported voice
problems.
Data analysis
The weighted mean, standard error, and percentile were
presented using a descriptive analysis. The weights for the
KNHANES were placed such that individuals participating in
the survey represented the entire population of Korea.3
Weighted one-way analysis of variance (ANOVA) and the
Rao-Scott chi-square test were used to compare the age, sex, education level, occupation, smoking, alcohol consumption, and
self-reported voice problem in the laryngeal pathology group
and the no laryngeal pathology group. The explanatory variables with a significance level below 0.1 were defined as the
related factors of laryngeal pathology and were included in
the decision tree model.
When the related factors of laryngeal pathology were
identified in the Rao-Scott chi-square test and weighted
one-way ANOVA, the related factors of laryngeal pathology
were statistically classified and a prediction model was
established, using a decision tree, which is a data mining
technique. The classification and regression tree (CART)
algorithm was used to predict the related factors in the decision
tree model. Measuring impurity with the Gini index, CART is
an algorithm that performs a binary split, where only two child
nodes are formed from the parent node.12 In the CART
algorithm, the alpha value for the criteria of splitting and
merging was set at 0.05. The number of parent nodes was
200 and that of child nodes was 80, and the number of branches
was limited to 5. To make up for the imbalance in data
distribution, the weights for misclassification costs were set
asymmetrically, considering the prevalence of laryngeal
pathology of Koreans.13 The validity of the model was first
tested using a misclassification table, and then the risks of the
model were compared using the 10-fold cross validation.
Decision Tree version 18.0 (SPSS Inc., Chicago, Illinois) was
used for all the analyses. Significance level was set at 0.01 in
the paired test only for the chi-square test to investigate the
related factors of laryngeal pathology and at 0.05 in the paired
test for all other analyses.
RESULTS
Characteristics of participants and prevalence of
laryngeal pathology
The general characteristics of the subjects by the presence of
laryngeal pathology are presented in Table 1. Of the entire
3135 subjects, 2887 (92.1%) were in the normal group and
248 (7.9%) were in the laryngeal pathology group. The mean
ages of the normal group and laryngeal pathology group were
43.4 and 47.9 years, respectively. The weighted one-way
ANOVA test result showed that the age of the laryngeal
pathology group was statistically significantly higher
(P < 0.001). The result of the Rao-Scott chi-square test
indicated that the normal group and laryngeal pathology group

Haewon Byeon

61

Data Mining Model for Risks of Laryngeal Pathology

TABLE 1.
The General Characteristics of the Subjects by Laryngeal Pathology
Characteristics
Age (weight mean SD)
Gender
Male
Female
Occupation
Economically-inactive
Non-manual
Manual
Smoking
Non-smoker
Ex-smoker
Current smoker
Alcohol consumption
1 time per weeks
23 times per weeks
4 times per weeks
Self-reported voice problem
No
Yes

Normal (n 2887)

Laryngeal Pathology (n 248)

P*

43.4 0.7

47.9 1.4

<0.001
0.001

1177 (90.5 1.2)

1710 (94.6 1.1)

133 (9.5 1.2)

115 (5.4 1.1)

77 (96.1 1.9)
1489 (93.9 1.1)
1246 (90.7 1.4)

4 (3.9 1.9)
108 (6.1 1.1)
128 (9.3 1.4)

1714 (93.8 1.2)

528 (92.2 1.5)
638 (90.4 1.4)

121 (6.2 1.2)

50 (7.8 1.5)
77 (9.6 1.4)

2271 (93.3 1.1)

419 (91.7 1.5)
188 (86.1 3.0)

179 (6.7 1.1)

42 (8.3 1.5)
27 (13.9 3.0)

2711 (93.3 1.0)

155 (80.1 2.9)

197 (6.7 1.0)

50 (19.9 2.9)

0.019

0.076

0.005

<0.001

Abbreviations: SD, Standard deviation.

Notes: Values in parentheses denote weighted % and standard deviation.
* Rao-Scott chi-square test.

showed a statistically significant difference in sex, occupation,

smoking, alcohol consumption, and self-reported voice
problem. The prevalence of laryngeal pathology was higher
in males than in females (P 0.001).
The prevalence of laryngeal pathology was the highest for
occupation in manual workers (P 0.019), for smoking in
current smoker (P 0.076), and for alcohol consumption
in drinkers with more than three drinks a week
(P 0.005). Also, the prevalence of laryngeal pathology
was three times higher in the group with self-reported voice
problem than no laryngeal pathology group (P < 0.001).
Because age, sex, occupation, smoking, alcohol consumption, and self-reported voice problem were the factors that
had a significant association with laryngeal pathology, they
were included in the explanatory variables of the decision
tree model.
Decision tree model of laryngeal pathology
The prediction model for laryngeal pathology using decision
tree analysis is presented in Figure 1. The variables set as the
related factors of laryngeal pathology were included in the
decision tree model, and a statistical classification model was
established using the CART algorithm. The result shows that
the classification variables that were the most significant
included sex, occupation, smoking, and self-reported voice
problem. The most significant primary prediction factor was
the recognition of a self-reported voice problem. Next, sex
was the related classification variable in the group that was
not aware of a voice problem. The third related classification
variable was age for both males and females. Fourth, alcohol

consumption was the related classification variable in females,

and occupation was the related classification variable in males.
Finally, smoking was the related classification variable in male
workers.
Table 2 is the gain diagram of the decision tree model
presenting the paths in the order of gain ratio. When deriving
the Gini index of each node to find the prediction path of
laryngeal pathology, four nodes were identified as the paths to
effectively predict etiological laryngeal pathology. The first
path, which had the highest Gini index in the prediction
of laryngeal pathology, was the group of patients who
self-reported a voice problem; in this group, 24.4% had
laryngeal pathology and their Gini index was 308.3%. The
second path with the second highest Gini index was the group
of patients who did not recognize that they had a voice problem
and were in their 50s; in this group, 16.7% had laryngeal
pathology and their Gini index was 211.7%. The third path
was the group who did not recognize a voice problem and
were male workers aged between 19 and 50 years who were
current smokers; in this group, 13.9% had laryngeal pathology
and their Gini index was 176.3%. Finally, the fourth path was
the group who did not recognize a voice problem and were
not economically active males in their 40s; in this group,
10.5% had laryngeal pathology and their Gini index was
132.6%.
When the analysis of the prediction model for laryngeal
pathology was completed, the model was tested using the
misclassification table, which compares the predicted value
and the observed value, and the misclassification table and
10-fold cross-validation test to assess the developed decision

62

Journal of Voice, Vol. 29, No. 1, 2015

FIGURE 1. The prediction model for laryngeal pathology using decision tree analysis.
tree model. The misclassification table shows that the
misclassification risk index was 0.22 and misclassification
rate was 22%. The result of the 10-fold cross validation to
compare the stability of the model indicates that the risk index
of cross validation was 0.20 and misclassification rate was 20%,
showing that the risk index of cross validation was not
significantly different from the risk index of the classification
model and misclassification rate.

DISCUSSION
Because diseases are caused by a cluster of risk factors, not by a
single risk factor, it is important for the prediction model of
diseases to identify the patterns and priority of related risk
factors. In this study, a prediction model was established to
identify the potential risk factors of laryngeal pathology and
their paths. The results show that the major prediction factors
with the potential for laryngeal pathology included age, sex,
occupation, smoking, drinking, and self-reported voice

problem. Of these factors, the most primary prediction factor

was the awareness of a self-reported voice problem. A large
number of studies have made comparisons between subjective
voice assessment and objective voice assessment and reported
contradictory results for subjective voice problems.1416
Wuyts et al16 concluded that because subjective voice
assessment and objective voice assessment are highly
associated, they are effective in the diagnosis of dysphonia.
In contrast, a dominant number of studies suggest that the
relationship between self-reported voice problem and objective
voice assessment is not significant. It was pointed out in
these studies that subjective voice assessment excessively
emphasizes the psychological aspect and cannot represent the
voice problem objectively.15,17 Hummel et al14 concluded that
self-reported voice problems and objectively reported voice
problems are not associated, and the voice problem recognized
by the subjects and the acoustic parameters are in a
mutually independent relationship. Although self-reported
voice problem is not significantly related to the objective

TABLE 2.
Gain Diagram of the Decision Tree Model
No

Node
n (%)*

Gain
n (%)y

Response %z

Gain
Index %x

1
7
16

205 (6.5)
209 (6.7)
251 (8.0)

50 (20.2)
35 (14.1)
35 (14.1)

24.4
16.7
13.9

308.3
211.7
176.3

14

143 (4.6)

15 (6.0)

10.5

132.6

* Node n (%); node number, % to 3135.

y
Gain n (%); gain number, % to 248.
z
Response (%): The fraction of the laryngeal pathology in subjects.
x
Gain index (%): 308.3 in node 9.

Description About Node

Self-reported voice problemyes
Self-reported voice problemno; sexmale; age5059
Self-reported voice problemno; sexmale; age1949, >60;
occupationnonmanual, manual; smokingcurrent smoker
Self-reported voice problemno; sexmale; age4049;
occupationeconomically inactive

Haewon Byeon

Data Mining Model for Risks of Laryngeal Pathology

assessment of voice problems, as pointed out in those studies, or

are excessively recognized compared with the objective
indices, the assessment of the subjective perception of health
is of importance in measuring the health status because it is
not only easy to measure but also has a relatively higher
influence on the form of medical treatment that patients use
than the judgment of medical professionals or the clinical
confirmation of the disease.18 In particular, it has been reported
in studies at home and abroad that self-reported voice problem
is a major risk indicator for predicting progression toward
dysphonia. According to Miller and Verdolini,19 the risk of
current voice problem was about 5.2 times higher in the
group that recognized a subjective voice problem in the past.
In an epidemiological study on the Korean community,
self-reported voice problem was a risk factor independent of
laryngeal pathology, and even after considering related factors,
those who recognize voice problem showed about a 10 times
higher risk of benign laryngeal lesion, approximately 4.7 times
higher risk of organic voice disorders, and about 3.9 times
higher risk of functional voice disorders.7,8 In summary,
although with self-reported voice problem, there is the current
limitation of not being able to identify the objective problem of
the voice, it is the most important factor for predicting
progression toward laryngeal pathology. It is thus considered
that the management of self-reported voice problem needs to
be prioritized for the prevention of laryngeal pathology.
Four significant paths were identified in the decision
tree model for the prediction of laryngeal pathology. Those identified as high-risk groups for laryngeal pathology included those
who self-reported a voice problem, those who were males in
their 50s who did not recognize a voice problem, those who
were not economically active males in their 40s, and male
workers aged between 19 and 50 or 60 and over who smoked.
Most studies in the field of dysphonia have only focused on
the individual risk factors. As documented in previous studies,
demographic factors, such as age, sex, and occupation, and
health-risk behaviors, such as smoking, were reported as
individual risk factors of dysphonia.2027 It is reported that
older age is associated with a higher risk of laryngeal
pathology,24 and the risk of benign laryngeal lesions is highest
in males aged 60 and over and in females who were in their
40s and 50s.26 These results support the prediction model of
this study, suggesting that the risk of laryngeal pathology is
higher in those who are in their 40s and over. In the meantime,
gender was also reported to be a major risk factor of dysphonia.
According to Smith et al,25 the current rate of self-reported vocal
problems in females was significantly higher than that of males,
and the past rate was similarly higher. In addition, the National
Ambulatory Medical Care Survey of the United States from
2005 to 2007 reported that the number of visits to medical institutions due to self-reported vocal problems was about twice as
high in female than males.21 However, the prediction model
for laryngeal pathology in this study was contradictory to the
findings of previous studies, because three groups out of the
fourwith a high-risk for laryngeal pathology were all identified
as male. This seems to be because the vulnerable group of
self-reported voice problems and laryngeal pathology vary de-

63

pending on gender. It is thought that the recognition rate of

self-reported voice problem is higher for females, because females assess subjective health conditions more negatively than
males and take a more preventive attitude toward health problems.28 On the contrary, health-risk factors such as smoking
and the occupation factor that can affect laryngeal pathology
were reported to be significant risk factors for vocal disorders
only in males.26 In addition, the prevalence of laryngeal pathology that included the suspected malignant neoplasm of the larynx was higher in males than females.29 Thus, although females
were identified as a more vulnerable group for self-reported
vocal problems, males are considered a more vulnerable group
to laryngeal pathologies when such risk factors as health-risk
behavior and occupational hazard were complexly combined.
Furthermore, this study suggests that the five risk factors that
include age, sex, occupation, smoking, and self-reported voice
problem identified in previous studies can individually
affect laryngeal pathology and cause laryngeal pathology in combination. Nevertheless, because most previous studies focused on
the individual risk factors, an analysis is needed to identify
multiple risk factors in future studies of laryngeal pathology.
Data mining techniques such as neural network analysis could
be used to identify the multiple risk factors of dysphonia.
This study has strength in that it identified high-risk groups of
laryngeal pathology, while considering multiple potential risk
factors. The limitations of the study are as follows. First, there
may exist other potential factors that can affect laryngeal
pathology in addition to those risk factors included in the study
model. Thus, studies are needed to investigate the risk factors of
laryngeal pathology in various aspects. Second, because this
study did not include muscle tension dysphonia, spasmodic
dysphonia, and conversion dysphonia, caution should be taken
in applying and interpreting the study model. Third, the
definition of laryngeal pathology in this study may be too broad.
As risk factors might differ according to type, the development
of a prediction model is required in the future that classifies
types of laryngeal pathology. Fourth, because this study is cross
sectional, it is difficult to make causal assumptions. Thus,
future studies should be longitudinal to explore causality.
CONCLUSION
The results of this study suggest that individual risk factors,
such as age, sex, smoking, and self-reported voice problem,
affect the onset of laryngeal pathology in a complex manner.
Based on the results of this study, early management of highrisk groups is needed for the prevention of laryngeal pathology.
Acknowledgments
The author would like to acknowledge the Korea Center for
Disease Control and Prevention that provided the raw data for
analysis.
REFERENCES
1. Byeon H. Association between occupation and self-reported voice problems
in the general population. Korean J Commun Disord. 2011;16:360371.
2. Cohen SM. Self-reported impact of dysphonia in a primary care population:
an epidemiological study. Laryngoscope. 2010;120:20222032.

64
3. Ministry of Health and Welfare. The Korea National Health and Nutritional
Examination Survey 2007-2009. Seoul, South Korea: Ministry of Health
and Welfare; 2010.
4. Verdolini K, Ramig LO. Review: occupational risks for voice problems.
Logoped Phoniatr Vocol. 2001;26:3746.
5. Roy N, Stemple J, Merrill RM, Thomas L. Epidemiology of voice disorders in
the elderly: preliminary findings. Laryngoscope. 2007;117:628633.
6. Roy N, Merrill RM, Gray SD, Smith EM. Voice disorders in the
general population: prevalence, risk factors, and occupational impact.
Laryngoscope. 2005;115:19881995.
7. Byeon H, Lee Y. Prevalence and risk factors of benign laryngeal lesions in
the adult population. Korean J Commun Disord. 2010;15:648656.
8. Byeon H. A study on the relationship between the self-reported voice
problems and voice disorders in the adult populations. Phon Speech Sci.
2011;3:111116.
9. Tan P, Steinbach M, Kumar V. Introduction to data mining. Boston, MA:
Addison Wesley; 2006.
10. Rosen CA, Gartner-Schmidt J, Hathaway B, Simpson CB, Postma GN,
Courey M, Sataloff RT. A nomenclature paradigm for benign
midmembranous vocal fold lesions. Laryngoscope. 2013;122:13351341.
11. Kim KH, Lee KH, Lee SM, et al. The proportional mortality ratios of
specific-cause mortality by occupationa and education among men aged
20-64 in korea (1993-2004). J Prev Med Public Health. 2007;40:715.
12. Breiman L, Friedman JH, Olshen RA, Stone CJ. Classification and
regression trees. Belmont, CA: Wadsworth; 1984.
13. Cristianini N, Shawe-Taylor J. An introduction to support vecter machins.
Cambridge, UK: Cambridge Univerity Press; 2000.
14. Hummel C, Scharf M, Schuetzenberger A, Graessel E. Objective voice
parameters and self-perceived handicap in dysphonia. Folia Phoniatr
Logop. 2010;62:303307.
15. Wilson JA, Deary IJ, Millar A, Meckenzie K. The quality of life impact of
dysphonia. Clin Otolaryngol. 2002;27:179182.
16. Wuyts FL, De Bodt MS, Molenberghs G, et al. The dysphonia severity
index: an objective measure of vocal quality based on a multiparameter
approach. J Speech Lang Hear Res. 2000;43:796809.

Journal of Voice, Vol. 29, No. 1, 2015

17. Lehto L, Laaksonen L, Vilkman E, Alku P. Occupational voice complaints
and objective acoustic measurements-do they correlate? Logoped Phoniatr
Vocol. 2006;31:147152.
18. Hunt M, McEwen J, McKenna SP. Perceived health: age and sex comparisons in a community. J Epidemiol Community Health. 1984;38:
156160.
19. Miller MK, Verdolini K. Frequency and risk fators for voice problems in
teachers of singing and control subjects. J Voice. 1995;9:348362.
20. Bermudez de Alvear RM, Baron FJ, Martnez-Arquerom AG.
School teachers vocal use, risk factors, and voice disorder prevalence:
guidelines to detect teachers with current voice problems. Folia Phoniatr
Logop. 2011;63:209215.
21. Best SR, Fakhry CF. The prevalence, diagnosis, and management of voice
disorders in national ambulatory medical care survey(NAMCS) cohort.
Laryngoscope. 2011;121:150157.
22. Chen SH, Chiang SC, Chung YM, Hsiao LC, Hsiao TY. Risk factors and
effects of voice problems for teachers. J Voice. 2010;24:183192.
23. Feierabend RH, Shahram MN. Hoarseness in adults. Am Fam Physician.
2009;80:363370.
24. Herrington-Hall BL, Lee L, Stemple JC, Niemi KR, McHone MM.
Description of laryngeal pathologies by age, sex, and occupation in a
treatment-seeking sample. J Speech Hear Disord. 1988;53:5764.
25. Smith E, Kirchner HL, Taylor M, Hoffman H, Lemke JH. Voice problems
among teachers: differences by gender and teaching characteristics. J Voice.
1998;12:328334.
26. Byeon H, Hwang Y. Gender differences in risk factors of self-reported voice
problems. Phon Speech Sci. 2012;4:99108.
27. Marcotullio D, Magliulo G, Pezone T. Reinke s edema and risk factors:
clinical and histopathologic aspects. Am J Otolaryngol. 2002;23:8184.
28. Han K, Lee J, Ryff C, Marks N, Ok S, Cha S. Health status and health
behavior of middle-aged Korean men and women. J Korean Home Econ.
2003;41:213229.
29. Byeon H, Lee Y. Laryngeal pathologies in older Korean adults and their
association with smoking and alcohol consumption. Laryngoscope. 2013;
23:429433.