Академический Документы
Профессиональный Документы
Культура Документы
The Research Unit for General Practice, Centre of Health and Society, University of Copenhagen, Copenhagen, Denmark
Department of General Practice, University of Copenhagen, Copenhagen, Denmark
c
Department of Biostatistics, University of Copenhagen, Copenhagen, Denmark
d
Section for Health Informatics, University of Aarhus, Aarhus, Denmark
e
Research Unit of General Practice, Department of Public Health, University of Southern Denmark, Jagtvej 20A, DK-8270 Hjbjerg, Denmark
b
A R T I C L E I N F O
A B S T R A C T
Article history:
Received 30 August 2015
Received in revised form 8 November 2015
Accepted 13 November 2015
Available online 24 November 2015
Purpose: We tested the Galker test, a speech reception in noise test developed for primary care for
Danish preschool children, to explore if the childrens ability to hear and understand speech was
associated with gender, age, middle ear status, and the level of background noise.
Methods: The Galker test is a 35-item audio-visual, computerized word discrimination test in
background noise. Included were 370 normally developed children attending day care center. The
children were examined with the Galker test, tympanometry, audiometry, and the Reynell test of verbal
comprehension. Parents and daycare teachers completed questionnaires on the childrens ability to hear
and understand speech. As most of the variables were not assessed using interval scales, non-parametric
statistics (GoodmanKruskals gamma) were used for analyzing associations with the Galker test score.
For comparisons, analysis of variance (ANOVA) was used. Interrelations were adjusted for using a nonparametric graphic model.
Results: In unadjusted analyses, the Galker test was associated with gender, age group, language
development (Reynell revised scale), audiometry, and tympanometry. The Galker score was also
associated with the parents and day care teachers reports on the childrens vocabulary, sentence
construction, and pronunciation. Type B tympanograms were associated with a mean hearing 56 dB
below that of than type A, C1, or C2. In the graphic analysis, Galker scores were closely and signicantly
related to Reynell test scores (Gamma (G) = 0.35), the childrens age group (G = 0.33), and the day care
teachers assessment of the childrens vocabulary (G = 0.26).
Conclusions: The Galker test of speech reception in noise appears promising as an easy and quick tool for
evaluating preschool childrens understanding of spoken words in noise, and it correlated well with the
day care teachers reports and less with the parents reports.
2015 Elsevier Ireland Ltd. All rights reserved.
Keywords:
Speech recognition in noise test
Day care children
Language development
Otitis media
Hearing
Non-parametric statistics
1. Introduction
The effect of the hearing loss often accompanying otitis media
with effusion (OME) on childrens development has been a focus
of intensive research, and the results of this research are diverging
[14]. A minor hearing loss accompanying OME may be debilitating
54
2.4. Hearing
Conditioned play audiometry was employed to test air
conduction with pure tones at 500, 1000, 2000, and 4000 Hz via
earphones according to described techniques [12]. Testing was
performed in a quiet room with an AS216 Interacoustics screening
audiometer. The validity of each session was ranked on a 14-point
scale based on the childs concentration, cooperativeness, and
perceived ambient noise level with 1 representing high validity.
Sessions with validity rankings of 1 or 2 with at least 2 measured
frequencies in both ears were kept for data analysis. A summary
index of hearing level (HL) for each ear was calculated by averaging
thresholds. Children were categorized according to the average
threshold of the best ear. Four groups were dened: HL less than
22 dB, 2225, 2629 dB, and 30 dB or more. Children who were
unable to carry out audiometry were categorized in a separate
group and included as a separate group throughout analysis.
In the graphic analysis, some of the variables were collapsed to
even fewer groups to be able do a more robust analysis; for
example, hearing groups were reduced to only three groups.
55
block includes the Galker test and the Reynell test scores and the
parents and the day care teachers evaluations of childs hearing
and speech comprehension.
The partial g-coefcient (GoodmanKruskal gamma coefcient) for ordinal categorical variables was calculated as a measure
of conditional dependence [19]. The gamma coefcient is a
measure of correlation based on the difference between concordant pairs (C) and discordant pairs (D). Gamma is computed as:
gamma = (C D)/(C + D). Thus gamma is the surplus of concordant
pairs over discordant pairs, as a percentage of all pairs, ignoring
ties. Two ordinal variables are associated with the extent to which
a relatively high value on one variable can be predicted from a
relatively high value on the other variable. The gamma coefcient
variates between one and minus one; a high value means
concordant (relation) and a low value means discordant (no
relation). Gamma is equivalent to Kendalls Tau, but gamma
statistics is preferable to Kendalls Tau when the data contain many
tied observations.
2.9. Analysis
3. Results
In all analyses, total scores based on the Rasch-validated
35 items of the Galker test were used [1]. For examination of the
noise effect on the Galker test, the data of children tested twice
were analyzed using a variance analysis in which the difference
between the Galker scores on the rst and the second test was used
as response variable. Presence of noise, child age, Reynell score,
and sex were used as explaining variables.
For examination of associations with the Galker test scores, we
used a non-parametric graphical model from a subclass of loglinear statistical models for multidimensional contingency tables.
In the graphical model, any relation is analyzed as a conditional
relation. This implies that a relation between two variables is
revealed only if the association remains signicant after controlling for the effect of any preceding or intervening variables. An
advantage of graphical modeling is the possibility of simultaneous
control for all other variables [17].
In a recursive graphical model, variables have been ordered in a
structure representing temporal or causal assumptions about the
relation between the variables [18]. In this study, the variables
were ordered in three recursive blocks assuming a causal structure.
Hence, variables from the rst block are assumed to potentially
inuence variables in the second and third block, but not vice
versa. The rst block includes the following variables: the childs
sex, age, result of tympanometry, and the parents educational
level. The second block includes results of audiometry. The third
Table 1
Questions to parents and teachers included in the graphical model.
How does the/your childs pronunciation compare with that of other children
at the same age?
Much ahead of/slightly ahead of/corresponding to/slightly delayed/very delayed
How is the/your childs vocabulary compared with that of other children at
the same age?
Much larger/slightly larger/corresponding to/slightly smaller/much smaller
How is the/your child able to construct sentences compared with other
children at the same age?
Much ahead of/slightly ahead of/corresponding to/slightly delayed/very delayed
How much of a problem do you think difculty of hearing has been to
the/your child during the past 4 weeks?
No problem/minor/moderate/quite a bit of/very much of a problem
Does the child have difculty hearing messages given to children in a group?
(teachers only)
No, never/rarely/sometimes/yes, often/yes, very often
Does the child have difculty following instructions in group activities
or games? (teachers only)
No, never/rarely/sometimes/yes, often/yes, very often
Overall, how much has your child been bothered by ear-problems?
(parents only)
Not at all/a little/moderately/very much/extremely
3.1. General
The 370 children represented an unselected sample of normally
developed Danish preschool children. However, one fourth of the
invited families did not respond to the invitation to participate in
the study. The non-participants did not differ from the participants
with regard to the gender ratio, but they tended to be slightly
younger. The day care teachers general impression was that the
non-participating children did not differ from the participating
children with regard to health, development, or behavior.
However, specic data on the non-participants could not be
collected.
The inclusion criterion was 34 years of age, but 30 children
turned 5 years during the intake procedure. They were kept in the
study, and the participating childrens mean age was thus 4.1 years
(standard deviation, SD = 0.65).
Table 2 presents background information on the participating
children and their families.
In general, the children had good health according to the
parents. About 28% were rated to have been bothered by moderate
to extreme ear problems during childhood. Almost just as many
(25%) had been treated with ventilation tubes at some point.
Results of tympanometry showed a high frequency of middle ear
pathology. Thus, only 178 (48%) of the children had normal
tympanometric curves (A or C1 curves) in both ears, 82 (22%) had
middle ear effusion in at least one ear (B curve) and 102 (28%) had
at least one ear with a C2 curve or a ventilation tube inserted.
Bilateral B curves were observed in 24 (6%) (Table 3).
Peripheral hearing acuity was measured by conditional play
audiometry down to a screening level of 20 dB. Although the
children were carefully conditioned to the procedure, overall 19%
remained unable to accomplish the procedure. As expected, this
was a particular problem among 3-year-olds (42%). Among 33% of
the children who had bilateral B curves, a hearing test was missing.
Among the children with bilateral B curves who successfully
completed audiometry, the hearing threshold was 30 dB or worse
on the best ear in 38% of the children. Overall, for children with a
complete examination including audiometry, the hearing level was
25 dB or worse bilaterally in 23 (6%) (Table 3).
3.2. The effect of background noise on the Galker test results
A subgroup of 107 children was retested with the Galker test.
Sixty-seven of these children were retested with the normal
background noise (Noise group) and 40 children were tested
56
Table 2
Characteristics of participants, N = 370.
N (%)
Age
68 (18)
3.0<3.5
3.5<4.0
80 (22)
4.0<4.5
90 (24)
4.5<5.0
102 (28)
5.0<6.0
30 (8)
Sex
Female
186 (50)
Male
184 (50)
Education
Mothers: Less than high school
57 (17)
Fathers: Less than high school
85 (23)
Siblings
One sibling
226 (61)
80 (22)
More than one sibling
Bilingual
35 (9)
Ever received speech therapy
38 (10)
Childs general health, rated by parents
Good or very good
345 (93)
Less good
15 (4)
10 (3)
Question unanswered
Child been bothered by ear problems, rated by parents
Not at all or a little
253 (68)
Moderately
62 (17)
Very much or extremely much
41 (11)
Question unanswered
14 (4)
Diagnosed with OME by doctor, ever
150 (41)
Ever treated with ventilation tubes
94 (25)
Table 3
Descriptive results of examinations, N = 370.
Otoscopy
Ventilation tubes present
Tympanometry in best ear
A
C1
C2
B/B
Grommets or perforation bilateral
Missing
Audiometry, hearing in best ear
2021 dB
2225 dB
2630 dB
30 dB or more
Missing
Reynell score by age groups
3.0<3.5
3.5<4.0
4.0<4.5
4.5<5.0
5.0<6.0
Missing
Galker score by age groups
3.0<3.5
3.5<4.0
4.0<4.5
4.5<5.0
5.0<6.0
Mean (SD)
370
4.1 (0.65)
52 (14)
165
92
58
24
25
6
(45)
(25)
(15)
(6)
(7)
(2)
119
120
39
31
61
(32)
(32)
(11)
(8)
(17)
40
Age, years
Number (%)
Mean (SD)
63
77
87
101
29
13
(17)
(21)
(24)
(27)
(8)
(3)
47
51
55
57
61
(8.3)
(6.8)
(5.7)
(5.9)
(3.4)
68
80
90
102
30
(18)
(22)
(24)
(28)
(8)
17
20
24
25
26
(7.3)
(6.6)
(3.7)
(4.6)
(4.3)
30
20
10
10
20
30
Noise on test 2
40
57
Table 4
Description of the effect of noise on Galker test scores.
Group
With noise
Test 2
No noise
Test 2
Noise group
compared with
No noise group
p-values
*
Girls%
67
60
40
53
Age,
years (SD)
4.1
(0.6)
4.6
(0.6)
Reynell score
Age-adjusted*
Mean (SD)
0.9
(5.8)
0.5
(5.0)
0.31
Mean (SD)
Median (IQR)
Mean (SD)
Median (IQR)
Mean (SD)
Median (IQR)
22.8
(5.2)
23.9
(4.5)
0.26
24
(6)
24
(6)
24.1
(5.8)
30.8
(3.4)
<0.0001
25
(6)
32
(4)
1.3
(3.0)
6.9
(3.4)
<0.0001
1
(4)
6
(3.5)
partial relation was also seen between the Galker test and the day
care teachers ratings of the childrens vocabulary attainment
(g = 0.26). This relation was not explained by the Reynell test,
whereas the Galker test explained the associations between the
Reynell test and teacher-rated vocabulary attainment. In other
words, even when information about performance on the Reynell
test was at hand, the Galker test still contributed more information
than the day care teachers vocabulary attainment rating.
Strong, unadjusted relations (Table 5) were also seen between
the Galker test and the day care teachers answers to all the other
language-related questions, in particular their rating of the childs
ability to follow instructions during group activities. These
associations were mediated in particular through the Reynell test
(g = 0.24) (Fig. 2). The analysis indicates that the Galker test and
the Reynell test measure two different, but associated phenomena.
Associations between all language-related questions and both
the Galker test and the Reynell test were closer for the teachers
ratings than for the parents ratings. The parents report on
language achievements added no value to the analyses when
teachers rating are known (Fig. 2).
3.5. Relations between the Galker test, ear status, and hearing-related
variables
Due to the missing data on audiometry for 61 children (17%),
any interpretation of the importance of this variable must be very
cautious. As noted in the Methods section, children with missing
data on audiometry were included as a separate group for analysis.
Table 5
Marginal (unadjusted) relation between variables of the graphical analysis and the Galker test groups.
Variable (number of groups)
Gamma coefcient
p-value
ANOVA F-value
p-value
0.19
0.49
0.08
0.11
0.20
0.30
0.59
0.11
0.26
0.35
0.37
0.34
0.34
0.09
0.03
0.15
0.15
0.13
0.005
<0.0001
0.116
0.034
0.011
<0.0001
<0.0001
0.13
<0.0001
<0.0001
<0.0001
<0.0001
<0.0001
0.26
0.32
0.01
0.01
0.02
6.6
27.0
2.4
2.8
2.2
12.6
57.4
2.7
5.8
11.0
14.1
11.7
9.3
1.1
1.7
6.5
4.5
2.4
0.01
<0.0001
0.07
0.04
0.06
<0.0001
<0.0001
0.048
<0.0001
<0.0001
<0.0001
<0.0001
<0.0001
0.36
0.14
<0.0001
0.001
0.05
58
Fig. 2. The graphical model of the relation between the Galker test, the Reynell test, audiometry, tympanometry, parents reports, day care teachers reports, and background
variables.
Partial (adjusted) g-coefcients are placed on the gure.
3.6. Relation between the Galker test, the childs age and gender, and
the parents educational level
A partial, highly signicant association between the Galker test
and the age of the child was demonstrated (gamma = 0.33) (Fig. 2).
A marginally, signicantly higher Galker score was seen among
girls; and a marginally, just signicant, higher score was seen
among children of fathers with a higher educational level (Table 5).
4. Discussion
4.1. The main ndings
The Galker test of speech reception in noise is a rating scale
designed to serve as a simple method for use in primary care to
quantify the impact of hearing problems often caused by middle
ear effusions in preschool children. No true gold standard exists for
this method, and its validity and characteristics have been
investigated from various angles. In a previous paper [1], the
internal construct validity was examined with the Rasch model.
Construct validity and reliability were demonstrated for 35 items
of the scale.
The present paper further explored the validity of the Galker
test by examining its associations with measures of middle ear
59
5. Conclusion
The present paper demonstrates the signicant inuence
background noise has on the Galker test of speech reception in
noise. A strong, signicant relation was found between the Galker
test and language comprehension evaluated by speech therapists,
and a signicant association was found between the Galker test
and language development evaluated by teachers in day care
centers. A weak, but signicant association was found between the
Galker test and middle ear status (bilateral type B tympanometry)
and audiometry. The relation with peripheral hearing was awed
by 17% missing tests.
The Galker test of speech reception in noise appears promising
as a tool in primary care for evaluating the ability of the preschool
child to hear and understand speech. Further research of its
validity is needed, in particular long-term follow-up studies to
determine the predictive value of the test and to further examine
clinical subgroups with known difculties.
Acknowledgements
We wish to thank the participating families and the staff from
the day-care centers; speech and hearing therapist Eve Galker for
her year-long participation in developing the Galker test, and for
her pilot-testing of the Galker test in day-care centers around
Aarhus. Willy Karlslund is acknowledged for technical assistance
60
[17]
[18]
[19]
[20]
[21]
[22]
[23]
[24]
[25]
[26]
[27]
[28]
[29]
[30]
[31]
[32]
[33]
[34]
[35]
[36]
[37]
[38]
[39]