Validation of the Paediatric Hearing Impairment Caregiver

Experience Questionnaire in Singapore Population

Hsueh Yee Lynne

LIM
1
, MBBS, FRCS, MPH, Ling XIANG
2
, MBBS, MMed, MSc, Ruijie Li
3
, MSc, Naomi Wong
1
,
BSpPath (Hons), Chi Pun Kevin Yuen
4
, PhD
1
Otolaryngology - Head Neck Surgery Department, National University Health System,
2
National University of Singapore,

3
Health Services & Outcomes Research, National Healthcare Group,
4
Department of Special Education and Counselling, The Hong Kong
Institute of Education
Corresponding author:
LI Ruijie (Ruijie_li@nhg.com.sg)
Senior Research Analyst,
Health Services & Outcomes Research,
National Healthcare Group
Background
Parents of children with hearing loss often experience increased parenting stress associated
with problems faced by the child. Due to the unique nature of the stressors, there are
currently no suitable instruments to measure the stress, hence, the Paediatric Hearing
Impairment Caregiver Experience Questionnaire (PHICE) was developed
1
. While this
instrument has been validated in the United States, its use was deemed inappropriate for
use locally due to the varying infrastructures and culture of the hearing impaired in the 2
countries. This study therefore aims to modify and validate the PHICE questionnaire to
make it relevant locally.
The PHICE questionnaire is a 68 item instrument spanning the domains: communication,
education, emotional well-being, equipment, financial, healthcare, social, and support. The
items are scored on an 8 point Likert scale. As it takes a considerable amount of time to
complete the questionnaire, another aim of this study is to create a shorter version of the
questionnaire.
Methods & Analysis
Expert panel review
An expert panel was convened to assess each item on the questionnaire for its suitability
for use in Singapore. The expert panel consisted of an otolaryngology surgeon, an
audiologist and 2 speech-language therapists. After a round of review, 7 questions were
removed because of the focus on sign language. In the Singapore, only a limited number of
children with hearing loss are attending the Singapore School for the Deaf, the only school
whose primary medium of instruction is sign language.
Administration of the questionnaire
125 caregivers of children with permanent hearing loss for at least 6 months attending the
otolaryngology, audiology and aural rehabilitation clinic were recruited into the study.
Informed consent was sought and a total of 125 completed PHICE questionnaires were
collected between January 2006 and December 2008. The questionnaire was self-
administered.
Exploratory factor analysis
All data analysis was conducted with R
2
. The amount of missing data was small at 0.72% of
the dataset and they were imputed via cross validation. Exploratory factor analysis was
conducted through principal axis factoring using oblique rotation3 (Oblimin). Non-
graphical solutions to the scree test4 including parallel analysis, optimal coordinates and
acceleration factor were used to help decide on the appropriate number of factors to
retain. The number of factors suggested to retain is 3, 3 and 1 respectively.
The suggested factor solutions were studied but none of them yield any interpretable
solution. A more thorough search for an interpretable solution was conducted for a 4 13
factor structure. A 6 factor solution was eventually adopted as the most interpretable factor
solution. A cross-loading difference threshold of less than 0.1 between the 2 highest factors
loading was set as the criteria for removal of items and this process eventually yielded a 5
factor solution: Adapting to hearing loss, Support, Education, Healthcare, Policy
and Expectation.
Using the 5 factor solution, changes were further made to it. 6 questions were reassigned
to other factors. This was done to ensure congruence between the meanings of the factors
and the questions contained within. This would also improve interpretability of the factors.
3 questions that were removed due to high cross loadings were reintroduced. This
reintroduction was a result of feedback from clinicians that these questions were pertinent
to the care of the patient. The resultant 5 factor solution is presented in Table 1.
Confirmatory factor analysis
Confirmatory factor analysis was conducted on the original 8 factor model, the 5 factor
model suggested in the exploratory factor analysis and the modified 5 factor model. Table
2 lists the various goodness-of-fit indices proposed by Hu & Bentler
8
for assessing
goodness-of-fit indices.
From the various indices, we can see that the original 5 factor solution is a superior fit for
the data and fulfils the criteria set out by Hu & Bentler for a good fit. The changes made to
the 5 factor model increased the levels of misfits as indicated by the indices. However, this
is a trade-off that has to be made to improve interpretability. Furthermore, the change
between the original and the modified 5 factor model is deemed to be minimal.
The confirmatory factor analysis suggests that after refactoring and reducing the number of
items, the new factor structure is able to better explain the underlying phenomenon as
expressed by the data.
Internal consistency
Cronbachs was computed for the new subscales within the 5 factor structure to
determine its internal consistency. Both the value for the original and the modified 5
factor solution was computed as presented in Table 4. All subscales had an alpha value of
more than 0.73 suggesting good internal consistency. The change from the original 5 factor
solution to the modified 5 factor solution is also minimal.
Discussion
The factor structure of the original questionnaire has been changed significantly from an 8
factor structure to a 5 factor structure with only 3 overlapping factors. Given the large
reassignment of items, it may be suggested that the new structure could an artefact of
factor analysis and may not be meaningful in clinical usage. This discussion would look at
the overlapping and non-overlapping factors (Table 3) and explore the qualitative changes
in the reassignment.
Overlapping factors
The overlapping factors are as shown in Table 4. The table is formulated after removal of
items removed in the new factor structure from the old factor structure. This allows for a
fairer comparison of the changes made to the new factor structure. The number of shared
items between these shared items varies between 2 5 items suggesting that qualitatively,
the factors share some similarity.
Non-overlapping factors
5 other subscales from the old factor structure have
been regrouped into 2 subscales. The reassignments of
the subscales help better define the meaning of each
subscale by the new composition of the items and are
supported by the confirmatory analysis.
Conclusion
In conclusion, the PHICE has been revised, reorganised in
terms of the subscales composition and the resulting
instrument is deemed to be structurally valid and
internally consistent.