Page 15 - IJPS-8-1

International Journal of
Population Studies Validity and reliability of Mini-Mental State Examination in older Chinese

produces a loss of information. However, this procedure the option to choose a subset of MMSE items that are more
was commonly found in studies that used MMSE (Deb related to the study does not restrict to a complete list of
and Braganza, 1999; Osterweil, Mulford, Syndulko, et al., MMSE items. Second, the creation of the various cognitive
1994; Yi and Vaupeo, 2002). More seriously, the deriving functions allows for addressing the effects separately from
of an optimal cutoff score (e.g., Wong and Fong, 2009) has the overall cognitive function or an individual cognitive
nothing to do with the validity of the MMSE inventory and subscale. For instance, examining the effect of the overall
its dimensionality, and the relationship to the cognitive cognitive function on a medical condition and concurrently
functions but a vaguely derived score is generated that exploring the effect of the calculation subscale on the same
contains measurement errors. medical condition. Third, removing the measurement
errors is automatically inbuilt into the latent model.
The results of the factorial invariance indicated the
MMSE inventory possessed a high degree of invariance In summary, this paper recommends validation of
between the two age groups which were seldom found MMSE using EFA, CFA, and factorial invariance test
in the factorial invariance literature. The configural and showed the results of the validation that MMSE is
invariance indicated that both the young-old and old-old more appropriate than the MMSE literature that only
have the same factorial structure. This gave the conclusion concentrated on EFA. While this systematic approach
that the structural form between MMSE items and their was commonly used and already established in the
cognitive functions is the same for the two age groups. The measurement, psychology, and education literature, it is
metric invariance further provided evidence that the factor recommended for future use of MMSE. These procedures
loadings were also invariant across the two age groups. avoid all the limitations discussed in the paper. It takes into
That is the weights that indicated the association between account measurement errors, relates the MMSE items more
the MMSE items and the cognitive function construct were appropriately to the theoretical cognitive function setting,
also maintained, reflected by the equality of their respective creates competing CFA models that are appropriately set
factor loadings. Scale invariance further qualified the up before testing, and tests for invariances.
latent MMSE construct was also with the same degree of Funding
measurement errors across the two age groups. The last
invariance, the strong factorial invariance, is a prerequisite This study was not supported by any grant.
to testing for the equality of latent means. In the presence Conflicts of interest
of this invariance, the comparison of latent means becomes
unambiguous (Cheung and Rensvold, 2002), indicating No conflicts of interest were reported by all others.
that the MMSE could be used with high confidence for
latent model analysis, where systematic group differences Authors’ contributions
in means matrices are due to group differences in common TKT conducted the analysis and drafted the introduction,
factor score distributions (Yoon and Millsap, 2007). In methods, results, and discussions of the manuscript. QF
summary, the results of the factorial invariance assured the reviewed, amended, and provided recommendations on
use of the MMSE as a seven latent dimensions construct the organization of the manuscript.
and suggested that moving away from the commonly used
summated score approach which contained measurement Ethical approval
errors to latent modeling is a better direction for further The human data used in our study are a publicly available
analysis using latent models. survey dataset that can be downloaded from the webpage:
For further analyses and follow-up after validation, https://cpha.duke.edu/research/chinese-longitudinal-
the recommendation is to set CFA as the base to establish healthy-longevity-survey-clhls.
the measurement component of MMSE. When the
structural component is considered, a latent approach is Availability of supporting data
recommended to use the various latent models for further The CLHLS dataset is in open access on the webpage:
analyses. For instance, using the structural equation https://cpha.duke.edu/research/chinese-longitudinal-
model to relate MMSE items and the cognitive function healthy-longevity-survey-clhls.
constructs to a group of covariates or examining the effect
of cognitive function constructs on a medical condition. References
The latent modeling followed after CFA has several Arevalo-Rodriguez I, Smailagic N, Roqué-Figuls M, et al.
advantages. First, the number of MMSE items to include (2021). Mini-mental state examination (MMSE) for the
in an MMSE inventory becomes a researcher’s choice that early detection of dementia in people with mild cognitive

Volume 8 Issue 1 (2022) 9 https://doi.org/10.36922/ijps.v8i1.1285