Page 9 - IJPS-8-1

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

their declining health conditions (Putnick and Bornstein, practice of CLHLS users tended to sum up these items for a
2016; Revuelta, Franco-Martinez, and Ximénez, 2021; score ranging from 0 to 30. In the present paper, all the 23
Schürer, van Ophuysen, and Behrmann, 2021). items were categorized as binary.
The main purpose of this study is to find out the MMSE 2.3. Procedure of validation
factorial structure of China’s older population using the
Chinese Longitudinal Healthy Longevity Survey (CLHLS). The validation process is broadly breakdown into three
While the reliability and validity of MMSE have been major stages: First, EFA to determine the number of
carried out for each wave of the CLHLS using EFA (Gu, dimensions; second, CFA to obtain the best factor structure
2005; CLHLS, 2020), the present paper is to reveal the factor of MMSE; and last, factorial invariance testing to examine
structure of MMSE from a more refined methodological the degree of invariance concerning age. These three steps
base. Using the most recent 2018 wave of the CLHLS, this are elaborated on in the following three subsections. The
study carried out EFA and CFA, further examined the various functions from the software package R were used
factorial invariance for the young-old and old-old, and for analyses.
reported reliability levels with the consideration of the 2.3.1. EFA
binary nature of the data, using an appropriate estimation
method. EFAs were carried out to determine the number of
dimensions and the factor structure of MMSE. First,
2. Data and methods factorability was carried out to assess the adequacy of
carrying out factor analysis. Bartlett test of sphericity
2.1. Data sources (Bartlett, 1951) and Kaiser–Mayer–Olkin (KMO) measures
The data were extracted from the CLHLS, which is a of sampling adequacy (MSA) (Kaiser, 1970; Kaiser and
national longitudinal survey of Chinese older adults. Rice, 1974) were generated to provide the evidence of
The CLHLS started in the year 1998 and accomplished a factorability of factor analysis, indicating the adequacy for
total of eight waves in 1998, 2000, 2002, 2005, 2008/2009, carrying out the factor analysis and providing evidence
2011/2012, 2014, and 2018/2019. More details about the about the sufficient numbers of significant correlations
survey were published in the literature (Gu et al., 2021). among the 23 MMSE indicators to justify undertaking
The present study used the most recent wave in the year factor analysis (Pett, Lackey, and Sullivan, 2003). For
2018/2019. The 2018/2019 wave had 15,896 older adults. individual MMSE items, individual MSA was reported.
For the purposes of this study, 7603 respondents were Function KMO and cortest.bartlett from the package
dropped for missing values in any of the MMSE items, psych were used to generate these three factorability
which led to a sample size of 8293 older adults. As the indices (Revelle, 2021).
sample size is large after deletion, it has little effect on the Next, the heterogeneous correlation matrix for the
validation results. All participants involved in the survey 23 binary coded MMSE items was generated using the
consented to take part in the survey, including answering function hector from the package polycor (Fox, 2019) as
the MMSE items. The sample was further divided into the an input to carrying out EFA using the package psych,
young-old aged 65 – 79 and the oldest-old aged 80+, with function fa (Revelle, 2021). The present study used several
the former consisting of 4751 and the latter having 4118 methods that were commonly recommended in the
persons. literature (Finch, 2020) to determine the dimension of the
MMSE inventory. These methods included the Kaiser’s
2.2. Measures for MMSE inventory greater than 1 rule (Kaiser, 1960), scree test (Cattell, 1966),
The MMSE inventory contained 23 items (Folstein, Folstein, Velicer’s minimum average partial procedure criteria
and McHugh, 1975). The CLHLS developed a Chinese (map; Velicer, 1976; Velicer and Jackson, 1990), Horn’s
version of the MMSE, similar to that of Yi and Vaupel parallel analysis (Dinno, 2009; Glorfeld, 1995; Hayton,
(2002), with revised items that take into account the Chinese Allen, and Scarpello, 2004; Horn, 1965), VSS criterion
cultural and socioeconomic context making the items easily (Revelle and Rocklin, 1979), BIC (Schwartz, 1978), sample
understandable and practically answerable. These items size adjusted BIC (SABIC; Sclove, 1987), root mean square
comprise the components to measure orientation to time, error of approximation (RMSEA; Browne and Cudeck,
place, and season, short registration, delayed registration, 1992), and standardized root means square of the residuals
calculation, instruction, visualization, attention, and (RMSR; Bentler, 1995). The package psych, function VSS.
language. The content, construct, and concurrent validities scree, and VSS generated the scree plot, the VSS, RMSEA,
were covered in the various waves of CLHLS which were RMSR, and the various information criteria indicators. The
not repeated in the present paper (e.g., Gu, 2005). The usual package paran and function paran produced the parallel

Volume 8 Issue 1 (2022) 3 https://doi.org/10.36922/ijps.v8i1.1285