Comparison Between the Mini-Mental State Examination and the Mini-Mental State Examination-2 in Patients With Mild Cognitive Impairment and Alzheimer’s Disease

Abstract

Background

This study aimed to compare the diagnostic utility of the Mini-Mental State Examination-2 (MMSE-2) and the Korean version of the Mini-Mental State Examination (K-MMSE) in differentiating normal cognitive aging, mild cognitive impairment (MCI), and Alzheimer’s disease (AD) within a Korean population.

Methods

A total of 226 individuals with MCI, 97 with AD, and 91 cognitively healthy older adults were recruited. Participants underwent the MMSE-2, K-MMSE, and a comprehensive neuropsychological assessment battery. Discriminant analysis was employed to compare the classification accuracy of each tool, while sensitivity and specificity were evaluated using receiver operating characteristic (ROC) curve analysis.

Results

Discriminant analysis revealed that the MMSE-2 accurately classified 71.1% of participants, including 68.6% of MCI patients, 78.4% of AD patients, and 72.5% of healthy controls. In contrast, the K-MMSE achieved an overall classification accuracy of 67.9%, with 83.6% accuracy for MCI, 68.0% for AD, and 28.6% for healthy controls. ROC analysis indicated that the area under the curve (AUC) values for the MMSE-2: Brief Version (BV) (0.708), Standard Version (SV) (0.720), and Expanded Version (EV) (0.728) surpassed that of the K-MMSE (0.703) in distinguishing healthy controls from MCI patients. However, the K-MMSE (AUC = 0.936) demonstrated superior performance compared to the MMSE-2:BV (0.930), MMSE-2:SV (0.925), and MMSE-2:EV (0.921) in differentiating MCI from AD.

Conclusion

The MMSE-2:SV and MMSE-2:EV exhibit greater sensitivity in detecting cognitive impairment between normal aging and MCI. Conversely, the MMSE-2:BV and K-MMSE demonstrate superior sensitivity in distinguishing between MCI and AD. These findings underscore the importance of selecting an appropriate cognitive assessment tool based on specific diagnostic objectives and clinical contexts.

Graphical Abstract

INTRODUCTION

The Mini-Mental State Examination (MMSE)1 is one of the most widely used cognitive screening tools worldwide. It is a quick and straightforward measure of various cognitive functions, with well-established validity and reliability in detecting moderate to severe stages of dementia.1,2,3 However, the MMSE demonstrates limited sensitivity in identifying mild cognitive impairment (MCI) or early stages of dementia. Additionally, it is not well-suited for detecting deficits in frontal executive function.4,5 Furthermore, a ceiling effect is often observed in individuals with higher levels of education or those with MCI, as the items of MMSE difficulty are relatively low.6

The Mini-Mental State Examination, 2nd Edition (MMSE-2)7 was developed to address the limitations of the original MMSE as a cognitive screening tool. Recently, translated into Korean, the MMSE-2 has demonstrated high reliability and validity in distinguishing healthy older adults from individuals with MCI and Alzheimer’s disease (AD). Moreover, the three versions of the MMSE-2—the Brief Version (MMSE-2:BV), the Standard Version (MMSE-2:SV), and the Expanded Version (MMSE-2:EV)—have demonstrated relatively high sensitivity and specificity in distinguishing between healthy older adults, patients with MCI, and those with AD. While the MMSE-2 has proven to be a clinically valuable cognitive screening tool in Korea, its sensitivity in distinguishing MCI from normal cognitive aging may not be as high as initially anticipated.8

The aim of this study was to compare the utility of the MMSE-2 and the Korean version of the Mini-Mental State Examination (K-MMSE) in determining which test is more sensitive in differentiating normal cognitive aging from patients with MCI and AD within a Korean population.

METHODS

Participants

Patients

A total of 323 patients who consecutively visited the Department of Neurology at Seoul National University Bundang Hospital between June 2012 and April 2013 were enrolled in this study. Of these, 226 patients were diagnosed with MCI, and 97 were diagnosed with AD. The diagnosis of MCI was based on Petersen’s criteria,9 while the diagnosis of “probable AD” followed the criteria established by the National Institute of Neurological and Communicative Disorders and Stroke and the Alzheimer’s Disease and Related Disorders Association.10 This study involves a reanalysis of data previously utilized in Baek et al.8

Control participants

A total of 91 healthy adults participated in this study between June 2012 and April 2013. Participants were either caregivers of patients visiting Seoul National University Bundang Hospital or individuals recruited from a healthcare center. They had no subjective memory complaints, no history of any of 29 exclusionary medical conditions, and no indications of cognitive decline.11 Additionally, the K-MMSE5 was administered, and their scores fell within the normal range based on age and educational level. The Korean Instrumental Activities of Daily Living (K-IADL)12 scale was also used to assess daily living activities, with all participants scoring below 0.43, the threshold distinguishing healthy older adults from those with dementia. This study represents a reanalysis of data previously utilized in Baek et al.8

Instruments

MMSE-2

The copyright for the MMSE-2 is held by Psychological Assessment Resources (PAR), Inc. Prior to the commencement of this study, PAR granted us permission to translate the MMSE-2 into Korean. Furthermore, PAR generously authorized the complimentary use of 520 test forms (260 Red Forms and 260 Blue Forms) for research purposes. This agreement was formalized on March 25, 2011. The MMSE-2 has also been validated as a reliable and effective screening tool for assessing cognitive impairment in Korean populations.8

The MMSE-2 includes two alternative forms, the red form and the blue form, designed to mitigate learning effects that may arise with repeated administration. Additionally, the MMSE-2 is available in three versions: the MMSE-2:BV, the MMSE-2:SV, and the MMSE-2:EV. The nine subtests included in the MMSE-2 are described in detail below.

The MMSE-2:BV (16 points) consists of four subtests administered in the following order: registration (3 points), orientation to time (5 points), orientation to place (5 points), and recall (3 points). The MMSE-2:SV (30 points) includes seven subtests: attention and calculation (5 points), language (8 points), drawing (1 point), along with the four subtests from the MMSE-2:BV. The MMSE-2:EV (90 points) comprises nine subtests, adding story memory (25 points) and processing speed (35 points) to the seven subtests from the MMSE-2:SV.

K-MMSE

The K-MMSE, a Korean adaptation of the MMSE developed by Kang et al.,5 was designed to closely follow the format of the original MMSE, with a total score of 30 points. Additionally, large-scale normative data for the Korean population have been established, supporting the validity and utility of the K-MMSE as a cognitive screening tool.13

While the K-MMSE retains most items from the original MMSE,1 some modifications have been made. For example, orientation is divided into separate categories for time and place. The overlapping pentagons used in the MMSE as part of the "Copy" task are included as a distinct item in the K-MMSE under "Visual construction." Furthermore, while the MMSE assesses "Attention and calculation" using both the serial 7s subtraction test and spelling the word "WORLD" backward, the K-MMSE includes only the serial 7s test.

As a result, the K-MMSE comprises seven subtests, administered in the following order: orientation to time (5 points), orientation to place (5 points), registration (3 points), attention and calculation (5 points), recall (3 points), language (8 points), and visual construction (1 point).

Other neuropsychological assessments

The MMSE-2 and K-MMSE were administered with a minimum interval of one hour to mitigate potential learning effects associated with performing both tests consecutively. During this interval, detailed neuropsychological assessments were conducted.

The neuropsychological battery included the Seoul Verbal Learning Test (SVLT)14,15 for evaluating verbal memory, the Rey Complex Figure Test (RCFT; copy)16 for assessing visuospatial function, the Semantic Word Fluency Test (SWF) and Phonemic Word Fluency Test (PWF)17 for evaluating frontal lobe executive functions, and the Korean-Color Word Stroop Test18 for assessing executive function. Naming ability was measured using the Korean version of the Boston Naming Test (K-BNT),19 while attention was evaluated with the forward and backward Digit Span Test.20

In addition to these domain-specific assessments, global cognitive and functional status were measured using the Clinical Dementia Rating (CDR)21 and CDR Sum of Boxes.22,23 This comprehensive approach provided a robust evaluation of participants’ cognitive and functional abilities.

Procedure

The MMSE-2 and K-MMSE included overlapping subtests, such as orientation to time, orientation to place, attention and calculation, writing, and drawing. These overlapping subtests were not repeated, and the scores from the first administration were applied equally to both tests. However, the attention and calculation subtest were administered twice because it required participants to learn and recall three words between tests, though only the score from the first administration was used for analysis. When identical subtest items were present in both tests but differed in content or format, responses were collected and scored separately.

Statistical analysis

First, an analysis of variance was conducted to compare age and education levels across the three groups (226 patients with MCI, 97 patients with AD, and 91 healthy older adults). Additionally, a χ² test was used to compare gender distributions among the groups.

Second, the results of the detailed neuropsychological assessments across the three groups were analyzed using analysis of covariance (ANCOVA), with age and education included as covariates to control for demographic differences.

Third, the results of the MMSE-2 (BV, SV, and EV) across the three groups were analyzed using ANCOVA, with age and education included as covariates to control for demographic differences.

Fourth, the results of the K-MMSE across the three groups were analyzed ANCOVA, with age and education included as covariates to account for demographic differences.

Fifth, discriminant analysis was conducted to evaluate the classification accuracy and ability of the MMSE-2 and K-MMSE to differentiate between patients with MCI and AD.

Finally, the sensitivity and specificity of the MMSE-2 and K-MMSE in differentiating among the three groups were evaluated using receiver operating characteristic (ROC) curve analysis and area under the curve (AUC) measurements.

All data were analyzed using SPSS version 18.0 (SPSS Inc., Chicago, IL, USA), with a significance threshold set at P < 0.05 for all statistical tests.

Ethics statement

This study was approved by the Institutional Review Board of Seoul National University Hospital (Approval Number: B-1612-376-111). Written informed consent was obtained from all participants prior to study procedures. For patients with AD who had impaired capacity to provide consent, written informed consent was secured from their legal caregivers in accordance with ethical guidelines.

RESULTS

The participants’ demographic data

The mean age of the patients with MCI (n = 226) was 71.05 ± 7.73 years (range: 70–72 years), while the mean age of the patients with AD (n = 97) was 75.38 ± 7.60 years (range: 73–77 years). The mean age of the healthy older adults (n = 91) was 67.05 ± 7.55 years (range: 65–69 years). The mean years of education were 11.45 ± 4.80 years (range: 10–12 years) for the MCI group, 9.63 ± 5.15 years (range: 8–11 years) for the AD group, and 10.98 ± 5.21 years (range: 9–12 years) for the healthy older adult group.

There were no significant differences in gender distribution among the participants, χ²(1, 414) = 1.76, P = 0.415. However, significant differences were observed in age, F(2, 411) = 27.82, P < 0.001, and years of education, F(2, 411) = 4.38, P = 0.013, across the three groups. Tukey’s post hoc analysis revealed that the mean age of patients with AD was significantly higher than that of patients with MCI and healthy older adults. Additionally, the mean age of patients with MCI was significantly higher than that of healthy older adults.

In terms of education, the mean number of years of education for patients with MCI was significantly higher than for patients with AD. No significant differences in years of education were observed between patients with AD and healthy older adults or between patients with MCI and healthy older adults. The demographic data are summarized in Table 1.

Table 1. Characteristics of participants.

Variables	All participants (N = 414)
	Normal (n = 91)	MCI (n = 226)	AD (n = 97)
Age, yr	67.05 ± 7.55	71.05 ± 7.73a	75.38 ± 7.60b
Education, yr	10.98 ± 5.21	11.45 ± 4.80	9.63 ± 5.15c
Male/Female	29/62	90/136	36/61

Values are presented as mean ± standard deviation.

MCI = mild cognitive impairment, AD = Alzheimer’s disease.

^aP < 0.001 for MCI vs. Normal.

^bP < 0.001 for AD vs. MCI and Normal.

^cP = 0.009 for AD vs. MCI.

The results of participants’ neuropsychological assessments

The neuropsychological assessment results were compared across the three groups (MCI, AD, and healthy older adults). Significant differences were observed among the groups in all assessed cognitive domains, including attention, verbal memory, visuospatial function, language function, and frontal/executive function (P < 0.05 for all domains).

Tukey’s post hoc analysis indicated that healthy older adults scored significantly higher than both patients with MCI and AD across the MMSE, SVLT, RCFT (copy), SWF, PWF, Stroop Color-Word Test (color naming), K-BNT, and forward and backward digit span tests. Additionally, patients with MCI scored significantly higher than patients with AD on these same measures. However, for the Stroop Color-Word Test (word reading), no significant difference was found between healthy older adults and patients with MCI, but both groups scored significantly higher than patients with AD. The mean subtest scores for each group and the results of Tukey’s post hoc analyses are detailed in Table 2.

Table 2. The results of neuropsychological assessments in the three groups.

Neuropsychological assessments	Normal	MCI	AD	P	F	Post hoc
K-MMSE	27.31 ± 2.30	25.75 ± 2.46	19.31 ± 3.80	< 0.001	229.273	N > M > A
SVLT-immediate recall	21.47 ± 4.06	17.15 ± 4.39	10.91 ± 3.71	< 0.001	111.870	N > M > A
SVLT-delayed recall	7.26 ± 2.00	3.49 ± 2.92	0.29 ± 0.92	< 0.001	152.141	N > M > A
SVLT-recognition	9.46 ± 1.68	7.42 ± 2.42	3.70 ± 2.87	< 0.001	104.727	N > M > A
RCFT-copy	32.41 ± 3.87	28.87 ± 5.32	22.50 ± 7.63	< 0.001	66.910	N > M > A
SWF-animal	17.40 ± 4.14	12.86 ± 3.75	8.84 ± 4.31	< 0.001	80.743	N > M > A
SWF-supermarket items	18.97 ± 5.32	14.46 ± 5.84	8.33 ± 4.57	< 0.001	61.570	N > M > A
PWF-ㄱ, ㅇ, ㅅ	26.35 ± 11.01	20.41 ± 9.78	13.83 ± 8.95	< 0.001	27.506	N > M > A
K-CWST-reading	110.89 ± 5.98	111.13 ± 4.03	105.31 ± 13.68	< 0.001	13.490	N = M > A
K-CWST-color reading	88.30 ± 17.91	70.41 ± 23.45	40.73 ± 24.30	< 0.001	63.859	N > M > A
K-BNT	48.15 ± 8.09	40.48 ± 9.97	28.70 ± 11.22	< 0.001	63.329	N > M > A
Digit span-forward	6.07 ± 1.51	5.63 ± 1.40	4.91 ± 1.45	0.002	6.440	N > M > A
Digit span-backward	4.10 ± 1.15	3.56 ± 0.97	2.92 ± 0.95	< 0.001	25.228	N > M > A

Values are presented as mean ± standard deviation.

MCI = mild cognitive impairment, AD = Alzheimer's disease, K-MMSE = Korean version of the Mini-Mental State Examination, SVLT = Seoul Verbal Learning Test, RCFT = Rey Complex Figure Test, SWF = Semantic Word Fluency Test, PWF = Phonemic Word Fluency Test, K-CWST = Korean Color Word Stroop Test, K-BNT = Korean version of the Boston Naming Test.

The results of the MMSE-2 (BV, SV, and EV) in the normal, MCI, and AD groups

The MMSE-2 (BV, SV, and EV) scores for the three participant groups (healthy older adults, patients with MCI, and patients with AD) are presented in Table 3. ANCOVA, controlling for age and education, revealed significant differences in MMSE-2 (BV, SV, and EV) scores across the three groups (P < 0.05).

Table 3. The results of the MMSE-2:BV, SV, EV and K-MMSE in the normal, MCI, AD groups.

Variables			Normal	MCI	AD	P	F	Post hoc
MMSE-2
	Registration		2.91 ± 0.44	2.91 ± 0.35	2.60 ± 0.59	< 0.001	12.381	N = M > A
	Orientation to time		4.80 ± 0.43	4.54 ± 0.77	2.40 ± 1.52	< 0.001	175.531	N = M > A
	Orientation to place		4.86 ± 0.38	4.73 ± 0.53	3.62 ± 1.06	< 0.001	91.732	N = M > A
	Recall		1.87 ± 0.85	1.20 ± 0.84	0.26 ± 0.51	< 0.001	77.302	N > M > A
	MMSE-2:BV		14.43 ± 1.32	13.37 ± 1.58	8.91 ± 2.54	< 0.001	230.625	N > M > A
		Attention and calculation	4.09 ± 1.17	3.77 ± 1.24	2.61 ± 1.56	< 0.001	28.292	N = M > A
		Language	7.69 ± 0.92	7.67 ± 0.57	7.19 ± 1.00	< 0.001	7.936	N = M > A
		Drawing	0.95 ± 0.23	0.90 ± 0.30	0.69 ± 0.47	< 0.001	13.542	N = M > A
	MMSE-2:SV		27.26 ± 2.66	25.71 ± 2.35	19.39 ± 3.94	< 0.001	205.003	N > M > A
		Story memory	10.46 ± 3.50	7.15 ± 3.21	3.34 ± 1.80	< 0.001	109.956	N > M > A
		Processing speed	12.24 ± 4.31	10.47 ± 4.03	6.35 ± 3.40	< 0.001	38.778	N > M > A
	MMSE-2:EV		49.84 ± 9.59	43.48 ± 7.81	29.06 ± 7.17	< 0.001	168.365	N > M > A
K-MMSE
	Orientation to time		4.80 ± 0.43	4.54 ± 0.77	2.40 ± 1.52	< 0.001	175.531	N = M > A
	Orientation to place		4.86 ± 0.38	4.73 ± 0.53	3.62 ± 1.06	< 0.001	91.732	N = M > A
	Registration		3.00 ± 0.00	2.98 ± 0.13	2.94 ± 0.24	0.045	3.133	N = M > A
	Attention and calculation		4.09 ± 1.17	3.77 ± 1.24	2.61 ± 1.56	< 0.001	28.292	N = M > A
	Recall		1.74 ± 1.07	1.12 ± 1.07	0.15 ± 0.36	< 0.001	48.008	N > M > A
	Language		7.88 ± 0.36	7.71 ± 0.61	6.90 ± 1.18	< 0.001	40.062	N = M > A
	Drawing		0.95 ± 0.23	0.90 ± 0.30	0.69 ± 0.47	< 0.001	13.542	N = M > A
Total			27.31 ± 2.30	25.75 ± 2.46	19.31 ± 3.80	< 0.001	229.273	N > M > A

Values are presented as mean ± standard deviation.

MCI = mild cognitive impairment, AD = Alzheimer's disease, MMSE-2 = Mini-Mental State Examination-2, BV = Brief Version, SV = Standard Version, EV = Expanded Version, K-MMSE = Korean version of the Mini-Mental State Examination.

Tukey’s post hoc analysis showed that the total scores of the MMSE-2:BV, SV, and EV were significantly higher for healthy older adults than for patients with MCI and AD. Additionally, patients with MCI had significantly higher scores than patients with AD. Among the individual items of the MMSE-2, recall, story memory, and processing speed scores were significantly higher in healthy older adults compared to both patients with MCI and AD, and patients with MCI scored significantly higher than those with AD on these items.

However, there were no significant differences between healthy older adults and patients with MCI on the registration, orientation to time, orientation to place, attention and calculation, language, or drawing items. Nevertheless, the scores for these six items were significantly higher in both healthy older adults and patients with MCI compared to patients with AD.

The results of the K-MMSE in the normal, MCI, and AD groups

The K-MMSE scores for the three groups (healthy older adults, patients with MCI, and patients with AD) are presented in Table 3. ANCOVA, controlling for age and education, revealed significant differences in K-MMSE scores among the three groups (P < 0.05).

Tukey’s post hoc analysis indicated that the recall and total scores of the K-MMSE were significantly higher in healthy older adults compared to patients with MCI and AD. Additionally, patients with MCI had significantly higher scores than patients with AD.

However, there were no significant differences between healthy older adults and patients with MCI on the subtests of registration, orientation to time, orientation to place, attention and calculation, language, or drawing. Nevertheless, scores on these six subtests were significantly higher for both healthy older adults and patients with MCI compared to patients with AD.

Comparison of the discriminant analysis of the MMSE-2 and the K-MMSE

MMSE-2

1) The discriminant analysis of the MMSE-2 to classify into the normal, MCI, and AD groups

Discriminant analysis using the simultaneous input method was conducted with the nine subtests of the MMSE-2 as independent variables and the three groups (healthy older adults, MCI, and AD) as dependent variables. The results indicated that the first discriminant function differentiated healthy older adults from the patient groups (MCI and AD), while the second function distinguished patients with MCI from those with AD.

The first discriminant function was statistically significant and accounted for 92.1% of the total variance in the model (Wilk’s lambda = 0.342; χ² = 436.80; P < 0.001). The second discriminant function was also statistically significant, accounting for 7.9% of the total variance in the model (Wilk’s lambda = 0.881; χ² = 51.51; P < 0.001). Wilk’s lambda is a statistical measure in discriminant analysis that evaluates the significance of differences between groups. A smaller Wilk’s lambda value (i.e., closer to 0) indicates a strong relationship between the independent variables and the dependent variable (group), suggesting that the discriminant analysis is highly effective.

According to the standardized canonical discriminant function coefficients, orientation to time was the most discriminative subtest in the first function, which distinguished healthy older adults from the patient groups (MCI and AD). In the second function, which differentiated patients with MCI from those with AD, story memory was identified as the most discriminative subtest. Canonical loading is a coefficient in discriminant analysis that indicates the strength of the relationship between each independent variable and the discriminant function. In other words, the closer the value is to ± 1, the stronger the correlation between the variable and the discriminant function.

The structure matrix of canonical loadings for the predictor variables and the two discriminant functions revealed that the first function was strongly correlated with orientation to time (canonical loading = 0.78), orientation to place (canonical loading = 0.58), recall (canonical loading = 0.55), processing speed (canonical loading = 0.42), and attention and calculation (canonical loading = 0.33). The second function was strongly correlated with story memory (canonical loading = −0.66). These results indicate that orientation to time was the most effective subtest for distinguishing healthy older adults from the patient groups, while story memory was the most effective subtest for differentiating between patients with MCI and AD.

The classification results based on the two discriminant functions are summarized in Table 4. Using the MMSE-2:BV, 55.3% of patients with MCI, 78.4% of patients with AD, and 72.5% of healthy older adults were correctly classified, resulting in an overall classification accuracy of 64.5%. For the MMSE-2:SV, 58.0% of patients with MCI, 79.4% of patients with AD, and 72.5% of healthy older adults were correctly classified, yielding an overall classification accuracy of 66.2%. Using the MMSE-2:EV, 68.6% of patients with MCI, 78.4% of patients with AD, and 72.5% of healthy older adults were correctly classified, achieving an overall classification accuracy of 71.7%.

Table 4. Classification rates and accuracy (%) by discriminant analysis in the normal, MCI, and AD groups (MMSE-2).

MMSE-2		Predictive groups
		Groups	Normal	MCI	AD	Total
MMSE-2:BV
	Frequency	Normal	66	24	1	91
		MCI	82	125	19	226
		AD	0	21	76	97
	%	Normal	72.5	26.4	1.1	100.0
		MCI	36.3	55.3	8.4	100.0
		AD	0.0	21.6	78.4	100.0
Classification accuracy, %						64.5
MMSE-2:SV
	Frequency	Normal	66	23	2	91
		MCI	79	131	16	226
		AD	0	20	77	97
	%	Normal	72.5	25.3	2.2	100.0
		MCI	35.0	58.0	7.1	100.0
		AD	0.0	20.6	79.4	100.0
Classification accuracy, %						66.2
MMSE-2:EV
	Frequency	Normal	66	23	2	91
		MCI	54	155	17	226
		AD	0	21	76	97
	%	Normal	72.5	25.3	2.2	100.0
		MCI	23.9	68.6	7.5	100.0
		AD	0.0	21.6	78.4	100.0
Classification accuracy, %						71.7

MMSE-2 = Mini-Mental State Examination-2, MCI = mild cognitive impairment, AD = Alzheimer's disease, BV = Brief Version, SV = Standard Version, EV = Expanded Version.

2) The discriminant analysis of the MMSE-2 to classify the healthy older adults vs. the patients with MCI

The discriminant analysis by the simultaneous input method was performed by using the nine subtests of the MMSE-2 as independent variables and the two groups (healthy older adults and MCI) as dependent variables. There was a significant difference between the two groups over nine independent variables (Wilk’s lambda = 0.790; χ² = 73.35; P < 0.001).

According to the results of the standardized canonical discriminant function coefficients in the discriminant function, story memory was the most discriminating subtest. Moreover, the structure matrix canonical loadings of the predictor variables and the discriminant function indicated that the function was strongly correlated with story memory (canonical loading = 0.88), recall (canonical loading = 0.70), processing speed (canonical loading = 0.38), and orientation to time (canonical loading = 0.34), but the function was not significantly correlated with attention and calculation, orientation to place, drawing, language, and registration. These results indicated that the best discriminating subtest between the healthy older adults and the patients with MCI was story memory.

According to the classification results, with the MMSE-2:BV, 73.6% of the healthy older adults and 64.2% of the patients with MCI were correctly classified, and thus the overall classification accuracy was 66.9%. With the MMSE-2:SV, 71.4% of the healthy older adults and 65.0% of the patients with MCI were correctly classified, and thus the overall classification accuracy was 66.9%. With the MMSE-2:EV, 73.6% of the healthy older adults and 74.8% of the patients with MCI were correctly classified, and thus the overall classification accuracy was 74.4%.

3) The discriminant analysis of the MMSE-2 to classify the patients with MCI vs. the patients with AD

Discriminant analysis using the simultaneous input method was conducted with the nine subtests of the MMSE-2 as independent variables and two groups (MCI and AD) as dependent variables. Significant differences were observed between the two groups across the nine independent variables (Wilk’s lambda = 0.129; χ² = 268.22; P < 0.001).

The standardized canonical discriminant function coefficients indicated that orientation to time was the most discriminative subtest. Additionally, the structure matrix of canonical loadings revealed that the discriminant function was strongly correlated with orientation to time (canonical loading = 0.81), orientation to place (canonical loading = 0.61), story memory (canonical loading = 0.53), recall (canonical loading = 0.49), processing speed (canonical loading = 0.43), and attention and calculation (canonical loading = 0.35). However, the function was not significantly correlated with registration, language, or drawing. These findings suggest that orientation to time was the most effective subtest for distinguishing patients with MCI from those with AD.

The classification results demonstrated high accuracy in distinguishing between patients with MCI and AD across the different versions of the MMSE-2. Specifically, the MMSE-2:BV correctly classified 89.4% of patients with MCI and 81.4% of patients with AD, resulting in an overall classification accuracy of 87.0%. The MMSE-2:SV showed even greater accuracy, correctly classifying 92.5% of patients with MCI and 82.5% of patients with AD, with an overall classification accuracy of 89.5%. Similarly, the MMSE-2:EV correctly classified 92.0% of patients with MCI and 82.5% of patients with AD, achieving an overall classification accuracy of 89.2%.

4) The discriminant analysis of the MMSE-2 to classify the healthy older adults vs. the patients with AD

Discriminant analysis using the simultaneous input method was conducted with the nine subtests of the MMSE-2 as independent variables and two groups (healthy older adults and AD) as dependent variables. The analysis identified one significant discriminant function, revealing a significant difference between the two groups across the nine independent variables (Wilk’s lambda = 0.247; χ² = 253.62; P < 0.001).

The standardized canonical discriminant function coefficients identified story memory as the most discriminative subtest in distinguishing between healthy older adults and patients with AD. Additionally, the structure matrix of canonical loadings revealed that the discriminant function was strongly correlated with story memory (canonical loading = 0.75), recall (canonical loading = 0.67), orientation to time (canonical loading = 0.61), orientation to place (canonical loading = 0.44), processing speed (canonical loading = 0.44), and attention and calculation (canonical loading = 0.31). However, the function was not significantly correlated with drawing, registration, or language. These findings indicate that story memory was the most effective subtest for differentiating between healthy older adults and patients with AD.

The classification results showed high accuracy in distinguishing between healthy older adults and patients with AD across the different versions of the MMSE-2. Specifically, the MMSE-2:BV correctly classified 91.2% of healthy older adults and 88.7% of patients with AD, resulting in an overall classification accuracy of 89.9%. The MMSE-2:SV correctly classified 91.2% of healthy older adults and 89.7% of patients with AD, achieving an overall classification accuracy of 90.4%. The MMSE-2:EV demonstrated the highest accuracy, correctly classifying 93.4% of healthy older adults and 97.9% of patients with AD, with an overall classification accuracy of 95.7%.

K-MMSE

1) The discriminant analysis of the K-MMSE to classify into the normal, MCI, and AD groups

Discriminant analysis using the simultaneous input method was conducted with the seven subtests of the K-MMSE as independent variables and the three groups (healthy older adults, MCI, and AD) as dependent variables. The analysis revealed that the first discriminant function distinguished healthy older adults from the patient groups (MCI and AD), while the second function differentiated patients with MCI from those with AD.

The first discriminant function was statistically significant and accounted for 97.4% of the total variance in the model (Wilk’s lambda = 0.395; χ² = 397.15; P < 0.001). The second discriminant function was also statistically significant and explained 2.6% of the total variance (Wilk’s lambda = 0.963; χ² = 15.28; P < 0.05).

The standardized canonical discriminant function coefficients indicated that orientation to time was the most discriminative subtest in the first function, which distinguished healthy older adults from the patient groups (MCI and AD), while recall was the most discriminative subtest in the second function, which differentiated patients with MCI from those with AD.

The structure matrix of canonical loadings showed that the first function was strongly correlated with orientation to time (canonical loading = 0.83), orientation to place (canonical loading = 0.61), language (canonical loading = 0.42), and attention and calculation (canonical loading = 0.35). The second function was strongly correlated with recall (canonical loading = 0.84). These findings suggest that orientation to time was the most effective subtest for distinguishing healthy older adults from the patient groups, while recall was the most effective subtest for differentiating between patients with MCI and AD.

The classification results based on the two functions are presented in Table 5. The analysis showed that 83.6% of patients with MCI, 68.0% of patients with AD, and 28.6% of healthy older adults were correctly classified, resulting in an overall classification accuracy of 67.9%.

Table 5. Classification rates and accuracy (%) by discriminant analysis in the normal, MCI, and AD groups (K-MMSE).

K-MMSE	Predictive groups
	Groups	Normal	MCI	AD	Total
Frequency	Normal	26	64	1	91
	MCI	25	189	12	226
	AD	0	31	66	97
Proportion, %	Normal	28.6	70.3	1.1	100.0
	MCI	11.1	83.6	5.3	100.0
	AD	0.0	32.0	68.0	100.0
Classification accuracy, %					67.9

K-MMSE = Korean version of the Mini-Mental State Examination, MCI = mild cognitive impairment, AD = Alzheimer's disease.

2) The discriminant analysis of the K-MMSE to classify the healthy older adults vs. the patients with MCI

Discriminant analysis using the simultaneous input method was conducted with the seven subtests of the K-MMSE as independent variables and two groups (healthy older adults and patients with MCI) as dependent variables. Significant differences were observed between the two groups across the seven independent variables (Wilk’s lambda = 0.905; χ² = 30.96; P < 0.001).

The standardized canonical discriminant function coefficients identified recall as the most discriminative subtest. Additionally, the structure matrix of canonical loadings revealed that the discriminant function was strongly correlated with recall (canonical loading = 0.81), orientation to time (canonical loading = 0.55), language (canonical loading = 0.43), attention and calculation (canonical loading = 0.37), and orientation to place (canonical loading = 0.36). However, the function was not significantly correlated with registration or drawing. These findings indicate that recall was the most effective subtest for distinguishing healthy older adults from patients with MCI.

The classification results showed that 69.2% of healthy older adults and 61.1% of patients with MCI were correctly classified, yielding an overall classification accuracy of 63.4%.

3) The discriminant analysis of the K-MMSE to classify the patients with MCI vs. the patients with AD

Discriminant analysis using the simultaneous input method was conducted with the seven subtests of the K-MMSE as independent variables and two groups (MCI and AD) as dependent variables. Significant differences were observed between the two groups across the seven independent variables (Wilk’s lambda = 0.443; χ² = 258.82; P < 0.001).

The standardized canonical discriminant function coefficients identified orientation to time as the most discriminative subtest. Additionally, the structure matrix of canonical loadings indicated that the discriminant function was strongly correlated with orientation to time (canonical loading = 0.83), orientation to place (canonical loading = 0.63), recall (canonical loading = 0.43), language (canonical loading = 0.40), and attention and calculation (canonical loading = 0.36). However, the function was not significantly correlated with drawing or registration. These findings suggest that orientation to time was the most effective subtest for distinguishing between patients with MCI and those with AD.

The classification results showed that 92.0% of patients with MCI and 81.4% of patients with AD were correctly classified, resulting in an overall classification accuracy of 88.9%.

4) The discriminant analysis of the K-MMSE to classify the healthy older adults vs. the patients with AD

Discriminant analysis using the simultaneous input method was conducted with the seven subtests of the K-MMSE as independent variables and two groups (healthy older adults and AD) as dependent variables. Significant differences were observed between the two groups across the seven independent variables (Wilk’s lambda = 0.294; χ² = 223.55; P < 0.001).

The standardized canonical discriminant function coefficients identified recall as the most discriminative subtest. Additionally, the structure matrix of canonical loadings indicated that the discriminant function was strongly correlated with orientation to time (canonical loading = 0.69), recall (canonical loading = 0.65), orientation to place (canonical loading = 0.50), language (canonical loading = 0.36), and attention and calculation (canonical loading = 0.35). However, the function was not significantly correlated with drawing or registration. These findings suggest that orientation to time was the most effective subtest for distinguishing between healthy older adults and patients with AD.

The classification results showed that 95.6% of healthy older adults and 94.8% of patients with AD were correctly classified, yielding an overall classification accuracy of 95.2%.

Diagnostic utility of the MMSE-2 and the K-MMSE

MMSE-2

To evaluate the diagnostic utility of the three versions of the MMSE-2, ROC curve analysis and AUC calculations were performed. The results for each version of the MMSE-2 are detailed below.

MMSE-2:BV

The diagnostic utility of the MMSE-2:BV was analyzed using ROC curve analysis and AUC calculations for three different comparisons: healthy older adults versus patients with MCI, patients with MCI versus patients with AD, and healthy older adults versus patients with AD.

First, in distinguishing healthy older adults from patients with MCI, the MMSE-2:BV demonstrated an AUC of 0.71, with a 95% confidence interval (CI) ranging from 0.64 to 0.77 (P < 0.001). A cut-off score of ≤ 14 out of 16 was identified as optimal for predicting MCI, yielding a sensitivity of 60% and a specificity of 75%.

Second, when distinguishing patients with MCI from patients with AD, the MMSE-2:BV showed a higher diagnostic performance, with an AUC of 0.93 (95% CI, 0.90–0.96; P < 0.001). Using a cut-off score of ≤ 10 out of 16, the sensitivity was 88% and the specificity was 87%, indicating strong discrimination between these two groups.

Finally, in distinguishing healthy older adults from patients with AD, the MMSE-2:BV achieved an AUC of 0.97 (95% CI, 0.96–0.99; P < 0.001), demonstrating excellent diagnostic accuracy. A cut-off score of ≤ 10 out of 16 resulted in a sensitivity of 98% and a specificity of 70%.

MMSE-2:SV

The diagnostic performance of the MMSE-2:SV was assessed using ROC curve analysis and AUC calculations for three group comparisons: healthy older adults versus patients with MCI, patients with MCI versus patients with AD, and healthy older adults versus patients with AD.

First, in distinguishing healthy older adults from patients with MCI, the MMSE-2:SV demonstrated an AUC of 0.72, with a 95% CI ranging from 0.66 to 0.79 (P < 0.001). A cut-off score of ≤ 26 out of 30 was identified as optimal for predicting MCI. At this threshold, the sensitivity was 74%, meaning 74% of patients with MCI were correctly identified, and the specificity was 59%, meaning 59% of healthy older adults were correctly classified as not having MCI.

Second, for distinguishing patients with MCI from those with AD, the MMSE-2:SV demonstrated higher diagnostic accuracy, with an AUC of 0.93 (95% CI, 0.89–0.96; P < 0.001). A cut-off score of ≤ 23 out of 30 provided a sensitivity of 84%, meaning 84% of patients with AD were correctly identified, and a specificity of 87%, meaning 87% of patients with MCI were correctly classified as not having AD.

Finally, in distinguishing healthy older adults from patients with AD, the MMSE-2:SV showed excellent diagnostic performance, with an AUC of 0.95 (95% CI, 0.92–0.98; P < 0.001). Using a cut-off score of ≤ 23 out of 30, the sensitivity was 92%, meaning 92% of patients with AD were correctly identified, and the specificity was 87%, meaning 87% of healthy older adults were correctly classified as not having AD.

MMSE-2:EV

The diagnostic performance of the MMSE-2:EV was assessed using receiver ROC curve analysis and AUC calculations for three comparisons: healthy older adults versus patients with MCI, patients with MCI versus patients with AD, and healthy older adults versus patients with AD.

First, in distinguishing healthy older adults from patients with MCI, the MMSE-2:EV showed an AUC of 0.73, with a 95% CI ranging from 0.66 to 0.80 (P < 0.001). A cut-off score of ≤ 46 out of 90 was determined to be optimal for predicting MCI. At this threshold, the sensitivity was 71%, meaning 71% of patients with MCI were correctly identified, and the specificity was 69%, meaning 69% of healthy older adults were correctly classified as not having MCI.

Second, for distinguishing patients with MCI from those with AD, the MMSE-2:EV demonstrated stronger diagnostic performance, with an AUC of 0.92 (95% CI, 0.89–0.95; P < 0.001). Using a cut-off score of ≤ 36 out of 90, the sensitivity was 82%, indicating that 82% of patients with AD were correctly identified, and the specificity was 85%, meaning 85% of patients with MCI were correctly classified as not having AD.

Finally, in distinguishing healthy older adults from patients with AD, the MMSE-2:EV showed an AUC of 0.94 (95% CI, 0.91–0.98; P < 0.001), demonstrating excellent diagnostic accuracy. A cut-off score of ≤ 34 out of 90 provided a sensitivity of 92%, meaning 92% of patients with AD were correctly identified, and a specificity of 71%, meaning 71% of healthy older adults were correctly classified as not having AD.

K-MMSE

The diagnostic accuracy of the K-MMSE was evaluated using ROC curve analysis and AUC calculations to differentiate among three groups: healthy older adults, patients with MCI and AD.

First, for distinguishing healthy older adults from patients with MCI, the K-MMSE showed an AUC of 0.70, with a 95% CI ranging from 0.64 to 0.77 (P < 0.001). A cut-off score of ≤ 26 out of 30 was found to be optimal for predicting MCI. At this threshold, the sensitivity was 70%, meaning 70% of patients with MCI were correctly identified, and the specificity was 59%, meaning 59% of healthy older adults were correctly classified as not having MCI.

Second, for distinguishing patients with MCI from those with AD, the K-MMSE demonstrated excellent diagnostic performance, with an AUC of 0.94 (95% CI, 0.91–0.96; P < 0.001). Using a cut-off score of ≤ 23 out of 30, the sensitivity was 84%, indicating that 84% of patients with AD were correctly identified, while the specificity was 88%, meaning 88% of patients with MCI were correctly classified as not having AD.

Finally, for distinguishing healthy older adults from patients with AD, the K-MMSE showed an even higher AUC of 0.97 (95% CI, 0.95–0.99; P < 0.001), indicating outstanding diagnostic accuracy. A cut-off score of ≤ 22 out of 30 was optimal for predicting AD, with a sensitivity of 95%, meaning 95% of patients with AD were correctly identified, and a specificity of 78%, meaning 78% of healthy older adults were correctly classified as not having AD.

Comparison between the MMSE-2 and the K-MMSE

Comparison between healthy older adults and patients with MCI

1) The MMSE-2 vs. the K-MMSE

The MMSE-2:BV and K-MMSE were compared using ROC curve analysis to assess which test was more effective in distinguishing between healthy older adults and patients with MCI. The AUC of the MMSE-2:BV was 0.708, while the AUC of the K-MMSE was 0.703; however, no significant difference was observed between the two tests.

Similarly, the MMSE-2:SV and K-MMSE were compared using ROC curve analysis to evaluate their sensitivity in differentiating healthy older adults from patients with MCI. The AUC of the MMSE-2:SV was 0.720, compared to 0.703 for the K-MMSE, but the difference was not statistically significant.

Finally, the MMSE-2:EV and K-MMSE were also compared using ROC curve analysis for their ability to discriminate between healthy older adults and patients with MCI. The AUC of the MMSE-2:EV was 0.728, compared to 0.703 for the K-MMSE; again, the difference was not statistically significant (Table 6).

Table 6. The AUC of the MMSE-2:BV, MMSE-2:SV, MMSE-2:EV, and K-MMSE in the normal, MCI, and AD.

Groups	Tests	Cut-off	AUC (95% CI)	Sensitivity, %	Specificity, %	P
Normal vs. MCI	MMSE-2:BV	14/15	0.708 (0.644–0.773)	60	75	0.866 (MMSE-2:BV vs. K-MMSE)
	MMSE-2:SV	26/27	0.720 (0.655–0.785)	74	59	0.458 (MMSE-2:SV vs. K-MMSE)
	MMSE-2:EV	46/47	0.728 (0.662–0.795)	71	69	0.406 (MMSE-2:EV vs. K-MMSE)
	K-MMSE	26/27	0.703 (0.638–0.768)	70	59
MCI vs. AD	MMSE-2:BV	11/12	0.930 (0.900–0.961)	88	87	0.705 (MMSE-2:BV vs. K-MMSE)
	MMSE-2:SV	23/24	0.925 (0.894–0.956)	84	87	0.273 (MMSE-2:SV vs. K-MMSE)
	MMSE-2:EV	36/37	0.921 (0.893–0.949)	82	85	0.181 (MMSE-2:EV vs. K-MMSE)
	K-MMSE	23/24	0.936 (0.910–0.961)	84	88
Normal vs. AD	MMSE-2:BV	10/11	0.973 (0.955–0.992)	98	70	0.861 (MMSE-2:BV vs. K-MMSE)
	MMSE-2:SV	23/24	0.952 (0.921–0.983)	92	87	0.084 (MMSE-2:SV vs. K-MMSE)
	MMSE-2:EV	34/35	0.943 (0.907–0.980)	92	71	0.068 (MMSE-2:EV vs. K-MMSE)
	K-MMSE	23/24	0.971 (0.950–0.993)	93	88

AUC = area under the curve, MMSE-2 = Mini-Mental State Examination-2, BV = Brief Version, SV = Standard Version, EV = Expanded Version, K-MMSE = Korean version of the Mini-Mental State Examination, MCI = mild cognitive impairment, AD = Alzheimer’s disease, CI = confidence interval.

Comparison between patients with MCI and patients with AD

The MMSE-2:BV and K-MMSE were compared using ROC curve analysis to evaluate their sensitivity in distinguishing between patients with MCI and patients with AD. The AUC of the MMSE-2:BV was 0.903, while the AUC of the K-MMSE was 0.936; however, the difference between the two tests was not statistically significant.

Similarly, the MMSE-2:SV and K-MMSE were compared using ROC curve analysis to assess their effectiveness in discriminating between patients with MCI and those with AD. The AUC of the MMSE-2:SV was 0.925, compared to 0.936 for the K-MMSE, but this difference was also not statistically significant.

Finally, the MMSE-2:EV and K-MMSE were compared using ROC curve analysis to determine their sensitivity in distinguishing between patients with MCI and patients with AD. The AUC of the MMSE-2:EV was 0.921, while the AUC of the K-MMSE was 0.936; again, no statistically significant difference was found between the two tests (Table 6).

Comparison between healthy older adults and patients with AD

The MMSE-2:BV and K-MMSE were compared using ROC curve analysis to evaluate their sensitivity in distinguishing between healthy older adults and patients with AD. The AUC of the MMSE-2:BV was 0.973, while the AUC of the K-MMSE was 0.971; however, no significant difference was observed between the two tests.

Similarly, the MMSE-2:SV and K-MMSE were compared using ROC curve analysis to assess their effectiveness in discriminating between healthy older adults and patients with AD. The AUC of the MMSE-2:SV was 0.952, compared to 0.971 for the K-MMSE; however, this difference was not statistically significant.

Lastly, the MMSE-2:EV and K-MMSE were compared using ROC curve analysis to determine their sensitivity in distinguishing between healthy older adults and patients with AD. The AUC of the MMSE-2:EV was 0.943, while the AUC of the K-MMSE was 0.971; again, no statistically significant difference was found between the two tests (Table 6).

DISCUSSION

The cognitive screening tools, MMSE-2 and K-MMSE, were compared to evaluate which test was more sensitive in distinguishing among three groups: healthy older adults, patients with MCI, and patients with AD. The findings highlighted several key points.

First, the results of the MMSE-2 (BV, SV, and EV) and the K-MMSE showed significant differences across the three groups (healthy older adults, patients with MCI, and patients with AD). In all three versions of the MMSE-2, scores on subtests such as recall, story memory, and processing speed were significantly higher in healthy older adults compared to patients with MCI and AD. Additionally, patients with MCI scored significantly higher than patients with AD. These findings indicate that deficits in verbal memory were particularly pronounced.

Additionally, in the K-MMSE, recall subtest scores for healthy older adults were significantly higher than those of patients with MCI and AD. Furthermore, patients with MCI scored significantly higher on the recall subtest than patients with AD.

When comparing the effect sizes of the recall test in the MMSE-2 (η² = 0.27) and the K-MMSE (η² = 0.19), the recall test in the MMSE-2 demonstrated slightly greater sensitivity in discriminating among the groups. Furthermore, among the verbal memory tests, the story memory test in the MMSE-2 (η² = 0.35) showed greater sensitivity compared to the recall tests in both the MMSE-2 (η² = 0.27) and the K-MMSE (η² = 0.19). These findings suggest that the verbal memory tests, particularly recall and story memory, included in the MMSE-2 are more effective in differentiating among the three groups (healthy older adults, MCI, and AD).

Second, the results of the discriminant analysis of the MMSE-2 were as follows. In the MMSE-2, the subtests that significantly discriminated between healthy older adults and the patient groups (MCI and AD) were orientation to time, orientation to place, recall, processing speed, and attention and calculation. Additionally, the subtest that significantly differentiated patients with MCI from those with AD was story memory.

The subtests orientation to time, orientation to place, recall, and story memory in the MMSE-2 primarily assess episodic memory, while processing speed and attention and calculation measure psychomotor ability and working memory associated with the frontal lobe. Consistent with previous studies, these findings suggest that assessing episodic memory is a critical factor for group discrimination, as episodic memory is one of the first cognitive functions to decline in patients with MCI or AD.24,25,26,27

Furthermore, in the MMSE-2, the subtests that significantly discriminated between healthy older adults and patients with MCI were story memory, recall, processing speed, and orientation to time. These findings highlight that tests related to episodic memory are the most sensitive for group discrimination, consistent with results from previous studies.24,25,26,27

The subtests that significantly discriminated between patients with MCI and those with AD in the MMSE-2 included orientation to time, orientation to place, story memory, recall, processing speed, and attention and calculation. These results suggest that as dementia progresses, not only episodic memory but also frontal lobe functions show significant decline.

Third, the results of the discriminant analysis of the K-MMSE were as follows. In the K-MMSE, the subtests that significantly discriminated among the three groups (healthy older adults, patients with MCI, and patients with AD) were orientation to time, orientation to place, language, and attention and calculation. Additionally, the subtest that significantly discriminated between patients with MCI and those with AD was recall. Similar to the MMSE-2, these findings indicate that assessing episodic memory is the most sensitive method for group discrimination.

Furthermore, in the K-MMSE, the subtests that significantly discriminated between healthy older adults and patients with MCI included recall, orientation to time, language, and attention and calculation. These subtests are highly sensitive measures of episodic memory and working memory. The subtests that significantly differentiated patients with MCI from those with AD were orientation to time, orientation to place, recall, language, and attention and calculation. These findings suggest that as dementia progresses, not only episodic memory but also frontal lobe functions decline. In conclusion, subtests related to episodic memory are the most sensitive for group discrimination in both the MMSE-2 and the K-MMSE.

A comprehensive comparison of the discriminant analysis results for the MMSE-2 and K-MMSE revealed that the overall classification accuracy of the MMSE-2 across the three groups was 71.7%, compared to 67.9% for the K-MMSE, indicating that the MMSE-2 is more accurate than the K-MMSE. Specifically, the classification accuracy for distinguishing healthy older adults from patients with MCI was 74.4% for the MMSE-2 and 63.4% for the K-MMSE, suggesting that the MMSE-2 is more sensitive and accurate than the K-MMSE for this comparison. For distinguishing patients with MCI from those with AD, the classification accuracy was 89.2% for the MMSE-2 and 88.9% for the K-MMSE, indicating similar performance between the two tests. Finally, for distinguishing healthy older adults from patients with AD, the classification accuracy was 95.7% for the MMSE-2 and 95.2% for the K-MMSE, again showing comparable results for both tests.

In summary, the MMSE-2 demonstrated greater sensitivity compared to the K-MMSE in discriminating among the three groups. Notably, the MMSE-2 was approximately 10% more accurate than the K-MMSE in distinguishing between healthy older adults and patients with MCI. However, there were no significant differences between the two tests in distinguishing between healthy older adults and patients with AD or between patients with MCI and those with AD.

Finally, when comparing the AUC values of the MMSE-2 and K-MMSE, no significant differences were observed across all classifications. However, the MMSE-2:SV and MMSE-2:EV appeared to be slightly more sensitive than the K-MMSE in distinguishing between healthy older adults and patients with MCI. Conversely, the MMSE-2:BV and K-MMSE appeared to be slightly more sensitive for distinguishing patients with AD from healthy older adults and patients with MCI.

These findings suggest that tests with higher difficulty levels are better for distinguishing healthy older adults from patients with MCI, while tests with lower difficulty levels are more effective for distinguishing patients with AD from the other groups.

This study has certain limitations. Notably, the MCI group was not subdivided in the present study. Future research would benefit from categorizing MCI into amnestic and non-amnestic subtypes, enabling a more refined and comprehensive analysis of the differences in MMSE and MMSE-2 scores. Despite these limitations, this study represents the first to compare the MMSE-2 and K-MMSE in patients with MCI and AD, providing valuable insights into their diagnostic utility.

Overall, the results indicate that for the developmental purpose of the MMSE-2, MMSE-2:SV and MMSE-2:EV are more sensitive and accurate than the K-MMSE or MMSE-2:BV in detecting early cognitive decline (e.g., distinguishing healthy older adults from patients with MCI). However, as dementia progresses, the K-MMSE or MMSE-2:BV may be more useful for group discrimination than MMSE-2:SV or MMSE-2:EV. In conclusion, the MMSE-2 demonstrates greater utility as a cognitive screening tool in clinical settings compared to the K-MMSE.