Normative Scores on the Norwegian Version of the Mini-Mental State Examination

Abstract

Background:

The Mini-Mental State Examination (MMSE), a simple test for measuring global cognitive function, is frequently used to evaluate cognition in older adults. To decide whether a score on the test indicates a significant deviation from the mean score, normative scores should be defined. Moreover, because the test may vary depending on its translation and cultural differences, normative scores should be established for national versions of the MMSE.

Objective:

We aimed to examine normative scores for the third Norwegian version of the MMSE.

Methods:

We used data from two sources: the Norwegian Registry of Persons Assessed for Cognitive Symptoms (NorCog) and the Trøndelag Health Study (HUNT). After persons with dementia, mild cognitive impairment, and disorders that may cause cognitive impairment were excluded, the sample contained 1,050 cognitively healthy persons, 860 from NorCog, and 190 from HUNT, whose data we subjected to regression analyses.

Results:

The normative MMSE score varied from 25 to 29, depending on years of education and age. More years of education and younger age were associated with higher MMSE scores, and years of education was the strongest predictor.

Conclusion:

Mean normative MMSE scores depend on test takers’ years of education and age, with level of education being the strongest predictor.

Keywords

Dementia mild cognitive impairment Mini-Mental State Examination normative score

INTRODUCTION

Using DSM-5 criteria, a large population study in Norway recently revealed that 14.6% of people at least 70 years old in the country have a major neurocognitive disorder (e.g., dementia), while one in three have a minor neurocognitive disorder, or mild cognitive impairment (MCI) [1, 2]. According to the most widely used definition of MCI, one criterion is a score of 1.5 units of standard deviation (SD) less than the normative score on a cognitive test, while another is the preservation of independence in activities of daily living (ADL) [3]. Scores 1.0–1.5 SD less than the normative score are also commonly used to define MCI [4]. By contrast, other definitions, including Winblad et al.’s criteria, do not require a cutoff on a cognitive test but do require preserved independence in ADL, along with evidence of cognitive dysfunction that significantly deviates from what is expected considering the person’s education and age [5].

MCI can stem from Alzheimer’s disease (AD), stroke, neurological disorders, depression, and/or other psychiatric disorders, alcohol or substance abuse, delirium, or a serious physical disorder [4]. According to a systematic review, approximately one of every four people with MCI will develop dementia in the next five years of their life [6]. Although some studies including patients from memory clinics have revealed higher conversion rates, lower rates have been reported in the general population [7, 8]. Given the relatively high conversion rate, people with MCI should be followed up to ensure correct etiological diagnoses [3 –5]. We advise anyone showing significant decline in cognition to be assessed, including individuals who score 1.0 or more SD less than the normative score on a cognitive test [4].

One of the most widely used tests to evaluate cognition is the Mini–Mental State Examination (MMSE). Since its publication in 1975 as a bedside test, the MMSE has gained popularity worldwide as a case-based tool for detecting global cognitive impairment, especially among older adults [8 –10]. The test consists of 20 questions and tasks and generates scores from 0 to 30, with higher scores indicating better performance. The first Norwegian version was published in 1988 and showed, comparable with the US version, that a score of less than 24 was the optimal cutoff to distinguish individuals with dementia from those without [9, 11]. In other studies, using various translations of the MMSE, the cutoffs to differentiate people with dementia from those without have varied between 20 and 27 [10 , 12–17].

However, regardless of the cutoffs used in different studies, the sensitivity and specificity of the MMSE have been reported to be fair; therefore, the MMSE should not be regarded as a diagnostic test for MCI or dementia but simply as a test for evaluating cognitive function [10 , 15–19]. For one, dementia is defined by a significant decline in cognition, behavioral changes, and impaired function in ADL, and a cognitive test cannot capture all such symptoms. For another, using a single cutoff point is problematic. After all, it is well known that more education and younger age are associated with higher scores on the MMSE; although studies have also shown that ethnicity and sex influence MMSE scores, evidence of such associations is poor [10 , 15–18]. Because education and age do influence test performance, using adjusted cutoff scores instead of a single score has been recommended [10, 19].

Altogether, though the MMSE is not a diagnostic test, we argue that it can be used to evaluate whether a test taker’s performance deviates from a norm adjusted for education and age and the extent of the deviation. To that purpose, we need to define a normative score for people in the same culture using the same translated version of the MMSE and adjusted for education and age. Similar research has been conducted in other countries, including a large population study conducted in the United States with more than 18,000 people [20]. That study revealed normative scores on the MMSE between 22 and 29 depending on level of education and age, and subsequent studies have reported similar variations [14 , 21–28].

In 2001, the copyright for the MMSE was handed over to Psychological Assessment Resources (PAR), an organization that requires a fee for using the test. To avoid the fee, new revised national versions of the MMSE that deviate from the original from 1975 have been developed in some countries, including Norway [12, 16]. In the third version of the Norwegian MMSE, used since 2017, two questions regarding country and county of residence are replaced with questions asking the test taker’s year of birth and age. Moreover, some small technical changes from the original version were made, and a comprehensive guide on how to use the test has been developed [29].

Various strategies can be applied to assess normative scores on cognitive tests. One involves including all individuals in a community survey without excluding anyone. A more widely used strategy, however, involves excluding people with disorders known to cause cognitive impairment, including stroke, depression, and various other mental and physical disorders. Following the first strategy, the estimated normative scores on a test conducted among the oldest individuals would likely be too low, for the prevalence of dementia and MCI is high among people more than 70 years old [1]. Following the second design, by contrast, the estimated normative scores could be too high if a large proportion of the surveyed adults are so-called successful agers [30]. Meanwhile, the choice of design is less important for estimating normative scores on cognitive tests among younger people, because the prevalence of MCI and dementia in younger populations is far lower [2].

Because normative scores for the MMSE do not exist in the Nordic countries, the aim of our study was to define such norms, adjusted for education and age, and to explore whether sex and/or marital status are associated with MMSE score.

METHODS

We used data from two sources. The first was the Norwegian Registry of Persons Assessed for Cognitive Symptoms (NorCog). Second, to enrich the study population with older adults, we included participants at least 70 years old from a small substudy of the fourth wave of the Trøndelag Health Study (i.e., HUNT4 Trondheim 70+), a community study conducted in Trondheim, Norway, including people 70 years old or older.

As for the first source, NorCog is a research and quality registry that collects and stores demographic, clinical, and biological data from patients who have undergone a comprehensive standardized assessment for cognitive impairment as part of specialized health care [31]. Patients or family carers provide their informed consent to participate, and 75% of patients assessed in at least one of 44 outpatient clinics have consented to be included in the registry [31]. Between January 2017 and January 2022, all 11,418 prospective participants tested with the third version of the Norwegian MMSE were eligible for inclusion. In terms of assessment strategy, we decided to use the second strategy described above—that is, to exclude all individuals with significant cognitive impairment—and a comprehensive exclusion procedure was performed. First, we excluded all patients diagnosed with dementia, MCI, any organic psychiatric disorder, alcohol or substance abuse, psychosis, depression, a neurologic disorder, or a learning disability. Second, we excluded individuals with significantly lower scores on the MMSE than the vast majority of the sample; they included people with self-reported stroke, self-reported depression, or a high burden of psychiatric symptoms as assessed in interviews with them and/or a family carer. Those exclusions resulted in the inclusion of 860 of 11,418 individuals whom we judged to be cognitively healthy (Fig. 1).

Fig. 1

Flow chart showing the procedure for excluding patients in the NorCog cohort from the sample. F and G diagnoses were made according to the 10th revision of the International Classification of Disorders and carers reported symptoms using the Neuropsychiatric Inventory (NPI). MMSE, Mini-Mental Examination Status; MADRS, Montgomery Aasberg Depression Rating Scale.

As for the second source, the HUNT4 Trondheim 70+ study was conducted in, Trondheim, Norway’s third-largest city, in 2018 and 2019. Of 504 home-dwelling people at least 70 years old who were invited to participate, 312 consented to be tested with the MMSE [32]. However, three had invalid data on the MMSE, 19 had dementia, and 100 had MCI, which left 190 cognitively healthy older adults for the sample. None of the 190 had any diagnosis or symptom burden, just as the individuals included from the NorCog cohort. Individuals who declined to participate tended to have lower scores on the Montreal Cognitive Assessment (MoCA), but no differences were found between the participants and the decliners regarding level of education, age, and sex [33].

Between the sources, whereas trained nurses and nursing students administered the MMSE in the HUNT Trondheim 70+ study, trained doctors, psychologists, or nurses administered it for the data in NorCog following standardized guidelines. In the NorCog cohort, all 860 individuals were assessed with an interview, a comprehensive battery of standardized cognitive tests, a physical examination, blood testing, and magnetic resonance imaging (MRI) or computed tomography (CT) imaging of the brain. In some cases, lumbar punction was performed to examine biomarkers of AD and/or positron emission tomography (fluorodeoxyglucose, or amyloid PET) were used to make a correct diagnosis (31). Last, depression was evaluated with the Montgomery Aasberg Depression Rating Scale and those scoring above 16 were excluded [34].

In the HUNT4 Trondheim 70+ cohort, participants were interviewed and tested with MoCA, the MMSE, and the 10-word test of the Consortium to establish a registry for Alzheimer’s disease (CERAD) battery [35] but not examined with MRI or CT. A family member of each participant was interviewed to capture possible changes in the individual’s cognition and behavior. The Depression subscale of the Hospital Anxiety and Depression Scale was used to identify depression, and individuals with scores greater than 7 were excluded [36]. To exclude individuals in both cohorts with serious psychiatric symptoms the Neuropsychiatric Inventory was used [37] (see Fig. 1).

Experienced doctors—in most cases, two—made diagnoses in both cohorts independently of each other [1, 31]. In the NorCog cohort, the ICD-10 criteria were applied for all disorders, while Winblad et al.’s criteria were used to diagnose MCI [5]. In the HUNT4 Trondheim 70+ cohort, DSM-5 criteria were used for major (e.g., dementia) and minor (e.g., MCI) neurocognitive disorders. The score on the MMSE was not used to diagnose MCI or dementia.

Statistical analyses

Associations between MMSE score and continuous age and years of education, categorized as 10 years (i.e., compulsory), 11–13 years (i.e., secondary), and ≥14 years (i.e., higher), were assessed with a linear regression model. Because age clearly showed a nonlinear association with MMSE scores, it was entered into the model as such. In addition, the model included all two- and three-way interactions between covariates. The Bayes information criterion (BIC), in which smaller values imply a better model, was applied to reduce the multiple model for excessive interaction terms. Thorough residual diagnostics was performed; the normality of residuals was considered by inspecting a histogram and Q–Q plot, a box plot, and Levene’s test were used to assess homoscedasticity. Due to some negative skewness of residuals, multiple transformations of MMSE scores were considered; however, none of them improved the model’s fit. The model with robust and bootstrap-based standard errors did not differ notably from the results of the originally estimated model. Due to a few missing values on covariates, only participants with complete observations were included in regression analyses. Sex and marital status were considered as potential covariates but did not improve the model’s fit according to the BIC.

To derive the normative MMSE scores, two approaches were employed. First, standardized (Z–) scores were derived from the residuals of the regression model described above. Due to deviations from normality assumption, these scores might be unreliable, particularly in the segment of a scale with largest deviations. This approach was therefore supplied by percentile ranks [38] derived from age intervals with midpoints at 49, 52, 55, 58, 61, 64, 67, 70, 73, 76, 79, 82, 85, and 88 years (midpoints 49 and 88 represent age groups ≤50 and ≥87 years, respectively). Each midpoint represents an interval of –/+1 year from midpoint for normative data. The percentiles were derived based on interval of –/+5 years from midpoint, a so-called overlapping cell procedure [39]. Percentiles corresponding to Z-scores of 0, –1, –1.5, and –2 were tabulated.

All analyses were performed in Stata version 17 and SPSS version 27.

Ethics and data protection

Informed consent was obtained from all participants, and the study was conducted in accordance with the Declaration of Helsinki. The study was approved by the Regional Ethics Committee for Health and MEDICAL RESEARCH in southeast Norway and by the National and Regional Data Protection Agency. A Data Protection Impact Assessment was also performed, revealed satisfactory results and was approved by the Regional Data Protection Agency.

RESULTS

Demographic characteristics and the MMSE scores in both cohorts appear in Table 1. As expected, the participants from HUNT4 Trondheim 70+ were older, a larger proportion of them had more years of education, and they generally had higher MMSE scores than participants from the NorCog cohort. No difference in MMSE score emerged between the sexes, with males scoring 27.6 (SD = 2.7) versus females: 27.6 (SD = 2.7, p = 1.0). Married or cohabiting participants scored higher on the MMSE (mean 27.7, SD = 2.7) than unmarried ones (mean 27.3, SD = 2.8; p = 0.014). Table 2 shows mean MMSE score with SD stratified by years of education and age, both coded as categorical variables. The results of a linear regression model assessing the associations between MMSE score (dependent variable) and continuous age and categorized years of education are shown in Table 3. In the multiple model, more years of education is associated with higher MMSE scores. Higher age is weakly associated with MMSE score up to age of approximately 65 years; for older persons increasing age is associated with lower MMSE score (Fig. 2). The results of this model were subsequently used to determine normative MMSE scores (Z-scores) as $Z - score = (MMSE - \hat{MMSE}) / 2.477$ , in which MMSE is the observed MMSE score, $\hat{MMSE}$ is the MMSE score estimated from the regression model in Table 3, and 2.477 is root-mean-square error. Figure 3 illustrates the normative MMSE scores for participants with varying age stratified by years of education, as well as scores representing 1.0, 1.5, and 2.0 SD less than the normative score. Table 4 presents the normative data for different age based on percentile ranks (the second approach) stratified by education. According to applied approaches, normative scores varied between 25 and 29, and differences between two approaches were minimal. Using the results shown in Table 3, Fig. 3, and Table 4 we constructed a calculator that estimates Z-score and percentile of a participant’s performance on the MMSE adjusting for their years of education and age (https://www.aldringoghelse.no/mmsenr3en). In case of different scores generated by the two approaches, we suggest that the most conservative score should be used.

Table 1

Characteristics of the participants

	All	NorCog	HUNT
	(N = 1,050)	(n = 860)	(n = 190)
Females, n (%)	530 (50.5)	434 (50.5)	96 (50.5)
Age (N = 1,046)
Mean (SD)	70.2 (10.9)	69.0 (11.5)	75.4 (4.5)
Min –Max	35–94	35–94	70–93
Age group, n (%) (N = 1,046)
< 65 y	298 (28.5)	298 (34.8)	0
65–69 y	112 (10.7)	112 (13.1)	0
70–74 y	239 (22.9)	135 (15.8)	104 (54.7)
75–79 y	198 (18.9)	145 (16.9)	53 (27.9)
≥80 y	199 (19.0)	166 (19.4)	33 (17.4)
Education, n (%) (N = 969)
≤10 y	326 (33.6)	287 (36.7)	39 (21.0)
11–13 y	225 (23.2)	202 (25.8)	23 (12.4)
≥14 y	418 (43.1)	294 (37.6)	124 (66.7)
Married or cohabitating, n (%) (N = 1,011)
Yes	687 (68.6)	555 (64.5)	132 (69.5)
MMSE score
M (SD)	27.6 (2.7)	27.4 (2.9)	28.5 (1.3)
Median (min.–max.)	28.0 (14–30)	28.0 (14–30)	29. (23–30)

NorCog, Norwegian registry of persons assessed for cognitive symptoms; HUNT, Trøndelag health study; MMSE, Mini-Mental State Examination.

Table 2

Mean score (SD) on the Mini–Mental State Examination among groups of people with different ages and years of education

Age group	Education, in years
	≤ 10	11–13	≥14
< 65 y	n = 36	n = 87	n = 141
	27.1 (3.3)	28.1 (2.3)	28.7 (1.9)
65–69 y	n = 29	n = 32	n = 39
	26.5 (3.0)	28.1 (2.2)	28.7 (1.5)
70–74 y	n = 69	n = 40	n = 119
	26.9 (2.9)	27.7 (1.6)	28.9 (1.4)
75–79 y	n = 82	n = 35	n = 76
	25.9 (3.7)	27.1 (3.1)	28.1 (2.3)
≥80 y	n = 108	n = 31	n = 41
	26.0 (2.8)	27.4 (1.8)	27.5 (2.9)
All	n = 324	n = 225	n = 416
	26.3 (3.2)	27.8 (2.3)	28.5 (2.0)

Table 3

Linear regression with Mini-Mental State Examination score as the dependent variable, education as the categorical variable, and age as the continuous variable (N = 965)

Variable	Bivariate model		Multiple model
	RC (SE ^* / 95% CI)	p	RC (SE ^* / 95% CI)	p
Intercept			24.33 (18.13, 30.52)	< 0.001
Age	0.22 (0.10)^*	0.024	0.16 (0.09)^*	0.090
Age×Age	–0.002 (0.0007)^*	0.004	–0.001 (0.0007)^*	0.040
Education
≤10 y	–2.16 (–2.53, –1.80)	< 0.001	–1.89 (–2.27, –1.51)	< 0.001
11–13 y	–0.72 (–1.12, –0.31)	0.001	–0.71 (–1.12, –0.31)	0.001
≥14 y (ref.)	0		0

RC, regression coefficient; CI, confidence interval; ^*Standard error (SE) is reported instead of CI because of the higher-order term in the model. Participants less than 50 years old were assigned the age of 49 years, whereas people more than 88 years old were assigned the age of 89 years.

Fig. 2

Association between MMSE score and age.

Fig. 3

Normative Mini–Mental State Examination (MMSE) scores and scores for 1.0, 1.5, and 2.0 SD less than the means, stratified by years of education and age.

Table 4

Normative data of MMSE based on percentile ranks (second approach)

z -score	Percentile	Education ≤10 y	Education 11–13 y	Education ≥14 y
Age ≤50 y (midpoint 49) estimated from ages ≤54 y, N = 92
		N = 7	N = 30	N = 55
0	0.50	27	29	29
–1.0	0.16	25	25	27
–1.5	0.07	–	23	25
–2.0	0.02	–	19	22
Age [51; 53] y (midpoint 52) estimated from ages [47; 57] y, N = 107
		N = 15	N = 35	N = 57
0	0.50	28	29	29
–1.0	0.16	25	27	27
–1.5	0.07	18	23	24
–2.0	0.02	–	23	22
Age [54; 56] y (midpoint 55) estimated from ages [50; 60] y, N = 145
		N = 24	N = 49	N = 72
0	0.50	27	29	29
–1.0	0.16	25	27	27
–1.5	0.07	23	23	24
–2.0	0.02	18	23	22
Age [57; 59] y (midpoint 58) estimated from ages [53; 63] y, N = 173
		N = 29	N = 60	N = 84
0	0.50	28	29	29
–1.0	0.16	25	27	27
–1.5	0.07	23	23	26
–2.0	0.02	18	22	23
Age [60; 62] y (midpoint 61) estimated from ages [56; 66] y, N = 203
		N = 41	N = 66	N = 96
0	0.50	28	29	29
–1.0	0.16	24	26	27
–1.5	0.07	22	22	26
–2.0	0.02	14	22	24
Age [63; 65] y (midpoint 64) estimated from ages [59; 69] y, N = 210
		N = 48	N = 68	N = 94
0	0.50	28	29	29
–1.0	0.16	23	26	28
–1.5	0.07	20	23	26
–2.0	0.02	14	22	24
Age [66; 68] y (midpoint 67) estimated from ages [62; 72] y, N = 292
		N = 70	N = 81	N = 141
0	0.50	27	28	29
–1.0	0.16	25	26	28
–1.5	0.07	21	23	26
–2.0	0.02	18	22	24
Age [69; 71] y (midpoint 70) estimated from ages [65; 75] y, N = 379
		N = 114	N = 84	N = 181
0	0.50	27	28	29
–1.0	0.16	24	26	27
–1.5	0.07	22	24	26
–2.0	0.02	18	22	25
Age [72; 74] y (midpoint 73) estimated from ages [68; 78] y, N = 434
		N = 150	N = 85	N = 199
0	0.50	27	28	29
–1.0	0.16	23	25	27
–1.5	0.07	20	24	26
–2.0	0.02	16	19	24
Age [75; 77] y (midpoint 76) estimated from ages [71; 81] y, N = 442
		N = 173	N = 79	N = 190
0	0.50	27	28	29
–1.0	0.16	23	25	27
–1.5	0.07	21	23	25
–2.0	0.02	16	19	21
Age [78; 80] y (midpoint 79) estimated from ages [74; 84] y, N = 362
		N = 168	N = 61	N = 133
0	0.50	27	28	29
–1.0	0.16	23	25	26
–1.5	0.07	20	21	24
–2.0	0.02	16	18	21
Age [81; 83] y (midpoint 82) estimated from ages [77; 87] y, N = 252
		N = 142	N = 38	N = 72
0	0.50	27	27	28
–1.0	0.16	23	26	26
–1.5	0.07	20	21	24
–2.0	0.02	16	18	21
Age [84; 86] y (midpoint 85) estimated from ages [80; 90] y, N = 170
		N = 100	N = 30	N = 40
0	0.50	27	27	28
–1.0	0.16	24	26	25
–1.5	0.07	22	24	21
–2.0	0.02	18	21	16
Age ≥87 y (midpoint 88) estimated from ages ≥83 y, N = 107
		N = 66	N = 17	N = 24
0	0.50	26	27	28
–1.0	0.16	23	26	24
–1.5	0.07	21	24	21
–2.0	0.02	21	–	16

DISCUSSION

Our results suggest that normative MMSE values depend on education and age but not on sex or marital status, and that education is the strongest predictor. As in similar studies, we found that more years of education and younger age were both associated with higher scores on the MMSE [14 , 16–28]. The association between MMSE score and age was not linear, however; MMSE scores for age up to about 65 years were quite similar, but among older persons increasing age showed clear decay in scores, which may indicate that the cognitive aging process starts at the age of approximately 65 years. Another explanation could be that the MMSE is too simple for distinguishing the performance of people less than 65 years old.

Those results align with the findings of studies from countries with similar education systems, which suggests that the test is robust. Translation and adaptation into different Western languages do not seem to have influenced the primary test results given that normative values found in the United States, Ireland, and Sweden are comparable to the values that we determined [19 , 28]. However, studies from South Korea and Japan, two countries with non-Western cultures, have shown lower normative scores on the MMSE [14 , 27]. Other studies from Asia have also revealed higher MMSE scores among males than females with similar education and of the same age [14, 26]. It is difficult to explain that difference between the sexes. However, men have also been shown to perform better on visuo-construction and mathematical tasks than women [40, 41]. Because the task serial seven of the MMSE test gives 5 points and drawing the two pentagons 1 point, those two tasks may explain why males have outperformed females in some countries. At the same time, females have scored higher on verbal memory tests than males [41 –43]. One task of the MMSE, awarding 3 points at most, involves recalling three words, which most people find rather easy to do. In a study examining normative scores on MoCA, where the test taker should recall five words, we found that females performed significantly better than males [41].

Normative MMSE scores are important when evaluating individual performance. Scores less than 1.5 SD of the normative score meet the criteria for the diagnosis of MCI in individuals with intact ADL, and many would argue that scores less than 1.0 SD should also be considered to indicate significant cognitive impairment (i.e., MCI) [3, 4]. If someone with significant cognitive impairment is dependent in ADL given a cognitive dysfunction, then a dementia disorder could be present. We argue that individuals scoring between 1.0 SD and 1.5 SD less than the normative score on the MMSE, adjusted for education and age, should be considered for further assessment. Meanwhile, anyone scoring more than 1.5 SD below the normative score should be offered further assessment to improve their etiological diagnosis. The use of a single cutoff score for the case-based detection of a significant cognitive disorder is not recommended, however. For example, according to our results, a score of 24 for an 81-year-old with 10 years of schooling implies normal performance (Z-score = –0.84 and 20th percentile according to the first approach and Z-score > –1.0 or being between 16th and 50th percentile according to the second approach). By comparison, the same score for a 71-year-old with at least 14 years of education indicates pathological performance (Z-score = –1.83 and 3.4% percentile according to the first approach and Z-score < –2.0 or being under 2nd percentile according to the second approach).

Limitations and strengths

Our study had some limitations. For one, the design was imperfect. Ideally, the best design would have been to include persons at random from a representative population sample and examine them comprehensively in order to exclude ones with cognitive impairment due to a defined diagnosis. The HUNT4 Trondheim 70+ cohort was nearly perfect in that respect, except for representativeness. Individuals who declined participation had significantly lower MoCA scores than the participants. Beyond that, 65% of participants had at least 14 years of education. Thus, the cohort arguably consisted of “supernormal” older adults, which could explain why the Trondheim cohort had higher mean MMSE scores than the NorCog cohort. However, because 860 of 1,050 of our participants (82%) were recruited from NorCog, we do not consider that limitation to have been a major one. The 860 participants from NorCog were clinical patients not selected at random, which was another limitation. Even so, they all had been subjected to a comprehensive assessment performed to exclude individuals with disorders possibly associated with cognitive impairment [31].

Our strict procedure for identifying cognitively healthy persons can also be interrogated, namely for having overlooked some such individuals. Not everyone with a neurological disorder, alcohol or substance abuse, depression, or other psychiatric or serious physical disorders had low MMSE scores, and some of them could have been included even if their scores at the group level were less than the scores of individuals without such disorders. We know who they are, but we do not know whether their MMSE scores on the day of testing represent their scores before they became ill with one of the disorders or conditions shown in Fig. 1. Such exclusion could have contributed to higher normative scores. To diagnose MCI, we used two definitions: Winblad et al.’s criteria for NorCog participants, which does not contain any cutoff on a cognitive test, and the DSM-5 criteria, which use –1.0 SD as the cutoff. Applying Winblad et al.’s criteria, which are based on the clinician’s judgment, could result in both the under- and overestimation of MCI, whereas using the DSM-5 criteria could result in an overestimation compared with the use of the Mayo criteria. It is difficult to ascertain whether that different could have biased our results, for we do not know the validity of the clinical diagnosis of MCI using Winblad et al.’s criteria. However, those criteria clearly express that such patients are “not normal” [5]. Our design and procedure for exclusion have been used in various studies on normative MoCA scores, which indicates that we used similar principles for selecting cognitively normal individuals [43 –45].

Furthermore, cognitively healthy persons should arguably not be assessed for cognitive impairment in specialist health services. We agree that they should not, but some individuals with anxiety and/or with relatives with dementia are nevertheless referred to outpatient clinics for the assessment of cognitive impairment. The argument has been supported by Medbøen et al.’s study [31], and clinicians are aware of that phenomenon from praxis.

Of course, our study also had strengths. The population was relatively large, and by including people more than 69 years old from HUNT Trondheim 70+, we were able to estimate normative scores for the “oldest old,” a group in which the prevalence of significant cognitive impairment is high. Many studies have not involved collecting normative data about the oldest old. However, we did not manage to include a larger group of people above the age of 89, which should be consider a limitation of the study. Second, our results align with normative scores on the MMSE achieved in countries with education systems comparable with Norway’s [19 , 28].

While most widely applied approach for deriving normative data based on Z-score derived from model residuals has shortcoming (skewed residuals might result in less reliable normative data), the second approach is more robust and not so dependent on strong assumptions, however, provides less nuanced picture and depends largely on sample size. The normative scores from two approaches agree well, and providing both sets strengthen the results.

In sum, normative scores on the MMSE should be adjusted for education and age and thereby made useful in daily clinical praxis to detect individuals with significant cognitive decline, a group that should be considered for further assessment in order to make etiological diagnoses of cognitive impairment. The MMSE is not a diagnostic test but merely a test that evaluates cognitive function on the day when the test is applied. In busy clinical practices, using a calculator (https://www.aldringoghelse.no/mmsenr3en) may therefore be valuable.

FUNDING

The study was financed by the Norwegian National Center for Aging and Health. There were no restriction in regard to research conduct.

CONFLICT OF INTEREST

The authors have no conflict of interest to report.

DATA AVAILABILITY

The data is available on request. Please contact the first author.

Footnotes

ACKNOWLEDGMENTS

We thank the boards of NorCog and HUNT Trondheim 70+ for giving us access to the data used in the study. The Trøndelag Health Study (HUNT) is a collaboration between the HUNT Research Centre (Faculty of Medicine and Health Sciences, Norwegian University of Science and Technology), Trøndelag County Council, Central Norway Regional Health Authority, and the Norwegian Institute of Public Health.

References

Gjøra

, Strand

, Bergh

, Borza

, Brækhus

, Engedal

, Johannessen

, Kvello-Alme

, Krokstad

, Livingston

, Matthews

, Myrstad

, Skjellegrind

, Thingstad

, Aakhus

, Aam

, Selbæk

(2021) Current and future prevalence estimates of mild cognitive impairment, dementia, and its subtypes in a population-based sample of people 70 years and older in Norway: The HUNT Study. J Alzheimers Dis 79, 1213–1226.

Kvello-Alme

, Bråthen

, White

, Sando

(2019) The prevalence and subtypes of young onset dementia inCentral Norway: A population-based study. J Alzheimers Dis 69, 479–487.

Petersen

(2004) Mild cognitive impairment as a diagnostic entity. J Intern Med 256, 183–194.

Albert

, DeKosky

, Dickson

, Dubois

, Feldman

, Fox

, Gamst

, Holtzman

, Jagust

, Petersen

, Snyder

, Carrillo

, Thies

, Phelps

(2011) The diagnosis of mild cognitive impairment due to Alzheimer’s disease: Recommendations from the National Institute on Aging–Alzheimer’s Association workgroups on diagnostic guidelines for Alzheimer’s disease. Alzheimers Dement 7, 270–279.

Winblad

, Palmer

, Kivipelto

, Jelic

, Fratiglioni

, Wahlund

, Nordberg

, Bäckman

, Albert

, Almkvist

, Arai

, Basun

, Blennow

, de Leon

, DeCarli

, Erkinjuntti

, Giacobini

, Graff

, Hardy

, Jack

, Jorm

, Ritchie

, van Duijn

, Visser

, Petersen

(2004) Mild cognitive impairment –beyond controversies, towards a consensus: Report of the International Working Group on Mild Cognitive Impairment. J Intern Med 256, 240–246.

McGrattan

, Pakpahan

, Siervo

, Mohan

, Reidpath

, Prina

, Allotey

, Zhu

, Shulin

, Yates

, Paddick

, Robinson

, Stephan

BCM

; DePEC team (2022) Risk of conversion from mild cognitive impairment to dementia in low- and middle-income countries: A systematic review and meta-analysis. Alzheimers Dement 8, e12267.

Engedal

, Barca

, Høgh

, Bo Andersen

, Winther Dombernowsky

, Naik

, Gudmundsson

, Øksengaard

, Wahlund

, Snaedal

(2020) The power of EEG to predict conversion from mild cognitive impairment and subjective cognitive decline to dementia. Dement Geriatr Cogn Disord 49, 38–47.

Chen

, Qian

, Zhang

, Su

, Huang

, Wang

, Chen

, Zhao

, Han

, Ma

(2022) Prediction models for conversion from mild cognitive impairment to Alzheimer’s disease: A systematic review and meta-analysis. Front Aging Neurosci 14, 840386.

Folstein

, Folstein

, McHugh

(1975) “Mini–mental state”: A practical method for grading the cognitive state of patients for the clinician. J Psychiatr Res 12, 189–198.

10.

Busch

, Chapin

(2008) Review of normative data for common screening measures used to evaluate cognitive functioning in elderly individuals. Clin Neuropsychol 22, 620–650.

11.

Engedal

, Haugen

, Gilje

, Laake

(1988) Efficacy of short mental tests in the detection of mental impairment in old age. Compr Gerontol A 2, 87–93.

12.

Mitchell

(2009) A meta-analysis of the accuracy of the Mini–Mental State Examination in the detection of dementia and mild cognitive impairment. J Psychiatr Res 43, 411–431.

13.

Rosli

, Tan

, Gray

, Subramanian

, Chin

(2016) Cognitive assessment tools in Asia: A systematic review. Int Psychogeriatr 28, 189–210.

14.

, Jia

, Yang

(2016) Mini–Mental State Examination in elderly Chinese: A population-based normative study. J Alzheimers Dis 53, 487–496.

15.

Bartos

, Raisova

(2016) The Mini–Mental State Examination: Czech norms and cutoffs for mild dementia and mild cognitive impairment due to Alzheimer’s disease. Dement Geriatr Cogn Disord 42, 50–57.

16.

Santiago-Bravo

, Sudo

, Assuncão

, Drummond

, Mattos

(2019) Dementia screening in Brazil: A systematic review of normative data for the Mini–Mental State Examination. Clinics (Sao Paulo) 74, e971.

17.

Arevalo-Rodriguez

, Smailagic

, Roqué-Figuls

, Ciapponi

, Sanchez-Perez

, Giannakou

, Pedraza

, Bonfill Cosp

, Cullum

(2021) Mini–Mental State Examination (MMSE) for the early detection of dementia in people with Mild Cognitive Impairment (MCI). Cochrane Database Syst Rev 7, CD010783.

18.

Huo

, Lin

, Bat

BKK

, Chan

JYC

, Tsoi

KKF

, Yip

BHK

(2021) Diagnostic accuracy of dementia screening tools in the Chinese population: A systematic review and meta-analysis of 167 diagnostic studies. Age Ageing 28, 1093–1101.

19.

Cullen

, Fahy

, Cunningham

, Coen

, Bruce

, Greene

, Coakley

, Walsh

, Lawlor

(2005) Screening for dementia in an Irish community sample using MMSE: A comparison of norm-adjusted versus fixed cut-points. Int J Geriatr Psychiatry 20, 371–376.

20.

Crum

, Anthony

, Bassett

, Folstein

(1993) Population-based norms for the Mini–Mental State Examination by age and educational level. JAMA 269, 2386–2391.

21.

Tombaugh

, McIntyre

(1992) The mini-mental state examination: A comprehensive review. J Am Geriatr Soc 40, 922–935.

22.

Dufouil

, Clayton

, Brayne

, Chi

, Dening

, Paykel

, O’Connor

, Ahmed

, McGee

, Huppert

(2000) Population norms for the MMSE in the very old: Estimates based on longitudinal data. Mini-Mental State Examination. Neurology 55, 1609–1613.

23.

Anstey

, Matters

, Brown

, Lord

(2000) Normative data on neuropsychological tests for very old adults living in retirement villages and hostels. Clin Neuropsychol 14, 309–317.

24.

Laks

, Baptista

, Contino

, de Paula

, Engelhardt

(2007) Mini-Mental State Examination norms in a community-dwelling sample of elderly with low schooling in Brazil. Cad Saude Publica 23, 315–319.

25.

Kenny

, Coen

, Frewen

, Donoghue

, Cronin

, Savva

(2013) Normative values of cognitive and physical function in older adults: Findings from the Irish Longitudinal Study on Ageing. J Am Geriatr Soc 61(Suppl 2), S279–S290.

26.

Kim

, Park

, Kim

, Chi

, Kim

, Moon

, Park

, Bae

, Woo

, Ryu

, Yoon

, Lee

, Chang

, Han

, Jhoo

, Han

, Cho

, Kim

(2002) Interactive influences of demographics on the Mini-Mental State Examination (MMSE) and the demographics-adjusted norms for MMSE in elderly Koreans. Int Psychogeriatr 24, 642–650.

27.

Sakuma

, Ura

, Miyamae

, Inagaki

, Ito

, Niikawa

, Ijuin

, Okamura

, Sugiyama

, Awata

(2017) Distribution of Mini-Mental State Examination scores among urban community-dwelling older adults in Japan. Int J Geriatr Psychiatry 32, 718–725.

28.

Kvitting

, Fällman

, Wressle

, Marcusson

(2019) Age-normative MMSE data for older persons aged 85 to 93 in a longitudinal Swedish cohort. J Am Geriatr Soc 67, 534–538.

29.

The Norwegian National Center for Aging and Health. https://www.aldringoghelse.no/mmsenr3en. Accessed 20 January 2023.

30.

Rowe

, Kahn

(1987) Human aging: Usual and successful. Science 237, 143–149.

31.

Medbøen

, Persson

, Nåvik

, Totland

, Bergh

, Treviño

, Ulstein

, Engedal

, Knapskog

, Brækhus

, Øksengård

, Horndalsveen

, Saltvedt

, Lyngroth

, Ranhoff

, Skrettingland

, Naik

, Soares

, Johnsen

, Selbaek

(2022) Cohort profile: The Norwegian Registry of Persons Assessed forCognitive Symptoms (NorCog) –a national research and quality registry with a biomaterial collection. BMJ Open 12, e058810.

32.

Åsvold

, Langhammer

, Rehn

, Kjelvik

, Grøntvedt

, Sørgjerd

, Fenstad

, Heggland

, Holmen

, Stuifbergen

, Vikjord

SAA

, Brumpton

, Skjellegrind

, Thingstad

, Sund

, Selbæk

, Mork

, Rangul

, Hveem

, Næss

, Krokstad

(2023) Cohort profile update: The HUNT Study, Norway. Int J Epidemiol 52, e80–e91.

33.

Nasreddine

, Phillips

, Bédirian

, Charbonneau

, Whitehead

, Collin

, Cummings

, Chertkow

(2005) The Montreal Cognitive Assessment, MoCA: A brief screening tool for mild cognitive impairment. J Am Geriatr Soc 53, 695–699.

34.

Engedal

, Kvaal

, Korsnes

, Barca

, Borza

, Selbaek

, Aakhus

(2012) The validity of the Montgo-mery-Aasberg depression rating scale as a screening tool for depression in later life. J Affect Disord 141, 227–232.

35.

Morris

, Mohs

, Rogers

, Fillenbaum

, Heyman

(1988) Consortium to establish a registry for Alzheimer’s disease (CERAD) clinical and neuropsychological assessment of Alzheimer’s disease. Psychopharmacol Bull 24, 641–652.

36.

Bjelland

, Dahl

, Haug

, Neckelmann

(2002) The validity of the Hospital Anxiety and Depression Scale. An updated literature review. J Psychosom Res 52, 69–77.

37.

Kaufer

, Cummings

, Ketchel

, Smith

, MacMillan

, Shelley

, Lopez

, DeKosky

(2000) Validation of the NPI-Q, a brief clinical form of the Neuropsychiatric Inventory. J Neuropsychiatry Clin Neurosci 12, 233–239.

38.

Crawford

, Garthwaite

, Slick

(2009) On percentile norms in neuropsychology: Proposed reporting standards and methods for quantifying the uncertainty over the percentile ranks of test scores. Clin Neuropsychol 23, 1173–1195.

39.

Pauker

(1988) Constructing overlapping cell tables to maximize the clinical usefulness of normative test data: Rationale and an example from neuropsychology. J Clin Psychol 44, 930–933.

40.

Aartsen

, Martin

, Zimprich

; Longitudinal Aging Study Amsterdam (2004) Gender differences in level and change in cognitive functioning. Results from the Longitudinal Aging Study Amsterdam. Gerontoloy 50, 35–38.

41.

Frias CM

, Nilsson

, Herlitz

(2006) Sex differences in cognition are stable over a 10-year period in adulthood and old age. Neuropsychol Dev Cogn B Aging Neuropsychol Cogn 13, 574–587.

42.

Asperholm

, Nagar

, Dekhtyar

, Herlitz

(2019) The magnitude of sex differences in verbal episodic memory increases with social progress: Data from 54 countries across 40 years. PLoS One 14, e0214945.

43.

Engedal

, Gjøra

, Benth

JŠ

, Wagle

, Rønqvist

, Selbv

(2022) The Montreal Cognitive Assessment: Normative data from a large, population-based sample of cognitive healthy older adults in Norway-The HUNT Study. J Alzheimers Dis 86, 589–599.

44.

Borland

, Nägga

, Nilsson

, Minthon

, Nilsson

, Palmqvist

(2017) The Montreal Cognitive Assessment: Normative data from a large Swedish population-based cohort. J Alzheimers Dis 59, 893–901.

45.

Thomann

, Goettel

, Monsch

, Berres

, Jahn

, Steiner

, Monsch

(2018) The Montreal Cognitive Assessment: Normative data from a German-speaking cohort and comparison with international normative samples. J Alzheimers Dis 64, 643–655.