Mini-Mental State Examination

Alzheimer’s/Dementia:

(Ashford, J.W., Schmitt, F.A., 2001 ; n = 981, Alzheimer’s/Dementia)

• As the severity of dementia increases, the standard error of measurement begins to decreases exponentially until 5 years and increases exponentially after 5 years. 

Alzheimer’s and Dementia: (Andrews et al, 2019); n=19,566 (normal cognition=8,404; MCI-Alzheimer’s Disease=2,815; mild AD dementia=3,163; moderate-to-severe AD dementia=1,509; remaining 3,675 did not meet criteria for cohorts)

• Across participants whose cognitive ability ranged from normal to moderate-severe AD dementia (n=15,891), on average, a 1-3 point decrease in MMSE was indicative of a meaningful decline.

Dementia: (Mitchell, 2009); 39 studies reviewed (dementia=34; MCI=5)

• Most commonly favored cutoff score for dementia was 23v24 (n=18)

Alzheimer’s/Dementia:

(Brugnolo, A., Nobili, F. et al, 2009); n = 524; mean age = 78.02 (6.07) years, Alzheimer's Disease/Dementia)

Alzheimer’s and Dementia: (Mougias et al., 2018); n=210 (AD=114; VaD=24; other/not-specified dementias=63); mean age=81.48±7.06 years

• Excellent internal consistency: Cronbach’s alpha 0.76

Alzheimer’s/Dementia:

(Freidl et al, 1996; n = 1947; ages 50-80 years; mean time post CVA = 8.5 days, Alzheimer’s/Dementia)

• The MMSE showed poor concurrent validity when compared to the Mattis Dementia Rating Scale (MDRS) (r = 0.29, p = 0.000)

Alzheimer’s and Dementia: (Mougias et al., 2018)

• Comparison of the SMMSE and MMSE showed excellent concurrent validity (r =0.820; p<0.001)

Convergent Validity:

Alzheimer’s and Dementia: (Mougias et al., 2018)

• Excellent convergent validity between MMSE and Global Deterioration Scale (r = -0.671)

Dementia: (Trzepacz et al., 2015; n = 100)

• Excellent convergent validity between MMSE and MOCA (r = 0.86)

Discrminant Validity

Alzheimer’s and Dementia: (Mougias et al., 2018)

• Adequate discriminant validity between MMSE and Katz Activities of Daily Living (r = 0.465)

Alzheimer’s and Dementia: (Mougias et al., 2018)

Poor floor effects: To evaluate floor effect, a variance index was calculated for the three groups, divided based on severity of cognitive decline as classified by MMSE scores. Group 1 (n=61; MMSE=0-6) and group 3 (n=37; MMSE=17-22) had the lowest variance index, (3.77 and 2.38). When considering the percentage of participants who scored into 0-6 (lowest MMSE quartile), 29% achieved these scores on the MMSE, indicating poor floor effects.

“Our findings reflect an inherent limitation of the MMSE to detect longitudinal changes in short‐term intervals in patients with severe dementia, which has been also reported even in long‐term intervals when functional decline is examined.” (Mougias et al., 2018)

Non-demented older adults:

(Jacqmin-Gadda, 1997; n = 2,537; mean age of participants = 74.4 (6.8) years, Non-demented older adults) 

• For patients aged 65 at initial assessment: SEM = 2.4 points at baseline and 2.0 points at 1, 3 and 5 years 
• For patients aged 85 at initial assessment: SEM = 3.5 points at baseline and 3.2 at year 1; 3.3 year 3 and 3.4 at year 5

Middle Aged and Older Adults  (Feeney, Savva, O’Regan, King-Kallimanis, Cronin & Kenny, 2016; n= 128; median age= 71)

• SEM for entire group (n=128): 1.02

Community Dwelling Older Adults:

(Rockwood et al, 2008; n = 3255; all participants older than 65 years; version = Modified MMSE (3MS), Community Dwelling Older Adults)

• 5 point change (or less) over a 5 to 10 year period

Middle Aged and Older Adults  (Feeney et al., 2016)

• MDC for entire group (n=128): 2.38 (90%); 2.84 (95%)

Elderly Patients:

(Lopez et al, 2005; n = 418, mean age = 77.73 (6.89) years; mean education = 12.53 (3.20) years; all patients were being screened for cognitive deficits, Elderly Patients)

• Lopez suggests that MMSE score should only be used to make one of two conclusions:
  • “needs no further evaluation” 
  • “needs a more extensive examination to determine if cognitive deficits exist

Cut-off scores suggested by prior research:

(Lopez et al, 2005)

• ≥ 24 = no impairment
• 18-23 = mild impairment
• ≤ 17 = severe impairment
• ≥ 27 = no impairment
• 21–26 = mild impairment
• 11–20 = moderate impairment
• ≤ 10 = severe impairment
• < 23 is generally accepted as indicating cognitive impairment and was associated with the diagnosis of dementia in at least 79% of cases (Lancu & Olmer, 2006)

Several sources of bias are known to exist with MMSE including: age, education, cultural and socioeconomic background. Various adjustments to MMSE scores have been suggested. (Mungas et al., 1996, n = 2983 individuals evaluated for possible dementia) For Hispanic individuals, a formula is MMSAdj = Raw MMS - (0.471 X [Education-12]) + (0.131 * [Age-70])

(Pedraza et al., 2011; n=3254 adults, 2819 Caucasian and 435 African-American, mean age 76.4 (7.3) years including 2048 cognitively normal and 1206 with some form of dementia)

• African American older adults had lower levels of education and significantly lower MMSE scores. A cut score of 23/24 was recommended for African-Americans to maximize classification accuracy (.96), with 92% sensitivity and 98% specificity.
• Attempts to identify an ideal cut point for patients with stroke have not been successful (Blake et al, 2002)

Norms for the MMSE and 3MS:

(Bravo & Herbert, 1997; n = 7754; mean age = 74.9 (6.8) years; 97.9% community dwelling, 2.1% institutionalized)

MMSE Score Pre & Post Rehab:

(Ozdemir et al, 2001; n = 43; mean age = 60.49 (8.1) years; mean rehab = 64.09 (18.27) days)

• Before Rehab: MMSE mean = 18.95; SD = 6.49 
• After Rehab MMSE mean = 22.79; SD = 5.46

Community Dwelling Older Adults:

(Andrew et al, 2008; version = Modified MMSE (3MS), Community Dwelling Older Adults)

• Longitudinal Modified MMSE scores:

  • 1 year mean = 90.4 (6.9)

  • 5 year mean = 89.1 (8.2)

  • 10 year mean = 85.8 (13.0) 

Non-demented Older Adults:

(Jacqmin-Gadda, 1997, Non-demented Older Adults)

• Mean MMSE score = 26.1 (3.1) points

• 20.2% had an MMSE score < 24 points

Participants 75 and Older:

(Aguero-Torres et al, 1998; n = 1745, Sweedish sample, Participants 75 and Older)

Older Adults:

(Molloy and Standish, 1997; n = 48 participants recruited from a nursing home and chronic care facility, assessed 3 times with 1 week intervals between testing -- age not reported, Older Adults)

• Adequate Interrater reliability (ICC = 0.69)

Middle Aged and Older Adults (Feeney et al., 2016)

• Excellent inter-/intra-rater reliability (ICC=0.75)

Community-Dwelling Patients (Brodaty et al., 2002; n = 283. Those >75 were included regardless of cognitive status. Subjects aged 50-74 had to have memory problems, but no existing diagnosis of depression or delirium)

• Adequate predictive validity for dementia (area under the curve of ROC was 0.85)

Community-Based Black Individuals: (Callahan et al., 2002; n = 344)

• Adequate predictive validity for cognitive impairment (area under the curve of the ROC was 0.85)
• Excellent predictive validity for dementia (area under the curve of the ROC was 0.96)

Referrals from Alzheimer Disease Center (Callahan et al., 2002; n = 651)

• Excellent predictive validity for cognitive impairment (area under the curve of the ROC was 0.93)
• Excellent predictive validity for dementia (area under the curve of the ROC was 0.95)

Convergent Validity

MCI Patients: (Trzepacz et al., 2015; n = 299)

·         Excellent convergent validity between MMSE and MOCA (r = 0.60)

Healthy Controls: (Trzepacz et al., 2015; n = 219)

·         Adequate convergent validity between MMSE and MOCA (r = 0.43)

Mixed Population: (Trzepacz et al., 2015; n = 618. Of this 618, n = 219 healthy controls, n = 299 MCI, and n = 100 dementia)

·         Excellent convergent validity between MMSE and MOCA (r = 0.84)

Community-Dwelling Patients (Brodaty et al., 2002; n = 283. Those >75 were included regardless of cognitive status. Subjects aged 50-74 had to have memory problems, but no existing diagnosis of depression or delirium)

·         Excellent convergent validity between MMSE and GPCOG patient section (r = 0.683)

Community-Based Black Individuals: (Callahan et al., 2002; n = 344)

·         Excellent convergent validity between MMSE and number of errors on Six-Item Screener (r = 0.77)

Referrals from Alzheimer Disease Center (Callahan et al., 2002; n = 651)

·         Excellent convergent validity between MMSE and number of errors on Six-Item Screener (r = 0.87)

Chinese Older Adults: (Tsai et al., 2016; n = 142)

• Adequate predictive validity for MCI (area under the curve of ROC was 0.88)
• Adequate predictive validity for dementia (area under the curve of ROC was 0.89)

Mixed Population: (Trzepacz et al., 2015); 618 cases: 219 cognitively normal healthy control, 299 MCI, and 100 AD dementia cases

• Poor ceiling effects: Ceiling effect of 28-30 points accounted for 71.4% of the healthy control and MCI group.

Neurocognitive Disorders and Parkinson's (Lucza et al., 2015; mild neurocognitive disorder due to PD- n=60; median age= 71; major neurocognitive disorder due to PD- n= 53; median age= 74)

• SEM for mild neurocognitive disorders due to PD= 2.32
• SEM for major neurocognitive disorders due to PD= 2.22

Neurocognitive Disorders and Parkinson's (Lucza et al., 2015);

• MDC (95% CI) for mild neurocognitive disorders due to PD=6.43
• MDC (95% CI) for major neurocognitive disorders due to PD= 6.16

Parkinson’s Disease:

(Djuradin et al, 2010)

• The best cut-off for distinguishing between demented and non-demented patients with PD was 26/30 for the MMSE.

Neurocognitive Disorders and Parkinson's (Lucza et al., 2015);

• For mild neurocognitive disorder due to PD= 27.5 (with a sensitivity of 0.602 and specificity of 0.706)
• For major neurocognitive disorder due to PD= 26.5 (with a sensitivity of 0.692 and specificity of 0.806)

This study determined cut-off scores being 27.5 and 26.5 for the MMSE when detecting mild and major neurocognitive disorders in PD respectively. These numbers represent an absence of mild and major NCDs.

Dementia in Parkinson’s Patients (Kaszas et al., 2012; PD patients with dementia- n=51; mean age= 62.7; SD= 8.4; PD patients without dementia- n=22; mean age= 64.6; SD= 10.6; )

• 26 points with sensitivity of 79.9% and specificity of 74.0%

This study found that most patients scored 26 points on the MMSE (which demonstrates an absence of dementia), with the support of the instrument having a higher ability to detect serious concerns.

Parkinson’s Disease: 

• MMSE score (n = 1120 PD with Dementia (n = 43): 19.6 (4.6), no dementia (n = 87): 28.5 (2.0) p < 0.001(Aarsland et al, 2001) 
• Patients with PD and Dementia had an annual decline from year 1 to year 4 of 2.3 (95% CI, 2.1 to 2.5; 9.1% change from year 1) (Aarsland et al, 2004) 
• Mean (SD) MMSE Score for PD and no dementia at year 1 was 27.3(3.7), was 23.8(8.0) at year 4, and 19.2(9.8) at year 8 (all P-values < 0.001). (Aarsland et al, 2004) 
• Mean (SD) Scores for PD with dementia at year 1 was 25.4(4.9), were 16.2(8.0) at year 4, and 7.2(6.5) at year 8. (Aarsland et al, 2004) 
• Comparison of Mean MMSE (28.8 + 1.1 and MoCA scores (24.9 + 3.1) in a chort of patients with PD who scored normal on the MMSE as part of the eligibility criteria- 52% met criteria for cognitive impairment based on the MoCA score (< 26). (Nazem et al 2009)

Parkinson’s Disease with Dementia and Parkinson’s Disease:

(Pagonabarraga et al, 2010; n = 70 for PD-ND and n = 32 for PDD, Parkinson’s Disease with Dementia & Parkinson’s Disease)

• Total scores were significantly lower in the PDD population
• MMSE scores: PD-ND = 27.9(2.1), PDD = 23(2.8) 
• P < 0.001

Parkinson’s Disease:

(Aarsland et al, 2004): 129 patients with PD (57%F)Mean age 70.0(8.1)yrs, duration of disease mean 8.6(4.9) yrs, H&Y stage mean 2.4(0.9). 

• MMSE negatively correlated with age (P < 0.001) and motor score B of the UPDRS (P < 0.001) 
• Also sig interaction of age by time (P < 0.01) and motor score B by time implying more rapid cognitive decline with higher age and higher levels of motor score B. 
• Predictors of more rapid cognitive decline in PD were: old age at study entry, hallucinations, and more severe motor symptoms, in particular, symptoms not associated with dopaminergic motor symptoms such as gait, speech, and postural disturbances. 
• In the model including UPDRS scores of bradykinesia and rigidity, age (P < 0.001) and rigidity score (P < 0.01) were negatively related to MMSE 
• Independent risk factors at baseline for dementia development in PD were age (OR, 1.2;95% CI, 1.1 to 1.2)/ MMSE Score < 29 (OR 3.3; 95%CI, 1.3 to 8.2) and H&Y Stage > 2 (OR, 3.4; 95% CI, 1.3 to 8.6). 
• At MMSE score < 24 at baseline the incidence of dementia in the PD was 74.8% (95% CI, 48.6 to 101.0) per 1000 persons. The OR associated with PD was 5.3 ((%% CI, 3.2 to 8.6) after adjustment for age, gender and education. 

Djudardin et al 2010: 

• Coefficient of the MMSE Scores < 26 was 2.21 (OR = 1 (reference), 95% CI, 3.03-27.28, p < 0.0001) for predicting PDD (parkinson’ s disease dementia) 

(Harvey, P.D., Ferris, S.H., et al, 2009, Parkinson’s Disease)

Study demonstrated criterion related validity for the ADAS-cog in people with PDD and showed strong correlation with the MMSE. This supports the validity of a previous clinical study’s results of patients with PDD that have used these measures. It also suggests that they are useful for future studies.

(Zadikoff et al, 2008):

• MMSE correlated slightly more with the UPDRS I item 1 (Spearmen correlation coefficients -0.34, P = 0.0013) than the MoCA (Spearman correlation coefficients -0.26, P = 0.0153) 
• More pronounced ceiling effect of the MMSE (27/88 subjects scored 30) compared to the MoCA (4.88 scored 30)

Dementia in Parkinson’s Patients (Kaszas et al., 2012); PD patients with dementia- n=51; mean age= 62.7; SD= 8.4; PD patients without dementia- n=22; mean age= 64.6; SD= 10.6; )

• Receiver Operating Characteristic (ROC) analysis- area under the curve: Adequate= 0.867 for dementia

Neurocognitive Disorders and Parkinson's (Lucza et al., 2015);

• Receiver Operating Characteristic (ROC) analysis- area under the curve: Adequate= 0.82 for mild neurocognitive disorder due to PD
• Receiver Operating Characteristic (ROC) analysis- area under the curve: Excellent = 0.90 for major neurocognitive disorder due to PD

Parkinson’s Disease:

(Aarslsand et al, 2004)

• Pearson correlation between MMSE and Dementia Rating Sale = 0.87 (P < 0.001) 

(Hoops et al, 2009, Parkinson’s Disease) 

Convergent Validity

Dementia in Parkinson’s Patients: (Kaszas et al., 2012)

• Adequate convergent validity between MMSE and Frontal Assessment Battery (r = 0.419)
• Excellent convergent validity between MMSE and Addenbrooke’s Cognitive Examination (r = 0.717)

Discriminant Validity

Dementia in Parkinson’s Patients: (Kaszas et al., 2012)

• Poor discriminant validity between MMSE and verbal fluency + language/orientation + memory (r = -0.251)

Parkinson’s Disease:

(Kandiah et al, 2009; n = 106, mean age = 61.2, Parkinson's Disease)

• The rate for those with no cognitive decline was 10.41 (SD = 0.97) points/year whereas the rate for patients with cognitive decline was 22.39 (SD = 2.07) points/year (Fig. 1)

Parkinson’s Disease:

(Lessig et al, 2012; n = 98, mean age = 68.8 (1.02), Parkinson's Disease)

• There were significant declines on the MMSE for both individuals with disease duration ≥ 10 and < 10 years (P < 0.001 and P = 0.01)
• This demonstrates that the MMSE measures cognitive decline over a prolonged period of time

Stroke: (Morris, Hacker, & Lincoln, 2011; n = 101; mean age = 76 years; mean time post stroke = 15 -18 months, Acute Stroke)

• At a cut-off score of 27 the MMSE showed good sensitivity (sensitivity, 0.80), low specificity (specificity, 0.20; positive predictive value, 0.84; negative predictive value, 0.16).

Stroke:

(Blake et al, 2002; mean age = 70.8 (12.2) years; < 4 weeks post stroke; mean Barthel Index score at recruitment = 10.5 (5.8); UK sample, Acute Stroke)

Stroke:

(Toglia et al, 2011; n = 72; mean age = 70 (17.0 years; mean time post CVA = 8.5 days), Stroke)

• The internal consistency of 11 MMSE items showed Cronbach alpha of 0.60, indicating that the items may not be measuring a unidimensional construct in persons with mild subacute stroke
• The internal consistency of 10 MoCA items showed stronger internal consistency with Cronbach alpha of 0.78, suggesting that it is a more reliable measure for persons with mild cognitive deficits

Stroke:

(Nys et al, 2005, Acute Stroke)

• Scores on the MMSE were no better than chance at detecting cognitive impairment

Stroke: (Morris, Hacker, & Lincoln, 2011; n = 101; mean age = 76 years; mean time post stroke = 15 -18 months, Acute Stroke)

• MMSE AUC = 0.53
• In the acute stroke population, the MMSE is not able to accurately discriminate between people with and without cognitive impairment.

Predictive Validity

Stroke: (Cumming et al., 2013; n = 60)

• Adequate predictive validity for cognitive impairment after stroke (area under the curve of ROC was 0.84)

After altering the criterion standard threshold to require 2 or more domain z-scores of <1.5 resulted in a classification of 32 of 60 (53%) patients as cognitively impaired, adequate predictive validity for cognitive impairment after stroke (area under the curve of ROC was 0.89)

Stroke:

(Godefroy et al, 2011; n = 95; mean age = 68.2 (13.7) years; mean time post CVA < 3 weeks, Stroke)

• Using raw scores, MoCA was more frequently impaired (P = 0.0001) than MMSE. MoCA showed good sensitivity (sensitivity, 0.94) but moderate specificity (specificity, 0.42; positive predictive value, 0.77; negative predictive value, 0.76), whereas an inverse profile was observed for MMSE (sensitivity, 0.66; specificity, 0.97; positive predictive value, 0.98; negative predictive value, 0.58) 

(Agrell & Dehlin, 2000, Acute Stroke)

• MMSE scores were found to significantly correlate with the BI, MADRS and Zung Depression Scale (p > 0.05)

Stroke: (Cumming et al., 2013; n = 60)

• Adequate reliability between MMSE and criterion standard which is determined on basis of neuropsychological battery (ICC = 0.57, CCC = 0.57, k = 0.57)
• After altering the criterion standard threshold to require 2 or more domain z-scores of <1.5 resulted in a classification of 32 of 60 (53%) patients as cognitively impaired, adequate reliability between the MMSE and criterion standard (ICC = 0.70. CCC = 0.70, k = 0.70)

Stroke:

(Toglia et al, 2011; n = 72; mean age = 70 (17.0) years; mean time post CVA = 8.5 days, Stroke)

• Compared to the MOCA, a more pronounced ceiling effect was noted for the MMSE

Stroke:

(Nys et al, 2005; n = 34; mean age = 64.7 (11.5) years; Dutch sample, Acute Stroke)

• MMSE AUC = 0.67* (standard error = 0.11)

  • Results indicate that the MMSE performed no better than chance at differentiating cognitive impairment from cognitively intact patients

*A measure with an AUC = 1 can be interpreted as having 100% sensitivity and specificity

The MMSE was NOT useful to assess memory problems or overall cognitive impairment after stroke. (Blake et al, 2002)

Meta-analytic Evidence:

(Tombaugh & McIntyre, 1992)

• Poor to Excellent test-retest reliability in MMSE's administered < 2 months (r = 0.38 - 0.99)

Mixed Diagnoses:

(Folstein et al, 1975; n = 63 normal elderly and 206 elderly with cognitive or emotional disorders, Mixed Diagnoses)

• Excellent 24 hour test-retest (same examiner) reliability (r = 0.88)
• Excellent 24 hour test-retest reliability (with different examiner; r = 0.827)
• Excellent 28 day (mean time between assessments) test-retest reliability (r = 0.988)

Meta-analytic Evidence:

(Tombaugh & McIntyre, 1992)

• Adequate to Excellent internal consistency (Cronbach's alpha = 0.54 to 0.96)

• Poor to Adequate MMSE and FIM correlations (Ozdemir et al, 2001; mean rehab = 64.09 (18.27) days)

Mixed diagnosis:(Folstein et

• Excellent convergent validity with:
  • WAIS (Wechsler Adult Intelligence Scale) Verbal IQ (r = 0.78) 
  • WAIS (Wechsler Adult Intelligence Scale) Performance IQ (r = 0.66) 

Traumatic Brain Injury:

(Srivastava et al, 2006; n = 43; mean age 66 (7.2) years; one year post TBI, mild to moderate severity; 60.5% sample mild, TBI) 

• Scores on MMSE were not useful to identify those with cognitive impairment, possible utility of attention items in identifying those who are not impaired. 

(de Guise et al, 2011; n = 162, 102 mild TBI, 30 moderate, 30 severe; see abstract below, TBI) 

• Other measures (clock drawing test and Trail Making Test - B) offered better utility for predicting outcome than the MMSE when tested early in inpatient rehabilitation