In an extensive review, Tombaugh and McIntyre (1992) reported moderate to high Test-Retest Reliability citing correlations of 0.38 to 0.99 in studies having a retest interval of<2 months (24/30 studies r>0.75).

Interobserver Reliability: Molloy and Standish (1997; elderly) reported an ICC of 0.69 for the traditional MMSE. Dick et al. (1984) reported K=0.63 and concordance correlation coefficient =0.87 between evaluations performed by GPs and those performed by psychologists (Fabrigoule et al. 2003).

Internal Consistency: Cronbach’s α coefficient of 0.54-0.96 has been reported by Tombaugh & McIntyre (1992).

Concurrent Validity: Tombaugh and McIntyre (1992) reported correlations of 0.70 to 0.90 between MMSE scores and other measures of cognitive impairment.

Construct Validity Correlations between ADL scores and the MMSE are 0.40-0.75. Tombaugh and McIntyre (1992) support the importance of cognitive status to functional outcome. Grace et al. (1995) reported significant association between FIM scores and MMSE scores (p<0.05), while Agrell and Dehlin (2000) (stroke) reported significant correlations between MMSE scores and BI, as well as between MADRS and the Zung Depression Scale (p<0.05). Lower MMSE scores are expected in stroke patients versus controls (p<0.001) and factor analysis revealed that 3 factors explained 53% of variance. The MMSE showed strong correlations with the WAIS-verbal (r=0.78) and performance- IQ (r=0.66) scores (Folstein et al. 1975). Dick et al (1984) reported r=0.55 and r=0.56 for verbal and performance IQ, respectively.

Construct Validity (known groups): MMSE scores could discriminate between groups based on categories of vocational recommendations (return to work, vocational training, supported work and continued remedial therapy; p<0.0001), and MMSE scores accounted for 36% variance between cell means (Mysiw et al. 1989; TBI). DePaolo and Folstein (1978; stroke) reported the MMSE was able to distinguish between patients with cerebral abnormalities and those with peripheral disorders only (p<0.0005).

Predictive Validity: Ozdemir et al. (2001; stroke) reported relationships between baseline MMSE scores and change in motor-FIM from admission to discharge among stroke rehabilitation patients (r=0.31; p<0.04), suggesting MMSE baseline scores are somewhat predictive of functional improvement.

Sensitivity & Specificity: Tombaugh & McIntyre (1992) reported an average sensitivity of 75% among dementia patients. Among general neurology and psychiatry patients, sensitivity was lower, ranging from 21-76%. A major variable in sensitivity was the level of impairment, as sensitivity of the MMSE increased with level of impairment. A low level of sensitivity is supported (Dick et al. 1984) as it is not sensitive to changes in patients with right-sided disease and is not useful in discriminating between focal versus diffuse disease, particularly among stroke patients (Grace et al. 1995) Sensitivity was reported as 44%, area under curve=0.7097 (Agrell & Dehlin, 2000). Agrell and Dehlin (2000) (stroke) reported MMSE could discriminate between patients with left-sided and infratentorial lesions (p<0.05) though not between right-sided and left-sided lesion groups. Tombaugh & McIntyre (1992) reported specificity of 62%-100%, while Agrell & Dehlin (2000) (stroke) reported 80%, and Grace et al. (1995) reported 84%. Blake et al. 2002 reported sensitivity=62% and specificity=88% in a group of stroke patients, where no suitable cut-off point could be identified if MMSE is used as a screening measure for verbal or visual memory deficits.

Modified Mini-Mental State Examination (3MS): Grace et al. (1995; stroke) compared the MMSE directly to 3MS. The test-retest stability of the 3MS was reported as r=0.80 and p<0.001).

Concurrent/Construct Validity: The 3MS correlated strongly with the MMSE at admission and discharge (r=0.84 and 0.85, respectively; p<0.001) and was also correlated with a battery of neuropsychological assessments (Controlled Oral Word Association, Boston Naming Test, Hooper Visual Organization Test, Logical Memory immediate and delayed, Visual Recall immediate and delayed, Wechsler Memory Scale Revised). Association with functional outcome (FIM) is stronger for the 3MS than for the MMSE (t=3.28, p<0.05). Using the standardized cut-off points for cognitive impairment and ROC analysis, the 3MS showed greater sensitivity than the MMSE (69% versus 44%) and similar specificity (80% versus 79%), area under the curve -0.7977 for 3MS.

3MS+Clock-drawing: To increase 3MS sensitivity among patients with right hemisphere stroke, Suhr & Grace (1999) (stroke) advocate the addition of the Wilson clock-drawing test. A clock-drawing task added < 2 min. to administration and increased sensitivity among stroke patients with right hemisphere lesions (87%). This testing format maintained a strong association with FIM scores (p<0.005).

Standardized MMSE: Molloy and Standish (1997) developed detailed instructions for administration and scoring of each item. Test retest variance was reduced by 86% and interobserver variance by 76% when the standardized MMSE was used. (Standardized MMSE:ICC=0.90; MMSE: ICC=0.69).

Telephone Version Adult Lifestyles and Functioning Interview: Includes 22/30 of the original MMSE items, the majority of which were removed from the last section (language and motor skills). Correlations between phone and face-to-face versions=0.85 (p<0.0001). Patients tended to do slightly better on in-person testing than on the telephone. Sensitivity (using a brief neurological screening test as the criterion) of 67% and specificity of 100% were reported in a population of elderly, community-dwelling individuals. This was similar to the sensitivity/specificity reported for screening with the traditional MMSE (68%, 100%) (Roccaforte et al. 19920)(elderly).

T-MMSE (26 item version of the Adult Lifestyles and Functioning Interview MMSE, Roccaforte et al. cited in Newkirk et al. 2004; dementia): T-MMSE correlated with the MMSE (r=0.88; p<0.001) and neither hearing impairment nor years of education were associated with T-MMSE scores. On the 22 points in common between the 2 scales, scores were correlated (r=0.88 p<0.001), but telephone scores tended to be higher than in-face scores (p<0.01) (Newkirk et al, 2004). The authors provide tables for the conversion of T-MMSE scores to MMSE scores.