INTERNATIONAL JOURNAL OF GERIATRIC PSYCHIATRY

Int J Geriatr Psychiatry 2002; 17: 134±139.

DOI: 10.1002/gps.541

Mini Mental State Examination in Sinhalese: a sensitive test

to screen for dementia in Sri Lanka
H. A. de Silva1* and S. B. Gunatilake2
1
Department of Pharmacology, Faculty of Medicine, University of Kelaniya, Sri Lanka
2
Department of Medicine, Faculty of Medicine, University of Kelaniya, Sri Lanka

SUMMARY
Background Although the Sri Lankan population is ageing rapidly, dementia has not been systematically investigated
here. The Mini Mental State Examination (MMSE) is a brief global instrument used to assess cognitive abilities in the
elderly.
Objective This study aimed to develop and validate a Sinhalese translation of the MMSE, which could be used as a screen-
ing instrument to detect impaired cognition in an epidemiological investigation of dementia in Sri Lanka.
Methods Due to the high literacy rate in the country, the MMSE was translated and modi®ed slightly without having to
make major changes to the original version. 380 randomly selected subjects over 65 years in a semi-urban area were
screened with the translated version of the MMSE. The cut-off score for cognitive impairment was taken as 17. Of the
380 subjects screened, 33 scored  17, and were thus considered cases of suspected dementia. All 33 who scored  17
and 24 randomly selected subjects who scored > 17 on the MMSE, thus considered cognitively normal, underwent a brief
clinical examination and neuropsychological assessment with the more comprehensive neuropsychiatric test battery, Cam-
bridge Cognitive Score (CAMCOG), to determine the presence of dementia.
Results Evaluated against the performance at the CAMCOG, the sensitivity and speci®city of the translated MMSE were
93.5% and 84.6% respectively.
Conclusion Therefore, the Sinhalese translation of the MMSE described here is a sensitive instrument to screen for
dementia in Sri Lanka. Copyright # 2002 John Wiley & Sons, Ltd.

key words Ð MMSE; Sri Lanka; Sinhalese; dementia; epidemiology; ageing

INTRODUCTION Dementia is one of the most distressing and bur-

densome health problems associated with an ageing
In most countries declining fertility rates and increas- population. The prevalence of dementia increases
ing life expectancy have led to an increase in the exponentially with age at least up to age 90 or 95.
population share of the elderly. In Sri Lanka too, the Studies in North America, the United Kingdom and
average life expectancy from birth for males has Europe have reported that dementia affects 1±6% of
increased from 62 years in 1963 to 70 years in the population over the age of 65 and 10±20% over
1991, and for females from 61 years to 74 years the age of 80 (Clark and Goate, 1993; Canadian Study
(Annual Health Bulletin, 1998). Thus, the percentage of Health and Aging, 1994; Ott et al., 1995) with the
of the population aged over 60 years in this country is annual incidence approximately doubling for every
expected to increase rapidly from 8% currently, to ®ve years of age between the ages of 75 and 89 years
reach 13% in 2010, and 21% in 2025 (Hsiao, 2000). (Paykel et al., 1994). Interestingly, evidence from a
few studies done in India (Chandra et al., 1998),
China (Li et al., 1989; Zang et al., 1990) and Nigeria
*Correspondence to: Dr H. A. de Silva, Department of Pharma-
cology, Faculty of Medicine, PO Box 6, Thalagolla Road, Ragama,
(Hendrie et al., 1995) suggests that dementia in gen-
Sri Lanka. Tel: 00 941 956188. Fax: 00 941 958251/958337. eral and Alzheimer's disease (AD) in particular may
E-mail: [email protected] be considerably rarer in developing than in developed
Received 16 November 2000
Copyright # 2002 John Wiley & Sons, Ltd. Accepted 26 June 2001
 mini mental state examination in sinhalese 135

countries. This could be due to differences in sam- and the university hospital are situated. Ragama was
pling procedures and the diagnostic criteria used for chosen as the sample area to take advantage of its
identifying subjects with dementia. Thus, con®rma- dense semi-urban population and its socio-economic
tion and elaboration of this discrepancy require more heterogeneity. The languages spoken are mainly
prospective studies in developing countries, using Sinhalese and English, and the great majority of
appropriately harmonized diagnostic procedures. elderly individuals living in this town are literate
The difference in disease frequency, if genuine, might and have had at least a few years of formal education.
suggest geographical variation in the distribution of Ragama is divided into eight primary healthcare
critical risk and protective factors for and against divisions (public health midwife areas). 380 subjects
dementia. over 65 years of age (mean age ˆ 68.2 years;
Dementia has not been systematically investigated SD ˆ 7.17) were selected randomly from two public
in Sri Lanka, and as a result there are no epidemiolo- health midwife areas (population of 8915) for the sur-
gical estimates of the disease in this country. Epide- vey. All 380 subjects were interviewed to obtain basic
miological studies of dementia require community demographic data, and scores on the MMSE (maxi-
screening of subjects for cognitive impairment with mum score ˆ 30). The subjects were all living in their
neuropsychiatric tests suitable for the cultural back- own homes or in that of a family member. As no local
ground of the study population. This requires transla- or national age-strati®ed household lists exist, the
tion and cultural adaptation of such screening initial contact with the subjects was made through a
instruments. The lack of such culturally appropriate Public Health Midwife (a community healthcare
screening instruments may have precluded studies worker). The assessment of all subjects was done by
on dementia in this country. the authors at a community centre in the neighbour-
The Mini Mental State Examination (MMSE; hood. All subjects had to be ¯uent in either English
Folstein et al., 1975) is used widely as a screening or Sinhalese, and those with major illnesses, deafness,
instrument to detect dementia both in the clinical set- blindness, mental retardation and major physical dis-
ting and in community based epidemiological studies, ability were excluded from the study.
and appears to be the best brief cognitive test cur-
rently in use (Brayne, 1998). The MMSE provides a
Instruments
measure of cognition, which is based on an inter-
viewer asking 17 questions each of which is noted The original MMSE was translated into Sinhalese and
as correct or incorrect. These questions cover a broad back translated by the authors (who are bilingual),
set of cognitive domains: orientation, registration, and the accuracy and cultural appropriateness of the
short-term memory, attention, calculation, visuo- translation were assessed by the Linguistics Unit
spatial skills and praxis. The MMSE, variously mod- and the Department of Sinhala, University of
i®ed and translated into several languages, has been Kelaniya, Sri Lanka. In order to make the Sinhalese
used successfully in several independent cross- translation of the MMSE applicable to subjects who
national studies of dementia epidemiology (Brayne, are illiterate or who have had little/no formal educa-
1998). The aim of this study was to develop and vali- tion as well, some modi®cations made in the Chinese
date a translated (to Sinhalese, the language used by (Katzman et al., 1988) and Hindi (Ganguli et al.,
more than 80% of the population) and culturally 1995) versions of the test were incorporated where
appropriate MMSE that could be used as a screening relevant. Table 1 provides a comparative brief
test for cognitive impairment in Sri Lanka. The result- description of the original and translated versions of
ing instrument will be used in an epidemiological the MMSE.
investigation of dementia to determine the prevalence Most items on the MMSE could be directly trans-
of, and risk factors for, dementia in a semi-urban lated to Sinhalese at the time of administering the test,
population in Sri Lanka. and a few required minor modi®cation. Attention was
given to make each item in the questionnaire appro-
priate for use in the Sri Lankan context in terms of
METHODS familiarity and cultural relevance. The item that
Subjects required major translation was the repetition phrase,
`no ifs, ands, or buts'. The modi®ed items in the trans-
The study was conducted in Ragama (a semi-urban lated version of the MMSE are brie¯y discussed
town 15 km North of Colombo, the administrative below in relation to items in the original English ver-
capital of Sri Lanka), where the faculty of medicine sion of the test.

Copyright # 2002 John Wiley & Sons, Ltd. Int J Geriatr Psychiatry 2002; 17: 134±139.
136 h. a. de silva and s. b. gunatilake

Table 1. Item-wise comparison of the translated original MMSE and translated version (®gures in parentheses in the table show maximum
score for each item)

Item MMSE Translated version

1. Orientation to time Date, day, month, year, season (5) Date, day, month, year, time of day (5)
2. Orientation to place Country, town, street, place, ¯oor (5) Country, town, street, place, ¯oor (5)
3. Three objects registration Apple, table, penny (3) Orange, table, rupee (3)
4. Calculation Subtract serial sevens backwards from 100 (5) Subtract serial threes backwards from 20** (5)
5. Recall Name the three objects learned earlier (3) Name the three objects learned earlier (3)
6. Language (Naming) Pencil, watch (2) Pencil, watch (2)
7. Repetition `No ifs, ands or buts' (1) `No ifs, ands or buts' in Sinhalese (1)
8. Language (Comprehension) Read and follow command Asked to follow verbal command,
`Close your eyes' (1) `Close your eyes'* (1)
9. Three step task Follow interviewer's instruction: `take paper Follow interviewer's instruction: `take paper
in right hand, fold in half and put on the table (3) in right hand, fold in half and put on the table (3)
10. Sentence construction Write a complete sentence (1) Say a complete sentence* (1)
11. Copying ®gure Overlapping pentagons (1) Overlapping pentagons (1)

*For illiterate subjects.

**For subjects who have had less than 5 years formal education.

Since there are no major seasonal variations in the accordance with other studies (Fillenbaum et al.,
climate in Sri Lanka, the question on `season' was 1984).
replaced with a question of `time of day'. Most Those screening positive for cognitive impairment
elderly residents in Ragama knew and kept track of (MMSE  17) and a randomly selected sample of
years according to the Roman calendar. Therefore, those screening negative (MMSE > 17) were asked
we did not have to replace this question with another to undergo a clinical examination and a cognitive
to test temporal orientation as done in the Hindi ver- assessment with the more comprehensive neuropsy-
sion of the MMSE (Ganguli et al., 1995). chiatric test battery, Cambridge Cognitive Score
`Apple', `table' and `penny' in the original MMSE (CAMCOG), to determine the presence of dementia.
were substituted with `orange', `table' and `rupee', The CAMCOG is the cognitive assessment section of
items more familiar to people in Sri Lanka. the Cambridge Mental Disorders of the Elderly
Those who had  ®ve years formal education were Examination (CAMDEX) (Roth et al., 1986). The
asked to subtract serial threes backwards from 20, family and cognitive history of the patient was
while those who had more than six years education obtained from the spouse or a family member using
were asked to subtract serial sevens backwards from section H of the CAMDEX. The clinical examination
100 as required in the original version of the MMSE. and the cognitive assessment with the CAMCOG
For the reading and writing tests, illiterate subjects were done in hospital. The CAMCOG was also
were asked to `close your eyes' instead of read and translated to Sinhalese at the time of administration,
follow a written instruction, and say a sentence and a few items in the test battery were modi®ed to
instead of write. This was done to guard against fail- make the test more appropriate to Sri Lankan
ure on these items due to illiteracy. subjects (e.g. `who is the current Prime Minister of
England' replaced with `who is the current President
Survey procedure of Sri Lanka'). The cut-off score for CAMCOG to
suggest cognitive impairment was taken as 80 out
Based on the results of two previous studies done in of a maximum possible score of 105 (Roth et al.,
China, the cut-off score on the MMSE suggestive of 1986).
cognitive impairment was taken as 17 out of a maxi-
mum 30 (Li et al., 1989; Zhang et al., 1990). The
RESULTS
lower cut-off, instead of the more standard cut-off
point of 23, has been found to be an appropriate score The age and sex composition of the sample is shown
to detect cases of cognitive impairment in community in Table 2. In the sample of 380 subjects, there were
surveys (Brayne and Calloway, 1990). Non-response 126 (33.1%) males and 254 (66.9%) females, more
or refusals on any speci®c item was scored as zero in than half (55.5%) of whom were between 65±70 years

Copyright # 2002 John Wiley & Sons, Ltd. Int J Geriatr Psychiatry 2002; 17: 134±139.
 mini mental state examination in sinhalese 137

Table 2. Age and sex composition of the sample studied

Age (Years) Male Female Total

Number % Number % Number %

65±70 75 19.8 136 35.7 211 55.5

71±75 30 7.9 73 19.2 103 27.2
76±80 20 5.3 33 8.6 53 13.9
81±85 Ð Ð 5 1.3 5 1.3
> 85 1 0.3 7 1.8 8 2.1
Total 126 33.1 254 66.9 380 100

of age. Education levels were generally low, espe- Table 3. Validity measuresÐsensitivity and speci®city, like-
cially among the very old subjects and women. lihood ratios and Youden's indicesÐfor some cut-off points for
the Sinhalese version of the MMSE validated against CAMCOG
More than half (54.2%) has had less than six years scores (n ˆ 57)
of formal education, 11.6% no formal education,
and 21 Subjects (5.5%) were illiterate. MMSE Sensitivity Speci®city Likelihood Youden's
Of the 380 subjects who were interviewed 33 cut-off % % ratio index
score
scored  17 on the MMSE (screened positive), and
were thus considered cases of suspected dementia. 23 100 69.2 3.3 0.69
All 33 subjects who screened positive, and 24 ran- 22 100 73.1 3.7 0.73
domly selected subjects who scored > 17 on the 21 100 76.9 4.3 0.77
20 100 80.8 5.2 0.81
MMSE (screened negative), thus considered cogni- 19 100 84.6 6.5 0.85
tively normal, underwent a brief clinical examination 18 96.8 84.6 6.3 0.81
and a neuropsychological assessment with the CAM- 17* 93.5 84.6 6.1 0.78
COG (total of 57 subjects). There were two illiterate 16 74.2 96.2 19.3 0.70
subjects among the screen positives and one illiterate *Cut-off point used by us.
subject among the selected screen negatives. There
were seven subjects with < six years of formal educa-
tion among the screen positives and ®ve with < six
DISCUSSION
years of formal education among the selected screen
negatives. The MMSE, variously translated and adapted to suit
After assessment with the CAMCOG and clinical many cultures, has been used as a screening instru-
examination, 29 out of the 33 subjects who screened ment among elderly people in many countries, which
positive on the MMSE showed evidence of dementia is generally considered as the ®rst step in prevalence
while four subjects scored above cut-off on the CAM- surveys in the identi®cation of individuals with a high
COG and showed no clinical signs of dementia. Of probability of being demented (Launer, 1992). The
the 24 randomly selected subjects who screened test takes very little time with an intact individual,
negative on the MMSE, 22 showed no evidence of and even with those who are severely impaired,
dementia while two scored below cut-off on the usually between 5±10 min, but can take longer with
CAMCOG and showed clinical evidence of dementia. those who are mildly impaired or who have auditory
Therefore, the sensitivity and speci®city of the trans- and communication dif®culties.
lated version of the MMSE used by us (when 17 was As the instrument has several items that require lit-
used as the cut-off point) were respectively 93.5% and eracy skills, performance at the MMSE is in¯uenced
84.6%. The sensitivity and speci®city and other valid- by level of education and literacy (Weiss et al., 1995).
ity measures such as, likelihood ratios for some cut- Such a precondition of literacy is impractical in most
off points of the Sinhalese version of the MMSE used developing countries where illiteracy and social back-
by us during this study are shown in Table 3. Since wardness are widely prevalent, particularly among the
this is the ®rst time this version of the MMSE elderly population. For example, in the study done in
has been used in any population, an item-wise Ballabgarh, India, the study population was largely
performance of the instrument in relation to CAM- illiterate (74%), many did not know or keep track of
COG scores obtained by all 57 subjects is shown in years, all buildings in the study area did not have
Table 4. street numbers and streets did not have names

Copyright # 2002 John Wiley & Sons, Ltd. Int J Geriatr Psychiatry 2002; 17: 134±139.
138 h. a. de silva and s. b. gunatilake

Table 4. Item wise performance by the two groups at the translated MMSE (®gures in parentheses for each subtest/item show maximum
score for each item)

Subtest/item Non-demented sample Demented sample

CAMCOG ( ) CAMCOG ( ‡ )
(n ˆ 26) (n ˆ 31)

Score % Score %

Orientation to time (5) 4.1 (0.8) 82 3.4 (1.2) 69

Orientation to place (5) 4.5 (0.7) 90 3.3 (1.0) 66
3-object registration (3) 2.6 (0.5) 86 1.9 (0.5) 63
Serial sevens/three (5) 3.8 (1.4) 76 0.5 (0.8) 10
3-object recall (3) 1.9 (1.1) 63 0.8 (0.6) 27
Name pencil and watch (2) 2.0 (0) 100 1.8 (0.6) 90
Repeat phrase (1) 0.8 (0.4) 80 0.3 (0.5) 30
3-step task (3) 2.7 (0.6) 90 1.7 (0.8) 57
Read/verbal command (1) 0.9 (0.3) 90 0.2 (0.4) 20
Write/say sentence (1) 0.6 (0.5) 60 0.2 (0.4) 20
Copy drawing (1) 0.6 (0.5) 60 0.2 (0.4) 20

(Ganguli et al., 1995). As a result, researchers have level without necessitating literacy skills for the ben-
had to greatly modify the MMSE to make it appli- e®t of the small percentage of illiterate subjects. Such
cable to largely illiterate rural populations in subjects were asked to follow a verbal command and
countries such as India (Ganguli et al., 1995) and also to say a complete sentence as were also the
Bangladesh (Kabir and Herlitz, 2000). We did not instructions given in the Chinese version of the
encounter this problem during our study because of MMSE used in the Shanghai study (Zhang et al.,
the high literacy rate in the country (90%) in general, 1990). These modi®cations were incorporated to the
and the study area in particular (95%) (Annual Health Sinhalese version of the MMSE to prevent illiteracy
Bulletin, 1998). Although 11.6% of the study popula- and poor education affecting performance at the
tion had no formal education and 54.2% had less than screening test.
six years of formal education, all cognitively intact Based on the results of two previous studies
subjects were familiar with writing instruments, and (Li et al., 1989; Zhang et al., 1990), one of which used
comfortable with items testing abstract mental arith- level of education-dependent cut-off points to screen
metic skills and copying overlapping pentagons as for dementia (Zhang et al., 1990), we used 17 as the
re¯ected in the scores obtained by them for these cut-off for our version of the MMSE instead of the
subtests (Table 4). As expected, subjects with more standard cut-off point of 23. We did this to mini-
impaired cognition fared particularly badly in the mize the likelihood of not detecting subjects with
same items requiring literacy skills (Table 4). cognitive impairment during screening. However, in
The selected study population in Ragama does not the population we studied, the best sensitivity and
vary signi®cantly in education, literacy, and cultural speci®city for our instrument would have been
sophistication from other urban/semi-urban areas of achieved if 19 was used as the cut-off point
the country, and is representative of the general Sri (Table 3). Although the cut-off used by us (17 out of
Lankan population. Therefore, the instrument was 30) gave a slightly lower sensitivity than 19, there was
translated to Sinhalese and made more appropriate no difference in speci®city for the two cut-off points
to our culture without having to make major modi®ca- (Table 3). Therefore, based on these results 19 is a bet-
tions to the original version. The abstract calculation ter cut-off point than 17 to screen for dementia with
of subtracting serial sevens backwards from 100, as the modi®ed MMSE in our study population.
required in the original MMSE, was substituted with The main purpose of the present study was to
subtracting serial threes backwards from 20 for those develop a version of the MMSE, translated to
with less than ®ve years of formal education. Items in Sinhalese and made more appropriate and relevant
the original MMSE which required the ability to read to the cultural context of Sri Lanka. In order to
and write such as following a command by reading a evaluate the sensitivity and speci®city of the modi®ed
written instruction (e.g. `close your eyes') and writing instrument, the MMSE scores of all subjects in our
a complete sentence respectively, were substituted study were evaluated against performances at the
with items which would test the desired cognitive CAMCOG, the cognitive section of the CAMDEX.

Copyright # 2002 John Wiley & Sons, Ltd. Int J Geriatr Psychiatry 2002; 17: 134±139.
 mini mental state examination in sinhalese 139

The CAMCOG allows observation of a broad range of Fillenbaum GG, George LK, Blazer DG. 1984. Scoring non-
response in the Mini Mental State Examination. Psychol Med
impairments associated with dementia in a relatively
18: 1021±1025.
short time (20±30 min). The total cognitive examina- Folstein MF, Folstein SE, McHugh PR. 1975. Mini mental state: a
tion with the CAMCOG consists of 68 items, which practical method for grading the cognitive state of patients for
assess orientation, language (comprehension and the clinician. J Psychiatry Res 12: 189±198.
expression), memory (recent and remote), new learn- Ganglui M, Ratcliff G, Chandra V, Sharma S, Gilby J, Pandav R.
1995. A Hindi version of the MMSE: the development of a cog-
ing, attention, calculation, praxis, abstract thinking, nitive screening instrument for a largely illiterate rural elderly
perception and executive function. Thus, in compari- population in India. Int J Ger Psychiatry 10: 367±377.
son to the MMSE, CAMCOG provides a wider cover- Hendrie HC, Osuntokun BO, Hall KS, Ogunniyi AO, Hai SL,
age of cognitive functions and detects milder degrees Unverzagt FW, et al. 1995. Prevalence of Alzheimer's disease
of cognitive impairment. This makes the CAMCOG a and dementia in two communities: Nigerian Africans and
African Americans. Am J Psychiatry 152: 1485±1492.
very comprehensive test battery to diagnose demen- Hsiao W. 2000. A Preliminary Assessment of Sri Lanka's Health
tia, and also to evaluate the sensitivity and speci®city Sector and Steps Forward. Institute of Policy Studies: Sri Lanka;
of screening instruments such as the MMSE (Roth 22.
et al., 1986). Kabir ZN, Herlitz A. 2000. The Bangla adaptation of the Mini Men-
tal State Examination (BAMSE): an instrument to assess cogni-
This survey is the ®rst time that a translated MMSE tive function in illiterate and literate individuals. Int J Ger
has been used as a screening instrument for cognitive Psychiatry 15: 441±450.
impairment in Sri Lanka, and this study has shown Katzman R, Zhang M, Ya-Qu O, Wang Z, Liu WT, Yu E, et al.
that the modi®ed version of the MMSE used by us 1988. Chinese version of the Mini Mental Examination: impact
to be simple to use and appropriate to the population of illiteracy in a Shanghai dementia survey. J Clin Epidemiol 41:
971±978.
studied. Therefore, we conclude that the slightly mod- Launer LJ. 1992. Overview of incidence studies of dementia con-
i®ed Sinhalese translation of the MMSE described ducted in Europe. Neuroepidemiol 11: 2±13.
here is a useful and sensitive instrument to screen Li G, Shen YC, Chen CH, Zhao YW, Li SR, Lu M. 1989. An epi-
for cognitive impairment in Sri Lanka. demiological survey of age-related dementia in an urban area of
Beijing. Acta Psychiatr Scand 79: 557±563.
Ott A, Breteler MM, van Harskamp F. 1995. Prevalence of
Alzheimer's disease and vascular dementia : the Rotterdam
REFERENCES study. B Med J 310: 970±973.
Paykel ES, Brayne C, Huppert FA, Gill C, Barkley C, Gehlhaar E,
Annual Health Bulletin. 1998. Ministry of Health, Sri Lanka. et al. 1994. Incidence of dementia in a population older than 75
Brayne C. 1998. The Mini Mental State Examination, will we be years in the United Kingdom. Arch Gen Psychiatry 51: 325±332.
using it in 2001? Int J Ger Psychiatry 13: 285±294. Roth M, Tym E, Mountjoy CQ, Huppert FA, Hendrie H, Verma S,
Brayne C, Calloway P. 1990. The case identi®cation of dementia in Goddard R. 1986. CAMDEX: a standardized instrument for the
the community: a comparison of methods. Int J Ger Psychiatry diagnosis of mental disorder in the elderly with special reference
5: 309±316. to the early detection of dementia. Br J Psychiatry 149: 698±709.
Canadian study of health and aging: study methods and prevalence Weiss BD, Reed R, Kligman EW, Abyad A. 1995. Literacy and per-
of dementia. 1994. Can Med Assoc J 150: 899±913. formance on the Mini Mental State Examination. J Am Geriatr
Chandra V, Ganguli M, Pandav P, Johnston J, Belle S, De Kosky ST. Soc 43: 807±810.
1998. Prevalence of Alzheimer's disease and other dementias in Zhang M, Katzman R, Salmon D, Jin H, Cai G, Wang Z, et al. 1990.
rural India: the Indo-US study. Neurology 51: 1000±1008. The prevalence of dementia and Alzheimer's disease in Shang-
Clark RF, Goate AM. 1993. Molecular genetics of Alzheimer's dis- hai, China: impact of age, gender and education. Ann Neurol 27:
ease. Arch Neurol 50: 1164±1172. 428±437.

Copyright # 2002 John Wiley & Sons, Ltd. Int J Geriatr Psychiatry 2002; 17: 134±139.