Cognitive Screening and the MMSE: The Evidence Behind the Tools

A 70-year-old woman is admitted from a village two hours away with a urinary infection. She is drowsy in the afternoons and muddled at night, and pulls at her cannula. On the second day a house officer gives her the Mini-Mental State Examination, the MMSE, and she scores 17 out of 30. By the evening the word “dementia” is in the notes. Her son is upset: she ran the household until a month ago. She went to school for two years. This is an illustrative case, not a real patient.

Two tools on the MEDiscuss CDSS deal with her. The Cognitive Screening Pathway asks the questions in the right order: delirium first, then mood and drugs, then a cognitive screen whose cut-off is set by her schooling, then the reversible causes. MMSE Score & Interpretation scores an MMSE that has been given, domain by domain, reads it against her years of schooling, and compares it with an earlier score. What the two tools no longer carry is most of the why. The study figures, the national numbers and one full table have moved here, so that the tools stay quick at the bedside. By the end of this article you will know what that score of 17 can tell you about her, and what it cannot.

1. Delirium first

Why should a memory test wait? Because a delirious patient scored on a cognitive instrument produces a low number that looks exactly like dementia, and the label follows her out of the ward. Delirium is an acute, fluctuating disturbance of attention and awareness, usually caused by an illness or a drug. Our patient has an infection, drowsiness and night-time confusion. That is delirium until shown otherwise.

The Cognitive Screening Pathway uses the 4AT for this. It takes about two minutes and needs no training. A score of 4 or above means possible delirium. “How good is it?” Across seventeen studies and 3,702 observations, its pooled performance was sensitivity 0.88, specificity 0.88 (Tieges 2021). Sensitivity is the share of patients with delirium whom the test picks up; specificity is the share without delirium whom it correctly leaves alone. Both near 0.9 is good for a two-minute bedside test.

So the 17 she scored on day two is not her baseline. It is a measure of her delirium. Treat the infection, and test her cognition again once she has settled.

2. Years of schooling decide what the score means

Four weeks later she comes to the clinic, alert and well, and scores 22 out of 30. Is 22 abnormal? Many of us were taught that anything below 24 is. Here is why that rule fails her.

Crum and colleagues published medians by age and by educational level from a population sample in 1993 (Crum 1993). A median is the middle score: half the people scored above it, half below. The education gradient is the striking one: a median of 22 at nought to four years of schooling, 26 at five to eight years, 29 at nine years or more. The age gradient across the whole adult range, 29 in the youngest group falling to 25 at 80 and above, is smaller than the gap between an unschooled and a schooled elder of the same age.

Our patient, with two years of schooling, has scored exactly the median for her education. A cut-off of 24 would call the middle of her own population abnormal. That is why the Ministry’s guidelines for elderly care at Health and Wellness Centres refer at below 21 for a patient schooled below eighth class and below 24 above it, and why both tools ask for years of schooling before anything else. The MMSE tool prints the median for the patient’s own schooling beside every score.

“Is it schooling, or is it living in a village?” Both, it seems. In 2,651 Marathi speakers, rural residents scored lower after age, sex and education were controlled for, particularly on visuoconstructional ability, immediate recall and calculation. The tool names no paper for that study on its reference list, and neither can this article; hold it as a reason for caution rather than as a correction factor.

3. The two attention methods are not the same test

The attention section of the MMSE can be done in one of two ways. The original paper offers serial subtraction and, for the patient who will not or cannot do it, backward spelling, and custom has treated the two as equivalent. Ganguli and colleagues tested that in a population sample in 1990 and found them only weakly associated, with serial subtraction the harder of the pair and both strongly related to educational level (Ganguli 1990).

This is the common mistake, and an easy one to make: a different house officer, at the next visit, uses the other method because the patient struggled with the first. The score moves by several points and nothing has happened to the patient. Write the method beside the number. The MMSE tool asks for it, and refuses to compute a rate of decline when two administrations did not use the same one.

4. How good is the MMSE, in numbers

The diagnostic accuracy literature has measured two cut-offs, 24 and 25. The Cochrane review of community and primary care populations reports both, with the operating characteristics below. This is the table the MMSE tool carried on 4 November 2026.

“A sensitivity of 0.85 and a specificity of 0.90. That sounds like a good test. Why do the tools keep saying it cannot diagnose dementia?” Because of who is being tested. Screening a general clinic where perhaps five in a hundred have dementia, those figures return roughly four true positives and ten false positives per hundred tested. Read that again: in a general clinic, most people who fail the test do not have dementia. The reviewers concluded that the MMSE contributes to a diagnosis of dementia in low prevalence settings, but should not be used in isolation to confirm or exclude disease (Creavin 2016). The score opens an assessment; it does not close one.

Look also at the last row. The education-adjusted cut-off catches nearly everyone with dementia, at a sensitivity of 0.97, but its specificity of 0.70 means many well people are called abnormal. Adjusting for schooling is right, and it still does not make the MMSE a diagnosis.

5. How much change between two scores is real

Our patient has two scores now: 17 in the ward and 22 in the clinic. A rise of five points. Is she getting better, or is this noise?

Work on reliable change indices in people aged 75 and above put the change that exceeds measurement error at two to four points, once practice and regression to the mean are allowed for (Hensel 2007; Hensel 2009). A reliable change index is the smallest difference between two scores that is bigger than the test’s own wobble. Her rise of five points is real. Degenerative disease does not improve, so the MMSE tool reads a real improvement as a sign that the first score was depressed by something: here, the delirium.

A real fall, in the other direction, is still not a diagnosis. In that cohort it predicted incident dementia with a sensitivity of about 61 per cent and a specificity of about 72 per cent (Hensel 2009).

“How fast does the score fall in Alzheimer disease?” Meta-analysis usually puts the untreated trajectory in Alzheimer disease at around three points a year. A specialist clinic cohort of a hundred patients measured 2.43 points a year with a standard deviation of 2.82, a wide distribution around a modest mean (Miyakawa-Liu 2022). The standard deviation is larger than the mean: some patients barely move, and some fall fast. The tool treats two to four points a year as the expected band, and flags a fall faster than six points a year as pointing away from a purely degenerative process, towards delirium, a drug, depression, a vascular event, normal pressure hydrocephalus or a subdural collection.

6. The MMSE did not travel well to rural India

Ganguli and colleagues developed the Hindi Mental State Examination for the Ballabgarh cohort in the Indo-US Cross-National Dementia Epidemiology Study, working item by item to remove the requirement for reading and writing and to make the content locally meaningful (Ganguli 1995). That work is the clearest available statement that three of the eleven MMSE domains are unusable in a largely unschooled population, and that translation alone does not fix it.

There is no validated Kannada MMSE. A Kannada MMSE improvised at the bedside is a reasonable clinical examination of cognition, and both tools ask you to record it as one, not as a score out of thirty. For a comparable number in a patient who never went to school, the Cognitive Screening Pathway uses three instruments that need no reading or writing, and gives the Indian cut-offs for the Addenbrooke’s Cognitive Examination, set by schooling.

“Why is the MMSE form itself not on the tool?” Because it is published commercially, its publisher asserts copyright over it, and asks for a written request even for a translation or a change of format. The tool gives the scoring rules, the thresholds and the interpretation, which are published freely, and asks you to use the form your department already holds. This article does the same.

7. How common dementia is, and why families come late

Her son asks a fair question: “Is this common, doctor?” Here are the figures the Cognitive Screening Pathway used to carry in its Evidence tab, under the heading “6. The Numbers Worth Knowing”.

• 7.4 per cent of Indians aged 60 and over have dementia, from LASI-DAD, the nationally representative study of 2017 to 2020 (Lee 2023). That is about 8.8 million people, projected to reach 16.9 million by 2036.
• Rural 8.4 per cent against urban 5.3 per cent. Women 9.0 per cent against men 5.8 per cent, but the sex difference loses statistical significance once age and education are adjusted for. Report that one carefully.
• Age and education are the two robust predictors. Sixty per cent of the LASI-DAD sample had no formal education at all.
• The treatment gap exceeds 90 per cent across most of the country. In a Goa study, of 81 people with dementia, 41 had seen a doctor in the previous three months and 4 had received a diagnosis and dementia-specific treatment. That figure is from 2009 and a current national one does not exist.

Why do families come late? The Dementia India Report recorded the local names for it: chinnan, childishness, in Kerala; nerva frakese, tired brain, in Goa; weak brain in Banaras. In none of those settings was there any awareness of dementia as a disease of the brain (ARDSI 2010). A family that has waited three years has usually been told it is age.

“And if it is dementia, could it be reversible?” Reported rates of reversible dementia range from 0 to 40 per cent across studies, averaging around 12 per cent, and potentially reversible causes are found in about 18 per cent of patients under 65 against 5 per cent of those over 65. The figure the tool asks you to hold on to is smaller: a potentially reversible cause in 9 per cent, and 0.6 per cent of cases that actually reversed (Siddiqui 2025). The yield is low. The tool still asks for vitamin B12 and thyroid tests in everyone, because the ones you find are the ones you can treat.

8. The informant: the Short IQCODE answers, whole

The question the MMSE cannot answer is whether she has changed. Only someone who knew her ten years ago can answer that, and the Short IQCODE asks them. It carries its author’s declaration that there is no copyright on it. Here is its answer table as the Cognitive Screening Pathway carried it on 4 November 2026; on a phone the tool now shows it as two narrow tables. The Kannada is the tool’s own translation, not a validated version, and the tool asks for the English wording where the informant understands English.

Each of the sixteen items is answered on this scale, comparing the person now with ten years ago, not with other people of the same age. Her son answers “not much change” to almost everything until a month ago.

Back to the ward

Our patient was delirious on day two, and her 17 was a measure of that. At four weeks she scored 22, the median for her two years of schooling, and above the referral threshold of 21 for a patient schooled below eighth class. The rise of five points is larger than measurement error, which points to the first score having been depressed by the delirium. Her son reports little change over ten years. The word “dementia” comes out of her notes. Her vitamin B12 and thyroid tests are checked anyway, and she is seen again in six months, with the same attention method written beside the number.

Try this

A 68-year-old retired college lecturer, with seventeen years of schooling, scores 28 out of 30. His wife says he has stopped managing their money and has made two odd decisions this year. Is the score reassuring? Illustrative case.

Answer. No. The median at nine or more years of schooling is 29, so 28 is not high for him. More importantly, the MMSE has no executive item: nothing in it tests planning, judgement or initiative, which is exactly what his wife describes. Someone well schooled with early change routinely scores 27 or 28. A normal MMSE does not close the question; he needs an instrument with an executive component, and an informant history.

Take-home points

• Screen for delirium first, with the 4AT. A cognitive score taken through delirium is not a baseline.
• Read every MMSE against the patient’s years of schooling. A score of 22 is the median at nought to four years.
• In a general clinic most people who fail the MMSE do not have dementia. The score opens an assessment; it does not close one.
• A change of one or two points between visits is not a finding. Use the same attention method every time, and write it down.
• Ask someone who knew the patient ten years ago. The Short IQCODE asks the question the MMSE cannot.

Declaration: the author built and maintains both tools. They are free to use. The figures in this article are those the tools carried in their Instruments and Evidence tabs on 4 November 2026, each from the source named beside it; where a tool names no source on its reference list, the article says so. Both tools are still in administrator preview, pending clinical review. No item of the MMSE is reproduced here or in either tool.

References

MMSE Score & Interpretation

1. Folstein MF, Folstein SE, McHugh PR. “Mini-mental state”: a practical method for grading the cognitive state of patients for the clinician. J Psychiatr Res. 1975;12(3):189-198.
2. Crum RM, Anthony JC, Bassett SS, Folstein MF. Population-based norms for the Mini-Mental State Examination by age and educational level. JAMA. 1993;269(18):2386-2391.
3. Creavin ST, Wisniewski S, Noel-Storr AH, et al. Mini-Mental State Examination (MMSE) for the detection of dementia in clinically unevaluated people aged 65 and over in community and primary care populations. Cochrane Database Syst Rev. 2016;2016(1):CD011145. The source of the operating characteristics printed here, and of the conclusion that the MMSE should not be used in isolation to confirm or exclude disease.
4. Arevalo-Rodriguez I, Smailagic N, Roque-Figuls M, Ciapponi A, Sanchez-Perez E, Giannakou A, Pedraza OL, Bonfill Cosp X, Cullum S. Mini-Mental State Examination (MMSE) for the early detection of dementia in people with mild cognitive impairment (MCI). Cochrane Database Syst Rev. 2021;7(7):CD010783. Eleven studies, 1,569 participants. The basis for this tool declining to offer a conversion risk.
5. Ganguli M, Ratcliff G, Huff FJ, Belle S, Kancel MJ, Fischer L, Kuller LH. Serial sevens versus world backwards: a comparison of the two measures of attention from the MMSE. J Geriatr Psychiatry Neurol. 1990;3(4):203-207.
6. Ganguli M, Ratcliff G, Chandra V, Sharma S, Gilby J, Pandav R, Belle S, Ryan C, Baker C, Seaberg E, DeKosky S. A Hindi version of the MMSE: the development of a cognitive screening instrument for a largely illiterate rural elderly population in India. Int J Geriatr Psychiatry. 1995;10(5):367-377.
7. Ministry of Health and Family Welfare, National Health Systems Resource Centre. Operational Guidelines: Elderly Care at Health and Wellness Centres. New Delhi. Annexure V. The source of the two referral thresholds this tool leads with: refer at below 24 above eighth class and below 21 below eighth class.
8. Tombaugh TN, McIntyre NJ. The Mini-Mental State Examination: a comprehensive review. J Am Geriatr Soc. 1992;40(9):922-935.
9. Hensel A, Angermeyer MC, Riedel-Heller SG. Measuring cognitive change in older adults: reliable change indices for the Mini-Mental State Examination. J Neurol Neurosurg Psychiatry. 2007;78(12):1298-1303.
10. Hensel A, Luck T, Luppa M, Glaesmer H, Angermeyer MC, Riedel-Heller SG. Does a reliable decline in Mini Mental State Examination total score predict dementia? Dement Geriatr Cogn Disord. 2009;27(1):50-58.
11. Miyakawa-Liu M, Feehan AK, Pai S, Garcia-Diaz J. Rates of cognitive decline in 100 patients with Alzheimer disease. Ochsner J. 2022;22(2):129-133.
12. Newman JC, Feldman R. Copyright and open access at the bedside. N Engl J Med. 2011;365(26):2447-2449, with the fuller argument in the Stanford Technology Law Review. The basis on which the assertion of copyright over the MMSE has been questioned, and the reason this tool nonetheless does not reproduce the items.
13. Shaji KS, Sivakumar PT, Rao GP, Paul N. Clinical practice guidelines for management of dementia. Indian J Psychiatry. 2018;60(Suppl 3):S312-S328.

Cognitive Screening Pathway

1. MacLullich A, Ryan T, Cash H. 4AT: Assessment test for delirium and cognitive impairment, version 1.2. 2011–2014.
2. Tieges Z, Maclullich AMJ, Anand A, et al. Diagnostic accuracy of the 4AT for delirium detection in older adults: systematic review and meta-analysis. Age Ageing. 2021;50(3):733-743.
3. Bellelli G, Morandi A, Davis DHJ, et al. Validation of the 4AT, a new instrument for rapid delirium screening: a study in 234 hospitalised older people. Age Ageing. 2014;43(4):496-502.
4. Jorm AF. A short form of the Informant Questionnaire on Cognitive Decline in the Elderly (IQCODE): development and cross-validation. Psychol Med. 1994;24(1):145-153.
5. Khobragade P, Nichols E, Meijer E, Varghese M, Banerjee J, Dey AB, Lee J, Gross AL, Ganguli M. Cognitive decline as reported by an informant in India: findings from LASI-DAD. Int Psychogeriatr. 2022;36(3):177-187.
6. Borson S, Scanlan J, Brush M, Vitaliano P, Dokmak A. The Mini-Cog: a cognitive “vital signs” measure for dementia screening in multi-lingual elderly. Int J Geriatr Psychiatry. 2000;15(11):1021-1027.
7. Lessig MC, Scanlan JM, Nazemi H, Borson S. Time that tells: critical clock-drawing errors for dementia screening. Int Psychogeriatr. 2008;20(3):459-470.
8. Shulman KI. Clock-drawing: is it the ideal cognitive screening test? Int J Geriatr Psychiatry. 2000;15(6):548-561.
9. Mekala S, Paplikar A, Mioshi E, Kaul S, Divyaraj G, Coughlan G, et al. Dementia diagnosis in seven languages: the Addenbrooke’s Cognitive Examination-III in India. Arch Clin Neuropsychol. 2020;35(5):528-538.
10. Bhattacharyya B, Paplikar A, Varghese F, Alladi S, et al. Illiterate Addenbrooke’s Cognitive Examination-III in three Indian languages: an adaptation and validation study. Arch Clin Neuropsychol. 2025;40(3):642-654.
11. Lee J, Meijer E, Langa KM, Ganguli M, Varghese M, Banerjee J, et al. Prevalence of dementia in India: national and state estimates from a nationwide study. Alzheimers Dement. 2023;19(7):2898-2912.
12. Department of Health Research, Ministry of Health and Family Welfare, Government of India. Standard Treatment Workflow for the Management of Dementia. New Delhi; October 2019.
13. Siddiqui MA, Garg S, Tikka SK. Clinical practice guidelines on addressing cognitive impairment in treatable dementias. Indian J Psychiatry. 2025;67(1):41-53.
14. Shaji KS, Sivakumar PT, Rao GP, Paul N. Clinical practice guidelines for management of dementia. Indian J Psychiatry. 2018;60(Suppl 3):S312-S328.
15. Ministry of Health and Family Welfare, National Health Systems Resource Centre. Operational Guidelines for Elderly Care at Health and Wellness Centres. New Delhi. The source of the field thresholds used here: refer at MMSE below 24 above eighth class, below 21 below eighth class.
16. Nasreddine ZS, et al. MoCA Test mandatory training and certification: what is the purpose? J Am Geriatr Soc. 2020;68(2):444-446.
17. Paddick SM, Gray WK, McGuire J, Richardson J, Dotchin C, Walker RW. Cognitive screening tools for identification of dementia in illiterate and low-educated older adults: a systematic review and meta-analysis. Int Psychogeriatr. 2017;29(6):897-929.
18. Chakravarty K, Ray S. Cognitive screening tools for adults in Indian setting: a commentary. Ann Indian Acad Neurol. 2023;26(4):347-348.
19. Alzheimer’s and Related Disorders Society of India. The Dementia India Report 2010: prevalence, impact, costs and services for dementia. New Delhi: ARDSI; 2010.
20. Hodges JR, Larner AJ. Addenbrooke’s Cognitive Examinations: ACE, ACE-R, ACE-III, ACEapp and M-ACE. In: Larner AJ, editor. Cognitive Screening Instruments. 2nd ed. Cham: Springer; 2017.