Comparing Detection of Alzheimer’s and Vascular Disease–Related Cognitive Impairment With Brief Cognitive Screens
Abstract
Objectives:
The study compared the accuracy of the Mini-Mental State Examination (MMSE) with its modified version (3MS) in distinguishing healthy older adults from adults with cognitive impairment due to suspected Alzheimer’s disease (AD) or vascular disease (VaD).
Method:
Participants were 98 veterans who underwent comprehensive neuropsychological evaluation due to concern for cognitive decline. Participants were selected via retrospective chart review on the basis of diagnosis. They had diagnoses of mild or major neurocognitive disorder due to suspected AD (N=20), mild or major neurocognitive disorder due to suspected VaD (N=44), or no neurocognitive diagnosis (i.e., healthy adult comparisons; HC, N=34).
Results:
The 3MS demonstrated superior detection of cognitive impairment. The extent of this enhanced detection was influenced by the suspected etiology of cognitive impairment. The 3MS and MMSE had comparable discrimination of AD and HC. With respect to VaD, the 3MS showed superior discriminability compared to the MMSE.
Conclusions:
Overall, results support the adoption of the 3MS over that of the MMSE. The 3MS is a superior (and free) tool for detecting cognitive impairment in geriatric populations. Its use is recommended for first-line screening of cognitive symptoms in older adult populations, especially those with concern for VaD.
Dementia, one of the leading causes of disability and death in older adults, is associated with substantial medical, social, and economic burden (1, 2). The estimated number of people currently living with dementia (47 million) is expected to increase to 75 million by 2030 (3). In 2022 alone, the expected health care costs associated with dementia are expected to exceed $300 billion (4). Although there are many types of dementia, Alzheimer’s disease (AD) and cerebrovascular disease (VaD) are the two most common etiologies and account for most of the dementia-related economic burden (5, 6). Prompt diagnosis of these dementias is key in providing opportunities for quality-of-life interventions and treatments that may attenuate cognitive decline and reduce socioeconomic burden (7). The gold standard for diagnosis of dementia remains a comprehensive neurocognitive evaluation (8); however, cognitive screens in various medical settings often represent the first and best method for early detection of cognitive decline and dementia. Despite this fact, cognitive screens are underused and dementia remains underdiagnosed in medical settings (4, 9, 10).
Many cognitive screening tasks have been developed to efficiently and accurately identify abnormal cognitive aging. One of the most widely used and best validated screens is the Mini-Mental State Examination (MMSE) (11, 12). Despite its clinical utility, it has been criticized for its limited coverage of executive functions, language, and visuocognitive abilities (13–15). It has also been criticized for a restricted scoring range with ceiling effects (14, 16). Finally, the MMSE is copyrighted intellectual property that cannot be used without financial cost (17).
In response to these and other limitations, the Modified Mini-Mental State Examination (3MS) was developed (18). The 3MS includes additional items covering executive functions, memory, and language abilities and uses an expanded scoring system to better capture a range of performances. Notably, the 3MS is free for clinical use (unlike the MMSE) and does not require costs associated with mandatory training/certification, unlike other popular screen alternatives (e.g., the Montreal Cognitive Assessment) (19). The clinical utility of the 3MS is supported by multiple studies indicating improved reliability over the MMSE across a range of clinical populations and settings (13, 20–23).
Given the practical and clinical advantages, some have argued for widespread substitution of the 3MS for the MMSE (15). However, research has been mixed regarding the differential sensitivity of the 3MS and MMSE in detecting dementia. Studies employing samples with heterogenous dementing etiologies have found the 3MS to be more sensitive not only to dementia (i.e., major neurocognitive disorder) but also to mild cognitive impairment (i.e., mild neurocognitive disorder) (13, 15, 18, 20, 24). In contrast, these screens have demonstrated relatively comparable sensitivities when detecting dementia due to probable and possible AD (23). Differences with respect to dementia etiology (i.e., mixed dementias vs. AD) across these study samples may be a driving force in their discrepant findings.
To date, no study has compared the classification accuracies of the 3MS and MMSE in detecting cognitive decline from the two most common neurodegenerative disease processes in a single sample. The primary objective of the study was to compare the differential utility of these measures in predicting an eventual diagnosis of AD or VaD. We hypothesized that 3MS classification accuracy would exceed that of the MMSE and that this advantage would be most pronounced in detecting VaD.
Methods
Participants
Participants consisted of 98 male veterans referred for comprehensive neuropsychological assessment due to concern for cognitive decline. Sixty-four adults met criteria for a mild or major neurocognitive disorder due to objective cognitive impairment and the extent of functional decline. Of these, cases were subdivided into two groups designated by predicted etiology: cognitive impairment due to AD (N=20) or VaD (N=44). Healthy adults in the comparison group (HC; N=34) were those with cognitive complaints who did not meet criteria for any cognitive disorder.
Procedures
All assessments took place at a southeastern Department of Veteran Affairs hospital from March 2, 2017, to December 21, 2017. All study procedures were approved by the local institutional review board and the research and development board of the Department of Veteran Affairs. Informed consent was waived due to the retrospective nature of the study.
Evaluations for clinical purposes were completed by licensed psychologists who were board certified or eligible for board certification in neuropsychology, with frequent involvement of a neuropsychology trainee under direct supervision (e.g., graduate students, interns, and postdoctoral fellows). All participants underwent neuropsychological evaluations consisting of a comprehensive neuropsychological test battery as well as clinical interview with the patient and, if available, collateral informants such as family. Prior to the evaluation, veterans completed the 3MS (with derived MMSE). Diagnoses and etiological conclusions were made by the licensed provider at the time of the evaluation with integration of medical record review, neuroimaging, laboratory work, neuropsychological test performance, and a clinical interview conducted with the patient and, if available, collateral informants, such as family members. Diagnoses were made in accordance with criteria put forth by the DSM-5 (25). For this study, cases were retrospectively selected from archival clinical data based on diagnosis after comprehensive neuropsychological evaluation.
Measures
The MMSE is a standardized cognitive screen that assesses orientation, attention, working memory, verbal memory, language, and visuoconstruction. Administration is brief (5–10 minutes), and there is empirical support for its sensitivity and specificity of identifying individuals with pathological cognitive impairment (11, 26). The 3MS is an enhanced version of the MMSE that was designed to be more standardized in its assessment, more granular in scoring, and to cover a broader variety of cognitive domains (18). It includes an expanded scoring system with four additional items: additional orientation questions, verbal fluency, verbal reasoning, and delayed recall (with semantic and multiple-choice cues).
Statistical Analysis
Descriptive statistics were run for demographic variables. Nonparametric statistics were employed for correlations (e.g., Spearman’s rho) and group comparisons of cognitive screen performance due to violations of parametric assumptions. Specifically, Kruskal-Wallis tests were run with 3MS and MMSE performance as the between-subjects factor. Mann-Whitney U tests were conducted for post hoc comparisons. Effect sizes were calculated with eta squared. Logistic regression models used cognitive screen performance (3MS, MSSE) as separate, individual predictors of group status (HC vs. cognitive impairment [CI]; HC vs. VaD; HC vs. AD). Area under the curve (AUC) for receiver operating characteristic (ROC) values were calculated in order to quantify and compare the relative discriminability of the cognitive screens. Discriminability was classified as acceptable (0.70≤AUC≤0.79), excellent (0.80≤AUC≤0.89), or outstanding (AUC≥0.90) (27).
Results
Table 1 summarizes descriptive statistics and univariate group comparisons for cognitive screen performance.
Healthy (N=34) | Vascular disease (N=44) | Alzheimer’s disease (N=20) | Significant contrastsa | |||||||
---|---|---|---|---|---|---|---|---|---|---|
Variable | M | SD | M | SD | M | SD | F | df | η2 | |
Age | 69.0 | 9.4 | 75.4 | 8.8 | 84.1 | 7.5 | 18.77*** | 2, 95 | 0.28 | HC<VaD<AD |
Education | 13.1 | 2.2 | 13.1 | 2.4 | 12.3 | 3.9 | 0.75 | 2, 95 | 0.02 | – |
MMSEb | 27.9 | 1.3 | 24.2 | 4.2 | 19.6 | 6.5 | 39.91*** | 2, 95 | 0.36 | HC>VaD>AD |
3MSc | 94.1 | 3.4 | 82.0 | 10.7 | 65.7 | 19.6 | 54.50*** | 2, 95 | 0.45 | HC>VaD>AD |
Descriptive statistics and group comparisons of cognitive screen performance for vascular disease, Alzheimer’s disease and healthy comparison groups
Kruskal-Wallis tests revealed 3MS and MMSE performance differed significantly across the three groups with large effect sizes (η2>0.30). Post hoc analyses revealed that HC performed significantly better than VaD across the 3MS and MMSE. Both HC and VaD performed significantly better than AD on both cognitive screens.
Ages of participants ranged from 45 to 95 and differed significantly across groups (F=18.77, p<0.001). The healthy control (HC) group (M=69.0, SD=9.4) was significantly younger than the VaD group (M=75.4, SD=8.8), and both HC and VaD groups were younger than the AD group (M=84.1, SD=7.5). Age is not thought to reflect a confounding variable in the current study. Evidence suggests the traditional pass/fail cut-score of the 3MS has equivalent, if not better, classification accuracy than the 3MS age-adjusted scoring when distinguishing healthy elderly adults from those with dementia or cognitive impairment (CI) (28). The HC (M=13.1, SD=2.1), VaD (M=13.1, SD=2.4), and AD (M=12.3, SD=3.9) groups were equivalent with respect to education.
Table 2 presents descriptive correlations for cognitive screen performance and demographic variables. As would be expected given the degree of test item overlap, the 3MS and MMSE were strongly correlated but not redundant (r=0.87). The 3MS and MMSE showed equivalent relationships with both age and education.
Variable | 1 | 2 | 3 | 4 |
---|---|---|---|---|
1. MMSEa | 1.00 | |||
2. 3MSb | 0.87** | 1.00 | ||
3. Age | −0.44** | −0.47** | 1.00 | |
4. Education | 0.26** | 0.22* | −0.06 | 1.00 |
Descriptive correlations for cognitive screens and demographic variables (N=98)
Table 3 presents logistic regression and ROC statistics for the 3MS and MMSE across all group comparisons of interest (HC vs. CI; HC vs. VaD; HC vs. AD). Both cognitive screens were significant predictors (p<0.001) of group status across all group comparisons. Table 4 shows classification accuracy results (e.g., hit rate, sensitivity, specificity) across 3MS and MMSE cutoffs for group comparisons of interest.
Group comparison | χ2 | R2 | OR | p | AUC |
---|---|---|---|---|---|
HC vs. CI | |||||
MMSEb | 42.96 | 0.49 | 0.53 | <0.001 | 0.86 |
3MSc | 63.75 | 0.66 | 0.67 | <0.001 | 0.93 |
HC vs. AD | |||||
MMSE | 43.49 | 0.76 | 0.40 | <0.001 | 0.92 |
3MS | 49.12 | 0.82 | 0.70 | <0.001 | 0.95 |
HC vs. VaD | |||||
MMSE | 31.15 | 0.44 | 0.52 | <0.001 | 0.83 |
3MS | 51.09 | 0.64 | 0.65 | <0.001 | 0.92 |
MMSEb cutoff | Hit rate | Sn | Sp | 3MSc cutoff | Hit rate | Sn | Sp |
---|---|---|---|---|---|---|---|
HC vs. CI | |||||||
MMSE | 3MS | ||||||
≤24 | 0.71 | 0.55 | 1.0 | ≤88 | 0.83 | 0.75 | 0.97 |
≤25 | 0.78 | 0.67 | 1.0 | ≤89 | 0.86 | 0.81 | 0.94 |
≤26 | 0.79 | 0.77 | 0.82 | ≤90 | 0.89 | 0.86 | 0.94 |
≤27 | 0.78 | 0.83 | 0.68 | ≤91 | 0.84 | 0.89 | 0.74 |
≤28 | 0.71 | 0.91 | 0.32 | ≤92 | 0.80 | 0.92 | 0.56 |
HC vs. AD | |||||||
MMSE | 3MS | ||||||
≤24 | 0.94 | 0.85 | 1.0 | ≤88 | 0.94 | 0.90 | 0.97 |
≤25 | 0.94 | 0.85 | 1.0 | ≤89 | 0.92 | 0.90 | 0.94 |
≤26 | 0.85 | 0.90 | 0.82 | ≤90 | 0.92 | 0.90 | 0.94 |
≤27 | 0.76 | 0.90 | 0.68 | ≤91 | 0.82 | 0.95 | 0.74 |
≤28 | 0.55 | 0.95 | 0.32 | ≤92 | 0.70 | 0.95 | 0.56 |
HC vs. VaD | |||||||
MMSE | 3MS | ||||||
≤24 | 0.67 | 0.41 | 1.0 | ≤88 | 0.81 | 0.68 | 0.97 |
≤25 | 0.77 | 0.59 | 1.0 | ≤89 | 0.84 | 0.77 | 0.94 |
≤26 | 0.76 | 0.71 | 0.82 | ≤90 | 0.88 | 0.84 | 0.94 |
≤27 | 0.74 | 0.78 | 0.68 | ≤91 | 0.81 | 0.86 | 0.74 |
≤28 | 0.64 | 0.89 | 0.32 | ≤92 | 0.75 | 0.89 | 0.56 |
Overall Classification Accuracy (HC vs. CI)
The 3MS showed outstanding discriminability (AUC=0.93), whereas the MMSE demonstrated excellent discrimination (AUC=0.86) of HC and CI. Moreover, the 3MS showed a better balance of hit rate, sensitivity, and specificity than the MMSE across a range of cutoffs.
Classification Accuracy×Etiology (HC vs. AD; HC vs. VaD)
Overall, the 3MS and MMSE showed comparable outstanding discrimination (AUCs> 0.90) of HC and AD. Classification accuracies for both measures demonstrated strong overall accuracy (hit rate> 0.90), sensitivity, and specificity (i.e., exceeding 0.95) at multiple cutoffs.
In contrast, the 3MS outperformed the MMSE when distinguishing HC from VaD. Specifically, the outstanding discriminability of the 3MS (AUC=0.92) exceeded the excellent discriminability of the MMSE (AUC=0.83). Additionally, the 3MS demonstrated a better balance of hit rate, sensitivity, and specificity than the MMSE across a range of cutoffs.
Discussion
Findings support the central hypothesis that the 3MS would demonstrate superior detection of cognitive impairment compared to the MMSE. Moreover, the extent of this enhanced detection was influenced by the underlying etiology of cognitive impairment. Consistent with our hypothesis, the enhanced classification accuracy of the 3MS was most pronounced when discriminating healthy adults from adults with cognitive impairment due to VaD. In contrast, the 3MS and MMSE showed relatively comparable discrimination of healthy adults and those with cognitive impairment due to AD. Taken together, findings support the incremental utility of the 3MS over the traditional MMSE as a screening measure for the two most common dementia etiologies.
The outstanding discriminability of the 3MS exceeded the MMSE’s excellent classification accuracy when distinguishing healthy adults from those with mild or major neurocognitive disorder due to AD or VaD. This finding is consistent with the large body of research showing that the 3MS is more sensitive than the MMSE in detecting dementias (i.e., major neurocognitive disorders) in samples of mixed etiologies (13, 18, 20, 24). Similarly, the 3MS has shown better detection of mild cognitive impairment (i.e., mild neurocognitive disorder) stemming from a variety of etiologies (15). Although modest, the increased discriminability of the 3MS in this study is meaningful when one considers the medical and societal costs associated with delayed and missed diagnosis of dementia.
The enhanced psychometric properties of the 3MS in detecting cognitive impairment are likely caused by a combination of factors. First, the addition of items and expanded scoring allows for greater variability in performance, increasing reliability potential. Indeed, the 3MS has shown increased reliability over the MMSE across a range of settings and clinical populations (13, 20–23). Beyond this pure psychometric advantage, research suggests the content of the additional items have afforded the measures increased discriminant validity. Specifically, the 3MS has shown stronger relationships with key cognitive abilities (e.g., memory, language) among individuals with mild cognitive impairment (15). Last, the incremental utility of the 3MS appears dependent on etiology of cognitive impairment.
Findings add to the literature by establishing that the incremental clinical utility of the 3MS varies across the two most common dementia etiologies. Consistent with prior research, both measures demonstrated comparable outstanding detection of individuals with AD (23). Moreover, research suggests that the verbal fluency item is the sole addition to the 3MS that contributes to its marginal advantage in detecting mild forms of AD (19). With AD progression comes significant global cognitive impairment that can be equally detectable by both measures. In contrast, VaD is more commonly associated with comparatively subtler cognitive impairments. In the context of cerebrovascular disease, the 3MS showed outstanding detection that exceeded the excellent discriminability of the MMSE. This finding is consistent with research showing that the MMSE has reduced detection of mild cognitive decline (14, 20), other subcortical dementia processes (e.g., Parkinson’s disease) (19), and cerebrovascular insult (e.g., right hemisphere stroke) (22). Moreover, the items added to the 3MS tap constructs that are particularly sensitive to cerebrovascular decline (e.g., processing speed, executive functioning) (29, 30). Specifically, verbal fluency requires processing speed and aspects of executive functioning, whereas verbal reasoning taps unique aspects of executive functioning. Finally, the addition of category and multiple-choice cues to memory items may permit the differentiation of retrieval versus consolidation deficits that are more characteristic of VaD versus AD memory profiles (31).
Several limitations should be considered when interpreting these results. First, this was a study of primarily male veterans from a single Veterans’ Affairs Hospital in the Southeastern United States. As such, the potential impact of other demographic factors (e.g., nonveteran status, sex) is unknown. Next, differential diagnoses were made clinically in the context of a full neuropsychological evaluation; however, comorbid neurodegenerative processes appear to be the rule rather than the exception (32–34). A portion of our AD group likely had neuropathic features of VaD processes and vice versa. Additionally, definite diagnosis requires neuropathologic confirmation of disease. Last, our study used a modest sample size. Despite this limitation, it is important to note that ROC results were strikingly similar to those obtained from studies employing significantly larger samples (23, 28).
The current study included mixed samples of individuals with mild and major neurocognitive disorders (i.e., mild cognitive impairment and dementia). Although this decision reduced homogeneity within groups, it maximized ecological validity. Specifically, practitioners treating elderly adults with cognitive complaints are often unable to distinguish mild cognitive impairment from mild dementia on the basis of interview alone. As such, research investigating cognitive screen cutoffs that combine these groups mirrors the clinical environment that practitioners face in their daily work.
Conclusions
Overall, results support the adoption of the 3MS over the MMSE. The added cost associated with the 3MS (i.e., approximately 5–10 minutes of additional administration time) pales in comparison to its incremental utility. The 3MS is particularly better than the MMSE at detecting VaD, the second most common dementing etiology. These classification enhancements have considerable medical, societal, and financial implications. Early and accurate detection of AD and VaD allows for targeted intervention to promote health behaviors, attempt to slow disease progression, and foster patient and caregiver preparedness and psychological coping. Moreover, early detection is critical to research that furthers understanding of these conditions and the future development of interventions targeting disease progression. Replication with larger, ethnically diverse veteran and civilian populations is strongly recommended. Additionally, future studies could examine item-level data to explore relationships to various cognitive abilities and relative contribution to enhance detection of VaD.
1. : The state of US health, 1990–2016: burden of diseases, injuries, and risk factors among US states. JAMA 2018; 319:1444–1472Crossref, Medline, Google Scholar
2. : Monetary costs of dementia in the United States. N Engl J Med 2013; 368:1326–1334Crossref, Medline, Google Scholar
3. Global action plan on the public health response to dementia 2017–2025. Geneva, World Health Organization, 2017. www.who.int/publications/i/item/9789241513487. Accessed Aug 4, 2020Google Scholar
4. : Alzheimer’s disease facts and figures. Alzheimers Dement 2019; 15:321–387. www.alz.org. Accessed Aug 4, 2020Crossref, Google Scholar
5. : Alzheimer’s disease facts and figures. Alzheimers Dement 2010; 6:158–194Crossref, Medline, Google Scholar
6. : Pathological correlates of late-onset dementia in a multicentre, community-based population in England and Wales. Lancet 2001; 357:169–175Crossref, Medline, Google Scholar
7. : The effect of a disease management intervention on quality and outcomes of dementia care: a randomized, controlled trial. Ann Intern Med 2006; 145:713–726Crossref, Medline, Google Scholar
8. : The diagnosis of dementia 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 2011; 7:263–269Crossref, Medline, Google Scholar
9. : The underdiagnosis of the vascular contribution to dementia. J Neurol Sci 2005; 229–230:3–6Crossref, Medline, Google Scholar
10. : Vascular dementia is underdiagnosed. Arch Neurol 1988; 45:797–798Crossref, Medline, Google Scholar
11. : “Mini-mental state”: a practical method for grading the cognitive state of patients for the clinician. J Psychiatr Res 1975; 12:189–198Crossref, Medline, Google Scholar
12. : Screening for cognitive impairment in older adults: a systematic review for the US Preventive Services Task Force. Ann Intern Med 2013; 159:601–612Medline, Google Scholar
13. : Mental status testing in the elderly nursing home population. J Geriatr Psychiatry Neurol 1995; 8:177–183Crossref, Medline, Google Scholar
14. : The Mini‐Mental State Examination: a comprehensive review. J Am Geriatr Soc 1992; 40:922–935Crossref, Medline, Google Scholar
15. : Comparing the Mini-Mental State Examination and the Modified Mini-Mental State Examination in the detection of mild cognitive impairment in older adults. Int Psychogeriatr 2019; 31:693–701Crossref, Medline, Google Scholar
16. : The Mini-Mental State Examination: pitfalls and limitations. Pract Neurol 2017; 17:79–80Crossref, Medline, Google Scholar
17. : Cognition, copyright, and the classroom. Am J Psychiatry 2005; 162:627–628Crossref, Medline, Google Scholar
18. : The Modified Mini-Mental State (3MS) Examination. J Clin Psychiatry 1987; 41:114–121Google Scholar
19. : Montreal Cognitive Assessment performance in patients with Parkinson’s disease with “normal” global cognition according to Mini-Mental State Examination score. J Am Geriatr Soc 2009; 57:304–308Crossref, Medline, Google Scholar
20. : Community screening for dementia: the Mini Mental State Exam (MMSE) and Modified Mini-Mental State Exam (3MS) compared. J Clin Epidemiol 1997; 50:377–383Crossref, Medline, Google Scholar
21. : Age- and education-specific reference values for the Mini-Mental and Modified Mini-Mental State Examinations derived from a non-demented elderly population. Int J Geriat Psychiatry 1997; 12:1008–1018Crossref, Medline, Google Scholar
22. : Folstein vs modified Mini-Mental State Examination in geriatric stroke: stability, validity, and screening utility. Arch Neurol 1995; 52:477–484Crossref, Medline, Google Scholar
23. : Mini-Mental State Examination (MMSE) and the Modified MMSE (3MS): a psychometric comparison and normative data. Psychol Assess 1996; 8:48–59Crossref, Google Scholar
24. : Comparisons between the Mini-Mental State Exam (MMSE) and its modified version: the 3MS test; in Psychogeriatrics: Biomedical and Social Advances. Edited by Hasegawa K, Homma A. Tokyo, Excerpta Medica, 1990, pp. 189–192Google Scholar
25. Diagnostic and Statistical Manual of Mental Disorders (DSM-5), 5th ed. Arlington, VA, American Psychiatric Association, 2013Google Scholar
26. : A meta-analysis of the accuracy of the mini-mental state examination in the detection of dementia and mild cognitive impairment. J Psychiatr Res 2009; 43:411–431Crossref, Medline, Google Scholar
27. : Applied Logistic Regression. New York, Wiley, 2000Crossref, Google Scholar
28. : Correcting the 3MS for bias does not improve accuracy when screening for cognitive impairment or dementia. J Clin Exp Neuropsychol 2004; 26:970–980Crossref, Medline, Google Scholar
29. : Vascular cognitive impairment. Lancet Neurol 2003; 2:89–98Crossref, Medline, Google Scholar
30. : Differentiation of vascular dementia from AD on neuropsychological tests. Neurology 1999; 53:670–678Crossref, Medline, Google Scholar
31. : Neuropsychological deficit in early subcortical vascular dementia: comparison to Alzheimer’s disease. Dement Geriatr Cogn Disord 2002; 14:26–32Crossref, Medline, Google Scholar
32. : Mixed neuropathologies and estimated rates of clinical progression in a large autopsy sample. Alzheimer’s Demen 2017; 13:654–662Crossref, Medline, Google Scholar
33. : The enigma of mixed dementia. Alzheimer’s Demen 2007; 3:40–53Crossref, Medline, Google Scholar
34. : Mixed brain pathologies account for most dementia cases in community-dwelling older persons. Neurology 2007; 69:2197–2204Crossref, Medline, Google Scholar