Identification of high-risk dementia cohorts in a community sample of Japanese elderly


*Shunichiro Shinagawa, MD, PhD, Department of Psychiatry, The Jikei University School of Medicine, 3-25-8 Nishi-shinbashi, Minato-ku, Tokyo 105-8461, Japan. Email:


Aim:  The aim of this study was to develop a simple diagnostic procedure for subjects at high risk of developing dementia using the Clinical Dementia Rating (CDR), which is applicable to community-based activities.

Methods:  This study divided 252 community-dwelling elderly with a CDR score of 0.5 into two groups based on the presence or absence of cognitive decline within the previous one year of the baseline, as assessed by a semi-structured interview. One hundred subjects were in the ‘previously progressive group’ (PP group) and 152 subjects were in the ‘previously stable group’ (PS group). After 6 years of observation, a total of 111 subjects were assessed in the follow-up investigation.

Results:  Among the 39 subjects from the PP group (82.9 ± 6.8 years old, 11 male, 28 female), 34 developed dementia (87%). Among the 72 subjects from the PS group (84.4 ± 6.0 years old, 22 male, 50 female), 44 developed dementia (61%). The relative risk of developing dementia for the PP group versus the PS group was 1.43. The rate of conversion to dementia was 12.9% per 100 person-years in the PP group, and 9.8% in the PS group. In the PP group, the Mini-Mental State Examination score was significantly lower and the CDR score was significantly higher than in the PS group.

Conclusion:  Although there have been many attempts to identify subjects with high risk of dementia, this preliminary study suggests that information about temporal changes in cognitive function is useful when performing community-based surveys.

THE CONCEPT OF mild cognitive impairment (MCI) has been proposed to describe the transitional state between normal cognition and dementia.1,2 The concept of MCI has become popular because identification of high-risk groups for dementia is important in the field of early prevention and intervention in public health. However, because the concept of MCI includes many heterogeneous states and may be unstable over time, not all subjects develop dementia.3 Standardized memory examinations, such as the Wechsler Memory Scale Revised can be used to identify subjects satisfying the strict definition of MCI.4 However, it may be very difficult to carry out such time-consuming examinations in community-based epidemiological surveys from the viewpoint of both costs and the burden on the participants. Therefore, it is debatable whether the concept of MCI should be used strictly in community-based interventions.5

The Clinical Dementia Rating (CDR) scale is an objective cognitive assessment tool completed by informants that can be easily used in the community.6,7 Some researchers have equated a CDR of 0.5 (questionable dementia) to MCI,8,9 suggesting that subjects with a CDR of 0.5 present specific pathological traits of Alzheimer's disease (AD) and neuropsychological impairments in several cognitive domains.8 However, subjects with a CDR 0.5 may also include an array of heterogeneous conditions, some of them showing progressive change and others not. It is important to identify groups at high risk of developing dementia using a brief and efficient screening method in a community setting.

Therefore, a 6-year observation of subjects with CDR 0.5 was conducted to examine the possibility that the development of dementia could be predictable. The study focused on the progression of cognitive decline within the previous year of the baseline using a semi-structured interview for subjects and informants in addition to the CDR scale. The aim was to devise a simple diagnostic procedure for subjects at high risk of dementia using the CDR, which could be employed for community-based interventions and public health activities.


Written informed consent was obtained after a complete description of the study had been given to all subjects or their informants. This study was approved by the ethics committee of Jikei University.

Baseline assessment

An epidemiological survey of dementia was conducted in Itoigawa city, a rural area of Japan, in 1998. There were a total of 33 120 inhabitants and 7847 of them were over 65 years old. Two-phased semi-structured screening tests were conducted, including items on education, occupation, medical history, risk factors such as hypertension, diabetes, and hyperlipidemia, activity of daily living, psychiatric and behavioral symptoms, CDR, and the Mini-Mental State Examination (MMSE).10 Subjects who met the criteria (scored ≦19 on MMSE, scored ≧20 on MMSE but had some psychiatric symptoms, had been diagnosed to have dementia before the survey, etc.) in the first phase were recruited into the second phase. The details of the design and methods of sampling in the baseline survey have been described elsewhere.11 The second phase was conducted with a total of 1114 subjects, using a door-to-door survey by a clinical team including a psychiatrist and a public health nurse. A total of 271 of these subjects were CDR ≧ 1, 252 of them were CDR = 0.5, and 306 were CDR = 0. The other 285 subjects did not participate in the second phase, due to refusal, death or as a result of moving to other communities.

For the 252 subjects with CDR 0.5, a semi-structured interview was conducted inquiring about changes in cognitive function within the previous one year of the baseline. Information was gathered from families familiar with the patients' daily lives. The interview was designed in accordance with the CDR and Clinical Global Impression of Changes (CGIC).12 On conducting CDR, we made additional questions whether subjects showed progressive decline, improvement, or no change within the previous one-year period (like global assessment in CGIC), in the domains of memory, orientation, and judgment/problem-solving abilities. In the memory domain, informants were asked whether the subjects showed a progressive episodic memory disturbance within the previous one-year period. For example, informants were asked to answer whether the subjects had begun to speak the same phrases repeatedly or had begun to look for personal belongings or had begun to forget important events in comparison to the one year previously. If the informants answered that subjects showed progressive memory decline within the past year, then they were categorized as being in the ‘previously progressive group’ (PP group). In the orientation domain, informants were asked to answer whether subjects had begun to forget dates or places in comparison to the previous year. If subjects had shown progressive cognitive decline in any other domain within the previous one year, they were categorized as being in the PP group. If subjects showed no cognitive changes even though they were diagnosed as CDR 0.5, then they were categorized as being in the ‘previously stable group’ (PS group). Well-trained psychiatrists made the final decision as to whether a subject should be included in the PP group or the PS group. The raters discussed each patient and checked each other in order to assure the accuracy of their decision. The inter-rater reliability was sufficient enough.

Of these 252 subjects with CDR 0.5, 100 were categorized into the PP group (those who showed progressive cognitive decline but did not meet the dementia criteria), while the other 152 were categorized into the PS group.

Follow-up assessment

After the 6-year observation period, follow-up assessment was conducted in 2004. There was no systematic intervention for subjects with CDR 0.5 during the 6-year observation period. Subjects with CDR 0.5 were assessed using the same process used for a previous survey in 1998. The psychiatrists who determined the CDR score in 2004 were blind to which group they were examining. The diagnosis of dementia was established according to the DSM-IV criteria and Consortium to Establish a Registry for Alzheimer's Disease (CERAD) criteria.13,14 The demented subjects were classified into subgroups according to the cause of dementia, on the basis of the DSM-IV criteria.13 The general design and flow chart of the survey is summarized in Figure 1.

Figure 1.

The general design and flow chart of the survey. CDR, Clinical Dementia Rating; PP group, previously progressive group; PS group, previously stable group.

Statistical analysis

Data analyses were carried out using the spss-PC software package. The significance of the differences between the two groups were assessed by t-test for age and MMSE; and by χ2-test with Fisher's exact test for sex, CDR, proportion of researched subject and the proportion of the cause of dementia. A significance level of 0.01 (two-tailed) was set for all analyses.

The conversion rate was calculated using the person-year method. A logistic regression analysis was also performed.


The demographic variables of the two subject groups are summarized in Table 1. There was no significant difference between the two groups in the MMSE score at the baseline. Out of the 252 CDR 0.5 subjects, 79 had already died after the 6-year observation period, 42 from the PP group, and 37 from the PS group. Sixty-two subjects had moved to other communities (due mainly to institutionalization) or refused to participate in the follow-up investigation, 19 from the PP group, and 43 from the PS group. There was no significant difference in the ratio of researched/dead/removed or refusal subjects between the two groups.

Table 1.  Comparison between the previously progressive group and the previously stable group for CDR 0.5 subjects
 The previously progressive group in CDR 0.5The previously stable group in CDR 0.5P-value
  1. AD, Alzheimer's disease; CDR, Clinical Dementia Rating; MMSE, Mini-Mental State Examination; VaD, vascular dementia.

Total number in 1998100152 
Mean MMSE score (1998)22.5 ± 2.723.2 ± 3.6n.s.
Researched/dead/removed or refusal in 200439/42/1972/37/43n.s.
Percentage of demented subjects87.1% (34/39)61.1% (44/72)P = 0.004
Annual conversion rate for dementia (%/100 person-years)12.9%9.8% 
Mean age (in 2004)82.9 ± 6.884.4 ± 6.0n.s.
Sex (male:female) (in 2004)11:2822:50n.s.
Mean MMSE score (in 2004)15.9 ± 6.318.8 ± 7.1P = 0.004
Diagnosis of demented subjects (AD/VaD/others)24/4/629/9/6n.s.
Number of subjects of each CDR score (0:0.5:1:2:3)0:5:11:12:113:25:17:15:12P = 0.006

Finally, 111 subjects participated in the follow-up investigation: 39 from the PP group and 72 from the PS group. Among the PS group, there was no significant difference in the baseline MMSE score between those who completed the follow-up investigation (n = 72) and those who did not (n = 80).

Among the 39 subjects from the PP group, 34 developed dementia (87%), of whom 24 had AD, four had vascular dementia (VaD), and the others had frontotemporal dementia, Parkinson's disease-related dementia [this includes dementia with Lewy bodies (DLB), because DSM-IV does not have the diagnostic criteria for DLB] and dementia not otherwise specified. In contrast, among the 72 subjects from the PS group, 44 developed dementia (61%), of whom 29 had AD, nine had VaD, and the others had Parkinson's-disease-related dementia (including DLB), alcohol-induced dementia and dementia not otherwise specified. The relative risk of developing dementia for the PP group versus the PS group was 1.43 [95% confidence interval (CI) = 1.14–1.78]. The rate of conversion to dementia was 12.9% per 100 person-years in the PP group, and 9.8% in the PS group.

After 6 years of observation, there were no significant differences in the mean age and sex ratio between the two groups. In the PP group, the MMSE score was significantly lower than in the PS group. In the PP group, the CDR score was significantly higher than in the PS group.

A logistic regression analysis was conducted in 111 subjects, defining conversion to dementia as a dependent variable, and age, sex, MMSE score (in 1998) and ‘previously stable or progressive’ as independent variables. Only ‘previously stable or progressive’ significantly affected conversion to dementia (odds ratio 1.67; 95%CI = 1.32–1.98; P = 0.014).


This study of community-dwelling elderly investigated the use of a simple method for identifying individuals at high risk of developing dementia, with a 6-year follow up. Several previous studies have suggested that most subjects who meet the MCI or other pre-dementia criteria will progress to AD or other forms of dementia. An overall MCI conversion rate of 10–15% has been reported. However, conversion rates have differed widely among studies.15–17 The sources of this discrepancy among studies include variations in the population characteristics (hospital-based vs community-based), length of follow up, and the definition of pre-dementia applied. Previous studies have noted that MCI patients who are older, have a low MMSE score, suffer from depression, show evidence of ischemic change, and the presence of the APOE4 allele have a higher risk of developing dementia.18,19 Other brain imaging studies and biomarker studies have also detected a high-risk group,20,21 but it is very difficult to apply these methods in epidemiological surveys.

As self-selected clinic attendees have a high conversion rate,16 the current results should not be compared with clinical-based results. To our knowledge, there are several community-based prospective cohort studies that have followed elderly community-dwelling MCI people for several years.5,22–24 In one of these studies,22 comprehensive test batteries including MMSE, visual memory, verbal fluency, visuospatial attention, simple logical reasoning and attention were adopted. This approach is not only costly and time-consuming but also places a large burden on participants, so it may be difficult to conduct this approach in epidemiological surveys. There is an increasing need for a brief and efficient screening method suitable for detecting individuals at high risk of developing dementia in a community setting. Although there have been many attempts to assess cognitive decline in MCI and other pre-dementia states, there are only a few brief assessments that use responses from informants to identify dementia in the community.25 The conversion from MCI to AD was 8.3% per 100 person-years22 and 8.5% per 100 person-years5 in these community-based prospective cohort studies, so the results of the current study are consistent with these results using a much simpler method, although a different definition of a pre-dementia state was used.

The current study conducted a semi-structured interview of subjects and informants that asked about changes in cognitive function within the previous year of the baseline in addition to the CDR scale, and thus this survey differed from other previous investigations. This interview was brief and imposed only a small burden on participants. Subjects who had shown cognitive changes within the previous year of the baseline developed dementia at a higher rate than those without cognitive changes. Furthermore, the PP group scored higher in CDR and lower in MMSE after 6 years of observation, which indicates that they included subjects showing faster progression of dementia. In subjects with a CDR of 0.5, information on previous progressive cognitive decline could be a meaningful predictor of the progression of dementia. There have so far been no previous attempts that focus on the progression of cognitive decline within the previous year of the baseline. It is therefore suggested that not only a cross-sectional cognitive assessment but also information about the progression of cognitive decline over time are important for the prediction of dementia.

There are a few methodological issues that should be considered when interpreting the current results. First, as this study was based on the retrospective recall of subjects and informants, thus there is a possibility that the informant's memories may have been inaccurate. Furthermore, our definition of the PP group and the PS group may even be considered inaccurate by others. However, using a clinical interview to obtain a medical history is the usual way of diagnosing dementia, so any possible bias introduced by the current methods is likely to be similar in routine clinical practice. Furthermore, the same methods have also been used to estimate the efficacy of antidementia drugs. Second, a relatively large proportion of subjects could not be followed up because of their death, change of residence or refusal to participate. In addition, among those lost subjects, an appreciable number may have developed dementia, and thus the rate of conversion to dementia may have been higher if these subjects had been taken into account.

In a conclusion, although many attempts have been made to identify individuals at high risk of dementia, the collection of information about changes in cognitive function over time might therefore have clinical utility.


The authors thank Professor K. Nakayama for his suggestions and encouragement for this study. This work is supported in part by a grant from St. Luke's Life Science Institute (to N.T.).