Prevalence of cognitive impairment in individuals aged over 65 in an urban area: DERIVA study

Study design

An observational, descriptive, cross-sectional population study.

Setting

The reference population was that of the city of Salamanca, with 172,375 inhabitants, of which 19.74% (34,020) were aged over 65. It includes 10 healthcare areas, each with a population of between 9,000 and 26,000 inhabitants.

Participants

We selected all those aged over 65 on 1 January 2009 and living in the city of Salamanca (urban setting). A door-to-door population-based survey was carried out during the months of May to November 2009. Two weeks before the interviews, letters were sent to the selected individuals, explaining the purpose of the study and requesting their cooperation. Confidentiality of data was guaranteed. Ten days after sending the letters, a telephone call was made to arrange a home interview.

The following exclusion criteria were applied

Ethical aspects

The protocol was approved by the Research Ethics Committee of Salamanca University Hospital. Participants signed the consent document after receiving the first explanatory letter providing information on the study.

Training of the evaluators

The principal researcher coordinated the entire process. The evaluators were four psychologists trained by one of the researchers (SMS) to carry out the interview with the programmed questionnaires. A manual was drafted, describing the appointment procedure and application of the interviews, and was reviewed with the interviewers over two sessions. We also used two recorded home interviews in the training sessions. During the study, communication was permanently maintained for resolving any doubts or dealing with incidents in relation to the questionnaires.

Data sources

The sample was taken from the Castilla y León Regional Health Service lists, which cover 99.5% of the population. The lists included both community dwellers and institutionalized elders.

Study size

Accepting an alpha risk of 0.05 and a beta risk of 0.20, estimating a CI prevalence of about 16% [12], with an error of 4%, and considering the current population aged over 65, a total of 320 patients was required. Assuming a loss rate of up to 50% due to non-responses, as observed in similar studies, the calculated sample size was 480 individuals. In the secondary analysis of cases and controls, with a 327 participants sample, the statistical power was of 79.4% to detect an Odds Ratio of 2.5 with a confidence level of 95% (Epidat 4.0). We carried out a stratified random sampling by health districts. In order to reach the required sample size, we made a replacement for lost participants. The sample size of each health district was proportional to its population over 65 years. In a first stage, 260 interviews were carried out, accounting for 80% of the required sample. Two months later, in a second stage, we replaced the losses within each health district and 67 more people were interviewed. The most common cause of losses was patient refusal to participate (83.0%). This was particularly the case among the younger individuals (mean age: 75.94 ± 7.01 years) (p < 0.001). In turn, 12 patients were excluded because they had moved out of the study district (7.8%), while 14 had died (9.2%) (Figure 1). There were no significant differences between sexes regarding the cause of losses, or between losses in the first and second recruitment stages. Non-responders represented 34.83% of the males and 30.13% of the females. The distribution by age groups is shown in Table 1. Mean age of the males was 76.61 years (SD: ± 7.65), versus 77.52 years (SD: ± 7.92) in the females. There were no differences in distribution between the different sex and age categories. A total of 327 participants were interviewed, representing 68.1% of those selected (Figure 1). Of these, 116 were men (35.5%) and 211 were women (64.5%), with a mean age of 76.35 years (SD:± 7.33) similar in the two sexes.

Measurements

A health questionnaire was administered (OARS Multidimensional Functional Assessment Questionnaire) [13] in the participant's home, for obtaining data on sociodemographics (age, sex, marital status), years of education, basic and instrumental activities of daily living, personal situation (people living with the patient, number of living offspring), morbidity (Charlson comorbidity index [14]), drug use and neuropsychological aspects. Marital status was classified according to whether the person was living with or without a partner (widowed, single, separated), while educational level was classified as follows: illiterate (failure to complete basic education), primary-secondary education (4-9 years of education) and higher education (over 9 years). Regular professional occupation before retirement was defined as domestic chores, full-time job, others (part-time job, long-term unemployment, etc.). The Katz Index of Independence in Activities of Daily Living (Katz ADL) was applied to assess functional status as a measure of patient ability to perform activities of daily living (ADLs) independently [15]. This test provides information on dependence or independence, not only in terms of the number of areas, but also identifying the specific areas. The information obtained is of a qualitative and descriptive nature, and does not provide a total score on the person's functional state.

At the beginning of the interview, neuropsychological assessment of patient cognitive status was carried out using a brief neuropsychological test battery including the following: Mini-Mental State Examination (MMSE) [16] in its validated Spanish version [17] to evaluate general cognitive state, with the possibility of assessing cognitive functions separately; the 7 Minute Screen [18] in its validated Spanish version [19] also to evaluate general cognitive state, as well as temporal orientation, memory, constructive praxias and language, separately; the Benton temporal orientation test [20] to evaluate temporal orientation; the Enhanced cued recall test [21] to evaluate episodic memory; the Clock drawing test [22] to evaluate constructive praxias; and the Categorical fluency task [23] to evaluate language. The cut-off points for cognitive impairment established for each test are as follows: MMSE<24; 7MS≤ percentile 20; Benton temporal orientation test ≤ 102; Enhanced cued recall test ≤ 12; Clock drawing test <3; Verbal fluency ≤ 10.

During the structured interview, participants were asked to present whatever relevant clinical records they might have, together with details of their current medications. In the 16 patients who were unable to complete the neuropsychological evaluation because of a deteriorated clinical condition - heart failure (2 cases), deafness (4 cases), severely impaired vision (2 cases) or severe mental impairment (4 cases) - the clinical and objective data were collected from the presented documentation, or by interviewing the caregiver or informant who knew the individual. We considered as reliable informants, in order of preference, a family member living in the same house as the individual (spouse, son/daughter, sibling); a person responsible for the care of the individual; someone living in the same house but not a family member; or a relative of the person not living in the same house. At the end of the interview the interviewers drafted a report on the quality of the information collected and on the social and health conditions of the person interviewed. Seventeen participants lived in homes for the elderly (5.2%). All the information was evaluated by the researchers (ERS, SMS, RGG and MVP) with a view to establishing the final diagnosis. The diagnoses were divided into three general categories: normal cognitive function, cognitive impairment - no dementia (CIND), and dementia.

Classified as cognitive impairment - no dementia (CIND)

CIND was defined as: 1) mild cognitive or functional impairment reported by the participant or informant that did not meet criteria for dementia; or 2) performance on neuropsychological or functional measures that was both below expectations and ≥ 0.5 standard deviations below published norms on any test [24].

Classified as dementia

A diagnosis of dementia was made based on the Diagnostic and Statistical Manual of Mental Disorders (IV Edition) criteria: the participant must present the development of multiple cognitive deficits including memory impairment and impairment in at least one other cognitive domain representing a decline from the previous level of functioning and of sufficient severity to cause impairment in function [4, 5]. At a functional level, the person must present dependence in at least two functional areas, leading to interference in basic activities of daily living. Alterations at both the cognitive and functional levels were indicated by low performance and scores below the cut-off points in the neuropsychological and functional tests. As regards functional state, the person must present a minimum level of alteration in at least two functional areas. All of this must be accompanied by concern on the part of the participants about a change at a cognitive level compared to his or her previous state [4, 5, 25].

Statistical analyses

The raw prevalence of CIND and dementia were calculated taking into account the total cases of CIND and dementia with respect to the total study sample. We estimated both global prevalence and specific prevalence per age group, sex, educational level, morbidity and functionality. Calculation was made of the 95% confidence intervals (95%CI) for the global and specific prevalence, together with prevalence adjusted for age and sex, using the European standard population [26] as a basis for adjustment (weighting: 36.4, 27.3, 18.2, 9.1 and 9.1 for age intervals of 65-69, 70-74, 75-79, 80-84 and = 85 years, respectively).

The continuous variables were expressed as the mean ± standard deviation (SD), while frequency distributions were used for the qualitative variables. Logistic regression analysis was used to explore the sociodemographic, functional and clinical factors independently associated with the presence of CIND/dementia. The enter method was used in a first step to include the adjusting variable (patient age and sex), followed in a second step by the stepwise method in application to the rest of the independent variables: years of education, educational level, Katz index, living with partner, restlessness, anxiety or depression, sleeping problems or insomnia, diabetes and Charlson comorbidity index. The dependent variable was cognitive impairment (CI)(code 0: no CI; code 1: presence of CI), while in the case of the independent variables the reference groups were male sex, illiteracy, absence of diabetes, no insomnia, no depression, and living without a partner. Patient age, the Katz index, years of education and the Charlson comorbidity index were taken as continuous variables. The data were analyzed using the SPSS/PC+ version 18.0 statistical package (SPSS Inc., Chicago, IL, USA).

The importance of certain personal factors

Our data coincide with those of practically all studies regarding the increase in CI with advancing age (Figure 2). It has been estimated that between 65 and 85 years of age, the prevalence doubles every 5.2 years, in an exponential manner [47]. However, there is no general agreement regarding prevalence in terms of patient sex. The condition is rarely associated with the male sex [48], while in contrast many studies have associated dementia with the female sex [2, 29, 31, 49, 50], since women predominate in descriptive studies, and a correlation is found in the bivariate analyses, in accordance with our own observations. However, on considering other variables such as age, educational level or comorbidity, this relationship disappears. Our data would support the idea that dementia is not associated with women [1, 25, 27, 33, 47]. In coincidence with our own study, many authors [7, 33, 51] have described an inverse relationship between the prevalence of CI and a lower educational level. Thus, it has been suggested that these differences could be related to certain biological mechanisms that would be responsible for this association [8, 27, 29, 52, 53]. However, a recent epidemiological study based on 875 necropsies revealed no protective effect of the years of education received in early stages in life in relation to the accumulation of neurodegenerative or vascular pathologies of the brain [54]. Even so, it has been suggested that when the disease affects people with higher educational level, the manifestations are milder, being mitigated by a greater coping capacity - thereby postponing their consequences for a period of time. It remains to be clarified whether educational levels developed in later periods in life may or may not affect the development of dementia. In an attempt to offer information on this issue, we analyzed the type of regular professional activity of the DERIVA study patients before retirement, though no relationships were found. Perhaps a different classification of activity, analyzing those which may contribute most to intellectual development, could help identify a relationship. Regarding the family situation, we found that those participants living with their partner showed a 13.7% lower prevalence of CI than those living without a partner (p = 0.007). This is in agreement with the observations of Helmer [55], who found single individuals to have a greater risk of suffering dementia than married people. In our case there were no differences related to the fact of living alone or with one or more people, in terms of the prevalence of CI. As ours is a cross-sectional study, however, neither of these characteristics offers information for clarifying whether the current situation is a consequence of or a risk factor for the development of CI.

Many chronic illnesses can be found in elderly people. The Charlson score was greater among the individuals with CI (p = 0.004), though the multivariate analysis did not find it to behave as a risk factor for CI. The association between diabetes (p = 0.045) and dementia has already been described in classic studies [52], though it must be mentioned that while arterial hypertension is accepted as being more closely associated with stroke [31, 51], diabetes is the disease found to be associated with CI in the logistic regression analyses. The association between CI and diabetes, as well as the absence of an association with arterial hypertension and hypercholesterolemia, coincide with the findings in another Spanish region [56]. It appears that brain damage would be related to vascular mechanisms [33, 37], and since the prevalence of these cardiovascular diseases is not homogeneous in all regions, they should be considered in CI prevalence studies with a view to establishing comparisons. People with CI more often have sleeping problems (p = 0.027) and anxiety-depression (p < 0.001), though only the latter was seen to behave as a CI risk factor in the multivariate analysis. In coincidence with Johansson et al. [57], we found an association between psychological stress and CI, though in the study by the mentioned authors this association was detected in middle-aged women.

Among the limitations of our study, mention must be made of the lack of consensus on the precise criteria involved in making comparisons between different epidemiological studies. Another common and important limitation in studies of this kind is the possible loss of the more seriously deteriorated individuals, since it has been shown that those people who refuse to participate are more likely to have more seriously impaired cognitive function [58] - a situation which may introduce bias and alter the prevalence data obtained. A further limitation of the study is that since the evaluations were carried out by four different psychologists, inter-observer reliability may be affected. However, each evaluation was followed by an appraisal of the interview with a view to tests correction and the reduction of possible bias. Among the strengths of the study, we should mention the inclusion of a representative sample from the city of Salamanca, as well the proposal to incorporate functional evaluation and the clinical processes influencing the prevalence of CI [5]. The study cohort would permit a longitudinal evaluation of CIND, and the long-term results may contribute to identifying the characteristics of those individuals who develop dementia.