| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
| ||||||||||||||||||||||||||||||||
RESEARCH ARTICLE |
1 Department of Clinical Psychology, Vrije University, Amsterdam, The Netherlands.
2 Research Institute Psychology and Health
3 Department of Clinical Epidemiology and Biostatistics
4 Institute for Research in Extramural Medicine (EMGO)
5 Department of Psychiatry, Vrije University Medical Center, Amsterdam, The Netherlands.
Address correspondence to Bas Steunenberg, Clinical Psychology, Vrije University, Van der Boechorstraat 1, Amsterdam, 1081 BT, The Netherlands. E-mail: b.steunenberg{at}psy.vu.nl
 |
Abstract |
|---|
 TOP Abstract MethodsResultsDiscussionReferences |
|---|
DOES aging only alter us physically or does our personality change as well? During the past three decades, the question of whether personality changes in later life has been the subject of debate in personality research. Some researchers maintain that personality is fully developed by the time a person reaches the age of 30 and that it becomes increasingly stable with time (Costa & McCrae, 1988
, 1994; Costa et al., 2000; Glenn, 1980; Martin, Long, & Poon, 2002; Roberts & Del Vecchio, 2000). Other researchers advocate the plasticity of personality as a function of contextual variables and compensatory behavioral changes to biological aging (Alwin, 1994; Baltes, Staudinger, & Lindenberger, 1999; Caspi & Roberts, 1999; Heatherton & Nichols, 1994; Roberts, 1997). In the present study we aim to examine the cross-sectional and longitudinal stability or change of the personality trait of Neuroticism over a 6-year period of time in a large community sample of Dutch elderly individuals aged between 55 and 85 years at baseline. We hypothesize that we will find general stability, with a minority showing significant changes in neuroticism scores in late life.
In general there are two different approaches to the study of personality stability (Costa & McCrae, 1994): stability of mean levels to estimate aggregate level changes in personality with age, and stability approached as an individual-differences phenomenon, assuming that some people change whereas many remain stable (Cattell, 1950, 1966; Mroczek & Spiro, 2003). A large body of longitudinal research on mean-level changes emphasizes the invariance of neuroticism in late life (Costa & McCrae, 1980, 1990, 1994; Martin et al., 2002; Smith & Baltes, 1999; Watson & Clark, 1984). However, recent results suggest that changes can be observed in old age (Small, Hertzog, Hultsch, & Dixon, 2003).
Studies on individual differences in change are scarce. A recent study by Mroczek and Spiro (2003) found significant variability in neuroticism with age. The younger elderly persons and men who recently experienced a life event showed a sharp decline in neuroticism over the follow-up period.
Neuroticism is a major personality trait (Eysenck, 1990) and one of the most well-established dimensions in the "Big Five" (Costa & McCrae, 1994). Research has shown personality in late life, and neuroticism in particular, to be strongly related to mental and physical health, level of social support, self-rated health, and functional limitations (Duberstein et al., 2003; Hooker, Monahan, Bowman, Frazier, & Shifren, 1998; Siegler & Brummett, 2000; Smith & Gallo, 2001; Watson & Tellegen, 1985). So, age-graded deteriorations in physical health and related declines in daily functioning may affect the association between aging and neuroticism. We hypothesize that these variables affect the individual trajectories of neuroticism in an aging population.
For several reasons, this study is rather unique. To our knowledge, this is the first study to analyze this association in a large community sample of elderly persons between the ages of 55 and 85 years at inclusion. Until now, most research has been performed cross-sectionally or on specific populations. To analyze the developments of neuroticism in aging populations, a prospective design is required. As far as we know, this study is the first using a 6-year longitudinal design with two follow-up measurements in a community population of elderly individuals between 55 and 85 years of age. In addition to this, the multidisciplinary design of this study allowed us to address the influences of physical health-related variables on this association. Finally, we are, to our knowledge, the first researchers looking at individual differences in change on the longitudinal trajectory of neuroticism in a large community sample of elderly people.
|
METHODS |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
). Full details on sampling and response are described elsewhere (Beekman et al, 1995; Penninx et al., 1997). In short, we drew a random sample of older (55–85 years) men and women from the population registers in eleven municipalities in three geographical regions of The Netherlands. We constructed the sample so as to reflect the national distribution of urbanization and population density. In order to be able to study age and gender differences, we stratified the sample for age and gender. We excluded persons aged 85 and older from the baseline sample, because the attrition rate of this age group was expected to be high over the 10-year study period. We weighted the baseline sample according to expected mortality in each age group, resulting in an overrepresentation of older age strata and a roughly even distribution of men and women. The respondents were visited at home by trained interviewers who worked with laptop computers. Interview and tests took approximately 1.5 hr. In the baseline LASA interview, which took place in 1992–1993, there were 3,107 participants. The first follow-up measurement was performed in 1995–1996 (n = 2,545, 82%), and the second in 1998–1999 (n = 2,076, 67%).
In the LASA interview, we measured the personality factor Neuroticism by using a self-report questionnaire. The baseline LASA cycle consisted of a face-to-face main interview, after which the interviewer left a questionnaire to be returned by mail. Not all respondents complied, and to be eligible for inclusion in the statistical analyses of the present study, the respondents should have answered more than 80% of all the items on the questionnaire. This cutoff point resulted in a sample of 2,117 participants at baseline (68.3%). The older old (
2 = 158.13; p <.001) and less educated (2 = 41.55; p <.001) were more often lost through item nonresponse. Nonresponders had a higher number of chronic diseases (2 = 12.7; p <.001), were more often cognitively impaired (2 = 217.17; p <.001), and showed more difficulties while performing activities of daily living (2 = 74.26; p <.001).
For the multilevel analyses, we used 2,117 cases at baseline, and 1,983 and 1,647 cases at first and second follow-up measurement. Data were available for 1,229 respondents on all three measurements. These 1,229 respondents had a higher mean age (2 = 328.86; p <.001), and they were more often male (2 = 4.827, p <.05) and higher educated (2 = 13.62, p <.001). They reported fewer chronic diseases (2 = 49.23, p <.001) and functional limitations (2 = 118.306, p <.001) at baseline in comparison with those respondents lost during follow-up. Finally, respondents lost during follow-up were more often cognitively impaired (2 = 83.25, p <.001), but we found no significant difference in mean neuroticism level at baseline (2 = 1.49, p =.22).
Measurements
Neuroticism
We operationalized neuroticism by using a subset of 25 items out of a list of 36 neuroticism items from the Dutch Personality Questionnaire (DPQ; Luteijn, Starren, & van Dijk, 1975, 2000). The pilot studies of the LASA program showed that the original 36-item scale measuring neuroticism can be abbreviated, because the original DPQ scale contained items that were less valid for an aging population and contained too many items for administration in older populations (de Beurs, Beekman, Deeg, van Dyck, & van Tilburg, 2000; Smits, Deeg, & Bosscher, 1995; Steunenberg, Beekman, Deeg, & Kerkhof, 2003). Persons scoring high on neuroticism items experience a broad range of negative moods, including emotions such as fear, sadness or depression, and self-dissatisfaction. High scorers are socially ill at ease and feel they cannot easily relate to other people, experience an inability to inhibit cravings, and are more vulnerable to stress (Costa & McCrae, 1980, 1984a; Luteijn et al., 2000). An example of a neuroticism item is "I often hate myself." Interviewers asked respondents to indicate whether various similar statements applied to them (yes = 2, do not know = 1, no = 0). Total scores range between 0 and 50. These DPQ items have strong negative relations with the Emotional Stability scale of the Revised NEO Personality Inventory (NEO-PI-R; Luteijn et al., 2000). We tested the utility and psychometric properties of the shortened scale, and it appeared to be a reliable (Cronbach's 
=.86) and valid instrument to measure neuroticism in the elderly population (Steunenberg et al., 2003).
Independent variables
We analyzed demographic variables and physical health-related variables for their association with stability or change of the level of neuroticism. We determined the presence of chronic diseases by asking the respondents whether they had any of the following diseases: cardiac disease; peripheral atherosclerosis of the abdominal aorta or the arteries of the lower limb; stroke; diabetes mellitus; lung disease (asthma or chronic obstructive pulmonary disease); malignant neoplasms; arthritis (rheumatoid arthritis or osteoarthritis); or any other major chronic diseases (Central Bureau of Statistics, 1989). We calculated the number of chronic diseases by summing up all specific diseases reported to be present. In a validation study, respondents' self-reports were compared with information obtained from their general practitioners, and this proved to be sufficiently reliable (Kriegsman, Penninx, Eijk, Boeke, & Deeg, 1996). For the present study, the presence of chronic diseases was indicated as 0 = no disease, 1 = one disease, and 2 = more than one disease.
We measured cognitive functioning by means of the Mini-Mental State Examination (MMSE; Folstein et al., 1975), a frequently used screening instrument for global cognitive dysfunctioning. On 23 questions and tasks, respondents received 1 or more points when they gave the correct answer or performed the task correctly. Scores ran from 0 (all answers incorrect) to 30 (unimpaired). We divided respondents into two categories: MMSE < 24 (impaired cognition, 0) and MMSE 
24 (normal cognition, 1).
We assessed functional limitations by asking the respondent the degree of difficulty he or she had with the following activities of daily living: climbing up and down a staircase of 15 steps without stopping, cutting one's own toenails, and using public transportation (Kriegsman, van Eijk, & Penninx, 1997; van Sonsbeek, 1988). Response categories were "yes, without difficulty," "yes, with difficulty," "only with help," and "no, I cannot." The number of limitations ranges from zero (low) to three (high). For this study, the presence of functional limitations was indicated as 0 = no functional limitations, 1 = one functional limitation, and 2 = more than one functional limitation.
The demographic variables included age group, gender, and level of education. Age ranged between 55 and 85 at baseline; for the purpose of data analyses, we categorized age in three age groups: 55–64, 65–74, and 75–85. We measured education on a 3-point ordinal scale: low (elementary or lower vocational education = 1), middle (general education = 2), and high (higher vocational, college, or university = 3).
Data Analysis
We analyzed data in three steps. First, we tried to find significant predictors of mean neuroticism level at baseline. To assess cross-sectional differences in the mean levels of the neuroticism scores, we performed univariate analyses of variance (ANOVA). In order to attain the most parsimonious set of predictors, we performed a forward-stepwise regression analysis. Although the observations on the neuroticism items were not normally distributed, because of the large number of observations the skewness had only a slight effect on the power of the ANOVAs and regression analysis (central limit theorism; see Bock, 1975).
Second, we assessed longitudinal changes in mean neuroticism score. To find out how many respondents show a clinically relevant change in mean neuroticism level during follow-up, we used a relevant change criterion. Clinical significance refers to the practical value or importance of the effect of an intervention, that is, whether it makes any real difference to the clients or to others in their functioning and everyday life. We defined a relevant clinical change as a decrease or increase of at least 8 points on the DPQ between two measurements. The difference of 8 points represents, in clinical terms, a large change (Lipsey & Wilson, 1993). It is greater than 5, which, on this scale, corresponds to the threshold for statistically significant change (Drenth, 1972; Jacobsen & Truax, 1991).
The third data analytic approach that we used in this study was multilevel analysis. We investigated the longitudinal stability or change of neuroticism with aging by using a multilevel analysis (also known as random coefficient analysis; see Goldstein, 1995; MlwiN, version 1.10.0007; Centre for Multilevel Modeling, Institute of Education, London). We performed this type of analysis because this statistical technique makes it possible to analyze longitudinal relationships by using all available longitudinal data, without summarizing the longitudinal development of each subject into one value. In contrast to the more traditional methods of longitudinal data analysis (i.e., a multivariate analysis of variance, or MANOVA, for repeated measures), which require a complete longitudinal data set, in a multilevel analysis, both the number of observations per individual as well as the time interval between observations may vary. Multilevel analysis does not use strict conditions concerning the type of missing data, because it assumes data missing at random: that is, given the observed data, the unobserved data are random.
In the present study, we defined a two-level hierarchy to form random regression models to describe the individual variability in the longitudinal development of neuroticism in an aging population. The first level is defined by age (range 55–85) and the second level by the respondents. Because the neuroticism score was highly skewed, we performed the longitudinal analyses by using a logarithmic transformation.
We entered both the outcome variable of neuroticism as well as age as continuous variables. We included age in all models, in order to test the influence of aging on the neuroticism score. We tried to find the best model for neuroticism in aging. After assessing the "crude" linear association of neuroticism with aging, we added a quadratic age effect to the model (Model 1).
Model 1:
|
|
Model 2 (adjusted for all covariates):
|
|
|
RESULTS |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
|
24) showed a significant higher level of neuroticism than those with a normal cognitive functioning (p <.01). We found significant associations with the number of chronic diseases (p <.001) and functional limitations (p <.001). The more chronic diseases and the more functional impaired, the higher the score on the neuroticism questionnaire was. The results of the multiple linear regression analysis are shown in Table 2. The regression equation with four of the predictors was significant, R =.25, adjusted R2 =.06, F(4, 2088) = 34.32, p <.001. The predictors cognitive functioning (p =.727) and age group (p =.183) showed no significant relation with neuroticism after the other factors were controlled for. Functional limitations showed the strongest association with the mean level of neuroticism. This predictor alone accounted for 4% of the variance of the baseline level of neuroticism, whereas the other variables contributed an additional 2%.
|
In Table 3, we present the neuroticism mean score differences between baseline and first or second follow-up, answering the question of how many respondents show a clinically significant change. We categorized respondents by the strength of the effect size of this mean-level change. A large majority of nearly 90% shows a small to medium effect size (
.90), indicating stability. A great effect size (difference of more than 8 points) is seen as a significant clinical change in mean neuroticism level; 4.6% of the respondents between baseline and first follow-up or second follow-up show a clinically significant decrease, and 7.8% of the respondents between baseline and first follow-up and 6.6% of the respondents between baseline and second follow-up show a clinically relevant increase of neuroticism level.
|
|
|
|
DISCUSSION |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
).
However, the longitudinal multilevel analysis reports a statistically significant change with aging. We found a U-formed course, showing a slight decrease until the age of 70 and a slight increase after this age. This U-formed course is very hard to explain. In general there are two possible explanations for changes in mean score in late life. The first possible explanation is a survivor effect. The longitudinal results indicate an effect of physical health on the stability of neuroticism, whereas adjusting the regression model for number of functional limitations slightly increases the stability of the development of neuroticism in an aging population. The more frail elderly individuals were more often lost by attrition. The respondents who were lost by attrition during follow-up showed significantly more difficulties while performing daily living activities. Costa & McCrae (1985) show that respondents who have a higher number of chronic diseases or functional limitations are higher in neuroticism. The decline in mean level of neuroticism might be explained by the attrition of the most frail elderly individuals, suggesting that those respondents who continue the study are the more healthy respondents. The slight increase after the age of 70 may be explained by an aging effect. Aging is known to be associated with an increase of physical illness and associated functional limitations, and a decrease of social resources. These factors are also known to be related with the mean neuroticism level (Costa and McCrae, 1985, 1988).
The mean neuroticism-level change for the total sample found in the longitudinal multilevel analyses was about 2 DPQ points, which proved to be statistically significant, but it is rather small and, according to our relevant change criterion, cannot be seen as a clinically significant change (Drenth, 1972; Jacobsen & Truax, 1991; Lipsey & Wilson, 1993).
The degree of change and the impact of that change are important for deciding clinical relevance. Merely examining the absolute level of change is not always enough to determine whether the difference or change really makes a difference. A mean change of 2 points on a scale with a maximum score of 50 is small, and the impact on everyday performance is negligible. Preceding these multilevel analyses, we analyzed how many respondents did change, using a clinically relevant change criterion. Of all the respondents, 5% showed a clinically significant decrease and 7% an increase.
Finally, we searched for significant predictors of individual differences in changes in neuroticism trajectories. None of the physical health-related variables were associated with individual differences in neuroticism trajectories. In other words, for those respondents who do change, the mean level of neuroticism is not affected by the number of chronic diseases, functional limitations, or level of cognitive functioning. A recent study by Mroczek and Spiro (2003) also found no significant association for functional limitations.
The strengths of this study are the large study population (n = 2,117) and the prospective follow-up design. The present study is one of the first 6-year longitudinal designs, investigating age changes and differences in the personality dimension neuroticism in a large, representative population of elderly persons. Because of the multidisciplinary design of the LASA study, we found it not only possible to examine the relationship between the level of neuroticism and the demographic characteristics, but we could also study whether physical health-related characteristics did influence the stability of neuroticism in the aging.
A limitation of the present study is the large number of respondents lost by attrition (31.7%). As in all community-based studies of elderly persons, the older old and less educated, those with cognitive impairments, and the respondents with more chronic diseases and difficulties while performing daily living activities were more often lost through attrition, which may threaten the generalizability of the results.
In summary, we can conclude that neuroticism remains rather stable in old age. In the cross-sectional analyses, age shows no significant association with neuroticism. The magnitude of the stability coefficient is high and in line with findings supporting the stability assumption (Costa & McCrae, 1988, 1994); 12% of the elderly individuals show a clinically relevant change. Finally, the longitudinal change found to be statistically significant cannot be considered clinically relevant (Drenth, 1972; Jacobsen & Truax, 1991; Lipsey & Wilson, 1993). In addition to this, we hypothesized that we would find significant influences of age-graded deteriorations in physical health and related declines in daily functioning on individual trajectories of neuroticism. However, our results showed that individual differences in change of neuroticism in late life are not affected by individual differences in physical health-related variables. In conclusion, neuroticism remains rather stable in very old age; although we did find significant individual differences in trajectories, we could not explain these differences by late-life increase in physical illness and associated functional limitations.
|
Footnotes |
|---|
Received for publication October 17, 2003. Accepted for publication July 16, 2004.
|
References |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
This article has been cited by other articles:
 |
|
 |
|
  A. Terracciano, R. R. McCrae, and P. T. Costa Jr Longitudinal trajectories in guilford-zimmerman temperament survey data: results from the Baltimore longitudinal study of aging. J. Gerontol. B. Psychol. Sci. Soc. Sci., March 1, 2006; 61(2): P108 - P116. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||
| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
