| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
ORIGINAL ARTICLES |
From the Laboratory of Personality and Cognition, National Institute on Aging, National Institutes of Health, Department of Health and Human Services, Baltimore, MD.
Address correspondence and reprint requests to Alexander Weiss, PhD, Laboratory of Personality and Cognition, National Institute on Aging, National Institutes of Health, Department of Health and Human Services, 5600 Nathan Shock Drive, Baltimore, MD 21224-6825. E-mail: weissal{at}grc.nia.nih.gov
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMETHODSRESULTSDISCUSSIONCONCLUSIONSNOTESREFERENCES |
|---|
Methods: Controlling for demographic and health measures, we used Cox regression to test whether the NEO Five-Factor Inventory (NEO-FFI) Neuroticism, Extraversion, Openness to Experience, Agreeableness, and Conscientiousness domains predicted all-cause mortality over 5 years in 1076 65- to 100-year-old participants who took part in a Medicare Demonstration study. Supplementary analyses on 597 participants aged 66 to 102 who were reassessed 2 years later were conducted to determine whether any of the Revised NEO Personality Inventory (NEO-PI-R) facets were related to mortality.
Results: When personality domains were treated as continuous variables, NEO-FFI Neuroticism and Agreeableness were significant protective factors. When personality domains were trichotomized, NEO-FFI Conscientiousness was a protective factor. In a third analysis, Agreeableness was not a significant predictor in a model that included the continuous Neuroticism and trichotomized Conscientiousness variables. Analysis of the NEO-PI-R Neuroticism, Agreeableness, and Conscientiousness factors showed that Agreeableness and Conscientiousness were protective and that there was a trend for a similar effect of Neuroticism. Facet-level analyses revealed that the Impulsiveness, Straightforwardness, and Self-Discipline facets of Neuroticism, Agreeableness, and Conscientiousness, respectively, were prospectively related to greater survival over a 3-year interval.
Conclusion: The effects of Neuroticism and Agreeableness on mortality are inconsistent across previous studies. This study indicates that, in a sample of older, frail participants, high Neuroticism and Agreeableness scores are protective and that more specific effects are primarily the result of the Impulsiveness and Straightforwardness facet scales. The Conscientiousness findings are consistent with those in earlier studies and demonstrate the importance of the Self-Discipline facet.
Key Words: mortality • conscientiousness • neuroticism • agreeableness • Five-Factor Model • elderly
Abbreviations: NEO-PI = NEO Personality Inventory, NEO-PI-R = Revised NEO Personality Inventory, NEO-FFI = NEO Five-Factor Inventory, MMPI = Minnesota Multiphasic Personality Inventory, ADL = activities of daily living, IADL = instrumental activities of daily living, HR = hazards ratio.
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONCONCLUSIONSNOTESREFERENCES |
|---|
Later studies using different measures also examined the relationship between mortality and personality. Two studies examined mortality risks conferred by Neuroticism and Extraversion. A large, early study by Huppert and Whittington (7) found no relationship between either Neuroticism or Extraversion, as measured by the Eysenck Personality Inventory (8), and mortality over a 7-year follow-up period. Korten and colleagues (9) tested whether scores on the Neuroticism or Extraversion scales of the Eysenck Personality Questionnaire–Revised (10) predicted mortality over a 3- to 4-year period in 897 older individuals (70+ years in age) and found that Extraversion was unrelated to subsequent mortality, but that, surprisingly, Neuroticism was protective.
Maier and Smith (11) tested whether their own abbreviated and unvalidated version of the NEO Five-Factor Inventory (NEO-FFI) (12) Neuroticism, Extraversion, and Openness domains predicted mortality. They found that high Neuroticism, low Extraversion, and low Openness were risk factors but that, after controlling for age, these predictors were no longer significant (11).
Friedman et al. (13) conducted a 60-year follow-up study on 1178 gifted children who were first assessed in the 1920s as part of the Terman Life-Cycle Study. Friedman and his collaborators found that "Social Dependability" was related to decreased mortality risk, even after controlling for several potential confounds (13). Friedman et al. also found that "Cheerfulness," but not "Sociability" was a risk factor for mortality (13). Moreover, in men, Friedman et al. found that "Permanency of Mood" was marginally protective (13). A later study (14) on the same sample examined correlations between these archival personality measures and the 5 domains of the NEO-FFI (12), a well-validated measure of the Five-Factor Model (FFM). Correlations were moderate in size and indicated that "Social Dependability" was related to Conscientiousness; "Permanency of Mood" was related to low Neuroticism; "Cheerfulness" was related to Agreeableness; and "Sociability" was related to Extraversion (14).
Two recent studies assessed whether the 5 NEO-FFI (12) domains were related to mortality. Christensen et al. (15) found that, even after controlling for relevant demographic and clinical predictors, patients suffering from chronic renal insufficiency who had low Conscientiousness or high Neuroticism scores were more likely to have died during a 4-year follow-up period. Recently, Wilson and colleagues (16) found that Neuroticism was a risk factor for mortality, and Conscientiousness was protective in a sample of older members of religious orders; they also found mixed results for Extraversion. [On page P110 of their paper, Wilson et al. (14) mistakenly state that Almada et al. (6) reported a relationship between neuroticism and mortality. In fact, Almada et al. only reported a relationship between cynicism and mortality.]
The NEO Personality Inventory (17) was examined by Boyle et al. (18) to determine whether Conscientiousness predicted coronary heart disease mortality or all-cause mortality in coronary patients and found that high Conscientiousness was associated with a reduced risk of mortality.
When previous studies are summarized within the FFM framework (19), the sole domain that consistently predicts mortality across studies is Conscientiousness. This is not surprising as a recent meta-analysis demonstrated that Conscientiousness is positively related to health-promoting and negatively related to health-harming behaviors (20). The results for Neuroticism and Agreeableness are mixed, and those for Extraversion and Openness are weak or nonexistent.
Given the relationship between Conscientiousness measures and mortality in previous studies, we tested the hypothesis that, even after controlling for demographic variables and health, NEO-FFI Conscientiousness (12) predicts all-cause mortality in a population of older, frail men and women who participated in a Medicare Demonstration study. Our study differs from previous studies in important ways. First, we used a community-dwelling sample and not a specific subgroup of the population such as the gifted (13), patients with chronic renal insufficiency (15), or members of religious orders (16). The present sample is also at higher risk, older, less educated, and likely lower in socioeconomic status (SES) than those used in previous studies. Additionally, when compared with other studies of personality and all-cause mortality, our sample size is comparable to the sample of Friedman et al. (13) and smaller than only 2 studies (6,7).
The FFM is a hierarchical structure of traits; the clusters of specific traits that define factors are called domains, and the traits making up these domains are called facets. (19) Thus, while summarizing previous results within the FFM framework and testing for the effects of the FFM domains, it does not enable one to examine which facets are related to mortality.
It is for this reason that we will follow up any significant relationship between mortality and NEO-FFI domains with supplementary analyses on a subset of the participants who completed the Revised NEO Personality Inventory (NEO-PI-R), which has 30 scales, 6 facet scales for each domain and the 5 domain scales derived by summing the 6 respective facet scales. The purpose of the supplementary analyses will be to determine whether risk or protective effects of a domain can be attributed to specific lower-order traits or facets of that domain.
|
METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONCONCLUSIONSNOTESREFERENCES |
|---|
At baseline, 247 participants were not administered the NEO-FFI, because they failed a cognitive screen consisting of being able to answer 3 questions about subjective health, functional status, and life satisfaction and recalling at least 1 of 3 words—book, watch, table—that had been presented 5 minutes earlier. An additional 51 participants who failed the cognitive screen completed the NEO-FFI because it was administered later by a category of personnel different from those that administered the cognitive screen. These 51 participants were excluded because they should not have been administered the NEO-FFI. We also excluded 11 participants who did not qualify for the Medicare Demonstration for other reasons, 1 participant who was not classified with respect to eligibility, and 1 participant who did not provide information on educational achievement.
Among remaining participants, 32 were eliminated because they did not report that they answered the NEO-FFI questions honestly and accurately (12). Additionally, 4 participants’ NEO-FFIs were not valid because they did not answer 10 or more questions, and 13 participants’ NEO-FFIs were not scored for other reasons.
Of the remaining 1084 participants, 4 did not provide information on diabetic or cardiovascular disease status, 2 did not have data on IADL restrictions, and 2 were later found to be ineligible.
The remaining 1076 participants ranged in age from 65 to 100 years (M age = 79.6; SD = 7.54) and included 292 males (M age = 78.7; SD = 7.41) and 784 females (M age = 80.0; SD = 7.56). The sample was primarily composed of non-Hispanic whites (96.3%), with the remainder being African-American (2.7%), white Hispanic (0.8%), or Native American (0.1%). Educational achievement in this sample was low: 38.8% did not finish high school, and only 12.7% completed a 4-year college degree or postgraduate studies. Additional information about this sample is provided in Table 1.
|
Interview Procedures
Trained interviewers orally administered all questionnaires shortly after participants entered the study. In case of hearing problems, interviewers spoke more loudly and slowly. Interviews were conducted in participants’ homes. Interviewers attempted to minimize distractions and conducted interviews mostly in private.
Although the NEO-FFI is usually administered via paper and pencil, the manual states that oral administration of the NEO-FFI is an accepted alternative (12), and oral administration has been used in at least one other study (21).
Measures
Mortality
Because of staggered enrollment times, mortality was determined between 3.10 and 4.97 years after the baseline phase of the Medicare Demonstration. During the course of the Medicare Demonstration, nurses, voucher specialists, or field data collectors, all of whom regularly visited participants, reported that 201 of the 1076 participants died. Mortality status and date of death for all participants was determined using the Social Security Death Index (22). When we queried the Social Security Index, it had last been updated on July 31, 2003; the most recent death in our sample was on July 23, 2003. A censoring date of July 31, 2003, was assigned to participants who were not found in the Social Security Death Index. Of the 424 deaths that occurred during the 5-year follow-up period, over half (56.6%) occurred within 2 years (104 weeks) after baseline; an additional 40.8% occurred in the third and fourth years (
105–208 weeks) after baseline; and the remainder occurred more than 4 years (208–238 weeks) postbaseline.
Covariates
Covariates were assessed on admission into the study. The covariates were chosen as they were related to health, personality, or mortality. We controlled for 3 demographic variables that have previously been demonstrated to be related to personality or mortality: gender (1 = female; 2 = male); age (0 = 65–74; 1 = 75–84; 2 = 85–100); and educational achievement (0 = did not complete high school; 1 = high school but no university degree; 2 = university degree or more).
We controlled for the presence of diabetes and cardiovascular disease (0 = absent; 1 = present). Diabetic status was ascertained by asking the participant whether they had diabetes mellitus. Cardiovascular disease status was assessed with the Health of Seniors Survey (23), which is sponsored by Medicare. Because of the large amount of comorbidity, cardiovascular disease was defined as present if a participants reported that a physician had informed them that they had any of the following conditions: high blood pressure, stroke, coronary artery disease, myocardial infarction, congestive heart failure, or "other heart conditions, such as problems with heart valves or the rhythm of his/her heartbeat."
We controlled for functional limitations as assessed by the ADL and IADL scales. The ADL and IADL scales used in this study were created using modified questions from the Home Care version of the Minimum Data Set (24). For ease of interpretability and to create similarly sized groups for analyses, we trichotomized ADL restrictions (0 = 0; 1 = 1; 2 = 2–5) and dichotomized IADL restrictions (0 = 0–4; 1 = 5–7).
We also controlled for self-rated health (0 = "fair" or "poor"; 1 = "excellent," "very good," or "good"); cigarette smoking status (0 = nonsmoker; 1 = former smoker or unknown smoking status; 2 = current smoker); and the presence of a major depressive episode (0 = absent; 1 = present) as assessed by the Mini-International Neuropsychiatric Major Depressive Episode Scale (25).
Personality
Personality at baseline was measured with the NEO-FFI, a 60-item questionnaire that was developed as a short form of the 240-item NEO-PI-R. There are 12 items for each NEO-FFI domain. These items were initially chosen by selecting the 12 items with the highest positive or negative loadings on each of the 5 NEO-PI (17) domains. After this first step, some substitutions were made so as to increase the degree to which different facets were represented in the NEO-FFI (12). Each question is accompanied by a 5-item Likert scale ranging from "strongly disagree" to "strongly agree." (12)
Two-week retest reliabilities of the NEO-FFI scales ranged from 0.86 to 0.90 in a college sample (26). A previous study showed that the internal consistencies of the NEO-FFI domains in the participants of the Medicare Demonstration were acceptable (27). Of the 1076 participants, 882 (82.0%) completed all 60 items. In the cases where a participant missed between 1 and 9 items, missed items were scored as neutral (12).
Adult combined gender norms (12) were used to convert raw domain scores into T-scores which have a mean of 50 and SD of 10. These T-scores are typically interpreted as low (T < 45), average (T = 45–55), or high (T > 55).
Statistical Analyses
Cox proportional hazards regression was used to assess predictors of time to event (28). This approach was selected because it is flexible and capable of handling censored data (28). Except where otherwise stated, SPSS 13.0 was used to conduct all analyses. The dependent variable was weeks from baseline assessment to death or, for participants that were not identified as having died by the Social Security Death Index, weeks from baseline to the censoring date. All predictor variables in the analysis were entered simultaneously, and thus effect sizes and levels of statistical significance were estimated after controlling for all other predictor variables.
According to adult norms, 31% of the participants should have T-scores in the high range. Table 1 indicates that, in the present sample, substantially fewer participants than expected scored high in Extraversion, Openness, and Conscientiousness. Also, preliminary examination of how deaths were distributed across low, average, and high T-scores (see Table 1) as well as an examination of the log-log graphs suggests possible violations of proportionality. Cox proportional hazards modeling is based on the proportionality assumption, which is violated when the relative risk of the outcome does not change in the same manner for equivalent changes in the levels of a risk factor or covariate (29). For example, the same per unit changes in a covariate or predictor at low values may be associated with smaller changes in mortality risk than the change in risk associated with per unit changes in covariates or predictors at high values. In cases of Cox regression, there are often circumstances, such as the present, where categorical variables may be appropriate for an otherwise continuous variable (30).
Because of these possible violations of proportionality, we ran 2 Cox proportional hazards regression models. In the first model, we assumed proportionality and personality predictors were treated as continuous variables. In the second model, we did not assume proportionality and personality predictors were trichotomized based on cut points specified in the manual (12): T-scores less than 45 were classified as low, between 45 and 55 were classified as average, and greater than 55 were classified as high. We chose the average group as the reference category. Trichotomization will reduce statistical power, but the effect of Conscientiousness on mortality is typically large, so this should not pose a problem. In addition to addressing possible violations of the proportionality assumption, trichotomized domain scores will be more readily interpretable.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONCONCLUSIONSNOTESREFERENCES |
|---|
|
|
|
|
Of the 5 trichotomized NEO-FFI domains, after controlling for all other predictors, only Conscientiousness was related to subsequent mortality. While there was no significant difference between participants with average and low Conscientiousness scores, the risk of mortality over the follow-up period for participants with high Conscientiousness scores was significantly lower than that of participants with average Conscientiousness scores (see Figure 3). In fact, participants with average Conscientiousness scores were, during the approximately 260 week follow-up period, more than twice as likely to die as those with high Conscientiousness scores. Because there was no significant difference between participants with average and low Conscientiousness scores, we compared participants with high Conscientiousness scores to the remaining participants. After controlling for all other significant effects, compared with participants who were average or low in Conscientiousness, those high in Conscientiousness were less than half as likely to die over the 5-year follow-up period (hazards ratio [HR] = 0.47; 95% C.I. = 0.29–0.76; p < .01). The mean difference in Conscientiousness between the 2 Conscientiousness groups was 14.78. Extrapolating, this suggests that each SD increase in Conscientiousness was approximately related to a 35.9% decreased risk of mortality.
|
The violation of the proportionality assumption likely explains why Conscientiousness was not a significant predictor of mortality when treated as a continuous variable. The purpose of the third analysis was to resolve the conflicting results of the 2 previous analyses. This analysis used the same participants and covariates as the primary analyses, but only included the continuous Neuroticism and Agreeableness predictor variables and the trichotomized Conscientiousness predictor variable. As before, even after controlling for all other effects, several covariates including gender, age, smoking status, cardiovascular disease status, self-rated health, and the number of ADL and IADL restrictions were significantly related to mortality. The protective effect of Neuroticism [HR = 0.99; 95% C.I. = 0.97–1.00 (the actual value of the high end of the 95% confidence interval was .997); p < .05] was still significant; each SD of Neuroticism was related to a 14.03% reduction in mortality risk. Also, participants high in Conscientiousness were less than half as likely to die over the follow-up period as participants who had average Conscientiousness scores (HR = 0.48; 95% C.I. = 0.30–0.78; p < .01). On the other hand, the effects of Agreeableness were no longer significant (HR = 0.99; 95% C.I. = 0.98–1.00; p > .05).
In the fourth analysis, we used PROC PHREG in SAS (31) to examine the extent to which the NEO-FFI Conscientiousness findings would generalize across differently chosen cut points. Every possible Conscientiousness cut point from 30 to 72 (Ts = 30–72) was examined in the present sample. The models included the same covariates and NEO-FFI Neuroticism and Agreeableness. Hazard ratios were not statistically significant along the T-score cut-point range from 30 to 52. However, cut points ranging from normal to high T-score ranges, ie, from T = 53 to 64 yielded statistically significant HRs. These results indicate that the Conscientiousness effect generalized across a wide range (12 possible) of cut points.
NEO-PI-R Analyses
We conducted supplementary analyses to determine whether NEO-PI-R Neuroticism, Agreeableness, or Conscientiousness factors were also related to mortality, and, if they were, whether any of the 6 facets of these factors were related to all-cause mortality. The 48-item NEO-PI-R domains are broader than the 12-item NEO-FFI domains, as they encompass all of the facets of each domain. While the facet-level analyses are exploratory, they represent the first attempt to examine the relationship between specific personality facets and mortality.
The NEO-PI-R analyses were based on a follow-up period of approximately 3 years from the start of the 24-month phase of the Medicare Demonstration and conducted on a sub-sample of 597 of the 1082 participants who, at baseline, had valid NEO-FFIs, passed the cognitive screen, and were not later found to be ineligible. Attrition was largely attributable to the fact that 324 participants died between the baseline and reassessment phase and that 67 failed the cognitive screen when it was readministered. Of these 691 participants, NEO-PI-R data were not available for 63 participants, invalid for 13 participants who reported that they did not answer NEO-PI-R questions honestly and accurately, and invalid for 1 participant who missed more than 40 items (12). In addition, 11 participants did not report whether they had diabetes or cardiovascular disease, and data on ADL restrictions or depression were not available for an additional 6 participants.
Ages of the 597 participants ranged from 66 to 102 years (age M = 80.7; SD = 7.21). There were 144 men ranging in age from 67 to 101 (age M = 79.7; SD = 6.74) and 453 women ranging in age from 66 to 102 (age M = 81.0; SD = 7.33). Other characteristics of this sample are provided in Table 4.
|
Of these 597 participants, the Social Security Death Index gave a date of death for 108. Approximately 45.4% of these deaths occurred within 52 weeks after the beginning of the 24-month follow-up phase (96–149 weeks postbaseline); an additional 44.4% of the deaths occurred between 53 and 102 weeks after the beginning of the 24-month follow-up phase (149–200 weeks postbaseline); and the remaining deaths occurred between 105 and 145 weeks after the beginning of the 24-month follow-up phase (201–238 weeks postbaseline).
Covariates identical to those assessed at baseline were assessed at the beginning of the 24-month follow-up phase using an identical interview procedure. Personality at the 24-month follow-up phase was assessed with the NEO-PI-R, a 240-item questionnaire that measures the 5 domains of personality and the 6 facets making up each of the domains. The only difference between personality assessment at the baseline and 24-month follow-up phase was that most of the 597 participants completed the NEO-PI-R in the standard written format, though, if participants could not complete the NEO-PI-R on their own, it was orally administered.
Responses on the NEO-PI-R are made on a 5-point Likert scale identical to that of the NEO-FFI (12). The manual presents evidence for the cross-observer validity of the NEO-PI-R domains (12); 2-year retest reliabilities of the NEO-PI-R domains ranged from 0.83 to 0.91 in the Baltimore Longitudinal Study of Aging sample (32). The internal consistencies of the 5 domains in the present sample ranged from 0.82 (Extraversion and Openness) to 0.90 (Neuroticism) and were comparable to values in the manual (12). Of the 597 participants, 490 (82.1%) completed all 240 items. If a participant missed between 1 and 40 items, missed items were scored as neutral (12). Adult combined gender norms were used to compute T-scores for the NEO-PI-R facets and all 30 facet T-scores were then used to create the 5 weighted factor scores, which provide a more accurate measure of the 5 factors than domain scores (12).
Based on the results of the previous analyses, the first analysis of the NEO-PI-R factors included continuous Neuroticism and Agreeableness predictor variables. The analysis also included a dichotomized Conscientiousness predictor variable, which distinguished between those with high (T > 55) scores and those with low or average scores (T = 45–55). Significant covariates included having 1 ADL (HR = 1.68; 95% C.I. = 1.00 (the actual value of the low end of the 95% confidence interval was 1.002)-2.81; p < .05) or 2 to 5 ADLs (HR = 2.10; 95% C.I. = 1.24–3.56; p < .01). Also, participants with 5 to 7 IADL restrictions were also at greater risk for mortality over the follow-up period (HR = 1.64; 95% C.I. = 1.04–2.60; p < .05).
After controlling for covariates (gender, age, educational achievement, smoking status, presence of cardiovascular disease or diabetes, number of ADL or IADL restrictions, self-rated health, and the presence of a major depressive episode), there was a trend indicating that NEO-PI-R Neuroticism was protective (HR = 0.98; 95% C.I. = 0.95–1.00; p < .06); each SD of Neuroticism was related to a 21% decrease in the mortality risk. The protective effects of NEO-PI-R Agreeableness (HR = 0.97; 95% C.I. = 0.95–1.00 (the actual value of the high end of the 95% confidence interval was .998); p < .05) were still significant; each SD in Agreeableness was related to a 20.8% decrease in the risk of mortality. Participants who were rated as being high in Conscientiousness also were less than half as likely to die over the follow-up period than those with average or low NEO-PI-R Conscientiousness (HR = 0.41; 95% C.I. = 0.18–0.92; p < .05). The mean difference in Conscientiousness between the 2 NEO-PI-R Conscientiousness groups was 17.15. Extrapolating, this suggests that each SD increase in Conscientiousness was approximately related to a 34.4% decreased risk of mortality.
The second analysis examined whether any of the Neuroticism, Extraversion, or Conscientiousness facets were related to mortality over the follow-up period. Because facets of any domain are highly intercorrelated, we conducted 3 sets of 6 analyses. The first 6 tested whether any of the Neuroticism facets were related to all-cause mortality, the second 6 tested whether any of the Agreeableness facets were related to all-cause mortality, and the last 6 tested whether any of the Conscientiousness facets were related to all-cause mortality. In the models used to test the Neuroticism and Agreeableness facets, the facets and factor scores for the other personality domains were treated as continuous variables. In the analyses used to test the Conscientiousness facets, the Conscientiousness facets were dichotomized to distinguish between participants scoring high (T > 55) and those scoring low or average (T 
55). Neuroticism, Extraversion, Openness, and Agreeableness were trichotomized in these models.
When controlling for the other predictors in the model, the only significant Neuroticism facet was N5: Impulsiveness (HR = 0.96; 95% C.I. = 0.94–0.99; p < .01); each SD increase in Impulsiveness was related to a 32.82% decline in mortality risk. The only significant covariate was number of ADL restrictions; the presence of 1 ADL (HR = 1.76; 95% C.I. = 1.05–2.94; p < .05) or 2 or more ADLs (HR = 2.19; 95% C.I. = 1.30–3.69; p < .01) versus no ADLs were both related to a greater risk of mortality. The only other significant predictor in this model was Agreeableness (HR = 0.97; 95% C.I. = 0.95–0.99; p < .05).
When controlling for the remaining predictors, the tests of the Agreeableness facets revealed that only A2: Straightforwardness (HR = 0.97; 95% C.I. = 0.95–0.99; p < .01) was a significant protective factor; each SD increase in Straightforwardness was related to a 25.49% decline in mortality risk. The only significant covariates that were risk factors included having 2 or more versus no ADL restrictions (HR = 2.00; 95% C.I. = 1.18–3.38; p < .05) and having 5 to 7 IADL restrictions (HR = 1.67; 95% C.I. = 1.05–2.66; p < .05).
The tests of the Conscientiousness facets revealed that only C5: Self-Discipline (HR = 0.45; 95% C.I. = 0.21–0.97; p < .05) was a large and significant protective factor even after controlling for all other predictor variables; participants high in C5: Self-Discipline were less than half as likely to die during the follow-up period than those with low or average scores. As in the previous analysis, having 2 or more ADLs versus no ADLs (HR = 2.01; 95% C.I. = 1.19–3.42; p < .01) and having 5 or more IADLs (HR = 1.68; 95% C.I. = 1.05–2.67; p < .05) were related to a greater mortality risk. There was a trend for the overall Neuroticism effects (p < .06) and the difference between participants with high and average Neuroticism scores was significant and indicated that those with high scores were at less risk for mortality during the 3-year follow-up period (HR = 0.58; 95% C.I. = 0.37–0.91; p < .05).
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONCONCLUSIONSNOTESREFERENCES |
|---|
Given the results of the NEO-FFI analyses, the second set of analyses tested whether the NEO-PI-R Neuroticism, Agreeableness, or Conscientiousness factors or any of their facets predicted mortality over a 3-year follow-up period. While the Neuroticism effects were no longer significant, there was a trend suggesting that it was protective. As in the NEO-FFI analyses, the effects of NEO-PI-R Agreeableness and Conscientiousness were significant and protective.
Analyses that examined NEO-PI-R facet-level predictors indicated that the Impulsiveness, Straightforwardness, and Self-Discipline facets of Neuroticism, Agreeableness, and Conscientiousness, respectively, were protective factors.
The protective effects for Neuroticism and the Impulsiveness facet were smaller than those of Conscientiousness and the Self-Discipline facet. Previous studies reported mixed results with respect to the relationship between Neuroticism and mortality, with 3 finding no relationship (6,7,11), 2 finding that Neuroticism is a risk factor (15,16), and 1 showing that Neuroticism is protective (9).
The finding that Impulsiveness scores confer lower mortality risk is puzzling. Impulsive individuals are unable to control cravings and urges; hence, they are expected to be more likely to engage in health-harming behaviors such as overeating and smoking (12). Research indicating that individuals with an impulsive personality style are more likely to adopt maladaptive coping styles such as behavioral disengagement or emotional venting and less likely to adopt adaptive coping styles such as acceptance, restraint, or suppressing competing activities (33) suggests that it is unlikely that the protective effect of Impulsiveness is not behaviorally mediated. To the extent that Neuroticism and its facets show mean level declines of about 1 SD decline, "high" Impulsivity scores in this advanced age group are relatively atypical and may possibly be a marker of biological resilience. High impulsivity scores in extreme old age may reflect a degree of "immaturity" or relatively slower rates of biological aging and senescence.
The existing literature on the relationship between Agreeableness and mortality is inconsistent. Some studies showed that MMPI-based measures of Antagonistic Hostility or Cynicism are risk factors (2,3,6), whereas 2 studies that used MMPI-based measures of low Agreeableness (4,5) and 2 studies that used NEO-FFI Agreeableness (15,16) found no evidence of a relationship between Agreeableness and mortality. In the present study, the protective effects of NEO-FFI Agreeableness and the NEO-PI-R Straightforwardness facet were modest. Also, the relationship between NEO-FFI Agreeableness and mortality was not significant after controlling for the effects of the continuous NEO-FFI Neuroticism and trichotomized NEO-FFI Conscientiousness domains.
The Agreeableness and Straightforwardness effects are more readily understandable than those of Neuroticism and Impulsiveness. First, there is ample evidence suggesting that agreeable individuals are at less risk for cardiovascular disease (1). Second, straightforward individuals are more likely to be frank, honest, and sincere and less likely to be manipulative (12). Hence, straightforward individuals may be more likely to have a relationship with their physicians that would lead to better health outcomes.
Even after controlling for demographic and health covariates and the other personality domains, the effects of Conscientiousness and, in particular, Self-Discipline were large and statistically significant predictors of mortality, even in older participants who had little education and were frail. Being low or average in Conscientiousness (HR = 2.14; 95% C.I. = 1.31–3.48) or Self-Discipline (HR = 2.21; 95% C.I. = 1.03–4.74) conferred approximately the same risk as being 85 to 100 versus 65 to 74 years old at baseline (HR = 2.33; 95% C.I. = 1.73–3.13).
The Conscientiousness findings are consistent with the study by Friedman et al. (13), even though their study did not use a standard measure of FFM Conscientiousness. Less surprisingly, these findings are also consistent with previous studies that assessed Conscientiousness using the NEO-FFI (15,16) and NEO-PI (18).
One new and important finding of this study is that it identifies the importance of Self-Discipline, as opposed to other Conscientiousness facets, eg, Order and Achievement Striving. Individuals high in Self-Discipline are characterized by the ability to begin and complete tasks. They are also motivated to accomplish tasks they set for themselves. This facet-level finding is an important addition to the literature in that it suggests a range of hypotheses about how the effects of Conscientiousness might be mediated. People high in Self-Discipline might have an advantage because they may be more proactive in engaging in a variety of health-promoting behaviors while avoiding or minimizing health-damaging behaviors. This hypothesis is consistent with findings of a recent meta-analysis on the relationship between Conscientiousness and health behaviors that demonstrates that Conscientiousness is positively related to behaviors that benefit health and negatively to behaviors that harm health (20).
We found no evidence for a relationship between Extraversion and mortality. This is largely consistent with previous research. Only 1 study that included a measure of Extraversion found evidence for a protective effect of Extraversion and, even in that study, the results were mixed (16). Our null findings for Openness are not surprising as they are consistent with null findings in other studies that included a measure of Openness (11,15,16).
Limitations
One possible limitation is our use of the Social Security Death Index. Cowper and colleagues (34) compared different mortality databases and noted that the Social Security Death Index is less sensitive than the National Death Index and that using the Social Security Death Index may result in missing between 3.4% and 14.9% more cases than one would miss using the National Death Index. However, there are 3 reasons why this is probably not a serious limitation. First, as noted by Cowper et al. (34) and others (35,36), the sensitivity of the Social Security Death Index is higher in older cohorts. Thus, the difference in sensitivity described by Cowper and her colleagues (34) may be an artifact of the low mean age of the cohorts in the studies they sampled. Second, other studies of personality and mortality have assessed mortality using the Social Security Death Index without problems (15). Third, there are no reports of bias in the distribution of personality factors among deaths that are missed by using Social Security Death Index. Therefore, using the Social Security Death Index should not affect the findings with respect to personality.
A second limitation of this study is that, given the mean age of the sample, it may be composed of highly resilient individuals. A third limitation was that the cardiovascular disease and diabetes measures were based on self-reports and not verified by physician records. Clearly it would have been better to have objective measures or physician reports. However, previous studies have indicated acceptable correspondence between self-reported medical conditions and disease diagnoses in the elderly (37,38). Additionally, at baseline, the self-report measure of cardiovascular disease was a risk factor for mortality. Hence, the self-report measures of health in this study were likely valid indices of health status.
A final limitation was that there were only a small number of minorities in the current study. As a result, we cannot determine whether these findings would generalize beyond non-Hispanic whites.
Future Directions
In the future there should also be an attempt to determine whether any specific characteristics of a study, notably, the age and health of the sample, influence the strength and direction of the relationship between Neuroticism and mortality. Similarly, there should be attempts to replicate the findings of a relationship between Impulsiveness and longevity and determine the behaviors underlying this relationship.
Future studies may also want to determine how Agreeableness and Straightforwardness lead to a lower risk of mortality. These studies can examine the relationship between these predictors and mortality and include potential mediators such as cardiovascular measures and the quality of rapport with physicians and other caregivers.
Future studies should also attempt to determine whether Conscientiousness is related to specific causes of death. Researchers interested in longevity may be well advised to try to identify the pathways by which Conscientiousness exerts its longevity effects; for example, identifying specific health-promoting behaviors related to Conscientiousness or Self-Discipline that lead to greater survival among the frail elderly.
|
CONCLUSIONS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONCONCLUSIONSNOTESREFERENCES |
|---|
We thank the Centers for Medicare and Medicaid Services for sponsoring the Medicare demonstration "A Randomized Controlled Trial of Primary and Consumer Directed Care for People with Chronic Illnesses," CMS 95 to 90467, Project Officers: Carolyn Rimes, Tamara Jackson-Douglass, and Donald Sherwood.
We also are grateful to the demonstration P.I., Gerald M. Eggert, Executive Director, Monroe County Long Term Care Program, Inc., East Rochester, NY 14445, and the Co-P.I., Brenda R. Wamsley, Executive Director, Center for Aging & Healthcare in West Virginia, Parkersburg, WV 26101, as well as coinvestigators Jurgis Karuza and Paul Duberstein and the staff who collected the data, as well as the participants in the demonstration. We also thank Bruce Friedman for his invaluable help in coordinating the collection of mortality data and management of the data set.
Finally, we woul thank R. R. McCrae and Corinna Löckenhoff for their helpful comments on earlier versions of the manuscript and Larry Brant and Shan Sheng for his advice on the statistical analyses.
|
NOTES |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONCONCLUSIONSNOTESREFERENCES |
|---|
Received for publication May 4, 2004; revision received June 2, 2005.
DOI:10.1097/01.psy.0000181272.58103.18
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONCONCLUSIONSNOTESREFERENCES |
|---|
This article has been cited by other articles:
 |
|
 |
|
  H. Iwasa, Y. Masui, Y. Gondo, H. Inagaki, C. Kawaai, and T. Suzuki Personality and All-Cause Mortality Among Older Adults Dwelling in a Japanese Community: A Five-Year Population-Based Prospective Cohort Study Am J Geriatr Psychiatry, May 1, 2008; 16(5): 399 - 405. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 G. H. Ironson, C. O'Cleirigh, A. Weiss, N. Schneiderman, and P. T. Costa Jr Personality and HIV Disease Progression: Role of NEO-PI-R Openness, Extraversion, and Profiles of Engagement Psychosom Med, February 1, 2008; 70(2): 245 - 253. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
B. A. Shipley, A. Weiss, G. Der, M. D. Taylor, and I. J. Deary Neuroticism, Extraversion, and Mortality in the UK Health and Lifestyle Survey: A 21-Year Prospective Cohort Study Psychosom Med, November 1, 2007; 69(9): 923 - 931. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 R. S. Wilson, J. A. Schneider, S. E. Arnold, J. L. Bienias, and D. A. Bennett Conscientiousness and the Incidence of Alzheimer Disease and Mild Cognitive Impairment Arch Gen Psychiatry, October 1, 2007; 64(10): 1204 - 1212. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
C. R. Jonassaint, S. H. Boyle, R. B. Williams, D. B. Mark, I. C. Siegler, and J. C. Barefoot Facets of Openness Predict Mortality in Patients With Cardiac Disease Psychosom Med, May 1, 2007; 69(4): 319 - 322. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
B. P. Chapman, J. M. Lyness, and P. Duberstein Personality and Medical Illness Burden Among Older Adults in Primary Care Psychosom Med, April 1, 2007; 69(3): 277 - 282. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
C. R. Cloninger How Does Personality Influence Mortality in the Elderly? Psychosom Med, November 1, 2005; 67(6): 839 - 840. [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
