Abstract
Research shows moderate agreement between different sources of information when assessing personality disorders (PDs) in older adults. The differences between measurement methods appear to be not only the result of measurement error, but also an indication that each source holds nonredundant information relevant to PD diagnosis. The current article examines three sources of diagnostic information (self-report, informant report, and clinical interview) and the utility of these instruments in predicting important outcomes in older adulthood. We analyzed data from 1,630 adults between the ages of 55 and 64 years participating in a longitudinal study of later life. PD symptomatology was assessed using multiple methods, which were then used to predict health, marital satisfaction, and cognitive decline. All three sources contributed significantly to the prediction of these outcomes, with important implications for the assessment of older adults in research and clinical settings.
Keywords
The assessment of personality disorders (PDs) in older adults can be challenging because the expression of PD symptoms varies over time. Both adaptive and maladaptive personality traits exhibit normative or expected changes across the lifespan, continuing through later life (Clark, 2009; Debast et al., 2014). It is often unclear to what degree these changes represent true changes in personality characteristics over a period of many years, and how much they represent age-related changes in social and occupational contexts altering the expression of (mal)adaptive personality traits (Balsis, Gleason, Woods, & Oltmanns, 2007; van Alphen, Derksen, Sadavoy, & Rosowsky, 2012). As such, items that capture personality pathology well during adolescence or young adulthood may have a very different, perhaps less pathological, meaning in later life. The diagnostic criteria in the Diagnostic and Statistical Manual of Mental Disorders, 5th edition (DSM-5; American Psychiatric Association, 2013a) were developed with young adults in mind (Rossi, Van den Broeck, Dierckx, Segal, & van Alphen, 2014), and cross-sectional as well as longitudinal studies show that PD symptoms exhibit lower prevalence rates as people age (Cohen et al., 1994; Johnson et al., 2000). Nevertheless, research into the outcomes associated with PDs in later life shows that these disorders have a long-lasting impact (Chen et al., 2009; Gleason, Weinstein, Balsis, & Oltmanns, 2014). The extent to which the PD criteria continue to capture true pathology, as indicated by associations with functional outcomes, in the face of age-related bias and change is still little understood.
An additional issue in assessing PDs in later life, that may have implications for the study of the association between personality pathology and various outcomes, is the question of whom to ask about an individual’s personality. Researchers have long advocated for the use of informant report measures in the study of normal-range personality (Funder, 1995; Hofstee, 1994). It may be particularly important to use these measures in the study of personality pathology, as individuals who exhibit high levels of PD symptoms may lack insight into the effects that their behaviors have on others (Klonsky, Oltmanns, & Turkheimer, 2002). Previous research has established moderate agreement between self- and informant report of personality pathology. Nevertheless, the two sources lead to different estimates of the prevalence rates of PDs in older adults (Oltmanns, Rodrigues, Weinstein, & Gleason, 2014). This divergence between self and other may not be due entirely to measurement error. Instead, both sources appear to supply unique information to the assessment of personality pathology (Vazire & Carlson, 2011). The self–other knowledge asymmetry (SOKA) model predicts that certain traits, such as internalizing problems, may be more accessible to the self, whereas informants may have preferred access to information about observable behaviors (Vazire, 2010). This asymmetry could lead to certain aspects of personality being emphasized more by one source than another. For example, individuals who are described as paranoid by those who know them well tend to describe themselves as being angry and hostile (Clifton, Turkheimer, & Oltmanns, 2004). Conversely, individuals who described themselves as paranoid tended to be viewed as being cold. As such, relying on one source over another may lead to discrepant conclusions about the course and impact of personality pathology. These results demonstrate the importance of considering multiple sources of information when assessing personality pathology in later life.
The fact that different sources of information provide somewhat inconsistent data regarding personality traits and pathology is now well accepted. The important question to be addressed is now concerned with the relative merits of different measures, with an eye toward establishing the functional impairment associated with personality pathology in later life. Which source is most useful for what specific purpose? Several domains represent clear targets for studying the impact of personality pathology in later life (Oltmanns & Balsis, 2011). Health is one such domain, as many chronic and serious illnesses become more prevalent in later life. Previous research shows that self-reported health is predictive of mortality, over and above more objective health indicators (Benyamini & Idler, 1999; Idler & Benyamini, 1997; Jylhä, 2009). In turn, personality pathology predicts self-rated health (Powers & Oltmanns, 2012). In terms of specific PDs, much of the literature on personality disorders and physical health has focused on borderline PD (BPD), characterized by emotional variability and impulsivity, which has been linked to sleep disturbances, obesity, and chronic illness (Dixon-Gordon, Whalen, Layden, & Chapman, 2015). On the informant side, there are a wide variety of studies that establish a link between informant reports of normal-range personality and health outcomes (Jackson, Connolly, Garrison, Leveille, & Connolly, 2015; Kneip et al., 1993; Smith et al., 2007), with informant reports sometimes outperforming even self-report. Mental health is also an important component of subjective health, and several studies have shown that informant reports of PDs add incremental validity to the prediction of lifetime major depressive episodes and the outcomes of patients with depressive disorders (Galione & Oltmanns, 2013; Klein, 2003). Overall, there is plenty of evidence that personality is predictive of subjective health; yet further research is needed to extend these findings to personality pathology in later life.
Relationship satisfaction is another important domain related to personality pathology, and it is of particular interest in later life. Several transitions (i.e., children leaving the home) during this time period have important implications for dyadic adjustment (Gorchoff, John, & Helson, 2008). Several studies have examined the ability of self- and spouse reports of normal-range personality traits to predict martial satisfaction (Cundiff, Smith, & Frandsen, 2012; Watson, Hubbard, & Wiese, 2000; Watson & Humrichouse, 2006). Judges’ ratings of marital satisfaction are strongly related to their ratings of their partner’s personality traits across the full range of adaptivity, but only weakly related to the self-rated personality traits of the partner (Brock, Dindo, Simms, & Clark, 2016). Conversely, both self- and spouse ratings of the targets’ personality uniquely predict the marital satisfaction of the target (Watson et al., 2000; Watson & Humrichouse, 2006). Another study found that spouse ratings of personality demonstrate incremental validity over self-ratings when predicting depressive symptoms and behavioral observations during a marital disagreement (Cundiff et al., 2012).
Relatively less research has been conducted on self- and spouse ratings of DSM-5 PDs and marital satisfaction. One study found that PDs are predictive of marital satisfaction, verbal aggression, and physical violence, and spouse reports of their partner’s personality contributes significant variance to the prediction of these outcomes, above and beyond self-report of personality alone (South, Turkheimer, & Oltmanns, 2008). Spouse-rated BPD, in particular, is associated with low satisfaction and high verbal aggression in heterosexual married couples (South et al., 2008). As such, it is important not only to measure the effect of a target’s personality on their own marital satisfaction but also the effect that their personality has on their partner’s marital satisfaction. Given the discrepancies between self- and informant report of personality pathology, it may be the case that the role of a target’s personality on marital satisfaction depends on whom one is asking.
Finally, cognitive decline and eventually dementia are later life outcomes with far-reaching personal and societal implications. Normal-range personality traits with links to PDs such as neuroticism and conscientiousness predict risk for the development of dementia of the Alzheimer’s type (DAT; Duberstein et al., 2011). There is also evidence that changes in neuroticism precede the early signs of cognitive decline associated with DAT (Balsis, Carpenter, & Storandt, 2005). Given the strong associations between neuroticism and PD symptoms (Samuel & Widiger, 2008), it is likely that PDs also convey risk for the early stages of dementia. These risks may be best captured by informant reports. Increasing recognition that close others may notice changes in an individual experiencing cognitive decline years before it could be detected by a clinician led to the development of an informant report screening instrument for nonspecific dementia (Galvin et al., 2005). Furthermore, research into the links between personality and cognitive decline shows that informant reports of normal-range personality discriminate between healthy aging and the earliest stages of DAT (Duchek, Balota, Storandt, & Larsen, 2007). As research into the role of personality in cognitive decline continues, it will be important to establish the predictive validity of self- and informant reports of personality in relation to measures of cognitive change.
The purpose of the current article is to examine the incremental validity of different sources of information in the prediction of several outcomes with particular relevance to later life: health, relationship satisfaction, and cognitive decline. Because method variance may obscure some of the unique relationships between personality pathology as reported by multiple sources and outcomes, where possible we included both self- and informant measures of outcomes as well as predictors. Few studies have examined both self- and informant reports of subjective health and relationship satisfaction in relation to both self- and informant reports of personality pathology. No studies, to our knowledge, have examined the relationship between personality pathology and informant reports of cognitive change. Examining the predictive validity of personality pathology obtained from multiple sources will not only provide a wider view of the impact of personality pathology in later life but will also provide researchers and clinicians with the knowledge they need to choose the most useful instruments for their purposes.
Method
Participants and Procedure
The St. Louis Personality and Aging Network (SPAN) is an ongoing longitudinal study investigating the relationships between personality, health, and aging in later life. The first phase (Phase I) of the study began in 2007. Individuals between the ages of 55 and 64 years (M = 59.5 years, standard deviation [SD] = 2.7) were recruited through phone records, with a total of 1,630 participants completing the baseline assessment (see Oltmanns et al., 2014). Sample characteristics were representative of the St. Louis area. Fifty-five percent of the participants were female (n = 889) and 65% were Caucasian (n = 1,060). Measures at baseline included a battery of questionnaires and interviews assessing the participant’s personality, health, life events, relationships, and basic demographics. Follow-up questionnaires assessing health, stressful life events, mood, and social functioning were administered every 6 months either online or through the mail. The fifth and final follow-up of Phase I began 2.5 years after baseline and involved bringing the participants back into the lab for another round of questionnaires and a semistructured diagnostic interview. This follow-up was interrupted by a loss of funding, but was begun again after additional support was obtained for Phase II.
At baseline, each participant was asked to nominate someone who knew him or her well who could serve as his or her informant. A total of 1,467 informants (68% female, n = 1,004) completed the baseline questionnaires either online or through the mail. Ages ranged between 18 and 92 years (M = 55 years, SD = 11.4), and the participants had known the informants for an average of 32 years (SD = 15). Informants were spouses/romantic partners (48%, n = 702), other family members (28%, n = 404), or friends (22%, n = 328), with the remainder falling into some other category, such as coworkers or neighbors (2%, n = 33).
Measures
Structured Interview for DSM-IV Personality (SIDP-IV; Pfohl, Blum, & Zimmerman, 1997)
A semistructured diagnostic interview for the assessment of PD symptoms, the SIDP-IV was administered to participants at both baseline and the fifth follow-up assessment. It consists of 80 items, which the interviewer rates on a scale from 0 (not present) to 3 (strongly present). Given the potential age bias in the criteria, interviewers were trained to rate the criteria from an age-neutral perspective. For example, items in the work style section of the interview were emphasized to apply to both jobs outside the home as well as household chores and volunteering. Items that seemed to tap stereotypes of aging and were frequently endorsed at a nonclinical level by this age group, such as rigidity/stubbornness and hoarding, were probed carefully to determine if these items reflected true pathology. Ratings for the criteria of a particular disorder were summed and divided by the number of items in the scale, producing a mean score. Due to low base rates of schizotypal, histrionic, and dependent PD symptoms, we included only the scales associated with paranoid, schizoid, antisocial, borderline, narcissistic, avoidant, and obsessive–compulsive PDs in the present analyses. Independent judges rated 265 randomly selected video-recorded interviews, and reliability was calculated using a one-way random, average measures intraclass correlation coefficient (Shrout & Fleiss, 1979). The interrater reliability of the total SIDP-IV was .67. Correlations between SIDP-IV scores at baseline and the fifth follow-up ranged from .34 (antisocial) to .68 (avoidant).
Multisource Assessment of Personality Pathology (MAPP; Okada & Oltmanns, 2009; Oltmanns, Turkheimer, & Strauss, 1998)
To obtain both self- and informant reports of personality pathology, we administered the MAPP at baseline and the fifth follow-up. The MAPP consists of 80 items that represent lay translations of each of the criteria for the 10 DSM-IV PDs; each item is rated on a 0 (I am/he or she is never like this) to 4 (I am/he or she is always like this) Likert-type scale. Although the MAPP was originally developed for use with younger adults (Oltmanns et al., 1998), the measure was reexamined in preparation for its use in the SPAN study. Items that had obviously age-biased content were either removed or rewritten to be age-neutral. As with the SIDP-IV, we calculated mean scores for each scale, including only the scales for paranoid, schizoid, antisocial, borderline, narcissistic, avoidant, and obsessive–compulsive PDs in the current analyses. Coefficient alphas for the self-report version of the MAPP at baseline ranged between .54 (antisocial) and .81 (avoidant). For the informant report version, baseline coefficient alphas ranged between .61 (obsessive–compulsive) and .82 (avoidant). As evidenced by these ranges, some of the MAPP scales, such as the antisocial scale, demonstrated low internal consistency. It may be that this finding is due to lack of consistency in the criteria themselves, with certain disorders being more or less uniform in content. In so far as this low consistency indicates error in measurement, it is likely that the present findings are an underestimate of the amount of variance explained by that particular scale. To further test reliability, test–retest correlations were obtained between baseline and the fifth follow-up. These correlations demonstrated good reliability, with self-reported MAPP scores exhibiting correlations ranging from .53 (antisocial) to .68 (avoidant). Informant-reported MAPP scores also exhibited good reliability, with correlations ranging from .60 (antisocial) to .70 (paranoid and narcissistic).
Rand-36 Health Status Inventory (HSI; Hays & Morales, 2001)
The HSI is a 36-item measure of subjective perceptions of mental and physical health. It is composed of eight subscales: physical functioning, role limitations due to physical health problems, pain, general health perceptions, emotional well-being, role limitations due to emotional problems, social functioning, and energy/fatigue. For the current analyses, these subscales were weighted and combined to produce a general health composite score. The reliability and validity of the HSI in community samples of adults is well documented (Moorer, Suurmeijer, Foets, & Molenaar, 2001; VanderZee, Sanderman, Heyink, & de Haes, 1996). Coefficient alphas for the eight subscales in the current sample ranged from .67 (social functioning) to .92 (physical functioning) at baseline.
Informant Report Version of the HSI
One important limitation of using a self-report instrument to assess subjective health is that it can be confounded with pathological personality traits. For example, an individual high on emotional instability may worry and complain about their health to a greater degree than a more emotionally stable individual. In order to address this limitation, we adapted 10 items from the HSI into an informant-report version of the measure. These items were selected based on the ease with which someone who knew the participant well could answer them. Example items include “In general, would you say his/her health is . . .” and “During the past 4 weeks, to what extent has his/her physical health (emotional problems) interfered with his/her normal social activities with family, friends, neighbors or groups?” Participants were provided with five or six response options for each item. For example, informants rated the question “In general, would you say his/her health is . . . ” on a scale from poor to excellent. The scale was scored in the direction of higher scores indicating better health, consistent with the self-report general health composite of the HSI. Coefficient alpha for the informant HSI was .87 at baseline.
Dyadic Adjustment Scale (DAS; Sabourin, Valois, & Lussier, 2005; Spanier, 1976)
In order to measure marital satisfaction we administered the four-item version of the DAS to those participants who were in a romantic relationship (n = 939), as well as those informants who were the spouse or romantic partner of the participant (n = 630). Informants filled out the DAS regarding their own satisfaction with the relationship. In order to avoid confusion, we will refer to the participant’s DAS as representing target dyadic adjustment and informant’s DAS as representing partner’s dyadic adjustment. Originally, the DAS was developed as a 32-item instrument. However, the four-item version used here has been well validated for use in epidemiological studies (Sabourin et al., 2005). Individuals respond to the items on a scale from 0 (Never/extremely unhappy) to 6 (All the time/perfect). Coefficient alpha for the target’s DAS at baseline was .78, whereas coefficient alpha for the partner’s DAS at baseline was .81.
The Ascertain Dementia Eight-Item Informant Questionnaire (AD8; Galvin et al., 2005).
The AD8 is an eight-item informant-report instrument that assesses cognitive change and has been validated as a screening instrument for mild dementia, regardless of etiology. Example items include “Repeats questions, stories, or statements” and “Difficulty remembering appointments.” Informants respond to each item with regard to change over the past several years, indicating either “YES, a change,” “NO, no change,” or “N/A, Don’t know.” Responses of “YES, a change” are summed to produce a total score. The AD8 was validated against the Clinical Dementia Rating (CDR), a commonly used classification system that differentiates between nondemented individuals (CDR = 0), individuals with very mild dementia (CDR = 0.5) and individuals with mild, moderate or severe dementia (CDR = 1-3; Galvin et al., 2005; Galvin, Roe, Xiong, & Morris, 2006). Using a cutoff score of ≥ 2 on the AD8, the sensitivity of the measure in discriminating between CDR = 0 and CDR ≥ 0.5 in a community sample was 85%, and the specificity was 86% (Galvin et al., 2005). In a clinical sample, the values for sensitivity and specificity were 92% and 46%, respectively (Galvin et al., 2006). Given that the current sample is remarkably similar to the community sample reported in Galvin et al. (2005), in that both are drawn from longitudinal studies of aging in the St. Louis area, the AD8 may be expected to demonstrate adequate discriminative validity in the current sample. The AD8 was added to the assessment battery while the fifth follow-up assessment was ongoing. In total, 368 informants completed the questionnaire, however, only 261 participants had complete data on both this instrument and all personality measures. All analyses using the AD8 also use data from the MAPP and SIDP obtained at the fifth follow-up. Coefficient alpha for the AD8 in the current sample was .77. Seventy-eight participants (21% of the subsample) were rated as having an AD8 score of two or greater.
Data Analytic Plan
After running a series of regression analyses without covariates to determine the total proportion of variance explained in each outcome by the personality variables, we conducted a series of hierarchical regression analyses controlling for gender and race. We first ran the analyses with the SIDP-IV mean scores for the personality disorders in the first step, followed by self-reported MAPP in the second and informant-reported MAPP in the third. Afterward, we ran additional analyses with the order of the steps switched, until we had modeled all possible orders.
Results
Means, SDs, and intercorrelations among the dependent variables are presented in Table 1. We obtained correlations between the AD8 and the target/partner DAS, HSI, and IHSI using data from the fifth follow-up assessment. All following analyses that use target/partner DAS, HSI, and IHSI scores use data from the baseline assessment. Self- and informant reports of the participant’s general health were strongly correlated. We also found a moderate correlation between the measures of the target’s and partner’s dyadic adjustment. Finally, the AD8 exhibited moderate-to-strong negative correlations with the measures of dyadic adjustment and health. Correlations between the MAPP and the SIDP-IV PD variables at baseline have been reported previously (see Oltmanns et al., 2014). Overall, the scales for the individual personality disorders are moderately correlated across self- and informant MAPP and the SIDP-IV, indicating moderate agreement between all three sources. The correlations between self-report MAPP and SIDP scores tend to be stronger than the correlations between these two measures and the informant-reported MAPP. This is consistent with the fact that both self-report and semistructured interviews rely on information from the self, whereas informants may have access to information that the self is unaware of or unwilling to disclose on a questionnaire or in the interview room.
Correlations Between Outcome Measures.
Note. N ranges from 155 to 1,079. All correlations significant at p < .01. SD = Standard deviation; HSI = RAND-36 Health Status Inventory; DAS = Dyadic Adjustment Scale; AD8 = Ascertain Dementia Eight-Item Informant Questionnaire.
Source: baseline assessment. bSource: fifth follow-up assessment.
Next, we ran a series of regression analyses examining how much of the variance in each outcome could be attributed to personality pathology across all three sources of information. The total proportion of variance explained is presented in Table 2, alongside the proportion of variance uniquely associated with the self, informant, and interview measures of personality pathology. Overall, the measures of personality pathology explained between 19% and 36% of the variance in the outcomes. In the cases of self-reported general health status, the target’s dyadic adjustment, and scores on the AD8, the majority of the variance explained was attributable to the shared variance between self, informant, and interview measures. However, on the informant-report HSI and the partner’s DAS, the proportion of variance uniquely associated with the informant’s report of the participant’s personality pathology was the largest.
The Variance Explained by Self-Report, Informant Report, and Semistructured Interview of Personality Pathology on a Series of Outcomes.
Note. HSI = RAND-36 Health Status Inventory; DAS = Dyadic Adjustment Scale; AD8 = Ascertain Dementia Eight-Item Informant Questionnaire.
Source: baseline assessment. bSource: fifth follow-up assessment.
Finally, we were interested in comparing the incremental validity of interview, self- and informant report of personality pathology. We ran a series of hierarchical regression analyses predicting each of the five outcomes, controlling for gender and race. First, we entered SIDP-IV scores in Step 1, self-reported MAPP scores in Step 2, and informant-reported MAPP scores in Step 3. We then switched the order in which we entered the PD measures, to compare the incremental validity of each measure against the other two in turn. Table 3 presents the R2 values associated with each of the regression models. All measures resulted in a significant increase in proportion of variance explained in self-reported health, regardless of the order in which they were entered (Fs ranged from 6.92 to 56.35, all p < .001). The same was true when predicting informant report of health, although the increase in variance explained by adding SIDP over and above self- and informant-reported MAPP was relatively weak (ΔR2 = .01, F = 2.37, p < .05; other Fs ranged from 3.84 to 76.04, all p < .001). Adding self-report to the model predicting target DAS scores after entering SIDP and informant report of personality pathology did not contribute significant variance (ΔR2 = .01, F = 1.90, p > .05); however, all other additions were significant at p < .01 (Fs ranged from 2.67 to 18.54). The results for partner DAS were less uniform. Adding self-reported MAPP scores did not contribute significant variance at any step in the model (Fs ranged from 0.51 to 1.74, all p > .05). SIDP scores contributed significant variance over covariates (F = 3.35, p < .01) and self-reported MAPP (F = 2.08, p < .05), but not informant-reported MAPP (Step 2: F = 0.67, p > .05; Step 3: F = 0.64, p > .05). Consistent with the analyses reported in Table 2, informant-report carried the weight of the predictive ability when examining target DAS, adding significant variance regardless of the step in which it was entered (Fs ranged from 35.87 to 40.04, all p < .001). Finally, the regression models predicting AD8 scores also showed a weaker effect for self-reported personality when entered after informant report (Step 2: F = 1.96, p > .05; Step 3: F = 2.24, p < .05), but otherwise all personality measures added significant variance regardless of the step in which they were entered (Fs ranged from 2.73 to 8.81, all ps < .01).
R2 Values for Regression Models Predicting Health, Dyadic Adjustment, and Cognitive Decline.
Note. SIDP = Structured Interview for DSM-IV Personality; MAPP = Multisource Assessment of Personality Pathology; HSI = Rand-36 Health Status Inventory; IHSI = Informant Health Status Inventory; DAS = Dyadic Adjustment Scale; AD8 = Ascertain Dementia Eight-Item Informant Questionnaire. Ns range from 261 to 1,413. Gender and race included as covariates in all models.
Regression models run using data from baseline assessment. bRegression model run using data from fifth follow-up assessment.
Tables 4 through 6 present the estimated regression coefficients for the final model of the regression analyses predicting self- and informant-reported health (Table 4), target and partner dyadic adjustment (Table 5), and informant-reported risk for cognitive impairment (Table 6). We found several PDs that consistently and uniquely predicted both self- and informant report of general health. Schizoid PD, as measured by the SIDP, was negatively related to health outcomes, whereas self-reported narcissistic PD exhibited a positive association. Only the BPD scales were uniquely predictive of both self- and informant-reported general health across all three measures of personality pathology. With regard to dyadic adjustment, both self- and informant-reported schizoid PD symptoms uniquely predicted lower target dyadic adjustment (b = −0.56, SE = 0.24, p < .001; b = −0.53, SE = 0.21, p < .05, respectively), as was clinical interview of avoidant PD symptoms (b = −1.27, SE = 0.43, p < .01). BPD symptoms, as measured by informant report and clinical interview, were also uniquely predictive of target adjustment (b = −0.78, SE = 0.32, p < .05; b = −3.89, SE = 0.68, p < .001, respectively). Informant-reported paranoid, schizoid, antisocial, and borderline scores all negatively predicted partner dyadic adjustment, whereas obsessive–compulsive PD symptoms exhibited a positive relationship with that outcome. Finally, clinical interview of borderline and avoidant PD predicted higher AD8 scores, indicating increased risk of cognitive decline (b = 2.14, SE = 0.61, p < .001; b = 0.87, SE = 0.41, p < .05, respectively). Clinical interview of antisocial PD and self-reported narcissistic PD actually predicted less cognitive change (b = −2.93, SE = 1.12, p < .01; b = −0.88, SE = 0.30, p < .01, respectively). Obsessive–compulsive PD was uniquely predictive of higher scores on the AD8, as assessed by clinical interview and self-report (b = −0.70, SE = 0.33, p < .05; b = 0.61, SE = 0.22, p < .01, respectively). The only specific PD scale that provided a unique contribution to the prediction of AD8 scores from the informant MAPP was BPD, predicting higher levels of cognitive change (b = 0.96, SE = 0.26, p < .001).
Semistructured Interview, Self-, and Informant Reports of Personality Pathology Predicting Self- and Informant Report of General Health.
Note. SIDP = Structured Interview for DSM-IV Personality; MAPP = Multisource Assessment of Personality Pathology; HSI = Rand-36 Health Status Inventory; IHSI = Informant Health Status Inventory. Regression coefficient estimates based on final model with gender and race entered as covariates. Regression models use data from baseline.
N = 1,413. bN = 1,091.
p < .05. **p < .01. ***p < .001.
Semistructured Interview, Self-, and Informant Reports of Personality Pathology Predicting Target and Partner Dyadic Adjustment.
Note. SIDP = Structured Interview for DSM-IV Personality; MAPP = Multisource Assessment of Personality Pathology; DAS = Dyadic Adjustment Scale. Regression coefficient estimates based on final model with gender and race entered as covariates. Regression models use data from baseline.
N = 865. bN = 624.
p < .05. **p < .01. ***p < .001.
Semistructured Interview, Self-, and Informant Reports of Personality Pathology Predicting Informant Report of Cognitive Decline.
Note. SIDP = Structured Interview for DSM-IV Personality; MAPP = Multisource Assessment of Personality Pathology; AD8 = Ascertain Dementia Eight-Item Informant Questionnaire. Regression coefficients estimates based on final model with gender and race included as covariates. Regression model uses data from fifth follow-up. N = 261.
p < .05. **p < .01. ***p < .001.
Discussion
The current findings highlight the relative merits of various perspectives regarding the assessment of PDs in later life. Self-, informant, and interview measures of personality pathology contribute both unique and shared variance to the prediction of important outcomes, such as health, relationship satisfaction, and cognitive decline. In general, this is true whether one uses self- or informant report to measure outcomes in the target individual. Examining the incremental validity of self-reported PDs over and above semistructured interview showed that self-report aided the prediction of the participant’s health, dyadic adjustment, and cognitive decline. Notably, informant report of PDs also contributed unique variance to the prediction of these outcomes when accounting for both interview and self-report. Furthermore, only informant report, not self-report, of PDs contributed unique variance above semistructured interview to the prediction of the partner’s dyadic adjustment. Informant report of PD symptoms may be particularly important to obtain when examining the outcomes of close others, in addition to informant-reported outcomes in the target individual. As such, comprehensive assessment of PDs in older adulthood may necessitate the use of informant reports.
Our first outcome of interest was the participant’s subjective general health. We found that BPD symptoms as measured by all three methods of assessment appear to be uniquely associated with negative health outcomes when controlling for the other two methods. These findings persisted across both self- and informant report of the participant’s health, suggesting that it is not simply biased reporting by individuals who exhibit BPD traits. We also found that schizoid, paranoid, and avoidant PD symptoms were uniquely associated with worse health outcomes, whereas narcissistic PD had the opposite association. However, unlike BPD, none of these PDs showed consistent patterns across all three methods of assessment. An extensive literature links PD symptoms with worse health (see, Dixon-Gordon et al., 2015), yet a complete understanding of the exact mechanisms in play and possible targets for interventions is still in development. The current study suggests that future research would benefit from the use of self-, informant, and interview instruments, as each provides unique information regarding the link between PD symptoms and health.
Second, we were interested in examining the ability of different methods of PD assessment to predict dyadic adjustment. Both self- and informant reports of PDs were useful in predicting the participant’s satisfaction with their relationship. However, when predicting the partner’s satisfaction, our results indicate that self-report of PDs contributes no significant variance above interview. Conversely, informant reports do contribute substantially to the prediction of this outcome, accounting for nearly a third of the variance. This finding is consistent with the previous literature on marital satisfaction (Brock et al., 2016; South et al., 2008; Watson et al., 2000; Watson & Humrichouse, 2006). One limitation of the current study is that the direction of this relationship cannot be determined. It may be the result of bias in the informant report (i.e., using marital satisfaction as a heuristic for describing the personality of their partner; Watson et al., 2000), but it could also potentially indicate that some aspects of the participant’s perceived pathological personality characteristics are uniquely captured by the informant’s perspective. Regardless, the current study demonstrates that the relationship between the partner’s perception of the target’s personality and dyadic adjustment does not appear to fade in later life. Rather, it is as important as ever to assess the informant’s perception of the target’s PD traits when investigating the informant’s outcomes.
Our final set of analyses were concerned with the link between personality pathology and the onset of cognitive decline, which is a particularly important aspect of health in later life. Using an informant measure of cognitive decline, we found that interview-rated and informant-reported BPD symptoms were uniquely associated with a higher likelihood of experiencing cognitive change indicative of the early stages of nonspecific dementia. Interview and self-report of obsessive–compulsive PD symptoms were also uniquely predictive of higher cognitive impairment, whereas self-reported narcissistic PD and interview-rated antisocial PD symptoms were predictive of a lower likelihood of experiencing cognitive impairment. It is clear from these findings that maladaptive personality traits, in addition to adaptive-range traits, may be important to assess when examining the role of personality in predicting the onset of dementia.
One limitation of the present analyses is that our outcome measures were also self- or informant reports, and therefore shared method overlap with our measures of personality pathology. It will be important for future research to examine more “objective” outcomes, such as behavioral observations and biomarker data. Although we strove to ensure that our measurements of personality pathology were free from age bias, we were also limited in that we did not use age-specific measures of personality pathology. It may be that self- and informant report instruments developed for explicit use in an older adult context may capture additional variance in the various outcomes examined in the current analyses.
An additional concern is that some of our participants scored highly enough on the AD8 to indicate possible cases of mild cognitive decline. It could be the case that minor changes in cognitive functioning alter the ability of individuals to self-report their own personality, although this possibility has yet to be fully explored. Nevertheless, informant reports of personality should be unaffected by the changes in cognitive ability in the target individual, unless these changes are associated with observable changes in target personality. Another important limitation, although outside the scope of the present research question, is that our data cannot address the direction of the associations between PDs and the outcomes in question. Future studies may explicate the nature of the associations between the various sources of information about PDs and later life outcomes. Future research may also expand on these findings in a variety of other ways. For example, the current study used data provided by an informant nominated by the participant as someone who knew him or her well. It would be interesting to examine differences among various informants and then determine which types of informants (and perhaps how many informants) contribute the most to the assessment of PDs in later life. Overall, the findings of the current study provide incentive to continue pursuing ever more detailed assessment of this form of pathology in the context of older adulthood.
In conclusion, the results of the current analyses support the use of self- and informant-report measures of PDs in later life contexts. Previous studies have established the benefit of obtaining informant report of personality pathology. It is also clear that it is difficult to use comprehensive semistructured interviews in some clinical and research settings. Therefore, we agree with the suggestion that clinicians and scientists alike should incorporate the use of informant-report measures into their work whenever possible (Miller & Lynam, 2015; Samuel, 2015). This recommendation may be even more apt in older adult populations, especially in the presence of cognitive impairment (American Psychological Association, 2013b). These findings provide researchers and clinicians with the tools that they need to make informed choices regarding their methods when assessing PDs and their outcomes. If one is interested in an outcome obtained using informant report, or in the outcomes experienced by close others of those suspected of having high levels of PD symptoms, it may be most appropriate to obtain an informant report of that person’s symptoms. Although the difficulties with assessing PDs in later life remain, more comprehensive assessment procedures may provide the key to understanding the role personality pathology plays in later life.
Footnotes
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by grants from the National Institute of Mental Health (RO1-MH077840-01), the National Institute on Aging (R01-AG045231-01A1S1), and the National Institutes of Health (NIH 5 T32 AG000030-39).