Abstract
Personality disorders (PDs) in the Diagnostic and Statistical Manual of Mental Disorders (DSM-5) are conceptualized as distinct clinical syndromes. However, debate persists about the clinical utility of this categorical model, with many researchers supporting a dimensional model that focuses on pathological personality traits and personality dysfunction. This model was published in Section III of DSM-5 and named the Alternative Model of Personality Disorders (AMPD). This study evaluated the AMPD by examining relationships between traits and dysfunction with traditional categorical PD constructs among older adults. Older adults (N = 202) completed the Personality Inventory for DSM-5, Levels of Personality Functioning Scale-Self-Report, and Coolidge Axis II Inventory. Results indicated that pathological personality traits do not relate to categorical PDs in directions predicted by the AMPD. Personality functioning related to categorical PDs in expected theoretical patterns according to the AMPD but lacked incremental validity above pathological personality traits. An implication of these findings is that the AMPD does not fully resolve the age-related issues with the traditional categorical PD model.
Historically, personality disorders (PDs) have been conceptualized according to a system in which an individual must meet certain criteria for a particular PD to receive a formal diagnosis. This system is known as the categorical model and has been officially endorsed in the current Diagnostic and Statistical Manual of Mental Disorders (DSM-5; American Psychiatric Association, 2013). Under this categorical model, ten discrete PDs were provided. However, several issues with this categorical model have been outlined, including high rates of comorbidity and heterogeneity (Krueger et al., 2014). Consequently, individuals often present in drastically different ways, despite sharing the same clinical PD diagnosis. It is difficult to generate new and effective therapies for PDs if clients significantly vary from one another, making the current diagnostic system especially problematic.
Because of these known problems, DSM-5 included two different models of PDs. In Section II, the categorical model of PDs was officially endorsed, upholding the same model from the DSM-IV. However, in Section III, the portion of the DSM-5 devoted to promoting future research, a different model was proposed. The alternative model of personality disorders (AMPD) conceptualizes PD symptoms as maladaptive extensions of normative traits. The AMPD requires assessment on two purportedly distinct dimensions: personality functioning (Criterion A) and pathological personality traits (Criterion B). The AMPD also provides proposed diagnostic criteria for six specific PDs (antisocial, avoidant, borderline, narcissistic, obsessive-compulsive, and schizotypal) in which each PD is given a unique profile of personality functioning (Criterion A) and pathological personality traits (Criterion B).
Criterion A of the AMPD
Research has indicated that generalized severity is an important predictor of the presence and expression of personality psychopathology, with studies finding consistent general impairments across all PDs (Bender et al., 2011). Subsequently, the DSM-5 included a personality functioning criterion in its Section III model and defines levels of personality functioning as distributed on a continuum, ranging from adaptive to maladaptive. Personality functioning is further broken down into two domains, each of which has two components: self (composed of identity and self-direction) and interpersonal (composed of empathy and intimacy). Each PD purportedly has its own typical style of impairments, and the DSM-5 Section III provides proposed criteria for six DSM-5 Section II PDs. The AMPD provides the Levels of Personality Functioning Scale (LPFS; a clinician-administered assessment) as a measurement of personality functioning.
Thus far, research supporting Criterion A has been mixed. Some support the clinical utility of personality functioning domains specified by the AMPD (Buer Christensen et al., 2020), whereas others suggest that the domains specified by the AMPD do not provide any additional clinical utility above and beyond assessing for general personality dysfunction (Sleep et al., 2019). Additionally, the incremental validity of personality functioning has been challenged, with some suggesting that it does not provide value above and beyond only measuring pathological personality traits (Criterion B) when assessed at only one time point (Few et al., 2012; Sleep et al., 2019). Subsequently, additional research is needed to clarify the clinical utility of Criterion A.
Criterion B of the AMPD
In addition to personality functioning, the AMPD also requires clinicians to assess for pathological personality traits. It defines personality traits as dimensional, placing maladaptive ones on the extreme end of the spectrum from normative features. The AMPD also endorses a hierarchical structure of personality wherein 5 larger domains, which describe broad personality features, are composed of 25 smaller facets that describe more detailed and specific behaviors and cognitions. The AMPD suggests assessing clients on both domains and facets to gain the most detailed picture. Each PD is hypothesized to have its own typical style of personality trait expression, and the AMPD proposed trait criteria for six DSM-5 Section II PDs. The Personality Inventory for DSM-5 (PID-5) was released alongside the AMPD as a self-report questionnaire for Criterion B.
Thus far, research has largely supported the clinical utility of the AMPD’s pathological personality trait model and the validity of the PID-5. Its five-factor structure has been widely replicated, but its lower-order structure has been debated, with many facets loading onto multiple domains and some failing to load on the domain they were designed to assess. This has led some researchers to suggest changes in the lower-order structure of the PID-5 (Watters et al., 2019). Other studies have examined the overlap between the PID-5 and Section II PD diagnoses, specifically examining the extent to which the six proposed trait PD models align with categorical diagnoses. Results have largely supported the proposed trait models and suggest that Criterion B can adequately capture the 10 traditional PD diagnoses (Fossati et al., 2013; Hopwood et al., 2012).
AMPD and Older Adults
Given that the proportion of older adults in the population will continue to grow in the coming decades, PDs in older adults are an increasingly important area of research. However, despite the recognized need for research, empirical evidence in this area is lacking, and core issues are still debated. The age neutrality of the DSM-IV/DSM-5 Section II PD diagnostic criteria has been challenged for a long period of time because some criteria seem to target a younger person’s experience while ignoring the context of old age. For instance, Segal et al. (2006) question the validity of Criterion 8 for dependent PD (is unrealistically preoccupied with fears of being left to take care of himself or herself). For some frail older adults with numerous medical problems, it would be entirely expected for them to be concerned about being left to care for themselves when they physically may not be capable. Balsis et al. (2007) estimate that one-third of the PD diagnostic criteria are inappropriate for older adults. This is a significant number of criteria that could impact whether an accurate PD diagnosis is given to an individual in later life. Despite the documented need for changes in PD criteria and the urging of van Alphen et al. (2015) for a geriatric subclassification, the DSM-5 Section II model retained the problematic criteria with no changes. However, Section III presents a promising alternative that may be more adaptable for use among older adults. The AMPD was designed to create diagnoses that are more flexible and personalized to individual clients. For instance, a PD diagnosis can be generated based on a specific trait profile displayed by a client, rather than criteria that are specifically based on younger adults (American Psychiatric Association, 2013). Personality trait profiles have been found to be highly related to both successful and nonresilient aging trajectories (Baek et al., 2016; da Rosa et al., in press). Subsequently, a model that incorporates the already-established utility of personality traits could possibly be more relevant to older adults and more sensitive to age-related changes. That being said, the AMPD does not contain geriatric variants in any of its diagnostic criteria, which could be a significant limitation.
Only a few studies have examined the AMPD in later life. For example, Van den Broeck et al. (2013) identified 33 items of the PID-5 that lack age neutrality, and Debast et al. (2018) identified that 25% of the PID-5-Brief Form was age biased against older adults. These studies suggest that the psychometric properties of the PID-5 are problematic in some respects and should be further scrutinized closely for use with older adults. Van den Broeck et al. (2014) is the only study to have examined the overlap between the PID-5 and previous measures of personality pathology, as measured by the Dimensional Assessment of Personality Pathology, in an attempt to establish construct validity among older adults. Factor analysis confirmed the five-factor structure of the PID-5, and substantial overlap was found in predictable patterns, suggesting that it has evidence for validity among individuals in later life. Cruitt et al. (2019) is the only study that has examined the AMPD’s specific conceptualization of personality functioning among older adults. They conducted life story interviews of older adults and coded the interviews for the AMPD’s personality functioning. Their study provided support for use of Criterion A among older adults using an interview measure of personality functioning, but research regarding the use of self-report measures of personality functioning among older adults is still lacking.
The Present Study
Overall, the existing evidence provides moderate support for the AMPD among older adults, indicating that Criteria A and B have some validity. However, Criterion A, as measured by self-report questionnaires, and its incremental validity have been yet to be examined, leaving significant research left to be done. Additionally, the PID-5 and measures of personality functioning have yet to be compared in combination to previous measures of Section II PDs in older adults, which would help to determine if the AMPD solves some PD measurement problems for older adults. In general, measurement and understanding of PDs is lacking in later life, and the present study aims to aid in understanding the basic nature of PDs among older adults. The purpose of this study is to evaluate the AMPD by examining relationships between personality functioning and pathological personality among older adults and comparing the AMPD to the Section II PD model among a later life population. Hypotheses for this study are:
There will be significant positive correlations between the ten Section II PDs and pathological personality traits (domains and facets) in directions suggested by the AMPD (see Table 1 for a listing of these directions). As only six Section II PDs are outlined in the AMPD, the hypotheses for the remaining four PDs (Schizoid, Histrionic, Paranoid, and Dependent) are based on a combination of prior research and impressions of face validity. For example, Dependent PD features should be theoretically related to the traits of Separation Insecurity, Submissiveness, and Anxiousness, and it is logical to hypothesize that Paranoid PD would be significantly related to Suspiciousness, Perceptual Dysregulation, and Unusual Beliefs and Experiences.
The PID-5 facets will be significantly predictive of the Section II PD scales in directions provided by the AMPD. See Table 1 for a full listing of subhypotheses based on the AMPD.
Overall personality functioning, as measured by the LPFS-SR, will be positively associated with the ten Section II PDs and the five PID-5 domains.
Personality functioning will show incremental validity above and beyond PID-5 pathological personality traits in predicting Section II PDs. Additionally, pathological personality traits will show incremental validity above and beyond personality functioning in predicting Section II PDs.
Predictions for Hypotheses 1 and 2 Based on the AMPD.
Note. AMPD = alternative model of personality disorders. *Indicates PDs that are not outlined in the AMPD; hypotheses for these PDs are based on prior research and impressions of face validity.
Method
Participants and Procedure
Participants consisted of older adults (N = 202) who were recrutied online via Amazon’s Mechanical Turk (MTurk). The average age was 67.46 years (SD = 2.98 years; range = 60–89 years), and the sample contained 79 women (39%) and 120 men (59%). Regarding ethnicity, 171 participants identified as not Hispanic or Latino (84%), 26 identified as Hispanic or Latino (13%), and 5 individuals wished to not identify themselves (3%). The distribution of race is as follows: 70% White, 16% Black or African American, 6% East Asian, 3% American Indian or Alaska Native, 1% Native Hawaiian or other Pacific Islander, 1% Multiracial/Multiethnic, 0.5% West Asian, 1.5% identified as Other, and 2% individuals did not wish to answer. The average years of education was 15.36 years (SD = 2.54 years). Approval to conduct the study was gained from the university’s Institutional Review Board, and informed consent was obtained before participants gained access to the online survey. The general MTurk sample was used, and any individuals who reported an age lower than 60 years were excluded from the sample. Additional exclusion criteria included living outside of the United States and not being proficient in English. Data collection occurred in July 2019. On average, it took participants 45 min to complete the study, and they were compensated $2.00 for their participation. Four attention/validity checks were included throughout survey, which asked participants to respond to items that are highly unlikely to be true (e.g., I was a member of the French Foreign Legion). A total of 26 participants were excluded from the sample because they failed one or more attention checks. All participants anonymously completed a series of counterbalanced self-report questionnaires, including the PID-5, Levels of Personality Functioning Scale-Self-Report (LPFS-SR), and Coolidge Axis II Inventory (CATI).
Measures
Personality Inventory for DSM-5
The PID-5 (Krueger et al., 2012) is a 220-item self-report measure that is designed to assess Criterion B (pathological personality traits) of the AMPD. The measure assesses 25 facets that belong to 5 overarching domains of Negative Affect, Detachment, Antagonism, Disinhibition, and Psychoticism. The PID-5 uses a 4-point Likert-type scale, ranging from 0 (very false or often false) to 3 (very true or often true), with higher score representing higher presence of that trait. Internal consistency of scale scores for the PID-5 is strong for the current older adult sample, with the domains’ Cronbach’s α scores all high: .93 for Negative Affect and Detachment, .94 for Antagonism, .95 for Disinhibition, and .97 for (Psychoticism). For the facets, Cronbach’s α was adequate to strong, ranging from .68 (Suspiciousness) to .95 (Depressivity).
Levels of Personality Functioning Scale–Self-Report
The LPFS-SR (Morey, 2017) is an 80-item self-report measure that assesses Criterion A (personality functioning), of the DSM-5 AMPD. The questionnaire assesses 4 domains (Identity, Self-Direction, Empathy, and Intimacy), and a total functioning score can also be computed. It is assessed on a 4-point Likert-type scale, ranging from 1 (Totally False, not at all True) to 4 (Very True), with higher scores indicating higher levels of dysfunction. Internal consistency of scale scores (Cronbach’s α) is high for the present sample: Empathy = .84, Intimacy = .89, Identity = .89, Self-Direction = .90, and total score = .97.
Coolidge Axis II Inventory
The CATI (Coolidge, 2020) is a 250-item self-report measure that assesses the ten traditional PDs. Participants answer questions on a 4-point Likert-type scale ranging from 1 (strongly false) to 4 (strongly true), and higher scores indicate a great likelihood of psychopathology. Internal consistency for PD scale scores is adequate in the present sample with Cronbach’s α ranging from .73 (Paranoid PD) to .91 (Antisocial PD). Additionally, the CATI has been widely used among diverse older adult populations (Edelstein et al., 2008).
Results
First, relationships between chronological age and gender with the PID-5 traits (Table 2) and with the LPFS-SR total score (Table 3) were examined by computing correlations. Regarding age, 4 of the 5 PID-5 domains (all except Detachment) were significantly and negatively correlated with chronological age, and 16 of the 25 PID-5 facets were significantly and negatively correlated with chronological age, with small to medium effect sizes. Regarding gender, all 5 PID-5 domains and all 25 PID-5 facets were significantly and positively correlated with gender (with medium effect sizes), indicating that increasing presence of pathological personality traits is associated with male gender. The LPFS-SR total score was significantly negatively correlated with chronological age (r = −.19, p < .001) but significantly positively correlated with gender (r = .27, p < .001), with small to medium effect sizes.
Correlations Between PID-5 Traits with CATI PD Scales and Age/Gender.
Note. AN = Antisocial; AV = Avoidant; BO = Borderline; CATI = Coolidge Axis II Inventory; DE = Dependent; OC = Obsessive-Compulsive; PA = Paranoid; PD = personality disorder; PID-5 = Personality Inventory for DSM-5; SZ = Schizoid; ST = Schizotypal. All correlations between PID-5 Traits and CATI PD scales are significant at p < .01. *Indicates p < .05. **Indicates p < .01. ***Indicates p < .001. Bold represents a correlation greater than |.30|. aCorrelation is point biserial r (r pb).
Correlations Between Personality Functioning Total Score with PID-5 Domains, CATI PD Scales, and Age/Gender.
Note. CATI = Coolidge Axis II Inventory; PID-5 = Personality Inventory for DSM-5. All correlations are significant at p < .001. aCorrelation is point biserial r (r pb).
Hypothesis 1
Overall, this hypothesis was partially supported in that all predicted correlations between the CATI PD scales and the PID-5 domains and facets were significant, positive, and at least a moderate effect size (>.30). However, all of the remaining, unpredicted PID-5 domains and facets also significantly correlated with the CATI PD scales at a significance level of p <.01 with medium to large effect sizes. See Table 2 for a full correlation matrix.
Hypothesis 2
Next, a series of hierarchical regression analyses were conducted, with gender and the PID-5 facets predicting the PD scales. Because of an unequal gender distribution in the sample and significant correlations with gender, gender was added as the sole predictor at Step 1 of each regression model. Then, gender and the 25 PID-5 facets were added as predictors at Step 2, thus controlling for the effects of gender. Because Hypothesis 2 is only concerned with the variance accounted for by the PID-5 facets, the change in R 2 from Model 1 (gender) to Model 2 (gender and PID-5 facets) will be interpreted. Overall, across the ten separate regression analyses, the PID-5 facets accounted for significant variance in the CATI PD scales, ranging from 64% (Avoidant PD) to 71% (Obsessive-Compulsive PD). Gender was a nonsignificant predictor in nine of the ten PD models, excluding the Dependent PD model. See Table 4 for a full table of regression coefficients.
Regression Coefficients with PID-5 Facets Predicting Each CATI PD Scale.
Note. AN = Antisocial; AV = Avoidant; BO = Borderline; CATI = Coolidge Axis II Inventory; DE = Dependent; HI = Histrionic; NA = Narcissistic; OC = Obsessive-Compulsive; PA = Paranoid; PD = personality disorder; PID-5 = Personality Inventory for DSM-5; SZ = Schizoid; ST = Schizotypal. Each PD scale was part of a hierarchical regression in which gender was controlled for at Step 1. Bold represents a correlation greater than |.30|. *p < .05. **p < .01, ***p < .001.
Antisocial PD
This hypothesis was partially supported. The PID-5 facets predicted 70% of the variance in the Antisocial PD scale, F(26, 171) =29.89, p < .001. As predicted, Callousness (β = .38) and Impulsivity (β = .16) were significantly predictive of the Antisocial PD scale. However, Manipulativeness (β = .10), Deceitfulness (β = .12), Hostility (β = .06), and Irresponsibility (β = .03) were not significant predictors, which is inconsistent with the AMPD. Additionally, Restricted Affectivity (β = −.14), Suspiciousness (β = .12), and Perceptual Dysregulation (β = .24) were also significant predictors of Antisocial PD.
Avoidant PD
The PID-5 facets predicted 64% of the variance in the Avoidant PD scale, F(26, 171) = 12.14, p < .001. This hypothesis was partially supported. Consistent with hypothesis, Anxiousness (β = .42, p < .001) and Withdrawal (β = .50) were significant predictors. However, Anhedonia (β = .16) and Intimacy Avoidance (β = .13) were not significant predictors. Although it was not predicted, Hostility (β = −.30), Suspiciousness (β = .14), and Unusual Beliefs and Experiences (β = −.28) were significantly predictive of the Avoidant PD scale.
Borderline PD
The PID-5 facets predicted 75% of the variance in the Borderline PD scale, F(26, 171) = 25.18, p < .001. This hypothesis was slightly supported. Emotional Lability (β = .18) was significantly predictive. However, Anxiousness (β = .11), Separation Insecurity (β = .03), Depressivity (β = −.16), Impulsivity (β = .11), Risk Taking (β = .01), and Hostility (β = .10) were not significant predictors of Borderline PD. Although not hypothesized, Restricted Affectivity (β = −.16), Irresponsibility (β = .28), and Rigid Perfectionism (β = −.14) were also significant predictors.
Narcissistic PD
The PID-5 facets predicted 70% of the variance in the Narcissistic PD scale, F(26, 171) = 20.29, p < .001. This hypothesis was half supported in that Grandiosity was a significant predictor (β = .28), but Attention Seeking was not (β = .06). There were also multiple other PID-5 facets that were significantly predictive: Anxiousness (β = .25), Emotional Lability (β = .17), Separation Insecurity (β = −.20), Depressivity (β = −.30), Suspiciousness (β = .30), Manipulativeness (β = .30), Irresponsibility (β = .27), and Withdrawal (β = −.14).
Obsessive-Compulsive PD
The PID-5 facets predicted 71% of the variance in the Obsessive-Compulsive PD scale, F(26, 171) = 22.62, p < .001. This hypothesis was not supported. Rigid Perfectionism (β = .001), Perseveration (β = .09), Intimacy Avoidance (β = .08), and Restricted Affectivity (β = .09) were not significant predictors of the Obsessive-Compulsive PD scale. However, there were a few significant predictors that were not predicted that had medium effect sizes: Anxiousness (β = .20), Depressivity (β = −.25), and Irresponsibility (β = .23).
Schizotypal PD
The PID-5 facets predicted 70% of the variance in the Schizotypal PD scale, F(26, 171) = 22.09, p < .001. This hypothesis was partially supported in that Eccentricity (β = .27) and Suspiciousness (β = .21) were significant predictors. However, Perceptual Dysregulation (β = .23), Unusual Beliefs and Experiences (β = .07), Restricted Affectivity (β = .03), and Withdrawal (β = .13) were not significantly predictive of the Schizotypal PD scale. There were also two significant predictors that were not part of this hypothesis, including Depressivity (β = −.24). Anxiousness (β = .19), and Attention Seeking (β = −.22).
Schizoid PD
The PID-5 facets predicted 69% of the variance in the Schizoid PD scale, F(26, 171) = 22.73, p < .001. This hypothesis was partially supported. Intimacy Avoidance (β = .15) and Withdrawal (β = .45) were the only significant predictors. Anhedonia (β = .13), Callousness (β = .008), and Restricted Affectivity (β = .04) were not significant predictors of the Schizoid PD scale.
Histrionic PD
The PID-5 facets predicted 65% of the variance in the Histrionic PD scale, F(26, 171) = 13.55, p < .001. This hypothesis was partially supported in that Emotional Lability (β = .25) and Manipulativeness (β = .23) were significant predictors. Attention Seeking (β = .15), Separation Insecurity (β = −.09), Impulsivity (β = .15), and Grandiosity (β = .07) were not significantly predictive of the Histrionic PD scale. Three PID-5 facets were significantly predictive of the Histrionic PD scale that were not predicted: Depressivity (β = −.26), Withdrawal (β = −.32), and Irresponsibility (β = .37).
Paranoid PD
The PID-5 facets predicted 68% of the variance in the Paranoid PD scale, F(26, 171) = 19.59, p < .001. This hypothesis was somewhat supported in that Suspiciousness was significantly predictive (β = .42). However, Unusual Beliefs and Experiences (β = .05), Perceptual Dysregulation (β = .06), Intimacy Avoidance (β = .04), and Withdrawal (β = .03) were not significant predictors. Although not hypothesized, Anxiousness (β = .25) and Depressivity (β = −.27) were significantly predictive of the Paranoid PD scale.
Dependent PD
Finally, the PID-5 facets predicted 74% of the variance in the Dependent PD scale, F(26, 171) = 21.97, p < .001. This hypothesis was mostly supported. Anxiousness (β = .26), Separation Insecurity (β = .17), and Submissiveness (β = .12) were significant predictors. However, Attention Seeking (β = .09) and Restricted Affectivity (β = −.01) were not significantly predictive. There was one unpredicted scale that was also significantly predictive: Anhedonia (β = .20). Finally, gender was a significant predictor of the Dependent PD scale (β = −.09).
Hypothesis 3
Overall, this hypothesis was supported. Personality functioning, as measured by the LPFS-SR total score (in which greater scores represent greater dysfunction), was significantly and positively correlated with every Section II categorical PD scale at large effect sizes, with correlation coefficients ranging from .52 (Avoidant PD) to .82 (Borderline, Dependent, and Obsessive-Compulsive PDs). Additionally, the LPFS-SR total score significantly and positively associated with every PID-5 domain, with values ranging from .79 (Detachment) to .87 (Disinhibition and Psychoticism). See Table 3 for a full correlation matrix.
Hypothesis 4
To examine incremental validity, LPFS-SR domains were first regressed onto the CATI’s PD scales. All models were significant, with R 2 ranging from .28 (Avoidant PD) to .69 (Dependent and Obsessive-Compulsive PDs). Next, LPFS-SR and PID-5 domains were regressed onto the CATI PD scales. The change in R 2 was significant for every PD scale once adding in the PID-5 domains, with values ranging from .06 (Obsessive-Compulsive PD) to .27 (Dependent PD). See Table 5 for full regression results. Overall, these analyses indicated that the PID-5 showed incremental validity above and beyond the LPFS-SR in predicting Section II PDs.
Incremental Validity of PID-5 Traits and LPFS-SR Functioning.
Note. CATI = Coolidge Axis II Inventory; LPFS-SR = Levels of Personality Functioning Scale–Self-Report; PD = personality disorder; PID-5 = Personality Inventory for DSM-5. *p < .05, **p < .001.
Next, the PID-5 domains were regressed onto the CATI PD scales. All models were significant, and R 2 ranged from .53 (Avoidant) to .74 (Antisocial and Obsessive-Compulsive PDs). Next, the PID-5 and LPFS-SR domains were regressed onto the CATI PD scales. Six of the ten models had significant changes in R 2 once adding in the LPFS-SR, but the changes were small, ranging from .02 (Histrionic) and .04 (Dependent and Paranoid PDs). The change in R 2 for the remaining four PDs were not significant. See Table 5 for full regression results, which indicate that the LPFS-SR did now show incremental validity above and beyond the PID-5 in predicting Section II PDs.
Discussion
This study evaluated the validity of the AMPD among older adults by examining its two diagnostic components and their relationships to the ten traditional PD constructs. Incremental validity of the two components was also investigated. Despite the growing proportion of older adults relative to the population, PDs in later life are an understudied topic and relatively little is known about the potentially unique expression of personality pathology. The extant research indicates that the traditional approach to PD classification fails to consider the unique biopsychosocial context of later life, leading age-biased diagnostic criteria (Balsis et al., 2007; Segal et al., 2006). Subsequently, examining PDs and new approaches to PD diagnosis (i.e., the AMPD) are needed steps. Overall, results from the present study provided some support for the AMPD’s use among an older adult sample, but findings also indicate that more research is needed to revise the model to reflect the unique context of later life.
AMPD’s Relationship to Age and Gender
Correlations between the PID-5, LPFS-SR total score, age, and gender were computed to determine the effect of basic individual difference variables on PDs in later life. Results indicated that age is negatively correlated with the majority of PID-5 scales and with the LPFS-SR. It is possible that some features of personality pathology decline with age, but it is also likely that the AMPD does not adequately reflect or consider the later life experience. Research indicates that some personality change occurs in older adulthood, including less overt anger or emotional lability and different interpersonal ways of relating to others (Magai, 1999). Both of these constructs appear in the AMPD’s conceptualization of PDs, and it is possible that negative correlations found in this study reflect such personality changes. However, the AMPD also may not reflect other ways in which older adults with PDs display the dysfunction differently (e.g., overly relying on others). Additionally, significant gender correlations were detected, with male gender being associated with greater personality pathology. These findings indicate that individual difference variables have an effect on the AMPD, despite the fact that such differences are not outlined or accounted for in the model.
Hypothesis 1: Simple Relationships Between Criterion B and Section II PD Model
Correlations between the PID-5 domains and facets and the CATI PD scales showed strong and significant associations between every personality trait and every PD scale. These correlational results are partially consistent with previous literature in that a greater number of significant correlations are found between Section II and III diagnoses than what is predicted by the AMPD. Two prior studies also reported correlations between Section II PDs and PID-5 traits, and although more significant associations were found than the AMPD predicts, there were substantially fewer correlations at a moderate to large effect size than what was found in the present study. For example, among undergraduate students, Anderson et al. (2014) found five PID-5 traits to significantly correlate with Schizoid PD at a modest effect size, whereas the present study found all 25 traits and 5 domains to significantly correlate with the Schizoid PD scale at mostly large effect size. Similarly, Orbons et al. (2019) used a Dutch clinical sample (aged 19–60 years) and reported both predicted and nonpredicted correlations between Section II PDs and PID-5 traits, but correlations were not found between every trait and Section II PD, such as in the present study. This is consistent with one previous study that examined the PID-5 and other constructs of personality pathology among a community sample of older adults, finding more overlap than expected (Stone et al., in press). Combined with the present findings, this indicates that the PID-5 has good specificity but lacks sensitivity in a later life sample, and that personality pathology in later life, as assessed by the AMPD, may represent more general dysfunction, rather than trait-specific impairment.
Hypothesis 2: Pathological Personality Traits Predicting Section II PDs
Next, a series of hierarchical regressions were computed with the PID-5 traits predicting the CATI PD scales while controlling for gender. Across PD scales, results were mixed. For each PD, the traits predicted by the AMPD partially emerged, with some indicating they are significant predictors of the PD that they are intended to be and others indicating they are not significant predictors. Additionally, for each PD, additional traits were found to be significant predictors, even though they are not predicted by the AMPD. Table 6 presents a visual description of the number of hypothesized traits that were significant and not significant for each PD scale and the number of traits that were not hypothesized but significant. Of the ten regression models, eight (Antisocial, Avoidant, Borderline, Histrionic, Narcissistic, Obsessive-Compulsive, Paranoid, and Schizotypal) had more nonpredicted traits emerge than predicted traits, suggesting poor concordance. Two models (Dependent and Schizoid) had more predicted traits emerge as significant than nonpredicted traits. These findings add only partial support for the AMPD, as the specific pathological personality traits that are intended to represent the ten PDs slightly replicated among this older adult sample, and many unintended traits emerged as significant predictors. It should be noted that six of the ten predicted PD models were based on the AMPD’s proposed trait criteria. Because the AMPD does not provide trait profiles for them, the remaining four PD models (Dependent, Histrionic, Paranoid, and Schizoid) were based on prior research and impressions of face validity. Despite not being specifically included in the AMPD, these four models came out similar to the six specifically provided by the AMPD.
Number of Predicted and Not Predicted PID-5 Traits That Emerged When Regressed Onto the CATI PD Scales.
Note. CATI = Coolidge Axis II Inventory; PD = personality disorder; PID-5 = Personality Inventory for DSM-5.
Findings from our regressions differ from previous research that has been conducted among predominantly younger adults. Anderson et al. (2014), Fossati et al. (2013), Hopwood et al. (2012), and Orbons et al. (2019) also found that the trait models did not emerge exactly as the AMPD predicts. However, across studies, they instead found fewer relationships than the AMPD states (i.e., one to three per PD, instead of four to seven), but the relationships that were detected were in predicted directions with predicted traits. None of the prior studies identified as many nonpredicted but significant traits that were found in the present study. This is consistent with results from Hypothesis 1 and indicate that the AMPD lacks sensitivity on the pathological trait level. From what can be gathered from the documentation provided by the PD workgroup (e.g., Krueger et al., 2012), older adults were not considered any more in DSM-5 than in previous iterations of the DSM. Subsequently, it is not surprising that pathological trait patterns among older adults appear to be different than among younger or middle-aged adults. Future research would benefit from identifying consistent patterns in which pathological personality traits differ in older adulthood, potentially leading to gero-specific trait patterns.
Hypothesis 3: Simple Relationship Between Criterion A, Criterion B, and Section II PDs
To examine relationships between personality functioning, pathological personality traits, and Section II PDs, correlations were computed between the LPFS-SR total dysfunction score with the PID-5 domains and CATI PD scales. Results showed that the LPFS-SR total score (in which a greater score indicates greater dysfunction) was significantly and positively correlated with every PID-5 domain and every CATI PD scale with large effect sizes. To date, only one study (Hopwood et al., 2018) has validated the LPFS-SR (among a sample with an average age of 35 years). The present findings are consistent with those of Hopwood et al. (2018) in that strong associations were detected between the LPFS-SR, the PID-5 domains, and Section II PDs. Subsequently, these findings suggest that the LPFS-SR performs similarly well among both adults and older adults. The LPFS-SR was designed to assess personality dysfunction that is related to all variants of PDs and personality pathology. Subsequently, the high associations found between both Section II and III models and the LPFS-SR total score make logical sense and suggest that personality dysfunction is a broader measure of impairment. Our results support further use and evaluation of the LPFS-SR with other samples of older adults to firmly establish this measure’s utility for older adults.
Hypothesis 4: Incremental Validity of Personality Functioning
In this hypothesis, the incremental validity of the PID-5 and LPFS-SR was examined. Results indicated that pathological personality traits (Criterion B) had incremental validity above and beyond personality functioning (Criterion A) in predicting Section II PDs. However, personality functioning did not show incremental validity above and beyond pathological personality traits. These results are consistent with studies using undergraduate and community samples that examined Criteria A and B of the AMPD at one time point (Bastiaansen et al., 2016; Few et al., 2012; Sleep et al., 2019). This lack of incremental validity raises questions about the clinical utility of personality functioning and about the validity of the AMPD, as its only two diagnostic criteria appear to measure the same underlying concept.
In theory, the construct of personality functioning has clinical utility for older adults. Later life is a period that has a unique biopsychosocial context. Experiencing death of friends and family, medical complications, and potential cognitive and sensory decline, among other stressors, has the potential to cause lifelong pathological personality traits to become increasingly dysfunctional (Segal et al., 2006). However, the present results indicate that the AMPD’s concept of personality functioning may be functionally measuring the same concept as pathological personality traits: more stable, trait-like concepts of inner and outer experience. Subsequently, it may not adequately capture the sensitive period of older adulthood in which biopsychosocial factors may cause lifelong traits to become dysfunctional. Future research would benefit from revising concepts of personality functioning and creating markers of personality severity that better capture the unique context of later life.
Limitations and Future Directions
Some limitations of the present study should be noted. Notably, the sample included community-dwelling older adults, resulting in overall lower levels of pathology than what would be found in clinical or treatment-seeking samples. Future studies should examine the validity of the AMPD among diverse treatment-seeking and psychiatrically impaired older adults. The study was also conducted online with a sample of older adults with an average educational level of about 15 years, limiting the current finding to computer-literate and more highly educated older adults. When compared to nationally representative samples, older adults recruited via MTurk are also more likely to have higher cognitive functioning, self-rated health, and depressive symptoms, meaning the current sample may also be limited in these ways (Ogletree & Katz, in press). Additionally, the sample primarily consisted of White older adults (70%), so one must exercise caution in generalizing results to more diverse late life populations. PD and pathology rates vary by race, so it would be important to extend this research into a more racially and ethnically diverse sample (Chavira et al., 2003). Furthermore, a large number of analyses were conducted in this study, potentially leading to questions of increased risk of Type I error, although our focus on effect sizes mitigates this potential problem. Finally, due to our cross-sectional design we are not able to draw conclusions about causality among the variables. Future studies should seek to use multimethod or multi-occasion data to determine more complex and directional relationships among personality pathology models.
One primary future direction for this type of research would be to use longitudinal methods. Longitudinal studies of PDs that include older adults are lacking, and it is largely unknown how manifestations of personality pathology change over time and into older adulthood, despite theorized geriatric variants (Segal et al., 2006). Unfortunately, neither Section II nor Section III of DSM-5 includes any information about age-specific manifestations of PD pathology. Future research should focus on longitudinally tracking pathological personality traits and personality functioning over time to determine the specific trajectories and changes in expression for each of the PDs into later life. It would be interesting to determine if traits and functioning can adequately capture these changes.
In conclusion, the Section II model of PDs has been criticized for the significant comorbidity and heterogeneity that result from categorical diagnosis, so the AMPD was proposed as a dimensional alternative intended to address these issues. Additionally, previous research has indicated that the Section II model of PDs is problematic when applied to older adults, including age-biased diagnostic criteria. Subsequently, the AMPD provided some promise in that it may be able to better adapt to the unique context of later life. Overall, results indicate that personality functioning (Criterion A) related to categorical PD diagnoses in expected directions. However, results also indicated that personality functioning, as conceptualized by the AMPD, lacks incremental validity above and beyond assessing pathological personality traits. Subsequently, additional research is needed to better distinguish between pathological personality severity and style. Additionally, pathological personality traits (Criterion B) do not relate to Section II PDs in directions explicitly predicted by the DSM-5, meaning that the prototypical trait profiles provided by the AMPD may not be valid for older adults. It is possible that personality pathology in later life is better expressed by general personality dysfunction, rather than trait-specific profiles. This poses a problem for the AMPD that heavily relies on identifying specific trait impairments for diagnosis. The present study suggests that the AMPD may be useful in some ways but that more research is needed to clarify the validity and clinical utility of applying trait-based profiles specifically to older adults. Future studies should seek to address these issues, which may help to clarify the exact nature of personality pathology in later life to inform future revisions in the DSM. We urge the developers of the AMPD to consider the context of later life more fully in revisions of the model.
Footnotes
Author’s Note
Lisa E. Stone and Daniel L. Segal, Department of Psychology, University of Colorado at Colorado Springs.
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) received no financial support for the research, authorship, and/or publication of this article.
Author Biographies