An Overview of the DSM-5 Alternative Model of Personality Disorders (AMPD)

Abstract

Authoritative classification systems for psychopathology such as the DSM and ICD are shifting toward more dimensional approaches in the area of Personality Disorders (PD). In this paper, we provide a brief overview of the dimensionally oriented DSM-5 Alternative Model of PDs (AMPD). Since its publication in 2013, the AMPD has inspired a substantial literature, underlining its generative influence on the field. Generally speaking, this literature illustrates both the reliability and validity of the constructs delineated in the AMPD. The literature also illustrates empirical challenges to the conceptual clarity of the AMPD, such as evidence of substantial correlations between indices of personality functioning (Criterion A in the AMPD) and maladaptive personality traits (Criterion B in the AMPD). Key future directions pertain to linking the AMPD literature with applied efforts to improve the lives of persons who suffer from PDs, and surmounting challenges germane to the evolution of the DSM itself.

With regard to psychiatric nosology, we live in quite interesting times. Prior to the neo-Kraepelinian revolution of the 1970s, psychiatric nosology was arguably somewhat of an intellectual backwater [1]. Early 20^th century efforts to classify psychopathology were largely derived from the need to keep official records, as opposed to the need to delineate a conceptual model to help direct mental health assessment and intervention. Arguably, the watershed moment that brought nosology into the limelight was the publication of DSM-III in 1980.

DSM-III was innovative in a variety of ways that are generally well known, such as the introduction of structured criteria for specific mental disorder diagnoses, which represented a notable change from the narrative approach of earlier DSMs. At that time, the multiaxial system of DSM-III was seen as an innovation that would help highlight the importance of categorical Personality Disorders (PDs) in characterizing a clinical case. DSM-III separated personality disorders from other clinical disorders by placing PDs on Axis II and other disorders on Axis I. This was done with the intent of highlighting the importance of personality in clinical case conceptualization, as noted by the chairman of the DSM-IV task force, Alan Frances [2]. Frances [3] subsequently noted that a shift to a more dimensional approach to PDs was inevitable; the question was not whether this would occur, but rather, when it would occur and what dimensional approach it would incorporate.

Consistent with Frances’ prescient perspective, in the post-DSM-III era, PDs became the poster child for the conceptual and practical problems that result from trying to characterize complex clinical presentations using relatively inflexible categorical rubrics. Given the multifaceted nature of human personality, problems such as comorbidity (multiple PD labels being truly applicable to the same patient) and within category heterogeneity (notable variation in key clinical features within putatively cohesive categories) were inevitable. Although these problems were well recognized prior to and during efforts to transition from DSM-IV to DSM-5 [4], solving these problems proved to be fraught with political and practical complexities that arose during the construction of DSM-5 [5]. Also, somewhat ironically from a historical perspective, DSM-5 deleted the multiaxial approach introduced in DSM-III.

The net result of this transition was that DSM-5 contains two approaches to conceptualizing and diagnosing PD, the model from DSM-IV reprinted, and the model developed by the DSM-5 PD workgroup, entitled the Alternative Model of Personality Disorders (AMPD). In this brief overview paper, our goals are to describe the AMPD, to give the reader a general sense of the voluminous research literature that has accumulated surrounding the AMPD since its publication in 2013, and to identify what we see as some key directions at the intersection of PD, authoritative classification systems such as the DSM, and clinical intervention.

The Content and Structure of the AMPD

The AMPD’s general criteria for PD consist of seven individual criteria, A through G. The first, Criterion A, refers to personality functioning. In the AMPD, the sine qua non of PD is evidence that the patient encounters difficulties with personality functioning, where functioning is understood as impairment in the sense of self and corresponding impairment in interpersonal functioning [6–8]. Criterion B refers to pathological personality traits. These traits are organized around five broad domains, Negative Affectivity, Detachment, Antagonism, Disinhibition, and Psychoticism. These domains are generally well understood as maladaptive variants of the domains of the Five Factor Model of normative personality structure (FFM [9]). In the AMPD, PD is defined as the combination of clinically significant problems in functioning along with one or more pathological traits. In addition to these two primary criteria, Criteria C and D refer to inflexibility and stability across time, respectively. Criteria E, F, and G refer to ensuring that the PD is not better explained by “another mental disorder” (E), the effects of a substance or a medical condition (F) and is not normative for the persons developmental stage or sociocultural environment (G).

The AMPD also recreates DSM-IV PD constructs, working within the AMPD’s framework. Combinations of functioning and traits are used to recreate Antisocial, Avoidant, Borderline, Narcissistic, Obsessive-Compulsive, and Schizotypal PD, where the target for these reconstructions was the DSM-IV descriptions of these PDs. The intent was to show how specific DSM-IV PD types could be understood as specific combinations of personality functioning and pathological traits, as opposed to categorical symptom lists. Finally, PD Not Otherwise Specified (PD-NOS, a common DSM-IV diagnosis [10]) is replaced in the AMPD by the concept of PD trait specified. Similar to PD-NOS the idea of PD trait specified is to capture presentations that do not fit neatly into specified types. PD trait specified allows the clinician to describe the PD in terms of a specified pattern of functioning and traits. The main difference from PD-NOS is that the clinical characteristics of the patient are specified (using concepts from the AMPD), as opposed to being “not otherwise specified,” which should be advantageous in terms of clinical conceptualization and treatment selection.

The Empirical Literature on the AMPD

The AMPD has inspired a complex and vast reaching literature, which tends to coalesce around the organization of the AMPD itself. That is, studies have tended to focus on the operationalization, reliability and validity of Criterion A and Criterion B, respectively. In addition, some studies have tackled the complex issue of associations between Criterion A and B, in other words these studies have examined associations between functioning and traits.

The AMPD literature was reviewed very recently by Zimmerman et al. [11] The intent of the Zimmermann et al. [11] paper was to provide a comprehensive review; indeed, 318 citations are provided. Rather than duplicate this thorough, recent, and comprehensive review, here, we summarize the basic conclusions reached by Zimmermann et al. [11], organized by research on Criterion A, B, and the overlap between Criteria A and B.

Criterion A: Personality Functioning

To aid the clinician in conceptualizing and assessing Criterion A of the AMPD, the DSM-5 provides a Levels of Personality Functioning Scale (LPFS). Self-functioning in the LPFS is subdivided into two aspects, Identity and Self-direction. Interpersonal functioning is subdivided into two aspects, Empathy and Intimacy. Each level of each of these four aspects of personality functioning is described along a five-point scale ranging from little or no impairment to extreme impairment [12 p775].

The LPFS has been operationalized and studied in a variety of ways, including self-report and other-report instruments, as well as structured interviews. Interestingly, inter-rater reliability tends to be modest but non-trivial (ICCs in the .4 to .6 range), even when raters are clinically inexperienced, and, perhaps unsurprisingly, training sessions and structured interview approaches appear to increase reported reliability coefficients [11]. This evidence is a bit surprising because the LPFS describes concepts that require some degree of psychological inference on the part of the rater (e.g., rating concepts such as a distorted self-image). Internal consistency reliability of the constructs is generally good and varies understandably by number of indicators (e.g., indices of Criterion A overall, containing numerous items, tend to show higher internal consistency compared with narrower assessments containing fewer items [11]).

Regarding validity, convergent validity of criterion A measures has been documented for a variety of clinically meaningful correlates, ranging from those that are conceptually very similar (e.g., indices of PD severity from a psychodynamic perspective) to those that are more distal, but still of clear clinical importance (e.g., treatment dropout [11]). Structural validity discussion in the Criterion A literature tends to focus on the organizational structure described in DSM-5. That is, there is debate regarding the extent to which Criterion A indices form a general dimension of personality functioning subdivided first into self and interpersonal aspects, and then further subdivided into identity and self-direction (under self) and empathy and intimacy (under interpersonal). In general, this exact proposed structure is not readily reproduced, perhaps because of the presence of a strong general factor uniting all elements of Criterion A [13]. Indeed, this general PD factor may represent elements common to all psychopathology, broadly conceived (as opposed to PD specifically) [14]. In addition, performance-based assessment measures (e.g., the Social Cognition and Object Relations scale [15]; SCORS) may provide some unique value in assessing Criterion A constructs (as opposed to self-report questionnaires [16]), and should be studied more frequently in efforts to clarify the assessment and structure of Criterion A.

Criterion B: Personality Traits

Criterion B of the AMPD refers to personality style and is conceptualized in terms of the five aforementioned trait domains of Negative Affectivity, Detachment, Antagonism, Disinhibition, and Psychoticism. In the AMPD, these domains are further fleshed out with 25 facets. For example, the Psychoticism domain is indicated by facets labeled Unusual beliefs and experiences, Eccentricity, and Cognitive and perceptual dysregulation.

Generally speaking, AMPD Criterion B tends to be operationalized by using the Personality Inventory for DSM-5 (PID-5) [17]. The PID-5 has been heavily researched; as of this writing, Google Scholar lists 1208 citations to the paper where this measure was introduced [17]. In the self-report PID-5 literature, facet and domain scales tend to show reasonable to good internal consistency reliability, and the factor structure of the 25 facets tends to delineate the five broad trait domains listed in DSM-5 (see meta-analyses reported in Somma et al. [18]; Watters & Bagby [19]), in English and in translations of this measure [19]. There is also an extensive literature on the validity of the PID-5 showing, for example, that DSM-IV PD features can be captured by the PID-5 domains [11].

Relations Between Criterion A and Criterion B

One key issue regarding empirical evaluation of the AMPD concerns notable correlations between Criterion A and Criterion B constructs. These two elements of the AMPD reflect distinguishable traditions in PD scholarship. Criterion A reflects clinical inferences about psychological mechanisms that go awry in PD patients (pertaining to coherent conceptualization of self and other), whereas Criterion B reflects the tradition of generating psychometric instruments designed to assess PD trait characteristics from the perspective of patient reports. In addition, forms of assessment other than self-report (e.g., informant and clinician reports) are applicable to both Criterion A and B concepts, although the literature remains tilted toward studies that rely on self-reports [20].

Empirically, at least based on the extant literature, Criterion A and B assessments are highly correlated, and debating the meaning of these correlations forms a key focus of the contemporary AMPD literature [11]. From our perspective, it is easy to see both sides of this debate. On the one hand, recognizing these distinguishable traditions in the PD literature by creatively synthesizing them in the AMPD represents a key opportunity for intellectual rapprochement [21]. On the other hand, the AMPD is indeed a complex model, and could be considerably simplified by somehow concatenating features described in Criterion A and B. These issues remain contested and are a focus of recent discussions in the literature [13,22,23].

Key Future Directions for the AMPD and Dimensional PD Models

Up to this point, we have focused on the DSM-5’s AMPD. In addition to the DSM, however, the other influential authoritative PD classification system is the International Classification of Disease (ICD), published by the World Health Organization (WHO). As of this writing, the ICD-11 will become the official ICD classification system on January 1, 2022 [24]. The PD classification system that is part of the ICD-11 is a significant departure from both the previous ICD model, and from categorical approaches to PD delineation [25]. Indeed, stated in terms of recent DSM models, the ICD-11 PD model is notably more similar to the AMPD than to the DSM-IV PD model.

Like the AMPD, the ICD-11 PD model is dimensional, with a focus on conceptualizing PDs in terms of overall severity (akin to Criterion A in the AMPD) and style (akin to Criterion B in the AMPD). If a PD might be applicable to the clinical presentation, the first step is to identify the overall degree of personality difficulty on an ordinal scale (with gradations of subthreshold, mild, moderate, and severe). One option is to stop there (concluding simply that a PD is present to a specific degree), but also has the option to further describe the features of the case, using specifiers akin to the domains of the AMPD. Specifically, the ICD-11 model describes trait domains of negative affectivity, detachment, dissociality (akin to AMPD antagonism), disinhibition, and anankastia (i.e., obsessive–compulsive features). Psychoticism is omitted not because it is conceptually irrelevant to describing personality, but rather, because psychotic phenomena are described in a separate part of the ICD. The ICD-11 PD model also allows the clinician to specify a borderline pattern, to provide some continuity with classical PD constructs [26].

The ICD-11 PD model is in some ways even further of a departure from the traditional PD typological model of e.g., DSM-IV, compared with the AMPD. It essentially requires the diagnostician to make a dimensional (ordinal) determination of the degree of personality difficulty, in order to render any type of PD diagnosis. The general situation then, with regard to PD nosology in authoritative classification systems, entails movement toward dimensional approaches, and away from classical PD categories. To our way of thinking, this makes scientific and practical clinical sense (see e.g., the discussion on clinical utility of the AMPD provided by Waugh et al. [21]). Nevertheless, the paradigm switch creates a notable disjunction with the extant PD treatment literature, and this disconnect underlines both challenges and opportunities.

Linking Developments in PD Research with Intervention Approaches

Intervention approaches for PD have historically centered around the Borderline PD concept. Indeed, this is a major reason this concept was added to the ICD-11 PD model, i.e. to provide some continuity with existing approaches [25]. Comparatively less attention has been paid to the need to develop interventions applicable to other classical PDs.

There are likely numerous reasons that this situation evolved, in which the majority of treatment-oriented effort is directed at one of the 10 DSM-IV categorical PDs. Borderline PD is a very serious and debilitating condition, and efforts to intervene that are tied to this concept are to be celebrated and encouraged. Moreover, Borderline PD criteria delineate the quintessence of PD broadly conceived. From the perspective of the AMPD, BPD criteria emphasize difficulties in self-other conceptualization and representation (Criterion A), which result in extensive manifest distress (e.g., qualities described the Negative Affect domain of Criterion B). From this perspective, in transitioning to newer dimensional models of PD, it is critical to both recognize successes with PD intervention to date, and to capitalize on opportunities to expand the armamentarium of effective prevention and intervention strategies to encompass the greater breadth of difficulties covered by AMPD criterion B. Encouraging developments along these lines can be seen in the work of Livesley and colleagues [27] on integrated approaches to PD treatment, and Kernberg’s [28] recent creative synthesis of individual differences such as temperament and intelligence with levels of personality organization that help to frame case conceptualization. Indeed, we would venture that most seasoned clinicians naturally develop multidimensional and nuanced case conceptualizations that are not readily reduced to single category labels.

The question is how to continue to pursue this progression successfully given political dynamics that naturally come into play during periods of paradigm transition. Consider, for example, some of the dynamics that have occurred in the context of the NIMH’s Research Domain Criteria (RDoC) project. Briefly, RDoC is an effort on the part of the NIMH to encourage researchers to focus more on understanding systems that may be relevant to psychopathology (with a particular focus on neuroscience), as opposed to having their efforts constrained by thinking of classical psychiatric categories as the primary targets of empirical inquiry [29]. Pickersgill [30] recently provided a fascinating narrative account of some of the controversy in the field surrounding RDoC, in which he interviewed a series of mental health opinion leaders and summarized their perspectives and comments about RDoC. Although opinions expressed and conveyed by Pickersgill [30] were not monolithic, many are well summarized by the following quote, where “Tom” is Thomas Insel, the NIMH director who promoted RDoC during his tenure:

I can tell you that since Tom has decided to leave nearly everyone that I’ve talked to has said something that they didn’t feel they could say when he was still there which was that he has done more damage to clinical psychiatry than can be repaired in five or ten years. [30 p628]

Clearly, many prominent psychiatric opinion leaders reacted badly to the idea that NIMH would essentially discard cherished categorical psychiatric diagnostic rubrics in favor of the RDoC approach. Returning then to the topic of PD intervention, in the context of the AMPD, a solution may lie in viewing the multidimensional nature of case conceptualization in the AMPD framework as an opportunity to demonstrate the usefulness of a dimensional model (cf. Hopwood [31]). The acute, distressing, and life-threatening behaviors associated with Borderline PD deserve the attention they have received, but also, most PD patients do not fit neatly into categorical rubrics. This leaves other characteristics (beyond those that form the understandable foci of Borderline PD intervention) relatively less well conceptualized from an intervention angle. For example, emotional lability (a key Criterion B facet) combined with risk taking (a separate Criterion B facet that tends to vary relatively independently of emotional lability) is a different presentation from emotional lability combined with rigid perfectionism (a third Criterion B facet). What is needed are case conceptualization flow charts linked to the multidimensional nature of PD variation. Reducing dimensional information to singular labels (e.g., “high on detachment”) may be as problematic for reliability and case conceptualization as classical PD labels. We see the development of integrated multi-dimensional case conceptualization and intervention as a major opportunity in connecting developments in PD nosology with intervention efforts [32].

Evolving the DSM

Efforts by the DSM-5 PD workgroup resulted in the AMPD, which, by any standard, has been generative of extensive discussion and empirical investigation. Clearly, there is great scholarly and scientific interest in evolving PD nosologies. Interestingly, the vast majority of the AMPD literature is in psychology journals, and not in psychiatry journals (a fact that is readily gleaned from perusing the citations given in Zimmermann et al. [11]). Perhaps this is unsurprising because psychiatry is often a conservative enterprise, for understandable reasons, such as a perceived need to preserve the tradition of the medical model as a basis for ensuring putative legitimacy. For example, it may be easier to add entirely new diagnostic categories to the DSM with only modest evidentiary support (e.g., Disruptive Mood Dysregulation Disorder [33]) than to challenge underlying assumptions (i.e., that legitimate psychiatric constructs are categorical by fiat). Other potential reasons might pertain to a preference for clinical observation as the basis for construct building (as opposed to patient reports), and the perceived clinical utility of classical category labels.

Steve Hyman, a former director of NIMH, describes this situation as “repairing a plane while it is still flying.” [34] In essence, patients and families present needing help on a daily basis and asking them to wait until we have the “perfect nosological system” is obviously impossible. From the perspective of medical authority, it is better to respond by reassuring consumers that extant approaches are valid. Nevertheless, we need to continue to provide ethical assessment and intervention, as we also strive to improve our conceptual models and corresponding intervention approaches.

From this perspective, the dynamics in play throughout the DSM-5 construction process are still active in the field. The field is slowly shifting toward more dimensional approaches to PD conceptualization. How will the DSM revision process correspond with this ongoing shift? One possibility is that the DSM continues to ossify and thereby comes to be seen as essentially irrelevant to both research and practice. Clinicians on the front lines probably do not use the DSM as intended [35], and from that perspective, the DSM being out of sync with the needs of front-line clinicians may not really constitute much of a change. Indeed, there is evidence that front-line clinicians prefer AMPD concepts (Criterion B in particular) to DSM-IV PD categories in numerous respects [36]. In the research arena, the literature is now replete with concepts and approaches that are fundamentally at odds with foundational categorical assumptions of recent DSMs, such as the NIMH’s dimensionally oriented RDoC project [29] and the Hierarchal Taxonomy of Psychopathology (HiTOP) [37,38]. Although each of these projects (the AMPD, RDoC, and HiTOP) have distinct origins and foci, they generally converge on remarkably similar broad dimensional domains for organizing the conceptualization of major individual differences relevant to psychopathology. If the American Psychiatric Association seeks to keep the DSM relevant, then the DSM will likely need to evolve along with the needs of both clinicians and researchers. Attending to these very real needs for evolution and change may be more important than seeking to promote and maintain historical psychiatric orthodoxy.