How borderline personality disorder BPD is diagnosed

The "new" borderline personality disorder: Dimensional classification in the DSM-5 and ICD-11


The decades-long and increasing dissatisfaction with the previous categorical classification of personality disorders (PS) in the American DSM-IV and in the ICD-10 of the World Health Organization has, among other things. to do with a lack of empirical support for many categories, the very high comorbidity of PS among each other or the large heterogeneity of symptoms within a diagnosis. In the last revisions of the two diagnostic systems, it has supported a radical change towards a dimensional classification system that is much more secure in empirical psychological research. The revolution did not take place in the DSM-5 because the dimensional model remained in the appendix, while the old DSM-IV classification was adopted unchanged. However, this “Alternative Model of Personality Disorders” (AMPD) has received considerable research interest since its publication. In the ICD-11, a dimensional assessment of the personality in the 5 domains "Negative Affectivity", "Detachment", "Dissociality", "Disinhibition" and "Anankastia" is already approved as a radical alternative to the 10 previous categories and will be approved worldwide from 2022 change the diagnosis. While all previous categories were initially eliminated, critical voices ultimately achieved that the borderline disorder was retained as the only qualifier in the ICD-11. The two systems are examined in terms of their central assumptions and the practical procedure. The discussions and the previous empirical evidence on the dimensions and clinical usefulness for practitioners are summarized.


Decades of increasing discontent with current categorical classification systems of personality disorders (PD) in the American DSM-IV and the ICD-10 of the World Health Organization are based on missing empirical support of many of the categories, high comorbidity of PDs and large heterogeneity of symptoms within a diagnosis. This discontent has paved the way for a radical change in the last revisions of these diagnostic systems to replace a categorical with a dimensional model of classification, which is strongly supported by empirical psychological research. For the DSM-5 the revolution was canceled, as the dimensional model was placed in Section III “Emerging models” whereas the old DSM-IV categorical system was adopted without any changes. However, this "Alternative Model of Personality Disorders" (AMPD) has received much attention in research since its publication. In the ICD-11, a dimensional assessment of personality on five domains (“Negative Affectivity”, “Detachment”, “Dissociality”, “Disinhibition” and “Anankastia”) as a radical alternative to currently existing 10 categories is officially accepted and will change diagnosis of personality disorders worldwide from 2022. Whereas all current categories have been eliminated in the proposal, at first, critics were able to reclaim a “borderline pattern” as the one and only qualifier for a specific category in the final version of the ICD -11. Both systems are illustrated in their central assumptions and practical procedures. Discussions on dimensions, their current empirical basis and clinical utility are summarized.

The borderline personality disorder in the DSM-IV and ICD-10: problems with the categorical diagnostic system

The categorical diagnosis of personality disorders (PS) in the currently valid diagnostic systems has become very familiar to us over the past three decades. The currently valid ICD-10 of the World Health Organization (World Health Organization 1992) and the DSM-IV or DSM-5Footnote 1 of American psychiatry (American Psychiatric Association 1994, 2013) describe nine or ten specific PS diagnoses. They are diagnosed when general criteria of PS (a markedly different, lingering pattern of cognitions, affects, interpersonal behavior, or impulsiveness that is inflexible and profound and begins in adolescence or young adulthood) and a sufficient number out of a set of eight until ten criteria for a specific PS are met.

For borderline personality disorder (BPD), five out of nine or ten criteria must be met: desperate efforts to prevent actual or suspected abandonment, emotional instability, intensive and unstable interpersonal relationships with alternation between idealization and devaluation, identity disorder, threats of suicide, self-harm, Difficulty controlling anger, impulsiveness in potentially self-harming areas, chronic feelings of emptiness (Falkai and Wittchen et al. 2015; WHO 1992).

Did one actually describe naturally occurring categories with these diagnoses and thus separate nature at their joints ("carving nature by its joints"), if at the same time more than half of the people who meet the general criteria of a PS do not fit into any of these categories (Zimmerman et al. 2005)?

How different people with the same PS diagnosis can be, their heterogeneity, has often been critically noted (Ofrat et al. 2018; Widiger and Trull 2007). For BPS, there are 256 different symptom constellations in the DSM and with five out of nine criteria, two patients may only be able to present a single symptom overlapping. For compulsive PS with four out of eight criteria, it is even possible not to have any overlap with another person and still get the same diagnosis (Samuel and Griffin 2015). This is not only possible mathematically, but also shows itself in clinical reality. In a sample of 252 patients who met the criteria for BPD, 136 different combinations of BPD were present (Johansen et al. 2004).

In the categorical system, the boundary between normal and pathological personality is an arbitrary one (Ofrat et al. 2018). In the “revolutionary” DSM-III (APA 1980), which implemented this type of diagnosis for the first time, no justification was given for the threshold between normal and sick for most personality disorders (Demazeux 2015). In addition, the ultimately untenable position was assumed that all criteria are equally important for the diagnosis (Balsis et al. 2011). At the BPS, the DSM-III established the five of the eight criteria at the time, at least as an attempt to maximize the agreement among clinicians. This form of decision-making was humorously referred to as the BOGSAT method (“bunch of guys sitting around a table”; H. Pincus, quoted in Demazeux 2015, p. 9). The central (arbitrary) threshold for whether someone suffers from disease-related BPD is the criteria fulfilled from 4 (healthy) to 5 (sick) (Samuel and Griffin 2015). Some authors argue that the problem could at least be reduced if the number of criteria to be met were increased to at least 7 (Paris 2020). This stricter approach would allow a more homogeneous group of people to be diagnosed with BPD.

Also very problematic is the result of longitudinal studies that these criteria are stable over time to very different degrees and thus a postulated central characteristic of a personality disorder is clearly called into question. Grilo et al. (2004, 2014) have shown that almost 30% of a sample of BPD patients followed for only 24 months fulfilled only 2 or fewer of the criteria in this period of 12 months or more. So you were very far from a BPS diagnosis for over a year in these two years! For example, self-harm is much less stable over time than emotional instability. Conversely, years of self-harm, emotional instability and severe dissociations would not be sufficient for a diagnosis of BPD, even if they result in persistent severe impairment (Cooper and Balsis 2009; Samuel and Griffin 2015).

Tyrer et al. (2019) summarize that the general criteria of a personality disorder (as a prerequisite for the diagnosis of a specific category) have simply been ignored (see also First et al. 2014). When clinicians made diagnoses, the spectrum used was reduced to the three categories of BPS, Antisocial PS, and Combined PS / Unspecified PS. And they emphasize that studies have only given 8% of psychiatric patients a PS diagnosis, although empirical evidence has shown that the real numbers are between 40% and 90%.

These and other criticisms of the categorical diagnosis of personality disorders have increasingly focused attention on dimensional models as a more appropriate alternative (Hopwood et al. 2018; Smith et al. 2020). Here, features are not viewed dichotomously as “present” or “not present”. Instead, continuous dimensions (traits) from mild to extreme are identified, which are differently pronounced in patients (Krueger et al. 2014). On the one hand, this approach corresponds to the results of neurobiology, which are much more compatible with dimensional models than with categorical models (Caspi et al. 2014). It also corresponds to the dominant models in normal psychology of personality. An extensive integration of personality disorders with the leading, empirically very well-established “Big Five” model of the normal personality (with the traits “Neuroticism”, “Extraversion”, “Openness”, “Compatibility” and “Conscientiousness”) was first published in 1994 (Costa and Widiger 1994; Widiger and Costa 2013; Miller and Widiger 2020). A trait was defined as a “variable on which a relatively stable disposition for certain behavior patterns is based” (Markon and Jonas 2015, p. 64, German translation HM).

The Dimensional Alternative Model of Personality Disorders (AMPD) in the DSM-5: the Fast Revolution

In the preparation of the DSM-5, the Personality & Personality Disorder Work Group (PPDWG), under the direction of Andrew Skodol, clearly worked towards the replacement of categorical diagnostics with a dimensional model of PS. The first concrete form (Skodol et al. 2011) envisaged the deletion of five empirically least reliable categories (paranoid, schizoid, histrionic, narcissistic and dependent PS) and described the remaining PS as prototypes on a grid of dimensions (Krueger et al . 2014). The personality of patients should be assessed dimensionally on 37 facets (e.g. emotional lability, fear of separation, withdrawal, hostility etc .; Skodol et al. 2011) and this assessment should be independent of the previous "drawers" (antisocial, borderline etc.) provide a differentiated profile of the person with which the treatment can be planned. The solution was a "hybrid model" of prototypes on personality dimensions, which was supposed to combine dimensional and categorical diagnosis and thus all, i.e. H. Critics and advocates of the old model, should satisfy and facilitate the transition (Krueger and Markon 2014).

The model triggered a large number of critical to negative reactions. Even Widiger (2011), as a long-standing proven representative of a dimensional approach, criticized the PPDWG for not using the secured Big Five model as the basis for the DSM-5 (Huprich 2015). One consequence of this violent criticism was the resumption or retention of the narcissistic PS (another example of the BOGSAT method?) and recent studies that resulted in a system of 5 trait domains with a total of 25 facets. Recording instruments for these facets now exist both as self-assessment, third-party assessment by relatives and as clinical assessment (PID-5; Personality Inventory for DSM-5; First et al. 2018, Krueger et al. 2012; Somma et al. 2019). The PPDWG now largely followed the Big Five model. “Openness” is not represented, while a domain “psychoticism” has been introduced instead, in order to, among other things. to be able to depict the symptoms of a schizotypic personality disorder (which is not a PS as a schizotype in the ICD-10; Fig. 1). Psychoticism, however, has repeatedly been discussed as a possible extreme variant of the “openness” trait (cf. Chmielewski et al. 2014).

Much more prominent in DSM-IV than in DSM-IV is the assessment of criterion A, the assessment of the severity of the impairment of personality (Morey and Bender 2014). Disorders of self-functions (identity and self-control) and interpersonal functions (empathy and closeness) are differentiated and assessed on a continuum (Tab. 1; cf. Widiger et al. 2019 and comments for an interesting discussion). For the first time, the operationalization of the “healthy” personality is also included. Clinicians use a “scale for measuring the functional level of personality” (SEFP) to assess the severity of the impairment from “0 = no or minor impairment” to “4 = extreme impairment” for the four areas mentioned. For the diagnosis of a personality disorder, at least two out of four areas must be rated as at least “2 = moderate”. To illustrate this, Tab. 1 shows the “moderate impairment” for the four areas (Falkai et al. 2015).

If the diagnosis of a personality disorder is justified by the severity of the impairment, criterion B describes the content of the predominant traits. The domains “negative affectivity”, “closedness”, “antagonism”, “disinhibition” and “psychoticism” are described, which are again differentiated into 25 facets (in order to avoid redundancies, only the domains of the more important for Europe are shown in Table 3 ICD-11 shown in more detail, see below). For example, negative affect can be divided into the sometimes heterogeneous facets of "emotional lability", "fearfulness", "separation anxiety", "submissiveness", "hostility", "perseveration", "depression", "mistrust" and (absence of) "lack of affect" differentiated (Falkai et al. 2015).

What does the diagnosis of BPS look like in this system? The hybrid model (dimensional and categorical) in the DSM-5 becomes clear when the criteria for the existence of a BPS are to be assessed as one of six possible prototypes. We are very familiar with this form of assessment (Tab. 2):

The BPS is thus described on the new dimensions, which on the one hand depict the severity of the impairment of the personality (self and interpersonal relationships), on the other hand those facets are named that are relevant for the diagnosis in the respective trait domains. Here there are four of seven characteristics that must be fulfilled.

Six of the previous 10 types were included in the hybrid model (antisocial, avoidant-self-insecure, borderline, narcissistic, compulsive, schizotype). If none of these prototypes are appropriate, a diagnosis of personality disorder, trait-specific (PS-MS) is possible. It corresponds most closely to a purely dimensional model: (1) general assessment of the severity of the personality disorder (criterion A; at least two out of four areas at least moderate) and (2) specific assessment of the five traits (criterion B; negativity, reticence, antagonism , Disinhibition and Psychoticism; Falkai et al. 2015, p. 1058).

The degree of severity (criterion A) and traits (criterion B) should ideally be independent of each other, which is unlikely to be empirically sustainable (Bastiaansen et al. 2016; Berghuis et al. 2014; Zimmermann et al. 2015). On the contrary, the total number of PS criteria met (A and B together as opposed to A alone) best reflects the current and future stresses in work, leisure and relationships (Hopwood et al. 2011). In addition to the specific factors, a kind of general factor emerges on which all Load criteria (A and B). Interestingly, the criteria of BPS seem to load on this general factor rather than on specific traits (Sharp et al. 2015; Wright et al. 2016). Some authors therefore criticize the fact that narcissistic or antisocial PS can very well be viewed as ego-syntonic extreme variants of traits, which, however, does not seem to be the case for BPS. A number of BPD symptoms cannot be mapped well on this system. These symptoms will also be experienced as severe ego-dystonic (e.g. chronic suicidality, depersonalization, pseudo-hallucinations; Paris 2020).

Specific application examples of the AMPD in assessment and treatment planning can be found e.g. B. in Bach et al. (2015), Pincus et al. (2016) or Waugh et al. (2017).

Self-assessment tools developed for the model for the clinical recording of the degree of severity (Levels of Personality Functioning Scale, Morey 2017; short version LPFS-BF, Bach and Hutsebaut 2018) and the trait domains (Personality Inventory for DSM ‑ 5 - PID ‑ 5 and short versions; Krueger et al. 2012; Somma et al. 2019) are currently being intensively investigated, but have not yet been standardized. The American Psychiatric Association has made these versions of PID available for research and clinical evaluation free of charge and in German on the Internet. Summarizing representations of measuring instruments for both diagnostic systems can be found in Hengartner et al. (2018), Zimmermann et al. (2019) and Evans et al. (2020).

Hopwood (2018) also tried for the first time to illustrate the use of AMPD for the selection of therapeutic techniques beyond individual case vignettes. A major criticism of categorical models was the lack of evidence-based treatment measures for most personality disorders with the exception of BPD (Cristea et al. 2017). Dimensional models should provide a much more meaningful (and ultimately evidence-based) basis for the choice of treatment methods. Hopwood (2018) has presented a very preliminary model, which at first glance may seem unusual and a bit angular. But we will perhaps (have to) get used to finding manuals with titles such as “Therapy of Negative Affectivity”, “Therapy of Seclusion” or “Therapy of Disinhibition” in the future. Even if it remains a very exciting discussion which aspects of a (temporally stable) trait are accessible to change through therapy (Ringwald et al. 2020), reviews show that interventions have an effect on traits (Roberts et al. 2017), and the first Guidelines and suggestions for the clinical handling of various traits have already been formulated (Bach and Presnall-Shvorin 2020).

What the Board of Trustees of the American Psychiatric Association rejected in full for DSM-5 and banned into Part III ("Emerging models"), while Part II still contains the old categorical diagnostics (Krueger and Markon 2014) , has been fully implemented in ICD-11 and has been approved by the WHO since May 2019. When the ICD-11 becomes the valid diagnostic system from 2022, the way in which personality disorders are diagnosed will change fundamentally around the world. This ultimately makes the official diagnosis of PS in the USA fundamentally inconsistent with the diagnosis made in the rest of the world (Bagby and Widiger 2020).

Paradigm Shift in the ICD-11: Dimensional Classification of Personality Disorders

Work on the diagnosis of personality disorders began in 2010 before the publication of the DSM-5 (Tyrer et al. 2019). Right from the start, the working group rejected a hybrid model as in DSM-5 and instead defined a single dimension for the severity of the PS from “normal” to “severe”. Impairment was measured in terms of the degree of interpersonal dysfunction, the consequences for social and professional roles, cognitive and emotional experiences and the risk of harming oneself or others. The main reason for the outstanding emphasis on a single dimension for the severity of PS was the previous high comorbidity of PS diagnoses, the arbitrary threshold between "healthy" and "sick" and the poor coverage of the symptoms of patients in the 10 categories ( Ofrat et al. 2018; see above). In addition, it was shown time and again that the number of PS diagnoses also increased with increasing severity. For the sake of simplicity, clinicians therefore (possibly additionally) selected the PS in complex cases (Tyrer et al. 2019).

Bach and First (2018) describe the application of the new model in clinical practice. The first step here is to check whether the general requirements for diagnosing a personality disorder have been met. What is required is a persistent disorder, characterized by problems in the functional level of aspects of the self (e.g. identity, self-esteem, accuracy of self-perception, self-control) and / or interpersonal dysfunction (e.g. the ability to establish close and mutually satisfactory relationships that Understanding the other's perspective and managing conflicts).

This disorder no longer needs to start in adolescence or young adulthood, it just needs to last longer than 2 years. This means that for the first time, PS that start at the age of 50 are also possible! If no abnormalities are recognizable before the age of 25, a qualifier “late start” is planned (Tyrer et al. 2015).

The disorder manifests itself in cognitive patterns, emotional experience and expression and is recognizable through various personal and social situations. It cannot be explained by development or by social or cultural factors (including political conflicts). It is not an expression of an illness, a drug effect or a withdrawal and causes considerable personal, family, social and professional stress.

In a second central step, the severity of a PS is assessed. Practitioners also have no other choice here to come to a diagnosis, because the 10 categories of the ICD-10 can no longer be assigned. This considerably reduces the problem of comorbidity and only evaluates from “mild” (6D10.0) to “moderate” (6D10.1) to “severe” (6D10.2). In addition, “Severity unspecified” (6D10.Z) and a subclinical category “Difficulty” can be assigned (Fig. 2). So there are anomalies here that cause some distress, but are not so clear that they justify the diagnosis of a disorder (Tyrer et al. 2015). The "Standardized Assessment of Personality Disorder" (SASPD; Olajide et al. 2018) developed for this purpose is a very short self-assessment tool with only 9 items to record the degree of severity (see also Bach and Anderson 2020 for a comparison of the instruments for the DSM-5 and the ICD-11).

In a third step, a person's personality style can be described. The traits in the ICD-11 largely correspond to the DSM-5 domains. Originally the four traits “Negative Affectivity” (6D11.0), “Closed-backness” (6D11.1), “Dissociality” (6D11.2) and “Compulsiveness (Anankastia)” (6D11.4) were proposed. The fifth dimension “disinhibition” (6D11.3) was added later (Tyrer et al. 2019; Tab. 3). The independence of the last two traits is, however, questioned. Empirical results indicate that obsession and disinhibition could be two opposite poles of the same dimension (Bach et al. 2020; Oltmanns and Widiger 2020).

This description of the style of personality is not necessary for making a diagnosis (in contrast to DSM-5; Tab. 2)! It is likely that as the severity of the PS increases, the number of conspicuous trait domains that contribute to the specific expression of the dysfunction also increases (Tyrer et al. 2019). But it is easily possible and permissible for a heavy PS to have only one or no domain prominently in the foreground. The domains are understood as continuous dimensions, not categories. Nevertheless, for the purpose of coding, the decision “available” vs. “not available” must be made (Bach and First 2018). In their illustration of the procedure, the authors describe five different patients, one of whom is shown to be severely impaired by borderline symptoms. The impairment was assessed as "severe". Of the trait domains, (1) negative affectivity, (2) dissociality and (3) disinhibition were rated as “present”, i. H. a severe personality disorder characterized by many negative emotions, such as shame, anger and mistrust, disinhibition as impulsive and harmful reactions to immediate stimuli and dissociality as suspicious aggression and tendency to manipulate others. This is how the dimensional description of a BPS could look like in the future.

While a clearly differentiated level of 25 facets was described in the AMPD in addition to the domains (Krueger and Markon 2014), the ICD-11 Work Group deliberately avoided facets in order to keep clinical application as simple as possible (Tyrer et al. 2011; Fig . 2). However, Oltmanns and Widiger (2020) z. B. for the domain "negative affectivity" clearly argued that many of the previous PS categories contain negative emotions (narcissistic, antisocial, insecure, compulsive, dependent, borderline PS) and a distinction between affects (externalizing: anger, distrust vs. Internalizing: anxiety, depression etc .; see also Smith et al. 2020) would contribute significantly to the clinical usefulness of the system. Therefore, only 2 years after the publication of their “Personality Inventory for ICD-11” (PiCD; Oltmanns and Widiger 2018), they published the “Five-Factor Personality Inventory for ICD-11” (FFiCD), an extension of the trait Domains recorded at the facet level (a total of 20 facets, 7 for negative affectivity, 3 each for dissociality, reservedness and compulsiveness, and 4 for disinhibition), most of which could also be confirmed in a factor analysis (Oltmanns and Widiger 2020). This extension is not relevant for a diagnosis according to ICD-11, but should prove to be clinically useful and relevant to treatment. Tyrer et al. (2019) emphasize that the proposed system without facets is sufficiently complex if, with three degrees of severity (plus the "difficulty" as a subclinical phenomenon) and five trait dimensions, almost 500 different diagnostic results can be achieved, as well as the BPS in the system can be mapped very well.

The acceptance of this procedure in the ICD-11 (as well as in the AMPD) will largely depend on whether and in what form treatment decisions will be facilitated in the future by assessing the severity or the domains (cf. Bach and Presnall-Shvorin 2020). The postulated DSM-5 and ICD-11 trait domains seem to be able to be empirically confirmed again and again and have also been replicated in different cultures (Bach et al. 2017; Lotfi et al. 2018; Pires et al. 2017). But there are also many critical voices, such as Paris (2020), which welcomes the development in principle, but sees decades of research in danger or even lost. This concerns v. a. the BPS, for which by far most of the research literature was created within the specific PS.

And that's why critics of the model (Herpertz et al. 2017) achieved here (and only here) at the BPS that it is retained as a "borderline pattern" in the ICD-11. The critics argued that a diagnostic system should have clinical usefulness beyond the scientific criteria and thus provide helpful information on the choice of treatments, prognosis and outcomes. It should help patients to be able to give informed consent or to be able to decide which treatments seem most helpful to them. It should help to use resources in the health system sensibly and should be supported by robust scientific results. The authors did not see these aspects really fulfilled in the proposal for the ICD-11 (Herpertz et al. 2017). These critics also fundamentally support a dimensional approach, but demand a hybrid model as in DSM-5 in order to guarantee a gentler transition.

Tyrer et al. (2019