The aim of this study was to investigate the role of personality functioning (PF) within the HiTOP framework of psychopathology using ambulatory assessed longitudinal data over two years in a sample of N = 27 173 mental health app users. We conducted a bass-ackwards analysis that yielded a hierarchical taxonomy of psychopathological symptoms, traits, and PF (research question 1), which we subsequently used for latent modeling of general and specific component variance (research question 2) and longitudinal prediction (research question 3). Using a very large sample with repeated measurements, this study achieves an unprecedented level of measurement and estimation accuracy [50] with respect to answering the present research questions.
Locating Personality Functioning in a Hierarchical Dimensional Structure of Psychopathology
In our sample, we replicated a hierarchical dimensional structure that largely aligns with the HiTOP model [3]. However, we identified a distinct PF component that was indicative of internalizing, externalizing, and general psychopathology. Our findings also support previous evidence on higher-order constructs, including an internalizing spectrum with subfactors of distress and fear, a thought disorder component with psychotic symptoms and schizotypal traits, and an externalizing spectrum with antagonistic and disinhibited components. We observed minor discrepancies regarding the location of eating disorders and detachment.
Most notably, 10 of the 11 PF scales formed a distinct PF component (N8) with the PF facets anticipating behavior of others, affect communication and internal model of relationships showing the highest loadings. While these scales were highly indicative of the p-factor in Wendt et al. (2023), all PF scales loading on the PF component were highly indicative of a general factor of personality functioning in another study [40]. In addition, we found small to moderate loadings of separation insecurity, suspiciousness, perseveration, eccentricity, and impulsivity on this component. Most of these trait scales have moderate to large correlations with the total score or subdomains of DSM-5 PF [31]. These previous findings indicate that the N8 component found in our study mainly captures variance that is due to PF.
Taken together, while our findings underline the comorbidity problem which led to the development of HiTOP in the first place, they also point towards PF as a construct that is identifiable as a distinct component which shows moderate to high correlations across spectra and hierarchical levels.
Personality Functioning as a Transdiagnostic Construct Capturing Higher-Order Component Variance and Predicting Future Outcomes
The use of symmetrical bifactor models to identify central indicators for higher-order factors in combination with a bifactor-(S-1) model for assessing the predictive validity of PF compared to other lower-order components incorporated both suggestions on “riskier tests” and “bringing theory to the fore” concerning research on higher-order factors of psychopathology [60, 61]. Using this exploratory approach, several PF indicators such as identity, affect differentiation, self reflection, affect tolerance, internal relationship models, and affect communication were identified to be pure markers of the internalizing, thought disorder, and externalizing spectra. Whereas for the externalizing spectrum, more interpersonal aspects of PF (affect communication, anticipation, internal model of relationships) defined the higher-order factor, for the internalizing spectrum, more self-related aspects of PF (identity, affect differentiation, self reflection, affect tolerance) defined the higher-order factor. Most of these scales that capture what could also be called “mentalizing impairments regarding one’s own mental states” were previously described as “pure markers of p” [33]. Our findings therefore substantiate that aspects of PF play a significant role in a broad range of mental disorders, empirically corroborating psychodynamic etiological theory underlying PF summarized by Bender and colleagues [14]: “Biological and environmental problems and their interactions can lead to maladaptive mental models of self and others, and to maladaptive patterns of emotional experience and expression, cognition, and behavior. These, in turn, may lead to the development of psychopathology in general and personality pathology in particular” (p. 344).
The basis for the higher-order structure, that is, the high correlation between the 14 factors on the bottom layer, seems to change significantly if variance that is attributable to PF is partialed out using a bifactor-(S-1) approach. While small to moderately correlated clusters of eating, fear, distress, externalizing, and thought disorder still remain, correlations between residualized components of internalizing and externalizing spectra are negative after partialling out PF. This indicates that PF may explain substantial parts of the variance of higher-order constructs in HiTOP. It could also indicate why patients with different disorders share common etiological pathways and respond to the same treatments. This hypothesis is also tentatively supported by the longitudinal prediction of affective well-being and psychosocial impairment using PF and residualized lower-order components in our sample. Specifically, in the longer run (up to two years), PF accounted for more than two-thirds of total variance explained in psychosocial impairment and more than half of the total variance explained in affective well-being in multiple regression models including 13 additional predictors based on 87 psychopathology scales.
Clinical Implications
Following our findings of the importance of PF in a hierarchical taxonomy of psychopathology, clinicians may assess identity problems, affect differentiation, and communication, along with internal relationship models, as etiologically informed indicators for a broad range of mental disorders. A number of well-validated PF measures may thus provide both parsimonious and reliable utility for assessment and treatment planning. This is in line with recent practical recommendations [62–64] which emphasize that adaptations of treatment indication, modality, and intensity may be based on individual PF assessments. Drawing on findings of PF impairments and their clinical relevance (e.g., higher drop-out rates, less therapy compliance, more risk for ruptures in the therapeutic relationship, and generally higher rates of comorbidity and chronicity), patients with mild impairments in PF may need comparably less intense or structured treatments, whereas patients with moderate and high PF impairments may need highly structured settings and more intense or process-oriented treatments, with a particular emphasis on reducing destructive tendencies towards the self and others [62]. Our findings also tentatively suggest that treatment of internalizing disorders may focus on self aspects of PF, while treatment of externalizing disorders may focus predominantly on interpersonal aspects of PF.
Directions for Future Research
First, research should investigate PF as a potential treatment indicator and target change in transdiagnostic PF features as outcome. While long-term interventions such as psychotherapy seem to be effective for changing PF [65] and traits [22], future research is needed to differentiate the impact on different PF facets. In addition, the development of scalable (digitally aided) interventions that help to support change in PF may be relevant for general healthcare.
Secondly, research on etiological processes should investigate links to transdiagnostic PF features [66]. Disentangling transdiagnostic from specific variance using psychometrically sound latent constructs could also advance studies on genetic mechanisms in psychopathology [8, 67]. Furthermore, longitudinal research that explores the development of PF from early childhood to young adulthood up to adult age with concurrent and HiTOP-conform assessment of psychopathology and allostatic load [68] over a long period of time is needed to investigate etiological questions of causality. A useful approach to disentangle “surface characteristics” from “core processes” [69] could be longitudinal bifactor-(S-1) models.
Finally, in addition to data-driven approaches such as HiTOP, it is also important to continue developmentally informed, theory-based, and theory-oriented research on psychopathology [24, 70–72]. Integrating personality functioning into these models can help to broaden existing approaches, with its descriptive approach on capacities helping to balance or integrate more specific conceptualizations. In the long term, it may also shed light on important empirical and conceptual questions regarding the p-factor [61]. Although progress has been made in differentiating symptoms and traits in the HiTOP model [11], further differentiation with regard to personality functioning is needed. The results of the current study supports other studies that argue that personality functioning may indeed not just index different ways of expressing maladaptive traits [31]. At the same time, the general HiTOP approach has the potential of including these different perspectives into one conceptual model.
Limitations
A number of limitations should be taken into account when interpreting the results of the current study. Approximations on sample characteristics can only be obtained from a separate assessment of a subsample of MindDoc users (see Supplemental material/OSF) and a previous clinical trial [73]). The scales or components of psychopathology that were used for the bass-ackwards-procedure differed with respect to their level of abstraction. For example, the schizophrenia spectrum was assessed with one scale whereas eating pathology was assessed with a total of 14 scales and other areas of interest such as PTSD were missing completely due to the given conceptualization of the mental health app. Further, the two-step answer format leads to high skewness. Although we were able to address this with the latent correlation procedure and the extracted hierarchical structure was similar to other studies using different methods, we cannot rule out consequences regarding our results. In addition, some artefactual components that were removed in the bass-ackwards procedure represent clinically valid phenomena. For example, a patient with predominant anorexic or bulimic features could primarily seek help because of depressive symptoms (K8).
Methodological issues may concern the assessments based on self-reports and the reliability of stepwise estimation using correlation matrices. Furthermore, causal etiological conclusions cannot be drawn unambiguously from our data as the bass-ackwards analysis was based on longitudinally averaged indicators. However, using repetitive longitudinally averaged assessments in a very large sample both removes confounding of within- and between-person variance and enhances estimation precision, both of which represent key shortcomings of cross-sectional self-report data [50]. We could also demonstrate substantial variance explanation by PF in longitudinal prediction. Furthermore, the two main outcomes on prediction differed in assessment methodology as affective well-being was assessed through ambulatory assessment (measurement 3 times per day) and psychosocial impairment included reverse coded items.