Authors: Lacey Chetcuti, Antonio Hardan, Emily Spackman, Eva Loth, James C. McPartland, Thomas W. Frazier, Eric A. Youngstrom, Robert F. Krueger, Mirko Uljarević
Categories: New Research, diagnostic specificity, heterogeneity, social behavior, social engagement, transdiagnostic
Source: JAACAP Open
Authors: Lacey Chetcuti, Antonio Hardan, Emily Spackman, Eva Loth, James C. McPartland, Thomas W. Frazier, Eric A. Youngstrom, Robert F. Krueger, Mirko Uljarević
Altered social drive is a common feature across psychiatric disorders and is embedded in multiple diagnostic criteria, underscoring the need for a transdiagnostic approach. However, the extent to which social drive alterations vary across diagnoses and clinical presentations remains poorly characterized. This study examined whether distinct social drive profiles—defined by differences in social reticence, seeking, and maintaining challenges—relate to variations in clinical features and show specificity to particular neurodevelopmental and neuropsychiatric conditions.
Data were drawn from the Healthy Brain Network (N = 2,380; ages 5-21 years, mean [SD] age = 10.27 [3.39] years; 68% male) and included youth with attention-deficit/hyperactivity disorder, anxiety disorders, autism spectrum disorder, oppositional defiant/conduct disorder, depressive disorders, obsessive-compulsive disorder, and tic disorders. Latent profile analysis identified distinct social drive profiles based on constellations of social reticence, seeking, and maintaining challenges. Profiles were compared across demographic, social functioning, and clinical measures, and the distribution of diagnostic categories within each profile was assessed.
Five profiles engaged (n = 1,530), inhibited (n = 477), aloof (n = 189), avoidant (n = 143), and constrained (n = 50). Profile differences were evident in demographic factors, social functioning, and clinical features. No single diagnosis mapped exclusively onto any profile; rather, participants with distinct neurodevelopmental or neuropsychiatric diagnoses were distributed across all 5 profiles.
Psychiatric diagnoses alone may not fully capture alterations in social drive, which appear to transcend diagnostic boundaries. These findings support a transdiagnostic framework and challenge disorder-specific models of social drive differences.
The exposure to diverse social experiences plays a critical role in development, overall health, and well-being.^1^^,^^2^ Conversely, pronounced variations in the intrinsic drive to engage with others and build relationships can have substantial implications for both concurrent and long-term outcomes.^3^^,^^4^ Indeed, altered social drive is a characteristic of a range of neurodevelopmental diagnoses (NDDs) and neuropsychiatric diagnoses (NPDs). Considering the pronounced heterogeneity within diagnostic conditions and frequent co-occurrence of discretely defined diagnoses related to reduced social drive, it has been suggested that a transdiagnostic perspective is essential for understanding the impact on mental health and well-being.5, 6, 7, 8 The extent to which social drive is similar or different within and across diagnoses remains poorly characterized, both empirically and conceptually.
DSM-5^9^ includes criteria for reduced social approach behaviors or heightened social avoidance in multiple diagnostic criterion A1 (“failure to initiate or respond to social interactions”) and criterion A3 (“absence of interest in peers”) for autism spectrum disorder, the negative symptom of asociality specified in criterion A5 (“lack of interest in social interactions”) for schizophrenia, and criterion D (“social situations are avoided”) for social anxiety disorder. This overlap and lack of conceptual clarity in criteria increases the likelihood of definitional comorbidity among psychiatric conditions, which in turn complicates clinical assessment and treatment planning. Addressing this issue is crucial for developing more precise diagnostic criteria and interventions that better address the association of social drive with diverse clinical presentations.
A central challenge in developing a more precise understanding of social drive across NDDs and NPDs is the lack of a unified conceptual and definitional framework, resulting in inconsistent measurement approaches across different literatures. Constructs such as social motivation, asociality, social anhedonia, and withdrawal are used to describe social drive in different conditions and are often inconsistently defined.^7^ For instance, the operationalization of social drive has ranged from general preferences for social engagement^10^^,^^11^ to more specific models that consider the types of behaviors directed toward social engagement (eg, ingratiating behaviors^12^^,^^13^) or the rewards derived during different stages of social engagement (eg, anticipation vs attainment^14^). This inconsistency precludes the integration of findings across studies, which is essential for identifying shared and distinct clinical presentations and underlying neural mechanisms. Consequently, there is increasing recognition of the need for a more nuanced characterization of social drive in terms of underlying processes that are biologically rooted and relevant across diagnostic and general populations.^7^^,^^15^^,^^16^
Fine-grained assessment of specific processes contributing to social drive is essential for the characterization of diagnostic heterogeneity. For instance, although reductions in social drive are posited to play a key role in the development and expression of autism, significant interindividual variability is apparent within this population. Adopting a person-centered approach, which focuses on identifying patterns of social drive at the level of the individual rather than relying solely on diagnostic categories, Wing and Gould^17^ and subsequently Uljarević et al.^18^ identified distinct subgroups within the autism population characterized by unique constellations of social seeking and maintaining behaviors. More recently, Chetcuti et al.^19^ revealed that children with autism could be characterized into 4 unique profiles, each exhibiting distinctive patterns of strengths and weaknesses across reticence, seeking, and engaged, characterized by the fewest challenges across all areas; inhibited, characterized by high reticence, mild challenges in seeking, and few challenges in maintaining; aloof, characterized by challenges with seeking and maintaining but relatively low reticence; and avoidant, characterized by the highest challenges across all areas. Outside of this handful of studies, the heterogeneity of social motivation in autism has predominantly been appraised using quantitative severity measures such as the Social Responsiveness Scale, Second Edition (SRS-2),^10^ which includes a dedicated social motivation subscale that, nonetheless, collapses distinct social drive processes and other aspects of social functioning into a single summary score. Moreover, the SRS-2 social motivation subscale does not discriminate between autism and anxiety.20, 21, 22 This is despite altered processing of the positive value of stimuli and experiences, with approach-oriented behavior playing a more central role in autism^23^^,^^24^ and altered processing of threats or losses and avoidance-oriented behavior playing a more central role in anxiety.^25^^,^^26^
A fine-grained assessment of social drive processes, which allows for a comprehensive characterization of heterogeneity, holds promise for elucidating both similarities and differences across clinical presentations. This approach has the potential to offer insights beyond those provided by existing instruments. For example, the above-described findings challenge the notion that reduced social drive is uniformly present across individuals with autism,^12^^,^^13^^,^^27^ instead suggesting that no singular profile of social drive difficulties is unique to all individuals with autism.^19^ Building on these insights, the present study aimed to characterize the nature and specificity of alterations in social drive processes—social reticence, seeking, and maintaining—across children and adolescents with an NDD and/or NPD. Following current recommendations for person-centered research,^28^ this study further aimed to appraise the robustness of the profiles across multiple methods and their associations with detailed external validators, including cognitive abilities, co-occurring psychopathologies, and other aspects of social functioning. Lastly, the specificity of social drive profiles to diagnostic categories was assessed by comparing the incidence of profile membership across different diagnostic groups. Given the role of social drive across various NDDs and NPDs and its proposed transdiagnostic relevance,^6^^,^^7^ it was hypothesized that identified profiles of social drive would not be specific to any distinct diagnostic category. Furthermore, identified profiles of social drive were expected to show patterns of association with other aspects of social functioning and co-occurring psychopathology corresponding to those observed in autism.^19^
Data were obtained from the Healthy Brain Network.^29^ The current sample comprised 2,380 participants ages 5 to 21 years (mean [SD] age =10.27 [3.39] years; 68% male) with an NDD and/or NPD based on clinician consensus and ICD-10 criteria^30^ (see Table S1, available online, for additional characteristics). A full description of the Healthy Brain Network study cohort is available elsewhere.^29^
The following diagnoses were assigned either individually or in attention-deficit/hyperactivity disorder (ADHD) (1,847 diagnoses), anxiety disorders (963 diagnoses), autism (520 diagnoses), oppositional defiant disorder (ODD)/conduct disorder (CD) (438 diagnoses), depressive disorders (300 diagnoses), obsessive-compulsive disorder (OCD) (138 diagnoses), and tic disorders (195 diagnoses). These allocations were not mutually exclusive, meaning that participants could receive multiple diagnoses, and all individuals with any combination of these disorders were included in the analysis. Certain diagnoses were infrequently assigned, limiting their inclusion as distinct diagnostic groups in the analysis. These included bipolar disorder (12 diagnoses), eating disorders (23 diagnoses), neurocognitive disorders (3 diagnoses), intellectual disability (60 diagnoses), personality disorders (1 diagnosis), schizophrenia spectrum disorders (11 diagnoses), sleep disorders (14 diagnoses), somatic disorders (2 diagnoses), and substance use disorders (19 diagnoses). Consequently, participants for whom these diagnoses were primary were excluded from the analysis.
Social reticence, seeking, and maintaining scores were calculated based on a previously established approach that combined information from multiple assessment tools^31^: the Social Communication Questionnaire,^32^ SRS-2,^10^ Strengths and Difficulties Questionnaire (SDQ),^33^ Child Behavior Checklist (CBCL),^34^ Screen for Child Anxiety Related Disorders (SCARED),^35^ and Extended Strengths and Weaknesses Assessment of Normal Behavior (E-SWAN).^36^ Social reticence was quantified as the average of 7 items related to experiencing unease or reticence in social situations; social seeking, as the average of 7 items related to the general affinity or preference for social interaction; and social maintaining, as the average of 12 items describing the tendency to employ specific strategies for initiating or maintaining social contact. This method for measuring social drive provided an excellent representation of the data (comparative fit index = 0.97, Tucker-Lewis index = 0.97, root mean square error of approximation = .04 [90% CI 0.037-0.041], standardized root mean square residual = 0.05), and the scores demonstrated good internal consistency reliability (α = 76-.79 and ω = 0.76-0.79).^31^
Latent profile analysis was conducted in Mplus version 8.10^37^ to identify profiles or subgroups of youth with NDDs and/or NPDs who share similar patterns of social reticence, seeking, and maintaining. Analyses used the maximum likelihood estimator with robust standard errors to account for non-normality in the data. The optimal number of profiles was determined using the Akaike information criterion, Bayesian information criterion, and adjusted Bayesian information criterion, where lower values indicate superior fit. Models with different numbers of subgroups were compared using the Lo-Mendell-Rubin likelihood ratio test, where p < .05 indicates that adding another profile improves model fit. Classification accuracy was evaluated using the entropy statistic, with values closer to 1 indicating clearer separation between profiles. Parsimony and interpretability were considered in arriving at the final solution. Next, bootstrapped (1,000 resamples) 1-way analyses of variance were used to examine differences in factor scores between the identified subgroups. To ensure convergence across subtyping methods,^28^ supplementary k-means cluster analysis was performed.
Bootstrapped (1,000 resamples) Pearson’s χ^2^ tests and 1-way analyses of variance were performed to explore profile differences in sex distribution, chronological age, full-scale IQ (FSIQ), and scores derived from the following 1.The SRS-2^10^ is a 65-item parent-report measure of social functioning used widely in the assessment of autism. Scores for the Research Domain Criteria constructs perception and understanding of mental states (UMS) and for production of nonfacial communication (PNFC) and production of facial communication (PFC) were derived from the SRS-2 according to prior factor analysis,^38^ with higher scores indicating more difficulties. To ensure the independence of the study’s independent and dependent variables, 1 item (“Offers comfort to others when they are sad”) was excluded from computation of the UMS scale due to its inclusion in the derived social maintaining factor.^31^2.The Social Aptitudes Scale (SAS)^39^ is a 10-item parent-report questionnaire measure of social aptitudes that involve rapidly interpreting social and emotional cues within complex contexts to guide socially skilled behavior. Items are summed to produce a total score, with higher scores reflecting better social aptitudes.3.The CBCL^34^ is a 118-item parent-report questionnaire used to evaluate emotional and behavioral problems. Internalizing and externalizing problem scores are used, as well as the Dysregulation Profile (CBCL-DP), calculated by summing t scores from anxious/depressed, attention problems, and aggressive behavior subscales.^40^ Higher scores indicate more difficulties. To maintain independence, 2 items (“Too shy or timid” and “Withdrawn, doesn’t get involved with others”) were excluded given their inclusion in, respectively, social reticence and seeking factors.^31^4.The SCARED^35^ is 41-item measure of anxiety symptoms, completed by both parent report and self-report. This study used scores from the school avoidance and separation anxiety subscales, where higher scores indicate greater difficulties.
In addition to examining differences across profiles in demographic and clinical measures, the distribution of individuals with various diagnoses within each profile was assessed. Diagnostic categories were defined based on developmental onset and clinical significance and included the •Autism (n = 520) with or without other diagnoses (ODD/CD, ADHD, OCD, anxiety and/or depression)•ODD/CD (n = 394) excluding youth with autism, with or without other diagnoses (ADHD, OCD, anxiety and/or depression)•ADHD (n = 1,187) excluding youth with autism or ODD/CD, with or without other diagnoses (OCD, anxiety and/or depression)•OCD (n = 73) excluding youth with autism, ODD/CD, or ADHD, with or without other diagnoses (anxiety and/or depression)•Co-occurring anxiety and depressive disorders (n = 41) excluding youth with autism, ODD/CD, ADHD, or OCD•Pure depressive disorders (n = 80) excluding youth with autism, ODD/CD, ADHD, OCD, or co-occurring anxiety•Pure anxiety disorders (n = 237) excluding youth with autism, ODD/CD, ADHD, OCD, or co-occurring depression
Goodness-of-fit indices for the latent profile analysis improved as additional profiles were added (Table 1). Although the LMRT value for the 5-profile model was not statistically significant (p = .167), inspection of the information criteria suggested meaningful improvements up to 5 profiles, with smaller gains thereafter (Figure S1, available online). Additionally, the 5- and 6-profile solutions exhibited the highest entropy values (both 0.871), indicating good classification accuracy. The 5-profile solution was chosen as optimal given that the fifth profile exhibited a qualitatively distinct pattern of social reticence, seeking, and maintaining scores, as well as unique associations with external measures, setting it apart from the other 4 profiles. In contrast, the sixth profile differed primarily in severity rather than pattern, showing similar associations with external measures (Tables S2 and S3, available online). Therefore, models with 7 or more profiles were not explored.Table 1Goodness-of-Fit Indices From Latent Profile Analysis ModelsNo. of subgroupsAICBICABICLMRTEntropy114,856.38614,891.03814,871.974—— n1 = 2,381213,130.85913,188.61213,156.8401,679.525∗∗∗0.936 n1 = 277 n2 = 2,140312,458.92812,539.78212,495.301658.750∗∗∗0.836 n1 = 470 n2 = 1,718 n3 = 193412,075.35712,179.31212,122.122379.373∗∗0.860 n1 = 1,678 n2 = 397 n3 = 49 n4 = 357511,777.27511,904.33111,834.432296.5470.871 n1 = 1,530 n2 = 477 n3 = 189 n4 = 134 n5 = 50611,615.01011,765.16711,682.559164.9610.871 n1 = 1,482 n2 = 434 n3 = 230 n4 = 121 n5 = 34 n6 = 80Note: ABIC = adjusted Bayesian information criterion; AIC = Akaike information criterion; BIC = Bayesian information criterion; LMRT = Lo-Mendell-Rubin likelihood ratio test.∗∗∗p < .001
Table 2 presents comparisons of reticence, seeking, and maintaining scores across the 5 profiles. The first profile (64% of the sample), engaged, demonstrated the fewest challenges across each area. The second profile (20% of the sample), inhibited, was characterized by high levels of reticence, mild challenges in seeking, and few challenges in maintaining. The third profile (8% of the sample), aloof, showed challenges with seeking and maintaining but relatively low reticence. The fourth profile (6% of the sample), avoidant, displayed challenges across all areas. The fifth profile (2% of the sample), constrained, exhibited moderate challenges with reticence and seeking and the highest challenges with maintaining (Figure 1). The 5 profiles identified through latent profile analysis were broadly replicated by a k-means cluster analysis, and consistent patterns in mean reticence, seeking, and maintaining scores were observed across the corresponding profiles (Figure S3, available online).Table 2Profile Comparisons Across Reticence, Seeking, and Maintaining Factor ScoresProfileF/χ^2^pη^2^Engaged (n = 1,530)Inhibited (n = 477)Aloof (n = 189)Avoidant (n = 143)Constrained (n = 50)Mean(SD)Mean(SD)Mean(SD)Mean(SD)Mean(SD)Reticence−0.39(0.45)1.03(0.54)−0.21(0.50)1.25(0.57)0.60(0.88)998.51<.0010.63Seeking−0.29(0.48)0.56(0.57)0.38(0.72)1.34(0.67)1.03(0.89)495.39<.0010.43Maintaining−0.21(0.20)−0.04(0.26)0.74(0.30)1.03(0.31)2.20(0.46)2,426.98<.0010.80Note: Profiles were derived from the full sample across all diagnostic categories. Pairwise comparisons reflect these profiles rather than comparisons within specific diagnostic groups. All means are significantly different from one another. Bias-corrected and accelerated 95% CIs for pairwise contrasts are presented in Table S4, available online.Figure 1Line Graph Showing Mean Values of Profiles***Note:***Line graph shows mean values of reticence, seeking, and maintaining for engaged (n = 1,530; blue), inhibited (n = 477; orange), aloof (n = 189; gray), avoidant (n = 143; yellow), and constrained (n = 50; green) subgroups. Error bars represent 95% CIs calculated from SEM. Scores for reticence, seeking, and maintaining factors were calculated as the average of z scores obtained from each constituent item in their original scales, with higher scores indicating greater challenges.
Table 3 and Figure S4, available online, show the sex distribution; chronological age; and FSIQ, SRS-2, SAS, CBCL, and SCARED scores across the profiles. The engaged profile was significantly younger than the aloof, avoidant, and constrained profiles, which did not differ from each other. The inhibited profile had significantly more girls, whereas the aloof and avoidant profiles had more boys. The engaged profile showed significantly higher FSIQ scores compared with the other profiles, and the inhibited profile had higher FSIQ scores than the avoidant profile.Table 3Profile Comparisons Across Demographic and Social Functioning CharacteristicsProfileF/χ^2^pη^2^/VPairwise comparisonsEn (n = 1,530)In (n = 477)Al (n = 189)Av (n = 143)Co (n = 50)Mean**(SD)Mean(SD)Mean(SD)Mean(SD)Mean(SD)Age, y10.01abc(3.26)10.62a(3.52)10.56(3.41)11.08b(3.66)11.72c(4.21)8.15<.001.01En < In, Av, Co = AlIn = Al = Av = Con****(%)n(%)n(%)n(%)n(%)**Sex, male1,052(67)295j(62)142j(75)102j(76)31(62)18.04<.001.09—FSIQ99.83abc(16.42)96.83ad(16.21)94.78b(17.94)92.86d(16.94)93.35c(18.74)9.37<.001.02En > In, Al, Co; = AvIn > Av; = Al, CoAl = Av = CoSAS20.12abcd(6.84)16.43aefg(6.17)13.72be(7.51)10.72cf(5.72)11.63dg(7.65)90.10<.001.16En > In, Al, Av, CoIn > Al, Av, CoAl = Av = CoSRS-2 UMS11.33abcd(5.09)13.04aefg(4.99)16.56be(4.71)16.48cf(5.04)17.37dg(4.94)78.87<.001.12En < In, Al, Av, CoIn < Al, Av, CoAl = Av = CoSRS-2 PFC1.79abcd(1.02)2.29aef(1.31)2.38bgh(1.21)2.97ceg(1.39)3.28dfh(1.67)26.71<.001.12En < In, Al, Av, CoIn < Av, Co; = AlAl < Av, CoAv = CoSRS-2 PNFC4.88abcd(3.87)6.81aefg(4.73)9.12beh(5.00)10.84cfh(5.32)10.49dg(5.64)104.44<.001.16En < In, Al, Av, CoIn < Al, Av, CoAl < Av; = CoAv = CoCBCL Int8.45abcd(6.84)14.64aef(8.31)11.12begh(8.18)17.12cfg(8.45)16.98dh(12.7A7)93.35<.001.14En < In, Al, Av, CoIn < Al, Av; = CoAl < Av, CoAv = CoCBCL Ext11.08abcd(9.37)12.09aefg(9.11)15.68beh(11.31)15.15cfi(10.73)19.66dghi(12.69)20.16<.001.04En < In, Al, Av, CoIn < Al, Av, CoAl < Co; = AvAv < CoCBCL-DP179.63abcd(20.67)190.36aef(23.61)189.91bgh(23.79)197.61ceg(25.03)201.09dfh(29.91)45.34<.001.08En < In, Al, Av, CoIn < Av, Co; = AlAl < Av, CoAv = CoSCARED Sch (PR)0.97abc(1.36)1.98ad(2.03)1.18de(1.58)1.81be(2.10)1.85c(2.31)39.16<.001.07En < In, Av, Co; = AlIn > Al; = Av, CoAl < Av; = CoAv = CoSCARED Sep (PR)2.67ab(3.04)4.70acd(3.76)2.77ce(3.19)4.57bde(4.02)3.75(4.09)38.03<.001.07En < In, Av; = Al, CoIn > Al, Av; = CoAl < Av; = CoAv = CoSCARED Sch (SR)2.03a(1.97)2.54abcd(2.21)1.97b(2.00)2.08c(1.90)1.64d(2.06)4.94<.001.01En < In; = Al, Av, CoIn > Al, Av, CoAl = Av = CoSCARED Sep (SR)5.00(4.03)5.29abc(4.04)4.25a(4.00)4.25b(3.66)3.61c(3.89)3.37<.01.01En = In, Al, Av, CoIn > Al, Av, CoAl = Av = CoNote: Profiles were derived from the full sample across all diagnostic categories. Pairwise comparisons reflect these profiles rather than comparisons within individual diagnostic groups. Bias-corrected and accelerated 95% CIs for pairwise contrasts are presented in Table S4, available online. Effect sizes are expressed as η^2^ for analysis of variance and Cramer’s V for χ^2^. Al = aloof; Av = avoidant; CBCL = Child Behavior Checklist; Co = constrained; DP = Dysregulation Profile; En = engaged; Ext = Externalizing; FSIQ = full-scale IQ; In = inhibited; Int = Internalizing; PFC = production of facial communication; PNFC = production of nonfacial communication; PR = parent-report; SAS = Social Aptitudes Scale; SCARED = Screen for Child Anxiety Related Disorders; Sch = school avoidance; Sep = separation anxiety; SR = self-report; SRS-2 = Social Responsiveness Scale, Second Edition; UMS = understanding of mental states.a-iMeans with the same superscript letters are significantly different.jObserved frequencies significantly differ from expected frequencies.
The avoidant and constrained profiles showed similarly low social aptitudes per the SAS, followed by the inhibited and aloof profiles, with the engaged profile showing the highest social aptitudes. For the SRS-2, the avoidant and constrained profiles showed the highest scores on UMS, PFC, and PNFC compared with the engaged and inhibited profiles. The constrained profile scored higher than the avoidant profile on UMS, though PFC and PNFC scores were similar. The avoidant profile scored higher than the aloof profile on PFC and PNFC, and the aloof profile in turn scored higher than the engaged profile. The aloof profile also scored higher than the inhibited profile on UMS and PNFC, whereas PFC scores were similar.
CBCL internalizing scores were higher for the avoidant and constrained profiles than for the engaged and aloof profiles, with no difference between avoidant and constrained profiles. The avoidant profile had higher internalizing scores than the inhibited profile, which scored higher than the engaged and aloof profiles. The aloof profile had higher scores than the engaged profile. CBCL externalizing scores were highest for the constrained profile, followed by the avoidant and aloof profiles, which scored higher than the engaged and inhibited profiles, but did not differ from each other. The inhibited profile scored higher than the engaged profile. CBCL-DP scores were higher for the avoidant and constrained profiles than the aloof, inhibited, and engaged profiles, with no difference between avoidant and constrained profiles. Both the aloof and inhibited profiles had higher CBCL-DP scores than the engaged profile, but did not differ from each other.
For the SCARED, parent-reported school avoidance was higher in the inhibited, avoidant, and constrained profiles compared with the engaged profile. The inhibited and avoidant profiles also scored higher than the aloof profile, which fell between the engaged and constrained profiles, though the differences were not significantly different. Parent-reported separation anxiety was higher for the inhibited profile compared with the avoidant profile, with both scoring higher than the engaged and aloof profiles, which did not differ from each other. No differences in parent-reported separation anxiety were found for the constrained profile compared with other profiles. When self-reported, profile differences on the SCARED were less pronounced. The inhibited profile showed higher school avoidance than all other profiles and higher separation anxiety than the aloof, avoidant, and constrained profiles, but not the engaged profile.
Figure 2 illustrates the distribution of examined diagnoses within each profile, and Figure S5, available online, presents the distribution of profile membership per diagnosis. No singular diagnostic category exclusively corresponded to any specific social drive profile; rather, youth with distinct NDDs or NPDs were distributed across all 5 profiles.Figure 2Bar Graph Showing Distribution of Examined Diagnoses Within Each Profile***Note:*Diagnostic categories include autism (n = 520), oppositional defiant/conduct disorder (ODD/CD; n = 394), attention deficit/hyperactivity disorder (ADHD; n = 1,187), obsessive-compulsive disorder (OCD; n = 73), co-occurring anxiety and depressive disorders (n = 41), 'pure' depressive disorders (n = 80), and ‘pure’ anxiety disorders (n = 237). Distinct categories were established based on developmental onset and clinical significance and are presented in the key in that order from left to right, allowing individuals in a given diagnostic category to also meet the criteria for any conditions listed to the right. Refer to theMethodsection for more details regarding exclusions from each category.
Autism was more common in the aloof (42%), constrained (46%), and avoidant (53%) profiles compared with engaged (14%) and inhibited (25%) profiles. Conversely, ADHD was more common in the engaged (57%), inhibited (42%), and aloof (34%) profiles compared with constrained (26%) and avoidant (24%) profiles.
Furthermore, depression (without co-occurring anxiety) was more common in the constrained profile (6%), relative to the remaining 4 profiles (each 1%). Conversely, anxiety (without co-occurring depression) was more common in the inhibited (13%) and engaged (9%) profiles, relative to the avoidant (5%), aloof (4%), and constrained (0%) profiles.
ODD/CD was similarly represented across all profiles (ranging from 12% in inhibited profile to 16% in constrained profile), as was OCD and tic disorders (ranging from 2% in both aloof and constrained profiles to 4% in inhibited profile), and co-occurring anxiety and depression (ranging from 1% in both avoidant and engaged profiles to 4% in constrained profile). Examination of the frequencies of mutually exclusive diagnoses across the 5 profiles revealed a consistent pattern of results (Figure S6, available online).
Given that altered social drive occurs across a range of NDDs and NPDs and features in multiple diagnostic criteria, adopting a transdiagnostic approach, which focuses on mechanisms that cut across diagnostic categories, is essential.5, 6, 7, 8 However, the extent to which alterations in social drive are similar or different across specific diagnoses and clinical presentations remains unclear. Addressing this question is crucial for refining clinical descriptions and tailoring psychiatric care to individual needs, ultimately improving patient outcomes. Yet, progress has been hindered by the limitations of current assessment tools, which lack validity across various clinical populations and fail to capture the nuanced, underlying processes contributing to the heterogeneity of social drive and its differential presentation.^7^ Using an innovative multi-instrument approach to approximate key social drive processes, the current study addresses this issue by investigating whether profiles of social drive—characterized by distinct patterns of reticence, seeking, and maintaining challenges—are associated with variations in other clinical features and are specific to certain NDDs and NPDs.
Consistent with previous findings,^19^ the engaged and avoidant profiles displayed a relatively uniform distribution of difficulties across social reticence, seeking, and maintaining and were quantitatively differentiated along a gradient of severity. In contrast, the inhibited and aloof profiles were qualitatively the inhibited profile demonstrated high reticence but fewer challenges in social seeking and maintaining, whereas the aloof profile showed low reticence and greater difficulties in social seeking and maintaining. Furthermore, comparison of the profiles with external validators—including other aspects of social functioning (including social aptitudes, production of facial and nonfacial communication, and understanding of mental states) and co-occurring psychopathologies (including internalizing, externalizing, and dysregulation problems)—revealed similar results to those obtained in autism.^19^ That is, among these 4 profiles, the engaged profile exhibited the fewest challenges, and the avoidant profile exhibited the greatest challenges. The aloof profile exhibited more challenges with social functioning than the inhibited profile, and the inhibited profile exhibited more challenges with internalizing than the aloof profile.
In addition to replicating the profiles identified previously in autism, the profiles derived from the current transdiagnostic sample also exhibit similarities to those found in the general population based on temperament and personality measures. Asendorpf^41^^,^^42^ and, subsequently, Coplan et al.^43^ characterized 3 profiles of social withdrawal based on varying approach-avoidance shy, characterized by social withdrawal due to social fears and social-evaluative concerns despite a desire to interact with others; unsociable/disinterested, characterized by a lack of interest in social interaction due to a nonfearful preference for solitude; and avoidant, characterized by a preference for solitude and an active avoidance of social interactions. These profiles closely resemble the inhibited, aloof, and avoidant subgroups identified here. Furthermore, there are similarities in terms of other correlates and outcomes, including more pronounced peer problems and mental health difficulties among avoidant and shy profiles.44, 45, 46 These apparent similarities in social drive profiles across various diagnostic categories and the general population, as well as across measures targeting both clinical challenges and trait-level variability, suggest that variations in social drive extend beyond traditional diagnostic classifications, possibly spanning a wider functional spectrum. Future investigations, including diverse clinical and nonclinical populations, should explore this possibility.
This study also identified a novel constrained profile, characterized by moderate difficulties in both reticence and seeking, along with significant challenges in maintaining social interactions. This profile has not been previously identified in autism17, 18, 19 and does not align with profiles found in the general population. However, individuals with autism made up a large proportion (46%) of its diagnostic composition, suggesting that its absence in prior work may have been due to insufficient statistical power. This profile closely resembled the avoidant profile in displaying the greatest challenges with social functioning and co-occurring psychopathologies relative to the engaged, inhibited, and aloof profiles. However, it exhibited more pronounced challenges than the avoidant profile in relation to externalizing and understanding mental states. Additionally, the constrained profile comprised a higher prevalence of individuals with depression diagnoses (10%, including those with or without co-occurring anxiety) compared with other profiles. Intriguingly, these clinical features of the constrained profile align with those typically preceding the onset of bipolar disorder,^47^^,^^48^ which commonly emerges in adolescence and early adulthood,^49^ beyond the age range of the current sample. Although not statistically significant, it is worth noting that participants in the constrained profile tended to be slightly older than participants in other profiles. However, the subsample of individuals with bipolar disorder diagnosis was too small (n = 12) to facilitate a meaningful comparison of profile membership. Nevertheless, the potential link between a constrained social drive profile and the emergence of bipolar disorder warrants further exploration.
Lastly, examination of the diagnostic composition of each profile indicated that social drive patterns do not align consistently with any particular NDD or NPD; rather, individuals with a given diagnosis were distributed across the different profiles. This finding supports a transdiagnostic conceptualization of social drive5, 6, 7, 8 and highlights the heterogeneity of social drive alterations within clinical conditions, which becomes evident when social drive is measured at a fine-grained level. Notably, while a greater proportion of individuals with autism were classified in profiles characterized by reduced social drive, a sizable proportion were classified as engaged, characterized by mild difficulties across all areas. Conversely, a sizable proportion of individuals with diagnoses for which altered social drive is not a well-recognized feature—including ODD/CD, ADHD, and tic disorder—were classified within profiles characterized by reticence, seeking, and/or maintaining difficulties. Consequently, psychiatric diagnosis, unto itself, may be an imprecise indicator of the degree or underlying nature of alterations within key social drive processes, and alterations within key social drive processes may not be exclusive to a single diagnosis, but rather shared across various clinical presentations. This idea is further reinforced by the fact that the profiles identified in the current analyses were based on severity measures that primarily focused on the clinical end of the functional continuum, not accounting for social drive strengths. This underscores the need to further refine identified profiles by using instruments that can capture broader variation in the comprehensive set of social drive processes.
Although further investigation is needed, it is plausible that distinct profiles of social drive may correspond to different support needs and responses to treatment. If so, interventions should be tailored to individual profiles rather than applied uniformly by diagnostic category, moving beyond a one-size-fits-all approach. For example, individuals with a constrained profile may benefit from interventions that build confidence in social interactions and strengthen communication skills, whereas those with an inhibited profile may require gradual, supported exposure to social situations alongside strategies for managing anxiety. By contrast, individuals with an engaged profile, who present with fewer social drive difficulties, might gain more from interventions that target other developmental or complex social skills to foster interpersonal growth. Identifying social drive profiles may thus help clinicians better match interventions to patient needs, promote equity in access to effective treatments, and optimize allocation of resources.
At the same time, it is important to recognize that most existing interventions were designed to address discrete diagnoses rather than transdiagnostic dimensions such as social drive.^50^ Consequently, we currently lack sufficient evidence to determine which therapeutic approaches or mechanisms are most effective for children and adolescents with specific social drive profiles. For instance, many behavioral interventions for autism presume reduced social reward processing and aim to enhance the salience of social input to increase opportunities for learning. However, our findings suggest that social drive is not uniform in autism or across other NDDs and NPDs. This raises the possibility that such interventions may be less suitable for some individuals; for instance, those with an engaged profile may benefit more from targeting other areas of functioning, whereas those with an inhibited profile may experience distress if social exposure is not introduced sensitively.
Findings reported here underscore the need to shift from prescribing interventions primarily on the basis of diagnostic classification to tailoring them according to individual social drive profiles (engaged, inhibited, aloof, avoidant, or constrained). By improving alignment between individual needs and interventions, such an approach has the potential not only to enhance quality of life, but also to reduce long-term economic costs for families, service systems, and society at large by directing clinical resources toward approaches most likely to be effective and ultimately encouraging greater social participation and independence. Further research is needed to test how tailoring interventions to social drive profiles influences patient outcomes, with the goal of advancing more personalized, effective, and sustainable approaches to care.
Although this study advances understanding of individual differences in social drive across NDDs and NPDs, some limitations should be noted. This study focused on a subset of participants within the Healthy Brain Network with clinician-identified NDDs and/or NPDs that occurred with sufficient frequency to allow meaningful comparisons. Consequently, the findings may not generalize to the broader population or to individuals with less common diagnoses for which differences in social drive have also been implicated. Future research should include a wider range of NDDs and NPDs to provide further insights into the transdiagnostic nature of social drive. Moreover, this study, based on informant-report scales at a single level of analysis, does not provide insights into individual differences at the neurobiological level, where cognitively mediated social drive processes, such as learning through social experience, might be more effectively captured. It also limited the exploration of brain-behavior correspondence across diagnostic categories. According to multifinality and equifinality principles,^51^ the same behavioral expression of social drive can arise from different underlying processes, whereas different behavioral expressions of social drive can result from the same underlying process. Relatedly, individuals with the same diagnosis may initially show similar social drive difficulties, but these can diverge over time due to the maturation of skills such as executive functioning and social cognition that influence social drive processes, as well as environmental interactions. However, the cross-sectional design of this study limits exploration of developmental pathways. Future research on social drive convergence and divergence across diagnoses should use a multimethod approach, integrating neurobiological and behavioral techniques, and a longitudinal design that spans key developmental periods, including early development and the transition from adolescence to adulthood.
Existing scales for measuring social reticence, seeking, and maintaining focus primarily on the clinical end of the functional continuum, limiting insights into normal-range variation. This restricts the ability of the study to identify subgroups based on strengths or the absence of difficulties in social drive processes. Additionally, even when combined through advanced multi-instrument approaches, these scales provide limited coverage of finer social drive processes, such as orienting (detecting social stimuli and activating interest), wanting (arousal in anticipation of reward), pursuing (decision making based on potential benefits and costs), liking (immediate pleasure from social stimuli), and learning (updating stimulus-reward associations), as outlined in a reward processing framework.^7^ However, to progress further in this direction, it is essential to develop and use rigorously validated informant- and performance-based tools that offer a more granular assessment of social drive across the full range of functioning.
Finally, current tools often overlook key contextual factors, such as the nature of relationships, which significantly impact social drive. For example, the motivation to seek or avoid interaction varies depending on whether one is engaging with a friend, stranger, or authority figure. Social drive is also shaped by environmental contexts, such as whether a situation feels supportive or threatening and whether the individual feels competent or overwhelmed. Without considering these influences, existing measures may fail to capture the complexity of social drive. Integrating these factors into finer-grained assessments could enhance our understanding of social drive, leading to more precise profiling across clinical presentations and more tailored interventions. Moreover, the current sample, drawn from families seeking psychiatric assessment through the US Healthy Brain Network,^29^ may not fully capture the experiences of diverse and underrepresented groups (see Table S1, available online, for demographic characteristics). Cultural norms, expectations, and socioeconomic factors may shape the manifestation of social drive profiles and the clinical needs they indicate. This highlights the importance of future research to ensure that findings are generalizable and that clinical applications are appropriate, culturally sensitive, and relevant across diverse populations.
This study provides a comprehensive characterization of individual differences in social drive, representing a significant advancement over much of the existing literature. In contrast to previous studies that relied on single scores or overall ratings of social drive,^28^ our study identified profiles based on nuanced social drive processes derived through multi-instrument factor analysis.^31^ These profiles were robust across different subtyping statistical approaches and associated with detailed external validators, adhering to best practice guidelines. Additionally, the inclusion of a large, well-characterized sample from various NDD and NPD groups, along with a transdiagnostic measurement approach, enabled the most detailed examination to date of social drive difficulties across diagnostic boundaries. Findings revealed no clear correspondence between NDDs and NPDs and the expression of social drive difficulties, as evidenced by their representation across distinctive engaged, inhibited, aloof, avoidant, and constrained profiles. Although replication and extension are warranted, these findings support a transdiagnostic conceptualization of social drive5, 6, 7, 8 and challenge diagnosis-specific models. Moving forward, the refinement of measurement tools to capture nuanced processes, such as social orienting, wanting, pursuing, liking, and learning, alongside the adoption of multimethod and longitudinal research designs, promises deeper insights into the diverse expression of social drive across diagnostic boundaries, ultimately facilitating more precise clinical care.
Lacey Chetcuti: Writing – review & editing, Writing – original draft, Visualization, Project administration, Methodology, Investigation, Formal analysis, Data curation, Conceptualization. Antonio Hardan: Writing – review & editing, Supervision, Resources, Project administration, Investigation, Funding acquisition. Emily Spackman: Writing – review & editing, Methodology, Investigation. Eva Loth: Writing – review & editing, Methodology, Investigation. James C. McPartland: Writing – review & editing, Methodology, Investigation. Thomas W. Frazier: Writing – review & editing, Methodology, Investigation. Eric A. Youngstrom: Writing – review & editing, Methodology, Investigation. Robert F. Krueger: Writing – review & editing, Methodology, Investigation. Mirko Uljarević: Writing – review & editing, Supervision, Project administration, Methodology, Investigation, Funding acquisition, Data curation.