This study aims to prospectively collect harmonized, quantitative, and dimensional psychiatric phenotypes (suicidality, anhedonia, and obsessive-compulsive symptoms) and information on discrimination, stigma, and unfair treatment in up to 27,500 individuals across diverse ancestries and clinical populations for genetic analysis within the NIMH Ancestral Population Network (APN). Five APN projects are participating in the current study (PUMAS-Paisa, PUMAS-NeuroGAP, KOMOGEN-D, LATINO, and SAX-II), spanning recruitment over 15 countries across the Americas, Africa, and Asia. Participants are being enrolled via the recruitment protocols specific to each participating project. The harmonized clinical battery includes the Columbia-Suicide Severity Scale, Snaith-Hamilton Pleasure Scale, and Florida Obsessive-Compulsive Inventory-II to assess suicidality, anhedonia, and obsessive-compulsive symptoms, respectively. Additionally, we use the Discrimination and Stigma Scale-Ultra Short to evaluate discrimination and stigma across cultures. This project is a pioneering effort to prospectively assess transdiagnostic phenotypes in large-scale, diverse ancestral and clinical populations (across different diagnoses) using a harmonized clinical battery, advancing genetic research in mental health in underrepresented populations.

Bipolar disorder's (BD) clinical heterogeneity has an unresolved genetic basis. We meta-analyzed genome-wide association studies (GWAS) of 16 BD subphenotypes in 226,032 individuals from 57 cohorts (38,022 cases); 10 advanced to multivariate and multi-trait analyses. Four factors (compulsive, psychotic, dysregulated, internalizing) explained 82.8% of shared genetic variance. BD1 and BD2 loaded on distinct factors despite a high genetic correlation; 87.0% of common-factor loci were significant in neither subtype. Unipolar mania aligned with psychosis over internalizing, and was distinguishable from BD1, and rapid cycling showed heritable cross-domain liability. We identified 356 risk loci, 158 novel, including the first univariate-GWAS associations for psychosis, unipolar mania, rapid cycling and schizoaffective disorder-and 249 credible genes (89 high-confidence), 12 with approved-drug or clinical-phase annotations. Cell-type association showed a midbrain dopaminergic-GABAergic gradient along the psychotic factor. BD's genetic architecture appears hierarchical-a general liability resolving into dimensions of course and comorbidity, beyond subtypes.

Abstract Background Free-text notes in electronic health records (EHRs) contain fine-grained psychiatric information that is essential for psychiatric research and clinical care, and often absent or under-recorded in structured codes alone. Clinical natural language processing (cNLP) can support extraction of this information from EHR notes, yet Spanish-language cNLP remains under-developed. Moreover, broad evaluations comparing multiple encoder-based language models across extensive, fine-grained psychiatric concept sets remain scarce, and it remains unclear how these models compare with traditional NLP (tNLP) systems and much larger generative large language models (LLMs). In addition, cross-site performance of fine-tuned models is rarely tested, and limited annotated training data remains a major challenge, especially for rare symptoms. Objectives We aimed to advance scalable, global psychiatric cNLP by fine-tuning multiple encoder-based models with differing architectures and pre-training strategies for detecting fine-grained psychiatric concepts in Spanish EHRs. We further evaluated the impact of augmenting the fine-tuning data with precision-weighted weak labels for less-frequent concepts, and compared the performance of the encoder-based models to that of tNLP and a fine-tuned generative LLM trained on the same data. Finally, we evaluated model cross-site generalizability on an external EHR dataset. Methods Three encoder-based models (BETO, XLM-RoBERTa-large, and bsc-bio-ehr-es) were fine-tuned on 1,642 clinician-annotated EHR documents from Colombia to detect 110 psychiatric concepts in Spanish text. To address the limited annotated examples available for less-frequent concepts, 12,000 additional documents were weakly-labeled for less-frequent concepts using tNLP, and incorporated into the fine-tuning data with labels weighted by pattern precision. Models were compared with tNLP and a generative LLM, and evaluated on an external EHR dataset from another psychiatric hospital in Colombia. Results Encoder model performance varied substantially, with macro-F1 ranging from 0.64 to 0.81. BETO achieved the highest macro-F1 (0.81; median F1=0.88 [IQR=0.77-0.96]). Adding precision-weighted weak labels for less-frequent concepts improved BETO’s overall macro-F1 to 0.83 and increased mean F1 for the 55 augmented concepts from 0.82 to 0.86. Under matched fine-tuning conditions, fine-tuned BETO and the tNLP method were equivalent in F1, whereas the LLM significantly outperformed BETO in F1. After weak-label augmentation, BETO significantly outperformed tNLP in F1 ( P FDR <.001) and narrowed the performance gap with the LLM, although equivalence was not established. Lastly, fine-tuned BETO maintained reasonably strong performance on data from an external hospital not used for model fine-tuning (out-of-domain macro-F1=0.78). Conclusions General-purpose pre-trained encoders had strong performance for psychiatric concept extraction from Spanish EHRs. Weak-label augmentation improved BETO’s performance and strengthened results relative to a tNLP baseline, while reducing, but not eliminating, the performance gap with a much larger fine-tuned generative LLM. These findings highlight the utility of these relatively lightweight models for scalable, accurate and reproducible detection of psychiatric concepts in Spanish-language EHRs.

Most genome-wide association studies (GWAS) of serious mental illness (SMI) have been conducted for categorical diagnoses in samples of primarily European ancestry. The portability of findings to non-Europeans, and to SMI-related symptoms/dimensional traits remains uncertain. In a sample of 8,666 SMI cases and controls from the Paisa region of Colombia we show that a primarily European schizophrenia GWAS polygenic risk score (PRS) predicted all SMI diagnoses in this sample, as well as symptoms (assessed in cases only) and traits assessed agnostic to SMI diagnosis: a one SD unit (SDU) increase in this PRS was associated to decreased risk in cases of suicidal thoughts (OR=0.89, 95% confidence interval 0.84-0.94), depressed mood (OR=0.90, 95% confidence interval 0.85-0.95), and increased risk of delusions (OR=1.12, 95% confidence interval 1.06-1.18) and to decreased cognition (in cases and controls) across five distinct domains (average decrease in cognition of 0.065 SDU, p<7e-05). We show that a published European GWAS of cognition predicted levels of executive function (average decrease in cognition of 0.06 SDU per unit increase in PRS, p<2e-04), but not diagnosis or symptoms. Specific loci identified in the SMI GWAS also showed association to multiple diagnoses, symptoms, and cognitive traits in Paisa. The most noteworthy result was for a locus on chromosome 7p22.3, associated in multiple SMI GWAS, that showed association in Paisa to increased risk of bipolar disorder, and to reduced complex cognition and social cognition. Our findings demonstrate wide portability from European GWAS to an admixed American sample, with associations to multiple transdiagnostic phenotypes.

Here we developed and deployed the blended genome exome (BGE) method, a DNA library approach that generates low-pass whole-genome (1-4× mean depth) and deep whole-exome (30-40× mean depth) data in a single sequencing run. BGE is cost-effective, empowers most genomic discoveries possible with deep whole-genome sequencing and captures global common single-nucleotide polymorphism diversity. We applied BGE to sequence >53,000 samples from the PUMAS Project (Populations Underrepresented in Mental Illness Associations Studies), including African, African American and Latin American populations. Imputed genotypes showed high concordance with Illumina Global Screening Array calls (R2 ≥ 95% for minor allele frequency ≥1%; ≥90% for minor allele frequency <1%), with consistent performance across local ancestries in admixed cohorts. For protein-coding copy number variants, deletions and duplications spanning at least three exons had a positive predicted value of ~90% relative to deep whole-genome data. At ~28% of the cost of deep whole-genome sequencing, BGE provides a scalable, reliable platform to expand genomic discovery and equitable access to sequencing in underrepresented populations.

Data passively obtained from smartphones and wearables can provide nearly continuous objective information that enables quantification of states and traits across broad physiological, behavioral, and emotional domains impacted in mental health conditions, including depression and anxiety. Widespread application of such digital phenotyping could transform the assessment of depression and anxiety in research and clinical care, but the field has lacked well-powered longitudinal studies demonstrating the utility of this approach. This paper describes the design and implementation of the Digital Mental Health Study (DMHS), which collected up to 12 months of sensor data from iPhone and Apple Watch in over 4000 consenting participants, a sample diverse by age, sex at birth, ethnicity, and depression symptom severity. To enable the use of these digital phenotypes to assay the complexity and heterogeneity of depression and anxiety, we designed a protocol of periodic self-report and interview-based scales optimized to assess elements of depression, anxiety, and perceived stress as broadly as possible while minimizing participant measurement burden. We report here the strategies used to recruit and enroll the DMHS sample, the process employed to develop study methods and protocols, and initial findings describing longitudinal symptom trajectories and demonstrating high participant engagement over 12 months.

BACKGROUND:Electronic health records (EHRs), increasingly available in low- and middle-income countries (LMICs), provide an opportunity to study transdiagnostic features of serious mental illness (SMI) and its trajectories. AIMS:Characterise transdiagnostic features and diagnostic trajectories of SMI using an EHR database in an LMIC institution. METHOD:We conducted a retrospective cohort study using EHRs from 2005-2022 at Clínica San Juan de Dios Manizales, a specialised mental health facility in Colombia, including 22 447 patients with schizophrenia (SCZ), bipolar disorder (BPD) or severe/recurrent major depressive disorder (MDD). Using diagnostic codes and clinical notes, we analysed the frequency of suicidality and psychosis across diagnoses, patterns of diagnostic switching and the accumulation of comorbidities. Mixed-effect logistic regression was used to identify factors influencing diagnostic stability. RESULTS:High frequencies of suicidality and psychosis were observed across diagnoses of SCZ, BPD and MDD. Most patients (64%) received multiple diagnoses over time, including switches between primary SMI diagnoses (19%), diagnostic comorbidities (30%) or both (15%). Predictors of diagnostic switching included mentions of delusions (odds ratio = 1.47, 95% CI 1.34-1.61), prior diagnostic switching (odds ratio = 4.01, 95% CI 3.7-4.34) and time in treatment, independent of age (log of visit number; odds ratio = 0.57, 95% CI 0.54-0.61). Over 80% of patients reached diagnostic stability within 6 years of their first record. CONCLUSIONS:Integrating structured and unstructured EHR data reveals transdiagnostic patterns in SMI and predictors of disease trajectories, highlighting the potential of EHR-based tools for research and precision psychiatry in LMICs.

The accurate detection of clinical phenotypes from electronic health records (EHRs) is pivotal for advancing large-scale genetic and longitudinal studies in psychiatry. Free-text clinical notes are an essential source of symptom-level information, particularly in psychiatry. However, the automated extraction of symptoms from clinical text remains challenging. Here, we tested 11 open-source generative large language models (LLMs) for their ability to detect 109 psychiatric phenotypes from clinical text, using annotated EHR notes from a psychiatric clinic in Colombia. The LLMs were evaluated both "out-of-the-box" and after fine-tuning, and compared against a traditional natural language processing (tNLP) method developed from the same data. We show that while base LLM performance was poor to moderate (0.2-0.6 macro-F1 for zero-shot; 0.2-0.74 macro-F1 for few shot), it improved significantly after fine-tuning (0.75-0.86 macro-F1), with several fine-tuned LLMs outperforming the tNLP method. In total, 100 phenotypes could be reliably detected (F1>0.8) using either a fine-tuned LLM or tNLP. To generate a fine-tuned LLM that can be shared with the scientific and medical community, we created a fully synthetic dataset free of patient information but based on original annotations. We fine-tuned a top-performing LLM on this data, creating "Mistral-small-psych", an LLM that can detect psychiatric phenotypes from Spanish text with performance comparable to that of LLMs trained on real EHR data (macro-F1=0.79). Finally, the fine-tuned LLMs underwent an external validation using data from a large psychiatric hospital in Colombia, the Hospital Mental de Antioquia, highlighting that most LLMs generalized well (0.02-0.16 point loss in macro-F1). Our study underscores the value of domain-specific adaptation of LLMs and introduces a new model for accurate psychiatric phenotyping in Spanish text, paving the way for global precision psychiatry.

Introduction:Clinical notes in electronic health records offer valuable insight into the symptom profiles and trajectories of patients with severe mental illness (SMI). However, systematically extracting symptoms at scale remains a challenge, especially in languages other than English. We developed a light, accurate, and interpretable natural language processing (NLP) algorithm to extract psychiatric phenotypes from Spanish clinical notes. Methods:We selected a set of 136 core psychiatric phenotypes and annotated 4,000 clinical note sections (e.g., Chief Complaint, Plan; called "documents") and 240 complete visit notes (called "entries") from two psychiatric hospitals in Colombia: Hospital Mental de Antioquia (HOMO) and Clínica San Juan de Dios Manizales (CSJDM). For phenotypes meeting frequency and inter-annotator reliability thresholds, we developed three NLP algorithms (HOMO, CSJDM, and COMBINED) for phenotype extraction and context labeling (e.g., negation, family history, uncertainty). We evaluated performance at the document and entry levels, as well as across hospitals. Results:Document-level performance at both hospitals was high (average F1 scores of 0.84 and 0.85). Moreover, on phenotypes meeting our document-level performance threshold of F1 ≥0.7, entry-level performance was high as well (average F1 of 0.75 and 0.78), as was the cross-hospital transportability of the algorithms (F1 of 0.75 HOMO-to-CSJDM and 0.77 CSJDM-to-HOMO). The COMBINED algorithm improved overall recall, without significantly decreasing precision (F1 of 0.78 and 0.77 on HOMO and CSJDM, respectively). The application of our algorithm for 50 high-performing phenotypes to the notes of 9,737 SMI patients highlighted the transdiagnostic nature of many core SMI phenotypes; 44/50 phenotypes were recorded in over 10% of patients across diagnoses. Multiple correspondence analysis further revealed variation in symptom space across diagnoses; while major depressive disorder and schizophrenia form distinct clusters, patients with bipolar disorder span the entire phenotypic spectrum. Conclusion:Our tool enables the systematic investigation of psychiatric symptoms from psychiatric notes, facilitating large-scale investigations in Spanish-speaking populations.

Bipolar disorder is a leading contributor to the global burden of disease1. Despite high heritability (60-80%), the majority of the underlying genetic determinants remain unknown2. We analysed data from participants of European, East Asian, African American and Latino ancestries (n = 158,036 cases with bipolar disorder, 2.8 million controls), combining clinical, community and self-reported samples. We identified 298 genome-wide significant loci in the multi-ancestry meta-analysis, a fourfold increase over previous findings3, and identified an ancestry-specific association in the East Asian cohort. Integrating results from fine-mapping and other variant-to-gene mapping approaches identified 36 credible genes in the aetiology of bipolar disorder. Genes prioritized through fine-mapping were enriched for ultra-rare damaging missense and protein-truncating variations in cases with bipolar disorder4, highlighting convergence of common and rare variant signals. We report differences in the genetic architecture of bipolar disorder depending on the source of patient ascertainment and on bipolar disorder subtype (type I or type II). Several analyses implicate specific cell types in the pathophysiology of bipolar disorder, including GABAergic interneurons and medium spiny neurons. Together, these analyses provide additional insights into the genetic architecture and biological underpinnings of bipolar disorder.

Tourette syndrome (TS) is a highly heritable childhood-onset neuropsychiatric disorder characterized by persistent motor and vocal tics. While both common and rare variants contribute to TS susceptibility, the role of rare de novo mutations (DNMs) remains incompletely characterized. Here, we report findings from the largest TS whole-exome sequencing study to date, analyzing 1,466 TS trios alongside 6,714 autism spectrum disorder (ASD) trios and 5,880 unaffected sibling controls from the Simons Simplex Collection (SSC) and SPARK cohorts. Leveraging a trio-based design across these cohorts enabled calibrated assessment of DNM burden while controlling for background mutation rates. We observed a significant exome-wide enrichment of protein-truncating DNMs in TS probands, particularly within genes intolerant to loss-of-function variation (pLI ≥ 0.9), with little contribution from damaging missense variants. Notably, TS probands did not exhibit enrichment in previously implicated ASD or developmental delay (DD) genes, but elsewhere in the genome, suggesting a distinct rare variant architecture. Using a Bayesian statistical framework that integrates both de novo and rare inherited coding variants, we identified three candidate TS risk genes with FDR ≤ 0.05: PPP5C , EXOC1 , and GXYLT1 . Literature shows that they have prior links to neurodevelopmental and psychiatric disorders. These findings reveal a rare variant burden in TS that is genetically distinguishable from ASD, underscore the importance of loss-of-function mutations in TS risk, and nominate novel candidate genes for future functional investigation.

BACKGROUND:Despite growing healthcare coverage, disparities in access to and outcomes of psychiatric care persist, even in countries with universal healthcare. How socioeconomic status (SES), travel time, and social support individually and jointly affect psychiatric clinical trajectories remains largely unexplored. METHODS:We analyze electronic health records (EHRs) from patients diagnosed with bipolar disorder, major depressive disorder, or schizophrenia at Clínica San Juan de Dios Manizales. Using zero-inflated and standard negative binomial regression, we quantify the effects of SES, travel time, and family/social support on utilization, clinical outcomes, and symptoms of mania, psychosis, and suicidality. A mixed-effects model examines how care-seeking patterns affect visit-to-visit variability in outcomes. RESULTS:Among 21,095 patients, utilization is lower for those with low SES (rate ratio [RR] 0.92, 95% CI: 0.90-0.95, p = 1.27e-10) and longer travel times (RR 0.94, 95% CI: 0.93-0.95, p = 1.19e-53). Patients with low SES are more likely to have severe symptoms (e.g., delusions: RR 1.28, 95% CI: 1.20-1.37, p = 2.57e-15) and require hospitalization (RR 1.10, 95% CI: 1.05-1.15, p = 1.94e-04), suggesting they primarily seek care when critical. Longer travel differentially affects those with low SES. However, the relationship between SES and adverse outcomes is less pronounced when living with family (e.g., hospitalizations: LRT, χ2 = 47.08, df = 3, p = 3.35e-10). Frequent outpatient care is associated with lower odds of hospitalization, suicidality, and other symptoms. CONCLUSIONS:Findings demonstrate use of EHRs to model patient outcomes, the important role of social support, and need for improved healthcare accessibility.

Importance:The clinical heterogeneity of bipolar disorder (BD) is a major obstacle to improving diagnosis, predicting patient outcomes, and developing personalized treatments. A genetic approach is needed to deconstruct the disorder and uncover its fundamental biology. Previous genetic studies focusing on broad diagnostic categories have been limited in their ability to parse this complexity. Objective:To test the hypothesis that clinically distinct subphenotypes of BD are associated with different underlying common variant genetic architectures. Design Setting and Participants:This multicenter study included a primary genome-wide association study (GWAS) of up to 23,819 bipolar disorder (BD) cases and 163,839 controls. These results were integrated via multi-trait analysis of GWAS (MTAG) with external summary statistics for BD (59,287 cases; 781,022 controls) and schizophrenia (SCZ; 53,386 cases; 77,258 controls). Sample overlap was statistically accounted for. Main Outcomes and Measures:The primary outcomes were the genetic dimensions underlying BD heterogeneity, differentiated by single nucleotide polymorphism (SNP)-heritability (h 2 SNP ), genetic correlations, genomic loci ( P ≤5×10 -8 ), and functional, cell-type, and gene-expression pathway analyses. Results:We identified four genetically-informed dimensions of BD: Severe Illness, Core Mania, Externalizing/Impulsive Comorbidity, and Internalizing/Affective Comorbidity. The analyses yielded up to 181 subphenotype-associated loci, 53 of which are novel. The Severe Illness Dimension was characterized by a unique neuro-immune signature (a protective association with HLA-DMB , P =2.50×10 -273 ) evident only when leveraging SCZ genetic data. The Internalizing/Affective dimension was associated with neurodevelopmental genes (e.g., DCC ). Notably, the rapid-cycling subphenotype showed a unique signature of strong negative selection, a finding not observed in other subphenotypes. Conclusions and Relevance:The clinical heterogeneity of bipolar disorder appears to be defined by a complex and multi-layered genetic architecture. The presented findings provide an empirical framework that may advance psychiatric nosology beyond its current diagnostic boundaries. These results may also inform future research to identify targets for personalized interventions. The delineation of these genetically-informed dimensions offers specific, biologically-grounded hypotheses for subsequent therapeutic discovery. Establishing such a framework is an essential step toward refining diagnostic criteria and developing more effective, personalized treatments. This work lays the foundation for a transition from a uniform treatment model to the paradigm of precision psychiatry. Key Points:Question: What are the distinct genetic architectures underlying the clinical heterogeneity of bipolar disorder?Findings: In this genetic study of 23,819 bipolar disorder (BD) cases and 163,839 controls, clinical heterogeneity mapped onto four genetically-informed dimensions. A severe illness dimension was defined by a neuro-immune signature ( HLA-DMB ) shared with schizophrenia. An affective comorbidity dimension was distinguished by neurodevelopmental pathways involving axonal guidance ( DCC ). Notably, the rapid-cycling phenotype showed evidence of purifying selection, suggesting influence by rare, highly penetrant alleles. Meaning: These findings provide a data-driven biological framework for bipolar disorder, guiding future research toward patient stratification and targeted therapeutics.

Perinatal depression (PND) affects up to 20% of women and is associated with significant impairment and disability in affected women. In addition, perinatal depression is associated with broader public health and multigenerational consequences. Innovative approaches are needed to reduce the burden of perinatal depression through identification, tracking, and treatment of depressive symptoms during the perinatal period. This study is a randomized clinical trial comparing the relative efficacy of a multi-tiered system of care, Screening and Treatment of Anxiety and Depression (STAND) to perinatal care delivered by a reproductive psychiatrist in reducing symptoms of depression and anxiety. A sample of 167 individuals was randomized between week 28 of pregnancy and 6 months postpartum. A secondary aim compares the original online therapy intervention used in the first half of the study to a newer online therapy program used in the second half of the study for individuals assigned to the STAND treatment. The study measures, intervention groups, and analysis methods are described, as well as expected implications. The findings from this study may improve the methods for tracking symptom changes over time, monitoring treatment response, and providing personalized care for individuals with PND. As such, this study may improve the lives of patients with PND and their families and lower the related health care costs to society.Trial registration NCT: 9/24/2021NCT direct link: https://www.clinicaltrials.gov/study/NCT05056454?term=NCT05056454&rank=1&a=1.

Over the last ten years, there has been considerable progress in using digital behavioral phenotypes, captured passively and continuously from smartphones and wearable devices, to infer depressive mood. However, most digital phenotype studies suffer from poor replicability, often fail to detect clinically relevant events, and use measures of depression that are not validated or suitable for collecting large and longitudinal data. Here, we report high-quality longitudinal validated assessments of depressive mood from computerized adaptive testing paired with continuous digital assessments of behavior from smartphone sensors for up to 40 weeks on 183 individuals experiencing mild to severe symptoms of depression. We apply a combination of cubic spline interpolation and idiographic models to generate individualized predictions of future mood from the digital behavioral phenotypes, achieving high prediction accuracy of depression severity up to three weeks in advance (R2 ≥ 80%) and a 65.7% reduction in the prediction error over a baseline model which predicts future mood based on past depression severity alone. Finally, our study verified the feasibility of obtaining high-quality longitudinal assessments of mood from a clinical population and predicting symptom severity weeks in advance using passively collected digital behavioral data. Our results indicate the possibility of expanding the repertoire of patient-specific behavioral measures to enable future psychiatric research.

Experimental cognitive tests are designed to measure particular cognitive domains, although evidence supporting test validity is often limited. The Consortium for Neuropsychiatric Phenomics test battery administered 23 experimental and traditional neuropsychological tests to a large sample of community volunteers (n = 1,059) and patients with psychiatric diagnoses (n = 137), providing a unique opportunity to examine convergent validity with factor analysis. Traditional tests included subtests from the Wechsler and Delis-Kaplan batteries, while experimental tests included the Attention Networks Test, Balloon Analogue Risk Task, Delay Discounting Task, Remember-Know, Reversal Learning Task, Scene Recognition, Spatial and Verbal Capacity and Manipulation Tasks, Stop-Signal Task, and Task Switching. Several experimental cognitive measures were insufficiently related to other tests and were excluded from factor analyses. In the remaining 18 tests, exploratory factor analysis and subsequent multigroup confirmatory factor analysis supported a three-factor structure broadly corresponding to domains of verbal/working memory, inhibitory control, and memory. In sum, several experimental measures of inhibitory control had weak relationships with all other tests, while the convergent validity of most tests of working memory and memory was supported.

OBJECTIVES:Most bipolar disorder (BD) patients initially present with depressive symptoms, resulting in a delayed diagnosis of BD and poor clinical outcomes. This study aims to identify features predictive of the conversion from Major Depressive Disorder (MDD) to BD by leveraging electronic health record (EHR) data from the Clínica San Juan de Dios Manizales in Colombia. METHODS:We employed a multivariable Cox regression model to identify important predictors of conversion from MDD to BD. RESULTS:Analyzing 15 years of EHR data from 13,607 patients diagnosed with MDD, a total of 1610 (11.8%) transitioned to BD. Predictive features of the conversion to BD included severity of the initial MDD episode, presence of psychosis and hospitalization at first episode, family history of BD, and female gender. Additionally, we observed associations with medication classes (positive associations with prescriptions of mood stabilizers, antipsychotics, and negative associations with antidepressants) and a positive association with suicidality, a feature derived from natural language processing (NLP) of clinical notes. Together, these risk factors predicted BD conversion within 5 years of the initial MDD diagnosis, with a recall of 72% and a precision of 38%. CONCLUSIONS:Our study confirms previously identified risk factors identified through registry-based studies (female gender and psychotic depression at the index MDD episode) and identifies novel ones (suicidality extracted from clinical notes). These results simultaneously demonstrate the validity of using EHR data for predicting BD conversion and underscore its potential for the identification of novel risk factors, thereby improving early diagnosis.

Digital sensing tools, like smartphones and wearables, offer transformative potential for mental health research by enabling scalable, longitudinal data collection. Realizing this promise requires overcoming significant challenges including limited data standards, underpowered studies, and a disconnect between research aims and community needs. This report, based on the 2023 Workshop on Advancing Digital Sensing Tools for Mental Health, articulates strategies to address these challenges to ensure rigorous, equitable, and impactful research.