Mei-Hua Hall
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Mei-Hua Hall has a total of 223 co-authors. Here are the top co-authors by number of shared publications:
- Robin MacGregor Murray
Department of Psychosis Studies, Institute of Psychiatry, Psychology and Neuroscience, King’s College London, London, UK
11 shared publications Dost Öngür
10 shared publications- Elvira Bramon
Division of Psychiatry, Department of Mental Health Neuroscience, University College London, London, UK
10 shared publications Deborah L. Levy
Psychology Research Laboratory, McLean Hospital, Harvard Medical School, Belmont, USA
6 shared publicationsEben Holderness
Psychosis Neurobiology Laboratory, McLean Hospital, Belmont, USA
6 shared publications- Fruhling Rijsdijk
Psychology Department, Faculty of Social Sciences, Anton de Kom University, Paramaribo, Suriname
6 shared publications
Latest Publications
- 14 August 2025
- Published by Springer Nature in Neuroinformatics
- Vol. 23 (3) , 43
Latent Growth Models of Longitudinal Changes in Functional Connectivity during Early Stage Psychosis
Resting state functional magnetic resonance imaging (fMRI) is a useful technique to characterize functional connectivity patterns between regions of the brain, based on the Fisher-transformed Pearson correlations in the BOLD signal. Pinpointing how connectivity patterns change in neuropathies like early-stage psychosis (ESP) can help understand the disorders and track progression. Using study data from 21 ESP subjects with complete data for three consecutive scans, we examined connectivity changes throughout the whole brain with a region of interest (ROI) to ROI-based approach for ROI defined by the Harvard-Oxford cortical and subcortical atlases, supplemented by the AAL atlas for the cerebellum, and by networks defined by the CONN toolbox independent component analysis of the Human Connectome Project. We applied latent growth modelling, which is a type of structural equation modelling, to these connectivity measurements across baseline and follow-up visits. The models use age, community functioning, and negative symptoms at baselines as the covariates for subject-specific slope and intercept of the longitudinal measurements. After stringent thresholding cutoffs of root mean square error of approximation, standardized root mean square residual, comparative fit index, and Benjamini-Hochberg corrected p-value, we found a subset of connectivity measurements with significant longitudinal slopes (N = 18 atlas, N = 6 network), and used the subject's slope estimates to stratify these subjects into three clusters based on how the ROI-to-ROI correlations of functional connectivity change over time. The connections with significant slopes include atlas level regions like the temporal lobe, fronto-parietal lobe, and cerebellum, and network level patterns like the DMN, FPN, and Salience Networks. The structural equation modelling approach identifies ROIs whose functional connectivity changes over time, indicating the ROIs most dynamic during ESP. This highlights the utility of latent growth models for the analysis of longitudinal functional connectivity measures across the whole brain with relatively small sample sizes.
- 14 August 2025
- Published by Springer Nature in Neuroinformatics
- Vol. 23 (3) , 43
Latent Growth Models of Longitudinal Changes in Functional Connectivity during Early Stage Psychosis
Resting state functional magnetic resonance imaging (fMRI) is a useful technique to characterize functional connectivity patterns between regions of the brain, based on the Fisher-transformed Pearson correlations in the BOLD signal. Pinpointing how connectivity patterns change in neuropathies like early-stage psychosis (ESP) can help understand the disorders and track progression. Using study data from 21 ESP subjects with complete data for three consecutive scans, we examined connectivity changes throughout the whole brain with a region of interest (ROI) to ROI-based approach for ROI defined by the Harvard-Oxford cortical and subcortical atlases, supplemented by the AAL atlas for the cerebellum, and by networks defined by the CONN toolbox independent component analysis of the Human Connectome Project. We applied latent growth modelling, which is a type of structural equation modelling, to these connectivity measurements across baseline and follow-up visits. The models use age, community functioning, and negative symptoms at baselines as the covariates for subject-specific slope and intercept of the longitudinal measurements. After stringent thresholding cutoffs of root mean square error of approximation, standardized root mean square residual, comparative fit index, and Benjamini-Hochberg corrected p-value, we found a subset of connectivity measurements with significant longitudinal slopes (N = 18 atlas, N = 6 network), and used the subject's slope estimates to stratify these subjects into three clusters based on how the ROI-to-ROI correlations of functional connectivity change over time. The connections with significant slopes include atlas level regions like the temporal lobe, fronto-parietal lobe, and cerebellum, and network level patterns like the DMN, FPN, and Salience Networks. The structural equation modelling approach identifies ROIs whose functional connectivity changes over time, indicating the ROIs most dynamic during ESP. This highlights the utility of latent growth models for the analysis of longitudinal functional connectivity measures across the whole brain with relatively small sample sizes.
- Open Access
- 1 August 2025
- Published by Springer Nature in Translational Psychiatry
- Vol. 15 (1) , 1-8
Machine learning in psychiatric health records: A gold standard approach to trauma annotation
Psychiatric electronic health records present unique challenges for machine learning due to their unstructured, complex, and variable nature. This study aimed to create a gold standard dataset by identifying a cohort of patients with psychotic disorders and posttraumatic stress disorder, (PTSD), developing clinically-informed guidelines for annotating traumatic events in their health records and to create a gold standard publicly available dataset, and demonstrating the dataset’s suitability for training machine learning models to detect indicators of symptoms, substance use, and trauma in new records. We compiled a representative corpus of 200 narrative heavy health records (470,489 tokens) from a centralized database and developed a detailed annotation scheme with a team of clinical experts and computational linguistics. Clinicians annotated the corpus for trauma-related events and relevant clinical information with high inter-annotator agreement (0.715 for entity/span tags and 0.874 for attributes). Additionally, machine learning models were developed to demonstrate practical viability of the gold standard corpus for machine learning applications, achieving a micro F1 score of 0.76 and 0.82 for spans and attributes respectively, indicative of their predictive reliability. This study established the first gold-standard dataset for the complex task of labelling traumatic features in psychiatric health records. High inter-annotator agreement and model performance illustrate its utility in advancing the application of machine learning in psychiatric healthcare in order to better understand disease heterogeneity and treatment implications.
- 1 August 2025
- Published by Springer Nature in Translational Psychiatry
- Vol. 15 (1) , 1-8
Open AccessMachine learning in psychiatric health records: A gold standard approach to trauma annotation
Psychiatric electronic health records present unique challenges for machine learning due to their unstructured, complex, and variable nature. This study aimed to create a gold standard dataset by identifying a cohort of patients with psychotic disorders and posttraumatic stress disorder, (PTSD), developing clinically-informed guidelines for annotating traumatic events in their health records and to create a gold standard publicly available dataset, and demonstrating the dataset’s suitability for training machine learning models to detect indicators of symptoms, substance use, and trauma in new records. We compiled a representative corpus of 200 narrative heavy health records (470,489 tokens) from a centralized database and developed a detailed annotation scheme with a team of clinical experts and computational linguistics. Clinicians annotated the corpus for trauma-related events and relevant clinical information with high inter-annotator agreement (0.715 for entity/span tags and 0.874 for attributes). Additionally, machine learning models were developed to demonstrate practical viability of the gold standard corpus for machine learning applications, achieving a micro F1 score of 0.76 and 0.82 for spans and attributes respectively, indicative of their predictive reliability. This study established the first gold-standard dataset for the complex task of labelling traumatic features in psychiatric health records. High inter-annotator agreement and model performance illustrate its utility in advancing the application of machine learning in psychiatric healthcare in order to better understand disease heterogeneity and treatment implications.
- 20 May 2025
- Published by Springer Nature in Molecular Psychiatry
- p. 1-10
Towards clinical subtypes in schizophrenia: integrating cognitive, functional, and digital phenotyping assessments
Heterogeneity in the clinical presentation of schizophrenia impairs both proper and preventative care. The digital phenotyping data gathered from an international multi-site cohort study in people with schizophrenia (SZ) offers a novel opportunity to explore clinically meaningful subtypes in the context of clinical, functional, and cognitive data. Using a set of behavioral features derived from smartphone digital phenotyping, clinical assessment of symptoms including PANSS, clinical assessment of cognition with BACS, and clinical assessment of functioning with the social functioning assessments over the target period of twelve months, we found that the international cohort of 74 patients were categorized into three well-defined clusters that suggest clinically actionable targets from differential correlations in each. Namely, the identified clusters seemed to share phenotypic traits with the affective psychosis with more severe symptomatic presentation, a non-affective SZ with functional impairment, and a higher functioning non-affective SZ cluster. Partial correlation analysis further highlighted the emergence of different features per cluster, where anxiety symptoms were most notable for one group, whereas psychotic symptoms were most notable for the other two. Importantly, we showcase an analysis pipeline that transparently addresses challenges of missing data and potential skew so that this research methodology can be applied to future prospective validation studies. This study hopes to build a foundation for future digital phenotyping clustering work by scaling up to new sites, and populations to uncover the nature and extent of heterogeneity in schizophrenia.
- 20 May 2025
- Published by Springer Nature in Molecular Psychiatry
- p. 1-10
Towards clinical subtypes in schizophrenia: integrating cognitive, functional, and digital phenotyping assessments
Heterogeneity in the clinical presentation of schizophrenia impairs both proper and preventative care. The digital phenotyping data gathered from an international multi-site cohort study in people with schizophrenia (SZ) offers a novel opportunity to explore clinically meaningful subtypes in the context of clinical, functional, and cognitive data. Using a set of behavioral features derived from smartphone digital phenotyping, clinical assessment of symptoms including PANSS, clinical assessment of cognition with BACS, and clinical assessment of functioning with the social functioning assessments over the target period of twelve months, we found that the international cohort of 74 patients were categorized into three well-defined clusters that suggest clinically actionable targets from differential correlations in each. Namely, the identified clusters seemed to share phenotypic traits with the affective psychosis with more severe symptomatic presentation, a non-affective SZ with functional impairment, and a higher functioning non-affective SZ cluster. Partial correlation analysis further highlighted the emergence of different features per cluster, where anxiety symptoms were most notable for one group, whereas psychotic symptoms were most notable for the other two. Importantly, we showcase an analysis pipeline that transparently addresses challenges of missing data and potential skew so that this research methodology can be applied to future prospective validation studies. This study hopes to build a foundation for future digital phenotyping clustering work by scaling up to new sites, and populations to uncover the nature and extent of heterogeneity in schizophrenia.
- Open Access
- 9 May 2025
- Published by Oxford University Press (OUP) in Schizophrenia Bulletin
- Vol. 52 (1)
Transdiagnostic Neurocognitive Endophenotypes for Schizophrenia, Bipolar I Disorder and a Broad Psychosis/Bipolar I Disorder Phenotype: A Mega-Analysis of Twin and Sibling Data
Background: Psychiatric research is increasingly embracing a paradigm shift from categorical diagnoses to neurobiologically meaningful dimensions that cross current diagnostic boundaries. This transposition calls for redefining endophenotypes to accommodate transdiagnostic vulnerabilities. We sought to identify shared and disorder-specific neurocognitive endophenotypes for schizophrenia, bipolar I disorder (BD-I) and a broad psychosis/BD-I phenotype in a mega-analysis of twin/sibling data. Study Design: We performed genetic model fitting to intelligence (IQ) and computerised neurocognitive data derived from 1050 twins/siblings from three research centres in the UK, Denmark and the Netherlands, affected (n = 257) or unaffected (n = 793) by schizophrenia, other primary psychoses and BD-I. We examined the endophenotypic status of IQ, spatial working memory (SWM), visual recognition, sustained attention/rapid visual processing (RVP), mental flexibility, and spatial planning/problem solving (all validated as endophenotypes for schizophrenia in previous studies) in relation to schizophrenia, BD-I and the broad phenotype. Study Results: After covarying for age, gender, education and research centre, IQ and SWM emerged as transdiagnostic endophenotypes, showing statistically significant heritabilities (h2 67–75% and 28–30%, respectively), phenotypic correlations (rph |0.14|-|0.25|) and genetic correlations (rg |0.18|-|0.42|) with all diagnostic phenotypes. Additionally, all remaining cognitive domains received validation as endophenotypes for the broad phenotype, and all, but RVP, for schizophrenia. Conclusions: IQ and SWM tap into transdiagnostic elements of the genetic vulnerabilities to psychosis and BD-I. Our findings add to emergent evidence which spurs cautious optimism that a psychiatric nosology based on aetiology rather than phenotypical classifications may be feasible in the future, enabling biotyping and novel approaches to treatment.
- 9 May 2025
- Published by Oxford University Press (OUP) in Schizophrenia Bulletin
- Vol. 52 (1)
Open AccessTransdiagnostic Neurocognitive Endophenotypes for Schizophrenia, Bipolar I Disorder and a Broad Psychosis/Bipolar I Disorder Phenotype: A Mega-Analysis of Twin and Sibling Data
Background: Psychiatric research is increasingly embracing a paradigm shift from categorical diagnoses to neurobiologically meaningful dimensions that cross current diagnostic boundaries. This transposition calls for redefining endophenotypes to accommodate transdiagnostic vulnerabilities. We sought to identify shared and disorder-specific neurocognitive endophenotypes for schizophrenia, bipolar I disorder (BD-I) and a broad psychosis/BD-I phenotype in a mega-analysis of twin/sibling data. Study Design: We performed genetic model fitting to intelligence (IQ) and computerised neurocognitive data derived from 1050 twins/siblings from three research centres in the UK, Denmark and the Netherlands, affected (n = 257) or unaffected (n = 793) by schizophrenia, other primary psychoses and BD-I. We examined the endophenotypic status of IQ, spatial working memory (SWM), visual recognition, sustained attention/rapid visual processing (RVP), mental flexibility, and spatial planning/problem solving (all validated as endophenotypes for schizophrenia in previous studies) in relation to schizophrenia, BD-I and the broad phenotype. Study Results: After covarying for age, gender, education and research centre, IQ and SWM emerged as transdiagnostic endophenotypes, showing statistically significant heritabilities (h2 67–75% and 28–30%, respectively), phenotypic correlations (rph |0.14|-|0.25|) and genetic correlations (rg |0.18|-|0.42|) with all diagnostic phenotypes. Additionally, all remaining cognitive domains received validation as endophenotypes for the broad phenotype, and all, but RVP, for schizophrenia. Conclusions: IQ and SWM tap into transdiagnostic elements of the genetic vulnerabilities to psychosis and BD-I. Our findings add to emergent evidence which spurs cautious optimism that a psychiatric nosology based on aetiology rather than phenotypical classifications may be feasible in the future, enabling biotyping and novel approaches to treatment.
- Open Access
Machine Learning in Psychiatric Health Records: A Gold Standard Approach to Trauma Annotation
Psychiatric electronic health records present unique challenges for machine learning due to their unstructured, complex, and variable nature. This study aimed to create a gold standard dataset by identifying a cohort of patients with psychotic disorders and posttraumatic stress disorder, (PTSD), developing clinically-informed guidelines for annotating traumatic events in their health records and to create a gold standard publicly available dataset, and demonstrating the dataset’s suitability for training machine learning models to detect indicators of symptoms, substance use, and trauma in new records. We compiled a representative corpus of 200 narrative heavy health records (470,489 tokens) from a centralized database and developed a detailed annotation scheme with a team of clinical experts and computational linguistics. Clinicians annotated the corpus for trauma-related events and relevant clinical information with high inter-annotator agreement (0.715 for entity/span tags and 0.874 for attributes). Additionally, machine learning models were developed to demonstrate practical viability of the gold standard corpus for machine learning applications, achieving a micro F1 score of 0.76 and 0.82 for spans and attributes respectively, indicative of their predictive reliability. This study established the first gold-standard dataset for the complex task of labelling traumatic features in psychiatric health records. High inter-annotator agreement and model performance illustrate its utility in advancing the application of machine learning in psychiatric healthcare in order to better understand disease heterogeneity and treatment implications.
Open AccessMachine Learning in Psychiatric Health Records: A Gold Standard Approach to Trauma Annotation
Psychiatric electronic health records present unique challenges for machine learning due to their unstructured, complex, and variable nature. This study aimed to create a gold standard dataset by identifying a cohort of patients with psychotic disorders and posttraumatic stress disorder, (PTSD), developing clinically-informed guidelines for annotating traumatic events in their health records and to create a gold standard publicly available dataset, and demonstrating the dataset’s suitability for training machine learning models to detect indicators of symptoms, substance use, and trauma in new records. We compiled a representative corpus of 200 narrative heavy health records (470,489 tokens) from a centralized database and developed a detailed annotation scheme with a team of clinical experts and computational linguistics. Clinicians annotated the corpus for trauma-related events and relevant clinical information with high inter-annotator agreement (0.715 for entity/span tags and 0.874 for attributes). Additionally, machine learning models were developed to demonstrate practical viability of the gold standard corpus for machine learning applications, achieving a micro F1 score of 0.76 and 0.82 for spans and attributes respectively, indicative of their predictive reliability. This study established the first gold-standard dataset for the complex task of labelling traumatic features in psychiatric health records. High inter-annotator agreement and model performance illustrate its utility in advancing the application of machine learning in psychiatric healthcare in order to better understand disease heterogeneity and treatment implications.
- 10 March 2025
- Published by Elsevier in Schizophrenia Research
- Vol. 277, 159-168
Unveiling distinct representations of P3a in schizophrenia through two-stimulus and three-stimulus auditory oddball paradigms
- 10 March 2025
- Published by Elsevier in Schizophrenia Research
- Vol. 277, 159-168
Unveiling distinct representations of P3a in schizophrenia through two-stimulus and three-stimulus auditory oddball paradigms
- Open Access
- 22 January 2025
- Published by JMIR Publications Inc. in JMIR Mental Health
- Vol. 12 (1) , e71630
Cross-site predictions of readmission after psychiatric hospitalization with mood or psychotic disorders (Preprint)
Background: Patients with mood or psychotic disorders experience high rates of unplanned hospital readmissions. Predicting the likelihood of readmission can guide discharge decisions and optimize patient care. Objective: The purpose of this study is to evaluate the predictive power of structured variables from electronic health records (EHRs) for all-cause readmission across multiple sites within the Mass General Brigham (MGB) health systems and to assess the transportability of prediction models between sites. Methods: This retrospective, multi-site study analyzed structured variables from EHRs separately for each site to develop in-site prediction models. The transportability of these models was evaluated by applying them across different sites. The predictive performance was measured using the F1 score, and additional adjustments were made to account for differences in predictor distributions. Results: The study found that the relevant predictors of readmission varied significantly across sites. For instance, the length of stay was a strong predictor at only three of the four sites. In-site prediction models achieved an average F1 score of 0.661, whereas cross-site predictions resulted in a lower average F1 score of 0.616. Efforts to improve transportability by adjusting for differences in predictor distributions did not lead to better performance. Conclusions: The findings indicate that individual site-specific models are necessary to achieve reliable prediction accuracy. Furthermore, the results suggest that the current set of predictors may be insufficient for cross-site model transportability, highlighting the need for more advanced predictor variables and predictive algorithms to gain robust insights into the factors influencing early psychiatric readmissions.
- 22 January 2025
- Published by JMIR Publications Inc. in JMIR Mental Health
- Vol. 12 (1) , e71630
Open AccessCross-site predictions of readmission after psychiatric hospitalization with mood or psychotic disorders (Preprint)
Background: Patients with mood or psychotic disorders experience high rates of unplanned hospital readmissions. Predicting the likelihood of readmission can guide discharge decisions and optimize patient care. Objective: The purpose of this study is to evaluate the predictive power of structured variables from electronic health records (EHRs) for all-cause readmission across multiple sites within the Mass General Brigham (MGB) health systems and to assess the transportability of prediction models between sites. Methods: This retrospective, multi-site study analyzed structured variables from EHRs separately for each site to develop in-site prediction models. The transportability of these models was evaluated by applying them across different sites. The predictive performance was measured using the F1 score, and additional adjustments were made to account for differences in predictor distributions. Results: The study found that the relevant predictors of readmission varied significantly across sites. For instance, the length of stay was a strong predictor at only three of the four sites. In-site prediction models achieved an average F1 score of 0.661, whereas cross-site predictions resulted in a lower average F1 score of 0.616. Efforts to improve transportability by adjusting for differences in predictor distributions did not lead to better performance. Conclusions: The findings indicate that individual site-specific models are necessary to achieve reliable prediction accuracy. Furthermore, the results suggest that the current set of predictors may be insufficient for cross-site model transportability, highlighting the need for more advanced predictor variables and predictive algorithms to gain robust insights into the factors influencing early psychiatric readmissions.
- Open Access
- 1 January 2025
- Published by Association for Computational Linguistics (ACL)
- Vol. 2025, 28025-28042
Aspect-Oriented Summarization for Psychiatric Short-Term Readmission Prediction
WonJin Yoon, Boyu Ren, Spencer Thomas, Chanhwi Kim, Guergana K Savova, Mei-Hua Hall, Timothy A. Miller. Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing. 2025.
- 1 January 2025
- Published by Association for Computational Linguistics (ACL)
- Vol. 2025, 28025-28042
Open AccessAspect-Oriented Summarization for Psychiatric Short-Term Readmission Prediction
WonJin Yoon, Boyu Ren, Spencer Thomas, Chanhwi Kim, Guergana K Savova, Mei-Hua Hall, Timothy A. Miller. Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing. 2025.
- Open Access
- 9 December 2024
- Published by Cambridge University Press (CUP) in Acta Neuropsychiatrica
- Vol. 37, 1-30
Uncovering key predictive channels and clinical variables in the gamma band auditory steady-state response in early-stage psychosis: a longitudinal study
Objective: Psychotic disorders are characterised by abnormalities in the synchronisation of neuronal responses. A 40 Hz gamma band deficit during auditory steady-state response (ASSR) measured by electroencephalogram (EEG) is a robust observation in psychosis and is associated with symptoms and functional deficits. However, the majority of ASSR studies focus on specific electrode sites, while whole scalp analysis using all channels, and the association with clinical symptoms, are rare. Methods: In this study, we use whole-scalp 40 Hz ASSR EEG measurements – power and phase-locking factor – to establish deficits in early-stage psychosis (ESP) subjects, classify ESP status using an ensemble of machine learning techniques, identify correlates with principal components obtained from clinical/demographic/functioning variables, and correlate functional outcome after a short-term follow-up. Results: We identified significant spatially-distributed group level differences for power and phase locking. The performance of different machine learning techniques and interpretation of the extracted feature importance indicate that phase locking has a more predictive and parsimonious pattern than power. Phase locking is also associated with principal components composed of measures of cognitive processes. Short-term functional outcome is associated with baseline 40 Hz ASSR signals from the FCz and other channels in both phase locking and power. Conclusion: This whole-scalp EEG study provides additional evidence to link deficits in 40 Hz ASSRs with cognition and functioning in ESP, and corroborates with prior studies of phase locking from a subset of EEG channels. Confirming 40 Hz ASSR deficits serves as a candidate phenotype to identify circuit dysfunctions and a biomarker for clinical outcomes in psychosis.
- 9 December 2024
- Published by Cambridge University Press (CUP) in Acta Neuropsychiatrica
- Vol. 37, 1-30
Open AccessUncovering key predictive channels and clinical variables in the gamma band auditory steady-state response in early-stage psychosis: a longitudinal study
Objective: Psychotic disorders are characterised by abnormalities in the synchronisation of neuronal responses. A 40 Hz gamma band deficit during auditory steady-state response (ASSR) measured by electroencephalogram (EEG) is a robust observation in psychosis and is associated with symptoms and functional deficits. However, the majority of ASSR studies focus on specific electrode sites, while whole scalp analysis using all channels, and the association with clinical symptoms, are rare. Methods: In this study, we use whole-scalp 40 Hz ASSR EEG measurements – power and phase-locking factor – to establish deficits in early-stage psychosis (ESP) subjects, classify ESP status using an ensemble of machine learning techniques, identify correlates with principal components obtained from clinical/demographic/functioning variables, and correlate functional outcome after a short-term follow-up. Results: We identified significant spatially-distributed group level differences for power and phase locking. The performance of different machine learning techniques and interpretation of the extracted feature importance indicate that phase locking has a more predictive and parsimonious pattern than power. Phase locking is also associated with principal components composed of measures of cognitive processes. Short-term functional outcome is associated with baseline 40 Hz ASSR signals from the FCz and other channels in both phase locking and power. Conclusion: This whole-scalp EEG study provides additional evidence to link deficits in 40 Hz ASSRs with cognition and functioning in ESP, and corroborates with prior studies of phase locking from a subset of EEG channels. Confirming 40 Hz ASSR deficits serves as a candidate phenotype to identify circuit dysfunctions and a biomarker for clinical outcomes in psychosis.
- Open Access
- 19 September 2024
- Published by Oxford University Press (OUP) in SLEEP
- Vol. 48 (2)
A spectrum of altered non-rapid eye movement sleep in schizophrenia
Multiple facets of sleep neurophysiology, including electroencephalography (EEG) metrics such as non-rapid eye movement (NREM) spindles and slow oscillations, are altered in individuals with schizophrenia (SCZ). However, beyond group-level analyses, the extent to which NREM deficits vary among patients is unclear, as are their relationships to other sources of heterogeneity including clinical factors, ageing, cognitive profiles and medication regimens. Using newly collected high-density sleep EEG data on 103 individuals with SCZ and 68 controls, we first sought to replicate our previously reported group-level differences between patients and controls (original N=130) during N2 stage. Then in the combined sample (N=301 including 175 patients), we characterized patient-to-patient variability. We replicated all group-level mean differences and confirmed the high accuracy of our predictive model (AUC=0.93 for diagnosis). Compared to controls, patients showed significantly increased between-individual variability across many (26%) sleep metrics. Although multiple clinical and cognitive factors were associated with NREM metrics, collectively they did not account for much of the general increase in patient-to-patient variability. Medication regimen was a greater contributor to variability. Some sleep metrics including fast spindle density showed exaggerated age-related effects in SCZ, and patients exhibited older predicted biological ages based on the sleep EEG; further, among patients, certain medications exacerbated these effects, in particular olanzapine. Collectively, our results point to a spectrum of N2 sleep deficits among SCZ patients that can be measured objectively and at scale, with relevance to both the etiological heterogeneity of SCZ as well as potential iatrogenic effects of antipsychotic medication.
- 19 September 2024
- Published by Oxford University Press (OUP) in SLEEP
- Vol. 48 (2)
Open AccessA spectrum of altered non-rapid eye movement sleep in schizophrenia
Multiple facets of sleep neurophysiology, including electroencephalography (EEG) metrics such as non-rapid eye movement (NREM) spindles and slow oscillations, are altered in individuals with schizophrenia (SCZ). However, beyond group-level analyses, the extent to which NREM deficits vary among patients is unclear, as are their relationships to other sources of heterogeneity including clinical factors, ageing, cognitive profiles and medication regimens. Using newly collected high-density sleep EEG data on 103 individuals with SCZ and 68 controls, we first sought to replicate our previously reported group-level differences between patients and controls (original N=130) during N2 stage. Then in the combined sample (N=301 including 175 patients), we characterized patient-to-patient variability. We replicated all group-level mean differences and confirmed the high accuracy of our predictive model (AUC=0.93 for diagnosis). Compared to controls, patients showed significantly increased between-individual variability across many (26%) sleep metrics. Although multiple clinical and cognitive factors were associated with NREM metrics, collectively they did not account for much of the general increase in patient-to-patient variability. Medication regimen was a greater contributor to variability. Some sleep metrics including fast spindle density showed exaggerated age-related effects in SCZ, and patients exhibited older predicted biological ages based on the sleep EEG; further, among patients, certain medications exacerbated these effects, in particular olanzapine. Collectively, our results point to a spectrum of N2 sleep deficits among SCZ patients that can be measured objectively and at scale, with relevance to both the etiological heterogeneity of SCZ as well as potential iatrogenic effects of antipsychotic medication.