A highly scalable deep learning language model for common risks prediction among psychiatric inpatients

Abstract Background There is a lack of studies exploring the performance of Transformers-based language models in common risks assessment among psychiatric inpatients. We aim to develop a scalable risk assessment model using multidimensional textualized data and test the stability, robustness, and b...

Full description

Saved in:
Bibliographic Details
Main Authors: Enzhao Zhu, Jiayi Wang, Guoquan Zhou, Chunbo Li, Fazhan Chen, Kang Ju, Liangliang Chen, Yichao Yin, Yi Chen, Yanping Zhang, Xu Zhang, Xinlin Zhou, Zongyuan Wang, Jianping Qiu, Hui Wang, Weizhong Shi, Feng Wang, Dong Wang, Zhihao Chen, Jiaojiao Hou, Hui Li, Zisheng Ai
Format: Article
Language:English
Published: BMC 2025-05-01
Series:BMC Medicine
Subjects:
Online Access:https://doi.org/10.1186/s12916-025-04150-7
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1850231244348981248
author Enzhao Zhu
Jiayi Wang
Guoquan Zhou
Chunbo Li
Fazhan Chen
Kang Ju
Liangliang Chen
Yichao Yin
Yi Chen
Yanping Zhang
Xu Zhang
Xinlin Zhou
Zongyuan Wang
Jianping Qiu
Hui Wang
Weizhong Shi
Feng Wang
Dong Wang
Zhihao Chen
Jiaojiao Hou
Hui Li
Zisheng Ai
author_facet Enzhao Zhu
Jiayi Wang
Guoquan Zhou
Chunbo Li
Fazhan Chen
Kang Ju
Liangliang Chen
Yichao Yin
Yi Chen
Yanping Zhang
Xu Zhang
Xinlin Zhou
Zongyuan Wang
Jianping Qiu
Hui Wang
Weizhong Shi
Feng Wang
Dong Wang
Zhihao Chen
Jiaojiao Hou
Hui Li
Zisheng Ai
author_sort Enzhao Zhu
collection DOAJ
description Abstract Background There is a lack of studies exploring the performance of Transformers-based language models in common risks assessment among psychiatric inpatients. We aim to develop a scalable risk assessment model using multidimensional textualized data and test the stability, robustness, and benefit of this approach. Methods In this real-world cohort study, a deep learning language model was developed and validated using first hospitalized cases diagnosed with schizophrenia, bipolar disorder, and depressive disorder between January 2016 and March 2023 in three hospitals. The algorithm was externally validated on an independent testing cohort comprising 1180 patients. A total of 140 features, including first medical records (FMR), laboratory examinations, medical orders, and psychological scales, were assessed for analysis. The outcomes were short- and long-term impulsivity (STI and LTI), risk of suicide (STSS and LTSS), and need of physical restraint (STPR and LTPR) assessed by qualified nurses or clinicians. Analysis was carried out between August 2024 and June 2024. Models with different architectures and input settings were compared with each other. The area under the receiver operating characteristic curve (AUROC) was used to assess the primary performance of models. The clinical utility was determined by the net benefit under Youden’s threshold. Results Of 7451 patients included in this study, 2982 (47.6%) were male, and the median (interquartile range) age was 42 (28–57) years. The overall incidence of outcomes was 635 (8.5%), 728 (10.5%), 659 (8.8%), 803 (10.8%), 588 (7.9%), and 728 (9.8%) for STPR, LTPR, STSS, LTSS, STI, and LTI, respectively. The multitask semi-structured Transformers-based language (SSTL) model showed more promising AUROCs (STPR: 0.915; LTPR: 0.844; STSS: 0.867; LTSS: 0.879; STI: 0.899; LTI: 0.894) in the prediction of these outcomes than single-tasked or multimodal language models and traditional structured data models. Combining FMR with other data from electronic health records led to significant improvements in the performance and clinical utility of SSTL models based on demographic, diagnosis, laboratory tests, treatment, and psychological scales. Conclusions The SSTL model shows potential advantages in prognostic evaluation. FMR is a strong predictor for common risks prediction and may benefit other tasks in psychiatry with minimum requirements for data and data processing.
format Article
id doaj-art-1005aeebd6f74eefbf7e44025652b70b
institution OA Journals
issn 1741-7015
language English
publishDate 2025-05-01
publisher BMC
record_format Article
series BMC Medicine
spelling doaj-art-1005aeebd6f74eefbf7e44025652b70b2025-08-20T02:03:36ZengBMCBMC Medicine1741-70152025-05-0123111510.1186/s12916-025-04150-7A highly scalable deep learning language model for common risks prediction among psychiatric inpatientsEnzhao Zhu0Jiayi Wang1Guoquan Zhou2Chunbo Li3Fazhan Chen4Kang Ju5Liangliang Chen6Yichao Yin7Yi Chen8Yanping Zhang9Xu Zhang10Xinlin Zhou11Zongyuan Wang12Jianping Qiu13Hui Wang14Weizhong Shi15Feng Wang16Dong Wang17Zhihao Chen18Jiaojiao Hou19Hui Li20Zisheng Ai21School of Medicine, Tongji UniversitySchool of Medicine, Tongji UniversityShanghai Putuo Mental Health CenterShanghai Key Laboratory of Psychotic Disorders, Shanghai Mental Health Center, Shanghai Jiaotong University School of MedicineClinical Research Center for Mental Disorders, Shanghai Pudong New Area Mental Health Center, School of Medicine, Chinese-German Institute of Mental Health, Tongji UniversityShanghai Changning Mental Health Center, Changning DistrictShanghai Changning Mental Health Center, Changning DistrictShanghai Changning Mental Health Center, Changning DistrictDivision of Gastrointestinal Surgery, Department of General Surgery, West China Hospital, Sichuan UniversityShanghai Jinshan District Mental Health Center, Jinshan DistrictSchool of Medicine, Tongji UniversityLakefield College SchoolSchool of Medicine, Tongji UniversityShanghai Putuo Mental Health CenterShanghai Putuo Mental Health CenterShanghai Hospital Development CenterClinical Research Center for Mental Disorders, Shanghai Pudong New Area Mental Health Center, School of Medicine, Chinese-German Institute of Mental Health, Tongji UniversityClinical Research Center for Mental Disorders, Shanghai Pudong New Area Mental Health Center, School of Medicine, Chinese-German Institute of Mental Health, Tongji UniversityEast China University of Science and TechnologyUniversity Clinic of Child and Adolescent Psychiatry, Psychosomatics and Psychotherapy, RWTH Aachen UniversityShanghai Key Laboratory of Psychotic Disorders, Shanghai Mental Health Center, Shanghai Jiaotong University School of MedicineDepartment of Medical Statistics, School of Medicine, Tongji UniversityAbstract Background There is a lack of studies exploring the performance of Transformers-based language models in common risks assessment among psychiatric inpatients. We aim to develop a scalable risk assessment model using multidimensional textualized data and test the stability, robustness, and benefit of this approach. Methods In this real-world cohort study, a deep learning language model was developed and validated using first hospitalized cases diagnosed with schizophrenia, bipolar disorder, and depressive disorder between January 2016 and March 2023 in three hospitals. The algorithm was externally validated on an independent testing cohort comprising 1180 patients. A total of 140 features, including first medical records (FMR), laboratory examinations, medical orders, and psychological scales, were assessed for analysis. The outcomes were short- and long-term impulsivity (STI and LTI), risk of suicide (STSS and LTSS), and need of physical restraint (STPR and LTPR) assessed by qualified nurses or clinicians. Analysis was carried out between August 2024 and June 2024. Models with different architectures and input settings were compared with each other. The area under the receiver operating characteristic curve (AUROC) was used to assess the primary performance of models. The clinical utility was determined by the net benefit under Youden’s threshold. Results Of 7451 patients included in this study, 2982 (47.6%) were male, and the median (interquartile range) age was 42 (28–57) years. The overall incidence of outcomes was 635 (8.5%), 728 (10.5%), 659 (8.8%), 803 (10.8%), 588 (7.9%), and 728 (9.8%) for STPR, LTPR, STSS, LTSS, STI, and LTI, respectively. The multitask semi-structured Transformers-based language (SSTL) model showed more promising AUROCs (STPR: 0.915; LTPR: 0.844; STSS: 0.867; LTSS: 0.879; STI: 0.899; LTI: 0.894) in the prediction of these outcomes than single-tasked or multimodal language models and traditional structured data models. Combining FMR with other data from electronic health records led to significant improvements in the performance and clinical utility of SSTL models based on demographic, diagnosis, laboratory tests, treatment, and psychological scales. Conclusions The SSTL model shows potential advantages in prognostic evaluation. FMR is a strong predictor for common risks prediction and may benefit other tasks in psychiatry with minimum requirements for data and data processing.https://doi.org/10.1186/s12916-025-04150-7TransformersDeep learningSuicide riskImpulsivityPhysical restraint
spellingShingle Enzhao Zhu
Jiayi Wang
Guoquan Zhou
Chunbo Li
Fazhan Chen
Kang Ju
Liangliang Chen
Yichao Yin
Yi Chen
Yanping Zhang
Xu Zhang
Xinlin Zhou
Zongyuan Wang
Jianping Qiu
Hui Wang
Weizhong Shi
Feng Wang
Dong Wang
Zhihao Chen
Jiaojiao Hou
Hui Li
Zisheng Ai
A highly scalable deep learning language model for common risks prediction among psychiatric inpatients
BMC Medicine
Transformers
Deep learning
Suicide risk
Impulsivity
Physical restraint
title A highly scalable deep learning language model for common risks prediction among psychiatric inpatients
title_full A highly scalable deep learning language model for common risks prediction among psychiatric inpatients
title_fullStr A highly scalable deep learning language model for common risks prediction among psychiatric inpatients
title_full_unstemmed A highly scalable deep learning language model for common risks prediction among psychiatric inpatients
title_short A highly scalable deep learning language model for common risks prediction among psychiatric inpatients
title_sort highly scalable deep learning language model for common risks prediction among psychiatric inpatients
topic Transformers
Deep learning
Suicide risk
Impulsivity
Physical restraint
url https://doi.org/10.1186/s12916-025-04150-7
work_keys_str_mv AT enzhaozhu ahighlyscalabledeeplearninglanguagemodelforcommonriskspredictionamongpsychiatricinpatients
AT jiayiwang ahighlyscalabledeeplearninglanguagemodelforcommonriskspredictionamongpsychiatricinpatients
AT guoquanzhou ahighlyscalabledeeplearninglanguagemodelforcommonriskspredictionamongpsychiatricinpatients
AT chunboli ahighlyscalabledeeplearninglanguagemodelforcommonriskspredictionamongpsychiatricinpatients
AT fazhanchen ahighlyscalabledeeplearninglanguagemodelforcommonriskspredictionamongpsychiatricinpatients
AT kangju ahighlyscalabledeeplearninglanguagemodelforcommonriskspredictionamongpsychiatricinpatients
AT liangliangchen ahighlyscalabledeeplearninglanguagemodelforcommonriskspredictionamongpsychiatricinpatients
AT yichaoyin ahighlyscalabledeeplearninglanguagemodelforcommonriskspredictionamongpsychiatricinpatients
AT yichen ahighlyscalabledeeplearninglanguagemodelforcommonriskspredictionamongpsychiatricinpatients
AT yanpingzhang ahighlyscalabledeeplearninglanguagemodelforcommonriskspredictionamongpsychiatricinpatients
AT xuzhang ahighlyscalabledeeplearninglanguagemodelforcommonriskspredictionamongpsychiatricinpatients
AT xinlinzhou ahighlyscalabledeeplearninglanguagemodelforcommonriskspredictionamongpsychiatricinpatients
AT zongyuanwang ahighlyscalabledeeplearninglanguagemodelforcommonriskspredictionamongpsychiatricinpatients
AT jianpingqiu ahighlyscalabledeeplearninglanguagemodelforcommonriskspredictionamongpsychiatricinpatients
AT huiwang ahighlyscalabledeeplearninglanguagemodelforcommonriskspredictionamongpsychiatricinpatients
AT weizhongshi ahighlyscalabledeeplearninglanguagemodelforcommonriskspredictionamongpsychiatricinpatients
AT fengwang ahighlyscalabledeeplearninglanguagemodelforcommonriskspredictionamongpsychiatricinpatients
AT dongwang ahighlyscalabledeeplearninglanguagemodelforcommonriskspredictionamongpsychiatricinpatients
AT zhihaochen ahighlyscalabledeeplearninglanguagemodelforcommonriskspredictionamongpsychiatricinpatients
AT jiaojiaohou ahighlyscalabledeeplearninglanguagemodelforcommonriskspredictionamongpsychiatricinpatients
AT huili ahighlyscalabledeeplearninglanguagemodelforcommonriskspredictionamongpsychiatricinpatients
AT zishengai ahighlyscalabledeeplearninglanguagemodelforcommonriskspredictionamongpsychiatricinpatients
AT enzhaozhu highlyscalabledeeplearninglanguagemodelforcommonriskspredictionamongpsychiatricinpatients
AT jiayiwang highlyscalabledeeplearninglanguagemodelforcommonriskspredictionamongpsychiatricinpatients
AT guoquanzhou highlyscalabledeeplearninglanguagemodelforcommonriskspredictionamongpsychiatricinpatients
AT chunboli highlyscalabledeeplearninglanguagemodelforcommonriskspredictionamongpsychiatricinpatients
AT fazhanchen highlyscalabledeeplearninglanguagemodelforcommonriskspredictionamongpsychiatricinpatients
AT kangju highlyscalabledeeplearninglanguagemodelforcommonriskspredictionamongpsychiatricinpatients
AT liangliangchen highlyscalabledeeplearninglanguagemodelforcommonriskspredictionamongpsychiatricinpatients
AT yichaoyin highlyscalabledeeplearninglanguagemodelforcommonriskspredictionamongpsychiatricinpatients
AT yichen highlyscalabledeeplearninglanguagemodelforcommonriskspredictionamongpsychiatricinpatients
AT yanpingzhang highlyscalabledeeplearninglanguagemodelforcommonriskspredictionamongpsychiatricinpatients
AT xuzhang highlyscalabledeeplearninglanguagemodelforcommonriskspredictionamongpsychiatricinpatients
AT xinlinzhou highlyscalabledeeplearninglanguagemodelforcommonriskspredictionamongpsychiatricinpatients
AT zongyuanwang highlyscalabledeeplearninglanguagemodelforcommonriskspredictionamongpsychiatricinpatients
AT jianpingqiu highlyscalabledeeplearninglanguagemodelforcommonriskspredictionamongpsychiatricinpatients
AT huiwang highlyscalabledeeplearninglanguagemodelforcommonriskspredictionamongpsychiatricinpatients
AT weizhongshi highlyscalabledeeplearninglanguagemodelforcommonriskspredictionamongpsychiatricinpatients
AT fengwang highlyscalabledeeplearninglanguagemodelforcommonriskspredictionamongpsychiatricinpatients
AT dongwang highlyscalabledeeplearninglanguagemodelforcommonriskspredictionamongpsychiatricinpatients
AT zhihaochen highlyscalabledeeplearninglanguagemodelforcommonriskspredictionamongpsychiatricinpatients
AT jiaojiaohou highlyscalabledeeplearninglanguagemodelforcommonriskspredictionamongpsychiatricinpatients
AT huili highlyscalabledeeplearninglanguagemodelforcommonriskspredictionamongpsychiatricinpatients
AT zishengai highlyscalabledeeplearninglanguagemodelforcommonriskspredictionamongpsychiatricinpatients