A generic causality-informed neural network (CINN) methodology for quantitative risk analytics and decision support.

In this paper, we develop a generic framework for systemically encoding causal knowledge manifested in the form of hierarchical causality structure and qualitative (or quantitative) causal relationships into neural networks to facilitate sound risk analytics and decision support via causally-aware intervention reasoning. The proposed methodology for establishing causality-informed neural network (CINN) follows a four-step procedure. In the first step, we explicate how causal knowledge in the form of directed acyclic graph (DAG) can be discovered from observation data or elicited from domain experts. Next, we categorize nodes in the constructed DAG representing causal relationships among observed variables into several groups (e.g., root nodes, intermediate nodes, and leaf nodes), and align the architecture of CINN with causal relationships specified in the DAG while preserving the orientation of each existing causal relationship. In addition to a dedicated architecture design, CINN also gets embodied in the design of loss function, where both intermediate and leaf nodes are treated as target outputs to be predicted by CINN. In the third step, we propose to incorporate domain knowledge on stable causal relationships into CINN, and the injected constraints on causal relationships act as guardrails to prevent unexpected behaviors of CINN. Finally, the trained CINN is exploited to perform intervention reasoning with emphasis on estimating the effect that policies and actions can have on the system behavior, thus facilitating risk-informed decision making through comprehensive "what-if" analysis. Two case studies are used to demonstrate the substantial benefits enabled by CINN in risk analytics and decision support.

Association between housework and the risk of dementia among older Chinese adults: a prospective cohort study.

Objectives: Physical activity (PA) is known to improve physical functioning and mental health and to reduce the incidence of dementia. However, studies of the effects of non-recreational PA on the incidence of dementia, especially in East Asian populations, remain limited. In this study, we evaluate the association of doing housework with the risk of dementia among participants in the Chinese Longitudinal Healthy Longevity Survey (CLHLS). Methods: The analysis was conducted with data from 7,237 CLHLS participants age over 65 obtained in 2008/2009, 2011/2012, 2014, and 2018. The frequency of housework performance was classified into four groups. A Cox proportional-hazards model was used to examine the association of the baseline housework frequency with the incidence of dementia, with adjustment for demographic and socioeconomic characteristics and lifestyle and health conditions. Results: The adjusted multivariate model showed that the incidence of dementia was lower among participants who did housework almost every day than among those who rarely or never did housework (hazard ratio = 0.49; 95% confidence interval, 0.39-0.61). The subgroup and sensitivity analyses yielded similar results. Conclusion: A high frequency of housework performance was associated with a reduced incidence of dementia among older Chinese adults, especially those who did not exercise regularly. The encouragement of engagement in housework would be a cost-effective measure promoting healthy aging in the Chinese population.

Early Diagnosis of Bipolar Disorder Coming Soon: Application of an Oxidative Stress Injury Biomarker (BIOS) Model.

Early distinction of bipolar disorder (BD) from major depressive disorder (MDD) is difficult since no tools are available to estimate the risk of BD. In this study, we aimed to develop and validate a model of oxidative stress injury for predicting BD. Data were collected from 1252 BD and 1359 MDD patients, including 64 MDD patients identified as converting to BD from 2009 through 2018. 30 variables from a randomly-selected subsample of 1827 (70%) patients were used to develop the model, including age, sex, oxidative stress markers (uric acid, bilirubin, albumin, and prealbumin), sex hormones, cytokines, thyroid and liver function, and glycolipid metabolism. Univariate analyses and the Least Absolute Shrinkage and Selection Operator were applied for data dimension reduction and variable selection. Multivariable logistic regression was used to construct a model for predicting bipolar disorder by oxidative stress biomarkers (BIOS) on a nomogram. Internal validation was assessed in the remaining 784 patients (30%), and independent external validation was done with data from 3797 matched patients from five other hospitals in China. 10 predictors, mainly oxidative stress markers, were shown on the nomogram. The BIOS model showed good discrimination in the training sample, with an AUC of 75.1% (95% CI: 72.9%-77.3%), sensitivity of 0.66, and specificity of 0.73. The discrimination was good both in internal validation (AUC 72.1%, 68.6%-75.6%) and external validation (AUC 65.7%, 63.9%-67.5%). In this study, we developed a nomogram centered on oxidative stress injury, which could help in the individualized prediction of BD. For better real-world practice, a set of measurements, especially on oxidative stress markers, should be emphasized using big data in psychiatry.

Screening of the college students at clinical high risk for psychosis in China: a multicenter epidemiological study.

BACKGROUND: To investigate a 3-stage screening procedure and explore the clinical features of subjects at Clinical High Risk (CHR) for psychosis in a representative sample of Chinese college students. METHODS: An epidemiological survey of the prevalence of the CHR syndrome in Chinese college students that was selected by stratified random sampling from Shanghai, Nanjing and Nanchang cities was done following a 3-stage procedure. Participants were initially screened with the Prodromal Questionnaire-brief version (PQ-B), and whose distress score of PQ-B exceeded 24 would be invited to a telephone assessment using the subscale for positive symptoms of the Scale of Prodromal Symptoms (SOPS)/Structured Interview for Prodromal Syndromes (SIPS). Lastly, participants who scored 3 or higher in any item of the subscale would be administered with the SIPS interview conducted by trained researchers to confirm the diagnosis of CHR syndrome. RESULTS: Twenty-three thousand sixty-three college students completed the survey during September 2017 to October 2018. Seventy-two students were diagnosed as CHR subjects, and the detection rate in the total sample was 0.3%. The peak age range for the first diagnosis of CHR was 17 to 20 years. Thirteen and forty-six were set as the cutoff points of PQ-B total score and distress score to balance the greatest sensitivity and specificity. Binary logistic regression revealed that 8 items in PQ-B showed significant distinction for detecting CHR subjects. CONCLUSIONS: The 3-stage screening method can be utilized in the detection of CHR subjects for psychosis in the general population, during which delusional ideas, perceptual abnormalities and suspiciousness deserve great attention.

Brain regional homogeneity and function connectivity in attenuated psychosis syndrome -based on a resting state fMRI study.

BACKGROUND: By combining regional homogeneity (ReHo) and functional connectivity (FC) analyses, this study aimed to explore brain functional alterations in Attenuated Psychosis Syndrome (APS), which could provide complementary information for the neurophysiological indicators for schizophrenia (SZ) associated brain dysfunction. METHODS: Twenty-one APS subjects and twenty healthy controls were enrolled in the data acquisition of demographics and clinical characteristics as well as structural and resting-state functional magnetic resonance imaging (rs-fMRI). ReHo analysis was conducted to determine the peak coordinate of the abnormal regional brain activity. Then, identified brain regions were considered as seed regions and were used to calculate FC between reginal brain voxels and whole brain voxels. Finally, potential correlations between imaging indices and clinical data were also explored. RESULTS: Four APS and two HC subjects were excluded because the largest dynamic translation or rotation had exceeded 2 mm / 2°. Compared with healthy controls (HCs), APS subjects exhibited higher ReHo values in the right middle temporal gyrus (MTG) and lower ReHo values in the left middle frontal gyrus (MFG), left superior frontal gyrus (SFG), left postcentral gyrus (PoCG), and left superior frontal gyrus, medial (SFGmed). Considered these areas as seed regions, the APS subjects showed abnormal enhancement in functional brain connections, predominantly in the frontal and temporal lobes. CONCLUSIONS: We concluded that the APS subjects had spatially regional dysfunction and remoted synchronous dysfunction in the frontal and temporal lobes of the brain, and changes in ReHo and FC patterns may reveal the mechanism of brain dysfunctions and may serve as an imaging biomarker for the diagnosis and evaluation of SZ.