The care of traumatized children would benefit significantly from accurate predictive models for Posttraumatic Stress Disorder (PTSD), using information available around the time of trauma. Machine Learning (ML) computational methods have yielded strong results in recent applications across many diseases and data types, yet they have not been previously applied to childhood PTSD. Since these methods have not been applied to this complex and debilitating disorder, there is a great deal that remains to be learned about their application. The first step is to prove the concept: Can ML methods – as applied in other fields – produce predictive classification models for childhood PTSD? Additionally, we seek to determine if specific variables can be identified – from the aforementioned predictive classification models - with putative causal relations to PTSD. ML predictive classification methods – with causal discovery feature selection – were applied to a data set of 163 children hospitalized with an injury and PTSD was determined three months after hospital discharge. At the time of hospitalization, 105 risk factor variables were collected spanning a range of biopsychosocial domains. Seven percent of subjects had a high level of PTSD symptoms. A predictive classification model was discovered with significant predictive accuracy. A predictive model constructed based on subsets of potentially causally relevant features achieves similar predictivity compared to the best predictive model constructed with all variables. Causal Discovery feature selection methods identified 58 variables of which 10 were identified as most stable. In this first proof-of-concept application of ML methods to predict childhood Posttraumatic Stress we were able to determine both predictive classification models for childhood PTSD and identify several causal variables. This set of techniques has great potential for enhancing the methodological toolkit in the field and future studies should seek to replicate, refine, and extend the results produced in this study.
23 July 2018 in BMC Medical Informatics and Decision Making
Cardiac Resynchronization Therapy (CRT) is an established pacing therapy for heart failure patients. The New York Heart Association (NYHA) class is often used as a measure of a patient’s response to CRT. Identifying NYHA class for heart failure (HF) patients in an electronic health record (EHR) consistently, over time, can provide better understanding of the progression of heart failure and assessment of CRT response and effectiveness. Though NYHA is rarely stored in EHR structured data, such information is often documented in unstructured clinical notes. We accessed HF patients’ data in a local EHR system and identified potential sources of NYHA, including local diagnosis codes, procedures, and clinical notes. We further investigated and compared the performances of rule-based versus machine learning-based natural language processing (NLP) methods to identify NYHA class from clinical notes. Of the 36,276 patients with a diagnosis of HF or a CRT implant, 19.2% had NYHA class mentioned at least once in their EHR. While NYHA class existed in descriptive fields association with diagnosis codes (31%) or procedure codes (2%), the richest source of NYHA class was clinical notes (95%). A total of 6174 clinical notes were matched with hospital-specific custom NYHA class diagnosis codes. Machine learning-based methods outperformed a rule-based method. The best machine-learning method was a random forest with n-gram features (F-measure: 93.78%). NYHA class is documented in different parts in EHR for HF patients and the documentation rate is lower than expected. NLP methods are a feasible way to extract NYHA class information from clinical notes.
Treating adolescent depression effectively requires providing interventions that are optimally suited to patients’ individual characteristics and needs. Therefore, we aim to develop an algorithm that matches patients with optimal treatment among cognitive-behavioral therapy (CBT), fluoxetine (FLX), and combination treatment (COMB). We leveraged data from a completed clinical trial, the Treatment for adolescents with depression study, where a wide range of demographic, clinical, and psychosocial measures were collected from adolescents diagnosed with major depressive disorder prior to treatment. Machine-learning techniques were employed to derive a model that predicts treatment response (week 12 children’s depression rating scale-revised [CDRS-R]) to CBT, FLX, and COMB. The resulting model successfully identified subgroups of patients that respond preferentially to specific types of treatment. Specifically, our model identified a subgroup of patients (25%) that achieved on average a 16.9 point benefit on the CDRS-R from FLX compared to CBT. The model also identified a subgroup of patients (50%) that achieved an average benefit up to 19.0 points from COMB compared to CBT. Physical illness and disability were identified as overall predictors of response to treatment, regardless of treatment type, whereas baseline CDRS-R, psychosomatic symptoms, school missed, view of self, treatment expectations, and attention problems determined the patients’ response to specific treatments. The model developed in this study provides a critical starting point for personalized treatment planning for adolescent depression.
Human self‐reports often indicate that changes in mood are a major contributor to drug relapse. Still, arguments have been made that instances of drug‐seeking following abstinence in animal models (i.e. relapse/reinstatement) may be outside of hedonic control. Therefore, the present study utilized ultrasonic vocalizations in the rat in order to evaluate affect during cocaine self‐administration and contextual reinstatement of cocaine‐seeking in a pre‐clinical model of drug relapse (abstinence‐reinstatement model). Results show that while subjects effectively reinstated drug‐seeking (lever pressing) following 30 days of abstinence, and spontaneously recovered/reinstated drug‐seeking following 60 days of abstinence, ultrasonic vocalizations did not increase over baseline levels during either reinstatement session. These results are consistent with previous results from our laboratory and current theories of addiction suggesting that cues that are weakly associated with drug consumption can motivate drug‐seeking behavior that is outside of hedonic processing.
1 November 2017 in IEEE International Conference on Bioinformatics and Biomedicine (BIBM)
Cardiac Resynchronization Therapy (CRT) is an established pacing therapy for heart failure patients. The New York Heart Association (NYHA) classification is often used as a measure of a patient's response to CRT. Identifying NYHA class for heart failure patients in an electronic health record (EHR) consistently, over time, can provide better understanding of the progression of heart failure and assessment of CRT response and effectiveness. However, NYHA is rarely stored in EHR structured data such information is often documented in unstructured clinical notes. In this study, we thus investigated the use of natural language processing (NLP) methods to identify NYHA classification from clinical notes. We collected 6,174 clinical notes that were matched with hospital-specific custom NYHA class diagnosis codes. Machine-learning based methods performed similar with a rule-based method. The best machine-learning method, support vector machine with n-gram features, performed the best (93% F-measure). Further validation of the findings is required.