Aims Most programs specializing in the treatment of first-episode psychosis in the United States focus on schizophrenia. However, many early psychosis patients do not fit into this diagnostic category. Here we describe McLean OnTrack, an intensive outpatient treatment program that accepts all-comers with first-episode psychosis. Methods We assessed baseline characteristics of patients in the 2.5 years since program initiation. We examined how initial referral diagnoses compare to current diagnoses, calculating the proportion of diagnostic changes. Results At 2.5 years, patients in McLean OnTrack consist of 30 (33.0%) individuals with primary psychotic disorder, 40 (44.0%) with affective psychosis, 19 (20.9%) with psychotic disorder NOS who do not meet full criteria for either category, and two (2.2%) individuals with no psychosis. While patients with affective psychosis had higher pre-morbid functioning, all three categories of psychosis had similar rates of prior hospitalizations and substance use. The retention rate in the psychotic disorder NOS group was lower than that in affective and primary psychotic disorders. Finally, diagnoses changed over the course of treatment in 50.5% of patients. Conclusions Diagnostic heterogeneity appears to be the norm among patients with first-episode psychosis, and diagnoses commonly evolve over the illness course. Baseline indices of illness severity were similar across categories and suggest the need for early intervention, irrespective of specific diagnosis. We discuss the benefits and challenges of a trans-diagnostic approach to early intervention in first episode psychosis, treating patients who share many but not all characteristics.

Recently natural language processing (NLP) tools have been developed to identify and extract salient risk indicators in electronic health records (EHRs). Sentiment analysis, although widely used in non-medical areas for improving decision making, has been studied minimally in the clinical setting. In this study, we undertook, to our knowledge, the first domain adaptation of sentiment analysis to psychiatric EHRs by defining psychiatric clinical sentiment, performing an annotation project, and evaluating multiple sentence-level sentiment machine learning (ML) models. Results indicate that off-the-shelf sentiment analysis tools fail in identifying clinically positive or negative polarity, and that the definition of clinical sentiment that we provide is learnable with relatively small amounts of training data. This project is an initial step towards further refining sentiment analysis methods for clinical use. Our long-term objective is to incorporate the results of this project as part of a machine learning model that predicts inpatient readmission risk. We hope that this work will initiate a discussion concerning domain adaptation of sentiment analysis to the clinical setting.

Objectives: Diagnostic shifts in first episode psychosis (FEP) are not uncommon. Many studies examining diagnostic stability use structured diagnostic interviews. Less is known about the stability of FEP diagnoses made clinically. Methods: We conducted a retrospective chart review of patients enrolled in a transdiagnostic FEP clinic. For the 96 patients followed clinically at least 2 years, we compared diagnoses at intake and 24 months. Results: Diagnostic stability was high for bipolar disorder (89%), schizoaffective disorder (89%), and schizophrenia (82%). Psychosis not otherwise specified (13%) was more unstable, with limited baseline differences that would enable clinicians to predict who would convert to a primary psychotic vs affective psychotic disorder. Conclusions: Our real-world clinical sample shows that FEP diagnoses, with the exception of unspecified psychosis, are diagnostically stable, even without structured diagnostic interviews.

Readmission after discharge from a hospital is disruptive and costly, regardless of the reason. However, it can be particularly problematic for psychiatric patients, so predicting which patients may be readmitted is critically important but also very difficult. Clinical narratives in psychiatric electronic health records (EHRs) span a wide range of topics and vocabulary; therefore, a psychiatric readmission prediction model must begin with a robust and interpretable topic extraction component. We created a data pipeline for using document vector similarity metrics to perform topic extraction on psychiatric EHR data in service of our long-term goal of creating a readmission risk classifier. We show initial results for our topic extraction model and identify additional features we will be incorporating in the future.

BackgroundReadmission after discharge from a hospital is disruptive and costly, regardless of the reason. However, it can be particularly problematic for psychiatric patients, so predicting which patients may be readmitted is critically important but also very difficult. Clinical narratives in psychiatric electronic health records (EHRs) span a wide range of topics and vocabulary; therefore, a psychiatric readmission prediction model must begin with a robust and interpretable topic extraction component.ResultsWe designed and evaluated multiple multilayer perceptron and radial basis function neural networks to predict the sentences in a patient’s EHR that are associated with one or more of seven readmission risk factor domains that we identified. In contrast to our baseline cosine similarity model that is based on the methodologies of prior works, our deep learning approaches achieved considerably better F1 scores (0.83 vs 0.66) while also being more scalable and computationally efficient with large volumes of data. Additionally, we found that integrating clinically relevant multiword expressions during preprocessing improves the accuracy of our models and allows for identifying a wider scope of training data in a semi-supervised setting.ConclusionWe created a data pipeline for using document vector similarity metrics to perform topic extraction on psychiatric EHR data in service of our long-term goal of creating a readmission risk classifier. We show results for our topic extraction model and identify additional features we will be incorporating in the future.