Limited research exists in predicting first-time suicide attempts that account for two-thirds of suicide decedents. We aimed to predict first-time suicide attempts using a large data-driven approach that applies natural language processing (NLP) and machine learning (ML) to unstructured (narrative) clinical notes and structured electronic health record (EHR) data. This case-control study included patients aged 10–75 years who were seen between 2007 and 2016 from emergency departments and inpatient units. Cases were first-time suicide attempts from coded diagnosis; controls were randomly selected without suicide attempts regardless of demographics, following a ratio of nine controls per case. Four data-driven ML models were evaluated using 2-year historical EHR data prior to suicide attempt or control index visits, with prediction windows from 7 to 730 days. Patients without any historical notes were excluded. Model evaluation on accuracy and robustness was performed on a blind dataset (30% cohort). The study cohort included 45 238 patients (5099 cases, 40 139 controls) comprising 54 651 variables from 5.7 million structured records and 798 665 notes. Using both unstructured and structured data resulted in significantly greater accuracy compared to structured data alone (area-under-the-curve [AUC]: 0.932 vs. 0.901 P < .001). The best-predicting model utilized 1726 variables with AUC = 0.932 (95% CI, 0.922–0.941). The model was robust across multiple prediction windows and subgroups by demographics, points of historical most recent clinical contact, and depression diagnosis history. Our large data-driven approach using both structured and unstructured EHR data demonstrated accurate and robust first-time suicide attempt prediction, and has the potential to be deployed across various populations and clinical settings.

Objective: Social determinants of health (SDOH) are nonclinical, socioeconomic conditions that influence patient health and quality of life. Identifying SDOH may help clinicians target interventions. However, SDOH are more frequently available in narrative notes compared to structured electronic health records. The 2022 n2c2 Track 2 competition released clinical notes annotated for SDOH to promote development of NLP systems for extracting SDOH. We developed a system addressing 3 limitations in state-of-the-art SDOH extraction: the inability to identify multiple SDOH events of the same type per sentence, overlapping SDOH attributes within text spans, and SDOH spanning multiple sentences. Materials and Methods: We developed and evaluated a 2-stage architecture. In stage 1, we trained a BioClinical-BERT-based named entity recognition system to extract SDOH event triggers, that is, text spans indicating substance use, employment, or living status. In stage 2, we trained a multitask, multilabel NER to extract arguments (eg, alcohol “type”) for events extracted in stage 1. Evaluation was performed across 3 subtasks differing by provenance of training and validation data using precision, recall, and F1 scores. Results: When trained and validated on data from the same site, we achieved 0.87 precision, 0.89 recall, and 0.88 F1. Across all subtasks, we ranked between second and fourth place in the competition and always within 0.02 F1 from first. Conclusions: Our 2-stage, deep-learning-based NLP system effectively extracted SDOH events from clinical notes. This was achieved with a novel classification framework that leveraged simpler architectures compared to state-of-the-art systems. Improved SDOH extraction may help clinicians improve health outcomes.

Background: Electrocardiogram (ECG) signal, an important indicator for heart problems, is commonly corrupted by a low-frequency baseline wander (BW) artifact, which may cause interpretation difficulty or inaccurate analysis. Unlike current state-of-the-art approach using band-pass filters, wavelet transforms can accurately capture both time and frequency information of a signal. However, extant literature is limited in applying wavelet transforms (WTs) for baseline wander removal. In this study, we aimed to evaluate 5 wavelet families with a total of 14 wavelets for removing ECG baseline wanders from a semi-synthetic dataset. Methods: We created a semi-synthetic ECG dataset based on a public QT Database on Physionet repository with ECG data from 105 patients. The semi-synthetic ECG dataset comprised ECG excerpts from the QT database superimposed with artificial baseline wanders. We extracted one ECG excerpt from each of 105 patients, and the ECG excerpt comprised 14 s of randomly selected ECG data. Twelve baseline wanders were manually generated, including sinusoidal waves, spikes and step functions. We implemented and evaluated 14 commonly used wavelets up to 12 WT levels. The evaluation metric was mean-square-error (MSE) between the original ECG excerpt and the processed signal with artificial BW removed. Results: Among the 14 wavelets, Daubechies-3 wavelet and Symlets-3 wavelet with 7 levels of WT had best performance, MSE = 0.0044. The average MSEs for sinusoidal waves, step, and spike functions were 0.0271, 0.0304, 0.0199 respectively. For artificial baseline wanders with spikes or step functions, wavelet transforms in general had lower performance in removing the BW; however, WTs accurately located the temporal position of an impulse edge. Conclusions: We found wavelet transforms in general accurately removed various baseline wanders. Daubechies-3 and Symlets-3 wavelets performed best. The study could facilitate future real-time processing of streaming ECG signals for clinical decision support systems.