Development of a machine learning model using electrocardiogram signals to improve acute pulmonary embolism screening.

Aims: Clinical scoring systems for pulmonary embolism (PE) screening have low specificity and contribute to computed tomography pulmonary angiogram (CTPA) overuse. We assessed whether deep learning models using an existing and routinely collected data modality, electrocardiogram (ECG) waveforms, can increase specificity for PE detection. Methods and results: We create a retrospective cohort of 21 183 patients at moderate- to high suspicion of PE and associate 23 793 CTPAs (10.0% PE-positive) with 320 746 ECGs and encounter-level clinical data (demographics, comorbidities, vital signs, and labs). We develop three machine learning models to predict PE likelihood: an ECG model using only ECG waveform data, an EHR model using tabular clinical data, and a Fusion model integrating clinical data and an embedded representation of the ECG waveform. We find that a Fusion model [area under the receiver-operating characteristic curve (AUROC) 0.81 ± 0.01] outperforms both the ECG model (AUROC 0.59 ± 0.01) and EHR model (AUROC 0.65 ± 0.01). On a sample of 100 patients from the test set, the Fusion model also achieves greater specificity (0.18) and performance (AUROC 0.84 ± 0.01) than four commonly evaluated clinical scores: Wells' Criteria, Revised Geneva Score, Pulmonary Embolism Rule-Out Criteria, and 4-Level Pulmonary Embolism Clinical Probability Score (AUROC 0.50-0.58, specificity 0.00-0.05). The model is superior to these scores on feature sensitivity analyses (AUROC 0.66-0.84) and achieves comparable performance across sex (AUROC 0.81) and racial/ethnic (AUROC 0.77-0.84) subgroups. Conclusion: Synergistic deep learning of ECG waveforms with traditional clinical variables can increase the specificity of PE detection in patients at least at moderate suspicion for PE.

Phase response analyses support a relaxation oscillator model of locomotor rhythm generation in Caenorhabditis elegans.

Neural circuits coordinate with muscles and sensory feedback to generate motor behaviors appropriate to an animal's environment. In C. elegans, the mechanisms by which the motor circuit generates undulations and modulates them based on the environment are largely unclear. We quantitatively analyzed C. elegans locomotion during free movement and during transient optogenetic muscle inhibition. Undulatory movements were highly asymmetrical with respect to the duration of bending and unbending during each cycle. Phase response curves induced by brief optogenetic inhibition of head muscles showed gradual increases and rapid decreases as a function of phase at which the perturbation was applied. A relaxation oscillator model based on proprioceptive thresholds that switch the active muscle moment was developed and is shown to quantitatively agree with data from free movement, phase responses, and previous results for gait adaptation to mechanical loadings. Our results suggest a neuromuscular mechanism underlying C. elegans motor pattern generation within a compact circuit.

Development and validation of techniques for phenotyping ST-elevation myocardial infarction encounters from electronic health records.

OBJECTIVES: Classifying hospital admissions into various acute myocardial infarction phenotypes in electronic health records (EHRs) is a challenging task with strong research implications that remains unsolved. To our knowledge, this study is the first study to design and validate phenotyping algorithms using cardiac catheterizations to identify not only patients with a ST-elevation myocardial infarction (STEMI), but the specific encounter when it occurred. MATERIALS AND METHODS: We design and validate multi-modal algorithms to phenotype STEMI on a multicenter EHR containing 5.1 million patients and 115 million patient encounters by using discharge summaries, diagnosis codes, electrocardiography readings, and the presence of cardiac catheterizations on the encounter. RESULTS: We demonstrate that robustly phenotyping STEMIs by selecting discharge summaries containing "STEM" has the potential to capture the most number of STEMIs (positive predictive value [PPV] = 0.36, N = 2110), but that addition of a STEMI-related International Classification of Disease (ICD) code and cardiac catheterizations to these summaries yields the highest precision (PPV = 0.94, N = 952). DISCUSSION AND CONCLUSION: In this study, we demonstrate that the incorporation of percutaneous coronary intervention increases the PPV for detecting STEMI-related patient encounters from the EHR.

Federated Learning of Electronic Health Records to Improve Mortality Prediction in Hospitalized Patients With COVID-19: Machine Learning Approach.

BACKGROUND: Machine learning models require large datasets that may be siloed across different health care institutions. Machine learning studies that focus on COVID-19 have been limited to single-hospital data, which limits model generalizability. OBJECTIVE: We aimed to use federated learning, a machine learning technique that avoids locally aggregating raw clinical data across multiple institutions, to predict mortality in hospitalized patients with COVID-19 within 7 days. METHODS: Patient data were collected from the electronic health records of 5 hospitals within the Mount Sinai Health System. Logistic regression with L1 regularization/least absolute shrinkage and selection operator (LASSO) and multilayer perceptron (MLP) models were trained by using local data at each site. We developed a pooled model with combined data from all 5 sites, and a federated model that only shared parameters with a central aggregator. RESULTS: The LASSOfederated model outperformed the LASSOlocal model at 3 hospitals, and the MLPfederated model performed better than the MLPlocal model at all 5 hospitals, as determined by the area under the receiver operating characteristic curve. The LASSOpooled model outperformed the LASSOfederated model at all hospitals, and the MLPfederated model outperformed the MLPpooled model at 2 hospitals. CONCLUSIONS: The federated learning of COVID-19 electronic health record data shows promise in developing robust predictive models without compromising patient privacy.

Federated Learning of Electronic Health Records Improves Mortality Prediction in Patients Hospitalized with COVID-19.

Machine learning (ML) models require large datasets which may be siloed across different healthcare institutions. Using federated learning, a ML technique that avoids locally aggregating raw clinical data across multiple institutions, we predict mortality within seven days in hospitalized COVID-19 patients. Patient data was collected from Electronic Health Records (EHRs) from five hospitals within the Mount Sinai Health System (MSHS). Logistic Regression with L1 regularization (LASSO) and Multilayer Perceptron (MLP) models were trained using local data at each site, a pooled model with combined data from all five sites, and a federated model that only shared parameters with a central aggregator. Both the federated LASSO and federated MLP models performed better than their local model counterparts at four hospitals. The federated MLP model also outperformed the federated LASSO model at all hospitals. Federated learning shows promise in COVID-19 EHR data to develop robust predictive models without compromising patient privacy.

Psychiatric Diagnoses and Treatment Preceding Schizophrenia in Adolescents Aged 9-17 Years.

OBJECTIVE: Our study aimed to examine psychiatric diagnoses and treatment preceding a schizophrenia diagnosis in adolescents, stratified by sex and race/ethnicity. METHODS: Using Medicaid physical and behavioral health and pharmacy claims data, we identified 1,459 adolescents who were aged 9-17 years and diagnosed with schizophrenia between January 2006 through June 2009. Psychiatric diagnosis, mental health service use including psychiatric hospitalization, residential treatment and outpatient therapy and psychotropic medication use preceding schizophrenia were identified. RESULTS: Forty-five percent of the adolescents were diagnosed with one or more psychiatric conditions. More than 40% of the adolescents were hospitalized or placed in a residential treatment facility for other psychiatric conditions preceding schizophrenia. Overall, 72% of the adolescents were prescribed with one or more psychotropic medications and 22% were prescribed with three or more psychotropic medications in the year prior to their first schizophrenia diagnosis. We found that sex and race/ethnicity influence preceding psychiatric conditions and psychiatric treatment use. CONCLUSIONS: Careful screening and evaluation to validate diagnoses is important as the presence of certain psychiatric morbidity is common among adolescents with schizophrenia during the prodromal period. Developing acceptable and accessible interventions that will reduce psychiatric hospitalization and residential treatment care and improve care connection for schizophrenia treatment is important to mitigate complexity in treatment for adolescents and reduce cost burden for families and the society. Integrating health claims data in the development of schizophrenia risk conversion models can be useful in effectively predicting ideal timing of tailored interventions for adolescents with preceding psychiatric conditions.

Assessing First Visits By Physicians To Medicare Patients Discharged To Skilled Nursing Facilities.

Although approximately one in five Medicare beneficiaries are discharged from hospital acute care to postacute care at skilled nursing facilities (SNFs), little is known about access to timely medical care for these patients after they are admitted to a SNF. Our analysis of 2,392,753 such discharges from hospitals under fee-for-service Medicare in the period January 2012-October 2014 indicated that first visits by a physician or advanced practitioner (a nurse practitioner or physician assistant) for initial medical assessment occurred within four days of SNF admission in 71.5 percent of the stays. However, there was considerable variation in days to first visit at the regional, facility, and patient levels. We estimated that in 10.4 percent of stays there was no physician or advanced practitioner visit. Understanding the underlying reasons for, and consequences of, variability in timing and receipt of initial medical assessment after admission to a SNF for postacute care may prove important for improving patient outcomes and particularly relevant to current efforts to promote value-based purchasing in postacute care.

Making Sense of the Cognitive Task of Medication Reconciliation Using a Card Sorting Task.

OBJECTIVE: To explore cognitive strategies clinicians apply while performing a medication reconciliation task, handling incomplete and conflicting information. BACKGROUND: Medication reconciliation is a method clinicians apply to find and resolve inconsistencies in patients' medications and medical conditions lists. The cognitive strategies clinicians use during reconciliation are unclear. Controlled lab experiments can explore how clinicians make sense of uncertain, missing, or conflicting information and therefore support the development of a human performance model. We hypothesize that clinicians apply varied cognitive strategies to handle this task and that profession and experience affect these strategies. METHOD: 130 clinicians participated in a tablet-based experiment conducted in a large American teaching hospital. They were asked to simulate medication reconciliation using a card sorting task (CaST) to organize medication and medical condition lists of a specific clinical case. Later on, they were presented with new information and were asked to add it to their arrangements. We quantitatively and qualitatively analyzed the ways clinicians arranged patient information. RESULTS: Four distinct cognitive strategies were identified ("Conditions first": n = 76 clinicians, "Medications first": n = 7, "Crossover": n = 17, and "Alternating": n = 10). The strategy clinicians applied was affected by their experience (p = .02) but not by their profession. At the appearance of new information, clinicians moved medication cards more frequently (75.2 movements vs. 49.6 movements, p < .001), suggesting that they match medications to medical conditions. CONCLUSION: Clinicians apply various cognitive strategies while reconciling medications and medical conditions. APPLICATION: Clinical information systems should support multiple cognitive strategies, allowing flexibility in organizing information.

Distributed rhythm generators underlie Caenorhabditis elegans forward locomotion.

Coordinated rhythmic movements are ubiquitous in animal behavior. In many organisms, chains of neural oscillators underlie the generation of these rhythms. In C. elegans, locomotor wave generation has been poorly understood; in particular, it is unclear where in the circuit rhythms are generated, and whether there exists more than one such generator. We used optogenetic and ablation experiments to probe the nature of rhythm generation in the locomotor circuit. We found that multiple sections of forward locomotor circuitry are capable of independently generating rhythms. By perturbing different components of the motor circuit, we localize the source of secondary rhythms to cholinergic motor neurons in the midbody. Using rhythmic optogenetic perturbation, we demonstrate bidirectional entrainment of oscillations between different body regions. These results show that, as in many other vertebrates and invertebrates, the C. elegans motor circuit contains multiple oscillators that coordinate activity to generate behavior.

Quantitative Assessment of Fat Levels in Caenorhabditis elegans Using Dark Field Microscopy.

The roundworm Caenorhabditis elegans is widely used as a model for studying conserved pathways for fat storage, aging, and metabolism. The most broadly used methods for imaging fat in C. elegans require fixing and staining the animal. Here, we show that dark field images acquired through an ordinary light microscope can be used to estimate fat levels in worms. We define a metric based on the amount of light scattered per area, and show that this light scattering metric is strongly correlated with worm fat levels as measured by Oil Red O (ORO) staining across a wide variety of genetic backgrounds and feeding conditions. Dark field imaging requires no exogenous agents or chemical fixation, making it compatible with live worm imaging. Using our method, we track fat storage with high temporal resolution in developing larvae, and show that fat storage in the intestine increases in at least one burst during development.