Post-acute Sequelae of COVID-19 (PASC), also known as Long COVID, is a broad grouping of a range of long-term symptoms following acute COVID-19 infection. An understanding of characteristics that are predictive of future PASC is valuable, as this can inform the identification of high-risk individuals and future preventative efforts. However, current knowledge regarding PASC risk factors is limited. Using a sample of 55,257 participants from the National COVID Cohort Collaborative, as part of the NIH Long COVID Computational Challenge, we sought to predict individual risk of PASC diagnosis from a curated set of clinically informed covariates. We predicted individual PASC status, given covariate information, using Super Learner (an ensemble machine learning algorithm also known as stacking) to learn the optimal, AUC-maximizing combination of gradient boosting and random forest algorithms. We were able to predict individual PASC diagnoses accurately (AUC 0.947). Temporally, we found that baseline characteristics were most predictive of future PASC diagnosis, compared with characteristics immediately before, during, or after COVID-19 infection. This finding supports the hypothesis that clinicians may be able to accurately assess the risk of PASC in patients prior to acute COVID diagnosis, which could improve early interventions and preventive care. We found that medical utilization, demographics and anthropometry, and respiratory factors were most predictive of PASC diagnosis. This highlights the importance of respiratory characteristics in PASC risk assessment. The methods outlined here provide an open-source, applied example of using Super Learner to predict PASC status using electronic health record data, which can be replicated across a variety of settings.

Background: The COVID-19 global pandemic has posed unprecedented challenges to health care systems all over the world. The speed of the viral spread results in a tsunami of patients, which begs for a reliable screening tool using readily available data to predict disease progression.Methods: Multicenter retrospective cohort study was performed to develop and validate a triage model. Patient demographic and non-laboratory clinical data were recorded. Using only the data from Zhongnan Hospital, step-wise multivariable logistic regression was performed, and a prognostic nomogram was constructed based on the independent variables identifies. The discrimination and calibration of the model were validated. External independent validation was performed to further address the utility of this model using data from Jinyintan Hospital.Results: A total of 716 confirmed COVID-19 cases from Zhongnan Hospital were included for model construction. Men, increased age, fever, hypertension, cardio-cerebrovascular disease, dyspnea, cough, and myalgia are independent risk factors for disease progression. External independent validation was carried out in a cohort with 201 cases from Jinyintan Hospital. The area under the curve (AUC) was 0.787 (95% confidence interval [CI]: 0.747–0.827) in the training group and 0.704 (95% CI: 0.632–0.777) in the validation group.Conclusions: We developed a novel triage model based on basic and clinical data. Our model could be used as a pragmatic screening aid to allow for cost efficient screening to be carried out such as over the phone, which may reduce disease propagation through limiting unnecessary contact. This may help allocation of limited medical resources.

1.4%) patients, UA in 132 (0.4%) patients, and 30,665 (98.2%) with non-ACS etiologies. After applying feature engineering and selection techniques, we arrived at 11 features in our analysis. The features found to be independently associated with NSTEMI included: systolic blood pressure, brain natriuretic peptide, congestive heart failure, coronary artery disease, creatinine, glucose, prior myocardial infarction, heart rate, nephrotic syndrome, red cell distribution width, and troponin. To address the low event rates, we incorporated the use of undersampling technique and cost-sensitive learning. Table 1 demonstrates a system engineering approach indicating the range of continuous variables most associated with NSTEMI. This model had the following test characteristics for the diagnosis of NSTEMI: sensitivity 81%, specificity 80%, and an AUC score of 0.89.Conclusion: In this study, we developed a Random Forest model with modest accuracy at predicting NSTEMI. Further work is ongoing to refine modeling with additions of electrocardiograms and clinical notes.

Study Objectives: More than 1.6 million individuals present to US EDs each year for care after physical assault. However, little is known about posttraumatic outcomes of this population. We assessed the incidence of adverse posttraumatic neuropsychiatric sequelae (APNS) in a prospective longitudinal sample of individuals presenting to the ED after physical assault.Methods: ED patients aged 18-75 who presented to one of 28 ED sites after physical assault were eligible for enrollment if they had no evidence of solid organ injury > grade 1, no significant hemorrhage, no indication for chest tube placement or operation under general anesthesia, and were unlikely to be admitted for more than 72 hours. Baseline assessments included sociodemographic characteristics, maximal AIS score, and characteristics of the assault. Three-month outcome assessments included assessment of substantial posttraumatic stress (PTS, PCL-5 38), pain (numeric pain scale score 4), and depressive (PROMIS-8b depression 60) symptoms. Associations between patient characteristics and APNS were evaluated using logistic regression.Results: A total of 271 patients were enrolled in the study following physical assault. The mean age was 33.5 (SD 11.7), and 48% were female. Most participants identified as non-Hispanic Black (64%), non-Hispanic White (21%), or Hispanic (11%). The majority n¼ 206 (76%) reported being attacked by another person, 46 (17%) reported starting or intentionally joining an altercation, and 19 (7%) reported another assault mechanism. Average maximal AIS score was 1.3 (range 1-3), 99% were discharged to home after evaluation. Three month follow-up results were available for 207/271 patients (76%). Substantial PTS symptoms 3 months after assault were present in 63/199 (32%), substantial pain was present in 107/206 (52%), and substantial depressive symptoms were present in 70/207 (34%). After controlling for age and sex, patients who were attacked had more severe PTS symptoms (b¼0.18, t¼2.40, p¼0.018) and depressive symptoms (b¼0.18, t¼2.45, p¼0.015) than those who reported intentionally joining an altercation, but not more severe pain symptoms (b¼0.12, t¼1.67, p ¼ 0.097). In 60% of cases (n ¼ 163/271) the assailant was unknown to the patient. After controlling for age and sex, individuals in an altercation with an unknown assailant did not report more severe PTS (b¼0.005, t¼0.06, p¼0.949), p¼0.142), or depressive symptoms (b¼-0.08, t¼ -1.05, p ¼0.292) than individuals in an altercation with a known assailant.Conclusion: Among patients treated in the ED after physical assault, even when injury severity is low APNS are common. Secondary preventive interventions are needed to reduce the incidence of PTS, pain, and depressive symptoms after physical assault.