Implementation of COVID-19 Surveillance Testing Procedures in Dental School

Background. Dental practitioners and students of dentistry are potentially at increased risk of COVID-19 infection due to frequent usage of aerosol-generating procedures. To mitigate risk to patients and providers, the University of Utah School of Dentistry began regular surveillance PCR testing of its patient-facing faculty, staff, and students in May 2020. Methods. Surveillance testing occurred every other week for non-vaccinated individuals and continued through February 2022. After May 2021, fully vaccinated individuals were tested monthly and encouraged to seek additional testing if symptoms or an exposure occurred. We assessed risk of positive test among faculty, student, and staff groups through a Cox proportional hazards regression, accounting for multiple events and time-dependent variables with the Andersen-Gill model. To account for inconsistent testing after vaccination, time was examined as number of tests rather than calendar time. Results. In total, 410 participants were followed during the observation period, with an average of 22 (SD 10.0, RNG 1-50) tests per person. A total of 9,452 tests were performed. There were 158 positive tests, with 60 (38%) occurring in January 2022 alone. When analyzed by themselves, staff and student groups were significantly more likely to test positive (HR 1.98, 95% CI 1.15-3.42;HR 2.16, 95% CI 1.29-3.63 respectively) compared to faculty. However, once additional covariates were accounted for, the relationship was no longer significant (Staff: HR 2.15, 95% CI 0.92-5.05;Students: HR 2.38, 95% CI 0.88-6.40). Risk of COVID-19 within Dental School Hazard Ratios for testing positive for COVID-19 among different groups within the dental school. Vaccination is accounted for as time since last vaccine, with separate categories for one dose, and 2 or more doses combined. Time examined as test number. Conclusion. More than a third of all positive tests during the 22-month study occurred during one month of the Omicron wave. This sudden increase in positive tests was not observed in previous surges, and demonstrates the intensity of the Omicron wave. Additionally, we did not find a significant difference between patient-facing groups who had different work exposures. While this may be due to effective preventative measures, within the dental setting we do not see evidence that work role and resulting exposures increase risk.

Impact of COVID 19 on urgent care diagnoses and stewardship metrics

Background. The percentage of all respiratory diagnoses prescribed an antibiotic is an outpatient stewardship metric and was introduced as a HEDIS measure in 2022. Given a stable case mix, this metric is not affected by differences in coding practices between clinicians or health systems since all respiratory diagnoses are considered together. The onset of the COVID-19 pandemic introduced a high number of viral illness episodes where antibiotics are not recommended. The impact of this shift in case mix on respiratory diagnosis coding and prescribing metrics has not been explored. Methods. We examined antibiotic prescribing rates for respiratory diagnoses in a network of urgent care clinics affiliated with the University of Utah during two periods. Pre-Pandemic was Mar 2019-Feb 2020 and Pandemic was Mar 2020-Mar 2022. Respiratory diagnoses were identified using ICD10 codes and further stratified into 3 Tiers (Tier 1: antibiotics indicated;Tier 2: antibiotics sometimes indicated;Tier 3: antibiotics not indicated). We examined trends in antibiotic prescribing across these periods including the percentage of all respiratory visits prescribed antibiotics and by Tier and the distribution of diagnoses by Tier. No formalized stewardship interventions were introduced during these periods. Results. There were 146,897 urgent care visits during the study period (47,423 Pre Pandemic and 99,474 Pandemic). The respiratory prescribing rate declined from 42.3% Pre Pandemic to 26.2% during the Pandemic (Figure). The distribution of respiratory diagnoses by Tier and prescribing within Tier are shown in the Table. Tier 3 diagnoses increased from 48% to 67%, while Tier 2 diagnoses declined from 47% to 31%. Antibiotic prescribing declined for both Tier 2 and Tier 3 diagnoses. 15,429 (23%) of Tier 3 diagnoses during the Pandemic were coded as COVID-19. 50% of the reduction in prescribing is attributable to changes in Tiers alone. Figure Table Conclusion. The COVID 19 pandemic was associated with a reduction in the percentage of respiratory diagnoses prescribed antibiotics. Half was due to an increase in Tier 3 encounters although declines in prescribing occurred with Tiers in addition. Using this metric for benchmarking requires accounting for the impact of case mix differences over time or between systems and clinicians.

Predicting next 7-day Discharges of Hospitalized COVID-19 Patients Using Ensemble Learning

Background. COVID-19 pandemic, especially during resurgences of cases in hard-hit areas, led to significant shortage of hospital beds. Such shortages may be alleviated through timely and effective forecasting of hospital discharges. The objective of this study is to predict next 7-day discharges of hospitalized COVID-19 patients using daily-based electronic health records (EHR) data. Methods. Using EHR data of hospitalized COVID-19 patients from 03/2020-08/ 2021, we employed ensemble learning to predict next 7-day discharges of individual patients. We used both baseline and daily inpatient features for model training, validation, and test. Baseline features include demographic and clinical characteristics, and comorbidities. The daily inpatient features were vital signs, laboratory tests, medications administered, acute physiological scores, use of ventilator, and use of intensive care unit. 1832 hospitalized patients were identified (12,397 hospital days). Samples were randomly split at patient level (7:2:1) into training set (N=1,283 patients with 8,704 hospital days), validation set (N=366 patients with 2,524 days), and test/ holdout set (patient N=183, and 1,169 days). Prediction models were trained on the training set and the validation set. We conducted the model training separately on the samples of admission day and the samples of days after admission day. The predictions were based on the ensemble learning from decision tree, XGBoost, logistic regression, and multilayer perceptron, long short-term memory (LSTM), bi-directional LSTM, and convolutional neural network. The combination of ensemble learning on the test/holdout set was used for final next 7-day predictions based on 'hard' voting (by majority). Where there was a tie, we used 'soft' voting (sum of probabilities) to break the tie. (Figure Presented) Results. The overall average hospital length of stay was 8.7 (SD=10.5) days. The ensemble learning accuracies for admission-day samples and after-admission-day samples were 0.781 and 0.793, and the F1-scores for were 0.761 and 0.789, respectively. Conclusion. EHR data of hospitalized COVID-19 patients can be used to predict next 7-day hospital discharges. Additional inpatient features and more advanced machine learning techniques are needed for prediction accuracy improvement.

Topic-to-Topic Modeling for COVID-19 Mortality

We examine a cohort of 4307 COVID-19 case fatalities from a de-identified national registry in the U.S. using Latent Dirichlet Allocation and group each patient by topic based on their pre-existing conditions in the years prior to infection and again during the last three weeks of life. We show that certain pre-existing condition topics have strong associations with certain COVID-19 mortality topics suggesting that the major clinical pathways leading to COVID-19 death may be through failures of already weakened organ systems. We then explore the demographics for these groups and generate several insights and hypotheses. © 2021 IEEE.