Predicting daily emergency department visits using machine learning could increase accuracy.

OBJECTIVE: Administrators and clinicians alike have attempted to predict emergency department visits for many years. The ability to predict or "forecast" ED visit volume can allow for more efficient resource allocation, including up-staffing or down-staffing, changing OR schedules, and predicting the need for significant resources. The goal of this study is to examine combinations of variables via machine learning to increase prediction accuracy and determine the factors that are most predictive of overall ED visits. As compared to a simple univariate time series model, we hypothesize that machine learning models will predict St. Joseph Mercy Ann Arbor's patient visit load for the emergency department (ED) with higher accuracy than a simple univariate time series model. METHODS: Univariate time series models for daily ED visits, including ARIMA, Exponential Smoothing (ETS), and Facebook Inc.'s prophet algorithm were estimated as a baseline comparison. Machine learning models, including random forests and gradient boosted machines (GBM), were trained using data from 2017 to 2018. After final models were created, they were applied to the 2019 data to determine how well these models predicted actual ED patient volumes in data not utilized during the model fitting process. The accuracy of the machine learning and time series models were assessed based on out-of-sample predictive accuracy, compared using root mean squared error (RMSE). RESULTS: Using root mean squared error (RMSE) to assess out-of-sample predictive accuracy of the models, the results showed that the random forest model was the most accurate at predicting daily ED visits in the 2019 test set, followed by the GBM model. These performed only slightly better than the simple exponential smoothing model predictions. The ARIMA model performed poorly in comparison. The day of the week (likely capturing differences between weekdays and weekends) was found to be the most important predictor of patient volumes. Weather-related features such as maximum temperature and SFC pressure appeared to capture some of the seasonality trends related to changes in patient volumes. CONCLUSIONS: Machine learning models perform better at predicting daily patient volumes as compared to simple univariate time series models, though not by a substantial amount. Further research can help confirm these limited initial results. Gathering more training data and additional feature engineering could also be beneficial to training the models and potentially improving predictive accuracy.

Possible exercised-induced rhabdomyolysis associated with terbinafine.

Introduction. Terbinafine is reported to be associated with rhabdomyolysis. We present a patient taking terbinafine who may have developed exercise-induced rhabdomyolysis. Case Report. A healthy 40-year-old female developed onychomycosis of the right first toe for which she was taking terbinafine. After an increase in her exercise regimen, she began experiencing notable myalgias of the triceps. During outpatient evaluation, the patient was found to have elevated and worsening creatine kinase (CK) and aspartate transaminase. At evaluation in the emergency department, CK was <5000 IU/L and had decreased. She did not have electrolyte abnormalities, kidney injury or kidney failure. Discussion. Patients may be at risk for exercise-induced rhabdomyolysis while on terbinafine and may need to be cautioned regarding the intensity of exercise.

The HAS-Choice study: Utilizing the HEART score, an ADP, and shared decision-making to decrease admissions in chest pain patients.

OBJECTIVE: The HAS-Choice pathway utilizes the HEART Score, an accelerated diagnostic protocol (ADP), and shared decision-making using a visual aid in the evaluation of chest pain patients. We seek to determine if our intervention can improve resource utilization in a community emergency department (ED) setting while maintaining safe patient care. METHODS: This was a single-center prospective cohort study with historical that included ED patients ≥21years old presenting with a primary complaint of chest pain in two time periods. The primary outcome was patient disposition. Secondary outcomes focused on 30-day ED bounce back and major adverse cardiac events (MACE). We used multivariate logistic regression to estimate the odds ratio (OR) and its 95% confidence interval (CI). RESULTS: In the pre-implementation period, the unadjusted disposition to inpatient, observation and discharge was 6.5%, 49.1% and 44.4%, respectively, whereas in the post period, the disposition was 4.8%, 41.5% and 53.7%, respectively (chi-square p<0.001). The adjusted odds of a patient being discharged was 40% higher (OR=1.40; 95% CI, 1.30, 1.51; p<0.001) in the post-implementation period. The adjusted odds of patient admission was 30% lower (OR=0.70; 95% CI, 0.60, 0.82; p<0.001) in the post-implementation period. The odds of 30-day ED bounce back did not statistically differ between the two periods. MACE rates were <1% in both periods, with a significant decrease in mortality in the post-implementation period. CONCLUSION: Our study suggests that implementation of a shared decision-making tool that integrates an ADP and the HEART score can safely decrease hospital admissions without an increase in MACE.

Unique method of ocular ultrasound using transparent dressings.

BACKGROUND: Utilizing bedside ocular ultrasound to aid in diagnosing pathology such as retinal detachment, lens disruption, ocular foreign bodies, or increased intracranial pressure is becoming more pervasive in the Emergency Department. To eliminate an air interface, one must apply ultrasound gel between the patient's skin and the probe. In ocular ultrasound, this practice results in discomfort for the patient as gel seeps into their eyes. To limit patient discomfort, many physicians do not apply a sufficient amount of gel for the examination. This can result in decreased image quality and may cause the ultrasonographer to apply greater pressure to the eye to obtain a satisfactory image. This can be harmful to patients with a ruptured globe and may also be painful to the patient. DISCUSSION: Traditionally, the first step in ocular ultrasound is to place a generous amount of water-soluble ultrasound gel on the eyelid to eliminate the air interface. The authors promote a different and simple technique. A transparent dressing is placed over a closed eye. A generous amount of ultrasound gel is applied to the dressing. A linear ultrasound probe is then placed on the gel and a standard ultrasound scan is obtained. Transparent dressings, which are used as sterile coverings for i.v. sites, have been found to allow satisfactory ultrasound transmission. These products remove the air interface between the eyelid and the dressing. This allows ultrasound gel to be placed on the transparent dressing and not directly on the eyelid, potentially eliminating discomfort for the patient, and creating an easier cleanup. Because a generous amount of ultrasound gel is applied, the ultrasonographer is able to apply minimal pressure on the eye to complete the study, which may decrease harm to the patient's eye. When finished, the transparent dressing is removed. There is no cleanup or patient irritation. CONCLUSION: This article demonstrates a unique method of ocular ultrasound. The technique can be easily incorporated into emergency bedside ocular ultrasound.