OBJECTIVE: The ability to forecast atypical emergency department (ED) volumes may aid staff/resource allocation. We determine whether deviations from short-term predictions of weather can be used to forecast deviations from short-term predictions of ED volumes. METHODS: In this retrospective study, we attempted to predict the volume of patient visits to an academic pediatric ED based on short-interval local weather patterns (2000). Local temperature and precipitation data in 1- and 3-hour increments were obtained. Precipitation was coded to be present if it exceeded 0.04 in and subclassified as cold rain/snow if the ambient temperature was lower than 40 degrees F. ED visits were categorized as injuries, emergent, or nonemergent visits. For each category of visit, Box-Jenkins Auto-Regressive Integrated Moving Average time-series models were created of natural trends and cycles in temperature and patient volumes. From these models, differences (residuals) between predicted and observed values of these variables were estimated. The correlation between residuals for temperature and ED volumes was derived for various kinds of ED visit, after controlling for type/volume of precipitation. RESULTS: Residuals for ambient temperature controlled for precipitation correlated poorly with residuals for patient volumes, accounting for 1% to 6% of the variability in the volume of injuries, emergent, and nonemergent visits (R2 = 1%, 1%, and 6%, respectively). CONCLUSIONS: Deviations from short-term predictions of temperature correlate poorly with deviations from predictions of patient volume after adjusting for natural trends and cycles in these variables and controlling for precipitation. These weather variables are of little practical benefit for predicting fluctuations in the rates of ED utilization.
OBJECTIVE: The ability to forecast atypical emergency department (ED) volumes may aid staff/resource allocation. We determine whether deviations from short-term predictions of weather can be used to forecast deviations from short-term predictions of ED volumes. METHODS: In this retrospective study, we attempted to predict the volume of patient visits to an academic pediatric ED based on short-interval local weather patterns (2000). Local temperature and precipitation data in 1- and 3-hour increments were obtained. Precipitation was coded to be present if it exceeded 0.04 in and subclassified as cold rain/snow if the ambient temperature was lower than 40 degrees F. ED visits were categorized as injuries, emergent, or nonemergent visits. For each category of visit, Box-Jenkins Auto-Regressive Integrated Moving Average time-series models were created of natural trends and cycles in temperature and patient volumes. From these models, differences (residuals) between predicted and observed values of these variables were estimated. The correlation between residuals for temperature and ED volumes was derived for various kinds of ED visit, after controlling for type/volume of precipitation. RESULTS: Residuals for ambient temperature controlled for precipitation correlated poorly with residuals for patient volumes, accounting for 1% to 6% of the variability in the volume of injuries, emergent, and nonemergent visits (R2 = 1%, 1%, and 6%, respectively). CONCLUSIONS: Deviations from short-term predictions of temperature correlate poorly with deviations from predictions of patient volume after adjusting for natural trends and cycles in these variables and controlling for precipitation. These weather variables are of little practical benefit for predicting fluctuations in the rates of ED utilization.
Authors: Rafael Calegari; Flavio S Fogliatto; Filipe R Lucini; Jeruza Neyeloff; Ricardo S Kuchenbecker; Beatriz D Schaan Journal: Comput Math Methods Med Date: 2016-09-20 Impact factor: 2.238