Correlation between Emergency Medical Services Suspected COVID-19 Patients and Daily Hospitalizations

Objective: We sought to determine if Emergency Medical Services (EMS) identified Persons Under Investigation (PUI) for COVID-19 are associated with hospitalizations for COVID-19 disease for the purposes of serving as a potential early indicator of hospital surge. Methods: A retrospective analysis was conducted using data from the Maryland statewide EMS electronic medical records and daily COVID-19 hospitalizations from March 13, 2020 through July 31, 2020. All unique EMS patients who were identified as COVID-19 PUIs during the study period were included. Descriptive analysis was performed. The Box-Jenkins approach was used to evaluate the relationship between EMS transports and daily new hospitalizations. Separate Auto Regressive Integrated Moving Average (ARIMA) models were constructed to transform the data into a series of independent, identically distributed random variables. Fit was measured using the Akaike Information Criterion (AIC). The Box-Ljung white noise test was utilized to ensure there was no autocorrelation in the residuals. Results: EMS units in Maryland identified a total of 26,855 COVID-19 PUIs during the 141-day study period. The median patient age was 62 years old, and 19,111 (71.3%) were 50 years and older. 6,886 (25.6%) patients had an abnormal initial pulse oximetry (<92%). A strong degree of correlation was observed between EMS PUI transports and new hospitalizations. The correlation was strongest and significant at a 9-day lag from time of EMS PUI transports to new COVID-19 hospitalizations, with a cross correlation coefficient of 0.26 (p < .01). Conclusions: A strong correlation between EMS PUIs and COVID-19 hospitalizations was noted in this state-wide analysis. These findings demonstrate the potential value of incorporating EMS clinical information into the development of a robust syndromic surveillance system for COVID-19. This correlation has important utility in the development of predictive tools and models that seek to provide indicators of an impending surge on the healthcare system at large.