A Comprehensive County Level Framework to Identify Factors Affecting Hospital Capacity and Predict Future Hospital Demand

Abstract

Background: As of February 19, 2021, our review yielded a small number of studies that investigated high resolution hospitalization demand data from a public health planning perspective. The earlier studies compiled were conducted early in the pandemic and do not include any analysis of the hospitalization trends in the last 3 months when the US experienced a substantial surge in hospitalization and ICU demand. The earlier studies also focused on COVID 19 transmission influence on COVID 19 hospitalization rates. While this emphasis is understandable, there is evidence to suggest that non COVID hospitalization demand is being displaced due to the hospitalization and ICU surge. Further, with the discovery of multiple mutated variants of COVID 19, it is important to remain vigilant in an effort to control the pandemic. Given these circumstances, the development of a high resolution framework that examines overall hospitalizations and ICU usage rate for COVID and non COVID patients would allow us to build a prediction system that can identify potential vulnerable locations for hospitalization capacity in the nation so that appropriate remedial measures can be planned. Method: The current study recognizes that COVID 19 has affected overall hospitalizations, not only COVID 19 hospitalizations. Drawing from the recently released Department of Health and Human services (DHH) weekly hospitalization data (or the time period August 28th , 2020 to January 22nd, 2021.), we study the overall hospitalization and ICU usage as two components: COVID 19 hospitalization and ICU per capita rates; and non COVID hospitalization and ICU per capita rates. A mixed linear mixed model is adopted to study the response variables in our study. The estimated models are subsequently employed to generate predictions for county level hospitalization and ICU usage rates in the future under a host of COVID 19 transmission scenarios considering the new variants of COVID 19 and vaccination impacts. Findings: We find a significant association of the virus transmissibility with COVID (positive) and non COVID (negative) hospitalization and ICU usage rates. Several county level factors including demographics, mobility and health indicators are also found to be strongly associated with the overall hospitalization and ICU demand. Among the various scenarios considered, the results indicate a small possibility of a new wave of infections that can substantially overload hospitalization and ICU usage. In the scenario where vaccinations proceed as expected reducing transmission, our results indicate that hospitalizations and ICU usage rates are likely to reduce significantly. Interpretation: The research exercise presents a framework to predict evolving hospitalization and ICU usage trends in response to COVID 19 transmission rates while controlling for other factors. Our work highlights how future hospitalization demand varies by location and time in response to a range of pessimistic and optimistic scenarios. Further, the exercise allows us to identify vulnerable counties and regions under stress with high hospitalization and ICU rates that can be assisted with remedial measures. The model will also allow hospitals to understand evolving displaced non COVID hospital demand.

Using Public Data to Predict Demand for Mobile Health Clinics

“…However, they do not consider mobile health clinics or temporal features. Similarly, there are several papers that use multiple regression models for demand/utilization prediction of traditional hospitals and healthcare centers using hospitalization rates and ICU beds (Bhowmik and Eluru 2021), radiology volume records (Côté and Smith 2018), surgeries and admissions volume (Khaldi et al 2017), and blood donation records (Drackley et al 2012). However, none of these studies can be directly used or mildly adapted for the task of predicting the demand of mobile clinics.…”

A Multi-Period Curve Fitting Model for Short-Term Prediction of the COVID-19 Spread in the U.S. Metropolitans

“…Placio et al ( 40 ) established that for Miami Dade county the COVID-19 infection is associated with economically disadvantaged population and shows no association with racial/ethnic distribution. Bhowmik et al ( 41 ) found a significant association of demographics, mobility, and health indicators with COVID-19 hospitalization and ICU usage. Bhowmik and Eluru ( 41 ) also developed a model framework to evaluate the impact of mobility on transmission rates in the county while accommodating county-specific features.…”

“…Bhowmik et al ( 41 ) found a significant association of demographics, mobility, and health indicators with COVID-19 hospitalization and ICU usage. Bhowmik and Eluru ( 41 ) also developed a model framework to evaluate the impact of mobility on transmission rates in the county while accommodating county-specific features. Iyanda et al ( 42 ) established that the case fatality ratio in the rural counties, and in people of color is higher than the national rate highlighting the health disparities in these groups.…”