Personal protective equipment (PPE) is needed to protect healthcare workers when treating COVID-19 patients. The parameters below define the average amount of PPE used per interaction with a COVID patient.
The total amount of PPE used depends on the number of interactions between a healthcare worker and COVID patient, where an interaction is defined as the worker having to wear a new set of PPE (under normal operationg procedure) to perform their duties. The frequency of patient interactions is likely correlated with patient symptom severity. The parameters below define the average number of interactions between a healthcare worker and a patient in a day by patient severity.
PPE is normally replaced between each patient interaction. However, the rapid spread of COVID-19 resulted in a surge of demand for PPE, in particular, N95 respirators. Given the constrained supply of some PPE products, healthcare facilities have implemented strategies to extend the use of PPE. The parameters below define the "Conserve" PPE usage policy, where each coefficient represents the number of times a PPE item is used on average before being replaced.
This model was developed to perform short-term predictions of future demand for personal protective equipment (PPE) needed by healthcare and other essential workers when in contact with COVID-19 patients. Model predictions are driven by the confirmed number of COVID-19 cases within a specified location.
As seen below in the process diagram, the model is comprised of three
modules including: prediction of COVID-19 cases, stratification of
COVID-19 patients by severity, and prediction of PPE demand. Prediction
of COVID-19 cases was achieved by curve fitting to historical data using
a weighted average between an exponential, quadratic, and logistic
function. These mathematical functions are widely used to capture the
different dynamics and phases in a disease outbreak. Stratification of
COVID-19 patients was done by specifying the expected proportion of
patients to be hospitalized, to be in the ICU but not on a ventilator,
and finally to be in the ICU and on mechanical ventilation. It is
assumed that the patient would remain at the hospital for different
lengths of time (7-14 days) based on their symptom severity. Prediction
of PPE demand is then finally calculated based on the number of COVID-19
patients and specified coefficients representing the average PPE used
per patient. Users can select between “Normal” (i.e., PPE replaced for
each patient interaction) and “Constrained” (i.e., PPE use extended to
conserve inventory) use models. Model Parameters were selected based on
literature, consultation with subject matter experts, published
guidelines from Centers for Disease Control and Prevention, and from
The model is intended for short-term forecasting PPE demand over a period 3 weeks. While predictions can be made for longer-term demand, there is a low level of confidence in the results. This limitation is due to the use of curve fitting to predict COVID-19 cases, which is well-suited for near-term but not long-term predictions.
PPE demand was also estimated for non-COVID hospital usage such as police, firefighters, emergency medical technicians (EMT), long term care facilities, and home health care. Demand estimates for these sources were calculated based on the total number of employees in each respective industry and their anticipated weekly PPE usage per employee. The number of employees in a given county or state was estimated based on the 2019 occupational employment statistics from the U.S. Bureau of Labor Statistics.
A surge in COVID-19 patients has led to an increase in demand for drugs commonly used in the ICU and commonly used during mechanical ventilation. To identify and prevent potential future drug shortages, this model provides the estimated demand for the use of drugs at hospitals to treat COVID-19 patients and for the use of drugs at hospitals for general non-COVID-19 use. This model was developed by analyzing electronic health records at health systems comprised of nearly 20,000 hospital beds dispersed throughout the United States.
The purpose of this tool is to assist decision makers in planning for a potential surge in COVID-19 cases by forecasting cumulative case counts over a 4-month window at different severity levels (mild, moderate, severe, and very severe). Surge scenarios were created using a SEIR model, a widely used compartmental model in epidemiology that involves a patient population transitioning through the following compartments: susceptible, exposed, infected, and death or recovered. Additional information on SEIR models can be found here. Parameters for the SEIR model were found by fitting the model to historical COVID-19 cumulative case counts and deaths for each county and state. Counties are then grouped together based on their 2013 NCHS Urban and Rural Code under the assumption that a surge within a metropolitan area would likely be different compared to one in a rural area. For each urban and rural code group, counties are ranked and ordered based on their historical surge severity. Surges for different scenarios are then generated based on the fitted model parameters of counties as following: mild (21-40 percentile), moderate (41-60 percentile), severe (61-80 percentile), and very severe (81-100 percentile). Upper and lower bounds for each scenario are also calculated by increasing/decreasing the SEIR model infection rate parameter by 5%.
This model and dashboard were developed as a collaboration effort between Mitre, Cardinal Health, Llamasoft, LogicStream Health, Sodexo Healthcare Services, and GHX.