Predicting the Effects of the COVID Pandemic On US Health System Capacity | Qventus, Inc.
TO TOP

Predicting the Effects of the COVID Pandemic On US Health System Capacity

Predicting the Effects of the COVID-19 Pandemic On US Health System Capacity

Qventus works with leading health systems across the country to improve operations and drive more efficient patient flow. Over recent weeks, we have been helping our partners with COVID-19 planning and preparation, including adapting the CDC Flu Surge model to COVID-19 and running it for key metropolitan areas and all states. 

We’re sharing the information below with the hope that health systems find it informative as they plan for the impact of COVID-19. This analysis is preliminary, and we will continue to refine it as we get more information.

Summary Visualization of Model Predicting Effects of COVID-19 Pandemic on US Health System Capacity:



This model projects impact on health system capacity over the course of the COVID-19 pandemic. A more granular analysis can also be found here.

Introduction

Flattening the curve” continues to be our best path forward to minimize the impact of COVID-19. The situation in Italy and elsewhere, though, serve to illustrate the impact of not containing the virus. The United States needs to be ready should our containment measures not have as great an effect as intended.

In order to advance the discussion regarding preparatory measures that hospitals need to begin taking immediately, in this article we evaluate the potential impact to the health system when the “mitigation” phase of the pandemic is reached in the United States, and we provide a set of recommendations to support mitigation efforts.

Methodology

The Johns Hopkins Center for Health Security has recommended all health systems run the CDC Flu Surge model to better prepare for an upcoming influx of patients. They have also estimated between 1 to 10 million hospitalizations over the course of this pandemic.

Our model is based on the CDC’s publicly available FluSurge 2.0 model, which estimates the impact of a moderate (1958/1968-like, or the “Moderate” scenario in our model) or severe (1918-like, or the “Aggressive” scenario in our model) influenza pandemic on communities. We then pulled in additional data sets (see sources below) and conducted further analysis to make the results more usable, including:

  1. Factoring the model into states and CBSAs, and preloading population age group data
  2. Adding available hospital ICU (removing NICU / PICU beds) and Non-ICU (often referred to as Med-Surg) bed capacity
  3. Adding typical occupancy rates for beds by geography

Based on the limited available research so far on COVID-19, we also updated assumptions for hospital length of stay, hospitalization rate, and percent of patients requiring ICU care (see sources below).

Using this information, our model predicts the number of patients that will be hospitalized for COVID-19, the resultant effects on hospital bed, ICU, and ventilator capacity, and the number of deaths caused by the disease during the pandemic, over time and by region (state and CBSA).

Results

The results of the model are visualized below (and can be accessed here). The “Week” control advances the tool to show projected impacts on key measures over time. Note that this visualization is best viewed in a desktop browser.



The visualization treats each state as being on the same timeline; in reality, they will reach that starting point at different times, and no states have yet reached the admission rates implied by the “Week 1” starting point. While the exact time when admission rates will reach this level is unknown, based on the current trajectory, we appear to be about 4 weeks away from that point across the country given the current rate of disease spread.

We have also provided a more detailed model by Core-Based Statistical Area (CBSA), visualized below and located here.



Impact at a National Level:

Using the “moderate” assumptions (1958/1968-like), if the pandemic occurs nationwide the model projects that:

  • Approximately 6.1 million people will be infected (1.86% of the population), and 1.17 million people will be hospitalized (0.35% of the population)
  • 326,000 Med-Surg beds and 24,500 ICU beds will be needed just to care for COVID-19 patients during peak load times (above typical patient needs)
  • At peak, there will be a shortage of 9,100 ICU beds and 115,000 Med-Surg beds nationwide. At typical staffing ratios this would require 325,000 additional staff, which would be a key constraint in a situation where childcare, infection concerns and quarantine will already place a strain on existing staff availability.
  • There could be roughly 200,000 deaths caused by COVID-19

State Level Findings

Using the “moderate” assumptions, the model projects that:

  • The overwhelming majority of states (42, or 84%) will encounter capacity shortfalls over the course of 10 weeks
  • The states expected to experience the greatest capacity shortfalls (higher demand than supply) in ICU beds are Vermont (151% of capacity), Hawaii (138%), Maryland (136%), New York (136%), and Delaware (133%)
  • In addition to the above, another 27 states are projected to reach between 110% and 130% of ICU bed capacity 
  • Large states are predicted to have high numbers of patients unable to access ICU beds. According to the moderate scenario in the model, during week five New York will have a shortfall of 1,110 ICU beds, California 1,107, Florida 630, Texas 528, and Pennsylvania 402

Implications for Health Systems

Based on our ongoing work with health systems on their COVID-19 preparations, the following four key strategies should be considered for managing resources during the pandemic.

  • Redirect the “Worried Well”: Screening patients that don’t need to enter a health system has multiple benefits. It preserves scarce resources and minimizes the chances of spreading infection amongst the most vulnerable — and critically, among healthcare workers who will be needed to care for patients. Although EMTALA considerations still apply, techniques here include:
      • Establish alternate MSE sites on campus for ILI (Influenza Like Illness) screening for patients and staff (EMTALA applies)
      • Establish alternate ILI screening sites at hospital-controlled non-ED locations off campus or Community-based ILI screening locations that are not hospital-controlled (EMTALA does not apply to either) and encourage the public to access these sites for ILI screening.
  • Aggressively manage Length of Stay (LOS) for patients who do not have COVID-19: Health systems have wide variability in avoidable days (clinically unnecessary LOS). Reducing LOS by half a day across the health system would create additional effective capacity of at least 76,000 beds across the country, bridging 65% of the projected shortfall at peak.
      • In addition to managing operational flow, this will likely require more timely discharge to alternative dispositions, including Hospital at Home, etc.
  • Continuously track & evaluate capacity of key resources for the surge of COVID-19 patients, including:
    • Negative pressure / airborne infection isolation room (AIIR) capacity
    • Ventilators
    • ICU beds
  • Decant admitted COVID-19 patients outside the hospital as soon as possible: The long LOS of COVID-19 patients thus far (~10 days) means that the strain on Med-Surg units could be severe. Therefore, whenever possible, discharging COVID-19 patients expeditiously and leveraging tools like remote monitoring is advisable whenever possible

Conclusions

Planning and preparedness will be crucial to mitigating the effects of COVID-19 on our health system. As we are seeing in Italy and Washington State, the number of COVID-19 cases can increase exponentially, quickly overwhelming hospital staff and consuming critical resources. Our models suggest that an overwhelming majority (84%) of states will face critical shortages in ICU beds and overall hospital capacity.

While our model does not specifically tie resource shortages to increased mortality rates, the implied unmet demand for ICU beds highlights the potential challenges faced by health systems and public health officials in the coming months.

We hope this model and visualization will help health systems better understand the needs of their geographic regions, so they can take appropriate actions to prepare for potential outbreaks.

In a subsequent article, we will also address specific ways we are working with partners during this pandemic to deploy software that enables them to more effectively understand and manage critical resources, such as ICU beds and negative airflow isolation rooms.

For additional information or to share feedback, please contact covid19@qventus.com. Additionally, we would be happy to discuss specific implications for your health system, as well as approaches you can take to manage critical resources.

Detailed Notes on Assumptions and Methodology

  • ICU occupancy rates: We used reported Med-Surg occupancy rates by geography and assumed ICU occupancy rates were 8% higher than that (conservatively based on our experience).
  • % of patients needing ICU care: 7.5% [ranging between 5%-10%]. However, some recent estimates have estimated as high as 26% [Sources 1, 2, 3]
  • % of patients needing ventilator care: Assumed to be the same as the need for ICU, since much of the ventilator capacity is attached to ICU beds
  • ICU days during stay: This was left as the default 1968 pandemic assumption of 10 days. We found limited data here but clinicians across the country suggested 10 days was in line with what they are seeing. (It is reported as 8 days here but that also indicates a 26% rate of ICU admission. Other research reported total LOS of 14.5 days for severe patients. Additional sources include The Lancet with a reported median 7 days to death from ICU admission)
  • Non-ICU days of stay: 10 days of hospital stay [reported between 11 days – 12 days in literature] Source 1,2

Note: These represent our best assumptions based on the available research & data on COVID-19. If you have any suggestions / expertise to improve the inputs used in this analysis, we would welcome your feedback.

Data Sources

  • CDC FluSurge 2.0 Tool – FluSurge is a spreadsheet-based model which provides hospital administrators and public health officials estimates of the surge in demand for hospital-based services during the next influenza pandemic. FluSurge estimates the number of hospitalizations and deaths of an influenza pandemic (whose length and virulence are determined by the user) and compares the number of persons hospitalized, the number of persons requiring ICU care, and the number of persons requiring ventilator support during a pandemic with existing hospital capacity.
  • Population EstimatesKaiser Family Foundation estimates based on the Census Bureau’s American Community Survey, 2008-2018.
  • Additional Data Sources – State and CBSA ICU/Med-Surg capacity estimates for Qventus model aggregated from Definitive Healthcare and blended with other public data sets.