Mapping US health system capacity (ICU care in particular) for COVID19 surge preparedness

Hi everyone,

Thanks for providing this forum space dedicated to COVID19 work!

I’ve been.working with a volunteer team of geospatial experts, engineers, data scientists, designers, & communicators to build data, analyses, and tools that help us better understand, anticipate, and act to support and ramp up our local health systems’ capacity (providers, supplies, ventilators, beds, meds) to effectively care for a rapidly growing flow of COVID19 patients.

We’re doing open-source data collection, geospatial analysis, visualizations, and scenario planning tools with a strong focus on the bottleneck of healthcare treatment capacity for critically ill COVID19 patients (i.e. ICU-level beds, equipment, staffing needs). Aiming for high spatiotemporal data resolution (down to individual facility or county level per day across the USA) for current actuals/estimates and scenario projections going forward. It’s USA-centric at the moment but data, tools and knowhow are fully open and intended to be applicable globally.


We’re looking for more volunteers and/or collaborations to help us achieve these high-level project goals as quickly and robustly as possible:

  • A. Publish a public dataset that describes the US healthcare system capacity in high spatiotemporal detail.

  • B. Perform an analysis of the healthcare system capacity as compared to disease spread forecasts and publish data about where, when, and how large capacity gaps are anticipated to be.

  • C. Visualize the result of the analysis as well as the disease forecast and healthcare system capacity data in a format that supports healthcare system preparedness and resourcing decisions.

If you’re interested to learn more, start contributing, or collaborate:

Happy also to answer any questions and share more thoughts and ideas…let me know what you’re thinking!

Thanks, looking forward, and be well,


I love the idea and will try to integrate the ICU beds data as a datapoint in our map/algorithm. Let me know if I can help you integrate any of our data as well. It’s been mostly just me and I only started a week ago so the whole project is in a state of flux, but I can get you a stable format for county level data (sourced through these guys and aggregated/cleaned by some friends of mine to match our existing county data.

Thanks so much for taking the time/care to lay out your reasoning, thoughts and methodology. I want to do the same for our project but I’ve been racing to build stuff and feel like I haven’t had time, but I see the value in it through your post.

Questions about the project:
-Do you plan to try to model the adjustment that will take place when ICUs reach capacity, such as adding new beds, repurposing sections of the hospital, building field hospitals…etc?
-What’s next?

1 Like

Hey, thanks for that link and your thoughts! We should have a cleaned up dataset merging the HIFLD and HCRIS datasets soon so there’ll be total licensed and staffed beds (distinguished between ICU and regular beds) and their historic occupancy rates per most geolocated facilities very soon. Will let you know when there’s a release so you can put it to use asap and also help be an external set of eyes to see if there are any dataset issues.

I’ve found that writing out the rationale and proposed methodology, at least a draft, and sharing that has helped improve with my own understanding and mental organization of what the problems and solutions are. And it has also been a very useful communication tool to rally people and onboard volunteers asynchronously (without me having to directly explain to every person the project every time). So I would definitely recommend some writing!


-Do you plan to try to model the adjustment that will take place when ICUs reach capacity, such as adding new beds, repurposing sections of the hospital, building field hospitals…etc?

Yup, that’s exactly right. That’s the next step once we have the total licensed ICU beds data (representing 100% possible capacity in a facility in usual surge planning) and staffed ICU beds and occupancy rates (business as usual availability of beds and staffing). We already have a methodology in mind for estimating the amt of equipment needs and staffing ratios required to go how far beyond that 100% capacity. These steps are starting to be laid out in these two issues:

1 Like

Hey, do you have any update on the timeline for when you’ll have a cleaned/merged dataset that you feel is clean enough for release and public consumption? I’m about to start down the path of tuning the risk algorithm and might use age and comorbidity data if I can find it but would rather not reinvent the wheel for ICU beds as it looks like you guys are doing great work. Please keep me posted on your progress and let me know if there’s any data I’ve aggregated that would be useful for you.

1 Like

Yes, aiming for 1st release of the data tomorrow. Will let you know and post here when that’s up.


Hi @MadeUpMasters and all:

Our COVID Care Map team (updated project name) just completed a 1st (version 0.1) data release of ICU or total bed counts and their typical occupancy rates for every hospital facility in the US (based on recent 2018 cost reporting and other data sources). Here’s the interactive map explorer of this free, open, analysis-ready dataset:

In addition to facility-level data, we’ve aggregated to County, Hospital Referral Region, and State for comparison and analysis at these levels. We also show numbers per capita for total population, adults (age 20+), and elderly (65+) in those same spatial aggregations.

We know that this first cut of data is not 100% complete, current, or accurate and it may never be fully so, given the urgency of what’s needed and not letting perfect be the enemy of good.

We’re working on ways to correct, fill in, and update data at scale to ideally to make this as close to a real-time understanding of health system capacity as possible. And have this data layer be readily usable for analytics and decisionmaking by us or anyone else who wants to work with the data. This is why everything is open and we’ve tried to document all of our sources and methodology as clearly as possible for review, reproduction, and improvements:

Let us know as you start to work with the data if there’s anything off or how we can make it better and more useful for you.

Besides looking at ICU and total bed availability, we also have parallelized work and volunteers starting to model current and projected needs re: staffing for intensive care, ventilators, supplies, and coordination systems. See our github project and issues for what’s going on in these various area:


You all have done a fantastic job on this, I will get to work on integrating the data into our project in the next 1-3 days. Let’s keep in touch as you progress and if I make any big leaps on processing health data, age and comorbidity stuff I’ll make sure to ping you.

1 Like

Wow @daveluo that looks awesome. I wonder if we could potentially use that in our tool? I have been building a simple map which isnt live yet using Altair and Vega but looks like you used mapbox? I suspect I might need to change what im doing to make it more flexible for clustering etc.

I am working with some guys who have built a tool to help reduce the risk of disease spread at the point of triage while also providing medical staff to access to high risk patient data.

Our beta with updated interface just went live.

We are hoping we can help reduce the risk that medical staff and community are being exposed to by easily allowing people to provide the result of their self screening directly to their doctor, and optionally connect it with their MyChart patient data, so that doctors can monitor or contact high risk patients and treat them while low risk patients are reassured they don’t need to get a test yet or see a doctor in person.

Does anyone have any thoughts or feedback on this?
We are building out some other features already but it would be great to combine our efforts with other teams and also to get some valuable feedback from other people doing similar things.

Quick description of what we are doing:
The COVID Assessment Tool for Community Health (CATCH) is a FREE Zero Configuration self-screening & hospital intake web application which provides global support to individuals, health systems & organizations who need to screen, manage and triage patients for COVID-19.

  • Patients can self screen using a questionnaire. It is based on guidelines from the CDC to identify PUIs (Patients Under Investigation). Based on their answers, they get a recommendation immediately that helps them receive appropriate care.
  • Medical staff can identify urgent cases immediately, based on patient self-screenings allowing them to ask only those specific patients to come in, reducing backlogs.
  • Surges in patient need can be identified rapidly and precious resources can be optimally allocated and shared between hospitals.

Why are we different?
Unlike most other COVID tools, we have a dashboard available for Doctors and Hospitals to access relevant data on high-risk patients in their health system/region.

How does it work?
Patients simply answer a questionnaire, add their email, and link their primary care provider or hospital. Next, medical staff are notified and can identify urgent cases immediately, based on a patient’s survey, medical records and/or our automated triage system.

Without any signup anyone with the following email can use the system: -> records -> Region Data that is located in.

1 Like

Have either of you @daveluo or @madhavajay seen any resources for real time hospital/ICU occupancy at the county level? The only thing I’m aware of is that the Covid Tracking Project has started including covid19 hospitalizations by state, but it’s only state level. Cheers.

@MadeUpMasters I haven’t seen anything like that yet, but I have heard that inter-hospital resource sharing tools are in high demand.

This may be old but do you still need help on this? I’m interested in this, especially calculating per county healthcare capacities, and disease forecasts