Develop estimate of ventilator capacity

[lossyrob]

We are targeting collecting data that can serve as inputs to the FluSurge model:

Even if we don't use FluSurge as part of #9, these inputs will most likely be useful or necessary for whatever modeling we end up using.

This includes ventilator availability. One estimate of ventilator counts is to just use the number of licensed ICU beds. However a better estimate could be made if there are ventilator count datasets available.

The goal of this issue is to establish a dataset that estimates ventilator availability for US healthcare facilities. This should be in a format that is join-able to the HRCIS data (usa_hospital_beds_hcris2018 data in the data folder).

See https://gitter.im/covid19-healthsystemcapacity/community?at=5e6efd3b1f0d285eb280873b

[daveluo]

There's an excellent figure from the study cited in #18 showing what components are essential to providing effective mechanical ventilation care. This includes the actual machine but also the appropriate providers roles and ratios of staffing needed:

https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4636910/figure/F1/

Having a good estimate of all these components at the spatiotemporal granularity we're working at ends up being important because as we try to ramp up critcare capacity, the bottlenecks that emerge are different:

The number of available critical care physicians was the most constraining key component at the conventional capacity level, limiting the maximum number of ventilated patients to 18,900. The number of available critical care and intermediate care beds was the constraining key component at the contingency capacity level, limiting the maximum number of ventilated patients to 52,400. At the crisis capacity level, the number of available respiratory therapists was the key constraining component, limiting the maximum number of ventilated patients to 135,000. (Table 2)

https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4636910/table/T2/

[daveluo]

This brief on Ventilator Stockpiling and Availability in the US cites a 2010 paper to estimate a total of 160K ventilators available in USA:

Based on the most recent publicly available data (from 2010), 1 study estimated that US acute care hospitals own approximately 62,000 full-featured mechanical ventilators.1 Calculations suggest that about 28,883 of these ventilators (46.4%) can be used to ventilate pediatric and neonatal patients. The study also reported an additional 98,000 ventilators that are not full-featured but can still provide basic function in an emergency during crisis standards of care.

[daveluo]

And here's that paper: https://www.cambridge.org/core/journals/disaster-medicine-and-public-health-preparedness/article/mechanical-ventilators-in-us-acute-care-hospitals/F1FDBACA53531F2A150D6AD8E96F144D

Relevant tables:

Given their estimation methodology is probably sound enough (to be cited so widely) and their data source is from the 2007 AHA survey, we could use their ratios and inputs and our more recent bed counts that we are collecting or estimating from years 2018-2020.

For example, we could use the same "Ratio of MV to total ICU beds" of 0.703 in Table 1 and our Total Licensed ICU Bed counts to get a more recent national vent supply estimation and then split it out spatially by state, county, etc

[ghost]

I am still ramping up on this stuff a bit, so forgive me any error.

However, i think using the strong correlation between total ICU beds and mechanical ventilators from the 2010 survey makes me feel that if we trust our ICU bed numbers we should be able to infer number of ventilators at appropriate granularity.

One thing that popped up in the literature review I did is that many of the papers in modeling crisis situations are seemingly underestimating staff exposure and absenteeism. Due to that fact that several papers also mentioned respiratory therapist availability as the primary bottleneck in ventilating patients in their crisis models, we should carefully consider the parameters we use in modeling staffing capacity.

One more note: there wasn't a ton of information on capacity/provisioning of ventilator accessories. I am far from an expert in this area, do ventilator vents/face-masks or other peripheral components have to be frequently replaced?

[daveluo]

Thanks @laurenmoos for the analysis! That sounds good then for us to use the same ratio of total ICU beds to vents to update the estimate for full-featured vents by facility once we're solid on the ICU bed number or its estimate.

One question raised in those readings and here is the activating of back-up/basic feature vents to augment the supply and/or even using one vent for multiple patients. Any of these factors, if successful, could change the supply calculation quite dramatically. Worth a quick look as to what's being discussed as viable options?

I don't personally know the answer re: the supply chain for vent accessories. Most non-durable hospital supplies generally speaking is treated as one-time use for hygiene and infection control reasons. Although in our situation now, that's clearly not the case (with the scarcity and thus conserving and reuse of personal protective equipment like face shield masks, N95s, etc). It's a great observation. I don't know if we need to go down the rabbit hole of this question just yet as the bigger question is whether we even have enough ventilator machines but it's definitely something to keep in mind as a possibly emergent constraint down the road.

Re:

One thing that popped up in the literature review I did is that many of the papers in modeling crisis situations are seemingly underestimating staff exposure and absenteeism. Due to that fact that several papers also mentioned respiratory therapist availability as the primary bottleneck in ventilating patients in their crisis models, we should carefully consider the parameters we use in modeling staffing capacity.

This is an absolutely critical point about staffing as a bottleneck. Of the models I've seen that look at estimating ICU care capacity, I don't think any of them have modeled what staff attrition looks like (or holds some assumption as a constant) for any of the roles (physician, nurse, resp therapist). Some understanding of what staff attrition looked like as % of infected in population or hospital case loads grew (in China, Italy, elsewhere?) and modeling of this change is definitely needed as we enter a pretty dynamic period where patients increase rapidly, beds are scarce, providers get exposed, etc. Probably it's own issue. I can create one or please feel free to go ahead to write it.

[andrewwgithub]

Has anyone found more updated numbers for ventilators other than from 2007?

[daveluo]

That 2010 study linked above seems to be at least the most widely cited estimate. I've not seen a newer one yet. The SCCM's COVID resource page did provide some additional high-level estimates about vent supply (and other needs) in the last few days:

https://sccm.org/Blog/March-2020/United-States-Resource-Availability-for-COVID-19:

Mechanical ventilators: Reports from ICUs worldwide suggest that the most common reason for COVID-19 patient admission to the ICU is severe hypoxic respiratory failure requiring mechanical ventilation.

Supply of mechanical ventilators in U.S. acute care hospitals: Based on a 2009 survey of AHA hospitals, U.S. acute care hospitals are estimated to own approximately 62,000 full-featured mechanical ventilators.10,11 Approximately 46% of these can be used to ventilate pediatric and neonatal patients. Additionally, some hospitals keep older models for emergency purposes. Older models, which are not full featured but may provide basic functions, add an additional 98,738 ventilators to the U.S. supply.10 The older devices include 22,976 noninvasive ventilators, 32,668 automatic resuscitators, and 8567 continuous positive airway pressure units.

Centers for Disease Control and Prevention Strategic National Stockpile (SNS) and other ventilator sources: The SNS has an estimated 12,700 ventilators for emergency deployment, according to recent public announcements from National Institutes of Health officials.12 These devices are also not full featured but offer basic ventilatory modes. In simulation testing they performed very well despite long-term storage.13 Accessing the SNS requires hospital administrators to request that state health officials ask for access to this equipment. SNS can deliver ventilators within 24-36 hours of the federal decision to deploy them. States may have their own ventilator stockpiles as well.14 Respiratory therapy departments also rent ventilators from local companies to meet either baseline and/or seasonal demand, further expanding their supply. Additionally, many modern anesthesia machines are capable of ventilating patients and can be used to increase hospitals' surge capacity.

The addition of older hospital ventilators, SNS ventilators, and anesthesia machines increases the absolute number of ventilators to possibly above 200,000 units nationally. Many of the additional and older ventilators, however, may not be capable of sustained use or of adequately supporting patients with severe acute respiratory failure. Also, supplies for these ventilators may be unavailable due to interruptions in the international supply chain. Alternatively, ventilator manufacturers could be encouraged to rapidly produce modern full-featured ventilators to allow experienced clinicians to use supplemental ventilators that are familiar to them and can be readily incorporated into the hospital ventilator fleet and informatics systems. An analysis of the literature suggests, however, that U.S. hospitals could absorb a maximum of 26,000 to 56,000 additional ventilators at the peak of a national pandemic, as safe use of ventilators requires trained personnel.15