Creating a plan to reprogram smart CPAP machines to become emergency ventilators
IMPORTANT NOTE: Several people have skimmed this article and gotten the mistaken impression it proposes using CPAP or BiPAP machines to treat patients who need a ventilator, using CPAP masks with open exhaust ports. It does not. It is about converting the hardware currently used for CPAP to be mechanical ventilators, not CPAP, for some patients, in a way very similar to the FDA approved ventilators which may be in very short supply.
As I'm sure, you've heard about the need that many Covid-19 patients have for ventilators which keep them breathing when their lungs fail, as they do during the "ARDS" (acute respiratory distress syndrome) phase of Covid-19 which is the thing that kills you. The problem is that there are at most around 200,000 ventilators (including tapping older models sitting in storage, a government strategic reserve and a military reserve.) It is feared that as many as 900,000 could be needed if the worst projections are true. Around the world, many more.
Companies that make ventilators are ramping up manufacturing as fast as they can. It still won't be enough. That's even true if one applies a special technique devised years ago to put 2, 4 and even 9 patients on the same machine, if the patients can be matched so they need the same pressures, airflow and oxygen. This technique was tested under fire during the Las Vegas massacre and saved hundreds of lives, but it's still only barely tested for something like this -- ARDS patients may need to be on the ventilator for up to a month, if they live.
One tragedy for the ventilator makers -- if they are able to produce hundreds of thousands of their FDA approved modern ventilators, they will save many lives -- but then no hospital, anywhere in the world, will need to buy a new ventilator for years, because there will be such a huge surplus after this is over. Stepping up means a big sales windfall, but then destroying their business, unless some special government intervention comes in. There is much debate if manufacturing at this volume is even possible.
Several interesting projects have sprung up to create designs for cheap ventilators, included from CPAP parts and there's even a contest with a prize. What's even more important than cheap is that they be reliable and that it is possible to make or obtain vast numbers of reliable ones quickly. These range from hacker projects to fully manual ventilators which literally require an attending person to squeeze an air bladder to breathe for the patient. They must do this 24 hours a day, and will be exposing themselves to the virus when they do so. If they let up, if they make a serious mistake, their patient might die. It's a last resort.
In any event, the cost is very worth it, so we don't need the machines to be cheap, but they must be simple enough to be made fast, from off the shelf or easily mass-manufactured in factories which can be brought up to full production without putting their workers at major risk. Not easy.
Smart CPAP hardware with new firmware = Ventilator?
One option is being explored by several people. It is not yet verified to be practical, but if it is, it has many attractive elements to it. Around 20 million people in the USA have a condition known as obstructive sleep apnea, which means their airway can close up when they sleep, temporarily shutting off air and waking them up. Millions of them treat this with what is called a CPAP machine -- which blows air into their nose via a mask at higher pressure, and keeps the airway inflated. Over the years, these machines have become much more sophisticated, and several million of them are made each year.
A CPAP is not a ventilator, but can the hardware in modern CPAPs help create a ventilator?
The current machines can blow air with a computer controlled blower that can change the level of pressure instantly. They have sensors to detect the air pressure, sound and air flow. The older machines did not do this, they just stuck one one basic pressure, but the newer ones tend to have this ability.
While CPAP machines are not ventilators, and don't have software to deliver assist control or pressure control ventilation, they may contain much of the hardware needed for certain types of ventilation. What they don't have is the software.
Normal CPAP machines deliver a max pressure of 20cm of water. (That's not much, about 2% of atmospheric pressure.) Many ventilation patients use much less than that, but serious ARDS patients need that and more. Ventilators usually are rated to deliver over 40cm, though they don't normally go to that level, as it's dangerous.
Researchers have discovered, though, that the blowers used in some CPAP models are capable of 45cm and up to 400 l/min of airflow, much more than is needed for CPAP, and enough for ventilating many ARDS patients. Less established is whether these blowers and their electronics might fail if used 24 hours a day at these levels. My intuition is that most medical devices are built with wide tolerances and this will not be the case, but it must be confirmed. If 24 hour operation is not practical, there is a solution -- patients can have two or more machines, and have their machine switched every 8-10 hours, which is the normal duty cycle of CPAP machines. Because there are so many millions of CPAP machines out there, this remains an option.
Some patients can be helped with the CPAP or BiPAP therapy these machines deliver with their current software, and that's good. However, under discussion is not using CPAP/BiPAP to treat patients, but rather converting such machines, by replacing their internal software, to apply ventilation.
In fact, if two machines are available, they can both be hooked up to the patient's air tube, through a valve that allows either one (or both) to blow air but which prevents blowback into the unit. This allows automatic handoff between two machines, and also means that if failure is detected in one machine, it can be turned off and the other activated to take up the slack, also making an alert to replace the bad machine in reasonable time. This could result in an extremely high level of reliability.
In particular Covid ARDS patients are being ventilated at a fairly high pressure. In many cases, the minimum pressure (PEEP) is set at 16cm, which is high for a CPAP. The upper pressure limit varies -- it is not set, but is whatever it takes, up to a limit, to assure sufficient air flow through the lungs.
To act as ventilators, the machines would be converted by loading new software in their controllers. Fortunately, most of them can be field upgraded with new firmware. Many of the leading CPAP vendors, such as Medtronic/Puritan Bennett and Phillips Respironics, also make ventilators and are already familiar with the algorithms and software needed.
Some of these machines feature return tubes which can be used to measure pressure at the interface instead of in the machine. This is superior, but was largely eliminated from newer machines as no longer necessary. It is not known if this is needed.
Monitoring and control
Most of these machines also have a USB port, which can be used to set up the machine and to upgrade its firmware. New ventilator firmware could also use this port to communicate with a control and monitoring computer, which could either be based on a standard mobile phone or an old laptop running a free operating system such as Linux from a USB stick. This monitoring computer would not be life-critical. It would be used to:
- Provide a GUI to set up the ventilator and all its parameters
- Receive data from the ventilator to display what's going on, including detecting any problems. Display a graph of pressure, volume and flow over time.
- Sending this data and any alerts to central nursing stations (more than one in case of failure) and anybody else that needs to know. In addition, they would send, and relay from the ventilator a "heartbeat" regular signal so that the failure of either device would be noticed at the nursing stations, causing somebody to be dispatched.
While the monitors would communicate status over Ethernet or WiFi, they probably would not receive commands back that way, or if so, only over a non-internet verified channel, to avoid risks of computer intrusion changing parameters of the machines. Of course that means problems require dispatching of a qualified staffer which has its own costs.
(Some modern CPAPs have a cellular radio in them, or wifi. This may also be usable.)
I have prepared a set of notes on functions of the monitoring system.
ARDS patients tend to require fairly high pressures. They are often run with a minimum (PEEP) pressure of 14 to 16cm. This holds the alveoli in the lungs inflated open. ARDS patients prefer low tidal volumes which means a very high concentration of oxygen -- as much as 95% to start -- is desired. Most ventilation is done in Assist Control mode where you set a flow volume (ie. ml per breath, with a minimum breath rate) and the machine tries to deliver enough pressure to deliver that volume. If it can't, it sounds an alarm. It also alerts if the volume is not delivered and high plateau pressures suggest the lung is not accepting the air well.
ARDS patients are often intubated (tube down the throat into the trachea) and heavily sedated so they will tolerate that. In may cases so sedated they don't initiate breathing on their own all the time. Fluids can also build up which must be detected and suctioned out.
What a ventilator does
At a very basic level, a typical AC mode ventilator blows pressurized air (which is usually 50% or more oxygen in an ARDS patient) into the lungs to inflate them, then reduces pressure to let them deflate out an exhaust tube or port. It will be set with a base pressure (high for ARDS patients,) a tidal volume (how much air to pump in per breath, for example 500ml,) an oxygen concentration, a minimum breathing rate and various alert thresholds.
It starts hoping the patient begins to draw breath (which drops the pressure in the system, detected by a pressure sensor.) It then increases the pressure over the base to deliver the volume of air. If the volume isn't going, it increases the pressure up to a limit. Then, it reduces pressure down to the base level and lets the air come out. If the patient doesn't draw breath again within the minimum breathing rate, it will increase pressure on its own to cause a breath anyway. It uses various algorithms to look at the pressure and flow curves to figure out what's going on, or if anything is going wrong.
This mode is the most common. There are other modes which aim for a certain pressure rather than a certain volume of air (and have alarms if the volume is too low) and several others.
This more detailed engineer's summary of ventilation gives you lots more detail.
It is also necessary to provide most patients with supplemental oxygen, often quite a lot of it. It is unknown if it would be safe to feed high concentration oxygen through the blower. Normally with CPAP lower levels of supplemental oxygen are added at the mask/interface. This problem needs resolving. CPAP hardware itself does not have anything to control mixing of oxygen and regular air.
Disturbingly, some patients are needing almost pure oxygen. Pure O2 is toxic long term and it must eventually drop to 50% or less, but this is very high, and if fed through the machine, we must be sure there is no fire risk, even if things overheat or humidifiers dry out.
Many CPAP machines do not have an air inlet port, rather they take in air through a filter and it is not easy to attach anything to the inlet. Some do have an inlet port. They all have a standard outlet port, however.
Full ventilators have internal air-oxygen blending and even oxygen concentrators. External air:oxygen blenders exist, but may be difficult to get in the quantities needed.
Fortunately, you can add O2 at the tube going into the patient. This is done with CPAP, and research shows that even at 20cm of CPAP, you can accomplish 50% concentration of Oxygen, so a similar level should be possible with ventilation in the 15-25cm range, one hopes. This means the Covid patients may not be able to start this way but can be weaned down to it. This will involve venting extra oxygen to the room, so the room should be sufficiently ventilated so as to not increase its own O2 levels too much.
Patient interface would not be a typical CPAP mask for most patients. In fact, for ARDS patients an intubation into the trachea will be necessary with sedation. That tube will need valves to control flow and to assure expired air is filtered for virus particles.
In operating rooms, ventilators typically run on a closed circuit using chemicals to scrub the CO2 and return the flow back to the patient.
Most of these CPAP machines, when operating in their normal pressures, are low power. They run at 12 to 24 volts and only draw an average of about 12 watts. Their humidifiers (which is important for ventilation) draw much more power but sometimes from a separate circuit. The basic machine, however, can run off a typical marine/RV battery for around 60 hours, less at 25cm. And for 30 hours from a typical car battery. They can be hooked up to such batteries and the battery can also be hooked up to a low-current slow charger so the machine runs off external power but can run for multiple days -- and most of a day with humidification -- if the power goes out.
Alternately, these machines can be plugged into standard "UPS" units sold for computers. They might only run for an hour on such a unit but it should be enough to find other solutions if preparations are made, such as generators.
Generally, we want to keep this simple, and the ventilator must be. However, the control computer which is pulling data from the ventilator and displaying and relaying it could have some more innovation. Today, monitoring a patient and adjusting their ventilation is still considered partly an art among doctors, being able to read the pressure and flow graphs and understand what's going on, predict risks and adjust settings. That's actually the sort of thing neural network AIs can get good at if given data from the actions of skilled doctors and respiratory therapists. And that data could be gathered from patients being treated through this new interface.
This could be of value because if there are hundreds of thousands ventilated, there are not enough doctors to monitor and fine tune their treatment. While we would not quite be ready for the AI to do that, it could make recommendations, and signal alerts about potential problems coming down the line, and doctors could make decisions. In particular, there are a number of interesting stories about how such networks have learned to spot problems even before the skilled doctors do -- you give them the data from lots of patients and what problems they had, and they can uncover clues not seen before.
Loaner machines must be fully sterilized after donation and before return, and between any two patients. Hopefully this can be done in bulk, buy putting lots of machines in a room running at low speed, and flooding the room with ozone or other killer of pathogens.
I believe that people would stup up to loan machines and donate old computers. A tax deduction could also facilitate this.
While questions remain unasked, the value of this is strong. These machines already exist and do not need to be manufactured. In addition to what sits in inventory, millions of people have new machines and also have old machines sitting around, in some cases several of them. They can happily use those machines and loan out their newer model. Large numbers of patients are prescribed CPAP and get one bought by insurance, but decide they don't like it and turn it off. It would not be surprising if the number of such available machines ready for donation could be very large. The hardware, while not designed for ventilation, is medical grade and it exists. The control stations can be the hundreds of millions of old laptop computers or phones also sitting on shelves, a large fraction of which can become a dedicated station by booting from a flash drive. With batteries, these devices also can handle power failures.
Still needed are the hoses, interfaces and filters, plus possible additional solutions for oxygen and humidification. But a large part of the problem may already be solved.
The current vendors, if they decide they can do this, will be afraid. After all, the hard truth is these are not ventilators. Some will fail. Patients will die. While they are probably patients who would have died anyway, this is still a risk companies are not willing to take. They probably won't take it unless they are granted a waiver from such liability. Even so, that's why the monitoring system is important, so that it is expected these machines will fail and it will be detected immediately and fixed immediately. With enough machines, a backup machine can be sitting nearby.
The other question is how they are allocated. In one scheme, patients can be triaged to use higher end FDA approved ventilators. The CPAP based devices would only be used on patients who would otherwise be told, "we don't have a ventilator for you." In Italy, sadly, a number of people have died when this was the case. The cheaper devices can generally only improve things.
The FDA has loosened regulations even declaring that modified BiPAP and CPAP can be used to maintain airways, but this only helps some patients.
On the other hand, if we decided that the CPAP based machines can do a job on a body of patients who have lighter needs -- they may need less pressure or less oxygen, for example, a rational approach might be to assign all such patients the lesser machine and reserve the more expensive official machines for those who must have them. This can be a good result but it might mean that a patient who would have passed the triage for other reasons now might face the failure of a lower end machine. It's a terrible choice, but still saves the most patients.
ARDS-ICU beds are still needed for these patients. It is unlikely vetilation could be done safely at home. It could be adapted to temporary ARDS ICUs currently being constructed in unused buildings like convention centers.
Upate: The UK has released a specification for simpler ventilators for this emergency. Most of these requirements can probably be met, though oxygen safety is unknown, and PEEP over 15cm will need confirmation.
To be researched
The Cuirass ventilator, a descendant of the iron lung, is a chamber that goes around the chest. You suck air out of it, creating a pressure differential that sucks air into the lung. This is bulky and rarely used, but has the tremendous advantage of not needing intubation or sedation. Because of the high minimum pressure ARDS patients need, it is possible it might not be enough on its own, but could be combined with an adapted CPAP which would provide the base pressure while the Cuirass system uses negative pressure to increase the differential.
- Can 95% oxygen, or even 50%, be safely inserted into the air flow?
- Can the output vent be safely filtered to avoid spreading viruses?
- Can CPAP blowers and electronics operate at 20-30cm pressure on a 24 hour basis for days and weeks at a time with very low failure?
- Can CPAP vendors be convinced to release new firmware, and when?
- Can independent parties write new firmware and reflash these units?
- Can enough air:oxygen blenders and patient interfaces and more be supplied?
- Do we need additional sensors (extra pressure or flow sensors, O2 sensors) which can be hooked into the laptop via USB?
- Can we pull off the automatic switch from one machine to another, including switching oxygen? Switching machines is perilous when done by hand.