- Detailed Info
This winter I noticed a magazine advertisement for an avalanche airbag pack that claimed “A 97 percent success rate in real world conditions.” What the advertisement didn’t mention was that people caught WITHOUT an avalanche airbag have an 80 - 90 percent success rate. In other words, most people caught in an avalanche will get a cheap lesson; they will either escape off the slab, grab a tree, dig into the bed surface, ride on top of the debris, it will be a small avalanche that wouldn’t burry them anyway, they could be saved by a beacon recovery or they could just get lucky. Most people caught in an avalanche will survive, which is very good news for all of us.
It’s like the example my college statistics professor presented in which Sanka advertised their decaf coffee as “97 percent caffeine free.” What they didn’t tell you was that regular coffee is 90 percent caffeine free. This, no doubt, sells much more coffee than saying that their decaf coffee has one-third the caffeine as regular coffee, which is much less misleading.
There seems to be no end of confusion about the effectiveness of avalanche airbags. I have heard numbers bantered about ranging from the aforementioned 97 percent all the way down to, “Avalanche air bags would save only 3 out of 100 who would have otherwise have died”, which was presented by a prominent avalanche professional in a national class. That’s a lot of confusion.
So…here I will try to explain where these various numbers come from and provide what I think is a more useful—and less misleading—sound bite when we talk about the effectiveness of avalanche airbags.
Most of the data come from Europe because avalanche airbags were developed there and over the past 15 years, they have become quite ubiquitous. There have been close to 400 cases so far of people getting caught with avalanche airbags (the exact amount is uncertain because many non-fatal cases likely go unreported).
The 97 percent success rate number comes from a database maintained by the Swiss Federal Institute of Snow and Avalanche Research (SLF). In collaboration with ABS brand avalanche airbags, they kept track of all the people that reported being caught in an avalanche with an avalanche airbag, whether they successfully deployed it or not until the fall of 2010. After the fall of 2010, the responsibility of this data collection has been passed on to the various national avalanche centers. (So if you get caught with a balloon pack, we would love to hear from you or send the report directly to http://www.avalanche.ca/caa/industry-services/balloon-pack-study.)
According to Brugger’s 2007 study of this dataset, the percentage of people caught who died in an avalanche decreased from 19% to 3% for those who successfully deployed an avalanche airbag. In other words, there is an 81% “success rate” for those without a deployed airbag and a 97% “success rate” for those that did.
Thus, many people latched onto the 97% figure. But we have to remember that this is 97% of those CAUGHT. Since most of people caught survive anyway, the number most of us are probably more interested in is what percentage would survive who would have otherwise been KILLED. (In statistics-speak this is the difference between “absolute reduction in mortality” and “relative reduction in mortality.” Back to the Sanka coffee example this is the difference between being “97% caffeine free” and “one-third the caffeine of regular coffee.”)
At the International Snow Science Workshop last fall, the well-respected Canadian researcher, Pascal Haegeli presented some preliminary results of his more up-to-date study on the effectiveness of avalanche airbag packs. He performed the study like standard medical research in which he compared the mortality rate of the treatment group with the control group (people who wore airbags vs. those who did not). As I mentioned before, there is a wide variety in the severity of the avalanche that can catch people. (Many people escape off the slab or it’s a small avalanche, etc.) So he only included the cases where people were “seriously involved”, meaning that the avalanche airbag “had a chance to be effective.” In addition, he only included cases with multiple victims in which—in the same avalanche—some wore airbags and some did not, and he also included cases where people failed to deploy the airbag. In other words, he wanted to fairly compare the success of the technology, warts and all. He still needs to add more data from Europe and publish it in a peer-reviewed journal, so I can’t present the numbers here, but his PRELIMINARY results suggest the following:
If you look at it with a glass-half-full approach, a deployed airbag saved about half of those who would have otherwise died. If you look it with a glass-half-empty approach, you would say that half of the people who deployed airbags died anyway.
For various reasons, not all people who wore airbags were able to deploy them. So if you include these cases, WEARING an avalanche airbag would have saved about a third of those who would otherwise have died.
Pascal was careful to mention that his study presented perhaps a worst-case scenario because, he eliminated less serious avalanches from the analysis and included only multiple burial incidents, and thus the data was biased towards larger, less survivable avalanches.
Indeed, in the April 2012 issue of the Avalanche Review, Jonathan Shefftz did a great summary of five different published data sets mostly from older European data sets and he found roughly similar numbers. Wearing an avalanche airbag would have saved from 35 to 81 people out of 100 who would have otherwise died. (The average of the 5 studies is 64.) So, it seems that in real-world experience, wearing an avalanche airbag will possibly save a little more than half of those who would have otherwise died.
At least from my perspective, saving half of avalanche fatalities is pretty darn good. Avalanche airbags are the best technology we have seen come along including the beacon. Although it's impossible to directly compare beacons with avalanche airbags because it's an apples-and-oranges comparison, most experts agree that the avalanche airbag will likely save more lives.
My pet peeve with this issue is that people who argue about the numbers often leave the most important part out of the discussion--terrain. If you get caught in un-survivable terrain then, guess what, you won’t survive no matter what kind of rescue gear you use. There have been a number of prominent accidents in which the victim with a deployed airbag died because he was either strained through thick trees and rocks, deposited in a terrain trap, buried deeply or went over a cliff. In zero-tolerance-for-error terrain, airbags don’t work, beacons don’t work, Avalungs don’t work. Nothing works. Save your money, buy a life insurance policy and a beacon or RECCO so rescuers don’t have to spend all night probing.
So at least for me, unless I’m 99.9 percent certain that the slope won’t slide, then I don’t go to un-survivable terrain. If I’m going to spend the money and carry the extra weight of an avalanche airbag pack, I want to ride in terrain where it has a chance to make a difference. In other words, choose terrain with no obstacles, no terrain traps or sharp transitions and avoid large avalanche paths.
The other part of the discussion is the often-overlooked issue of what we call “risk homeostasis.” Each gizmo we buy to increase our safety usually cause us to increase our level of risk at the same time. For instance, when we added seat belts and airbags to cars, yes fatalities decreased, but it also allowed us to drive faster, farther, crazier and talk on our mobile phones at the same time. So safety measures usually work but not nearly as well as we would hope because people just increase their risk (and “utility”) at the same time. In avalanche airbag case, we will also get more powder, more fun, and more risk in the bargain.
Ignore the 97% number and the 3% number. My best guess is that avalanche airbag packs will probably save a little more than half of those who would have otherwise have died in an avalanche. They will never save all of them because 1 out of 4 will likely die from trauma of hitting trees and rocks on the way down and an additional 1 out of 4 will probably end up in a terrain trap (deep burial), buried by a secondary avalanche or caught in an avalanche that does not travel far enough for the inverse segregation process to work (larger objects rise to the surface).
In addition, people will increase their exposure to risk because of the perception of increased safety, which will cancel out some, but not all, of the effectiveness of avalanche airbags.
As usual, our choice of terrain is far more important than rescue gear. Un-survivable terrain will always be un-survivable. In terrain with few obstacles, terrain traps, sharp transitions and smaller paths, avalanche airbags have the potential to save significantly more than half of those who would have otherwise died. And that sounds pretty good to me.
I would love to hear your comments: http://utahavalanchecenter.org/email-bruce-tremper
Various Comments Emailed:
Disclaimer: The following statements reflect my own personal beliefs and
research into ABS packs and not that of my employer or ABS Systems.
I am not going to be whiny here, but feel compelled to state that "best
guess" numbers aren't always the best way to convey fact. I have looked
extensively into ABS packs because it is one of the products I clinic shops
on in my line of work. I get these questions all day long when speaking with
shop employees, and need to say that you seem to be downplaying the benefits
of an ABS pack in an avalanche situation.
When ABS did the study listed on their website (which the 97% number in the
ad came from), they looked at multiple slides where both ABS users and
non-users were impacted in a significant manner (aka, if you were on the
apron and ran away from the slide path without incident, you weren't counted
because an ABS pack or not, you weren't going to be buried.)
The bottom line is that in the 262 activations that occurred during the ABS
testing cycle, 3% died, 97% were visible on the surface, and 84% were
uninjured and able to self rescue.
In the SAME AVALANCHES, 67 folks were involved that were not wearing ABS
packs. 25% of those folks died (so 75% survival, NOT 81%) and only 57% of
those folks were visible on the surface.
I understand that you want to keep folks from feeling like superman, but to
talk down on a technology that has been proven to save lives is a long reach.
I have found you guys to be extremely credible in the past, but feel this
type of spin on the topic makes me wonder if your end goal is to save lives,
or something else.
Bruce's Note: As we can see, doing statistics on avalanche airbags is tricky business because it all depends on the design of the study and how you filter the data. Pascal chose to only include those who were "seriously involved" and exclude those in which the avalanche airbag probably did not make a difference in the outcome. And as such it skews the data towards less survivable avalanches. In addition, he included the modern Canadian and U.S. data in which more people tend to die from trauma than in Europe. So it represent perhaps the worst case scenario, as he stated. If you look at 5 other data sets of mostly European data, the numbers range from 35 to 81 (average of 64) people out of 100 who would be saved by an avalanche airbag, which is in the same ballpark as Pascal's study (see the April 2012 issue of the Avalanche Review).
It seems to me that if we all used risk reduction measures taught in avalanche classes, such as proper choice of terrain, low-risk travel ritual (one at a time, etc.) there is no reason why avalanche airbags (along with beacons) can't save the lives of most of the people who would have otherwise died. But we are all human, after all, and we often don't follow the rules or make the best decisions. Plus risk homeostasis is usually responsible for more modest gains in safety than we expect.
Great analysis and article. I've read those studies and have come to similar
conclusions. I'd like to add some thoughts on the subject that helped me
arrive at my views towards airbags. Namely how terrain type, avalanche hazard
and personality/experience interact with airbag function.
Low/mod hazard and/or simple and low consequence terrain: Either cautious
novices use this terrain or experts following safe terrain choices in high
hazard. Airbags not needed due to low likelihood of avalanche and easy rescue
probability, are likely not effective as the slide is too small to get into
the fluid particle physics needed to float the bag.
Considerable hazard and/or challenging and moderate consequence (survivable)
terrain. This is the vast majority of avalanche hazard and good ski terrain.
Folks who have advanced skills can safely ski here making good use of
terrain. Slides are big enough to make airbags effective (fluid particle
physics), and the terrain is survivable. If people are eating risk (which I
caution against) the bag might just same them here, likewise if someone doing
all the right things just makes a mistake.
High hazard and/or high consequence terrain. This is either the foolish
novice being where they have no right to be, or an expert who has eaten
hazard risk on challenging terrain. Either way a bag is not going to help on
the high consequence terrain, and if you're using it as a get out of jail
free card on the odds are going to catch you sooner or later.
My conclusion, air bags are most effective and needed at the junction of
considerable hazard and challenging terrain, which is where most of us ski
most of the time.
As always thanks for cutting through the BS to try to get down to what
matters- coming home safe.
I thought I would relate my experience in deploying my airbag this december
in a skier triggered slide in West facing Scotties (this avalanche accident
was reported to UAC by myself and Lindy on December 26, 2012)
I had a decent size slab, about 2.5 feet deep by 200 feet wide break off mid
slope, roughly 38 degrees. UAC incorrectly recorded the slide as 200'
vertical, but it was 200' in width. I immediately deployed the airbag as I
saw the slab fracture and begin to slide. The crown line was just a few feet
above me when the slab broke. The width of the fracture meant that I could
not attempt to ski off of the slab, I was instantly sliding downhill with
large blocks of snow all around me. Because I was at the very top of the
slab, my instinct was to self-arrest, and let the debris slide downhill
beneath me. The airbag seemed to help keep me above the moving debris, and
to keep me upright. I dug my edges will all my life, sliding faster and
faster down the mountain as the slab accelerated rapidly. I managed to slide
slower than the avalanche, and after five seconds or so, the majority of the
moving snow accelerated away from me. I was able to stop completely, and
watch as the slide exploded into a stand of aspens a few hundred feet down
I am certain that without the buyancy of the airbag, I would have been fully
englufed by the avalanche, and would have certainly suffered severe injury
due to trauma hitting the trees downslope. As it was, my bindings held, and
all I suffered was a sore knee and a good dose of ptsd. In hindsight, I
think that a whippet would have really helped my efforts to self-arrest on
the bed surface. I was able to self-arrest by the narrowest of margins,
clawing and edging like my life depended on it. A whippet or axe would have
made a huge difference. But it was definately the airbag that gave me a
fighting chance to stay upright, and allow the heavier avy debris to flow out
Please that others may learn from this close call of mine. FWIW, I would
rather have an airbag than a beacon at this point in my backcountry career.
The airbag can really change the trajectory of an accident, whereas a beacon
only helps to pick up the pieces once the shit hits the fan.
Thanks again to UAC for all you do.
I really appreciate your breakdown on airbag effectiveness. This is spot on
and you are right about people getting a superman complex with this new gear.
I believe airbags are a great product - but nothing is better than common
sense and continuing to be a safe and smart in the backcountry. Thanks!
A wise avalanche forecaster once told me a story of silver bullets (you may
know the guy...)
the airbag is the newest one. people that cling to silver bullets also cling
to a 5 color danger rating system, and easy answers in general. they want to
ski that slope and by god unless something smacks them in the head and says
dont do it, theyre going to probably try it.
just like unsurvivable terrain, there are unsavable people. i do not envy
your position of trying to save everyone, because from the POV of a 27 year
old guy in the backcountry, i realize that there are a LOT of people out
there that are alive simply because they have yet to come up short in russian
roulette, though they play it all the time.
my point here is im glad youre making the points people need to hear- airbags
do not save 97% of people caught in avalanches. thats a ludicrously
misleading statement. people (especially in the wasatach) need to be scared
straight a little bit. saturday i was hiking patsy marley hoping to find that
magic 30 minutes of corn before the snowpack turned into a swamp but knowing
that i was going to stay on low angle stuff regardless, and i found a
snowboarder by himself who said "i only go out alone on low risk days....
like today". please continue to highlight the ugly, the scary, the bleek and
above all, the realistic. (during the "signage surveys" earlier this year i
even suggested highlighting the total body count at places like 9990 with a
big scary sign)
having said all that, i very much like having my airbag because i dont think
im infallible in my decision making, and id like to do anything i can to pad
the odds in the event that i screw up. Thanks.
Thanks Bruce for all that you do. This is good information in its simplified form. I always enjoy reading what you write.
I thought this article did a great job summing up the "true" effectiveness of
rescue gear, such as airbags. I find that people tend to buy as many
avalanche rescue accessories as possible after taking an avalanche course and
deciding to spend more time in avalanche terrain. I think you describe the
increased risk factor beautifully. Another aspect that I am aware of is
improper use of rescue gear. Many individuals use "Avalung" devices, but the
air hose seems to annoy them, so they keep the hose stowed away in a zipper
pocket while touring. What good does that do? With the same equipment, even
if the hose is accessible, once a slid rips, it may be difficult for an
individual to actual put the hose in their mouth.
Back to avalanche airbags, I feel many people watch videos of huge avalanches
involving pro riders who claim the avalanche air bag saved their life. When I
analyze these situations I notice several huge factors that influence the
rescue. (One particular video I am thinking of involves Xavier De La Rue).
When some of these huge slides happen, involving pros, they have a crew of
people sitting in a helicopter with eyes on the situation at all times. This
allows communication from the point the crown releases "head left head left,
slide slide etc" and allows rescuers to keep an eye on the victim from a very
safe spot. Even when touring with my friends, we typically have set up spots
that allow us to watch each other ski from safety zones, but we are typically
managing terrain wisely and avoiding "high risk" terrain, as you mentioned.
But in many cases if a slide does release, especailly a large slide, it will
pull the victim down-slope and possibly out of eyesight. I think the point I
am getting it is skiing media is changing the perception of avalanche rescue
gear (the people operating the rescue gear are not shown) and the emphasis on
terrain management is severely downgraded. Another example is that video of
the guy at the freeskiing championships (I think) who does a massive
back-flip while outrunning the avalanche. Had he not stomped the landing, he
would have been buried, but fortunately with tons of rescuers available for
As a young person (24) who has been backcountry skiing for 6 years now, I
have completed Avalanche 1 and 2 courses as well as a few classes in snow
science at my former undergraduate institution, I notice a few trends in
1) Many people complete avalanche 1, buy the gear and think they are ready to
ski big lines and epic steeps right after a storm cycle (I live in Colorado,
so that is 99.9% of the time a bad idea).
2) It takes an accident and a moment of panic before people are willing to
better learn how to use rescue gear and make more cautious choices in regards
to terrain and snow management.
3)Conditions begin to become "normal" to people, leading to poor decision
making. Confidence comes from skiing low angle terrain, which suddenly makes
steeper terrain feel safe, with out analyzing terrain. I recently lost a
friend in a slide near Silverton, rescue gear such as an avalanche airbag
would have done very little other than push the individual into many trees
and confidence from low angle terrain was a big error in the decision making
4) "Skin highways" and the focus of getting to the top of a line distract
people from analyzing snow conditions in places that are safe to do so.
People read the online forecast and think they have enough information,
especially on "moderate" days. While backcountry skiing should be a
continuous collection of information related to weather, snow and terrain,
people seem to assume they are safe going up and they can analyze conditions
once they are on top of their line (a seemingly large misconception of skiing
is dangerous, hiking is not seems to exist). When I tour, I never see any
tracks venturing from the skin track to see how the snow feels on no
consequence rolls, or even just to take a feel on flats. I see skin tracks
that travel directly beneath very large slide paths with people who do not
even hesitate about where they are traveling on them.
My point is, you are right, but how can we encourage people that money can
not buy safety?
Bruce: I enjoyed your blog article on airbags, and you raise some good
points. However, it seems the data you present indicate a much higher
effectiveness for airbags in preventing fatalities. Looking at the first set
of "bar" exhibits the post, in avalanches without airbags, 19% of the people
die. With airbags, 3% die. As you note, the difference (16%) is "Saved by
airbag." Thus, 84% (16%/19%) of people who would otherwise die are saved by
an airbag. Obviously, the samples differ in the two bars and this is not a
randomized trial. But, unless the avalanches in the "airbag" sample were
systematically less dangerous than those in the "no airbag" sample, then the
84% improvement in survival should be due to the treatment (the airbag) and
not differences in the sample. And, if the hypothesis of risk homeostasis is
correct, it is likely that the avalanches in the "airbag" are more (not less)
dangerous, so the true effect of the airbag would be greater than 84%.
I will admit that this increase in survival seems too high, since airbags do
nothing to protect from trauma, and many avy fatalities are due to trauma (I
do not have the figure handy, but I have seen data on this and I seem to
recall it is not a small percentage).
Yes, you're correct. That study, based on data collected by ABS brand airbags, showed that airbags saved 81 people out of 100 would otherwise have died. (For simplicity, I rounded the numbers in the chart so you came up with a slightly different number.) But that data set included all incidents involving an avalanche airbag including many people who would probably have survived anyway (skied off the slab, small avalanche that would not have buried someone, self-arrest, etc). These incidents were filtered out for Pascal's study because he wanted to only compare incidents where the airbag "had a chance to make a difference".
That's why doing statistical studies like this are tricky because so much depends on the study design, the assumptions made and how the data is filtered. As Pascal said, his data set is biased towards larger, less survivable avalanches, so the conclusions will be more pessimistic than some of the other studies.
After I published the blog, I realized that I should have mentioned the results from other data sets so I added the following paragraph to the blog:
"In the April 2012 issue of the Avalanche Review, Jonathan Shefftz did a great summary of five different published data sets mostly from older European data sets and he found roughly similar numbers. Wearing an avalanche airbag would have saved from 35 to 81 people out of 100 who would have otherwise died. (The average of the 5 studies is 64.) So, it seems that in real-world experience, wearing an avalanche airbag will possibly save a little more than half of those who would have otherwise died."
Another thing to keep in mind is most data comes from Europe, which has a lower incidence of trauma. Here in North America we often ski in or above trees, so our trama rate is about 25% vs about half that for Europe (depending on the study, of course). So survival rates will be lower here.
As for the risk homeostasis issue, that is probably a subject for another blog. According to some studies, a gizmo like an avalanche airbag has less potential for risk homeostasis than others. So I'm cautiously optimistic.
Thanks for a great blog, including the
discussion of how airbag pack effectiveness also relates to terrain choices.
There has been a disturbing trend for some people to point to fatalities
involving airbag pack users, and to claim or imply that the effectiveness of
airbag packs overall is called into question because they are not 100%
effective. Saving roughly half the people who would otherwise have died is a
very impressive thing. As regards terrain choice, I love your clear graphic.
Interestingly, Haegeli did not find an increased risk of trauma from airbag
packs. While I believe there are some isolated instances where any gear can
cause issues, it intuitively makes sense to me that sliding on top of a slide
through trees could on average be no riskier than sliding a potentially
shorter distance at the bottom of a slide, through the same trees. So, it
may well be that terrain choice can impact the effectiveness of both airbag
packs and beacons more or less equally in many cases.
I thought you did a great job sifting through
the numbers Bruce, thanks. I was simply astonished that Xavier de la Rue
ended up on top wearing a snowboard still on his feet in that long gully with
all that weight behind him (equally amazed he didn't cut right at the start
of the whole thing). Same for Elyse Saugsted, amazing that she made it
through the timber at Tunnel Creek without fatal trauma to be found on top.
That was enough to open up my wallet and strap on a few more pounds, even as
a pretty conservative pick-the-right-days-tourer with 25 years of touring
Great blog on avalanche airbag fatality statistics. My partner Will Paden and
I make the WARY AviVest and the AviPack airbag systems, and we cringed last
year when we saw our competitor come out with the 97% survivability
advertisement. We knew it wasn't true when we read it, and Pascal pointed
that out in Anchorage. When we do our testing in British Columbia and Nevada
the results are completely dependent on the terrain. The only airbags we have
ever seen fail in testing were destroyed by terrain. In 50 or so class two
and three avalanche tests the only bags we ever lost were shredded by trees
and possibly in one case by a rock band. We have never had a secondary slide
bury a test dummy but it is a matter of time.
The "war of words" between airbag manufacturers at the various trade shows
about who's system has greater survivability or blunt force trauma protection
is ridiculous. All of these systems slightly increase your chances of
surviving an avalanche, and no more.
Numbers, numbers, numbers. And then confusion. Or interpretations. Let’s
keep in mind that we’re having all these discussions and confusion in the
first place because we are trying to fix something that is broken. What is
broken is the ‘experimental design’ of the various airbag studies.
What we would like to have, to perform a proper airbag study, is a huge
laboratory where we can quietly run hundreds of trials with subjects with and
without airbags. Where we can change one thing at a time: type of avalanche,
size of avalanche, position of the victim on the slab at trigger time, etc.
etc. And where we wouldn’t loose a single life in the process!
Unfortunately this type of laboratory will, for obvious reasons, never be
built. We have seen some efforts to approach this type of ‘pure’
experimental design in Switzerland and the Czech Republic, using dummies,
helicopters and bomb induced slides. Cost will keep the number of these
experiments (and thus data) probably too limited to be conclusive.
So we have to do with second (or even third?) best. When discussing airbag
datasets we should always keep this in mind.
To add to the confusion:
- in the ‘97% study’ the difference in numbers between people caught with
airbag (262) vs no airbag (67) is enormous. This makes comparison between
these groups maybe difficult, or even doubtful. Would we still accept the
same outcome if the dataset of people with and without airbags would be
reversed (so 67 with and 262 without airbags)? How would we then feel about
the same results?
- Pascal Heageli is afraid that his dataset might give a too pessimistic
outcome for the airbags. But his approach is imo biased in favor of airbags,
and he is very honest about it, as he includes only bigger slides ‘were an
airbag can make a difference’. So slides were an airbag is unlikely to make
a difference are left out of the equation and therefore the data are biased
in favor of airbags. (but I understand that bigger slides are obviously more
likely to kill people)
- what should avalanche educators advice the novice that has only, say, $ 500
to spent? Buy an airbag first? (as experts feel airbags save more lives then
transceivers and you don’t have to rely on your partners for rescue) Buy
transceiver, shovel and probe first? (because this is were it all starts)
Stay out of the backcountry? (what he or she will never do anyway)
- by the way I am a believer in airbags! ☺
Big thanks to Bruce Tremper, Jonathan Schefftz and Pascal Haegeli for their
inspiration and their arduous work on the various datasets!!