I see 3 ways of pulling this off, and just doing a little mental gymnastics tells me that 2 of them would be just as accurate or inaccurate for different reasons.

It would be possible to manipulate the government data mathematically to get the information you need, but it would require a lot of assumptions and ball park estimates. For instance, there is data for all the accidents involving death of a cyclist and data for how many were wearing helmets (1 in 6 I believe). You could make a reasonable assumption that all triathletes wear helmets, so take 5/6 of the deaths out. But you can't assume that no one else wears a helmet, so ... Then you would have to make all sorts of other assumptions based on the knowledge of the sport. The margin of error would quickly add up to something that is outside of what you are looking for as useful.

Or you could take a poll, but that is going to create some errors as well. You would need to include enough information that repeats did not occur, and a method for someone entering the data to check for a repeat. Again, the data errors are going to add up pretty quickly, probably to the point where it would soon be worthless again.

The 3rd option, the one that is the most labor intensive but potentially the most accurate, would be to dump the [government] data in small chunks to this community and let the community sort out the data into the correct categories. For instances, the data for Long Island is going to show, if my memory serves me, 7 people killed by a vehicle in the last 24 months. I know, as do several other people on this board that live here, that none of them were Triathletes. In fact none of them were cycling for exercise. The same would hold true for just about any region in the US - everyone of them is represented by someone on this board. The problem with this approach is the government is slow in releasing data, so it would be many months behind. There is probably a better clearing house somewhere on the internet with up to date data, and someone here probably knows where it is. As long as it is recent enough, with enough detail for a local to recognize the event, your data should be pretty damn close to what you want, easily within the margin of error.

---

"...the street finds its own uses for things"

I think if a name is in the public domain. Providing that list would be fine

I think the point would be that detailing all the factors surrounding it should be annonymised and aggregated.

No one would want to see a conclusion that uncle bob headed out at dusk dressed in black riding in to on coming traffic. Whether he did something stupid or whether he, as a result of the data, is demonstrated to have increased his personal risk. No family is going to want to read it. I suspect that the info could also be used in legal cases against their families interests if it was in the public domain and there were witnesses who provide responses that at the conclusion of the analysis show the rider increased their risk.

You may want to reach out to the 'League of American Bicyclists' as they have done this in 2014 (based on data collected during 2012-2013), and found that the specific data collected and reported by the government was quite limited. Their own data collection included: time of day; land use; road type; collision type; driver data; alcohol use; source to links of information; legal status; charges. They also wrote bios of the deceased as a memorial.

They came to some interesting conclusions.

http://bikeleague.org/...ery-bicyclist-counts

But the events you are more specifically interested in, which would include all triathlete deaths, would be a much smaller number than the ~700 or so nationwide deaths/year in the US. I've read through the FARS data and NHTSA stuff (and the entire thread, FTR) and as you have noted, the data is somewhat limited. Collecting data as you are suggesting is somewhat different than the above noted 'Every Bicyclist Counts' initiative as they got the information from the internet-newpaper/TV reports and blogs. You are suggesting more direct contact and I would urge caution in dealing with grieving families.

The data that the 'Every Bicyclist Counts' initiative found was largely consistent with FARS. They suggested that national performance measures for safety need to be set and met. The most interesting information that I took away from their data collection was that the highest percentage of fatal crashes were 'hit from behind' and that high-speed urban and suburban streets with no provisions for bicyclists are over-represented locations.

They also noted that when there was an additional factor reported for the driver-3 things stood out. 42% were operating their vehicle in a careless manner, 36% were reported to have committed 'hit and runs' and 12% were under the influence of alcohol/drugs. When an additional factor was reported for the cyclist-23% were riding the wrong way, 17% failed to yield the right of way and 9% were riding on the sidewalk.

They also noted a surprising lack of information, lack of outrage over these deaths and lack of action.

But using the numbers that are known, about 700 cycling deaths/year=a risk of 2.2 deaths per 1 million people. That number has not changed over the past 20 years. The circumstances certainly may have changed but the total numbers haven't changed much.

There does appear to be a lack of uniformity of data collection and a paucity of overall data but the specific death numbers are known. The numbers do not give us reasonable quantification of time/hours/miles cycling either.

I remain interested to see what ST can do to help.

"I see 3 ways of pulling this off"

or the 4th option? the one i suggested? which seems to me the simplest most logical, straightforward way of doing it. maybe you could tackle any of your 3 options, i'll tackle mine, and we'll compare results and see if they agree. that might be powerful.

---

Dan Empfield
aka Slowman

"But using the numbers that are known, about 700 cycling deaths/year=a risk of 2.2 deaths per 1 million people."

i would be willing to be the incidences of death by wingsuit are even less. this is why the denominator is important. it's not death per the population. it's death per the user population. if there are 700 cycling deaths per year, there are probably only a dozen triathlon cycling deaths per year. or fewer. this is why i think our own cohort risk should be pretty easy to estimate.

but we don't know unless we investigate.

---

Dan Empfield
aka Slowman

So if USATriathlon lists 500,000 members, and there are only a dozen triathlete deaths per year, then the odds are around 40,000-to-1 that a given member will die in a cycling incident.

Is that not a sufficient estimate? I don't know that you'd be able to change the order of magnitude of that estimate, other than in the "closer to zero" direction. For instance, show that half the deaths occurred in urban zones: you change it from 1/40,000 to 1/20,000. Does that matter?

(it's like buying two lottery tickets instead of one. While your odds of winning just doubled...)

---

----------------------------------
"Go yell at an M&M"

"So if USATriathlon lists 500,000 members,"

it's got more like half that or fewer, but when you add in triathletes who just didn't buy a USAT membership or one-day (triathletes who are still training but are on a racing sabattical) i think it's more like 450,000 in the U.S.

"then the odds are around 40,000-to-1 that a given member will die in a cycling incident."

in a year. but if i race and train for 40 years then it's 1 in 1000, yes? unless i do something to make my odds better than the mean.

but i think we need to count the number of deaths, just to be sure we have a hard number and not a guesstimate.

i just got of the phone with a USAT vice president, and they just had a meeting about this. i don't think it's the wars, but the rumors of wars, that present an optics issue. we need to either solve or validate the optics by telling ourselves the truth, that this is a dangerous sport, or that it appears to be a dangerous sport but, really, isn't. then we need to figure out how to make the sport even less dangerous through best practices.

---

Dan Empfield
aka Slowman

Dan, that's now how the stats work. Your probability is 1 in 40,000 per year every year. So is mine. You could go out once this year for 15 minutes, 365 times this year for 800 hours and your odds of death on any ride remain the same. It's not cumulative and year over year, it's not cumulative...it is constant every year (whatever the global number is)....in Poisson distributions the arrival rate of the next occurrence of the event is not correlated with the most recent one and your odds don't increase the more you are out there. This is the type of statistic that's "wrong place wrong time". Same deal with a plane crash.....whether you flew 100,000 miles this year or 1 when 2 people step on the next plane, their personal odds are identical to the global odds....every passenger has those odds every moment they are flying, but they don't change the more you fly.

In the age of the internet, everything SEEMS dangerous. Back in the day, you heard of every plane crash on the 3 networks. People feared flying because they heard about every plane crash in the world. Meanwhile, they never thought twice about driving their car.

In that same time in the 70's and 80's we'd just get out and ride our bikes. Once in a while someone we don't know of in Holland, or South Africa, or Nova Scotia, or Japan would get hit by a car and die. We never heard about those guys. We only heard of the very few in our immediate community that got hurt/died. It was like car accidents....you just heard about the local ones, so you never thought it was unsafe because it was exceedingly few.

But now due to the internet, bike crashes are like plane crashes in 1980. It does not matter where it happens, we hear about it instantly. It seems like there is someone dying every day, and yes, around the country or around the globe that's what is going on. But, for example in my local tri community, I am not aware of a single person who has died in a bike crash in 10 years. Probably 20, likely 30 years. I'll take a stretch and say in Ottawa Canada, a city with over 1M people with around 200 IM finishers every year and several 1000 triathletes doing races, no one gets hits and dies...EVER (someone jump in if you are aware of a single triathlete who died in this city since tri was invented as a result of getting hit). People get hit and get hurt.

The point being, triathlon did not get unsafe. From a mortality perspective its pretty darn safe. I look forward to the stats so we can prove this out and dispel the fears people may have because they see the aggregated incidents on the internet.

One more thing, I would like to apologize if earlier in this thread any of my commentary sounded mechanical and heartless. I was unaware when this thread started that one of your peers was on the wrong end of the bike-car crash and worse yet, that she's a co founder of Triathlon Today that I would eagerly read every month in the late 80's and early 90's thanks to her and Lew. Her and Lew touched more of us than they may be aware of.

"that's now how the stats work."

fine. i'm not a statistician. i'll generate the data, with your help and the help of other slowtwitchers. you do the stats.

---

Dan Empfield
aka Slowman

I'm not a stats guy, but had to take enough of those courses in engineering school....the main education point is "the more you ride, your odds don't go up"....they always stay constant. That's a good thing for the education process. You can improve your personal odds within the odds by doing things better than the median, but the chance of a texting driver plowing into your back won't go up the more you ride...it's wrong place at the wrong time....but we can improve our odds over the median, by riding when there are less cars, better light, better weather etc etc. I believe this can be an important part of the USAT/ST education once the raw data is compiled....thanks for doing this. I DON'T want to see people abandoning riding out of fear. If they do that, they should abandon getting in their car to go to the hardware store too.

Explain it like I'm stats-challenged, because I am. Knowing the 1/40000 per year figure, what are the odds of getting hit and killed in a 40 year career? Is it the same probability as during a 1 year career? Intuitively, I say no, but suspect this is somewhat like the gamblers fallacy, but in reverse.

I spent a little time doing some searching.

The most notable study was by someone you've quoted back in 2012, Dr. Creswell. He looked at 109 deaths during triathlon events between 1985 and 2015 and there was an article which came out in ESPN earlier this year at/after the ACC (American College of Cardiology) meeting which I did not go to (but I am a member), however several of my colleagues went, so I heard about this earlier this year. Of these deaths, 19 happened during the bike leg, most involving crashes/trauma according to the article (the study itself is not published yet).

http://www.athletesheart.org/...m-the-acc16-meeting/
http://www.espn.com/...eed-health-screening
http://www.slowtwitch.com/...y_Released_3190.html

I don't think finding specifics about these 19 deaths during a race are what you are looking for.
Finding and describing triathlete deaths recreationally, not during a race, IS what you are looking for and that will be a challenge to get details.

I still contend that you will not get much new information to help further reduce the risk of death, other than sensible things that are already known. Generating the data could prove useful, however, but isn't known.

As a cardiologist, I'm more worried about the fact the majority of deaths during triathlon are from undiagnosed ischemic heart disease. There should be a push for better screening to find heart disease in triathletes before they die during ANY race or in training (but to be honest, in general, athletes are at much lower risk than the general population).

But you still have to remember, we are dealing with a VERY LOW RISK population cohort-triathletes during cycling in training and during races.

"As a cardiologist, I'm more worried about the fact the majority of deaths during triathlon are from undiagnosed ischemic heart disease. There should be a push for better screening to find heart disease in triathletes before they die"

http://www.slowtwitch.com/...the_Water_j3378.html
http://www.slowtwitch.com/...Screening_j3400.html
http://www.slowtwitch.com/...wim_Alive_j3374.html
http://www.slowtwitch.com/...D_in_Tri__j3382.html
http://www.slowtwitch.com/...u_to_Move_j3437.html

and as a result:

http://www.slowtwitch.com/..._Ironmans_j3599.html

---

Dan Empfield
aka Slowman

Those are great articles and all relevant, even though all came out in early 2013. I did not say you weren't an advocate of screening for CAD and those are a testament to the fact you are moved to use your position here to do a 'greater good'.

Keep up the good work, Dan.

(I'm just trying to help out with data interpretation)

MadTownTRI wrote:

Explain it like I'm stats-challenged, because I am. Knowing the 1/40000 per year figure, what are the odds of getting hit and killed in a 40 year career? Is it the same probability as during a 1 year career? Intuitively, I say no, but suspect this is somewhat like the gamblers fallacy, but in reverse.

I'm not a statistician, but part of my job is data analysis sooo maybe I'll get it right? Your intuition seems correct.

Think of it this way:

You have a 20 sided die, and if you roll an 11 you're dead.

• Day one: 1/20 chance you'll roll and die. You roll a 9, whew.
• Day two: 1/20 chance you'll roll and die. You roll a 15, whew.

Tomorrow? You'll have a 1/20 chance again. But if you roll once per day over the next 10 days? That is more than a 1/20 chance of rolling an 11 when looked at as a group. I don't need to get into the statistics of what the odds are, because the only thing that is important is that when you look at the opportunity to roll 11 beyond the scope of an individual occurrence the probability increases.

So of course your odds of dying on the bike are greater if you bike 365 days per year vs. 10 trips to the store.

---

Too old to go pro but doing it anyway
http://instagram.com/tgarvey4

Hello Slowman and All,

I understand that you want to reinvent the wheel here but sometimes helpful valuable information can be gleaned from the past efforts ...

Three studies that are interesting for methodology and results ... though the data input is not from a triathlete base ..

I have read this entire thread and apologies if these reports have been noted already ...

You could use some of the methodologies as a template and input the triathlete crash data.

I think is should be noted that the origins of fatal crashes data can be quite different than the origins of injury crashes data ... often because of the speed of the motorists traveling on long straight sections as compared to the speed of motorists in and near intersections ... so fatal and injury reports are often developed from different sources ..... injury crashes from hospital and EMS data and fatal crashes from police reported data .... although there is some overlap.

It would be useful if you could entice someone working on their PhD or other academic endeavor (with the influence of a University) to join the project ....... and could make getting records from hospitals/EMS or coroners (with names redacted if required) easier.

The City of Boston Cyclist Safety Report http://www.cityofboston.gov/news/uploads/16776_49_15_27.pdf is quite transparent in the methodology and show results in easy to understand color graphs .. for instance on page 62/94 .. Appendix C - Selected Maps .... all reported bicycle collisions for period 2009-2012 are depicted on the map so 'hot spots' are easily detected.

For example some of their approach is shown here:

"Approach ...

The 2013 Cyclist Crash Report is comprised of multiple sections.
The Findings section consolidates information from the BPD, EMS and Boston Bikes to provide a holistic picture of crashes in Boston.

The report then makes recommendations based on the findings.

Lastly, the document provides the original crash report prepared by the BPD and EMS. By offering the original information from each department, the reader can most accurately and fully understand the data.

Data for this report comes primarily from two sources: the BPD and EMS.
Supplementary data comes from Boston Bikes.

Boston Police Department

The BPD collects standard variables from Boston Police incident reports related to each collision, such as da te and time of occurrence, details on involved parties and their property (i.e., motor vehicles), and an open-ended description of the circumstances surrounding a collision. The majority of these variables are collected into fields that are easily transferred into the CAD database, with the exception of the narrative section, written by the police officers that respond to the scene. As such, in order to extract relevant data for this section of the report, the group reviewed and coded thousands of narrative files originating from these incident reports.

The BPD, in collaboration the Boston Area Research Initiative at Harvard’s Radcliffe Institute, the Harvard Injury Control Research Center, and the Boston Cyclists Union reviewed BPD crash data to generate baseline crash estimates and
maps for planning purposes.
•
BPD’s Office of Research and Development provided senior research analysts and interns.
•
The Radcliffe Institute and the Rappaport Institute provided funding for a graduate level research consultant to conduct bicycle and pedestrian injury research studies as a part of her PhD dissertation.
•
Harvard University’s Institute for Quantitative Social Science also provided an in kind PhD level computer scientist to assist the student with data manipulation, programming, and advanced modeling.
•
Boston Cyclists Union supplied volunteer data coders for the project."

Examples shown below with some excerpts on the most common fatal motorist/cyclist crash ...

While the Cross/Fisher report is dated (1977) ..... it is the Rosetta Stone for bicycle crash methodology.

http://ntl.bts.gov/...9/DOT-HS-803-315.pdf

Excerpt: [page 229]

CLASS D PROBLEM TYPES

Class D includes five problem types that occurred when (a) a vehicle
overtook and collided with a bicyclist traveling in the same direction, or
(b) the threat of an overtaking motor vehicle caused the bicyclist to
collide with an object that obstructed the path he would have taken if the
obstruction had not been present. Class D does not include cases in which
the bicyclist turned or swerved into the path of an overtaking motor
vehicle.

Table 36 lists the problem types and subtypes for Class D and shows
the proportion of fatal and non-fatal cases that were classified into each
problem type and subtype. It can be seen in Table 36 that Class D accounted
for nearly 38% of all fatal cases and that nearly one-fourth of all fatal
accidents were classified into Problem Type 13.

http://bikeleague.org/...EBC_report_final.pdf

Every Bicyclist Counts

It is interesting to note that this data below on rear end collisions is still similar to the findings of the Cross Fisher report of 1977.

Excerpt:

Common Collision Types

Through Every Bicyclist Counts, we were able
to find out significantly more about fatal bicycle
crashes than is publicly available in FARS. This
reinforces the idea that better data collection and
reporting is possible and should reaffirm the commitment
of NHTSA to improve FARS reporting.
In 2012, FARS experimented with providing data
based upon the Pedestrian and Bicycle Crash
Analysis Tool (PBCAT), but abandoned that effort
in the hope to resume it in 2015 or 2016 when it
reports data from 2014. The crash typology used in
our initiative does not exactly match that of PBCAT,
but it is broadly smiliar. For example, the most common collision type in
our Every Bicyclist Counts data is a rear end collision.

Approximately 40% of fatalities in our data
with reported collision types were rear end collisions.

This is higher than what was found in the
2010 FARS release that included PBCAT-based
crash types (27% of fatal crashes with reported
collision types), although the crash type “motorist
overtaking bicyclist” was the most common collision
type in that data as well.

---

Cheers, Neal

+1 mph Faster

For very low odds like this, I would just add then up. Every year you have a 1/40000 chance, so 40/40000 or 1/1000.

Dev said earlier to Dan in the thread that 1/1000 wasn't how the stats worked, though, so maybe he can clarify. To me, 1/1000 vs 1/40000 is enough to get my attention. It's basically someone in my section vs someone in the entire football stadium. Feels more proximate, though still very insignificant chance.

"Hello Slowman and All, I understand that you want to reinvent the wheel here but sometimes helpful valuable information can be gleaned from the past efforts"

you haven't read the thread, have you?

---

Dan Empfield
aka Slowman

I started triathlon coming from running and I have been also scared of going out with the bike.
I live in a safe place but I always wanted to minimize the risks, training 90% of the time indoor with a powermeter. it was very effective but long term I started to love cycling. I started then cycling more and more outside, seeing bigger improvements with this type of training.
I went into Granfondo events and similar cycling races just to use them as ironman preparation.

Then, all of a sudden, one morning, alone on the road, no cars, nothing around me, straight road with cycling path....bum, 17mph full ahead against a tree. nobody knows and I don't remember why.
4 operations, shoulder, collarbone, lungs... but legs and arms were safe
now it is 3 months from the crash and i can bike and run but swimming is at the moment impossible because of shoulder operation.

What will happen now is that I will go back training 90% indoor as I was doing at beginning

I think it is a matter of minimizing the risks. if you are not into cycling racing, you can train most of the times indoor without any issue. of course, your mind needs to be ready for it.

good luck

Bit late to the thread but my thoughts:

- Suggestion to try and get data from Strava seems like a great one. If they were prepared to share it (a big if...) it would offer a pretty comprehensive insight into the riding habits of the majority of people riding for sport, which would help to quantify how risks vary by road type, region, time of day, etc
- I think a cohort of "triathletes" is too narrow. A lot (most?) triathletes will also have a road bike and take part in non-tri events, and vice versa for cyclists doing the occasional triathlon. I'm not even sure if I would define myself as a "triathlete" at the moment (have done a lot in the past, still S, B and R regularly, but haven't done an actual triathlon in a while). Seems to me that riding X miles per year on a road/TT bike in a certain area carries the same risks regardless of whether or not I'm running and swimming as well. I knew a guy who died cycling last year, he did a lot of tris but for the last year or so of his life was focused on pure TTing and wasn't swimming or running much. Would he fall into the "triathlete cohort" or not?
- Crowd-sourcing the accident data is an interesting idea. I'd look to do that but also pull in any other stats you can as well and try to cross-reference. Fatal accidents nearly always lead to a Facebook page and typically some fund-raising efforts for example.
- Fatalities is certainly going to be the easiest to establish. Life-changing damage (brain damage, paralysis, etc) shouldn't be much harder though, and arguably is at least as relevant for most of us in terms of evaluating the risk

For me I think how worthwhile the stats would be depends on how much you can break them down. Stats by country or by region are interesting in terms of understanding the impact of cycling policy and infrastructure on risk, but it's not going to mean much for individuals. I'm certainly not going to convince the wife that we move to a different country or even region in order to reduce my chances of being killed while cycling from 1 in 15,000 to 1 in 20,000. If you can start breaking them down by road type and type of day that gets interesting though. E.g. if my commute (done at rush hour on fairly urban roads) was 5x riskier than my weekend rides (done early in the morning in light traffic on fairly rural roads) that would really make me start to evaluate whether that was a smart choice.

Most interesting of all for me, but probably hardest to establish, is cause of accident. When I see the stats on cycling deaths, I like to reassure myself (and my wife) by comparing my riding to some of the people I see out on the roads who are flying through red lights, undertaking moving traffic at junctions, etc. Would be great to have numbers to back up how much (or little) we can actually mitigate risks through our own behaviour vs how many deaths are completely out of the cyclist's control.

I get where you are headed, Dan. And I agree - we are a different subset of riders than those collected and tracked through big agencies - the ones who aren't wearing helmets, are on BMX bikes, riding to the store, or whatever. We are a very small slice of the total population and if we figure there are 500,000 (or 450k or whatever) triathletes, how many of them are laying down the hours on the road (versus inside or inactive or ??)? In your data collection I'd wonder if you could collect average hours/week victim cyclist spent riding outdoors. This would be very telling in how much exposure the individual had.

Also, regarding "cyclist equipment: blinking rear light? " you might broaden it to be "safety equipment in use" because if they are hit from the front, did they have a front-flashing or steady light? Any reflectors? Was their jersey neon green? My husband looks like a neon christmas tree while I only have a rear flasher any my jersey is whatever I pulled out of my drawer.

I'm most curious about gender of rider, which you have listed. Nationwide, it is 88% or something that are male, which leads me to believe that as a female I have less probability, yet two recent deaths were women. I'm not sure the % of triathletes that are women, but I wonder if the national average of 88% male would hold in our population.

I think that if you are successful in collecting this data, overlaying it on a map, and comparing it to total tri population you would get very different statistics than the national stats. On the total tri population you could probably extrapolate number of active athletes by extrapolating race data and applying some assumptions on amount of training time might be included. It would be a lot of work, but maybe an engineer could data mine that for you. And perhaps some polling to get within an acceptable standard deviation of assumption parameters.

Also, it would be very interesting to see the risk exposure on hours on road versus total triathlete road hours. For instance, if of 500,000 athletes, 300,000 are actively training through the year, and each athlete trains 16 weeks at an average of 6 hours/week and half of that is outdoors, there are 14.4 million road hours of exposure. Obviously, the extrapolation would be more complex and hopefully scientifically derived. I'm actually not sure where to take that data point, but perhaps it would be "athletes training 100 hours for a race of which 50% of the time is on the road risk death by 2%" or something. I dunno. Just spitting stuff out here... but your project is intriguing and I think it has merit to evaluate our unique population and our exposure to death or injury.

The information is out there.... just need to go get it.

---

Hillary Trout
San Luis Obispo, CA

Your trip is short. Make the most of it.
https://www.slogoing.net/

"I think a cohort of "triathletes" is too narrow."

i agree. but are you tied into all the road racing cyclists in your state? the issue here is getting reliable, comprehensive stats.

"Crowd-sourcing the accident data is an interesting idea. I'd look to do that but also pull in any other stats you can as well and try to cross-reference."

how? if you use general stats on bicycle fatalities, only 1 in 6 wore a helmet. does this sound like your cohort? so, what accident data are you referring to?

so, again, and again, and again, what we need are:

1. stats that are RELIABLE.
2. stats that are RELEVANT.

if you can help me figure out a way to broaden our cohort while not wrapping our arms around people on bicycles who are not really relevant to our style of riding, i'm all for it. open to suggestions.

---

Dan Empfield
aka Slowman

I'm as tied into the road racing cyclists in my area as I am the triathletes. If anything I might be more tied in in terms of being aware of any serious accidents, since road cyclists who race are pretty much all members of a club and go out on group rides, whereas triathletes are more likely to train solo. There are 6 cycling clubs in my immediate locality that I can think of, most have some sort of triathlon wing or at least a subset of riders who also do tris, and there's also a fair bit of cross-fertilisation with riders who go out on rides with multiple clubs to suit their schedule. So although I only do rides with one of those clubs I'd be confident that I'd hear about any serious accidents involving any rider from any of those clubs. You obviously have a lot more data on the make-up of your membership than I do, but from the posts I see I'd be surprised if this forum didn't have very good ties into the road racing community. If you're prepared to crowd-source, you could also consider crowd-sourcing through other sites, though I think cycling sites are a bit more numerous and fragmented, I can't think of one that has the same pre-eminence that ST has in the triathlon community.

You can use the general stats as a sense check to gauge the reliability and coverage of the numbers you're getting from your crowd-sourcing. The 1 in 6 wearing helmets you quote is a great example, I'd agree that anybody who races even occasionally in any kind of cycling event is almost certainly going to be wearing a helmet most of the time, so straight away you've established an upper limit for how many fatalities you'd expect to uncover. It's also why I thought the suggestion of Strava was a great one for establishing what the denominator is in terms of your stats. Certainly from the people I know, a heavy majority of those who I would consider to be in "my cohort" are on Strava (i.e. they either race or are at least treat cycling as a sport not as a convenient means of transport or family activity), and equally people who only use their bikes for popping to the local shops or going out with the kids are almost certainly not on Strava.