There was a glowing article in theatlanticcities.com with the tantalyzying headline Dedicated Bike Lanes Can Cut Cycling Injuries in Half,  referring to this study, published in a peer-reviewed, albeit public health and not a transportation, journal:

Route Infrastructure and the Risk of Injuries to Bicyclists: A Case-Crossover Study Kay Teschke, PhD et al.  Am J Public Health. 2012;102:2336–2343. doi:10.2105/AJPH.2012.300762  [pdf]

The design of the study is intriguing: it’s based on randomly choosing a “control site” along the participant’s (i.e. the crash victim) route.

Cycle Tracks are NINE TIMES safer?

Undoubtedly, the incredibly safety differential of “cycle tracks” will be the main take-away. The study found them to be NINE TIMES safer compared to their reference street (essentially a “worst case”: a mulitlaned arterial with on-street parking and no bicycle facilities whatsoever). The actual result is OR 0.11 (0.02, 0.54) — that is to say Odds Ratio of about 9 times safer, compared to the reference road.

Ok, so I don’t understand a lot about statistics, but the wide range between the lower and upper confidence interval (27X) is a clue. In short there is not very much/many cycle tracks in the study, mentioned only as “despite their (cycle track’s) low prevalence in Toronto and Vancouver”. There were two reported collisions, and 10 control sites on cycle tracks (out of N=648). In the critique of the study by John Forester he found during the study period there was apparently only one cycle track, the Burrard Street bridge, in both cities — my that is a “low prevalence” — here is his take-away:

In the much more impressive cycle-track issue, the authors proclaimed enormous crash reduction without informing the readers of the two relevant facts. First, that their data came from only one installation. Second, that that installation was not along a typical city street but in the only situation in which a plain cycle track could possibly be safe, a place without crossing or turning movements by motorists, cyclists, or pedestrians…

And even regarding the Burrard Street Bridge cycle-track, the timeline seems to conflict/overlap somewhat with the study dates. According to a surprisingly detailed account on wiki a test of what sounds to be the cycle-track was “to begin in June 2009. The proposed trial began on July 13. It saw the southbound motor-vehicle curb lane and the northbound-side sidewalk allocated to bicycles, with the southbound-side sidewalk allocated to pedestrians. The reassigned lane was separated from motor vehicles by a physical barrier” The timeline of the study was for bicyclist injuries presenting to the ERs “between May 18, 2008 and November 30, 2009″.

But wait? According to this (from mid-2011, i think, the date is unclear), Tesche said there are other cycle tracks:  “However, we were able to examine separated bike lanes elsewhere in the city, including Burrard Bridge, Carrall Street, and other locations that met our definition: that is, a paved path alongside city streets that’s separated from traffic by a physical barrier,” Teschke told councillors.

Some Other Things i Noticed

The highest median observed motor vehicle speed along major roads was 44kph (27mph)! This is comically low compared to what I am used to here in Phoenix. Interesting trivia answer: 27.79mph —  the fastest time on record for a person running.

One-third of the incidents involved collisions with MVs. The balance were various types of falls or collisions with objects. The one-third number is pretty close to the 26% reported by another ER-based survey of bicyclist injuries (  Injuries to Pedestrians and Bicyclists: An Analysis Based on Hospital Emergency Department Data.  linked here ); though this isn’t directly comparable, e.g. in the former case, mountain biking was not eligible for the the study, whereas in the latter it was any sort of injury incurred on a bike.

There was a bunch of interesting data collected in the survey (which the author’s are nice enough to give a link to) that are not in the final study. I’m not sure why. I would have been interested to see various spins on lightness/darkness vs. cyclist’s light usage.

The Injury Prevention Article

and here’s another similar article, or perhaps pretty much the same(?):

Comparing the effects of infrastructure on bicycling injury at intersections and non-intersections using a case–crossover design Inj Prev doi:10.1136/ip.2010.028696 M Anne Harris, Conor C O Reynolds, Meghan Winters, Peter A Cripton, Hui Shen, Mary L Chipman, Michael D Cusimano. Shelina Babul, Jeffrey R Brubacher, Steven M Friedman, Garth Hunte, Melody Monro, Lee Vernich, Kay Teschke

NYC Protected Bike Lanes on 8th and 9th Avenue in Manhatten

According to a report (it’s really a brochure) by NYC DOT cited by  americabikes.org; these are the “First protected bicycle lane in the US: 8th and 9th Avenues (Manhattan)”…”35% decrease in injuries to all street users (8th Ave) 58% decrease in injuries to all street users (9th Ave) Up to 49% increase in retail sales (Locally-based businesses on 9th Ave from 23rd to 31st Sts., compared to 3% borough-wide)”. I don’t know if or what the data are to back up these claims. I also don’t know much about how these are structured, what was done with signals, how long these are,  or how long they have been in place… here is a google street view at 9th/23rd. (these segments show up in Lusk’s May 2013 AJPH article, discussed below)

Study of Montreal Cycle Tracks

Likewise, Harvard researcher Anne Lusk, et. al (includes Peter Furth, Walter Willett among others) has claims of safety increases  Risk of injury for bicycling on cycle tracks versus in the street, brief report Injury Prevention. Streetsblog.org is expectedly uncritical, but a through rebuttal by mathemetician M Kary can be found hosted on John Allen’s site(older, 2012), and more recently (Jan2014) including links to Kary’s two original unedited letters, as well as the published commentary in Inj Prev. , which includes a rebuttal from the authors. There is some other rebuttal from Ian Cooper, in a comment below.

Methodology aside, though the study claims an increase in safety, it found only a modest increase: “RR [relative risk] of injury on cycle tracks was 0.72 (95% CI 0.60 to 0.85) compared with bicycling in reference streets”. I.e. a 28% reduction in crashes.

They had an interesting reference to Wachtel and Lewiston 1994, a much-cited sidewalk study.

More Lusk, July 2013 Article in AJPH

Oh, it’s like it never ends:

Bicycle Guidelines and Crash Rates on Cycle Tracks in the United States
Anne C. Lusk, PhD, Patrick Morency, MD, PhD, Luis F. Miranda-Moreno, PhD, Walter C. Willett, MD,  DrPH, and Jack T. Dennerlein, PhD Published online ahead of print May 16, 2013; it was in the July printed edition of American Journal of Public Health. “For the 19 US cycle tracks we examined, the overall crash rate was 2.3 … per 1 million bicycle kilometers… Our results show that the risk of bicycle–vehicle crashes is lower on US cycle tracks than published crashes rates on roadways”. What are published rates? Later they say “published crash rates per million bicycle kilometers range
from 3.75 to 54 in the United States”. The first number is footnoted to Pucher/Irresistible (which is discussed and linked here), and the second to, if you can believe it, a study of Boston bicycle messengers (Dennerlein, 2002. I haven’t bothered to look that one up). In Pucher, it’s in Fig 10  where they quote US injuries at 37.5 per 10 million km for the period 2004-2005, sourced to US Department of Transportation (2007), which is/are Traffic Safety Fact Sheets according to the footnotes. Pucher does, um, mention that injury rates comparisons across countries are particularly suspect; Figure 10 would lead on to believe the UK and US have similar fatality rates, whereas US injury rates are quoted as SEVEN TIMES higher. (Pucher’s claim/point is that NL and DK are very safe, while US and UK are very dangerous). In any event TSF does not list injury rates per unit of travel, only number of injuries, e.g. TSF 2005 quotes 45,000 injuries (these are presumably some sort of statistical estimate?).  To get the rate estimates, he uses one of the surveys (household trans survey?).

Paul Schimek gathered data on the 19 cycletracks listed in table 3; he added another column “intersections per km” and sorted them into two groups, 1) Urban Side Paths and 2) Side Paths with Minimal Crossflow. And as would be predicted by traffic engineering principles, the former had very high (7.02) versus the latter which had very low (0.57) crashes per 1 Million bicycle kilometers. The published letter-the-editor of AJPH is available in full on pubmed (or draft version on google docs) which is well worth reading. He, by the way, provides an estimate for whole US bike crashes at 3.5 per 1M bike km’s; which fits rather nicely between the high/low cycletrack numbers. The bottom line is that the AASHTO guidelines (which prohibit the on-street barriers; but permit bicycle paths adjacent to the roadway where there is “minimal cross flow by motor vehicles”) , contrary to Lusk’s assertions, are well-founded. This blog post at  bicycledriving.org also discusses the same AJPH article, with links to both Schimek’s published letter, and Lusk’s published response. This is wrapped up in an article the Paul wrote A Review of the Evidence on Cycle Track Safety, Paul was kind enough to send me draft copy dated October 10, 2014.

Oh, and here is John Forester’s review of Lusk’s May AJPH article. In summary, Forester says “This review does not evaluate Lusk’s method of calculating car-bike collision rates. However, the cycle tracks with high collision rates are all in high-traffic areas with high volumes of crossing and turning traffic, while the cycle tracks with low collision rates are all in areas with low volumes of turning and crossing traffic. That is what should be expected, but it says nothing about any reduction in collisions that might have been caused by the introduction of cycle tracks. The data of this study provide no evidence that cycle tracks reduce car bike collisions”.

What about Davis, CA?

The article/thesis paper Fifty Years of Bicycle Policy in Davis, CA 2007
Theodore J. Buehler has a deep history. Davis, home of course to UC Davis, installed and compared designs including what we would now call a cycle track in the late 1960’s as “experimental” designs, (emphasis added):

Lane location relative to motorized traffic
The early experiments included three different types of bike facilities (see examples at the top of this section):

1. bike lanes between car lanes and the parking lane (Third St.),
2.  bike lanes between the parking lane and the curb (Sycamore Lane), [what we now call a cycle track, or protected bike lane] and
3. bike paths adjacent to the street, between the curb and the sidewalk (Villanova Ave.).

… The on-road lanes worked best, the behind-parking lanes were the worst, and the adjacent paths were found to work in certain circumstances.

Perhaps telling, perhaps not, I have archived the .pdf referenced above as I can no longer find it on the bikedavis.us website. There is a similar version of Buehler’s paper that was published through TRB with the same title (but with a co-author, Susan Handy); its conclusions are worded somewhat differently; instead of best and worst, they just say “Eventually all lanes were converted to the now familiar configuration of the bike
lane between the moving cars and parked cars” without saying why.

Notations from the City of Davis website says (retrieved 1/19/2017. Emphasis added):

Other Critiques

Ian Brett Cooper offers this critiques of a number of papers involving  segregated infrastructure, e.g.:

2012 Teschke: Route Infrastructure and the Risk of Injuries to Bicyclists: A Case-Crossover Study
Selection bias: uses comparison streets instead of a before-after situation; study claims greatly increased safety on cycle tracks, but the cycle tracks chosen for the study were not representative of a typical cycle track, in that all were on roads with limited or nonexistent road intersections. It is not surprising that bicycle facilities that have little or no possibility of interaction with motor vehicles are safer than those that have many such possibilities, and if all bicycle tracks were completely separated from turning and crossing traffic, they would indeed be safer than cycling on the road. The problem is, cycle tracks with few road intersections are very rare indeed.

2011 Lusk: Risk of Injury for Bicycling on Cycle Tracks Versus in the Street (Montreal, Canada)
The infamous Lusk study. Selection bias: study claims increased safety on bicycle specific infrastructure, but its street comparisons are flawed – the streets compared were in no way similar other than their general geographic location. Busy downtown streets with multiple distractions per block were twinned with bicycle tracks on quieter roads with fewer intersections and fewer distractions..

.

IIHS (2019) (#IIHS)

Some protected bike lanes leave cyclists vulnerable to injury

and the study page, more fully titled;

Not all protected bike lanes are the same: infrastructure and risk of cyclist collisions and falls leading to emergency department visits in three U.S. cities

Since Washington D.C. was a primary study site, and the home of a significant amount of separated bikelanes/cycletracks, The WashPo ran a news item on the study under the title D.C.’s oldest and most popular protected bike lane has ‘highest injury risk,’ study says, two-way cycletracks being unsurprisingly the most problematic.

 Highlights (part 1):

* Same “case-crossover” method as Teschke et al. Bicycling in Cities Study, which is the only reliable one yet used with N. American data.

* Unlike that study, there were actually PBLs installed in the cities where data was collected.

* They found NO safety benefit for one-way PBLs and a significant indication of higher risk with two-way PBLs.

* They reproduced the Teschke finding of higher risk for going downhill and VERY high risk due to streetcar tracks (all in Portland, OR).

Highlights, part 2:

* Only 40% of bicyclist injuries were due to moving motor vehicles (data are from emergency department visits).

* 12% of injuries were due to non-moving motor vehicles. These include dooring, but it is not presented separately. The figure rises to 20% of injuries on “major-roads.” They do not separate roads with and without on-street parking.

* Ordinary bike lanes appear to reduce risk BUT the presence or lack of on-street parking may be a confounding factor. The risk reduction is only AWAY from intersections. At intersections there was a 4-fold increase in risk (but not enough data to be statistically significant).

* Further, there was no evidence that bike lanes reduce the risk of collision with moving motor vehicles (see Table 6).

* There were only 18 injuries on one-way PBLs, which was not enough to determine how they affect risk EXCEPT that there was enough to say that they increase the risk of bike-ped injuries.

* Even with only 21 injuries on two-way PBLs, there was enough evidence to show that they increase risk by an order of magnitude, specifically collisions with bicyclists and pedestrians (Table 6).

The study is yet more evidence that looking at motor vehicle crash statistics (which ignore incidents not involving mvs) will not give a true picture of the safety effects of bike facilities.

.

Why cities with high bicycling rates are safer for all road users

Another Marshall and Ferenchak study Why cities with high bicycling rates are safer for all road users; was commented on in this letter by Paul Schimek. In the study, F&M claim safety-in-numbers was not shown, but that “Better safety outcomes are instead associated with a greater prevalence of bike facilities – particularly protected and separated bike facilities”. Schimek observse that they mixed “trails” (off-street paths removed from roadways) with true protected and separated bike facilities. He also points out “Third, a significant p-value does not imply a causal relationship. With 112,918 observations, it is not difficult to find coefficients that pass conventional significance tests”. (is that “p-hacking“?). As well as some other observations.

An earlier F&M paper originally titled The Relative (In)Effectiveness of Bicycle Sharrows on Ridership and Safety Outcomes seems to have used the same dense statistical techniques they say establishes their premise; it involves census tract block groups.

Time marches on… #FHWA2023

This is a fhwa report style white-paper(not peer reviewed, I assume?)

FHWA-HRT-23-078
Developing Crash Modification Factors for Separated Bicycle Lanes
September 2023
Karen Dixon, et al.

FULL report .pdf (122 pages!)
Tech Brief (8 pages)
ODOT summary sent like a newsletter

The headline number, “up to 53 percent” reduction presumably comes from  Table 66. p.104 CMFs for converting to an SBL; Before condition: Traditional or flush buffered bicycle lane / After condition: SBL with flexible posts CMF 0.468 (i.e. a reduction of 52.8%) along with some statistical significance factors.

The “53% reduction” touted by the FHWA is flawed and/or misleading… the report specifically excludes intersection crashes entirely, and only looks at segments between… this rather remarkable fact isn’t well-publicized — notably it’s not mentioned in the abstract, which is the only thing most people read(!) — here is what the report states

… The second challenge is the inconsistent nature of SBL applications, particularly at approaches to intersections. Initially, the research team hoped to develop CMFs for segments and intersections, but their attempts to model intersections and/or entire corridors were unsuccessful.
Consequently, the team focused on developing robust CMFs for segments. Future work may be to conduct research that estimates the safety effect of the various types of SBL-to-intersection transitions. 
— p.104 “Future Research” in the full report

This was confirmed by email with one of the authors, who stated:

We did a quick check on the intersection-related crashes to develop CMFs for intersections as well. But since it was out of the scope of the project and because of the funding restrictions, we could not extensively explore intersection-related crashes. Depending on the availability of resources, we might analyze intersection crashes as well in the future.

The report doesn’t seem to have any data with regard to bicyclist crashes/falls with or caused by the vertical element (whether it be a bollard, flexpost, parked car or whatever); these crashes are known to happen but are not reportable as traffic crashes, and as such are a known unknown that we really need the answers to before putting up barriers everywhere (on urban streets).

Keri Caffrey of cyclingsavvy has a longer critique, which is, as usual, loaded with excellent graphics depicting common crash scenarios; here’s the crux of it:

The FHWA document states that 1/3 of Killed/Serious Injury (KSI) crashes are caused by overtaking motorists. In the data from the Orlando Metro area, overtaking crashes accounted for 10% of KSIs and overtaking in bike lanes was 2.1% of KSIs (14 crashes in 7 years). The percentages for overtaking KSIs will certainly be higher when adding rural roads to the dataset. Overtaking crashes are worse on rural roads. But that’s not where they’re wanting to put separated lanes.

Which is to say, real-world metropolitan crash data varies dramatically from overall averages; the places where the SBL projects are being proposed are mostly or virtually all distinctly metro/urban settings. An SBL, at best, reduces the mid-block crashes where a driver drifts into a (an UN separated) BL and KSIs a bicyclist. And worse yet, is the junctions (both intersections and driveways) can introduce new crashes — and remember, the referenced study EXCLUDED intersection crashes

Some discussion here (this may be a private group?)  https://www.facebook.com/groups/SupportersOfFullLaneRightsForBicyclists/posts/7605228416231380/?comment_id=7605372829550272

Related and by the way, I had a lot of trouble finding the “Green Lane Project” spreadsheet, lots of dead links, which is referenced in the FHWA report:

• • I found it linked here
  • Here is the direct link to a google doc sheet (last update was Oct 2021 according to the file):
  • Here is COPY of the spreadsheet in case the original disappears.