Don’t let data lead you astray

Parkrun rules, OK?

Parkrun, the free, timed 5km run that takes place at over 1,100 places in the UK each Saturday, is brilliant. I recently completed my 100th Parkrun so have run a lap or two. The events are popular with people of all ages, shapes, sizes and running abilities. The mantra is that it’s a run not a race, and all are welcome. But this week Parkrun announced a change of rules.

Some people who attend Parkrun run with their dog. And many do so using a harness attached to their waist, with their pooch running directly ahead of them connected to the harness via a special bungee-like lead. Running with a dog in this way is a sport in itself, popularly known as CaniCross (or CaniX). The setup and allows both human and canine to run safely and comfortably.

My wife and our dog frequently run with a local CaniCross group, and I occasionally enjoy a dog-powered tow on local runs too.

However, from 2 April, this CaniCross-style setup will be banned from Parkrun. Instead, any dogs must stay at their owner’s side on a short hand-held lead.

According to Parkrun’s blog post about this, the change “is based on clear evidence contained within our incident database” which shows that “over ten percent of incidents at Parkrun events involve dogs.”

Parkrun go on to say “Our evidence shows that, when participants use waist harnesses, there is an increased risk of serious incidents, particularly trips and falls, compared to when using handheld leads.”

As I say, I’m a huge fan of Parkrun and the organisation is to be applauded for its myriad of achievements.

But the nature of this announcement flicked some professional switches and got me questioning the basis for it – especially the relevance of the stats and the inferences that are perhaps being made.

Barking up the wrong tree?

Earlier today, I tweeted to ask Parkrun to share the data upon which they are basing their decision.

Publishing the evidence being referred to would either help to show the solid thinking behind the decision, or enable the Parkrun community to contribute more meaningfully to the debate.

At the time of writing, I’ve had no response, but will update this post if one arrives.

As it stands, much seems to hang on that ten percent.

At face value, if ten percent of incidents involve dogs, this feels like something worth tackling.

My experience of 100 Parkruns is that less than ten percent of participants bring dogs, so the incident rate does seem unduly high. And if there is evidence that waist harnesses lead – no pun intended – to more trips and falls than handheld leashes, there is a strong reason to consider change.

But the problem is that there is so much we don’t know.

To start, we don’t know what that ten percent means in terms of actual numbers

We know that 120,000 people took part in a Parkrun within the UK last weekend.

In my experience, trips and falls are rare. But let’s say one person in every hundred was involved in an incident last weekend. That’s 1,200 bumps or bruises. The ten percent stat means that 120 involved dogs in some way.

In turn, this says that if you were one of the 120,000 people at Parkrun, you had a 1 in 1,000 chance of getting a grazed knee or twisted ankle due to tangling with a terrier or tripping over a whippet.

And that sounds pretty low. Yet at the same time, my starting-point stat of one person in every hundred being involved in some kind of incident (i.e. one that would warrant it being reported and recorded) feels pretty high.

Anecdotally, I’ve run 100 Parkruns and have not been involved in any such incidents and only seen one or two trips and falls. My gut says that something that merits being logged as an ‘incident’ is likely to occur to fewer than one in a hundred people. Maybe one in every two, three, four or five hundred, which makes your risk of a spaniel-induced sprain or a collie-caused cut two-to-five three times less likely – or once in every 2,000 to 5,000 Parkruns.

But currently we just don’t know.

And what even qualifies as an incident?

While an incident could include the likes of complaints by participants and issues spotted by marshals, let’s assume it refers to occasions involving an injury that merits at least minor first aid or assistance (for example, a bleeding cut or a sprained ankle).

Has cause or responsibility been taken into consideration – and is this recorded in a relevant way?

If Larry trips over a Lucy’s Labrador, Lucky, was this Lucy’s fault, or was it Larry’s for not looking where he was going?

If Lucky trips up Larry due to Lucy’s long lead and lack of control, this is a dog-related incident.

If Larry is not looking where he is going and trips over Lucky, this should not be recorded as a dog-related incident – or at least not in the same way. (Clumsy Larry may have been just as likely to trip over another runner.)

Insurance companies distinguish between at-fault and no-fault claims as the difference has an important bearing on the way they set your premium.

Has Parkrun made a similar distinction?

Are all incidents recorded, or is there a propensity for injured parties to report them more in some circumstances?

Maybe when Betty accidentlaly bumps Bob to the ground and bruises his bum, Bob gives Betty the benefit of the doubt. “It’s one of those things, hey-ho, let’s get up and jog on!” But maybe when Peter ends up in a pile after a tangle with Paula and her Pointer, he’s more likely to be pissed off and moan to a marshal.

In a similar way, we all tend to be more foregiving if an old lady seems to push in front of us in a supermarket queue than if a rowdy teenager does the same thing.

But again, where Parkrun’s data is concerned, we don’t know whether there is any such bias.

What evidence is there that running with a dog on a handheld lead will reduce the number of incidents?

Is there a record of the number of ‘Barkrunners’ at each event and what control method they use? Without this, any figure that says approach A leads to more incidents than approach B is meaningless.

Let’s say eight out of ten people who run with dogs currently use a harness versus two out of ten using a lead. With all other things being equal, you would expect four times the number of dog-related incidents to involve harnesses than handheld leads.

But if all you know is the number of incidents and the restraint method in those instances, you lack the context to infer anything of value.

Not only that, but who’s to say that forcing people and dogs who are used to running with a harness to change and run with handheld leads will not throw both parties out of kilter and result in more indicents rather than fewer.

In the UK, there are many, many more accidents involving right-hand-drive cars than left-hand drive ones (by virture of there being many, many fewer cars with steering wheels on the left). But if we inferred from the accident stats alone that left-hand-drive cars were therefore safer, and we gave everyone a left-hand-drive car, could we expect a reduction in accidents?! Would people switching from right- to left-hand-drive cars become more or less safe on the road?

What do we actually want to achieve?

To finish, let’s just say that all the yet-to-be-revealed stats are sound, and there is no unintentional bias or miscalculation due to the recording methodology. Does that mean that a ban is the best solution?

Exponents of CaniCross have suggested that imposing a restriction on the length of bungee – maybe by selling an official Parkrun bungee – would be a better solution. Getting folk with dogs to start a few minutes ahead of the rest of the Parkrunning pack might be good too.

Both are workable – or at least worth trying – and could be two of many alternatives that manage the challenge based on consultation with the people in question, rather than maybe simply inferring problems and deriving solutions from a spreadsheet.

And anyway, what is the aim in any case?

Why the focus on the apparent cause of ten percent of incidents? What causes the other ninety percent?!

Should the aim be to reduce the total number of incidents by ten percent, or just that ten percent where dogs are involved?

Would it be seen as a success to remove 120 dog-related incidents at Parkrun every week if that came at the expense of losing 1,200 participants who no longer feel welcome or able to comfortably (and safely) take part with their dog?

There is so much here that interested members of the Parkrun community like me do not know – and will not know until the data is shared with us.

We know what we don’t know, but have Parkrun asked and answered questions like these themselves?

Do they know what they don’t know?!