Sunday 13 January 2013

Number Needed to Educate (NNE)

As someone who teaches both general clinical dentistry and, increasingly, evidence-based dentistry to postgraduates and undergraduates, I have been interested in how we actually measure the effectiveness of our teaching. But I have recently become very interested in how we as clinicians are encouraged to move toward more evidence-based practice and how as postgraduate educators we facilitate that in the people who attend our courses.

There's a group within the Cochrane Collaboration called the Effective Practice and Organisation of Care Group (EPOC) and these people have produced a number of fascinating reviews about how effective (or ineffective) different interventions are in producing a change in professional practice of healthcare workers. Here's a biased sample that have interested me:


Now, the meta-analyses that have been conducted in these reviews result in the authors giving a summary risk difference. This is the absolute difference in the risk of a professional changing behaviour when they are subjected to, say, a CPD course compared to the risk of them changing their behaviour if, say, they didn't attend the CPD course. Granted, it's a bit odd to describe the chance of converting to a positive behaviour as a risk but that's the way it is. 

So what's interesting to me is that in the systematic review on educational meetings above there was a risk difference of 6% overall of a positive change in practice if people attended educational meetings. But when they looked at studies that tried to change complex behaviour this came down to 0%.

Now I've been thinking that if we were to treat risk difference (which is absolute) like absolute risk reduction we'd be able to determine a 'Number Needed to Educate' (NNE) that would result in a single change in practice.

You may be familiar with NNT (number needed to treat), which I have blogged about previously. Basically it describes for interventions the number of patients we would need to treat with a given treatment before we'd expect one positive result. The lower the NNT the more effective the treatment. 
"17 people sit for a whole meeting about a best practice and yet only one of them changes their behaviour."

The NNT is determined by dividing 1 by the absolute risk reduction. So if the absolute risk reduction is 0.1 (or 10%) then we'd need to treat 1/0.1=10 patients to see a benefit.

So could we start to do the same with different ways of changing our behaviour to one that is more evidence-based? If educational meetings have a risk difference of 6% overall, then the NNE (Number Needed to Educate) would be 1/0.06=17 (rounded to nearest whole number). So, get this, 17 professionals would have to sit through an educational meeting for one of them to change their practice to what is considered best practice. That's astounding. 17 people sit for a whole meeting about a best practice and yet only one of them changes their behaviour.

This has totally changed the way I see the group of professionals who attend my EBD courses. I have a 1 in 17 chance if I simply stick to a single educational meeting of changing their behaviour or it could be worse, since the risk difference was 0 for complex behaviours (i.e. NNE=1/0=infinity - I'd have to educate all the world's dentists in EBD and still I wouldn't expect one to change their behaviour...)

I think it's time to change the way we think about what we aim to do as educators. Whether we're teaching periodontics, endodontics, restorative dentistry - whatever - we need to look at those EPOC reviews and think about how best we facilitate changes in our own and others' behaviour so that as a professional group our practice becomes more evidence-based. It may be that we have to think out of the box and to engage other practices such as finding 'opinion leaders', developing outreach programmes, and encouraging the use of audit whilst providing feedback. 

Anything to make the chances of success better than 1 in 17.