Big Data and orthodontic research
This is the first of the guest posts on my blog. It is by Professor Phil Benson, who is Professor of Orthodontics at the University of Sheffield (North of England). We have worked together for many years. His post is on the use of “big data” in orthodontic research as a potential source of very useful research information.
I am really pleased to be invited to contribute to Kevin O’Brien’s blog. Like many people, I subscribe and read, with interest, Kevin’s posts each week. I am an academic based in Sheffield. This is a city that is about 40 miles (64 km) east of Manchester in the UK. Sheffield is not as well-known as Manchester, probably because it doesn’t have a football team in the English Premier League, (but does have what claims to be the world’s first football club). It is a very green city, in the proud English county of Yorkshire and is the only UK city with a national park (the country’s first, the Peak District National Park, where I live) within its boundary. I actually did my postgraduate orthodontic training in Manchester, way back when Kevin was a lowly lecturer.
What is big data?
Kevin and I both enjoy cycling and often commute to work by bike (although Sheffield is a much hillier city than Manchester), so I was pleased to see this article, which was recently published in the British Medical Journal. The report confirms the substantial health benefits of walking or cycling to work, rather than taking a car. The findings are based on data from a project called UK Biobank, which involves a large cohort, of about half a million participants, who volunteer to undergo multiple tests and examinations and have their health regularly monitored over time.
UK Biobank is an example of ‘Big Data’, which is exciting researchers in both medicine and the social sciences (as well as being the basis of much of the success of large internet companies, such as Facebook and Google). The term ‘Big Data’ refers to information about, usually, a large number of individuals, which enables the use of sophisticated statistical techniques to examine relationships between multiple variables.
The exploitation of ‘Big Data’ has been made possible because of improvements in the storage of large amounts of digital information. There is also an increasing willingness to make all research data open and available for analysis (Research Data Alliance).
What about orthodontic big data?
Seeing the results being generated through large datasets has made me think about the numbers of people worldwide who have orthodontic treatment. If we were able to collect just a fraction of the information about these patients then surely this would help us answer some of the questions that we face as clinicians and that are regularly discussed and debated in this blog.
There would undoubtedly be challenges and problems with starting an orthodontic data bank. The first hurdle would be obtaining some agreement about what data to collect. However, people are starting to think about this. Just as in the 1990s, when the group, then known as the Cochrane Collaboration (now just Cochrane), was established to help the development and delivery of systematic reviews examining the effectiveness of healthcare interventions, there is now an organisation to support the development and delivery of core outcome sets for use in research and clinical audit (the COMET Initiative).
What do we do about bias?
Another potential problem is sampling. It is natural for clinicians to only present data from patients with the most favourable outcomes. This would clearly lead to a biased sample (much discussed in this blog), which would not be very helpful. There would need to be some way of ensuring that data from all patients who start orthodontic treatment are collected and entered at the end of treatment, regardless of whether the outcomes are favourable or not. Further criticisms of ‘Big Data ‘have been voiced, including the ethics of collecting and storing large amounts of data about people, as well as the underlying assumptions of any analysis undertaken. However, if these problems could be addressed, then this could be an enormously valuable resource for clinicians and the orthodontic patients of the future.
A final word about cycling..
Those of you interested in cycling might like to look at this report of a cycling related RCT. The author wanted to find out if there was a difference in the cycling time between riding to work (a 27 mile or 43.5 km round trip) on his carbon frame bicycle, which was then worth approximately £1000 or €1180 or $1560 (weighing 20.9 lb or 9.5 kg), compared with a steel frame bicycle, acquired for just £50 (weighing 29.75 lb or 13.5 kg). He tossed a coin every morning to decide which bike to ride (I know, not ideal) and collected data over 6 months. He concluded that the difference in journey times was neither statistically or ‘clinically’ significant (mean difference 32 seconds; 95% CI –3 mins 34 secs to +2 mins 30 secs; P=0.72).
He points out in the discussion that although the difference in the weight of the two bikes might seem large (30%), when you take into account the weight of the cyclist, the difference is actually quite small (4%). He goes on to state that other factors, such as traffic and weather conditions, gravity and friction will also have an effect. Perhaps those of us who simply want to commute to work on a bike should not be swayed by clever marketing to acquire a more expensive bike than we actually need. Possibly a lesson there for non-cyclists as well.
Emeritus Professor of Orthodontics, University of Manchester, UK.
Have your say!
“Possibly a lesson there for non cyclists as well” – love it!
as all cyclists know there is a formula for the number of bikes you need:
Bikes needed =Current No Bikes +1
but this is countered by the reality of
Divorce = No of bikes needed – 1
Ha Ha love it (and very true!)
why is coin toss not a good way for randomising a choice between two methods of transport over 180 tosses?
PS that’s a long cycle
If the cheap bike involved losing 4 teeth I’d pick the expensive one !