Could machine-learning algorithms predict the patients most at risk?
Anyone who has read my blogs before will know how keen I am about the latest technology in the industry. There’s something about the fact that we, as professionals, are living in an age of real innovation that is very exciting indeed, especially as so many of these breakthroughs have the potential to transform the way we provide care for patients moving forwards.
In my usual perusal of the headlines, I recently came across an article detailing how machine-learning algorithms could be the way forward when it comes to predicting which individuals are at highest risk for tooth loss – all without the need for a dental exam.[i] Researchers at the Harvard School of Dental Medicine suggested algorithms that not only took into account people’s health status but also socio-economic variables often outperformed those that only evaluated dental indicators.
In one study performed by this research group, data from 12,000 individuals was used to test five different machine-learning algorithms that considered this broad spectrum of factors. The results clearly suggested that once socioeconomic factors were taken into account, these algorithms could more accurately predict risk of tooth loss.
But how could such technology impact how we provide treatment in the future?
First of all, if these algorithms could be employed to vet patients before an appointment, this could take a lot of strain off professionals. We’re already moving towards an approach where we recommend a longer time between appointments for patients that are unlikely to have any issues, so having an algorithm to assess this without the need for a dental exam could help us to make this decision with more confidence. Plus, by freeing up appointment schedules to have more time for those that actually need to be seen, this could help high-risk individuals to receive a more tailored and regular standard of care.
On the other hand, the danger of trusting our decisions solely to algorithms is that people, no matter how they are categorised, are still individuals. We cannot reliably predict people’s behaviour, and there are a number of elements that could come into play that could render the algorithm less effective. For example, what if the patient starts smoking but doesn’t want anyone to know? What about drug addiction? Eating disorders? The behaviour of individuals is impossible to fully predict using technology, and this leaves weaknesses in the use of these algorithms that could be tantamount to professional neglect if we rely on them too heavily.
Still, in the end, these algorithms do pose an interesting question: will technological advancements shape our patient lists in the future? As ever, we won’t know until we get there, but I personally think that if algorithms continue to become more sophisticated and we continue to view patients’ health more broadly by taking into account more socio-economic factors, this could be the case.
For further information please call EndoCare on 020 7224 0999
Or visit www.endocare.co.uk
[i] Harvard Medical School. Predicting Tooth Loss. Link: https://hms.harvard.edu/news/predicting-tooth-loss [Last accessed June 21].
