Introducing
Artificial Intelligence (AI) across primary care
could help GPs identify patients at increased risk for lung
cancer
up to four months sooner, a study has found.
Researchers at the Amsterdam University Medical Centre and other medical institutions developed an algorithm to identify those patients and tested it by analysing the files of 525,526 patients in the Netherlands, of whom 2,386 (0.5 per cent) were diagnosed with lung cancer.
The algorithm was able to identify those patients whose risk of having lung cancer was almost 10 times higher than the general population – and thus would have referred 1,480 (62 per cent) of the patients four months earlier.
However, the authors admitted that the system came at a cost. They said that for every patient diagnosed four months earlier, 33 patients will need to undergo a diagnostic test because of presumed increased risk – such as a chest X-ray or low-dose CT scan – and about 40 per cent of all patients with lung cancer will be missed and referred four months later following usual care.
This is because AI algorithms are only as good as the data – such as age, sex and symptoms – and assumptions they are fed. Previous research has found that biased representation of patient populations and medical scenarios in the training datasets can lead to inaccuracies.
“The development of machine learning-based decision support systems is a promising but complex process,” the authors wrote. “This is also true for its road to implementation. Besides the methodological, technical, and epidemiological challenges, ethical, societal, legal, and privacy issues should also be considered in future research.”
The researchers said their methodology might also be helpful to detect other “silent” types of cancers with a poor prognosis, such as pancreatic, oesophageal, and ovarian cancer, although it would need to be checked in different countries and healthcare systems.
They also suggested the observational study, published on Tuesday in the
British Journal of General Practice
, could be particularly relevant to the UK, where the five-year survival rate is 16 per cent and has improved only slightly during the past 40 years. This survival rate deteriorates with higher stages at diagnosis, from 56.6 per cent at stage 1 to 2.9 per cent at stage 4.
Read Next:
How to cut your cancer risk in each stage of life – from teens to old age
In countries with a gate-keeping system, like the UK and the Netherlands, patients can only access specialist care after referral by their GP, or in case of emergency. In the Netherlands, 90 per cent of patients with lung cancer are diagnosed after referral by their GP.
Although the authors did not carry out a costs analysis of introducing the algorithm throughout primary care, they said earlier identification of lung cancer will probably not only save lives but also reduce the necessity for aggressive and costly therapies.
Lung cancer remains one of the most prevalent cancers and, despite advancements in treatment, continues to have a poor prognosis. The five-year mortality rate still exceeds 80 per cent with around 48,000 new cases diagnosed in the UK each year. It is the UK’s biggest cancer killer, causing approximately 35,000 deaths annually, and it accounts for a large percentage of cancer deaths.