Cost-effectiveness of immediate prioritisation of chest X-rays using artificial intelligence in lung cancer in England: the LungIMPACT study
This health economic analysis of the LungIMPACT study concludes that using AI to prioritize chest X-rays for lung cancer in England is not cost-effective compared to AI-assisted interpretation alone or current practice, as it increases costs without reducing time to diagnosis.
Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer
Imagine the human body as a bustling city, and inside that city, a silent, sneaky troublemaker is trying to set up shop: lung cancer. This troublemaker is tricky because, in the early days, it doesn't make a lot of noise. It doesn't scream for attention; it just whispers, maybe causing a little cough or a bit of tiredness that people often ignore. Because it's so quiet, it often gets caught only after it has grown big and strong, making it much harder to defeat.
To catch this troublemaker early, doctors use a special "X-ray camera" called a Chest X-ray (CXR). It's like taking a quick snapshot of the city's skyline to see if there are any dark clouds forming. Every year in England, General Practitioners (GPs) send millions of these snapshots to the radiology department. But here's the problem: the radiology department is like a massive, overflowing mailroom. Thousands of photos arrive every day, and the doctors who read them (radiologists) have to sort through the pile one by one. If a dangerous cloud is hiding in the middle of the stack, it might get missed or delayed while the doctors look at the harmless ones first.
Recently, scientists invented a digital helper: Artificial Intelligence (AI). Think of this AI as a super-fast robot assistant that can scan every photo in a split second. Its job is to spot the "dark clouds" (suspicious spots) and shout, "Hey! Look at this one first!" This is called "prioritisation." The big question everyone wanted to answer was: If we let this robot assistant shout and move the scary photos to the front of the line, will it actually help catch the troublemaker faster without costing the health service a fortune? This is the story of the LungIMPACT study, which tried to find out if this high-tech sorting hat was actually worth the price tag.
The Great Sorting Experiment
In England, a massive experiment called the LungIMPACT study was launched to test this very idea. The researchers gathered data from nearly 87,000 patients who had their chest X-rays taken after visiting their GP. They split these patients into two groups to see what happened.
In the first group, the AI was turned on to do its "shouting" job. It scanned the X-rays and flagged the ones that looked suspicious, telling the human doctors, "Prioritise these! Read them immediately!" In the second group, the AI was also used to help read the images, but it was told to stay quiet. It didn't shout or move photos to the front; the doctors just read the X-rays in the order they arrived, with the AI's help as a second opinion.
The goal was simple: see if the "shouting" group got their diagnoses faster. If the AI could cut the time from taking the X-ray to finding out if someone had lung cancer, it would be a huge win for patients. But there was a catch. The researchers also had to count the money. They looked at every penny spent: the cost of the X-rays, the CT scans, the extra tests, and even the fee to rent the AI software and the cost of plugging it into the hospital's computer systems.
The Surprising Result: The Robot Didn't Save the Day
After crunching the numbers, the study found a result that might surprise you. The "shouting" AI did not significantly speed up the diagnosis.
Here is the breakdown of what happened:
- Time: The time it took to get a diagnosis was almost the same in both groups. In the group where the AI shouted to prioritise the photos, the average time to a diagnosis was about 75 days. In the group where the AI just helped read without shouting, it was about 77 days. That tiny difference wasn't enough to say the shouting AI actually made a real difference.
- Money: Because the AI software costs money to license and costs money to install into the hospital computers, the "shouting" group actually cost the health service more money. The study calculated that using the prioritisation feature added about £1 per patient compared to just using the AI without the shouting feature.
When the researchers looked at the big picture for the whole of England, the math got even starker. If every GP in England (who orders about 2.2 million X-rays a year) used the AI to prioritise X-rays, it would cost the NHS an extra £2.2 million per year just to get that tiny, almost non-existent time saving. If they compared the AI method to the old way of doing things (no AI at all), the cost would jump to between £16 million and £35 million per year.
Why Didn't It Work?
The researchers suggest that the problem wasn't the AI itself, but where it was trying to help. Imagine a highway where traffic is jammed. If you have a helicopter that can spot the slowest cars and tell them to move to the fast lane, that's great. But if the whole highway is gridlocked because there aren't enough lanes or enough police to manage the flow, moving one car to the front doesn't clear the traffic jam.
In the case of lung cancer diagnosis, the "traffic jam" happens after the X-ray. Even if the AI moves the scary X-ray to the front of the line, the patient still has to wait for a CT scan, then wait for a specialist to look at the results, then wait for a team meeting to decide on treatment. The study found that the AI couldn't fix the bottlenecks that happened after the X-ray was read. The system was already so busy that moving the X-ray to the front didn't make the whole process faster.
The Bottom Line
The study concludes that, at least for now, using AI to immediately prioritise chest X-rays is not a good use of money for the health service. It costs more than it saves, and it doesn't actually get patients diagnosed any faster.
The authors are very clear about this: they aren't saying AI is useless forever. They are saying that this specific way of using it—shouting to move X-rays to the front of the line—doesn't work well enough to justify the cost. Instead of spending millions on this specific tool, the study suggests that the health service might get better results by fixing the other parts of the process, like hiring more staff or making sure there are enough CT scanners and doctors to handle the patients once they are flagged.
In short, the robot assistant tried to be a hero by shouting "Look at me first!" but the rest of the team was too busy to listen. Until the whole team can move faster together, just shouting at the front of the line won't solve the problem.
Drowning in papers in your field?
Get daily digests of the most novel papers matching your research keywords — with technical summaries, in your language.