Infectious Diseases

Pfizer pays almost $120 million for app that detects COVID from a cough

Early trials found ResApp's technology could detect 92% of COVID cases solely from the sound of a cough
Early trials found ResApp's technology could detect 92% of COVID cases solely from the sound of a cough

Pharma giant Pfizer has shelled out nearly US$120 million to acquire a small Australian company claiming to have developed a smartphone app that can accurately diagnose COVID-19 by analyzing the sound of a cough.

For around a decade small Australian digital healthcare company ResApp has been working on developing an algorithm that can diagnose respiratory illnesses by simply studying the sound of a patient’s cough. Initially the system was trained to diagnose pneumonia, but by 2019 the researchers had shown the technology could effectively distinguish asthma, croup and bronchiolitis.

When the pandemic struck in 2020 the team unsurprisingly quickly pivoted to incorporate COVID-19 diagnoses into its cough-recognition technology. By early 2022 the first data from a pilot trial testing the COVID algorithm revealed impressively good results.

The trial found the system could accurately detect 92% of positive COVID cases solely from the sound of a cough. The system also recorded 80% specificity, meaning only two out of every 10 people screened received false positive results.

Soon after ResApp revealed these results pharma giant Pfizer began circling, initially offering around $65 million for the technology. Now, in a formal acquisition announcement, a deal has been finalized for Pfizer to buy ResApp for a massive $116 million.

In a statement, a Pfizer spokesperson said the preliminary data was encouraging and the deal expands the company's footprint into the sphere of digital health.

"We believe the COVID-19 screening tool is the next step to potentially provide new solutions for consumers that aim to quell this disease," the spokesperson said to ABC news. "We look forward to refining this algorithm further and working with regulators around the world to bring this important product to consumers as quickly as possible."

The ResApp team hopes the acquisition by Pfizer helps the technology grow and be widely deployed in remote parts of the world. Udantha Abeyratne, one of the original developers of the algorithm, said the goal of the project was to help bring better diagnostic tools to communities around the globe.

"From the very beginning, I had a big vision to develop scalable, cheap technologies to diagnose pulmonary diseases all over the world – not only in remote sub-Saharan Africa, but even in developed urban cities like New York and Brisbane," said Abeyratne. "I hope they will be able to diagnose killer diseases like pneumonia in very remote communities in Africa and Asia because they don't have access to sophisticated hospitals."

Source: UQ

  • Facebook
  • Twitter
  • Flipboard
  • LinkedIn
9 comments
Brian M
Can understand how this works, but maybe the price paid is a little on the high side unless there is something very clever there (and other assets), just seems like a case of applying an AI neural network to the problem and applying a lot of training data.

Wouldn't have thought you could get a patent for it either, as listening to coughs is as old as medicine itself and neural network applications are a known technology
i.e Something a person knowledgeable in the field could do!
paul314
I wonder whether the price paid is about Covid, or all the other diseases that can potentially be diagnosed. And the target market might not be so much individual consumers as clinics and medical offices -- if you can do the cough test with less-highly-trained and highly-paid people, that's a couple of minutes per patient, which adds up to a nice sum over the course of a year.
Winterbiker
One potential place for this app is in public screening at airports and transit terminals, major event entry points, there are a lot of public places where this could be useful. It could also be used at the entrance to your office building. Since you can detect other diseases as well maybe you send your workers home if they test positive for the common cold.. I think they will be making a lot more than 120 million in return for their investment.
see3d
@Brian M, Solving a problem in a way that has not been done before, even with old tech is patentable. In fact, some of the best patents are obvious -- after the fact. It is the old "Why did I not think of that" head slap type of invention. They cleverly combined different known ideas and devices to solve this problem in a way that had not been done before. My patents were conceived the same way, as were a host of others.
NMBill
Sounds fun. I get a little tickle in my throat at a restaurant and have 15 people sticking their phones in my face. No thanks.
Username
20% false positives seems completely unacceptable.
Jinpa
Pfizer wouldn't have paid $116M for this if they hadn't calculated how to make multiples of it by leasing it. Do a followup to learn with the charge will be to medical offices and hospitals.
DaveWesely
We really need to ask, "Why would Pfizer, a medical supply company, want to buy a software startup?". Especially if that startup's software could reduce the need for conventional medical products provided by Pfizer.
Pfizer could:
A Sell the app to reduce demand (and profit) for testing kits.
B Reduce availability of app with higher cost and restricted access. This maintains profits from both the app and conventional testing kits.
I sense a conflict of interest or a subversion of the competition needed for a healthy market.
Aladdin Connolly
I agree with "username" 20% false positives is ridiculous. But it actually seems much worse than that. If it detects 92 percent of covid cases and yet gives false positives to 20% of people. It could simply be that the majority of people given this test were positive for covid, so they likely had other symptoms that made the company chose them for the test. Twenty percent of those subjects didn't have covid, but were told they did.
It sounds like some real playing with statistics, and not random double blind testing. raw data would be far more clear.