Healthcare and AI are a common match.

Often we hear about software to help doctors with diagnostic tasks like finding anomalies on scans, but today’s startup wants help them take action before there’s even any cause to take a scan in the first place.

Read on to learn all about Sarvas Health.

– Martin

Here’s Why Over 4 Million Professionals Read Morning Brew

• Business news explained in plain English

• Straight facts, zero fluff, & plenty of puns

• 100% free

See for yourself

Sarvas Health wants to tackle chronic disease before it even starts to look like a problem

Much has been made of AI’s ability to help analyse medical tests and scans, but surely it would be even better if it spotted potential health problems well before they became problems?

Looking to tackle preventative healthcare with AI is Sarvas Health.

“What we've set out to do is develop AI tools that can predict the risk of onset of various chronic diseases within the primary care environment and within the community.,” explains co-founder Martin Ashby.

“There's more positivity to outcomes, both in terms of lives saved and longevity, and it saves the NHS the hospitalisation costs for diseases that could otherwise have been found earlier resulted in cost reductions and better outcomes.”

To do this, the Manchester-based startup has developed a way to use AI to predict chronic diseases from electronic health records.

“We've employed novel architectures, such as transformer models, so they learn the language of health records… We can then teach it what the pattern looks like in a heart failure, or diabetes, or COPD, whatever it is that we want to predict.”

While other models exist to spot one disease, Sarvas Health wants to be a one-stop-shop to identify multiple health conditions.

“With our model, what we'll find is as we train it on more disease types, it will inevitably become more accurate on the disease types it's already been trained on. And more traditional models can't do that. They can only train it on one disease time,” Ashby says.

But how accurate can this approach be? It doesn’t have to be bang-on perfect every time to be useful, but it certainly needs to meet a threshold of accuracy.

Ashby says the startup’s initial proof-of-concept, focusing on heart failure, has been trained on data from around 200,000 patients. He says it’s achieving more than 90% accuracy. That sounds impressive, but Ashby’s fellow Sarvas Health co-founder, Clare Simpson, adds that far more data is required:

“We need to get more data with more breadth, because [the current data is] from one data set, so there's not a lot of variability. The 90% accuracy comes from a technical validation. We're in the stage now where we're talking to some large companies about supporting us with a clinical trial. It'll be a small scale clinical trial, just to prove that it works, before then we go out and prove it on a larger scale.”

Simpson says the tech is intended to be used by GP surgeries who could analyse their patient lists to find people at high risk of certain conditions, and then invite those people in for tests.

According to the British Heart Foundation, there are around 200,000 new diagnoses of heart failure in the UK each year. And up to 80% of heart failure diagnoses in England are made in hospital, despite 40% of patients having symptoms that should have triggered an earlier assessment

Ashby envisions being able to scan an entire adult cohort of a GP’s patient data in 10 minutes, helping to get the most at-risk patients on a care pathway long before they would otherwise end up in hospital.

“Their lives are longer, they live better-quality lives, but ultimately, it saves the NHS a lot of money as well,” Simpson adds.

The story so far

Sarvas Health isn’t the first time Ashby and Simpson have worked together. They previously both worked at two companies in the wearable healthtech space.

At the most recent of these, Prevayl, they began to look at AI for healthcare, and heart health in particular. After the company shut down operations last year (it still exists as a licensing house for its patents), the pair were inspired to continue in that direction.

“We had a real desire to pursue the electronic health record review, specifically around heart failure, on the basis of the feedback we got from the consultant cardiologist. Of the people we'd spoken to, none had told us we were stupid, so it just motivated us to work as a small team to develop a proof-of-concept for heart failure,” Ashby explains.

The founding team numbers five, with three of them working full-time on the startup.

They’re currently going through the process of getting approved for access to NHS data to further train their model on

“Data in the UK is really hard to come by. So one of our frustrations is we see the government agenda, we see them talking about the fact they want people to work in primary care in the community, but they make it almost impossible for you to get access to data to train your model. And as a startup, that's tough,” Ashby says.

To complement the UK data they will be able to access, Sarvas Health is looking to access training data from overseas, too.

“We don't want a lack of UK data to stop us from developing something that could, we believe, save countless lives,” Ashby adds.

Subscribe to Premium to read the rest.