Tag: Digital

Five hurdles to digital health innovation in the UK|||
By Cambridge Design Partnership
November 30th 2023

Five hurdles to digital health innovation in the UK (and how to overcome them)

CDP recently led an investigation into how to advance innovation in digital health in the UK for the CPI, UKRI/Innovate UK and ABHI. Our aim was to find out how best to enable the UK to be the place of choice for enabling high-risk digital health innovation, improving patient outcomes. 

Our work with 50 leading healthcare professionals and entrepreneurs revealed that the UK has an enviable record in early-stage innovation, a highly regarded healthcare system and a potential treasure trove of high-quality data. 

However, we also found several hurdles that trip up many innovations before their potential can be truly realized. 

In this article, we describe our top five hurdles to success and signpost the resources available to help innovators overcome them. 

1. Offering the wrong product at the wrong price

Let’s start with perhaps the most obvious: you need to get your offering right. This was one of the more frequent topics to emerge in our discussions. True, it tended to come from the industry and investors, rather than entrepreneurs themselves. But perhaps this is the point; those closest to the concept are so captivated by the opportunity to solve a problem that they are rarely the best judge of commercial success.

“The biggest problem is developing stuff we don’t need, at the wrong price point.”

Life Sciences Lead, multinational consultancy

“People have struggled with finding the right balance between fixing a problem not just for the sake of it because it’s going to add value, but also there is a market attached to it.”

Medical Director, AI dermatology revenue-earning startup

Getting the right product at the right price is not easy. Regardless of how it is funded, healthcare everywhere is a complex system of separate entities with conflicting priorities. One of the biggest challenges for digital health offerings in particular is that the person paying the bills is rarely the direct beneficiary. This is as true across the NHS in the UK as it is in insurance-led services in the US. 

Digital interventions are regularly shown to make significant positive impacts on diagnosis, therapy, adherence and behavior change. To date, the FDA has approved, authorised or cleared 171 AI/ML-enabled medical devices. However, digital means adding overheads (electronics, batteries, software or new digital services) to an already overstretched budget that tends to bring value much further down the pathway. 

To get the right product at the right price, you need to be crystal clear about the value you bring and who you bring it to so that you can ensure the price is right. 

Helpful Resources

For the UK market, we found the following resources helpful in crossing this important hurdle:

  • The NHS Innovation Service provides an innovation guide that explains how to build a value proposition
  • The NICE Advice Service provides personalized advice on the value propositions for a fee 
  • The NHS Clinical Entrepreneur Programme (CEP), launched in 2016, provides training for NHS staff on the skills required to build a healthcare startup, all without them needing to leave the NHS

Indeed, this is such an important area that we at CDP are looking at how recent advances in Generative AI might make this easier to get right from the outset – not just for offering services within the UK, but how UK-based innovation can provide the right offering in the larger markets of the US and EU.

2. Neglecting the needs of key stakeholders

Digital products and services are still a novelty in healthcare. Even the regulation is taken from a device mindset – consider the terms SaMD (Software as a Medical Device) and now even AIaMD (Artificial Intelligence as a Medical Device). The digital-first mindset is to move fast, learn and repeat to get the best user insight and optimum benefit to market as fast as possible. This is not an easy marriage for healthcare, where verification and validation are critical steps to approval.

“If you are manufacturing a digital health product, you have three sets of policies to navigate right now [in the UK].”

CEO, digital health SME (referring to NHS DTAC, NICE and MHRA)

“The regs are written to cover all medical devices. They’re not very specific; it’s very high level and quite hard to interpret what we should actually be doing as an individual company.”

Medical Director, AI dermatology revenue-earning startup

This is not simply about the regulator; it is also about who will receive, who will administer and who will pay for your digital offering. On top of proving safety and efficacy, payers and adopters want to see evidence that your technology works under real-world conditions and produces sufficient benefit relative to current clinical practice to justify its cost. Not only do you need to convince your investor you have the right product at the right price point; you also need to convince them you have access to reimbursement. 

This need has led us here at CDP to build a strategy and insight team that explicitly looks across the spectrum of stakeholders including the end-user, practitioner and payer.

Helpful Resources

The following resources are helpful when considering the regulatory and UK purchaser stakeholders:

  • The NHS’ AI and Digital Regulations service offers a developer’s guide that leads you through the various regulatory and NHS requirements for digital technologies 
  • The NICE Evidence Standards Framework is designed to help ensure NHS stakeholders are adopting robust technologies that are likely to provide the expected performance, and are good value for money. The framework can be used by developers to understand their customer needs. NICE also offers an assessment of current/planned evidence via their META tool
  • The NICE Early Value Assessment can also help indicate the value your product can bring, and allow you to get support to understand what further evidence needs to be generated.
  • Similarly, NHS’ Digital Technology Assessment Criteria (DTAC) are designed to assess suppliers at the point of procurement, or as part of a due diligence process, ensuring digital technologies meet minimum baseline standards. The criteria can also be used by developers to understand what is expected for entry into the NHS and social care
  • FDA’s list of approved, authorized or cleared AI/ML-enabled medical devices

3. Testing, verifying and validating

The regulatory pathway will force you to verify and validate. It will be rigorous. It will take more time than you or your investors want. So, you will need to test, test and test again as early as possible to build the evidence you need for investment. And, importantly, test both the medical efficacy of your offering and its likely commercial success.

“[It] can take longer than six months, ridiculously, to build a cohort of data.
Getting people to step away from frontline service in the NHS is a fundamental challenge of getting access to that data. Even if you offered to pay, they’d say” ‘I don’t care; it’s not the money I’m short of, it’s people’.”

President, medical imaging multinational

“We really struggle to work with SMEs because we’re not able to move at the pace that they require for their cash flow.”

Director of Innovation, NHS trust

As these sentiments show, however, gathering data takes time and patience. The NHS is indeed a treasure trove of data, but unlocking it is a real hurdle. Existing NHS data typically needs preparation – cleaning and anonymizing – before you can access it. And there simply may not be the staff available to do this, meaning you may need to build additional paths to gather test data. 

The fidelity of the test can start low-fi, but will need to increase as you develop. CDP typically starts with insights research and human factors studies using UI sketches/descriptions of the product to explore the true user journey, before moving onto trials with real-life samples and wizard-of-Oz demonstrators. This builds a body of evidence that reinforces your expectation of efficacy with the all-important usability and the commercial viability of your offering, before embarking on the summative human factors, clinical and market trials.

Helpful Resources

The following resources can help you prepare your plans for testing, verification and validation:

  • NIHR study support service provides guidance and advice
  • The HDR UK Gateway portal helps researchers find existing data sets and connects them to relevant stakeholders
  • Trusted Research Environments (TREs) are a new initiative to facilitate access to NHS data for R&D. Only a few TREs currently exist and there is no guarantee they will have the data you’re looking for. However, the teams involved are well placed to advise you on next steps. Even though this requires approval from HRA and notification to MHRA, consider if it might be better to just do your own trial to collect fresh data instead. This is where an experienced external innovation partner can be very helpful

4. Navigating healthcare as a ‘system of systems’

The benefits of digital health typically require systems integration. Yet, healthcare everywhere is a complex system of systems, each element with its own approaches, tools and requirements.  

“[Different hospitals are] probably using different systems, different levels of maturity with different versions, with different level plugins. That probably means, even if I create it using the standard, it won’t automatically fit. It needs modification, adaptation and someone to do the translation.”

Digital Health Advisor, ex-NHSX

“You can often have very inflexible contracts with your electronic health records supplier; for example, if you want them to make one change or open up in an API or something like that, it can be prohibitively expensive.”

Director of Innovation, NHS trust

While there are only a few dominant providers of Electronic Health Record (EHR) systems, each installation is likely to be different. Moreover, the EHR providers will guard access jealously. Microsoft, Google, Amazon and others provide integration services to structure and translate data, but that is likely to be only a small part of the problem, and only useful if you are ingesting unstructured data from multiple sources. 

At CDP, we encourage our clients to focus on providing easy-to-use, yet secure APIs built around well-structured data that map well to the established data standards such as FHIR. Taking ownership of your own data in this way makes it easier to deploy, integrate and support. 

Helpful Resources

The following resources can help you prepare your digital health services for system integration:

5. Building a strong team

Innovation is rarely one guy in a garage. This is especially true in the digital health space. You will need to build a great team led by experienced professionals across the disciplines. Get this right and everything else will fall into place. Work out your strengths and weaknesses and actively seek resources to complement your team. 

“The most useful thing to an innovator is access to an actual practicing frontline clinician who understands the problem that they want solved. It’s a real challenge to get to these people. I might spend months trying to find someone who would talk to me.”

President, medical imaging multinational

“The UK does not have enough engineering capacity… does not have enough people with product skills… does not enough people with this sort of legal regulatory skills.”

Digital Health Advisor, ex-NHSX

Helpful Resources

The following resources can help innovators looking to build a world class team:

  • Many Health Innovation Networks (HINs) and NHS trusts have innovation teams who may be able to help matchmake with clinical champions. The NHS Innovation Service is a good place to start, but it’s worth seeing what individual trusts are doing. 
  • HINs and NHS trusts often also support innovation and hackathon events which are a great way to find those with a similar innovative mindset – but a complementary skill set.   
  • And then there are organizations such as CDP, who bring end-to-end product development services with the hard-won experiences of how to navigate this exciting but often frustrating area of innovation.

In addition, many of your digital needs are engineering and operational ones. Recruiting experienced people from the finance and technology sectors where the UK is strong will bring you good skills and expertise in algorithm development, handling personal data and building scalable secure systems.

It’s tempting for you (and your investors) to under-resource your team and compromise in the early phases. But as the hurdles above clearly show, this rarely leads to success. Build the great team you need from the outset, to make sure you truly have the right product at the right price, that meets the expectations of the key stakeholders, is properly tested and ready to integrate into the healthcare system of systems.

At CDP, we continue to follow up our insights working with clients and partners to find practical solutions to complex problems. To find out more about what successful innovation in digital health looks like, please do get in touch. 

In the meantime, download the full action plan for digital health innovation in the UK here.

How data and AI are changing bioprocessing
By Cambridge Design Partnership
October 30th 2023

How data and AI are changing bioprocessing – and why it’s needed

After numerous insightful talks and engaging conversations with industry leaders at this year’s BioProcess International, the key theme was clear: data, data and more data. 

Data has always been important, but now it is being collected to model current processes, understand how they work, and improve them. This is a trend that is only likely to accelerate in the future as AI becomes part of everyday life – both in and outside of work

Using data-based modeling to optimize well-established industrial processes 

There are many traditional processes that are used in the manufacture of antibodies, mRNA vaccines and cellular therapies. Companies are now collecting extensive data from these processes and using modeling to create their ‘digital twin’. 

The processes modeled range from relatively simple tasks such as optimization of freezing/thawing product intermediates, freeze-drying and automated buffer preparation, to more complex procedures such as bioreactor scale-up. Although these used to be manual ‘craft’ processes run by a combination of experience and pre-existing data, there is now a trend for them to be tested and optimized using in silico methods.

Using modeling to improve purification methods

Bioprocessing is used to create many therapeutic products, from molecules such as protein, DNA and RNA to much larger entities such as viruses and eukaryotic cells. Their production has many different steps that often require extensive purification before the next step can proceed. Common purification methods include clarification, chromatography, ultrafiltration/diafiltration and sterile filtration. 

These methods were typically used in an empirical way based on experience with similar products. Now however, use of modeling has led to a much more detailed understanding of how these separation/purification methods work. It allows the prediction of when column/membrane capacity is reached, and when “breakthrough” of contaminants is likely to occur. It has also led to the development of alternatives to standard resin-based column chromatography such as the incorporation of new reactive chemical groups on membrane filters that can then act like traditional resin-based columns.

Benefits of Process Analytical Technology (PAT)

PAT refers to on-line/at-line measurement of critical product quality and performance attributes so that real-time direct data collection can be used to control and optimize manufacturing processes.

PAT is being augmented by a much wider range of analytical techniques than before and now includes many different types of spectroscopy including variable path length, Fourier-transform infrared, Raman and Dynamic Light Scattering, as well as Nuclear Magnetic Resonance. The use of PAT for direct data collection that links to immediate process control is only likely to accelerate.

Inexorable rise of disposable closed cell processing systems

In addition to the data theme, it was clear to see that the number of automated closed cell handling and processing systems – from cell selection to expansion and harvesting – is rapidly increasing. Companies aim to offer end-to-end solutions to traditionally manual processes, either by offering modular components or a single complete system. 

The options for choosing automated disposable bioreactors/cell expansion systems are also increasing, with many players recently entering the market. It is clear why this option is advantageous; traditional stainless-steel bioreactors are complex, expensive, and laborious to clean and maintain.

Just how large these systems can grow is shown by ThermoFisher’s 5000L disposable Dynadrive bioreactor, which is offered as a fast-to-install option compared to stainless-steel alternatives. However, the environmental impact of the disposable route is a long-term concern and is expected to be a point of contentious discussion over the coming years.

Bioprocessing technology is developing (but not fast enough for demand) 

The technological developments described above are certainly needed as advances in eukaryotic culturing methods are allowing higher and higher cell densities to be realized, which makes purification more challenging. Furthermore, the pipeline for products that use these technologies is growing at a dizzying rate with over 1,500 cell and gene therapy and 700 mRNA trials listed on the US Clinical Trials site. New higher throughput processing techniques will need to be developed to accommodate this demand. 

The industry clearly recognizes this and companies were very open in sharing their results at BioProcess International – both good and bad! They are also keen to work with the process equipment manufacturers to optimize performance. Overall, improvements have been made, but there is a long way to go. 

Performance can be improved by a virtuous circle of data generation, data modeling and innovative design and engineering – something we at CDP are already doing to help our clients succeed.

If you would like to see how innovative engineering and automation could help increase bioprocessing throughput, please get in touch.

Insights into GenAI data scientist perspective - Cambridge Design Partnership|The data scientist’s perspective|Insights into GenAI product owner's perspective - Cambridge Design Partnership
By Cambridge Design Partnership
October 18th 2023

The product owner’s perspective: Five practical insights to accelerating innovation with GenAI

Here at CDP, we’ve delivered a range of Generative AI (GenAI) projects that use Large Language Models (LLMs). Each has been a journey of discovery, and sometimes frustration. But ultimately each has reinforced the potential for GenAI to dramatically accelerate innovation.

In an attempt to provide a useful contribution that cuts through the noise, we’ve distilled our learnings into a four-part series on how businesses, data scientists and product owners can leverage GenAI for success with a final perspective from a GenAI-powered ChatBot.

In this second article, we draw from our experiences implementing GenAI from a product owner’s perspective. To get a high-level view of LLMs check out Part 1 or for a deeper dive in the technology from a data scientist perspective check out Part 3.

1. Start at the end and work backwards

As with all truly transformative innovation, start by understanding what you are offering your users and work back from them. Ignore the undoubted magic of the technology at this stage – you can rely on that coming later.

You will need to set your success criteria, and this is where to start. Delighting your user base and measuring how they will benefit will do more to drive adoption than any shiny AI tech that might be going on behind the scenes.


Choose your project carefully.

  • Choose an area that you already know well or for which you have a good way of measuring success. This will ensure you see beyond the magic of the black-box and can truly judge the performance and value that LLMs bring.
  • Choose an area where LLMs work to their strengths by taking advantage of at least one of the core competencies they have been shown to do well; summary, expansion, inference and analysis.

2. Don’t forget the basics

Make good use of Service Design techniques to define what success looks like. Map the User Journey and spend time defining the touchpoints and modelling the semantic information architecture.

And then strip it back. Cut away absolutely everything that isn’t vital to the successful outcome you plan for. Don’t let the designers loose until this is done. And consider any investigative work with the technology up to this point as exploratory and should almost certainly be archived.

You’ll then have a clear set of priorities, requirements, information flows and use-cases that everyone understands, and everyone can support. The whole team will be clear about what they are aiming for. Keeping their eye focussed on the prize makes the Product Owner’s primary catch-phrases more effective: “No that is not in scope” and “This is lower priority”.

And if this is starting to sound like the start of any solid digital project – good, it should.

3. Experiment

Give your team as much time as possible to try things out. Build the time into the plan and break the experiments down into small and well-defined steps to learn and iterate.


Look to experiment with the following:

  • How the structure of prompts changes the output.
  • How the different LLMs compare when asked to respond to the same prompt.
  • How to extend the LLMs by adding training to embed your own information data.

Aim to build the experimental steps around the core competencies of Generative AI. And later, bring these together to form an overall solution using your favourite AI automation tool chain.


There will be surprises. There will be frustrations. And there will be changes in the way that you approach the use of the LLMs. Don’t be afraid to pivot on how you use the technology; or indeed ‘if’ you use the technology. But remember the basics, keep your eye on what success looks like and don’t let the team get carried away with ‘shiny object syndrome’.

4. Get lots of feedback

While using AI, remember to share your work with real humans as early as possible: People outside your team who can give you useful feedback. Set up demos within the team to share learnings and put on regular show-and-tell sessions with your target audience. And, as soon as possible, let them try it out – on their own, without you there. They will learn to see beyond the magic, and you will quickly find out what works and what doesn’t.

Your priorities will change – but the fundamental definition of success won’t (hopefully). And don’t forget the importance of plain old testing. The outputs from a LLMs can vary widely with only the smallest changes in training data and prompts. Fortunately, LLMs can come to the rescue here – they are great at evaluating the output from other models through peer- review. Use that capability to help you test. This is also useful for building into the architecture of your solution. Where you have the resources; double up the LLMs to interact and increase the quality of output for a production system.

5. Don’t underestimate the time you need


Don’t underestimate the time it will take to gather, prepare and refine your data. When it comes to data, quality and variety are just as important as quantity. With demographic information, a good distribution of variety is vital to represent your users truly and ethically. And don’t forget to set aside at least 10% randomly selected from the training set so that you can properly test the results.


To save time and increase the training and test data available, explore opportunities to synthesise data to add to your original data set. Also, don’t underestimate the time it will take to test and refine the prompts and LLMs settings to achieve the repeatable outcomes you are looking for. Prompt engineering is an art as well as a skill and takes time to learn.


Finally, know when to stop. It will always be possible to make it a bit better. Be clear about what is good enough and recognise when you get there. The impulse for the team to keep tweaking will never end – it’s simply too absorbing.

Interested in exploring how GenAI can accelerate your innovation?

Come and join us in Cambridge, UK, and Raleigh, NC, where we’ll be running a series of in-person workshops to help clients identify the opportunities (and threats) of GenAI and plan a path to accelerate their innovation.