Now, Medrano says, the industry finally has the technology to analyse the large-scale patterns arising from these variables.
But not all forms of RWE are created equal, he says, and in fact current methods of collecting RWE data don’t always produce insights that are useful to researchers or HTA bodies.
“One problem is that most RWE is contained in inflexible, siloed databases. Companies have to pay to access these databases, but they can’t always be sure that the exact data they need will be found within.
“That’s a pity, because when the information isn’t there, it’s usually not because it doesn’t exist – in fact it’s being generated every single day by hospitals and electronic health records (EHRs). It’s just not in that particular database.”
Luckily, the sector is evolving to a point where life science companies can use AI tools to directly analyse electronic health records and extract the exact information they need for their particular market access purpose.
This is the focus of Savana’s own technology, which uses natural language processing (NLP) to extract and read unstructured information buried within the free text of clinical notes and EHRs.
The technology and clinical research methodology can therefore enable clinicians and health researchers to analyse vast amounts of previously inaccessible clinical data.
The idea came from Medrano’s personal frustrations while working as a neurology consultant.
“I got weary of entering data in the electronic health records and not getting anything in return. There was a lot of valuable information that was lost to me.”
Now, computational linguistics is making this information readily available to pharma companies and regulators – and the benefits could go beyond ease of access.
“One of the problems with value-based contracting at the moment is that gathering RWE on how well a drug is working often involves doctors manually populating databases with each outcome case, usually at the end of the working day,” says Medrano.
“Why don’t we automate that? Why don’t we apply natural language processing to extract the outcomes automatically, allowing us to adjust the reimbursement with the support of AI? That’s now completely doable, and in fact we’ve already begun working on that in North America and some Eastern countries.”
On top of this, while Medrano says traditional databases can be “incredibly limited” for observational studies across wide populations, AI tech is allowing this kind of research to thrive in new ways.
“Observational studies can be extremely expensive and very slow. Now with this technology, we are able to quickly connect massive populations across primary care, specialised care, inpatient, outpatient, pathology – every single time the patient touches a point of care, we can capture that. For the first time, we can gather all those variables – tens of thousands of them – and connect them to understand what is happening.
“That can be extremely useful for market access purposes.”
Predictive modelling based on machine learning will also be a huge boon to RWE collection – and a useful tool for HTA bodies.
“That can be really important for positioning in a pathology, by understanding the risk of an outcome not based on classical statistics, but on neural networks that have already been used by other industries for years and can now be applied in medicine.
“That might even include ‘deep screening’ to find patients that are undiagnosed.”
“There’s really not a technological or methodological reason not to trust these RWE databases for making decisions. We have evaluation methodologies where we can demonstrate how reliable the databases are, and with random sampling we can check how good the information extraction is. It’s very transparent.
“The only progress that needs to be made is from a purely human, regulatory perspective – but we are already proving how useful and high-quality the data can be, which is important for being able to move forward in this regard.”
Medrano adds that regulatory offices are generally “open, positive and optimistic” with regards to RWE and future AI applications.
“Of course, they still have to be cautious – that’s their job, after all – and companies working in this space will have to continue to prove their reliability. There have been a good number of poorly designed machine learning-based algorithms, and we see many low-level publications that don’t follow the standard guidelines for scientific publications. That’s always something we have to be aware of, and make sure we’re not using these technologies to somehow bypass mandatory regulatory steps.
“But regulators are certainly interested in the potential of these new methods and collaboration is starting to bloom. The situation is very clearly moving in this direction.”
Dr Ignacio H Medrano, CMO and founder, Savana is a consultant neurologist with training in healthcare management and experience in clinical research strategies – formerly responsible for over 500 researchers. A Singularity University graduate, he is also a founder at Mendelian in the UK which is utilising AI in the diagnosis of rare diseases. Ignacio is in demand as an international speaker at digital health, clinical research, science and technology events and congresses.
Founded in 2014, Savana is an international medical company that has developed a scientific methodology that applies artificial intelligence (AI) to unlock all the clinical value embedded within the free text of electronic health records (EHRs). With the largest AI-enabled, multi-language, multi-centre research network in the world, Savana generates customised descriptive and predictive, Deep Real World Evidence research studies. Engineered by doctors for doctors, Savana is built following the highest privacy-by-design standards. Savana constitutes a clinical research ecosystem that aims to advance personalised and precision medicine worldwide.
George Underwood is the editor for pharmaphorum’s Deep Dive digital magazine. He has been reporting on the pharma industry for seven years and has worked at a number of leading publications in the UK.