The Need for Smart Data
James Weatherall's predictions and why the marriage of data science/analytics and human judgment creates the best decisions.
James Weatherall works at the cutting edge of pharma’s Big Data debate. As Head of AstraZeneca’s Advanced Analytics Centre, he knows first-hand what the challenges are in analyzing millions of electronic health records and reams of clinical data on medicines. His team works hard at identifying patterns and trends which could be turned to commercial advantage for pharma development, using a blend of scientific and technical skills such as statistics, informatics and scientific computing to solve science and business problems.
“There are two key issues for the industry,” he begins. “First, we need ‘smart data’ not Big Data – the rudiments of analysis have not gone away. You need to answer questions that can be made into something that can be usefully applied in the business. You shouldn’t get carried away with the fact that there is more data – we need the right data in order to power good analytics and good business decisions.”
Lack of skills
That brings us on to his second point: that there are, at present, simply not enough skilled people to take advantage of Big Data – for instance by carrying out the appropriate modelling and simulation. “The world is hungry for people to turn it into business action because most of the world’s data lies untapped, with its full potential yet to be revealed,” Weatherall says.
McKinsey's report, "Big Data: The Next Frontier for Innovation, Competition & Productivity" estimated in 2011 that between 140,000 and 190,000 deep analytical positions, and 1.5 million more ‘data savvy managers’ were required to take full advantage of big data – and that is just in the US alone. Given this shortfall, how are things going to pan out? Weatherall has been thinking about this, and puts forward half a dozen predictions for where pharma, and other companies, will have to move to take advantage of a rapidly changing landscape.
More diversity is needed
Firstly, companies will increasingly remodel or expand their quantitative skills base to be ever more diverse – in particular, this means more data scientists, rather than pure statisticians, will be required to get the most out of sources which demand different approaches. “The data we can get our hands on is ever more diverse,” he says. “Twenty years ago there wasn’t the wealth of data [we have now] on the clinical development process, for instance.” Access to electronic health data, digitized information from publications and the ability to data mine large cancer registries or genomic databases means that a wide range of skills is required. “We need different domain knowledge,” he goes on. “Will it be sufficient to just be a pure-play clinical statistician or would you need a purview over other domains to bring those other data sources into play? Statistics alone may not be sufficient to get full value.” There will also be much more use of machine learning and artificial intelligence, he thinks.
Third party help required
Pharma companies will also sign more deals with data analytics companies, Weatherall believes, simply because they cannot meet all the challenges of Big Data on their own. “This is always going to be a headache,” he laughs. “Pharma companies don’t necessarily make good software developers or data analytics companies.” Business-focused professional data science outfits are beginning to emerge, working across very different business sectors – such as banking, high tech manufacturing and consumer goods – from which pharma can learn. “We will need to carefully pick and choose how those alliances work,” he goes on. “I can see companies specializing in data analytics for pharma, or certain parts of pharma R&D, coming to the fore.” Issues such as the difficulty of dealing with multiple data sources still need to be ironed out: many in pharma are sceptical about the value of trying to make sense of information which may come in different formats. “It’s our job to sort data into a sensible state,” Weatherall believes. “Some types of data, such as from clinical trials, adhere more to international standards – but where standardization is not there, it is up to the data generator to create order.”
The way we store data is changing, he continues, and it will increasingly live ‘in the cloud’, leading to different hardware infrastructures. “It seems like we’ve been talking about this forever, but it certainly won’t go away,” Weatherall says. One of the major perceived bugbears is security, but he is not convinced this concern holds water. “Your data is possibly safer with experienced cloud data hosts than it would be within your own walls,” he begins. “There are organizations which put a lot more time, effort and ingenuity into understanding how data can be safeguarded and encrypted.” Such organizations may begin to offer incentives, such as money back on a contract if there is a data breach within five years. Either way, the attraction of moving costs such as data storage, software hosting and processing power off the capital assets of a company’s balance sheet is clear - but data privacy and security policies may need to evolve in order to accommodate such changes. IT departments within pharma are certainly having to become experts in the intricacies of managing cloud-based contracted databases in what Weatherall calls “a continuation of what data and IT staff are already doing to add value”.
Getting more from patients
But it could allow us to develop medicines in areas where patients themselves are illustrating an unmet medical need. We must prepare ourselves for a world in which patients have a much greater say over what medicines are financed and approved in future – we’ve got to listen because what they say is going to count.
Weatherall’s fourth prediction for the future is that direct-from-patient data will increase exponentially as the technology boom in wearables, health sensors, smart medical devices, mobile software applications and web 2.0 gathers pace. “I’m not sure the world is ready to take full advantage of the data being generated,” he cautions. “But it could allow us to develop medicines in areas where patients themselves are illustrating an unmet medical need. We must prepare ourselves for a world in which patients have a much greater say over what medicines are financed and approved in future – we’ve got to listen because what they say is going to count.” Harnessing all this will be a mixed experience because of issues such as data quality and concerns over privacy and ethics, but it is undoubtedly something which pharma cannot ignore.
Data as differentiator
Weatherall also predicts that data integration will become a key differentiator for pharma, with those companies which can integrate successfully heterogeneous data sets finding themselves able to gain access to insights that provide a competitive edge. “Those who will win are the firms which can take advantage of different domains,” he explains. “People are beginning to crack the problem of merging these.” The real trick, he goes on, is to figure out what he sees as the grey area between human and machine. “There is a natural tendency to either greatly overestimate or greatly underestimate what a computer or algorithm can do in this space,” he suggests. “But you can start to open doors by asking questions linking the molecular world to the world of organisms, allowing you to make serendipitous discoveries which you couldn’t before.”
The power of subjectivity
Despite all the technological changes, subjective and qualitative decision making will continue to play a key role, thinks Weatherall. “A common misconception is that ‘computers will make decisions for us’,” he says. “I see research and rhetoric about A.I. having ‘gone one step too far’ in some cases, for example in creating completely automated decision making pipelines – and I’d be surprised if that were ever the right way to go. I tend to be healthily sceptical about what we can ever expect computers or algorithms to do. At the end of the day, you need a human being to make a decision and be accountable for it.”
However, he says, bringing the two together is the ideal: “The pure data-driven view should be debated but the qualitative piece has to be continued with – and the optimal result is when you knit the two together.” The marriage of data science/analytics and human judgment creates the best decisions. “One emerging way of forging such a ‘perfect union’ in pharma, is the ever growing implementation of Decision Science as a dedicated capability,” he adds. Weatherall points to successful predictors of recent US presidential election results, which employed both data and some element of subjectivity. “Some human factors were conspicuously taken into account,” he explains. “They had to be part of the mix and my hunch is that they can never be ignored.”
Areas for improvement
These six predictions point a clear direction when it comes to grappling with Big Data. So if Weatherall would fill in a report card for pharma, what would it say? “Could do better,” he says immediately. “But I suspect most companies would have to say the same. Some companies naturally have more opportunities to do these things, in the same way that those industries at the forefront in using data naturally have more opportunities.”
He cites Google, Amazon and Facebook as having obvious routes into exploring the possibilities of this milieu – routes which are not open to life science and healthcare groups. “Pharma doesn’t have that entry point at every interaction with its end users, often because it is either unethical or even illegal,” Weatherall concludes. “There is still a lot of scepticism from executives who quite like the idea but can’t quite see what’s in it for them. So pharma is behind and there are genuine reasons why. However, the challenge posed by the need to accelerate in this space, is what excites me and makes my job enjoyable on a daily basis.”
Since you're here...
... and value our content, you should sign-up to our newsletter. Sign up here