With the end goal of reimaging global drug development, Novartis several years ago positioned themselves to “Go Big on Data” and fundamentally change how they operate. Writing in the journal American Society for Clinical Pharmacology and Therapeutics, CEO Vas Narasimhan, and VP Luca Finelli, discuss the evidenced benefits and remaining challenges of Nerve Live, their purpose-built analytics platform for insight-driven decision making.
The authors begin by contemplating the individual elements behind our ever-more digital world and workplaces. An acceleration in the rate at which data is being produced has been supported by increasingly powerful cloud computing. Whilst artificial intelligence and machine learning have been hugely important, crucial to their success, has been the will of people to accept these technologies into their everyday lives.
Novartis several years ago analysed exactly where data and analytics could improve their processes in a transformative manner, by cutting onerous manual tasks and lowering costs. In this review, Narasimhan and Finelli focus on the potential that their operational data holds for the company. As opposed to patient data, it’s subject to much fewer privacy regulations and data protection policies, so made sense as a starting point for digitalisation initiatives. Operational data refers to project management, finance, quality, trial management and drug manufacturing data etc that together support the drug development process.
However, when it came to using operational data effectively huge challenges stood in the way, primarily from the fact that the data typically resided in siloes, locked into dedicated and fragmented internal systems.
Continuous access to data, logical integration, proper data curation – combined with novel data science and engineering tools, would be crucial to transforming the goldmine of inaccessible operational data into actionable insights. To do so, Novartis four years ago invested in developing Nerve Live, a first-of-its-kind advanced analytics platform, with a hybrid architecture and local data ingestion server supported by cloud-based storage.
What does Nerve Live do?
- The platform ingests data from disparate sources, cleans it, links it, and makes it ready for analytics to be applied
- Enables sophisticated ML algorithms to be built that can analyse the data and generate actionable insights
- Enables web-based solutions/applications for employees to use in their workflows to improve their decision making
Image Source: Wiley Online Library
“Left panel: High‐level schematic representing sample data sources and the Nerve Live platform through its main logical components: data lake, analytics layer, data access application programming interface (API) and end‐user interfaces (modules or solutions). The current eight Nerve Live modules are indicated by coloured round icons with an acronym (see Table 1 for a complete description). Right panel: Illustration of the Nerve Live program through example screenshot from four Nerve Live modules (solutions).”
What challenges did they face developing Nerve Live?
- A step away from the norm – as the platform architecture was novel to the organisation, approval was far from guaranteed and took many rounds of discussions.
- Data ownership – aside from the challenges of data siloes, variation in data display and formatting across different functions added to the difficulties in accessing this data.
- Connecting and curating data – as operational data tended to be generated for temporary tasks it rarely lent well to doing analytics, so to enable its use across the operational landscape required a monumental effort. Decades of data stored in various systems with inconsistent formatting and identifiers needed to be reprocessed before it could be analysable.
The result of Nerve Live (so far):
In setting out to access experience, create intelligence and unlock value across their R&D value chain, Novartis has unveiled eight different solutions, (akin to mobile apps they say) that are currently being used by over 4000 stakeholders. The first they developed, for example, DCN Insights, uses ML to predict the optimal end of enrolment time across 300+ studies simultaneously. The module uses visualisations and ML-based risk to analyse events across trial sites globally. The tool acted as a “technology demonstrator” for the company, helping to gain organisational trust in it’s use to augment decision-making.
The review includes a summary table of the other seven Nerve Live modules developed:
- FPO (Footprint optimizer): Enables study enrollment scenario planning and identification of optimal study locations (countries and sites) based on historical data
- RP (Resource planner): Forecasts resource requirements, such as staffing and time commitment needed, for each clinical study
- ETP (Early trial pricer): Enables early prediction of clinical trial costs for various scenarios, allowing teams to choose the most suitable one
- SENSE: The “control tower” for Novartis clinical studies, helping to monitor the portfolio of studies and identify any potential risks to timelines or costs
- RC (Resource cockpit): Visualizes project portfolio across technical units, and enables prediction and optimization of resource allocation (internal and external)
- Nucleus: Enables design of optimal study drug supply plans and minimizes supply risk
- DYNAMO (Dynamic allocation with machine optimization): Forecasts total and phased costs for every new (planning) and ongoing (tracking) clinical trial
What makes it successful?
In reflecting on the Nerve Live program as a whole, the authors pinpoint three factors that have been crucial to its success within the organisation’s digital transformation strategy:
- Senior management sponsorship was widespread across functions and seminal to the digitalisation agenda. Without broad endorsement, it wouldn’t be possible to break the status quo of drug development
- The core team setup of talented product owners, each delivering solutions aligned with the strategic vision and the needs of the future users of the Nerve Live module. Such individuals led product development, from ideation to data science discovery and delivery. The authors noted the difficulties in defining the ideal product owners and desirable capabilities and qualifications, but generally, they tended to have data science expertise, business problem translation to “computably tractable questions” and high learning agility. Additionally, product owners should be able to navigate IT process governance and software applications deployment.
- An entrepreneurial approach with a product mindset allowed Nerve Live developers to think big but start small with a single platform working on just one module. This small team, working in a lean model delivered tangible results quickly and thus drove later investments to build additional solutions
Novartis is only beginning to systematically assess the benefit of the Nerve Live program, but so far they estimate that they have achieved productivity gains of around 10% across their portfolios since it’s application. But the authors were also keen to highlight the gains beyond efficiency, in terms of transparency, where the integrated systems have allowed data to be more accessible at all levels across the organisation. This, they believe, will allow greater collaboration in the approach to big, strategic questions, and moreover, transform how they seize the opportunities presented by data science and AI in the future of drug development.