We sat down with Milind Kamkolkar, Chief Data Officer of Sanofi, and Carlos Ariza, healthcare expert at PA Consulting, to discuss how to build effective analytics organizations, and how to take advantage of new technologies such as Machine Learning, Artificial Intelligence (AI), and blockchain.
Q: What are the challenges in building an analytics organization, particularly around attracting and retaining top talent, as well as getting buy-in from other functional areas?
Milind: A universal challenge in keeping staff motivated is the disproportionate amount of time analysts spend cleaning the data, at nearly 80-90%, instead of driving insights from the data, a mere 10-20%. Oftentimes this cleaning leaves the reports out of date before they even hit the business user’s inbox.
Carlos: I agree with Milind; we see many organizations using a cesspool of spreadsheets based on legacy systems and out of sync with each other, rather than relying on modern BI platforms to support their internal clients. But the challenge to get there is still in data cleaning and data management.
Milind: Exactly—what if we could use Machine Learning for data cleaning, using the MDM principles of the organization to begin driving it? This would free up analysts’ time to work with the business and help them move away from Excel spreadsheets and adopt newer, more up to date models to inform their work.
Q: So, what could developers/data analysts do with the time not spent cleaning data?
Milind: The analysts could use this extra time to move away from the creation of MVPs (Minimum Viable Products) to what I call MLPs, or Minimum Loveable Products. MLPs are low-cost technology applications with an attractive UX interface that will accelerate initial adoption. Most pilots to the business begin with MVPs, resulting in rejection from the business users because the tools are hard to use. Building a bit more UX into pilots can avoid these bad first impressions—and we all know the value of a first impression!
Another important point when building new analytics applications is to make sure that FAIR data principles are applied throughout (i.e. all data is Findable, Accessible, Interoperable and Re-useable). These open data principles will allow for internal transparency and cross sharing that will reduce internal redundancies and pave the road for richer insights in the future—delivered from both internal and external data sources.
Innovation Matters: What are the 'innovation leaders' doing right?
Q: It seems like a lot of the challenges in healthcare require better collaboration and data sharing across organizations. However, regulatory restrictions around PHI and valid concerns about data security are slowing down adoption. How do you think this is going to evolve in the next few years?
Milind: Regional and international privacy regulations make it difficult today to collaborate and share Healthcare and Life Science data in a secure, timely way. Looking forward, however, possibilities exist using technologies like blockchain for individuals to own, control and grant access to their own data, potentially charging premiums to those who wish to mine it. For example, blockchain is currently being used to aid displaced individuals, such as Syrian refugees, to help keep track of their healthcare records in this time of political disarray. The hard part is finding a balance between the benefits of having all your healthcare information in one place, and the negative “big brother is watching” implications that deters much of the adoption.
Carlos: I agree with your points Milind, we’ve heard examples from our clients where they’ve tried to implement data sharing solutions but ran into many regional and local regulations. At the moment there are experiments to use blockchain to allow EMR interoperability. I have been following this story and would love to see the implications going forward.
Q: Assuming that you have the right organization to take advantage of data, and access to timely information, the next step is applying advanced algorithms to find new insights. This is where new technologies like Deep Learning come in. Where do you think the healthcare industry is on this front today, and where will we be in 5 years?
Milind: Let’s remember that the recent advances in computing power have driven the changes in data processing that we see now in the Life Sciences and Healthcare industries. We’ve seen other innovative technologies in the past take as long as 20 years to truly be adopted. The adoption rate is much faster now, and what we’ve seen around deep learning is impressive, but there is also significant hype in this field. At the end of the day, deep learning is an evolution of existing predictive modeling techniques, and as such it requires data preparation and skills in statistical analysis, visualization, etc. that were also required for earlier predictive modeling tools. We’re seeing many pilots emerge around “cognitive” and “AI,” but it will be interesting to see which of those fade away, and which become truly transformative in the Healthcare and Life Sciences fields.
Milind and Carlos went on to facilitate a roundtable on the topic, and heard many of these same issues echoed from leaders across the Life Science and Healthcare industries.
Does your firm experience the same challenges in their business intelligence pilots and workflows? Where do you see your company in 5 years—eagerly embracing AI and Machine Learning, or struggling to adopt these technological advances?