Teh Seng Khoon
Staying ahead in business: using data science in the medtech industry
While technology has catalysed the improvement of healthcare solutions over the past few decades, improvements in patient health outcomes are growing at a slower pace than costs. In the medical technology (medtech) industry, a reformation is underway, with new players attempting to disrupt the healthcare system to make it more sustainable. Meanwhile, established medical suppliers are turning to data science to glean new business insights and explore other sustainable opportunities in the healthcare ecosystem.
Medtronic Asia Pacific (APAC), where I work, is the world’s leading medtech company. It develops biomedical engineering applications, including the world’s first implantable pacemaker back in the 1950s. I became the first principal data scientist at Medtronic two years ago. The crux of my job as a data scientist is to drive deep into the heart of business problems, investigate how resources can be deployed efficiently and effectively, and recommend the next best course of action to stakeholders. The solutions data scientists offer could help an organisation stay ahead of the curve, by marrying business knowledge and ideas with sound data science analyses and solutions.
Working as a data scientist: what do we do?
Data scientists work across multiple disciplines. This requires a breadth and depth of working knowledge, as well as good communication skills to draw out what a business unit needs and formulate solutions supporting their objectives. Having the necessary skills to identify problem areas is as important as tackling the problem itself. A common misconception is that data scientists simply analyse data, produce machine learning models and employ artificial intelligence to address an issue. There is, in fact, a good deal of soft skills required to distill a business problem into data science questions, before drawing out the metrics that the team should focus on.
One of my initial tasks in Medtronic was to make data science easily understood and accessible to business stakeholders. With this in mind, I designed and introduced a chart to explain data science to the average layperson (refer to Chart 1 below).
I use this to trace how I typically analyse a project. First, I start figuring out the data science problem and communicate closely with the data owner. After clearly establishing the problem hypothesis and business metrics to focus on, I then recommend data-driven machine learning or artificial intelligence solutions useful to the business stakeholders in the form of prototypes. With a minimally viable product, I design suitable cross-over experiments to test the efficacy of the prototype solution through a small-scale study. If it is promising, I work towards applying it for scale and impact.
Some business cases may not require on-the-fly artificial intelligence solutions to improve operational decisions. In these scenarios, I typically apply casual inference or probabilistic Bayesian inference to understand how modifying leading indicators can impact business outcomes before making recommendations. I employed this approach in a related project, where health economics modeling was used to determine whether a product could benefit healthcare stakeholders cost-effectively, prior to running real world studies to confirm our hypothesis. Here, I used decision theory techniques such as decision tree analytics to simulate one-off clinical decisions, and the Monte Carlo Markov Chain (MCMC) technology to simulate clinical scenarios with similar, repeated decisions. The cutting-edge Markov network in MCMC, in fact, lays the foundation for (Deep) Reinforcement Learning in decision-making in the application of data science to health economics modelling.
Even though employing data science is not new in medtech, the industry is still grasping how it can be wielded to gain a competitive edge in customer-centric sales and marketing activities. This is perhaps unsurprising as medtech companies conventionally employ product-centric and territory-focused sales strategies to achieve their top and bottom-line growth. In this respect, data science is gaining ground as a key enabler that could enormously transform a company.
It was fulfilling for me to start and lead a project that successfully prototyped a customer purchase behavior-based segmentation engine. This solution helps sales and marketing units channel resources to engage and prioritise existing customers, as well as identify those we need to reach out to. I employed design-thinking methodology that factored in user pain-points to identify problem areas, before outlining essential parameters to track. Data science and engineering techniques, plus business process modelling were also engaged to iteratively prototype a solution for users. These were eventually presented on a dashboard that the team could now use to stream data and glean customer insights. We are now launching outcomes of this work in a few countries and are measuring the impact of this solution.
These above on-going achievements would not be possible without my Singapore Management University (SMU)-Master of IT in Business (Analytics) (MITB) studies. SMU-MITB provided me with a very broad-based education linking business strategy to analytics. Very often, I could almost immediately apply my SMU-MITB learnings on my professional work. I am still ardently pursuing knowledge and technical proficiency, to build on my prior training in business analytical applications and classical healthcare analysis approaches, as well as in data science and engineering methodologies. Currently reading my part-time Ph.D. in Data Science and Engineering at SMU, it is fascinating to witness how state-of-the-arts data science is developing and overlapping with healthcare. These are exciting times for data scientists and data engineers indeed, with growing opportunities abound in medtech companies, and the industry in general.
Author: Teh Seng Khoon, Principal Data Scientist, Medtronic-APAC
Last updated on 31 Oct 2019 .