It is without a doubt that the recent movement in big data will escalate to greater heights in years to come. McKinsey and Co. projects a large shortage of data scientists of 140,000 to 190,000 with the necessary analytical expertise.
Kevin Fickenscher, president of CREO Strategic Solutions and healthcare executive, calls healthcare data scientists the “sexiest” members of the healthcare delivery team. A bold statement made by this leader. Big data has a wide array of entry points into the healthcare domain. With the adoption of the Affordable Care Act, the incentives regarding the healthcare costs, quality and delivery will dramatically alter the way physicians and other healthcare providers operate day to day. Soon enough, physicians will be prescribing mobile apps and wearable sensors to their patients. In fact, there is a projected $3B market for connected healthcare devices by the year 2019.
Despite the interest and overbearing hype about the potential of disrupting healthcare with big data analytics, the current approaches being employed by healthcare leaders employ trivial computational rigor and carry limited impact. Lately, the potential for applying big data and software to medicine has been touted by business leaders across the world. Many proponents of applying machine learning to healthcare focus on just that: applying. Biomedical leaders have used existing machine learning methodoogies for initiatives in predictive disease modeling, pharmacogenoimcs, targeted chronic disease management, and electronic health record management, among others. However, the application of existing technology can only go so far. Applying is not enough; doctors have to invent new computer science to create the greatest impact. Healthcare data is messy, complex, sparse and unpredictable. Approaches such as time series analysis for financial instrument pricing, or collaborative filtering for targeted advertising, cannot be directly translated to healthcare problems.
Solving the most complex, pressing problems in healthcare, should not just involve borrowing computer experts from other domains and asking them to implement their technology on healthcare data. Fundamental computer science needs to be invented in order to solve the most intricate problems, many of which are currently unforeseen. To address such problems, the healthcare domain has a need for robust methods in stochastic process modeling, efficient data compression, privacy and security, just to name a few. Simply thinking about problems from a high level systems perspective is insufficient for addressing the urgent problems in healthcare — a rigorous understanding of computational methodologies and their limitations is crucial. It is the duty of physicians, data scientists, and healthcare executives to acquire the robust technological skills necessary for alleviating the crises in our healthcare system today.
Robert Chen, MD,PhD Candidate