informa
/
Commentary

Why Healthcare is Behind the Data Curve

The healthcare sector has been trailing industries such as banking and retail when it comes to adoption of data analytics, but a growing number of success stories in healthcare provide proof of concept for other organizations to get on board.

Nearly every industry is searching for ways to create value from rapidly expanding amounts of data, but few sit on as much untapped opportunity as today’s healthcare providers. Electronic health records and connected medical devices have created an unprecedented volume of data. Most of this data sits unanalyzed, leading to operational inefficiency and missed opportunities for improved patient care. However some healthcare organizations are getting value from their data by finding applications that provide quick wins, and they are using their early successes to build a data-driven culture.

Health Catalyst recently released a survey of hospital executives on the role of predictive analytics in the healthcare industry. Though 80% of the executives surveyed believe forward-looking data analytics could significantly improve the industry, only 31% of respondents have integrated these advantages in their organizations. Even worse, 19% have no plans to do so.

What explains the slow adoption of predictive analytics in healthcare? It is largely historical. In contrast to industries like banking and insurance, which treat data analysis as a core competency, medicine has long been practiced as an art. Physicians accepted that medical information was locked in hard-to-analyze handwritten notes. As a result, few healthcare providers bothered to assemble analytics teams. Many still feel unprepared to do so today.

A multitude of examples show that a successful first project can build the confidence and momentum needed to create a data-driven and highly optimized organization.

As a case in point, one large Midwestern hospital system struggled with high costs arising from patients who were discharged too early and subsequently readmitted. Doctors broadly understood the risk factors for most patients, but they sometimes missed critical information because they lacked the time to pore over every detail of every patient’s medical record.

The hospital built a predictive model that physicians now reference when making discharge decisions. The model shows both a probability of readmission and the medical or social factors driving that risk. In some cases, the model confirmed the doctor’s understanding. In other cases, it flagged issues they would have missed, leading to better medical decisions.

This hospital system noticed an immediate financial return on their analytics investment. Building on this success, they created a new model to forecast staffing needs for each day. With the new model, they reduced both costly overstaffing as well as dangerous understaffing situations. Because they already had the data, they found model building and business optimization remarkably quick.

Their growing confidence in modeling has motivated dozens of new applications. Their newest predictive modeling project will improve how they escalate care from outpatient departments to the emergency department.

Are these highly optimized healthcare organization the wave of the future? We think so, but we also see many providers getting derailed by focusing on the wrong applications. Organizations betting on moonshots -- like replacing doctors with technologies that aren’t ready for prime time -- may become disillusioned and miss out on clear wins that are readily available today.

Healthcare organizations are at a critical juncture. They have the data to become dramatically more efficient if they start in the right direction. As one success leads to the next, they will soon look back and wonder why they waited so long to improve their patients’ health outcomes and their bottom line.

 

Jeremy Achin is a data scientist turned entrepreneur. As CEO of DataRobot, Jeremy sets the direction of the company, products, and the culture. He's passionate about helping organizations become more efficient by deploying machine learning everywhere. Prior to DataRobot, he was Director of Research and Modeling at Travelers Insurance where he built predictive models for pricing, retention, conversion, elasticity, lifetime value, customer behavior, claims and much more. A data science enthusiast, Jeremy spends his spare time building predictive models, usually on the data science competition platform Kaggle.com. Jeremy studied math, physics, computer science, and statistics at University of Massachusetts, Lowell.