…Real-time analytics, that is.
One of the challenges in healthcare today is the lag in time for the acquisition, storage, analysis and delivery of data back to the physician or other healthcare provider for decision-making. The majority of transactional information about a patient that is produced on reports or presented in terms of business intelligence is simply “old” or retrospective and, as such, limits its usefulness for decision-making. The average reporting solution attached to the modern EMR software typically batches up information every 24 hours and makes it available for reporting and analysis. In many healthcare decision-making settings like an Emergency Department, this data lag is frustrating and potentially dangerous from a patient safety point of view. It’s not that EMR solutions can’t immediately reproduce the collected data quickly, but to analyze the data and present actionable information to a medical professional, the data lag for real business intelligence limits its return on investment and ongoing use.
The Challenge
The problem is simple – we need to perform complex real-time analytics on data, both structured and unstructured, that is in motion. The ability to continuously analyze data, especially in healthcare, could make the famous “life and death” difference. Let’s take a basic example. Today, there are mobile devices that can stream electronic data like an EKG while you are in the ambulance headed to the ER. The challenge is combining the real-time information fed from the telemetry with your existing electronic health record, filtering out extraneous data and presenting the key facts to the physician and the support team for decisions and preparation when you arrive. This issue is especially true in modern trauma units where minutes count and integration of real-time data, especially lab results, is critical. Instead of the doctor scanning all of the values for ones that are out of range, the real-time solution would highlight the out of range values and possibly recommend alternatives to address them.
Data Analysis Differences
The concept of analyzing data in motion is call “stream computing” and it is a relatively new paradigm. In traditional data processing, you typically run queries against relatively static sources of data, which then provide you with a query result set for your analysis. With stream computing, you execute a process that can be thought of as a continuous query, that is, the results are continuously updated as the data sources are refreshed or extended over time. So, traditional queries seek and access static data to be queried and analyzed, but with stream computing, continuous streams of data flow to the application and are continuously evaluated by static queries. You are flipping things around to manage the real-time data more effectively.
The beauty of analyzing data streams “on the fly” is that the data can originate from sensors, cameras, news feeds, stock tickers, or a variety of other sources, including traditional databases. Just like in our simple example, real-time or continuous applications are composed of individual queries or operations that interconnect and operate on multiple data streams. Those data streams can come from outside the existing computer system or be produced internally as part of an application.
Analytic Capabilities
There are powerful tools available today to build complex real-time analytics on data that is in motion, especially for meeting this need in healthcare. One example is IBM InfoSphere Streams. IBM InfoSphere Streams provides an execution platform and services for user-developed applications that ingest, filter, analyze and correlate potentially massive volumes of continuous streams of data. InfoSphere Streams was a key part of the “magic” that made IBM’s Watson’s win on Jeopardy possible in the now famous man versus computer match. Multiple instances of Streams handled the high speed analysis of the queries in real-time for playing the game. One more interesting fact: Streams was designed to work on both traditional structured data and unstructured data like, doctor or nurse notes.
The potential to develop real-time analytic applications on continuous data for healthcare is exciting and an opportunity to step forward with innovation. Why worry about landing data in data models, or meeting shifting standards for healthcare integration when this technology is available? Let’s be the game changers and not chase traditional ways of thinking. What are your thoughts?
