Walking through the vendor area at HIMSS14, it was impossible to avoid being bombarded with analytics, analytics, analytics; everywhere you turned a vendor was advertising and pushing analytics. Unfortunately, most were overselling an underwhelming product. They look pretty, but they only scratched the surface of analytics and they will not be the partner that will carry your organization forward for the long run.
Before you pick an analytics partner and make a significant investment of time and money, you need determine if your vendor will be able to deliver long-term value. To do that, you need to understand the three phases of analytics and determine if your vendor will be able to support all three.
The first phase of analytics is descriptive analytics. Descriptive analytics use data visualization, statistics and mathematical models to show the current or historic state of an organization or population. The math may be fancy, but it only shows what has or is happening. This phase of analytics is best used for identifying patterns and correlations and is often used in management decision making. If your health system does not have transparency into its data, descriptive analytics can be a very powerful step forward. However, it is only phase one.
Predictive analytics is the second phase. We are all familiar with predictive analytics even if we don’t realize it. According to my phone, tomorrow is supposed to be sunny with a high of 70 degrees. Meteorologists have been using predictive analytics for decades, but it is relatively new technology to healthcare delivery. Predictive analytics are best used for relatively short term predictions of future state. In healthcare, that may mean predicting the number of nurses needed to staff a unit based on the number of expected admission and discharges, or risk stratification of an individual patient. I saw very few vendors that could clearly demonstrate meaningful predictive analytics.
The third phase is prescriptive analytics. Prescriptive analytics combines predictive analytics with pre-defined actions (or rules). The best known use of this type of analytics is automated trading on the stock market. The predictive algorithm predicts the price of the stock will rise if x or y occurs, so when x or y does happen the computer buys a predetermined amount of stock without human intervention. Similar types of algorithms are used by Amazon to give you suggestions while you shop. No human is involved when they recommend you buy conditioner to go with your shampoo.
There is extremely limited use of prescriptive analytics in healthcare, but its use will grow as we get better at population health. At this time, the best use of prescriptive analytics is to identify a high risk population and then automate a low risk intervention. For example, if you were able to identify a group of patients at high risk for a heart attack you could then use an automated calling system to screen them for evidence of chest pain or shortness of breath. Those that screened positive could receive follow-up care and those that screened negative had a very low risk intervention.
As you consider an analytics vendor, look past the how pretty their graphs are and determine if they will be able to carry you through to prescriptive analytics. It will be hard to justify why, after months of developing data models and making charts you need to do it again with a different vendor because your first one only did descriptive analytics.