Those in and around the healthcare industry have heard a lot about predictive health analytics recently. The reason for this was recently summarized well by B. Vindell Washington, MD, National Coordinator for HIT in the U.S. Department of Health and Human Services, who said, "Digitization has allowed us to have a more meaningful conversation around how to use predictive analytics to improve patient outcomes." (Source: Healthcare Executive Magazine, Volume 31, Number 5, September/October 2016). The adoption of Electronic Medical Records is ushering in a new digital age in Healthcare where organizations have the opportunity to fundamentally re-architect workflows to be more data-driven, dynamic, and patient-centric.
The digital hospital of the future will incorporate predictive analytics as a core component of its overall data strategy. A recent report by Grandview Research estimates the 2015 global healthcare market for predictive analytics at $1.48B, with a CAGR of 29.3 percent forecasted through 2025.
Early use cases of predictive analytics in areas such as population health, predicting admissions levels and acuity to improve staffing, predicting the occurrence of hospital-acquired conditions like sepsis and identifying security breaches have demonstrated the potential of this capability. As hospitals improve the quality and integration of their data into new analytics-oriented data stores and enrich that data with new streams of unstructured data sources (i.e. images and clinical notes) the value of predictive analytics will no-doubt continue to grow.
Much of what is possible today with predictive analytics has only recently been enabled by advances in big data technology. Since the early 2000's there have been significant advances in big data processing frameworks (i.e. Hadoop) and storage, compute, and networking technologies. As the volume, variety, and velocity of data continue to grow, specialized technologies are increasingly deployed from data ingestion through model development to improve the speed and scale of predictive models. The desire to incorporate a growing number of unstructured data types to identify new correlations and insights has caused massive refactoring of previous techniques and driven exciting advances in analytics practice.
However, big data technology isn't much good without big data. In healthcare, the adoption of Electronic Health Records (EHRs) has ushered in a new digital era which has made predictive analytics possible. Much like big data technology, the use of EHRs is still relatively new. This chart from the HealthIT.gov shows adoption of EHRs through 2015.
Not surprisingly the adoption of EHR systems corresponds closely the CMS incentive payments shown above. As healthcare organizations start to emerge from the era of EHR adoption, they are looking for ways to get more value from the systems, and resulting data sets, that they have invested in.
Using predictive analytics can be an inflection point in an organizations' journey to become data driven. Compared to traditional BI, which shows what happened in the past, predictive analytics indicates what may happen in the future. This forces interesting questions about the organization's readiness to proactively change workflows based on what might happen in the future. How would the output of a predictive model be put into action? Is there the requisite agility to reallocate resources or change workflows as a result of a real-time prediction? Predictions, by nature, won't be correct 100 percent of the time, so how do workflows comprehend false positives or negatives? There are also questions about technology infrastructure ranging from the availability of high-performance computing resources to storage and networking bandwidth to deliver real-time results to clinical workers. Some of these questions will be addressed in the following posts, and I look forward to community input on other key questions that should be considered.
More and more we're seeing statements like, "every company is a data company" and "data is the new gold.” These increasingly popular phrases reflect the growing realization that leveraging data to optimize and compete will be table stakes for organizations to succeed in the future, and healthcare is no exception. I would offer that for healthcare organizations to be successful, they will need advanced analytics a core capability (more on this topic in the next post). I'd suggest one useful framework for thinking about how to develop advanced analytics as a capability is treating it like an internal startup. One useful startup framework to consider is the Startup Development Phases from StartupCommons.org. This framework presents three stages of maturity: Formation, Validation, and Growth. Each of these stages is analogous to the readiness of an organization to truly embrace and leverage advanced analytics as a core capability and to modernize workflows to be data driven as a new standard method of care delivery.
This post is the first in a four-part series. In the next three blogs, I’ll dive deeper into each of these stages and highlight some of their key characteristics that could be most relevant to healthcare. Read Part II in the series.