In the coming year, both the opportunities and challenges for predictive analytics will likely increase. The statistical and predictive data analyst role is slated as one of the top 10 most in demand — yet hardest to fill — this year, according to Forrester.
As the market continues to expand, 10 chief medical information officers at hospitals systems around the country shared their thoughts on the opportunities, challenges and evolution of analytics technology in healthcare.
Q: In your opinion, what do you need for a successful predictive analytics program?
C. Mason Brown, MD, MBA, CMIO, Baptist Health (Montgomery, Ala.). Predictive analytics as a concept sounds like just what the doctor ordered. It is actually in the "ordering" where we can face dilemmas. Ideally, predictive analytics would present data to the healthcare provider at the point of care within his or her normal workflow. When predictive analytics presents clear, actionable data, there is obvious value. However, presenting ever increasing data points to providers without clear, evidenced-based responses can have an adverse effect on care. There will be increasing demand for predictive analytics at the bedside as long as it is tempered with research-based interpretive options. Establishing formal governance to establish predictive analytics directives will be critical to successful integration of these processes.
Louis Capponi, MD, CMIO, Cleveland Clinic. Analytics products and capabilities are rapidly expanding, so we need to understand predictive analytics alongside other data innovations such as artificial intelligence, natural language processing and computer learning.
To develop an accurate prediction model, you often need large datasets from multiple sources, so partnerships can be incredibly important. Additionally, models need to be kept up to date so they continue to be predictive as population risks and treatments evolve. Finally, predict what you can impact.
Robert Budman, MD, MBA, CMIO, Piedmont Healthcare (Atlanta). Other than scads of money and amazing employees? We built a small team of carefully selected data scientists with level-headed managers to set up data collection with rationally constructed queries within the platforms to run the predictions, and close monitoring of results to measure success and drive refinements. It is all too easy to focus on financial targets and outcome improvements, but you have to look at resources, hours and workflow, too. Rule No. 1: When you find good reporting folks, treat them right!
Sajjad Yacoob, CMIO, Children's Hospital Los Angeles. For predictive analytics to succeed, you need both data scientists and clinicians who use large amounts of data to provide guidance on clinical care decisions. Specifically, you need data scientists, physicians, nurses, pharmacists, database experts, computational scientists and EMR experts.
They must work as a team to examine the data, and then present it to clinicians in the EMR. For example, we utilize a system to help us predict — using real-time labs, vitals and other clinical data — which patients are at risk for sepsis and provide clinical guidance to the physicians. This group must constantly look at how the EMR is using this data, if the algorithms that were created are sensitive and specific enough and if the clinical processes are providing the outcome desired.
Q: What do you see as the greatest opportunity in using analytics to improve clinical quality?
Neil Kudler, MD, CMIO/COO, Baystate Health (Springfield, Mass.). Analytics holds the promise to predict risk and identify interventions for the right patient at the right time. Real-time actionable knowledge, derived from the integration of disparate data sources, can lead to significant improvements in quality. Understanding the population, however defined, can help clarify what resources are required to ensure and improve quality. We are working with a company that has demonstrated the ability to predict hospital-acquired conditions, like pressure ulcers, sepsis and line infections, more precisely and with greater sensitivity than what the EMR has to offer. Analytics can also help assess for gaps in care in the community and provide insights that are more meaningful than basic demographics. This can be very useful in defining objectives for performance and quality improvement.
Steven Orlow, MD, CMIO, Lutheran Hospital of Indiana (Fort Wayne). I continue to believe that the greatest potential for analytics centers on better understanding of the basics of healthcare — blocking and tackling, as they say. We are seeing benefits as we look at the variables that affect most patients instead of analytics around rare events or very unique, patient-specific activity.
Luis Saldana, MD, CMIO, Texas Health Resources (Dallas). I believe analytics are critical to improving clinical quality. You can't improve what you can't measure. The challenge is designing the right metrics to help answer the question, "Are we improving?" To get to improved outcomes, you have to design and iterate effective processes, and you need to use the right process metrics to get to better outcomes.
Q: How do you see big data and analytics evolving in healthcare in the next 5 years?
John Kravitz, CIO and Interim Chief Data Officer, Geisinger Health System (Danville, Pa.). I think population health involving advanced analytics has increased over the last year. Also, the interoperability of diverse platforms to support the analytic process to support the population health initiatives. Genomics is becoming a hot topic this year as well, especially as genome sequencing is beginning to identify risks to patients that save their lives. The integration of genomics into the provider workflow as this information becomes available is an item that a lot of providers are wrestling to accomplish successfully.