Moving healthcare analytics from an event to a process
I have spent the better part of my 40+ year career analyzing data. Lots and lots of data. And the great thing about data is that it can contain answers to meaningful questions, answers that usually lead to incredibly impactful improvements. Key though is applying the proper analytics to the right data. Now, the fact that I've had that long career also means I've reached an age where, let's just say, I have developed a great personal interest in the quality of healthcare. And so I find myself fascinated by the advancements in data through "big data" and simultaneously thinking about how that data might be capitalized upon to improve healthcare on a society-wide basis. The good news is this: Great strides are being made in applying analytics to big data to improve healthcare quality and reduce costs!
In this article I examine the steps that can be taken to move healthcare analytics from an event to a process, from simply examining a single outcome to instead expanding the analysis to look at what I have come to define as the analytics question space. By creating programs rather than executing projects we can leverage the data aggregation and cleansing efforts to create synergy between studies and speed the delivery of results.
While healthcare data is building exponentially, the industry's ability to process, extract and analyze that data in a meaningful way hasn't kept pace. At the same time, the availability of technology to collect and analyze large volumes of data has provided opportunities for managing population health at levels never before possible.
This convergence of big data and analytics enables healthcare organizations to answer the question, "How do we glean as much useful information as possible out of the data?" In large part, the answer depends on how effectively it is structured for analytics.
"Having volumes of data can be good if you have a plan and architecture to really use it effectively," says Tom Parry, PhD, president and co-founder of the Integrated Benefits Institute in San Francisco.
"It's about using the data we generate to analyze how we improve healthcare efficiency, performance, and delivery," says Rob Schmitt, CEO of Gibson Area Hospital and Health Services in Gibson City, Ill. "We generate tons and tons of data — patient data, financial data, medical records, pharmaceutical. The real challenge is, how do you utilize that data efficiently?"
Dr. Parry adds, "One of the biggest challenges for employers is to think creatively and pragmatically about data metrics and then put big data and sophisticated analytics in that context."
Outcomes research: Reducing costs by making People healthier
Dr. Parry uses analytics to help employers broaden their view of health and understand the financial impact of outcomes beyond healthcare costs. He finds that employers have been so focused on healthcare costs, they've often missed the outcomes that flow from better health — particularly as it is germane to their business performance.
His research shows that healthcare costs are typically only 40-50 percent of the total economic impact of health on a company, with the remaining costs being the financial consequences of lost time and lack of performance that result when employees aren’t healthy. Dr. Parry says the real opportunity for employers is to find ways to prevent disease.
Mr. Schmitt also looks to big data analytics to help his hospital improve population health — from preventing that next readmission to managing people’s health before they even get to the hospital.
"What we would like to do, and what we believe you need to do to be successful for population health, is break down the data even further into social economic factors that impact people's lives."
For example, if a patient keeps getting readmitted with pneumonia, the hospital could use data to analyze the external factors that could be contributing to the recurrence, asking questions such as, 'Are their surroundings contributing to their ill health?'
"You really can get down to that core basic level of each individual need. I think that's where hospitals in general have to get to to truly impact population health," Mr. Schmitt says.
Mandi Bishop, owner of Adaptive Project Solutions in Jacksonville, Fla., sees big data analytics going even further to paint a picture of overall health. "Genomics and proteomics are the two biggest opportunities that are presenting themselves from a patient wellness perspective,” Bishop says. “When we talk about health, health is all of the aspects of your life."
Ms. Bishop says that big data analytics can connect a person's lifestyle data, habits, activities — even financial health — and help a patient better understand how all of those external factors can play a role in preventing disease. Armed with that knowledge, patients can apply healthy lifestyle factors appropriately and potentially avoid turning on the genes that would cause disease in the first place.
Predictive analytics: Improving the odds of good outcomes
Where the convergence of big data and analytics really starts to grow together is in predictive analytics — the ability to use unstructured data and various algorithms to become more and more aware of the outcome.
Can genome sequencing, for example, tell you how well a particular pharmaceutical works for a specific population? Can adverse drug effects be avoided if you can predict how a particular person, given their genetic make-up, would react?
With predictive modeling, you can compare similar cases, come up with a statistical model of what’s happening to a particular population, and assign some probability of outcomes. It allows healthcare providers to better answer the question, "What's the most likely thing to work?"
Useful data: Getting it right
A critical component of making healthcare data useful is getting it organized in such a way that it can be viewed across large populations in an analyzable form.
"The ideal is to have data integrated at the person level across all of these programs. That really puts the user in a very powerful position to understand what’s going on with individual people, and how they're changing over time," Dr. Parry says.
That's easier said than done. From a data delivery standpoint, marketplace vendors are experts with data in their silo, but they're not integrated with other healthcare data providers. "You have the group health experts and the disability experts and the worker's experts and the EAP experts and the wellness experts — but they're not integrated. And that's where employers really get frustrated because they really don't have a single partner that has brought all this information together for them and tell them what it means," Dr. Parry says.
While building an integrated data warehouse can be an expensive business proposition, Dr. Parry says starting with high-level metrics to describe the population and the leading indicators of health, medical delivery and outcomes helps employers think about the use of data and metrics, and how information plays into their business case.
However, you cannot create an integrated data set to investigate big problems by simply doing individual projects and hoping they will grow together. The key to creating this integrated data set (or "data lake" as it is now being commonly called) quickly and effectively is to truly understand in advance where you are going. My experience has shown me that the best way to do this is to define your analytics question space, identify the information you need to answer these questions, and determine the source(s) and degree of availability of that information as the starting point.
Questions First: Assembling and analyzing the right data
At its core, analytics involves questions, data and analysis — deciding on the right questions, assembling the right data to analyze, and then performing the analysis. While it is tempting to jump straight to the analysis, if you don’t do the first two things you’re not going to get a successful result.
"We're doing analytics, but are we doing analytics on data that is meaningful?" asks Mr. Bishop say.
Defining that analytics question space is critical for any successful big data implementation. It focuses the discussion on gathering data that would be useful, rather than just gathering as much data as possible and hoping to make sense of it.
Asking the right questions upfront also helps reduce people’s anxiety surrounding big data. Dr. Parry says it's easy for employers to get overwhelmed with data, especially when it's coming from multiple vendors dealing with different populations of employees and different cuts of the data. Thus, establishing an analytics framework (or a similar alternative) is critical to making the entire process visible and understandable to the business.
Collaboration: Joining forces to enable better outcomes
Rather than go it alone, healthcare experts are finding that reaching out to other organizations can offer tremendous opportunities for implementing big data analytics.
Dr. Parry recently piloted a program to help employers gather the right data from vendors, including providing a data template that helps employers identify areas where they may have too much data, or not enough.
Mr. Schmitt's hospital is partnering with the county health department and school district in a joint effort to make an impact on population health. "We're actually looking at a new model of how healthcare can be delivered through public and private collaboration," he says.
Ms. Bishop likes the idea of collaboration and she sees consumer involvement as an important avenue for big data analytics — introducing patient-generated data into the mix and having patients become aware of the impact of their healthcare data on their lives, as well as the country's broader economic health.
As more and more of these collaborative efforts take shape, healthcare organizations will rely more and more on the ability to apply big data analytics to drive solutions and better health outcomes.
Wayne Applebaum is vice president of analytics and data science at Avalon Consulting, LLC. He has worked with Fortune 1000 companies to develop analytics programs that impact their bottom line.