The Democratization of Data Analytics

SAP Health's CMO on ways to make the most out of analytics.

The healthcare ecosystem is becoming more integrated and data centric, as payments shift to reward outcomes across the continuum of care rather than just performance of activities in a non-integrated fashion. The goal is to achieve high quality care at affordable cost in a reproducible fashion. Of course, this is more a journey rather a destination.

The good news is that other information-intensive sectors have faced and overcome similar challenges. In any industry, once data is gathered and measured, it can be personalized and used in beneficial ways. Retail industry giant Target is a good example. Target’s mobile apps, including Cartwheel, gather customer data and preferences, giving users a very personalized shopping experience. The app has been an enormous success and has been downloaded 40 million times, saving consumers around $1 billion on deals via customized discounts. By smartly using and personalizing data, Target is rewarded with more sales, along with a deepening engagement and loyalty from its base.

Patients should expect that same kind of personalized approach to their healthcare. Providers and hospitals have the opportunity to borrow from the learnings in other industries, rather than reinventing each wheel along the way.

From Data Wrangling to Insights

The ability to capture—and use—data is starting to come together in healthcare. Delivery organizations are making progress in finding ways to improve patient outcomes, value, and quality of care. Initial work has been slow, as hospitals and healthcare organizations work to integrate different types of data from fragmented sources, including research, diagnostic and clinical imaging, and a growing array of Internet of Things (IoT) devices and wearables.

The work of interfacing, cleansing, and integrating this data has been far from trivial. In fact, data scientists and analysts often spend 80% or more of their time in data prep work rather than actual data science, where value is created. This has made analytics time consuming, expensive, and therefore largely reserved for the most pressing and strategic of questions. In essence, analytics has been the purview of the corner office, leaving the front lines to make decisions only with the data they have at hand, supplemented with gut and intuition.

However, evolving standards are easing interoperability. This, together with the speed and simplifying power of in-memory computing, are revolutionizing the use analytics across the enterprise. Data scientists and analysts are able to shift from time-consuming data wrangling to creating insights. This shift is key, since the current approach to analytics simply cannot be scaled to the enterprise level, especially given the shortage of people who are skilled in data science and analysis.

Perhaps even more revolutionary is the ability to leverage a graphical user interface (GUI) as a visual interface to the data, enabling democratization of data-driven decision-making across the enterprise. Embedding analytics in frontline decision-maker’s workflow--and allowing them to easily ask and answer questions--will have as profound and far-reaching impact as any preceding breakthrough in healthcare. Ultimately, with easier access and better data query tools, all types of practitioners and administrators will be able to use data analytics routinely to make informed decisions at the point of care.

This interoperability of data, empowerment of data analysts, and democratization of analytics at scale will power wholesale transformation of care delivery at the system level. As participants become more adept at leveraging analytics in their daily workflow, the next wave of capability leverages machine learning tools, to help healthcare organizations glean deep insights from ever larger, more complex, and disparate data sources. These tools will help them optimize systems and make evidence-based decisions, truly moving the meter toward cost savings and delivery of value-based care.

Focusing on Value-based Care

The healthcare ecosystem cannot move to value-based care without overcoming several challenges. One of the major impediments is the inability to accurately gauge the quality of care on all levels, as providers can’t manage what is not measured. The ability to integrate clinical and financial claims and operational data at the service-line level is an essential building block that we didn’t previously have.

Adding to that is the fact that patients and caregivers are becoming more involved in the management of their health, and hospitals are now looking for ways to provide the best-quality care at a reasonable cost. The ability to set, understand and improve pricing relies on the collective ability to understand care delivery and care variation.

We have already had key wins in data sharing, but we need more. Hospitals need to progress in their thinking beyond holding on to data silos and become more willing to share. Providers are moving toward open data access, as they realize the value of sharing data via the cloud. Healthcare records of all types are increasingly digitized. New data sources, such as genome profiles and personal health information recorded automatically from wearables and other IoT devices are becoming more commonplace. Patients’ electronic health records (EHRs) are also becoming more available, discoverable, and understandable by analytics systems.

Using Analytics to Derive Better Outcomes

Analytics are key to a value-based care system. Real-time business analytics allow providers to react more organically to what is happening. The emergence of predictive analytics allows providers to move from reactive to proactive. With an in-memory computing platform that can be accessed by all departments, users can ask a question, apply their own lens to the data filters, and drill down to find answers. Visualization layers provide GUI-based views of data, making it more accessible and usable to all people in the healthcare ecosystem.

By combining clinical, financial, claims, and operational data, organizations can understand care delivery variances. Usage of unstructured data is increasing. For example, a tumor marker note in a physician file can now be mapped using natural language processing. That information is then usable in the data set that can be compared from one patient to the next.

This improvement allows a physician to use real-world data from a similar case at the point of decision for treatment. Overall, this shift provides better operations and analytics in billing and in hospital management, helping everyone in the health delivery system find inroads into value-based care.

Case Study: Mercy

A prime example of a transition to an evidence-based care model is Mercy, a 40-plus hospital system headquartered in St. Louis. One of the largest Catholic healthcare systems in the U.S., Mercy serves millions of patients annually with acute care centers, specialty hospitals, and 300 clinics in multiple states.

Mercy realized that delivering evidence-based and personalized medicine is crucial to improving care. To do this, they had to fully leverage not only their data, but external data. Mercy needed a health IT infrastructure with integrated analytics that would enable it to support complex business and clinical processes. These initiatives were critical for the transition to a value-based payment environment of improved quality and reduced costs, which led them to On-Demand Patient Data with SAP HANA.

Mercy is now leveraging the SAP HANA platform and analytics solutions process data in real-time, helping to improve patient care and save millions. The working model at Mercy is yielding remarkable results: from $1.2 million in savings for total knee replacement costs in the first fiscal year to $13 million in overall savings in less than two years.

Curtis Dudley, Vice President of Performance Solutions at Mercy, explains that Mercy has now organized millions of records from multiple sources—EHRs, financial systems, and external data. What used to take weeks deliver, it embeds where and when it is needed—back into the EHR or through custom dashboards, which provide on-the-spot exploration for instant answers to the Mercy staff.

Standards for Interoperability

To build functional data-sharing as efficient as Mercy has, a common standard is essential. One emerging example is Fast Healthcare Interoperability Resources (FHIR), which is a draft standard for exchanging healthcare information electronically. FHIR allows for data exchange between healthcare applications and supports multiple data types. It is less expensive than previous models and has a strong focus on implementation. With the capability of providing Web-based queries, the shared data is more accessible to all practitioners.

Automated clinical decision support and machine-based processing requires structured data formats, which FHIR can provide. Machine-based learning and artificial intelligence (AI) tools will augment the value of collected data sets even further. Machine learning helps analysis, as machines can look faster, deeper, quicker and learn from massive amounts of collected healthcare and research data.

What’s Ahead

For health providers to inch toward a value-based system, we need data to determine appropriate care and cost of care models. You can’t manage what you don’t measure--it is impossible to measure quality of care without data. In addition, the value model is a large disruption to the healthcare delivery system, and subsequently has inherent risks.

As we leverage in-memory computing and visual tools, the data engineering component shrinks. When that happens, we can democratize the ability to leverage data to make better decisions in all areas of healthcare: practitioners, business, insurers, and hospitals. With a focus on data creation and self-serve analytics, we can get positive results for all pieces of the healthcare ecosystem.