Enabling predictive healthcare analytics through better workflows

The shift to value-based care and the establishment of machine learning in support of data analysis is transforming healthcare. We may only be at the start of this journey, but a foundation is being set for predictive healthcare analytics in health systems that will allow physicians and care teams, hospitals, and health plans to mine diverse sets of data to identify patterns in support of advancing improved medical decisions.

A recent survey of healthcare organization executives by Deloitte found that 94% are moving toward value-based care. While only 27% of this group have completed pilots or rolled out technology, the intent is clearly there … and so are the incentives. Quality data reporting provisions of the 2015 Medicare Access and CHIP Reauthorization Act (MACRA) will base reimbursement on how well physicians and care teams access, share, and use data to improve the quality of care they provide.

The foundation for this change was set by the 2009 Health Information Technology for Economic and Clinical Health (HITECH) Act, which established financial incentives for physicians and hospitals that adopted meaningful use of EHR systems. “Meaningful use” has put us on this path.

Investment in EHR systems has been massive. A large hospital can spend up to half a billion dollars implementing these large-scale systems, which are analogous to enterprise resource planning (ERP) systems, the information technology (IT) backbone for transactions in other industries. Between 2012 and 2013, according to statistics from the Centers for Disease Control and Prevention, physicians in the United States increased their use of EHRs by 21%, with 78% using the systems in 2013, up from 54% in 2011. While frustrations with these systems abound, they have played a critical part in our societal move toward data-driven healthcare. We’ve gone well beyond moving data from the filing cabinet to digital formats and to a point where we now can consider data quality and ask:

• Is the EHR providing actionable information?
• Are physicians supported with meaningful clinical protocols that are easily accessible, and are they in a position to receive data on care costs or on clinical care and quality measures?
• How easily can this data be transferred to care teams in other hospitals working in the same area or to other points within the health system?

The need for richer data

In a value-based world of healthcare, care teams will need access to a variety of rich data sets from a multitude of systems both within and outside their own continuums of care. Benchmarking data accumulated from peers offers valuable insight into how any one team is performing on a broader level. In addition, patient data that is accessible across the entire care community supports a shared perspective of an individual’s or population’s health for a more coordinated approach to care.

This call for accessible data will require systems that are easier to use, with improved workflows that enhance, rather than deter, from clinical care. Many physicians find EHR systems frustrating and clumsy and complain that they divert attention away from patient care. According to a 2016 survey of physicians by the Deloitte Center for Health Solutions, 78% of respondents find EHRs most useful for analytics and reporting capabilities (but less important in supporting value-based care or improvements to clinical outcomes). In addition, 75% believe they increase costs, while 70% think they reduce productivity. The research found that 62% would like to see improved interoperability, while 57% would like to see better workflow and productivity in electronic healthcare systems.

When well executed, healthcare IT and meaningful analytics can provide critical insights to care teams, health plans, and patients alike into both the care they give and receive. Vendors supporting value-based care initiatives will need to work closely with end users to find ways to build solutions and tools that can be embedded into the daily practice of the user. If information doesn’t support, and isn’t incorporated into, the end-user’s workflows, at best the technology is not used, and at worst, we’ve made a difficult job even harder.

Less is often more

“Less is more” was first used by architect Ludwig Mies van der Rohe during the minimalist design movement in the 1960s. Later, famed Braun industrial designer Dieter Rams adapted the phrase to “less but better.” As humans, we have limited resources to make decisions before we become overwhelmed. When analytic solutions are at their most valuable, they are taking vast amounts of complex data and simplifying the information into cogent patterns that guide the end user in making decisions. This often requires taking features out of the product as opposed to putting features in. Some basic suggestions for clean and concise analytic dashboards with health-related data might include:

• Highlighting problem areas and allowing users to review those areas first;
• Removing words, replacing them with color, blocks, and arrows;
• Showing progress, rather than entire trend lines (allowing users to drill down if they want to); and
• Communicating and highlighting action areas—such as gaps in care, site of care, or formulary guidance—to facilitate the right decision.

Normalizing messy data

A fundamental technical issue facing product developers today is the need for data normalization. Interoperable engines can pull information from disparate sources into one location, but the data itself, input differently by different users, often is messy. For example, within the EHR, unstructured data on diabetes tests may be written in the physician’s notes but missing when we search for it in a specific data field. Vendors need to work closely with clients to show them where their data and recordkeeping problems live. In some cases, standard ways of inputting information might be needed. In others, some training may be necessary to optimize data collection and subsequent usefulness.

Being able to look through large swaths of information and identify trends is critical. Thus, pattern prediction can inform and improve preventive healthcare. Predictive healthcare analytics can only improve healthcare decisions, but together we must invest in the associated usability that ensures these analytics exist within enhanced workflows for better care decisions.