Following several years of intense debate and industry saber rattling, healthcare reform became the law of the land, hinging on personal and professional mandates as well as copious use of electronic capabilities for improved access, data collection and analytics.
Under the auspices of President Obama’s healthcare initiatives and finally reaching predecessor President Bush’s electronic health record deadline in 2015, is the healthcare industry where it was envisioned? Where did it go right? Where did it go wrong?
To examine the healthcare industry’s progress to date with clinical data analytics, Health Management Technology reached out to a variety of industry experts to share their insights.
HMT: Clinical data analytics are used to measure trends in disease prevalence, the effectiveness of care management programs and identifying population risk profiles. How much progress have healthcare organizations made in these areas within the last year?
Foster: Progress in this area has been slow. Some providers are dealing with EMR and HIS that don’t share data well with systems from other vendors, limiting their ability to have a truly full view of the populations for which they are taking on risk. Additionally, data analyst and data scientist shortages are impacting the ability of healthcare organizations to leverage their valuable data assets for transformation. A survey published [in early 2014] showed that the largest healthcare organizations are concerned about their ability to conduct the deep analysis needed under value-based reimbursement and population health management, both from a technology and a talent standpoint. However, the mid- and small-sized providers think they are well prepared. This suggests that the mid- and small-sized providers may be underestimating the complexity they are facing.
Bailey: Progress has been made, but there is a long way to go. This would be an easier question to answer if there was a common definition of what constitutes “clinical data.” Unfortunately, the market is overwhelmed with a perspective that claims data is clinical data and can have the same impact. Claims data alone cannot. Healthcare must progress beyond this.
Still, the market relies too heavily on quality measures (sometimes claims-driven, sometimes EHR-driven) that are not true outcomes and really only indicators and proxies for true outcomes. Healthcare still struggles to measure outcomes that really matter in business, which are a sense of wellness or wellbeing, functionality and functional status, time to return to work, and activities of daily living (ADLs) and quality of life, for example. These true patient-reported outcome measures are critical inputs to creating a local and trusted risk model at the health-system level.
Yet incremental progress has been made in the march toward ideal value, as demonstrated by publicly reported Medicare Shared Savings Program (MSSP) results and an increase in the number of public and commercial value contracts. Unfortunately, even with these two examples there is still more work to do on measuring what really matters to healthcare consumers and creating real value in the business.
Aminzadeh: There are several crucial data characteristics that ultimately determine how we can develop effective clinical analytics capabilities. These characteristics include, but are not limited to:
- Complete patient profile: Do we have access to 100 percent of healthcare data for a given individual across various healthcare settings?
- Comprehensive clinical and cost information: Do we have access to useful clinical (EMR) and cost information?
- Timelines: How quickly can we have access to data?
- Population perspective: Do we have access to data for all individuals for a given population?
- Data standards.
- Methodology.
Health plans were pioneers of clinical data analytics because of their access to complete patient claims history for the entire population. However, historical claims data lack clinical depth and timeliness. On the other hand, earlier versions of clinical workflow tools (such as EMRs) lacked effective analytical capabilities, and as a result, the development of analytical capabilities in clinical settings was significantly delayed.
Historically, clinical data analytics has evolved into two disparate analytical silos:
- Point-of-care perspective: This focuses on clinical settings and supporting clinical decision making at the point of care.
- Administrative perspective: Population-based analytics focuses on organizational settings and supporting administrative decision support, population-based health and program management.
Over the past several years, we have managed to overcome many of the structural and technology barriers (i.e., connectivity, lack of analytical capabilities and data access) to effective clinical analytics via changes in reimbursement policies, financial incentives, regulations and advances in informatics. These improvements have intensified over the past 12 months, allowing organizations to integrate both point-of-care and administrative perspectives in a comprehensive set of clinical analytics. For example:
- Disease registries have become a common functionality for most clinical and population health management tools, and the accuracy and speed of these registries has improved significantly.
- The development and availability of clinical outcomes is shifting the focus of program evaluation efforts from using process-of-care measures to true clinical outcomes.
- Analytical advances, such as the deployment of machine-based learning algorithms and Big Data techniques, are making the deployment of real-time predictive models for populations and individuals feasible, shifting the focus from retrospective to prospective and predictive.
- A new generation of clinical analytics capabilities designed to understand patient behaviors is enabling organizations to improve patient engagement and increase adherence. These capabilities assess a member’s clinical, utilization, psych-social and consumer profile in the context of their engagement with the healthcare ecosystem and provide guidance on what the healthcare organization can do to address the member’s specific barriers to engagement.
HMT: What strategies and tactics will it take for them to progress even more in 2015?
Foster: There are three keys to making progress relative to healthcare analytics:
- Data discovery tools that help end users explore data for root causes and unique correlations that might be missed in typical reporting. These tools also help end users visualize the story that the data is telling them for greater impact and understanding.
- Enterprise data governance. It is imperative that healthcare organizations move analyses out of individual departments and into an enterprise governance structure so that full organizational impact can be assessed through the same data set for every part of the organization.
- Expert consultants with technology, analysis, change management and healthcare expertise. With talent shortages and complexity as key issues related to healthcare analytics, consultants are critical to a healthcare organization’s success. Additionally, all of the analysis in the world cannot create impact unless changes are made based on the stories the data tell. Experts can help guide change in resistant organizations.
Bailey: More important than measuring trends in disease prevalence, care management effectiveness and identifying population risk profiles is taking a broader approach and focusing on value as a network. This is required in order for healthcare organizations to make more progress in 2015. Real value is equal to meaningful outcomes over the true cost of care, where meaningful outcomes are not proxies and costs are not charge-based. Value is also about creating high-functioning teams across the care continuum focused on: key disease and prevention programs, meaningful outcomes for every patient, true costs in support of comprehensive episodes of care, integrating care across all settings, clinical network growth and optimization, and implementing enabling technology.
One of the biggest challenges in meeting value objectives is having all essential and longitudinal data. There are business, clinical and technical aspects to this challenge. On the business side, data-sharing agreements and resources are needed. On the clinical side, validation and buy-in are both key. Finally, on the technical side, many organizations rely too heavily on subsystems for measurement and monitoring. Enterprise data modeling is essential. Most organizations are just beginning to recognize this need and implement central data repositories for their operational data needs and enterprise data warehouses for their analytic needs supporting their health system.
Aminzadeh: Organizations need to continue to:
- Eliminate or address structural barriers such as linking additional, disparate and member-centric data sources.
- Invest in emerging analytical and Big Data techniques such as the ability to use both structured and unstructured data sources.
- Develop and deploy analytical methods that can address emerging analytical fields such as patient engagement and behavioral analytics.
- Learn how to take advantage of non-traditional data sources such as social networks, wearables, consumer data and mobile apps.
While claims and/or clinical data provide a rich view of a member’s utilization patterns and clinical history, other data sources provide a different and very telling view of a member’s behavior. For example, call-center data may provide a view into a member’s engagement with the health plan; consumer data may provide a view into a member’s interests, buying patterns and household structure; and physician data may provide insights on a member’s relationship and proximity to their doctor. All these data sources can work together to provide insights into the patient that go well beyond traditional clinical assessments, and they can provide important clues into a member’s barriers to health.
HMT: Collecting claims-based data, which highlight actual utilization, as well as clinical data, which focus on individual and collective physiology, may not be enough without integrating the two. How successful have healthcare organizations been with this integration, and what will it take to drive them in 2015?
Foster: Some organizations are being extremely successful in integrating claims and clinical data to impact patient and financial outcomes. These organizations have implemented technology tools that can consume multiple data sets from virtually any source, and have implemented enterprise-level data governance. Organizations that keep data analysis at the departmental level will progress more slowly (or not at all) as department heads choose data that meet their objectives rather than setting objectives based on the story that an enterprise-level, single-source-of-truth data set tells.
Bailey: Healthcare is just scratching the surface of the possibilities here. Some larger challenges include:
- What does “integration” mean? The audience for the analysis is important to consider. Staff that interacts with patients on varying levels require actionable outputs from analyses to close the loop, whereas executives and business intelligence teams require an ability to look for patterns and trends in data.
- Claims and clinical data are insufficient to model value. Cost and outcomes data must also be included. Cost data require a data-driven approach to measure and monitor true cost. Outcomes data requires instruments such as those from the International Consortium of Health Outcomes Measurement (ICHOM) and Patient Reported Outcomes Measurement Information System (PROMIS) and must be automated to be used for a true outcomes analysis.
The real focus for data integration should be on using all data possible to tackle the huge waste in healthcare while improving outcomes. This must be a core business focus of health systems as it is a primary concern for all payers and the impetus driving value-based contracts. Over $700 billion spent on healthcare annually is considered waste (overuse, misuse, variation, inefficiency, harm, etc.). Waste is spending that could be eliminated without harm or reducing quality.
Aminzadeh: The integration of clinical (i.e., EMR, biometric devices) and non-clinical (claims, consumer, service) data is critical to developing and deploying effective clinical analytics capabilities. The degrees of success vary significantly across various healthcare organizations and, unfortunately, the majority of healthcare organizations have been marginally successful in integrating claims and clinical data to date. That being said, I’m optimistic about incremental improvements in this area in 2015 and 2016. There have been an increasing number of significant and in-depth collaborations between health plans and health systems, which has necessitated collaborations for data sharing and data integration. Until recently, there has not been a strong and compelling business case for this type of integration. However, because of changing reimbursement policies, emerging delivery models, revenue sharing and at-risk business relationships, there’s a greater need to operate as transparently as possible, while also using the richest data sources to positively manage patient health.