Advancing Healthcare through the Application of Predictive Knowledge Management

 Background

What is Predictable Healthcare? A predictable healthcare organization is one that has progressed from rote data capture and regurgitation to one that uses, reuses, and massages information to identify probable outcomes, enabling the healthcare entity to “act” in a more predictive fashion throughout the care process (before, during, and after a patient care experience).  One primary driver fostering the development of Predictive Knowledge Management (PKM) is the move toward transparency within the industry. Transparency is pushing the migration of healthcare organizations from simply collecting independent data events to creating coherent interdependent information environments that must actively support increased quality and patient safety, improve health outcomes, and enhance the patient experience across the entire healthcare continuum.

Where is Healthcare Today? In many respects, the move toward PKM in healthcare is an inevitable outcome of the move toward creating electronic databases.  Other industries (e.g., retailing, finance, automotive) have adopted predictive modeling techniques as a standard practice.  The platform for moving in this direction across all of these industries has clearly been the adoption of electronic information capability related to core functions within the specific organizations of the industry.

For healthcare, the core functions related to clinical care are the foundation for the evolving PKM capabilities of the industry.  Unfortunately, due to the historically paper-based medical environment and the backlog of nonintegrated information residing in healthcare organizations, the move toward implementation of effective PKM will be both costly and time-consuming.

To meet the needs of predictive modeling, healthcare information must evolve from the simple sequential capture of discrete pieces of data to a model that facilitates the gathering of integrated data into information packages.  Such an approach supports the creation of knowledge, which can eventually be used for driving improvements in the care process.

How Do Healthcare Organizations Achieve PKM?  Enabling this PKM evolution requires a delivery platform that assimilates several information technology stages that include the transition from disconnected applications with independent functions and capabilities.  The stages include the following:

• Connecting business strategy where visibility into specific applications, people, and partners is feasible
• Integrating business and information models where the end‑to‑end processes are managed
• Eliminating time delays of real-time operations in handoffs associated with the processes
• Modeling on a predictive basis the process requirements and acts that will optimize care delivery opportunities and prevent the occurrence of problems 

A number of environmental forces are precipitating the need for adoption of a predictive modeling capability by healthcare organizations.  The following is a brief overview of the forces that are contributing to the development of PKM.

What is the Future of PKM in Healthcare?  Many forces commonly addressed today—such as globalization of healthcare services, public dissatisfaction, the move to consumerization, clinical workforce shortages, as well as outcomes reimbursement and other financial issues—are precipitating a notable requirement for new discovery within the healthcare industry.  With the advent of the electronic health record, new opportunities for uncovering patterns of care—patterns of care we did not even know were in existence—will come to the forefront of medical knowledge.  As the healthcare industry continues its drive toward enhanced quality of care, promotion of better service, and reduction of cost, these undiscovered patterns of care will become increasingly transparent, first for physicians, nurses and other clinicians, and ultimately for all consumers of healthcare.

Provider Community.  While global trends will be influential on all segments of the healthcare industry, several key factors are evident for the provider community.

The approach and venue for delivering care will change:

• A much deeper understanding of the actual care process—along the entire continuum— will be required to support change efforts.
• Consumers will have a much larger voice in healthcare and demand more and better information on the outcomes of care.
• Businesses will no longer tolerate annual increases in healthcare expenditures that exceed general inflationary trends.
• Healthcare providers will shift their focus from the back office to the clinical setting to extract value.
• Demand for services will increase as the population ages.

For providers, these factors are driving an increasing recognition that more and better information is required to manage the “care process.”

Why?  First, in today’s healthcare world, much of the inference extrapolated on the quality of care is derived from secondary data.  Secondary data is interpretive in nature and consists primarily of financial information and billing code data.  Essentially, the ability to derive good information on care delivery patterns is extremely limited and, from a clinical perspective, nonexistent.  While such an approach to data extrapolation has been acceptable in an era of paper-based systems, in an era of available real-time, integrated, granular clinical and business operations information systems, such an approach is not only unacceptable, it is deemed archaic.  Therefore, the growing digitalization of clinical data and deployment of Clinical Information Systems (CIS) creates a requirement for a deeper understanding of the care process and demand for more detailed analytic services.

Second, extracting such clinical information from disparate legacy healthcare information systems can be extremely difficult.  Issues related to lexicon, semantics, normalization of data, and a host of other technical and standards-derived considerations must be managed to effectively compare data sets between disparate systems.

Third, from a pragmatic perspective, it is the rare healthcare organization that has deployed a single integrated health or clinical information system to meet all information needs.  In addition, most healthcare organizations have not fully depreciated their investments in information technology and, as a result, they cannot simply engage in wholesale replacement of existing siloed legacy systems with integrated information systems.  Therefore, these organizations frequently require an approach that ties disparate information together from a wide variety of sources to create a much more robust picture of care across a diverse, often geographically disconnected, healthcare delivery organization.

Notwithstanding an increase in the use of information, healthcare is still far behind other industries in creating integrated, longitudinal, client-focused databases that can serve as repositories for data mining and analytics.  At the same time, the quality-service-cost triad has driven many other industries to adopt new methods for driving “business intelligence” from massive stores of available data.  Healthcare is only now beginning to apply these proven intelligence‑gathering technologies, which will create the PKMsystems of the future. In the changing electronic world of healthcare, the abundance of massive data sets is creating the imperative among providers who increasingly recognize the value of data mining as a tool for driving better outcomes.

Payer, Pharmaceutical, and Medical Device Communities. The trend toward business intelligence is also affecting the Payer, Pharmaceutical, and Medical Device (PPMD) segments of the industry.  For the insurance industry, it is increasingly recognized that disease management models provide significant value-add contribution to enhanced service and reduction in costs.  Such models require, a priori, extensive data on the results or outcomes of care delivered by healthcare organizations and providers over time, which includes the availability of longitudinal, patient-centric information.  Again, analytics on the care process has proven to be an invaluable resource for enhancing quality, increasing patient safety, promoting consistency, and reducing overall costs.

In addition, the market discrimination required by the PPMD segments of healthcare involves effective analysis of medical, health, and demographic trends over larger populations, both for purposes of understanding the impact of their products, as well as for driving unit sales.  Intense marketing competition, shrinking margins, escalating research costs, increasing sales costs, and rising marketing costs are all contributors to the need for better business intelligence by the PPMD healthcare segments.

Companies are, therefore, seeking competitive advantage through data that gives their product an “edge,” or helps their sales and marketing effort fine-tune their respective strategy and message.  Traditional data sources cannot deliver the types of information required for discovering new, care-related competitive leverage points.  As an example, the pharmaceutical industry has historically used clinical data from sources such as medical chart abstractions, claims, and other publicly available data sources (e.g., Medicare database)—all secondary data sources.  These secondary data sources represent extrapolations from actual clinical data.  Extrapolated data excludes the causal detail that has immense analytical value.  In addition, these data sources have not traditionally provided a longitudinal view on the patient.  As a result, data analysis frequently fuses “extrapolated” databases to create additional “extrapolations” derived from nonintegrated sources (e.g., tying independent lab data to independent pharmacy data to independent radiology data, and so on). The best currently available sources of patient-focused data share common deficiencies:

• The data is generally old (i.e., > 6 months) when researchers finally gain access to the information.
• The most widely used data sources are based on insurance claims from pharmacies and payers; and, represent singular “encounters” with the healthcare system, rather than an integrated picture of a patient’s total medical experience.
• Diagnosis codes and other common identifiers are frequently unreliable.
• Clinical details are missing.
• Data is frequently of a poor quality because it is manually derived.
• Data on use patterns within a hospital is inadequate (for pharmaceuticals) or completely unavailable (devices).

Integrated data sources, on the other hand, provide the opportunity for these companies to conduct research related to pharmaco‑vigilance, pharmaco-economics, sales and marketing, drug research and development, and post-marketing activities, among other research initiatives.  Patient-level data that can be longitudinally analyzed—from the point where the patient presents with a chief complaint to the final outcome of a particular healthcare problem—represents the “Holy Grail” of competitive intelligence for the PPMD segments.

Where Does the Road Lead? PKM in the healthcare marketplace is continually evolving and is not a solution that will happen overnight.  There remains a fair amount of time, hard work, and investment related to infrastructure investment and data integration to make PKM a true reality in healthcare.  However, some significant milestones are making healthcare foresight possible.

First, purveyors of secondary data sources recognize the need to shift toward more patient-centric data.  As a result, many purveyors of solution data sets are in the formative stages of developing more clinically focused database structures.  In addition, clinical information system vendors are also attempting to develop data warehousing products.

Second, the federal government, through a variety of initiatives, is stimulating interest in creating patient-centric data models.  The Agency for Healthcare Research and Quality (AHRQ), Centers for Medicare & Medicaid Services (CMS), the Department of Defense (DoD), the Veterans Administration (VA), the National Cancer Institutes (NCI), the National Institutes of Health (NIH), and a variety of state efforts (e.g., Pennsylvania) are fostering more interest in developing patient-centric systems.

Third, various collaborative initiatives are evolving.  Healthcare professionals, healthcare systems, and other groups are recognizing the power and value of such information in conducting analytic work on their care delivery patterns.  As an example, several specialty hospital associations are examining the potential of a common data warehouse for purposes of research for members (e.g., pediatrics, oncology, and cardiovascular).  In a similar trend, evident among other industries that have already deployed business intelligence initiatives, healthcare organizations increasingly recognize that the more one knows about oneself, the better able they are at directing or predicting their future.

Fourth, regulatory agencies (e.g., The Joint Commission, URAC, specialty accreditation organizations) are seeking models that provide them with quality data on care delivery patterns.  The move toward outcomes and results is fostering the need for more patient data, which is integrated over time.

Fifth, the evolution of the Regional Health Information Organization (RHIO) movement will precipitate a need for PKM initiatives.  As these organizations begin to move patient-centric data on a regional basis, the natural progression is to begin asking the question, “So, what is happening from a care delivery standpoint in our community?”  This question is the next extension of the RHIO movement.  While the movement continues to be in its early growth stages, it will clearly raise issues for local and regional communities on the type of care being delivered for people.

Finally, until very recently, the question, “Who would be willing to pay for quality?”, represented a rhetorical question, at best.  However, the reimbursement landscape is changing very quickly.  Today, it is clear—pay-for-performance incentives, external measurement, and recognition based on comparative performance, risk management, and medical error revelations are clearly forces that provide measurable ROI for quality.  This shift to paying for quality also shows the change in the overall mentality from “the system is king” to “the data is king.”

Continued >