Diagnosing Medical Data

Are Data Warehouses The Cure For More Efficient Healthcare?

Even as modern healthcare becomes more complex, we want our healthcare providers to bemore accurate in predicting who's at risk of harboring and spreading disease while stilllowering costs. Physicians, hospitals and insurance companies are turning to datawarehouses for help.

The consolidation of hospitals, physician practices, clinics, pharmacies, home care andlong term health facilities is making the business of healthcare difficult. The complexnature of healthcare warrants the necessity of conducting retrospective, prospective andreal time analysis on integrated, accurate and reliable clinical and financial data.

There are many applications that enable such healthcare measurements, however, theeffectiveness of these decision support and business intelligence applications is whollydependent on the information that feeds them. Therefore, the data warehouse is critical tothe successful implementation and use of most healthcare applications.

These applications must enable physicians and other healthcare executives to conductimportant healthcare measurements, such as the following: Outcomes/Quality of CareManagement; Benefits Management;

Physician Profiling; Managed Care Contracting; Case Management and Referral Management;and Wellness Management.

The healthcare industry, like many others, is reliant upon many different kinds of data-- enrollment, clinical outcomes, medical records, pharmaceutical, claims, eligibility,geographic, demographic, contracts, etc -- that reside in numerous disparate systems.These disparate systems, while sufficient for their intended use, must be integrated inorder to enable healthcare organizations to conduct retrospective and prospectiveanalysis.

For example, a clinical system most often will not facilitate detailed financialanalysis on a case-by-case, episode of care, or per physician basis; conversely, afinancial system does not enable the comparison of clinical outcomes to cost. The natureof this stovepipe architecture limits the capabilities of all decision support andbusiness intelligence applications. For example, decision support tools currently useclaims, enrollment, financial and pharmaceutical data for retrospective analysis, butagain, all of this data resides in different systems.

A Healthcare Retrospective

Utilization management, physician profiling and outcomes/quality of care management areall types of retrospective analysis. Prospective analysis requires some retrospectivecomparisons, as well. This type of analysis simply means that you are viewing andcomparing episodes of care that have already occurred for the purpose of determining thebest quality of care with the lowest cost, most economical and effective physicianpractices, greatest outcomes of specific treatments, etc.

Retrospective analysis tools rely on the following data: decision support data(including claims, drug and eligibility data), clinical outcomes and medical recordsresults, pharmacy claims process system output, etc. Retrospective systems allowclinicians and other healthcare executives to take historical data and analyze past trendsbased on the cost and use of healthcare services.

While the retrospective tools on the market are adequate for their intended use, almostall vendors in this market utilize a subset of fields populated from operational systemsstored in a data mart. Programmers supporting National Committe for Quality Assurance(NCQA) and Health Plan Employer Data and Information Set (HEDIS) requirements find itnecessary to use information locked in multiple disparate systems. Therefore, withoutaccurate, integrated information, these analysis tools cannot be fully effective.

Retrospective analysis most likely will remain a part of the healthcare landscape, butthe trend is towards disease management with long term goals of managing demand andpredicting potential cost based on populations of people. Systems and applications willfollow this trend, as well.

Demand management allows a health plan to apply predictive modeling algorithms toidentify patient populations with medical characteristics that indicate potential futurehealth issues. Insurance organizations are able to identify members with potential healthrisks and begin managing their care with preventive programs.

This paradigm shift will eventually lead to improved quality of care and lowerhealthcare costs, heading off potential diseases with less expensive, preventativetreatments.

Good Prospects

Disease and, especially, demand management rely on prospective analysis. Prospectivesystems allow health plans to predict future health problems of its population, allowingfor early intervention strategies that will affect the future demand of health care.Health executives are able to predict which members are predisposed to developing certainconditions.

However, organizations must develop the infrastructure to properly manage theinformation necessary for prospective analysis and most organizations are not far enoughalong the healthcare continuum to properly utilize the models.

For example, when specific members are determined to be at risk of acquiring ordeveloping certain medical conditions, how is that list of members handled? Who within theorganization will manage those cases and what course of action should they take?

Similar to retrospective systems, prospective applications rely on many different typesof data -- demographic, geographic, member satisfaction surveys, SF36 based on theStandardized medical outcome questionnaire, market analysis, outcomes, historical trends,etc. Again, the enterprise data warehouse is the most effective system for integrating thenecessary data to feed the business intelligence and decision support applications usedfor the many kinds of analysis.

How Can a Data Warehouse Help?

Often times, decision support and business intelligence application vendors create datamarts populated with predefined types of information to feed their specific applications.Instead of eliminating the issue of disparate systems and inaccessibility to data, thisperpetuates the problem, only creating more systems that cannot talk to one another.

Therefore, the data in one data mart may or may not be useful to a different decisionsupport or business intelligence application.

An enterprise-wide data warehouse helps eliminate the recreation of the stovepipe,disparate system architecture in which numerous data marts are built to support individualapplications. The warehouse combines all of these systems, enabling the businessintelligence and decision support applications to pull from one source, not from a limitedsubset of data that populates a data mart. Therefore, a retrospective application and aprospective application could be fed from the same source -- the enterprise datawarehouse.

It is also possible to create an enterprise infrastructure in which smaller data martsreside, so that each mart pulls the same set of integrated financial, clinical andoperational data. The meta data in the enterprise warehouse ensures that business rulesare applied to data and that the data feeding the warehouse is updated as often asnecessary.

-- Jane Griffin is the vice president of Business Intelligence Solutionsand director of the Business Solution Center of Excellence at Prism Solutions, Inc. inAtlanta, GA.