EBM and Data Warehousing

Introduction
Evidence based medicine (EBM) replaces the traditional model of “medicine by authority” with a scientific model that avoids the use of underutilized and unsystematic information [1]. It encourages healthcare professionals and managers to be aware of and make use of published peer research. EBM requires that decisions makers have access to methodically compiled research information. It enhances knowledge and speeds the introduction of new therapies and withdrawal of ineffective ones. Thus, EBM represents a knowledge management system in the world of healthcare. Knowledge comes from information, which results from processing data. In the current age, information come a variety of sources including published scientific and technical literature, clearinghouses for public domain information, and internet. Information technology contributes to EBM in the following ways [2]:
•Reference databases – contains formal publication of clinical trials and reviews. Conventional databases are not usually effective in providing contextual, timely, precise information sought by physicians. However, entities like the Cochrane Collaboration Library do a good job of meeting the clinician criteria [3].
•Contextual and case-specific information – collected from current and past encounters with the healthcare system. Usually it consists of transaction data with the insurance provider and health provider’s administrative information system. It also accounts from the epidemiological and social factors for a particular site [4].
•Clinical and administrative data repositories – Data captured at the point of care through hospital information systems and electronic medical records
•Decision support software – Creation of clear and reproducible rules to facilitate clinical decision making
•Internet based interactive health information – Contains multimedia resources, unstructured content, unendorsed, and non-validated data.

EBM Data Warehouse cardinality
Evidence based medicine needs the support of organizations that provide means to search, store, and retrieve contextual patient relevant data to the clinician in a timely manner. This translates to the creation of a place where data is constantly maintained and updated, classified and populated to study variables uniformly under a nosological or classification system [5]. Using various data mining techniques the medical community can understand patterns of care, causal pathways, profile best practices, and create benchmarks. According to Kimble and Inmon (1996), the DW is Subject-Oriented, Integrated, Time-Variant, Non-Volatile data in support of management decisions. Data warehouse is also not only a product or a collection of subsystems, but also a process. The following table shows the relationship between
EBM functions and data warehouse

EBM feature------------------------------>Data Warehouse feature
A diagnosis needs to be established------>Query
A therapy mode or care pathways needs
to be determined------------------------->Query
A prognosis is needed-------------------->Prediction
Possible causes need to be verified------>Pattern recognition

An Architectural proposition
For the remainder of the paper, we will discuss the data warehouse part of the EBM criteria. Using the corporate information factory model [7] (Refer to illustration A) as a baseline, a nationally managed central enterprise data warehouse is not a feasible solution. EBM needs the following
•Information integration from heterogeneous systems
 Today, several repositories and data marts exist within the boundaries of institutions. The critical success factor is to have them share the data keeping in mind the questions of privacy, software regulation, and ethical and legal aspects of telecommunication in healthcare.
 Neutral standards like Simple Object Access Protocol (SOAP) combined with Service oriented architecture framework provide a paradigm to integrate information seamlessly
•Standardized facades for information exchange
 The single most important factor in information integration is the contract for information exchange. All institutions need to model their repositories off a reference implementation, which will have a skeletal, minimal set of fact and dimension tables like Patient, Clinic, Clinician, Drug, Therapy, Diagnosis, Prescription, and Encounter [8].
 From a governance standpoint, key stakeholders of medical data repositories must be accountable to participate in scheme of Electronic Health Record (EHR). EHR lays the foundation for inter institutional exchange of clinical data.
•A nationally managed online gateway to a network of repositories
 Bill Inmon defines this as a “virtual DW” which has the ability to farm out a query serviced in parallel by two or more distributed databases, aggregate and join results from those databases, and deliver a unified result set to the requester.

Hurdles
There are “hard” and “soft” hurdles for data warehouse based EBM. Some of them are below:
•Complex nature of the of the care delivery which includes interdependence of staff skills, clinical equipment, patient risks, guidelines and drugs and possibly other factors (Refer to illustration B)
•Need for continuous assessment of data warehouse rules and reconciling the outcomes and metrics with the medical advisory committee. This is a labor-intensive task.
•Need to alter physician mind mindset regarding EBM as cookbook medicine
•High volume of data and intricate integration needs bring forth technology and governance challenges.
Conclusions
Even if a national EBM program is technically available, getting the healthcare community to use it, is different matter altogether. Branded as cookbook medicine, health professionals feel that EBM will curtail their individual autonomy to practice. To the hospitals, the patient is market share and sharing patient information is counter intuitive. The patient is worried about privacy issues. Evidence based Practice Center program launched by the Agency for Healthcare Quality and Research has identified organizations to participate in its program. It has identified the priority conditions like cancer, diabetes, dementia, Ischemic heart disease, stroke, and hypertension to name a few. With the introduction of the new “meaningful use guideline” electronic data collection will be faster. The key to effective use of EBM rests on the ability to integrate disparate silos of knowledge built so far. Some of the recommended steps for major health and insurance providers would be to follow standards for information exchange, allow information query, mask patient identifiable information, create new data mart provide wrappers for existing ones, create IT governance organization, and create compliance programs. There is cost considerations involved in participating in these programs. As new incentives emerge for knowledge sharing and as technology-costs continue to come down, medical data warehousing or EBM will move closer to being a practical and viable solution.

References
[1] Evidence-Based Medicine Working Group. Evidence-based medicine: a new approach to teaching the practice of medicine. Journal of the American Medical Association, 1992, 268: 2420–2425.
[2] Information systems: the key to evidence-based health practice. Roberto J. Rodrigues. Bulletin of the World Health Organization, 2000, 78 (11)
[3] The Cochrane Collaboration. Valuable resource for family physicians. Becker L. Canadian Family Physician, 1997, 43: 403–404, 412–414.
[4] Enkin MW, Jadad AR. Using anecdotal information in evidence-based health care: heresy or necessity? Annals of Oncology, 1998, 9: 963–966.
[5] Healthcare Informatics Research: From Data to Evidence-Based Management Thomas T. H. Wan Received: 25 March 2005 / Springer Science + Business Media, Inc. 2006
[6] Kimball. R. and Inmon, W.H. (1996). The Data Warehouse Toolkit. John Wiley: New York
[7] Corporate Information Factory. Bill Immon. http://www.inmoncif.com/library/cif/
[8] Towards a sustainable data warehouse for evidence based medicine.Vienna University of Technology, Nevena Stolba
[9] The relevance of data warehousing and data mining in the field of evidence-based medicine to support healthcare decision making. Nevena Stolba and A Min Tjoa
[10] Sen, A. and Sinha, A. P. (2005): A Comparison of Datawarehousing Methodologies, Communication of the ACM, 48(3), 79-84