HIMSS coverage: Warehousing may link structured clinical data

A strategy to link clinical data with other types of medical information in data warehouses was proposed Wednesday at the Healthcare Information and Management Systems Society meeting.

Clinical Gateway combines electronic information stored in a clinical repository with financial information stored in a data warehouse. It was developed by Dr. R. Bharat Rao, head of clinical systems R&D in the Computer Aided Diagnosis and Therapy Group of Siemens Medical Solutions.

"There is a need for an enterprise-wide healthcare data warehouse that includes both clinical and financial information," Rao said.

Rao explained that creating this new class of combined information makes it possible to perform functions and analysis that are currently only possible with financial (billing) data: outcomes analysis, profiling, compliance, diagnosis support, quality assurance, and decision support.

Existing patient records can be a valuable clinical resource. Yet because key clinical information is not available in a structured form, they have limited use for QA, strategic allocation resources, and regulatory compliance monitoring.

Existing healthcare data warehouses are excellent for extracting and storing data from operational financial systems. Their financial data, like charges, are already highly structured and normalized. Operational clinical systems, however, have very poor data quality from the standpoint of access and analysis.

Rao said data in clinical repositories is often messy, and thus only a small fraction of the clinical data goes into the data warehouse, typically the part linked to financial data, such as ICD9 and CPT codes.

"If we can create highly structured clinical data from existing patient records, then the same kind of automated analysis is possible," he said.

Clinical Gateway has two components:
? A Repository Interface provides access to a limited predefined amount of clinical information in the Clinical Repository.
? A Reliable Extraction and Meaningful Inference from Nonstructured Data (REMIND) system produces a high-quality structured clinical database. REMIND uses clinical domain knowledge about the disease in question for this task.

REMIND is a data mining framework that combines all available evidence in a principled fashion over time, according to Rao.

"Our focus is not the extraction of information from individual parts of the clinical repository - we use readily available methods for that purpose - but to combine all available information," he said.