10th annual utah's health services research conference - a data model for representation and...

12
A Data Model for Representation and Storage of Biomedical Data Quality Naresh Sundar Rajan MS, Biomedical Informatics Core (BMIC), Department of Biomedical Informatics, University of Utah. This work has been supported in part by the National Center for Research Resources award UL1RR025764 and the Agency for Healthcare Research and Quality

Upload: university-of-utah-patient-centered-research-methods

Post on 07-Aug-2015

87 views

Category:

Health & Medicine


0 download

TRANSCRIPT

Page 1: 10th Annual Utah's Health Services Research Conference - A Data Model for Representation and Storage of Biomedical Data Quality. By: Naresh Sundar Rajan

A Data Model for Representation and Storage of Biomedical Data Quality

Naresh Sundar Rajan MS,

Biomedical Informatics Core (BMIC), Department of Biomedical Informatics,

University of Utah.

This work has been supported in part by the National Center for Research Resources award UL1RR025764

and the Agency for Healthcare Research and Quality award HS019862.

Page 2: 10th Annual Utah's Health Services Research Conference - A Data Model for Representation and Storage of Biomedical Data Quality. By: Naresh Sundar Rajan

Overview

• Data Characterization• Data Characterization Model• Architecture• Quality Analysis Framework• Examples• Dimensions• Preliminary Conceptual Model

Page 3: 10th Annual Utah's Health Services Research Conference - A Data Model for Representation and Storage of Biomedical Data Quality. By: Naresh Sundar Rajan

Data Characterization

• Clinical research studies such as Health Services Research (HSR), Comparative Effectiveness Research (CER) etc., rely on EHR data.

• EHR data is prone to data quality issues• Need for systematic and generalizable

methods to characterize data.

Page 4: 10th Annual Utah's Health Services Research Conference - A Data Model for Representation and Storage of Biomedical Data Quality. By: Naresh Sundar Rajan

Data Characterization Model

• Clinical studies might require the federation and integration of data from multiple data sources to create large cohorts.

• Addressing multi-source data characterization requires a common representations that support the semantic and syntactic differences in data sources.

• In order to make this representation computable, these quality assessments need to be stored in a data model that comprises all the dimensions of data quality.

Page 5: 10th Annual Utah's Health Services Research Conference - A Data Model for Representation and Storage of Biomedical Data Quality. By: Naresh Sundar Rajan

Architecture

Query Tool

MetadataRepository

VIRGOQuality Analysis

ADAPT

ADAPT

ADAPT

ADAPT

Counts&Data

Secur

ity

Security

TerminologyServer

DataSources

Page 6: 10th Annual Utah's Health Services Research Conference - A Data Model for Representation and Storage of Biomedical Data Quality. By: Naresh Sundar Rajan

Quality & Analytics Framework

Quality Service

Quality Analysis Repository

Data SourcesAdapters

TerminologyServices

MetadataServices

Page 7: 10th Annual Utah's Health Services Research Conference - A Data Model for Representation and Storage of Biomedical Data Quality. By: Naresh Sundar Rajan

Example – Categorical Variable - Completeness

• Completeness – Extent which data are not missing and is of sufficient breadth and depth.

Data Sources

Quality Service

OpenFurther

Page 8: 10th Annual Utah's Health Services Research Conference - A Data Model for Representation and Storage of Biomedical Data Quality. By: Naresh Sundar Rajan

An Example – Continuous Variables.

• Serum creatinine level on a random sample of synthetic data.

• Example Dimensions that fall under: Anomaly, Correctness, Accuracy.

Page 9: 10th Annual Utah's Health Services Research Conference - A Data Model for Representation and Storage of Biomedical Data Quality. By: Naresh Sundar Rajan

Data Quality (DQ) Dimensions

– Literature survey for various data quality dimensions and concepts.• >1100 research articles reviewed.• Dimensions/Concepts extracted manually by

reviewing.

– Issues such as Accuracy, Completeness, Timeliness, Believability, Objectivity, Volume of data, and etc.

– About 50 dimensions extracted from literature.

Page 10: 10th Annual Utah's Health Services Research Conference - A Data Model for Representation and Storage of Biomedical Data Quality. By: Naresh Sundar Rajan

DQ - Dimensions

Page 11: 10th Annual Utah's Health Services Research Conference - A Data Model for Representation and Storage of Biomedical Data Quality. By: Naresh Sundar Rajan

Conceptual Model

• Dimensions computable conceptual model

Page 12: 10th Annual Utah's Health Services Research Conference - A Data Model for Representation and Storage of Biomedical Data Quality. By: Naresh Sundar Rajan

Thank You!