AbstractElectronic Health Record Systems play an integral role in healthcare practice, enabling health organisations to collect, access, and manage data consistently. These data are used to support clinical care ranging from better health information exchange to general improvements in clinical practice, efficient hospital management, and accelerating the development of medical interventions. However, like all information systems, they are prone to data quality problems such as incomplete records, values outside expected ranges and implausible relationships, which have far-reaching impacts on patient safety, data consumers’ confidence, decision-making, and reusability.
This study aims to improve health data quality by exploring whether current efforts can be more effective if backed by appropriate technology in intelligent systems. In particular, it focuses on the identification of problems in health data, which studies note is often non-trivial, and it proposes a conceptual framework for conducting systematic assessments and evaluating the trustworthiness of data quality assessment programs. A robust automated tool for assessing health data quality, the Healthy- Data Toolkit, has also been developed to evaluate the utility of the proposed framework. The tool includes a novel method for enabling interoperable data quality assessments based on loose coupling and automated schema matching. Also, a systematic method for acquiring reference measurements for facilitating data quality assessments in the absence of gold standards was introduced. The validity and reliability of this implemented tool and methods have been evaluated using real-world datasets.
Overall, this study envisages that the framework proposed can help reduce the confusion surrounding health data quality and enable reproducible and credible assessments, which is essential to improving data quality practice, whether it be root cause investigations, documentation training or data cleansing. Also, this study anticipates that the implemented tool will benefit researchers and organisations that do not have dedicated teams or are unsure how to approach data quality assessments in preparing datasets for meaningful use and shedding more light on data quality problem types that need attention, which can help shape their improvement strategies.
|Date of Award||21 Jul 2022|
|Supervisor||Adrian Hopgood (Supervisor) & Philip James Scott (Supervisor)|