Data Quality in General Practice EMRs - Is it sufficient for building Clinical Decision Support Tool for Type 2 Diabetes Risk?

Project Details

Patron ID: PAT1024

Lead Investigator: Professor Dougie Boyle

An estimated 30% of Australians with type 2 diabetes mellitus remain undiagnosed, despite readily identifiable risk factors for the condition and at least 85% of Australians visiting a doctor each year (1, 2). Recognition of this issue has led to the development of a number of risk assessment tools (such as AUSDRISK and the Framingham model) to identify those at high risk of developing type 2 diabetes while avoiding the burden of unnecessary tests for individuals at low risk (3). However, time constraints in already busy general practice (GP) consultations prevent the widespread use of such manually deployed tools which require further time and data entry (4).

Electronic medical records (EMRs) with inbuilt clinical decision support tools have the potential to navigate this issue, by running risk prediction algorithms on existing patient data to provide in-consult advice to GPs regarding those at high risk of type 2 diabetes. Already, early risk prediction models on sample EMR datasets (such as test datasets or comprehensive hospital research datasets ) have shown increased accuracy as compared to traditional risk calculators by using a wide range of biochemical and demographic data (5). However, as general practice EMRs are optimised for clinical use rather than secondary purposes such as research or predictive algorithms, the data within a non-sample EMR dataset can often be of insufficient quality, incomplete or buried in free text, making it difficult to use for such purposes (6).

The Patron dataset contains a large volume of data extracted from general practice EMR software, including patients diagnosed with type 2 diabetes and information pertaining to their risk factors for the disease. This study aims to look at those data fields and assess them for data quality and completeness. With this information (beyond the scope of the current project), we hope to develop a risk prediction algorithm integrated into general practice software which can provide individual type 2 diabetes risk advice using data values that are more likely to be complete.

References:

1.            Sainsbury E, Shi Y, Flack J, Colagiuri S. Burden of disease for diabetes in Australia: It’s time for more action. Sydney: The University of Sydney, 2018.
2.            Jones JL, Lumsden NG, Simons K et al. Using electronic medical record data to assess chronic kidney disease, type 2 diabetes and cardiovascular disease testing, recognition and management as documented in Australian general practice: a cross-sectional analysis. Fam Med Community Health. 2022 Feb;10(1)
3.            Buijsse B, Simmons RK, Griffin SJ, Schulze MB. Risk assessment tools for identifying individuals at risk of developing type 2 diabetes. Epidemiol Rev. 2011;33(1):46-62.
4.            Chiang J, Furler J, Boyle D et al. Electronic clinical decision support tool for the evaluation of cardiovascular risk in general practice: A pilot study. Aust Fam Physician. 2017 Oct;46(10):764-768.
5.            Riihimaa, Päivi. (2020). Impact of machine learning and feature selection on Type 2 diabetes risk prediction. J Med Artif Intell. 2020 Jun;3(10)
6.            Botsis T, Hartvigsen G, Chen F, Weng C. Secondary Use of EHR: Data Quality Issues and Informatics Opportunities. Summit Transl Bioinform. 2010 Mar;2010:1-5.