Ford, Elizabeth, Lupton, Grace, Rooney, Philip, Oliver, Seb and Cassell, Jackie (2018) Development of a model for finding unlabeled cases of rheumatoid arthritis in UK primary care patient records. In: Medical Informatics Europe 2018, 24th-26th April 2018, Gothenburg, Sweden.

PDF (short paper) - Accepted Version Restricted to SRO admin only Download (98kB)

Abstract

When using electronic patient records (EPR) from UK primary care for research, it is not possible to tell the difference between “negative” and “positive, but unlabeled” cases. Using the exemplar of rheumatoid arthritis (RA), we developed a logistic regression model which could be used to identify cases of RA which are unlabeled. Combining symptom, referral and test information from codes and free text, our model discriminated between RA cases and controls with an AUROC of 0.923. This method for identifying “positive, unlabeled” cases in patient records has the potential to improve case ascertainment for a range of EPR studies.

Item Type:	Conference or Workshop Item (Paper)
Keywords:	Patient records, primary care, rheumatoid arthritis, data quality
Schools and Departments:	Brighton and Sussex Medical School > Primary Care and Public Health School of Mathematical and Physical Sciences > Physics and Astronomy
Research Centres and Groups:	Astronomy Centre
Subjects:	R Medicine > R Medicine (General) > R858 Computer applications to medicine. Medical informatics
Depositing User:	Users 8646 not found.
Date Deposited:	26 Mar 2018 15:59
Last Modified:	05 May 2021 11:43
URI:	http://sro.sussex.ac.uk/id/eprint/74659

View download statistics for this item

📧 Request an update