‘Big data’ from medical records transforms health-related research
Information provided in free-text sections of electronic patient records can help researchers to detect cases of a particular condition, according to a new study by Brighton and Sussex Medical School (BSMS) and the University of Sussex.
Dr Elizabeth Ford, BSMS
Results could lead to a major improvement in case detection rates, improving research and potentially saving lives.
Research using health ‘big data’ relies on good methods of data extraction so that researchers find the right patients to use in their studies.
So far such research has mainly relied on coded data in medical records, with free-text notes made during consultations generally overlooked due to the difficulty of anonymising the text.
Dr Elizabeth Ford from BSMS worked in collaboration with Professors John Carroll and Donia Scott in the Department of Informatics at Sussex to review all available literature on finding cases in electronic medical records where anonymised medical text had been used in the study – in addition to the easier ‘coded’ data.
The researchers found that where text was incorporated in the methods for choosing cases, there was a significant improvement (16%) in the sensitivity of case detection.
Dr Ford says: “Our finding that text significantly improves case-detection is important in public health research. If cases are missed from studies using electronic medical records, the research may get the wrong answers.
“Medical researchers and computational linguistics experts must work together to get the most out of health data in the new era of health big data.”
This study has been published in the Journal of the American Medical Informatics Association.
BSMS is a partnership between the Universities of Sussex and Brighton together with NHS organisations throughout the south-east region.