Studies using electronic health records(EHR) are frequently used to inform people’s healthcare. Determining people’s health conditions is central to EHR research, through the creation of lists of codes corresponding to diseases diagnosed and medications prescribed. Unfortunately, the creation of these codelists differ in methodology and reporting and in turn can lead to varying study results and even introduce bias.
There is a need to improve the transparency of codelists, incorporating common, clear steps to facilitate sustained codelist re-use. Gaps in efforts to incorporate this is most prominent for drug codelists. There are unique considerations for creating drug codelists, particularly when there is missing information describing each drug. . Furthermore, numerous brand names per drug and ever-changing status of drugs new, existing, and no longer in-use; makes codelist creation difficult.
We will describe a process for creating drug codelists that considers this missing information and when drug use-status is ever-changing, using two codelists: a codelist for drugs administered for high blood pressure and heart failure (British National Formulary Chapter 2.5) and a codelist for inhaled therapies for COPD (BNF Chapter 3.1;3.2). Our goal is to apply the codelists to a group of people with Chronic Obstructive Pulmonary Disease (COPD) in the Clinical Practice Research Datalink (CPRD) from 1 January 2010 to 31 December 2019.
Research using routinely collected electronic health records (EHR) are increasingly used to support patient care. In the UK, this has been occurring for over two decades using longitudinal primarycare records across a breadth of databases, including Clinical Practice Research Datalink (CPRD).
Determining exposures, outcomes, and covariates is central to EHR research through generation of medical and drug codelists. Unfortunately, both creation methodology and reporting vary across studies, together forming source of bias – manifesting in inappropriate exclusions and inclusions of codes.
The Reporting of studies Conducted using Observational Routinely-collected health Data (RECORD) statement has called for transparency, to consider potential for misclassification bias when using codelists to identify patients’ true prescription use, and for researchers to provide “sufficient detail” to enable codelist (and study) reproducibility. Approaches systematic and standardisable are important, to enable not only codelist reproducibility, but generalisability to other studies and their contexts.
Two previous studies have published reproducible methods for codelist development, with the paucity of literature most prominent for drug codelists. Unique considerations are present when creating drug codelists, particularly when there are missing data among covariates describing each drug (e.g. chemical name, route of administration, BNF code). Furthermore, numerous proprietary names and the status of new, existing, and discontinued drugs being constantly in-flux makes codelist generation complex.
We describe a process for developing drug codelists considering this missing data and when use-status of codes are ever-changing.
Our goal is to apply two codelists using this methodology, to a cohort of people with Chronic Obstructive Pulmonary Disease (COPD) in the Clinical Practice Research Datalink (CPRD) Aurum from 1 January 2010 to 31 December 2019, a cardiovascular (CV) codelist consisting of drugs indicated for hypertension and heart failure (BNF Chapter 2.5), and a COPD disease-specific codelist for inhaled therapies for COPD (BNF Chapter 3.1;3.2).
1. Product codes and drug issues using our drug codelist generation methods (prodcodeid, drugissues, N, %) - for the CV codelist and for the COPD inhalers codelist.
2. Patients prescribed any of the drugs determined using our drug codelist generation methods (N, %) - for the CV codelist and for the COPD inhalers codelist.
Jennifer Quint - Chief Investigator - Imperial College London
Emily Graul - Corresponding Applicant - Imperial College London
Alexander Adamson - Collaborator - Imperial College London
Georgie Massen - Collaborator - Imperial College London
Nicholas S Peters - Collaborator - Imperial College London
Philip Stone - Collaborator - Imperial College London
Sara Hatam - Collaborator - Imperial College London
Spiros Denaxas - Collaborator - University College London ( UCL )