Second Meeting of the Routine Data Section

This group was formed to address the issues encountered by NIHR statisticians working on complex, routine datasets. We aim to provide a networking group for statistical researchers involved in the analysis of either established databases or routine data that has not been pre-processed.

Links to slides from the day are provided below.

GP consultations

John Edwards   Arthritis Research UK Primary Care Centre, Keele University  

Content: I aim to illustrate how an appointment with a GP turns into to coded information in the electronic health records.

Objectives:  

1. to map the range of processes involved in converting a consultation to a record, including the level of training they primary care clinicians are given, and the difficulties faced.
2. give some real examples that illustrate how it works in practice, with different kinds of patients/conditions.

CPRD

Dan Dedman CPRD, MHRA (Dedman CPRD.pdf)

Content: I will give a brief introduction on CPRD data, data access, and ISAC applications. I will talk some examples of common problems and issues in the ISAC applications. In addition, I will talk about challenges when requesting CPRD linkage data.

Daniel Prieto (Prieto EHDEN.pdf)

Content: I will give a brief introduction on the approach and strategy of the European Health Data & Evidence Network.

Objectives: From this, attendees will not only have an idea of primary care consultation databases for research, but also know process of obtaining data from two different data providers.

Primary care consultation data

Antonella Delmestri CSM, NDORMS, University of Oxford (Delmestri BigData automation.pdf)

Content: I will show the advantages of automation in big clinical data management, curation and extraction by using a DataBase Management System (e.g. MySQL) and a programming language (e.g. Python).

Rosa Parisi University of Manchester (Parisi rEHR.pdf)

Content: I will demonstrate how a R package could manipulate and analyse electronic health record data (as described in this PLOSOne paper). During this session, you will find out how to use the package rEHR in order to extract ready-for analysis dataset, including creating a longitudinal cohort or perform matching. It could be centrally by a Data Manager, or with the use of existing programming package.

Handling missing data in the primary care consultation database

Irene Petersen University College London (Petersen Missing data.pdf)

Content: I will discuss the scale of missing data in the primary care consultation database, and discuss typical approaches to handle missing data. I will also introduce the two-fold approach for multiple imputation for longitudinal electronic health record data.

Additional reference on missing data and multiple imputation and recording of primary care electronic health records in electronic health records.

Objectives: From this, attendees will have ideas of handling missing data in the primary care consultation data.

15.30-16.00 Jessica Harris, University of Bristol (Harris CPRD HES.pdf)

Content: I will present a CPRD/HES linked study and go through how I have used various codelists, to define exposures and outcomes.