Routine Data in Primary Care research: Utopian dream or nightmare reality?
Rationale – Routinely collected healthcare data, in theory, offers an inexpensive, fast and voluminous resource to apply in a range of study designs across an array of clinical areas. However, the reality of using routine data for primary care research is often very complex. In this symposium we will explore some of the complexity in using routine data as well as some potential solutions to overcome the problems. Aims – The aims of the symposium are explore some of the complexities of using routine data as well as highlighting some potential solutions and ways forwards. Symposium Content – The symposium will cover a reflection of the reality of using routine data for primary care research, including some of the pitfalls and unintended consequences. It will present a number of pieces of work which illustrating various challenges and inconsistencies found between different routine data sources and the potential impact they may have on conclusions. The symposium will also explore ways to overcome some of the difficulties including posing the question how might we recognise good quality routine data and presenting some statistical methodology which might be used. Format of symposium – There will be a series of four talks followed by discussion Intended audience – Researchers using, or interested in using, routine data for primary care research. Plans for discussion – As well as a chance to ask questions after each talk there will be an extended discussion section led by a discussant drawing together the themes presented in the four talks. A particular focus will be given to understanding what we can take from previous studies and new methods such that the future research with routine data can be conducted to provide the best evidence possible for improving primary healthcare.
Presentation 1 - Randomised Trials and Routine Data: the promise, the reality and the potential
Rebecca Cannings-John Cardiff University
Randomised Controlled Trials (RCTs) have become more expensive and challenging to undertake and there are concerns about bias with respects to loss to follow-up, population selection, and the co-intervention effects from data collection. Whilst using routinely collected datasets (RCD) might seem like a panacea to RCTs, the practicalities of using them is far more complex. This discussion will focus on how the promise of accessibility to link to research data marries up with the reality by using examples of how it has been used in trials over the last 10 years and how certain aspects have improved whilst there are many unresolved challenges.
Presentation 2 - When is a consultation not a consultation? The hidden meaning of clinical system appointments data.
Jenni Burt, University of Cambridge
Extracting and using information on the contacts patients have with health care professionals at their primary care practice is a common strand of many studies. Frequently, there are limited opportunities to double-check the quality of such data. In one study recruiting patients who had had a recent face-to-face contact with a GP, queries were raised when some GPs were apparently seeing upwards of 55 patients per day. In this talk, we will present the results of further investigations into the ‘busy GP’ mystery, and consider the implications of using appointment data for research purposes.
Presentation 3 - Differences in Routine Datasets
Rupert Payne, University of Bristol,
The manner in which data are recorded by clinicians and administrators in routine practice can result in important differences between datasets, with discrepancies not necessarily immediately apparent to researchers. In this presentation we consider examples including differences in disease incidence and patient characteristics between primary and secondary care records, inaccuracies in timing of events in GP records, disparities in patient self-report and administrative records, and agreement between free text and coded clinical data. We will examine the key factors leading to differences between datasets, and potential ways in which datasets can be compared to provide an indicator of data quality.
Presentation 4 - Assessing real-world effectiveness and risk of side-effects for common treatments in Primary Care: Application of the Prior Event Rate Ratio (PERR) method.
William Henley, University of Exeter
There is increasing interest in using electronic health record data to study the effectiveness of interventions and the associated risk of side effects in a real-world setting. However, the lack of randomised treatment allocation can result in an imbalance in the characteristics of patients receiving different treatments. These prescribing biases may result from unmeasured or unrecognised processes, and failure to account for them may lead to errors in treatment estimates. This talk will introduce Prior Event Rate Ratio adjustment, a promising approach to addressing this unmeasured confounding bias, and illustrate its application to assessing side-effects to Type 2 diabetes medications.