Linking health care records for statistical analysis: care.data, possibilities and problems
Download audio [mp3]
Chair: Professor Harvey Goldstein (University of Bristol)
The recent debate about the role of The Health and Social Care Information Centre (HSCIC) in producing and making available linked GP and hospital records (care.data) has raised a number of political and ethical issues. The meeting will start with a brief history of the controversy, setting the background for a discussion of the statistical issues involved. These will include the balance between anonymisation and data degradation, how a possible 'safe laboratory' for researchers might work, how linkage errors can be dealt with and whether pseudonomysing data at source, to improve data security, is compatible with efficient linkage. Issues concerning trust in Government handling of data confidentiality will form an important background to the various contributions, but these are not the prime focus of the meeting.
Speaker, Titles and Abstracts:
Chris Roebuck (Director of Benefits and Utilisation at The Health and Social Care Information Centre (HSCIC))
‘Real-life challenges in data linkage’
There has been much discussion recently around the appropriateness of data sharing, but less around the practical challenges that confront the user/researcher. Garry Coleman is responsible for data access for the HSCIC, and the Information Asset Owner for Hospital Episode Statistics, will cover some of the very practical issues in linking and sharing data. Using examples from the HSCIC, he will cover some of the current issues being faced around pseudonymisation & anonymisation, challenges around quality, and ensuring appropriate access to data. Slides are available here.
Katie Harron (University College London)
‘Challenges in data linkage: error and bias’
Data linkage is becoming an increasingly important tool for observational research as individual-level information can be combined relatively quickly and inexpensively, avoiding the need for new data collection. However, lack of unique identifiers, particularly in administrative data, means that linkage is not always straightforward. Errors occurring during linkage (false-matches and missed-matches) can lead to substantial bias in results based on linked data. This talk will give an overview of different methods for linkage: deterministic linkage, probabilistic linkage and imputation for uncertain links. It will then address why it is important to evaluate the quality of linkage in terms of the impact of linkage error on results. Slides are available here.
Liz Little (Director of Business Development at Dr. Foster)
‘Care.data for quality measurement: possibilities, challenges and a global perspective’
Dr Foster has been at the forefront of publishing hospital quality outcomes for over 10 years. Using national administrative data we continue to provide NHS trusts and commissioners with benchmarking information; highlighting areas of poor and excellent care and guiding decision making inside and outside the hospital. Care.data offers a world of opportunity for more and better analysis of quality of care in the NHS. The development of metrics that incorporate clinical information and the ability to measure variation in outcomes across a whole patient pathway are but two of the exciting possibilities. However, collecting a richer patient-level data layer will do nothing unless it is robust and accessible. We will discuss the potential of care.data and Dr Foster’s experience in working with data at a local level. We will also put this debate into an international context with some examples from our experience with national datasets from around the world. Slides are available here.
Alison Macfarlane (City University London)
‘Care.data: Bungled opportunity or unjustified intrusion?’
A national population-based system to compile data about care given in general practice, where the majority of health care takes place, would be invaluable for a range of health service, public health and research purposes, especially if linked to hospital and other data. Systems have been implemented in countries such as Scotland, Wales, the Nordic countries, using a collaborative approach, usually involving, health services, universities and other public sector organisations. In contrast, the attempt to establish care.data in England has met opposition from many directions and campaigns to encourage the public to opt out, call into question the quality, completeness and usefulness of the data to be compiled.’ Slides are available here and a report from the meeting has been published here.
We are experiencing a high demand for places for this event so pre-registration is essential
Organiser Name George Leckie
Organising Group(s) RSS Social Statistics Section