A&E admission data

Status: 
Closed
Outcome: 
Data cannot be released
Notes: 

The extract service is operated on a cost recovery basis and is the mechanism for customers to access record level bespoke extracts of HES data

Due to the detail required within your request, the HES team is unable to provide the data free of charge.

Data request description: 
Admitted patient, outpatient, A&E and adult critical care datasets that are collected by HES, and the PROMs dataset. These include those submitted directly by the data providers and also fields we derive and add to the datasets. Currently this is used for a number of public reports. However, there is no access to granular data for analysis, and report-level data is prohibitively expensive.
Reason: 
The data is published but not in a format I can download and use (e.g. only displayed onscreen or only downloadable as a PDF rather than CVS)
Suggested use: 
Community Work
Research
Other
Make it easier for patients to make informed decisions about their use of A&E services in order to reduce pressure on those services. Currently there is no access to granular data for analysis. Report-level data - which is not detailed enough - is prohibitively expensive despite clear public interest use cases for the data, and the potential to save public money as a result.
Benefits overview: 
Given the increasing costs of A&E to the health service, and increased emphasis on reducing demand, there are clear opportunities to use granular level A&E admissions to create innovative ways to raise awareness of misuse of A&E, and to reduce the costs for those attempting to find ways to reduce demand. Publication of granular level data would also reduce pressure on the NHS Information Centre, further reducing costs.

Comments

Anonymisation?

Paul,

Is this data that you think NHSIC should be able to release with a sufficient standard of anonymisation? How granular is "granular"?

It sounds as if the dataset you want would contain information at the individual patient level. If so I can see that the barriers to release, at least as open data, might be considerable. (At the very least, privacy campaigners would need reassurance.)

-- Owen Boswarva, 21/02/2013

Health Episode Statistics (NHSIC)

 Flag as offensive 

Anonymisation

The key thing would be to identify fields which might make individual rows identifiable. I'd suggest for example that the time of admission and discharge could be easily limited to the hour. Or you could omit entries where codes apply to a single individual, for example (e.g. one person admitted with a particular cause during one hour).

You could also limit the whole dataset to a few key fields. But given the sheer numbers of people passing through A&E in any given hour I imagine time is going to be the key factor in anonymising entries.

 Flag as offensive 

Access in a secure environment?

If the benefits are as outlined, you would think that the NHSIC would be prepared to commission research, in which case access to the microdata would be permitted - for use in that project only and quite possibly "on site" rather than in the researcher's own environment.

ONS has a Virtual Microdata Lab in which approved researchers can access microdata.  They cannot remove the microdata or write anything down which can be construed as disclosive about the businesses or individuals concerned.  I believe that some other government departments have established similar arrangements but I don't know whether NHSIC is one of them.  I suspect probably not because patient privacy is one of the biggest issues around in this field. 

My personal opinion is that the Microdata Lab is one of ONS's most positive achievements and those involved in establishing it and pushing it forward deserve praise.  The flipside is that all involved must take data confidentiality absolutely seriously.  Perhaps it is easier for ONS to demonstrate its commitments to confidentiality than it would be for some administrative departments, but the potential gains in terms of informative analysis are huge.

 Flag as offensive 

Anonymisation

It is usually the conjunction of a number of data items which causes confidentiality problems.  Whatever you do with time, there may still only have been one person admitted with a particular issue and treated in a particular way at a particular hospital.  Yes, you can limit the dataset or combine data codes to reduce the overall number of possible values, but you may be losing something that actually matters as a result.  Preliminary work with extremely granular data in a secure environment might provide a basis for constructing with confidence a less granular dataset which could perhaps be analysed outside that secure environment.

Of course it may be that the "one person" referred to above cannot realistically be identified as a named individual.  But if it is not actually impossible (and it is hard to see how it could be completely impossible) the usual practice is, quite rightly, to err on the side of caution.  If there is a unique record, it can't be released.

 Flag as offensive