Advice on injury data access and sharing

Introduction

Aggregate data

- Publically available data
- Professional-only access: putting more in the public domain
- Small numbers and disclosure risk
- Accessing new aggregate data

Sharing person-level data

Introduction

There have been repeated calls for access to better information on injuries. There is certainly a need to improve the information that is collected, whether through enhanced collection of information on the causes of injury or improving the completion and quality of existing data collections such as information on emergency care (A&E) department attendances.

However, given these constraints, there is also a need for improved access to, and use of, the large amount of potentially useful injury-related data that is already collected by a wide range of agencies. Where there are data quality problems, it is still possible to extract useful information through careful data manipulation.

This brief guide addresses selected issues regarding accessing and sharing data in aggregate or person-level form. The presentation and method of interrogating data, however, is not addressed.

Aggregate data

Publically available data

Selected aggregate data (counts, rates, percentages, change over time) may be published at particular geographical levels in tables, charts or within commentaries. The Injury Observatory website aims to increase awareness of such data by signposting information resources through the ‘Research and Statistics’ website links, ‘Injury Prevention News’, ‘Other Useful Links’ and the search function on this and linked websites.

For example, see the continually updated England signposting guide to injury-related data sources, compiled indicators and selected publications, with links to on-line resources. This guide includes information on the lowest geographical level available in published indicators and also where age breakdown is, or is potentially, available.

Where local level data is not available, higher geographical level data can still provide a useful overview, and rates or percentages enable generation of local estimates of ‘expected activity’.

Professional-only access: putting more in the public domain

Aggregate data may be routinely circulated only within a professional group. Where there is no clear reason for restricting access, data providers could make this more accessible online. Data may be restricted because they involve small numbers (see below) or because those outside the group may not understand the data. In both cases, the output could be suitably adapted for publication with little effort.

For example, in England there are several injury-related indicators in the NCHOD ‘compendium’. However, an age breakdown at local level is only available to NHS professionals via the NHS-only access version (nww.nchod.nhs.uk) due to small numbers and disclosure risk (see below). This is overly restrictive.

Small numbers and disclosure risk

The UK Office for National Statistics (ONS) advises that care should be taken with regard to publishing small numbers of events in person-related statistics. This is due to the need to preserve confidentiality as there may be a risk that individuals could be identified, and these new details disclosed, if information held elsewhere (e.g. known by an acquaintance, available in a newspaper report or obtained by a ‘motivated intruder’) can be linked.

Small numbers may need to be suppressed in tables and any other values that can reveal them. Alternatively, data can be aggregated in a different way to avoid revealing small numbers or manipulated in an acceptable way to prevent revealing true numbers.

The risk of disclosure and cut-off number for suppressions (e.g. counts of 1 and 2, or 1 to 4, etc.) will depend on: the geographical level at which the data is published; the sensitivity of the information; legislation; specific data use agreements and whether there may be several different cuts of the same data, possibly published on different occasions.

Assessing risk and putting in place appropriate disclosure control measures may not be a straightforward process. Public health analysts in England, for example, have also found it difficult dealing with differing advice issued from the centre and the need to ensure a practical working approach. A typical rule of thumb used by England Public Health Observatories for geographies lower than Region, for example, is to suppress counts 1 to 4, except in the case of Hospital Episode Statistics (counts 1 to 5 as agreed with the Information Centre for Health and Social Care) and abortions (counts less than 10 suppressed at source). This may appear to be a more restrictive approach than that suggested in the ONS guidance, depending on the relevant statistics.

For further guidance, please refer to the ONS ‘Disclosure Control Policy for Tables’ guidance, the ONS Disclosure Control of Health Statistics guidance (which should be used for all publications in England), the England Information Centre for Health and Social Care’s Small Numbers Procedure (located in the right hand panel) and Scotland ISD’s 'Statistical Disclosure Control Protocol'.

Accessing new aggregate data

A large amount of data on injuries is collected by a wide range of agencies but only a fraction is made available in aggregate form on a routine basis. This is particularly the case with healthcare data. Those in the injury prevention field may want information on a particular age group, injury cause, health condition, geography or socio-economic group. They may want comparisons between categories or over time, or to see the same data cut in different ways. Of course, they will also have to bear in mind small number constraints at local level as outlined above.

However, it is not always clear what underlying data is collected, how reliable it is, who is able to provide the data, how to request it and the extent of analysis data providers are prepared to carry out. In the case of health data based on person-level data collection, local and regional public health analysts can discuss the specific requirements. The England signposting guide to injury-related data, for example, also provides detailed information on the person-level databases held by the regional Public Health Observatories.

Sharing person-level data

Those who have expertise in manipulating data may wish to request person-level data from other organisations so that they can flexibility analyse it or because they wish to generate useful information by matching this to person level data from another source. Alternatively, agencies may need access to person-identifying data to carry out particular checks.

In the UK, holding, sharing and manipulating person-identifying (‘personal’) data is strictly governed by the Data Protection Act which is overseen by the UK Information Commissioners’ Office. Personal data would include items such as name, unique system identifier, date of birth and postcode of residence. Organisations should consider first whether ‘anonymised’ person-level data sharing will suffice. For example, single age or age group could be used instead of date of birth, and partial postcodes or small area geographies, such as ‘lower super output areas (LSOAs)’, derived from postcodes can be used. In Wales, a method of linking anonymised health-related data sets has also been devised – see below.

In the UK, where organisations wish to share personal data they should refer to the Information Commissioner’s ‘Data sharing code of practice consultation paper’ published October 2010. This covers two main types of data sharing:’ i) ‘systematic routine data sharing where the same data sets are shared between the same organisations for an established purpose and ii) exceptional, one-off decisions to share data for any of a range of purposes.’ (p.2). Section 14 (pp30-33) of the consultation paper 'Suggested contents of a data sharing protocol' is particularly helpful and contains mini templates for data requests & responses.

In the Republic of Ireland, the Central Statistics Office (CSO) Data Protocol covers data sharing and linkage.

For healthcare data, reference should also be made to confidentiality codes of practice. For example see: England’s ‘NHS Confidentiality Code of Practice’, Confidentiality: Code of Practice for Health and Social Care in Wales’ and Scotland’s NHS Code of Practice on Protecting Patient Confidentiality'.

Further details on linking person-level anonymised data in Wales

Anonymous patient level injury records within Wales are collected and collated by the All Wales Injury Surveillance System (AWISS). The data within AWISS originates from the multiple data sources incorporated within the Secure Anonymised Information Linkage (SAIL) system run by the Health Information Research Unit (HIRU) based at Swansea University. The healthcare related datasets within SAIL include emergency department (ED), inpatient, outpatient and mortality datasets, and each encompass an encrypted Anonymised Linkage Field (ALF_E). This field, uniquely generated from a patient’s NHS number or accompanying demographic information (forename, surname, date of birth, address), allows each of the datasets within SAIL to be linked together at a patient level. Hence, AWISS represents an integrated surveillance system which can anonymously track individuals treated at ED departments through hospital inpatients, outpatients, and mortality. Ambulance dispatch and treatment data are not yet fully computerised in Wales but there are plans to incorporate such data when they become available.

Download guidance on injury data access and sharing

To download the guidance above on injury data access and sharing please click here.