Data linking – Project update 15 th May 2012 – Homecare & SDS event Atlantic Quay Ellen Lynch & Euan Patterson.

Data linking – Project update

15th May 2012 – Homecare & SDS event Atlantic QuayEllen Lynch & Euan Patterson

Overview

• Objective of project• Timescales• Data Sharing Agreements & other approvals• Local Authority instructions• Metadata• Statistical Disclosure Control• Next steps• Further information

Objective of project

• Evidence the interactions between social care, housing support and health– Are our social care, housing support & health policies

working?• Evidence for service planning and service evaluation• Evidence the outcomes of people who use these services• Evidence for indicators

Pilot project

• Analytical Services Division (Scottish Government) wrote to Directors of Social Work in Autumn 2011 asking for volunteers to participate in the pilot project.

• 18 local authorities have so far indicated their interest to participate in the pilot project.

Timescales

Milestone Timescales

Data Sharing Agreements signed & Privacy Advisory Committee approval in place

May/Jun 2012

Sharing and linking of data Jun/Jul 2012

Initial analysis (including statistical disclosure control guidance/methodology)

Jul- 2012

Access for researchers via safe haven Summer 2012

Pilot project review Sept/Oct 2012

Roll out across remaining local authorities Nov 2012-

Plan future projects On going

Data Sharing Agreements & other approvals

• Currently draft agreement with the SOLAR (Society of Local Authority Lawyers and Administrators in Scotland) subgroup on data protection.

• Aiming to get agreements in place by early June.• Should receive a response from the Privacy Advisory

Committee (PAC) (about linking to health data) by the end of May.

• Once the agreement and PAC approval are in place, the sharing and linking of data can proceed.

Local Authority instructions

• Local authorities will be provided with instructions about what is required and when.– For example the required method to share data

• Templates will be provided for:– Sending personal identifiers to ISD Scotland– Sending the linking ID, project code and unique ID to the Scottish

Government (via ProcXed.Net)

Process of uploading data to create a linked dataset

Metadata

• Information about the datasets being utilised in the project.

• Analytical Services Division are looking for feedback:– Is the layout clear?– Anything missing?

• Metadata demonstration

Statistical Disclosure Control

• Could any of the outputs lead to the identification of individuals, or allow people to learn information about individuals that they did not already know?

• If there is a risk, then we have to look at applying Statistical Disclosure Control to the data before it is publically released.

• Statistical Disclosure Control involves changing the data in some way, either by minimising the level of detail published or by introducing a degree of ambiguity into the data.

• Fundamental aspect of Statistical Disclosure Control is finding a balance between risk and utility.

Statistical Disclosure Control (cont.)

Disclosure RiskAssessment

(Subjective) Data Environment What is already out there?

What intruder scenarios exist?

(Subjective)Impact of DisclosureWhat would the damage be?How sensitive is the data?

(Objective)Dataset Risk

How disclosive are the data?

Statistical Disclosure Control (cont.)

• Impact of Risk AssessmentBased on the results of the risk assessment disclosure controls may need to apply.

Common approaches to Disclosure Control

Aggregating

band ages e.g. 16 – 24

Use larger geographies

Collapse categorical variables e.g. compulsory education / post-compulsory education

Suppression

Remove data for individuals at risk of disclosure

Suppress cells in tables and consider secondary

suppression

Manipulation

Swap the variable values for individuals from different areas

Add noise to the data e.g. inflate numerical values by 3%

Statistical Disclosure Control in Practice

Data from the Home Care Census 2010 & 2011 will be released shortly.The data will be available in two forms:– A licensed file held by the UK Data Archive, which you are required

to register for.– A public release file hosted on the Scottish Government’s website.

We are currently exploring the disclosure control measures need for the linked data, and the utility and user experience of this data are central considerations. The concentration of these controls will be on the outputs that are released from the safe haven, as there is no intention to release raw data.

Statistical Disclosure Control cont……• Statistical Disclosure Control in the safe haven:

– Implications for the data users: Do I need this output in this form? Do I need this level of detail to support my argument/publication? A very basic rule of thumb, if cells in tables represent less than 5

individuals, this could pose a disclosure risk .– Implications for the project coordinators:

A risk analysis for the data as a whole, and the subsets that underlies the output needs to be undertaken.

This involves both subjective judgements and some measurable risk metrics.

– For subjective judgements, guidelines and procedures will be used to ensure that a consistent approach to risk is taken.

– Software can help us to produce risk metrics for datasets, to better inform the decisions taken above.

Next steps (for local authorities)

• Data sharing agreements to be issued to all participating local authorities for completion and signing (by Information Asset Owner).

• Instructions and templates to be issued to all participating local authorities.

• Local Authorities populate templates and share required data by ISD Scotland and then subsequently the Scottish Government.

• Complete safe researcher training (to be available Summer 2012) for access to linked datasets (if you have indicated you want to analyse the linked data yourself).

Further information

• http://www.scotland.gov.uk/Topics/Statistics/Browse/Health/Datalinking/HealthSocialCareandHousin

• Ellen Lynch / Euan Patterson

• [email protected]