Skip to content

Policy for OpenSAFELY Access by Platform DevelopersπŸ”—

IntroductionπŸ”—

OpenSAFELY securely mediates access to pseudonymized patient data for the purpose of analysis (be that research, service evaluation or audit). Our objective is to minimise direct access to pseudonymised data wherever possible, to protect privacy; while facilitating high quality, efficient analytics on the data by a wide range of users. A small number of OpenSAFELY team members need more direct access to the pseudonymised data inside the secure environment. This access is required:

  • to develop and operate the OpenSAFELY platform
  • for data curation purposes

What this policy describesπŸ”—

This policy describes:

  • the conditions under which individuals with these roles may access the OpenSAFELY platform
  • the controls in place to ensure that such access cannot be used to subvert the OpenSAFELY security model

All such access conforms to the OpenSAFELY platform’s principles of security, transparency and minimal access.

The specific access activities this policy covers are:

  • running OpenSAFELY test pipelines
  • ad hoc access to backend systems
  • data curation activities

Running OpenSAFELY test pipelinesπŸ”—

WhyπŸ”—

For software maintenance purposes, it is sometimes necessary to execute OpenSAFELY project pipelines. The test project pipelines are necessary to:

  • ensure that the entire OpenSAFELY platform is operating correctly
  • verify that changes to the software work correctly in deployed systems

WhatπŸ”—

The OpenSAFELY team maintains a small set of test studies in the OpenSAFELY Internal project. These studies have pipelines that use the standard OpenSAFELY tooling, but which have been specially designed to test and assess the platform’s functionality rather than to answer research questions.

Because they are standard OpenSAFELY pipelines, they benefit from all the same security controls as OpenSAFELY research pipelines, including:

  • publication of all code that is run
  • logging of all operations
  • manual checking of outputs for disclosivity before they are released from the backend

Access to the OpenSAFELY platform will also be required to verify the pipeline outputs are as expected.

WhoπŸ”—

The test studies can only be run by OpenSAFELY Core Developers.

WhenπŸ”—

These test pipelines are run periodically as is necessary to check that the system is working properly. They are also run when updated software has been deployed. However this deployment only happens after new versions of the software have gone through careful testing with simulated data to ensure that they are functioning correctly and that the security characteristics of the platform are maintained.

Ad hoc access to backend systemsπŸ”—

Note

This section applies to OpenSAFELY-TPP and OpenSAFELY-EMIS only.

WhyπŸ”—

To ensure the correct operation of the platform, it is sometimes necessary to directly access OpenSAFELY backend systems so that:

  • software can be reconfigured or updated
  • problems which only manifest themselves in the deployed system can be investigated

WhatπŸ”—

Access to backend systems is via secure, authenticated, encrypted channels. All access to the system and database operations are logged. This process gives access to:

  • the OpenSAFELY software running in the backend
  • logs that the OpenSAFELY software produces
  • pseudonymized patient data
  • unvetted intermediate data
  • unchecked, unreleased study outputs.

There is no facility for extracting patient data or study outputs from the system via this process; the data can only be studied within the system itself. Some operational data may be released under controlled conditions, as detailed in the Operational Data Policy.

WhoπŸ”—

Direct access to the OpenSAFELY backend systems is only available to OpenSAFELY Core Developers.

WhenπŸ”—

This process is routinely used for updating the platform with new software versions. It is exceptionally used for investigating outages or other problems, and only when the information necessary for the investigation can’t be accessed from outside the system.

Data curation activitiesπŸ”—

Note

This section applies to OpenSAFELY-TPP and OpenSAFELY-EMIS only.

WhyπŸ”—

Characterizing datasets and checking data quality.

WhatπŸ”—

Access to backend systems is via secure, authenticated, encrypted channels. All access to the system and database operations are logged. Because data curation requires patterns of data access that are not supported by the standard OpenSAFELY pipelines, this process allows custom code to be run with access to the pseudonymized patient data, outside the pipelines.

These data curation activities benefit from all the same security controls as standard OpenSAFELY pipelines, including:

  • publication of all code that is run
  • logging of all operations
  • manual checking of outputs for disclosivity before they are released from the backend

Data curation activities are also subject to the standard OpenSAFELY research policies.

WhoπŸ”—

Only OpenSAFELY Core Developers and OpenSAFELY Data Investigators have permission to carry out data curation activities.

WhenπŸ”—

Data curation activities may be carried out:

  • when new datasets are added to OpenSAFELY
  • when datasets are updated
  • when previously unused fields in existing datasets are put to use
  • periodically for data quality assurance and sense checking purposes