6.6. Specific databases

6.6.1. Servelec RiO

Servelec RiO data exports come in several formats, including:

  • raw;

  • “RCEP”: preprocessed by Servelec’s RiO CRIS Extraction Program.

Individual organizations may process these too. CRATE provides a preprocessor (crate_preprocess_rio) to convert a RiO database to a format suitable for anonymisation via CRATE.

6.6.2. PCMIS

There is a specific preprocessing tool for PCMIS, namely crate_preprocess_pcmis.

6.6.3. TPP SystmOne

TPP provide a “strategic reporting extract” (SRE) containing SystmOne data. This contains structured data, but can contain free text too.

The structure of the SRE is good from CRATE’s perspective; it does not require reshaping for anonymisation.

The crate_preprocess_systmone will index a SystmOne source database (without which, anonymisation is very slow). It will also, optionally, create a view to add blurred geographical information, if you have used the the crate_postcodes tool to import UK Office for National Statistics geography data into a database.

Use the crate_anon_draft_dd tool to create a data dictionary from SystmOne. CRATE knows something about the structure of a typical SystmOne database.

NHS numbers, which are 10-digit integers incorporating a checksum, are represented in our database copy of the SRE by the VARCHAR(10) data type (clearly a little suboptimal). It remains OK to use these in your anonymiser config file:

sqlatype_mpid = BigInteger
#
# Within CPFT, we have some locally created columns with string versions of
# the primary SystmOne ID, and so forth, so we use:
#
# sqlatype_pid = String(100)
# sqlatype_mpid = String(100)

However, you will see some warnings during config checking. See sqlatype_mpid.

When generating a data dictionary, use these settings for your source database:

ddgen_omit_by_default = False
# ... or use "--systemone_include_generic" with crate_anon_draft_dd
# ... or use True if you want to hand-review everything

ddgen_per_table_pid_field = IDPatient
# ... largely cosmetic; improves the warnings if your local database
# modifications have an odd structure.