Essential information > Understanding UK Biobank > Data providers and dates of data availability

Data providers and dates of data availability

This page gives details of the various data providers for linked death, hospital inpatient, cancer and primary care (GP) records. It also gives the period for which each type of data is available, giving the earliest date for which we have data and an attempt to determine a censoring date for each type of data and data provider as described in the next sections.

Censoring dates

The censoring date for each data provider is the date up to when UK Biobank estimate that data received from that provider is mostly complete. Our general rule for estimating an appropriate censoring date is as follows:

The censoring date is the last day of the month for which the number of records is greater than 90% of the mean of the number of records for the previous three months, except where the data for that month is known to be incomplete in which case the censoring date is the last day of the previous month.

The censoring dates are not applied by UK Biobank to the data made available to researchers which will always contain the latest data regardless of censoring dates, and may include incomplete data after the dates below. These dates are intended for guidance only. Once a researcher has received their data, they should censor outcomes based on their own research protocol.

The Data Portal vs Showcase fields

Hospital inpatient data and death data are available to researchers in two format: record-level fields on the Data Portal and Showcase fields, which in the case of inpatient data give summary-level information about the hospital episodes. In addition, there are Showcase fields such as the Algorithmically-defined outcomes (Category 42) and the First Occurrences (Category 1712) which combine information from multiple sources (such as self-report at Biobank assessment centres, inpatient, death, and GP records).

These derived Showcase fields are updated less frequently than the Data Portal tables, and so two censoring dates are given below. The "Data Portal censoring date" gives the date the tables on the Data Portal are complete to, and the "Showcase censoring date" gives the date to which derived Showcase fields use complete data from that source. In both cases "complete" here means as determined by the above rule.

The Data Portal will in general be more up to date than the Showcase fields available in a main dataset. Researchers requiring the most current data should use the record-level fields.

Data available for all research purposes

Death data

Death	Data Provider	International Classification of Diseases (ICD)		Period of data currently available	Showcase censoring date	Data Portal censoring date
		ICD9	ICD10
England & Wales	NHS Digital		2006 onwards	April 2006 onwards	28 February 2021 *	28 February 2021
Scotland	NHS Central Register, National Records of Scotland		2006 onwards	April 2006 onwards	28 February 2021 *	28 February 2021

* Not including the Algorithmically-defined outcome fields (Category 42).

Hospital inpatient data

Hospital Admissions (Inpatients)	Data Provider	International Classification of Diseases (ICD)		Classification of Interventions and Procedures (OPCS)		Period of data currently available	Showcase censoring date	Data Portal censoring date
		ICD9	ICD10	OPCS3	OPCS4
Hospital Episode Statistics for England (HES)	NHS Digital		1997 onwards		1997 onwards	1997 onwards, with critical care data from 2011	31 March 2021 *	31 March 2021
Scottish Morbidity Record (SMR)	Information and Statistics Division (ISD), Scotland	1981 - 1996	1996 onwards	1977 - 1988	1989 onwards	1981 onwards	31 March 2021 *	31 March 2021
Patient Episode Database for Wales (PEDW)	Secure Anonymised Information Linkage (SAIL), Wales		1999 onwards		1999 onwards	1998 onwards	28 February 2018 *	28 February 2018

* Not including the Algorithmically-defined outcome fields (Category 42).

We have held back a very small proportion of English inpatient data for April 2017 onwards (approximately 0.25%, or around 600 episodes per year) due to an incomplete linkage match. After they have been scrutinised further, some of these records may be released at a future date.

The Scottish hospital inpatient data does not currently include psychiatric or maternity admissions.

Primary care (GP) data - for all research

Note that the GP data available for all research covers a total of approximately 45% of the UK Biobank cohort. The coverage dates are based on the value of the field event_dt (event date) in the gp_clinical table.

GP dataset	Data provider	Participant coverage (approx.)	Coding systems	Period of data currently available	Censoring date
England	TPP	165,000	See Resource 591	1938 onwards *	31 May 2016
	Vision	18,000	See Resource 591	1940 onwards *	31 May 2017
Scotland	Vision/EMIS	27,000	See Resource 591	1939 onwards *	31 March 2017
Wales	Vision/EMIS	21,000	See Resource 591	1948 onwards *	31 Aug 2017

* For each provider the number of records per year is very low initially and gradually increases.

Cancer data

Cancer	Data Provider	International Classification of Diseases (ICD)		Period of data currently available	Censoring date
		ICD9	ICD10
England & Wales	NHS Digital *	1979 - 1994	1995 onwards	1971 onwards	31 July 2019
Scotland	National Records of Scotland, NHS Central Register	1980 - 1996	1997 onwards	1957 onwards	31 October 2015

* Please refer to the cancer registry information page for further details on the data.

Data available only for COVID-19 research

COVID-19 test results

For information on a censoring date for the COVID-19 test results data, please see the Data Portal updates table on the timelines page. The censoring date will generally be approximately one or two weeks before the "Last Update" date.

Primary care (GP) data - for COVID-19 research only

GP dataset	Data provider	Participant coverage (approx.)	Coding systems	Period of data currently available	Censoring date
England	TPP	190,000	See Resource 3151	1938 onwards *	31 March 2021
	EMIS	260,000	See Resource 3151	1938 onwards *	30 April 2021

* For both providers the number of records per year is very low initially and gradually increases.