Code sets for Electronic Health Record research

Primary and secondary care code sets for Electronic Health Record research. The code sets were developed primarily for use with UK Biobank data.

🧑‍🎓 Please cite this work if you use it!

Primary care

Clinical event codes are provided using Read v2 and Clinical Terms Version 3 (CTV3) classifications.

Conditions (csv/rds)

Variable	Value	Level	Description
angina	diagnosis	stable	Stable angina
angina	diagnosis	unstable	Unstable angina
bipolar	diagnosis	-	Bipolar disorder
diabetes	diagnosis	-	Diabetes (type unknown)
diabetes	diagnosis	type1	Type 1 diabetes
diabetes	diagnosis	type2	Type 2 diabetes
diabetes	diagnosis	gestational	Gestational diabetes
diabetes	diagnosis	secondary	Secondary diabetes
diabetes	diagnosis	remission	Diabetes remission
diabetes	diagnosis	resolved	Diabetes resolution
diabetes	family_history	-	Family history of diabetes
hypertension	diagnosis	-	Hypertension
learning_disabilities	diagnosis	-	Learning disabilities
mi	diagnosis	-	Myocardial infarction/heart attack
pcos	diagnosis	-	Polycystic ovarian syndrome
schizophrenia	diagnosis	-	Schizophrenia
stroke	diagnosis	haemorrhagic	Haemorrhagic stroke
stroke	diagnosis	ischaemic	Ischaemic stroke
tia	diagnosis	-	Transient ischaemic attack

Biomarkers (csv/rds)

Variable	Value	Level	Description
blood_glucose	fpg	-	Fasting plasma glucose
blood_glucose	hba1c	-	Glycated hemoglobin
blood_glucose	ogtt	2hour	2 hour oral glucose tolerance test
blood_glucose	random	-	Random blood sugar
blood_glucose	unknown	-	Glucose test (unknown type)
anthropometric	bmi	-	Body mass index
anthropometric	height	-	Height
anthropometric	weight	-	Weight
anthropometric	waist	-	Waist circumference

Demographic/other (csv/rds)

Variable	Value	Level	Description
smoking	current	trivial	Current trivial smoker
smoking	current	light	Current light smoker
smoking	current	moderate	Current moderate smoker
smoking	current	heavy	Current heavy smoker
smoking	current	very_heavy	Current very heavy smoker
smoking	current	-	Current smoker (level unknown)
smoking	former	trivial	Former trivial smoker
smoking	former	light	Former light smoker
smoking	former	moderate	Former moderate smoker
smoking	former	heavy	Former heavy smoker
smoking	former	very_heavy	Former very heavy smoker
smoking	former	-	Former smoker (level unknown)
smoking	never	-	Never smoked
smoking	non	-	Non-smoker (assumed current)
smoking	passive	-	Passive smoker (assumed current)
smoking	consumption	-	Cigarette consumption

Drug prescriptions

Around 76% of UK Biobank prescription records have a BNF code. 99.7% of records have a BNF and/or Read v2 code. Prescription codes are therefore provided using British National Formulary (BNF) and Read v2 classifications.

prescriptions.rds is a named "list of lists" for the following drug categories:

Drug category	Name
Anti-diabetes drugs	diabetes
Anti-hypertensives	hypertension
Atypical anti-psychotics	antipsychotic
Steroids	steroids
Statins	statins

Further details are provided here.

⚠️ UK Biobank guidance highlights issues including incomplete and/or inconsistently formatted BNF codes, missing Read v2 codes and missing drug names. The ukbb-ehr-data repository includes example code to handle these issues and extract drugs using these code sets.

Secondary care

Secondary care diagnoses are provided using ICD-9 and ICD-10 coding classifications. Procedures are provided using OPCS-3 and OPCS-4 classifications.

Conditions (csv/rds)

Variable	Value	Level	Description
diabetes	diagnosis	-	Diabetes (type unknown)
diabetes	diagnosis	type1	Type 1 diabetes
diabetes	diagnosis	type2	Type 2 diabetes
diabetes	diagnosis	gestational	Gestational diabetes
diabetes	diagnosis	secondary	Secondary diabetes

Other resources

Open repositories

The majority of diagnosis records in the interim EHR data release use the CTV3 coding classification. The code set repositories below typically only cover Read v2 diagnostic codes and limited prescription coding.

• https://phenotypes.healthdatagateway.org/
• https://www.opencodelists.org/
• https://clinicalcodes.rss.mhs.man.ac.uk/
• https://caliberresearch.org/portal is no longer updated

Kuan et al (2019) includes a map of 308 physical and mental health conditions. Read v2 codes are available at CALIBER and https://github.com/spiros/chronological-map-phenotypes.

Prescription coding

• https://openprescribing.net/bnf/ includes a browsable BNF with high-level prescribing trends
• https://www.thedatalab.org/blog/161/prescribing-data-bnf-codes/ summarises the BNF coding structure

Code mapping

• https://biobank.ndph.ox.ac.uk/showcase/refer.cgi?id=592
• https://isd.digital.nhs.uk/

Citing this work

If you use this work, please cite it as below:

@article{10.1093/jamia/ocab260,
    author = {Darke, Philip and Cassidy, Sophie and Catt, Michael and Taylor, Roy and Missier, Paolo and Bacardit, Jaume},
    title = "{Curating a longitudinal research resource using linked primary care EHR data - a UK Biobank case study}",
    journal = {Journal of the American Medical Informatics Association},
    volume = {29},
    number = {3},
    pages = {546-552},
    year = {2021},
    month = {12},
    issn = {1527-974X},
    doi = {10.1093/jamia/ocab260},
    url = {https://doi.org/10.1093/jamia/ocab260},
    eprint = {https://academic.oup.com/jamia/article-pdf/29/3/546/42333190/ocab260.pdf},
}

Licence

Made available under a Creative Commons Attribution 4.0 International License.