Unexpected ICD‑10 Umbrella Codes in Field 41270: Case of I48 (Atrial Fibrillation)
Hello,
While setting inclusion/exclusion criteria for Atrial fibrillation and flutter (ICD‑10 code I48), we noticed that field 41270 contains individuals coded with both the umbrella (parent) code I48 and one of its more specific subcodes (e.g., I48.0, I48.1, etc.). Our understanding is that hospital Episode Statistics should typically record specific ICD‑10 subcodes, and that umbrella codes should only appear when they exist as stand‑alone categories (e.g., M35).
In addition to I48, we identified 25 other umbrella ICD‑10 codes that also appear in field 41270 despite having defined subcodes. According to the ICD‑10 classification, I48 is not selectable as a standalone diagnosis—only its specific subcodes are valid entries—so the presence of the umbrella code may lead to inflated case counts, as some diagnoses are effectively counted twice.
We identified ~7.4k individuals with the umbrella code I48 in their hospital records. The total number of unique affected individuals across all 26 codes is ~13.9k.
Has anyone encountered this issue before, and is there guidance on how these umbrella ICD‑10 codes should be interpreted or filtered when constructing phenotype definitions using field 41270?
Thank you.
List of 26 umbrella ICD‑10 codes that also appear in field 41270 despite having defined subcodes:
['A09 Diarrhoea and gastro-enteritis of presumed infectious origin',
'C80 Malignant neoplasm without specification of site',
'E05 Thyrotoxicosis [hyperthyroidism]',
'F06 Other mental disorders due to brain damage and dysfunction and to physical disease',
'F10 Mental and behavioural disorders due to use of alcohol',
'F20 Schizophrenia',
'F23 Acute and transient psychotic disorders',
'F32 Depressive episode',
'G40 Epilepsy',
'I48 Atrial fibrillation and flutter',
'I73 Other peripheral vascular diseases',
'K21 Gastro-oesophageal reflux disease',
'K29 Gastritis and duodenitis',
'K43 Ventral hernia',
'K85 Acute pancreatitis',
'L89 Decubitus ulcer',
'M75 Shoulder lesions',
'O60 Preterm delivery',
'R17 Unspecified jaundice',
'U51 Admission status on psychiatric patients',
'V49 Car occupant injured in other and unspecified transport accidents',
'V85 Occupant of special construction vehicle injured in transport accident',
'W18 Other fall on same level',
'W45 Foreign body or object entering through skin',
'W79 Inhalation and ingestion of food causing obstruction of respiratory tract',
'X58 Exposure to other specified factors']
Comments
1 comment
Dear Mitja,
Thank you for your query.
The HES data (and other linked data) occasionally include invalid or incomplete codes within the dataset. The data provider will sometimes provide umbrella codes such as I48 which are still valid codes, even though they may not be selectable. These could be also be codes like I48X, I48-, I48? etc. We perform some verification and cleaning rules for each data-field. These rules check whether the ICD and OPCS codes correspond to the correct version of the coding system, and that values are within the given coding range and are in a standard format. Where ICD or OPCS codes contain trailing characters (such as dashes and X’s) or other additional characters that are not part of a valid code, UK Biobank applies cleaning rules to strip the trailing characters. Asterisks and daggers, which have a specific meaning in ICD-10, are made available in the addendum fields, however a portion of Scottish data is missing data in the addendum fields. We are looking into addressing this in due course.
This is the reason why you see umbrella ICD-10 codes with defined sub-codes having counts against them. This feature of the data originates from the original data source. You can also refer to Resource 138483 for more information on HES data processing.
Broadly, any umbrella ICD-10 code in field 41270 is meant to be interpreted as "there are x cases of code XX.X either as the umbrella term (where a more specific diagnosis is not available) or as the subtype of the ICD-10 code". I understand that this might be confusing for case selection. We are seeking clarification from the data providers on this, and will update our resources if we receive any correspondence regarding this.
We cannot provide advice on whether to include or exclude these codes for your case selection or guidance on research methodology or techniques, as these are specific to individual research projects. For advice on research methodology and techniques, we open the question to researchers and experts in the field who may be able to offer valuable insights and support.
Please sign in to leave a comment.