Eating disorder-related clinical events in linked ALSPAC primary and secondary care records

This phenotype identifies eating disorder-related clinical events in linked primary and secondary care health records in the Avon Longitudinal Study of Parents and Children (ALSPAC). It includes clinical codes and dataset-specific ascertainment criteria used to identify linked eating disorder-related clinical events across primary care, Hospital Episode Statistics and Mental Health Services Data Set records.

Primary-care concepts are separated by coding system, including Read v2, CTV3/Read v3 and SNOMED CT. These codelists comprise clinically reviewed codes judged to be eating-disorder diagnostic codes, rather than codes for suspected eating disorders, symptoms, referrals or screening.

The Read v2 and CTV3/Read v3 concepts are ALSPAC-focused lists, restricted to eating disorder diagnostic codes identified in relation to ALSPAC linked primary-care data and cross-checked against ALSPAC/UKSeRP Read-code resources. External Read-code catalogues were used during code identification and review, but the submitted concepts should be interpreted as ALSPAC-focused implementation codelists rather than exhaustive Read v2 or CTV3 eating disorder codelists.

The ALSPAC primary-care SNOMED code catalogue available to researchers was searched for eating disorder-related codes, which were then reviewed and rated. This should be interpreted as an ALSPAC-specific SNOMED CT implementation list, not as an exhaustive current SNOMED CT UK Edition eating disorder codelist.

ICD-10 F50 codes were used in HES and MHSDS datasets to define eating disorder-related clinical events. F50.4, “Overeating associated with other psychological disturbances”, and F50.5, “Vomiting associated with other psychological disturbances”, were excluded.

In MHSDS-VARS, eating disorder-related events were also identified using dataset-specific service-context fields, including eating disorder main specialty and eating disorder/dietetics team type. These service-context criteria are implementation rules rather than standard clinical codelists.

This phenotype should therefore be interpreted as an ALSPAC-focused linked-record eating disorder clinical-event phenotype, rather than a comprehensive national eating disorder phenotype covering all possible contemporary clinical terminologies.

This phenotype can be used to identify eating disorder-related clinical events in primary and secondary care datasets linked to ALSPAC. In practice, each clinical concept should be applied separately within the relevant data source and coding system: Read v2, CTV3/Read v3 and SNOMED CT for primary care; ICD-10 F50 codes for HES and MHSDS diagnosis fields; and MHSDS-VARS service-context fields for eating disorder main specialty or eating disorder/dietetics team type.

The phenotype is intended to support both event-level and participant-level analyses. In the ALSPAC B3832 implementation, the phenotype was used to identify linked eating disorder-related clinical events between ages 13 and 24 years and to examine overlap with questionnaire-derived disordered eating and self-reported eating disorder diagnosis or treatment. Researchers applying this phenotype in other datasets should check which coding systems, data fields and code versions are available, and should interpret the SNOMED CT and Read-code components as ALSPAC-focused implementation lists rather than exhaustive contemporary UK clinical terminology codelists.

Where datasets contain multiple records per participant, the recommended implementation is to first identify all matching event-level records within each individual dataset, then derive dataset-specific participant-level flags and dates, before merging with an analysis backbone. This preserves the ability to analyse individual event records while also allowing reproducible participant-level ascertainment.