Identifying cerebral palsy from routinely-collected data in England and Wales.

Purpose: An observational study using routinely-collected health care data to describe the extent to which children and young people (CYP) with cerebral palsy (CP) can be identified and the prevalence of CP can be estimated. Patients and methods: Routinely-collected anonymized data, for CYP (aged 0-25 years old between 1 January 2004 and 31 December 2014) were analyzed in two linked datasets, from England and Wales respectively. Datasets included National Health Service; General Practitioner (GP), inpatients, outpatients, and national mortality records. CP was identified using ICD-10 codes G80.0-G83.3 and equivalent Read v2 codes. Ascertainment rates of CP were identified for each data source and compared between countries. Frequency and consistency of coding were investigated, and prevalence of CP estimated. Results: A total of 7,113 and 5,218 CYP with CP were identified in the English and Welsh datasets respectively. Whilst the majority of CYP with CP would be expected to attend their GP, 65.3% (4,646/7,113) of English and 65.1% (3,396/5,218) of Welsh cases were ascertained from GP datasets. Further cases were identified solely in inpatient datasets (2,410 in England, 1,813 in Wales). Few cases were coded for CP within outpatient datasets. Four character codes that specified CP type were rarely used; one in five health care records were coded both with G80 codes (explicitly CP) and with G81-83 codes (other paralytic syndromes) or equivalent Read codes. Estimated period prevalence of CYP with CP was 2.5-3.4 per 1,000 in England and 2.4-3.2 per 1,000 in Wales. Conclusion: In England and Wales, coding of CP in routine data is infrequent, inconsistent, non-specific, and difficult to isolate from conditions with similar physical signs. Yet the prevalence estimates of CP were similar to those reported elsewhere. To optimize case recognition we recommend improved coding quality and the use of both primary and secondary care datasets as a minimum.