The prevalence of hidradenitis suppurativa is shown by the Secure Anonymised Information Linkage ( SAIL ) Databank to be one per cent of the population of Wales

tumours that are strongly associated with specific gene disorders can be used to detect respective patients with genetic conditions from electronic health records. A related approach has been described, where diagnostic coding data are refined using multiple electronic records such as prescription data to improve identification of patients with diabetes. Patients with CCS in this series who had not been clinically diagnosed stand to benefit from genetic testing, counselling and skin surveillance. Secondly, the presence of associated tumours in CCS such as salivary gland tumours and malignant transformation can be monitored. The generalizability of this approach to other genetic disorders beyond CCS, where cancer and skin appendageal tumours may occur in combination, is important. Muir–Torre syndrome, where sebaceous carcinoma and bowel cancer develop, Birt– Hogg–Dub e syndrome, where trichodiscomas and renal cell carcinoma present, and Reed syndrome, where leiomyomas and renal cell carcinoma present, are pertinent examples. In these conditions, screening for malignancy could be initiated in relevant individuals, and in the case of Muir–Torre syndrome chemoprevention with aspirin can reduce cancer risk. Recently in the UK, national registration has been initiated for skin cancer. As we transition to a digital health service, it is conceivable that all skin pathology cases will be searchable, including skin appendageal tumours. We propose that the application of filtering algorithms could be applied to increasingly comprehensive electronic health records, allowing patients with an underlying genetic condition to be detected. Limitations of this proof-of-principle study are that some patients with CCS will have only one biopsy at the time of study, and typographical errors in free-text coding can result in omission of patients. Our method found that most patients with potential CCS identified from electronic pathology records during this interval had a confirmed diagnosis of CCS. Future studies, as genetic testing for CCS becomes established as a standard of care, should also investigate the number of patients with a confirmed diagnosis of CCS who are not identified using our method in the electronic pathology records. Nonetheless, these data are relevant, at a time when artificial-intelligence-based analysis of large datasets such as national cancer registries is underway. The ability to detect any potential genetic disease by proxy methods raises issues regarding consent, similarly to those raised in the context of conventional genetic testing. Public perception of genetic testing is changing, and patient group engagement will be important in guiding how such algorithms may be acted upon when they detect a patient with a genetic predisposition. For example, it may involve notification only when a helpful intervention is possible. It would be beneficial to conduct qualitative interviews in individuals with conditions such as CCS to capture their perceptions of the risks and benefits of being diagnosed in this manner. Patients have the right not to know about a genetic diagnosis, yet some may consider it a failing of a digital healthcare system if data point to a genetic diagnosis where screening or chemoprevention is possible, and yet no action is taken. We highlight that electronic health records are a rich resource, and continue to offer new facets that can impact on patient care.