Sepsis recording in primary care electronic health records, linked hospital episodes and mortality records: Population-based cohort study in England.

Explicit recording of 'sepsis' is inconsistent across healthcare sectors with a high proportion of non-concurrent records. Incidence estimates are higher when linked data are analysed.