Add hosp-from-A&E outcome

[wjchulme]

Hospitalisation data is slow but we can get close(ish) using A&E attendance + discharge location + positive test.

See report on this here: https://docs.google.com/document/d/1Ek6x1PUyVcdQeULWWWsjRBZFS_G3IsxJeKnwNWfytDM/edit

It's not as easy as defining the outcome once, as there may be multiple relevant attendances, positive tests, etc.

The study definition needs:

• Dates of any A&E attendance where discharge destination was hospital ward, emergency short stay ward, intensive care unit
• Dates of COVID-19 related A&E attendance where discharge destination was hospital ward, emergency short stay ward, intensive care unit

See emergency_attendance_date_X function in the study definition and adapt as appropriate. Also see https://docs.opensafely.org/study-def-variables/#cohortextractor.patients.attended_emergency_care for relevant filtering options.

Once extracted, define the proxy outcome as one of:

• attended A&E with COVID-19 diagnosis & discharged from A&E to hospital
• attended A&E within 2 weeks of a positive test and discharged from A&E to hospital

which will need to be defined in the process_data_long.R script.

The first definition will be more specific but less sensitive than the second. The second will have particularly low specificity when infection prevalence is high.