This issue of missing the ‘non-admitted’ patients has been a massive source of anxiety for us. It is only relevant if you are following the ‘all referrals’ or ‘prospective data flow’ approach. In which case, then we have to completely depend on the quality of the initial data capture.

If a month’s data can only be completed by ‘backfilling’ (i.e. pulling notes) then we will flag that as a time period where it is likely that the patients not admitted are a biased sample. Where there is little ‘backfilling’ then we can be more sure that we have not missed patients seen but not admitted.

We will not discard any data but this will permit us to run a sensitivity analysis and makes sure that our conclusions are not affected by months with different levels of data completeness.