Every working day, general practices receive discharge summaries, outpatient letters, results and reports. Each one may contain something the practice must do: change a medicine, arrange a blood test, book a review, record a diagnosis. Much of it needs only to be filed. The difficulty is that it is not obvious, at a glance, which is which.
Most correspondence does not need a GP — but some of it really does
Practices across England have redesigned how they handle correspondence so that trained staff process routine items and pass on only those that need a clinician. NHS England has published case studies of practices doing this. In one, an audit of 1,687 items over six weeks found that 21% needed clinician action and 79% needed only filing; the practice estimated about an hour of clinician time a day was freed. In another, the practice reported a 78% reduction in correspondence going to its GPs after an audit, a written protocol and staff training.
These approaches work because they are designed: clear rules for what a clinician must see, training, and checking. Defence organisations have also described what can go wrong — for example, a letter filed without a GP seeing it, where the patient was vulnerable and the letter mattered. The lesson is not that delegation is unsafe; it is that delegation needs a system.
The risk is in the requests
A letter is not just information. It often contains requests addressed to the practice. A UK study of discharge summaries sent to general practices found that, among summaries that asked the practice to do something, 46% (112 of 246) had at least one request that was not carried out. Requested medication changes were not made in 17% of cases (124 of 750), and around a quarter of requested tests and follow-ups did not happen.
None of this suggests carelessness. It reflects how the work is structured: the request sits in paragraph three of a two-page document; the person who reads it is not always the person who can act on it; and there is rarely a check that the action happened.
What a better system needs to do
From the practices that handle correspondence well, and from the research, a few requirements emerge:
- Identity first. Before anything else, confirm the document belongs to the patient it is attached to.
- Find the requests, not just the topics. A summary of the letter is less useful than a list of the things it asks the practice to do, each linked to the exact sentence.
- Compare with the record. A dose change only matters if the record does not already reflect it.
- Treat uncertainty as uncertainty. "Consider stopping" is not "stop". Conditional advice needs a clinician's judgement.
- Route to the right person. Clinical decisions to clinicians, bookings to the team that books.
- Check that it happened. The result received and reviewed, the record updated, the appointment attended — confirmed from the source, not from a ticked task. A booking on its own shows only that something is planned.
Where software can help, and where it should not
Software can do the reading at scale, find the requests, show the passage and the record side by side, and keep track of whether the work was completed. It should not decide what a letter means for the patient, and it should not file or change anything without a person's decision.
That distinction is the basis of how Reedley is being designed. It is also why we describe what exists today carefully: Reedley has been built and tested on synthetic documents only, and its accuracy on real correspondence has not yet been measured.
Sources
- NHS England. Improvements in document management frees up one hour of clinician time a day (Parkview Surgery case study).
- NHS England. 78% reduction in correspondence going to GPs — Northgate Medical Centre.
- MDDUS. Time-saving or risky business? Risk alert, April 2019.
- Spencer RA, et al. Processing of discharge summaries in general practice. British Journal of General Practice 2018;68(673):e576–e585.
This article is general information about UK primary care. It is not clinical, legal or regulatory advice. Figures are taken from the cited sources and are not Reedley results.