How to Reduce a Referral Backlog Without Losing New Work

Give the recovery project a precise boundary

A backlog project needs a start date, an accountable lead, a defined population and an administrative endpoint. "Clear referrals" is too vague: it could mean finish intake, obtain a booking or confirm returned documentation.

Choose one endpoint the team can verify. For example: every included intake episode reaches an accepted scheduling handoff or another authorized administrative disposition. That is a narrower promise than completing the patient's entire referral journey.

Record the cutoff and the systems included. Capture the work currently outside the central queue, such as an approved portal awaiting collection or a maintained manual list. New arrivals after the cutoff belong to a separate cohort even if the same staff handle them.

Use the inbound referral management playbook for ordinary receiving operations. This plan addresses a temporary accumulation that needs additional reconciliation and capacity.

The worksheets and calculations below are original proposed operating tools. Clinical priority and any required professional decisions remain inputs supplied through your organization's approved process.

Reconcile the opening list before setting a target

A raw record count is not necessarily a count of unique, unfinished referrals. Compare each item with its source and destination evidence.

Assign a stable episode reference. Link repeated transmissions without losing their receipt history. Check whether the expected action already occurred elsewhere. Preserve uncertain matches for designated review rather than silently merging them.

Hypothetical opening review: a team finds 260 records in its selected intake backlog. Twenty are repeated receipts linked to other records. Thirty represent work already completed before the project, with evidence available in another system. The opening active cohort therefore contains 210 episodes.

That reduction from 260 to 210 is record reconciliation. It is not 50 newly completed referrals or a measured improvement in patient access.

The 210 active episodes are classified as 120 with an administrative action available now, 60 awaiting an identified external dependency, and 30 whose status or association still needs investigation. These categories total 210. Each group remains in the project ledger until it reaches the stated endpoint.

Use a recovery ledger with a reason for every exit

The ledger should allow a covering colleague to understand both the remaining work and the evidence required to finish it.

Use the established referral tracking state dictionary instead of inventing a second permanent status system. The project ledger adds cohort membership and recovery evidence to that existing record.

Do not allow "aged out" to become a successful outcome. If the responsible team changes or withdraws a request, record that decision under the approved process. It is different from scheduling acceptance, and neither event should imply attended care.

Keep records excluded during reconciliation in a separate history tab or report. Someone reviewing the initial total should be able to find all 260 source records and explain what happened to each.

Protect new work before estimating clearance time

A project that clears old work by neglecting current receipts simply changes the age of the problem.

Estimate the capacity required to keep up with ordinary arrivals, then identify additional minutes available for the recovery cohort. Include staff coverage, training, review and expected exceptions in the plan. Do not treat every paid hour as queue-processing time.

The NHS England and NHS Improvement core model guide distinguishes required from available capacity. It provides UK service-planning context, not US staffing benchmarks or clinical-priority rules.

Hypothetical planning calculation: the team has 480 coordinator minutes available each day for the defined intake work. New arrivals average 40 episodes, requiring an assumed eight active minutes each. That uses 320 minutes and leaves 160 minutes for the opening backlog.

If the selected recoverable backlog tasks average ten active minutes per episode to reach the agreed endpoint, the plan supports 16 backlog exits per day. The 120 currently actionable episodes would require 7.5 such working days, rounded up to eight for a simple schedule.

This is a conditional workload estimate, not a promised clearance date. The other 90 episodes require separate dependency or investigation plans. Their future availability and effort are not established by the ten-minute assumption.

Plan for a busy day before it arrives

Repeat the calculation under a higher arrival scenario. In the same hypothetical model, 55 new episodes consume 440 of the 480 available minutes. Only 40 minutes remain, enough for four of the modeled backlog tasks.

At 60 new episodes, the entire 480-minute allocation is consumed by current intake. There is no modeled capacity left for backlog reduction.

Write down the response before those conditions occur. The operations lead might arrange qualified temporary coverage, redistribute work within available skills, reduce the planned recovery batch or revise the project timeline. The choice depends on actual resources and approved service commitments.

Do not solve an arithmetic deficit by deleting protected breaks, assuming unpaid overtime or counting a manager as available without agreement. If the project needs additional capacity, name its source and when it becomes available.

Keep the immediate support decision separate from a longer-term staffing or automation evaluation. Staff should not have to wait for a future project to receive necessary coverage today. The referral coordinator staffing model covers the weekly workload worksheet behind that longer-term evaluation.

Work dependencies without manufacturing progress

A referral awaiting information can still require administrative work, but repeated checking is not the same as advancing it.

For each dependency, record what is needed, who supplies it, the last request, the approved follow-up point and who investigates a missed response. Group similar requests where appropriate, while preserving individual episode references.

Before contacting anyone, recheck current evidence. A document may have arrived, an appointment may have been recorded, or the responsible team may have changed the request since the backlog was exported.

If an association remains uncertain, resolve that uncertainty before taking an action that could duplicate outreach or attach information incorrectly. The referral intake reconciliation guide covers source and episode relationships.

Keep the external dependency visible even after a reminder is sent. A sent message completes an attempt, not the dependency. Record the response or accepted next owner before counting the episode as an exit.

Reconcile daily movement using two cohorts

Use one equation for the active population:

Closing active work = opening active work + new arrivals + reopened episodes - verified exits.

Apply it to unique episodes within the same scope. Keep record corrections in a separate reconciliation line so they do not look like service delivery.

Continuing the hypothetical example: the project begins with 210 active episodes. Across three working days, the counts are below. The middle column shows opening active work + new arrivals - verified exits for each day.

There are no reopenings or further corrections in this illustration. Across the period, 210 + 120 - 168 = 162.

Now inspect cohort membership. Day 1 exits include 16 opening-backlog episodes and all 40 new arrivals. Day 2 exits include 12 opening-backlog episodes and all 45 new arrivals. Day 3 exits include 20 opening-backlog episodes and all 35 new arrivals. Each day uses exactly 480 minutes under the earlier assumptions of eight minutes per new episode and ten per backlog episode.

The project has resolved 48 of the original 210 episodes. It has also handled all 120 new arrivals on their receipt days. Reporting 168 "backlog referrals cleared" would overstate the opening-cohort result.

Check whether the recovery is harming another queue

Pair the opening-backlog count with new-arrival age, unowned work, reopened episodes, correction tasks and staff effort. Inspect the oldest unresolved record alongside the total.

IHI describes balancing measures as a way to check whether improving one part of a system creates problems elsewhere. Here, the practical question is whether recovery work displaces necessary current work.

Use the referral operations dashboard guide for consistent metric definitions. The temporary project view should show opening-cohort progress separately from the everyday performance report.

Review a sample of exits against source evidence. If scheduling cannot find a supposedly transferred record, restore the unresolved state and investigate. Preserve the correction rather than protecting the project's headline.

Close the project with accepted residual ownership

The final report should account for the original cohort, new work and any remaining dependencies. State how many episodes reached each authorized endpoint, how many remain unresolved and which team accepted them.

A project can end while some legitimately pending work continues, provided its continued ownership and follow-up are explicit. Do not describe that as every referral completed.

Record which fixes belong in ordinary operations: a corrected collection process, clearer handoff, maintained directory or accepted capacity change. Give each an owner.

For the continuing referral coordination workflow, the objective is a process that accounts for new work after the temporary recovery team leaves.

FAQ

Sources

• NHS England and NHS Improvement: Demand and capacity models, core model user guidance, May 2018, pages 18 and 21. Required and available capacity in UK service planning.
• IHI: Establishing Measures, balancing measures. The recovery ledger and examples in this article are original proposals.