Skip to main content

Clinic Recall Metrics: Measure Reviewed Care, Not Messages Sent

A useful clinic recall measure starts with a defined group of care tasks that became due, a fixed reporting date and evidence of what happened next. Count a reviewed result or completed clinical review separately from an appointment, a delivered message or a patient's decision to decline. Track unresolved work and staff effort alongside completion; a higher return rate alone does not show better care.

By Bractis · Updated · 10 min read

Clinic owners, doctors, nurses and recall coordinators building a small, auditable follow-up dashboard for an Indian outpatient practice.

1. Define the care task before counting the reminder

A recall episode is one identifiable follow-up obligation arising from a care plan: for example, a clinician-requested review of a result or a reassessment with specified information. It is not each WhatsApp message, each phone attempt or every visit by the same person. The intended action and its evidence of completion should be written when the episode is created. Otherwise staff can improve a dashboard simply by changing what they call finished.

Begin with a bounded question: of the episodes due during a particular week, what had reached their agreed next step by the reporting date? Freeze the eligible cohort and record both dates. A task first becoming due after that week belongs to a later cohort. Correct genuine data errors with an explanation; do not quietly remove difficult cases from the denominator.

A person can have two independent episodes, such as a medicine review and a referral requiring an outside report. Keep their clinical identities separate while coordinating contact. This avoids both double-counting people and losing one task when another is completed. A paper register can do this with an episode reference and patient reference; a new software dashboard is not a prerequisite.

  • Eligibility: a documented action is due under an agreed care plan; marketing invitations use a separate measure.
  • Unit: people, care episodes, appointments or contact attempts, named explicitly.
  • Observation: the due-date window and the later date on which status was checked.
  • Completion: task-specific evidence, such as documented clinical review and a next plan, with an accountable reviewer.

2. Use a complete status table before calculating a percentage

The following invented example contains 40 due episodes across 32 people. Eight people have two episodes each and 24 have one: 16 + 24 = 40. The table gives each episode one current reporting status at the cut-off. The underlying history can contain several events, but the summary must not count the same episode in every row it passed through.

Here, 'reviewed with a plan' means the intended clinical review occurred and its next decision was recorded. It does not claim cure or improved health. The six attended episodes still awaiting review remain unfinished. A reported outside transfer counts in the last row only when acceptance of the handover has been verified; merely handing the patient a referral letter is insufficient.

On a narrow screen, scroll the table horizontally to read all columns.

Original fictional cohort: 40 due episodes, assessed at one reporting cut-off
Current statusEpisodesWhat the count establishes
Reviewed with a documented next plan18The intended review is documented; health outcomes require separate assessment.
Attended, but intended clinical review still pending6Attendance occurred; the outstanding review still has an owner.
Booked, not yet attended5An appointment exists; the care task is still open.
Contacted, without a booking or agreed next step4Contact occurred; it did not resolve the task.
Unreachable at the cut-off3Status is unknown; absence of a reply does not prove refusal.
Informed decision to decline the proposed action2The decision and any agreed further action are documented.
Transfer accepted by the receiving service2The local handover obligation is documented; outside care is not assumed complete.

Source: Institute for Healthcare Improvement: Establishing Measures

3. Publish the numerator, denominator and limitation together

Clinical review completion in this cohort is 18 ÷ 40 = 45%. Attendance is 24 ÷ 40 = 60%, because it includes the six episodes that reached a visit without the intended review being finished. Labelling 60% as completed care would conceal the main bottleneck. Neither fraction is a national benchmark or a prediction for another clinic.

A wider administrative measure could be 'documented disposition': 18 reviewed episodes + 2 informed declines + 2 accepted transfers = 22, or 55% of the cohort. Label that measure literally. Declining an action is not the same clinical outcome as completing it, and an accepted handover does not prove treatment at the destination. Keep these component counts visible so a rise in dispositions cannot hide a fall in reviews.

The remaining 18 episodes need a defined next action: six reviews, five appointments not yet attended, four contacts without a next step and three unresolved contacts. A clinician decides how urgency and further contact should vary. The oldest administrative entry does not automatically have the greatest clinical priority.

Respect an informed decision without repeatedly pressuring the person to improve a completion score. If a decline includes an agreed later discussion or an outstanding safety action, that work needs its own owner. Use the reporting categories to describe care, not to erase the patient's preferences.

4. Separate cohort performance from today's outstanding queue

A fixed-cohort report and a live workload report answer different questions. The first follows a defined set of episodes. The second shows what staff must still address now, including older episodes and newly due work. Do not splice the numerator from one into the denominator of the other.

Consider a second, separate fictional period: the clinic opens with 18 unresolved episodes, adds 12 newly due episodes and records 10 dispositions under its published closure rules. Its closing queue is 18 + 12 − 10 = 20. Ten closures can therefore coexist with a growing backlog. Investigate the mix: reviewed care, accepted handovers and informed decisions should still be distinguishable.

Contact activity has another unit. If the original 40-episode cohort generated 60 attempts, that is 1.5 attempts per episode, not per person and not a completion rate. Shared-phone contact may concern several tasks, and an unanswered call is still an attempt. Keep a consistent recording rule for one call covering multiple episodes so effort is not inflated by duplication.

IHI describes outcome, process and balancing measures for improvement. For a recall workflow, useful companions include the intended care result, the steps leading to it and possible burdens such as staff time, unwanted contact or patient travel. The measures proposed here are local operational definitions, not a validated clinical score.

Source: Institute for Healthcare Improvement: Establishing Measures

5. Use the numbers to investigate one change at a time

Start by reading the unfinished records. If visits happen but results remain unreviewed, more reminder messages may increase the wrong activity. If people cannot travel on clinic days, an evening slot or a suitable local assessment may address a different obstacle. Establish the reason with patients and staff rather than inferring it from a status code.

Compare successive cohorts using the same definitions and observation period. A cohort allowed four weeks to mature cannot fairly be compared with one observed after three days. Record service cancellations, staffing changes and known data gaps. Where numbers are small, show counts alongside percentages; one episode can move the rate substantially.

Examine access barriers only where the information is appropriate to collect and sufficiently complete. A blank language field is unknown, not the default language. A patient using a shared phone is not necessarily unwilling to engage. Avoid exposing identifiable small groups in circulated dashboards. Operational review needs enough detail to act without distributing unnecessary health information.

A before-and-after rise is a reason to investigate, not proof that software caused better outcomes. Patient mix, available appointments, outside care and recording quality can all change. Agree a small improvement, describe what was actually tested and look for both benefit and burden before expanding it.

Why a busy recall desk can still have unfinished care

In the fictional 40-episode cohort, 24 episodes reached attendance but only 18 had the intended review and a next plan. An attendance-only dashboard would show 60%; reviewed care is 45%. The practical investigation starts with the six attended episodes awaiting review. The figures do not show that patients improved, that a particular reminder caused attendance or that the clinic should target a prescribed percentage.

Define a recall improvement measure

Use these prompts in your clinic’s own notes. Record the result, responsible person and next action for each item.

Define one denominator
Choose one recall type and write its eligibility, episode reference, due-date window and reporting cut-off. Identify how two tasks for one person will be counted.
Reconcile the whole cohort
Place every eligible episode into one current status. Check that all status counts sum to the denominator and retain an explicit category for unresolved information.
Locate the unfinished work
Distinguish booked, attended and reviewed episodes. Name the person responsible for the next action and the clinician who decides its priority.
Check benefit and burden
Choose one proposed change and a consistent observation period. Record staff effort or unwanted contact as well as the intended care measure.

Questions doctors ask

What should a clinic count as a completed recall?

Define completion for the specific care task before measuring it. For a result-review recall, attendance or report receipt alone is insufficient: the intended clinical review and its next plan need documentation. Record informed declines and accepted transfers separately so an administrative disposition is not mistaken for the same clinical outcome.

Should the denominator be patients or follow-up episodes?

Use the unit that answers the question and label it. People measure reach; episodes measure distinct care obligations. One person may have more than one obligation. Keep a patient reference and an episode reference so reporting can avoid counting repeated messages as additional patients or closing two tasks when only one was addressed.

Why can the pending list grow while staff complete more follow-ups?

Newly due work can exceed documented closures. In a separate fictional period, 18 opening episodes plus 12 new episodes minus 10 dispositions leaves 20 pending. A fixed-cohort completion measure and a live workload measure should therefore be reviewed together, with their different denominators made explicit.

Can unanswered messages be recorded as declined care?

No. An unanswered message leaves the person's status unknown. An informed decision to decline requires an attributable decision and appropriate documentation, including any agreed further action. Clinical priority, contact preferences and proportional further attempts should be reviewed through the clinic's care process.

Does a higher recall rate prove the clinic's software improved health outcomes?

No. The rate may reflect appointment availability, patient mix, outside care, a longer observation period or changed recording. Use stable definitions, visible counts and appropriate patient outcome measures. Describe observed changes honestly; do not turn a local before-and-after comparison into a causal or marketing guarantee.

Sources and scope

Updated on . This guide proposes original operational definitions and fictional arithmetic, not clinical outcome benchmarks or a validated measurement instrument. Clinicians remain responsible for individual care priorities; data collection and reporting should respect patient choices and appropriate access. No independent clinical or legal review is claimed.

Try the workflow
with Bractis.

In a Bractis demonstration, ask the team to trace a due care task through booking, attendance, review and a documented next step. Check whether exports preserve episode identifiers, uncertainty and the reporting dates required by your own measure. Confirm the current product behaviour before relying on it for a clinic report.

Request a clinic demo