A risk register that separates signals from harm
6 min read
Lesson objectives
- Apply a risk register that separates signals from harm to a service case with evidence and decision limits.
Please sign in to use all the tools
Once you sign in, the reading tools are yours: larger or smaller text, typeface, line spacing, focus mode, highlighting and bold, and notes saved to your account and gathered in My study.
A useful risk register is not a red-colored incident list. It is a decision tool for finding where a service should examine and prevent a possible failure before every vulnerability becomes an event. In the referral service, nine requests may lack visible acknowledgement. That is an operational observation. It does not yet tell us how many were lost or whether anyone was harmed. The risk worth recording is that a request requiring follow-up could remain without an owner because transfer of responsibility cannot be demonstrated. This describes a testable failure mechanism, not a personal accusation or an unmeasured clinical outcome. A risk register differs from an incident log. An incident is an observed event or signal at a particular time. A risk entry describes a possible pattern of failure, plausible consequences, existing controls and gaps that could permit recurrence. Actual incidents, near misses and prospective process reviews can all inform priority, but each source has limitations. More reports alone do not prove more harm; detection or confidence in reporting may have improved.
Begin a register entry at one step in the service map. Describe the step's purpose, how it could fail, possible causes, how the failure would become visible, and its plausible effect. This is the logic of failure mode and effects analysis: examine a process prospectively rather than waiting for every weakness to cause an event. Suppose receipt requires acknowledgement against a referral identifier. A failure mode is no acknowledgement that can be matched to the request. Possible causes include coordinator absence, identifiers that differ across systems, or failed synchronization. These are hypotheses, not one established cause. The administrative effect is an unresolved task. An effect on an individual patient's care requires qualified assessment; it cannot be inferred from missing acknowledgement alone. Record the current control, perhaps a daily exception queue. Then ask whether it detects failure before a locally critical interval and whether someone responds. An alert without a response owner is incomplete control. Risk scoring also needs restraint. Ratings can structure a discussion, but with weak data they do not measure the probability of harm precisely.
How should risks be ordered when a trustworthy probability is unavailable? Consider potential consequence, exposure, signal frequency and the chance of detecting failure before it has an effect, while stating assumptions. If two clinics rate the same risk differently, compare their severity definitions, referral mix and detection time before combining scores into a common list. Multiplying estimated ratings can suggest precision the process does not have. Do not let a score of 48 automatically outrank 45 as though both numbers were calibrated measurements. Give particular attention to severe potential consequences even when reports are rare, and to gaps that current controls cannot see. When an estimate is unsupported, label it uncertain and specify an evidence-gathering test, such as sampling referrals across shifts or checking whether an alert reaches the on-duty backup. The resulting decision is not to defer safety. It is a proportionate precaution alongside a plan to improve knowledge.
The action field should not end the entry. Describe the control mechanism: reconciliation by a shared identifier, a work queue that stays open until receipt is proven, or absence coverage for the coordinator. Name an implementation owner and, where appropriate, a separate impact reviewer. Set a check date and success evidence beyond saying training was delivered. Instructions can be published while four requests still have no receipt evidence. Track matched acknowledgement and exception-resolution time, with a balancing measure for duplicate referrals if resending is introduced. An individual case that may need urgent action follows the clinical and safety escalation channels immediately; it does not wait for the next risk-priority meeting. Separating those routes preserves the register's value as a service management tool without allowing it to obstruct individual escalation.
Treat register entries as revisable hypotheses. If two weeks of review show that most exceptions are caused by known synchronization delay, do not erase the entry and announce the risk has vanished. Repair the measure definition, distinguish a technical signal from genuine absence of receipt, and check that synchronization work has not concealed real exceptions. If staff interviews reveal manual copying of a work list at shift change, add that route to the map instead of preserving an obsolete diagram. Date each change to the risk description, exposure and control so that ratings before and after the revision are not presented as an uninterrupted series. The purpose of a register is not to win a dashboard contest for the least red shading. It is to improve the organization's ability to anticipate a failure, see evidence of it and implement a control whose effect can be checked. An entry should therefore be concise enough to use and specific enough for someone else to challenge.
Read the entry aloud at a review meeting:
- a request may have no owner because acknowledgement cannot be matched to sending
- four of 400 requests remain unresolved
- a named coordinator checks their destinations and returns with evidence
If the entry cannot be explained plainly, it may combine several causes or effects that one owner cannot test. Split distinct mechanisms into testable entries, but do not multiply the same failure into ten rows to manufacture ten scores. Check who bears the control's side effects as well. Delay may fall for the sending team while workload moves to receivers. Examine that effect before calling the intervention successful.
Lesson summary
Apply a risk register that separates signals from harm to a service case with evidence and decision limits.