A service risk register: from score to control decision
6 min read
Lesson objectives
- Prioritize service risks using likelihood, impact, detectability and control strength without treating a score as clinical authority.
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 risk register helps when it turns a broad concern into an event that can be observed and controlled. 'Privacy is high risk' does not tell the committee what to do. Describe the event: a message goes to the wrong identity record; a follow-up request remains in an unmonitored queue; or an appointment request cannot finish during a connection outage. For each event, identify a plausible cause, consequence, existing control, evidence that the control works, and decision owner. This is an operational design, neither a patient diagnosis nor a legal compliance ruling.
A committee may define an internal one-to-five scale for likelihood and impact, then multiply them for an initial comparison. That arithmetic is not a clinical risk standard. A qualified specialist reviews situations that might cause health harm. Record detectability and control strength separately; their quality cannot be inferred from a written policy. If an identity check is claimed, ask how often it detects mismatch before dispatch and how exceptions are handled. Do not subtract an invented control percentage from the score merely to make the risk look lower. Only a test can support a claim of control performance. Rank events with uncertainty visible, then choose whether to remove a function, narrow it, separate its pathway, or accept a bounded pilot with monitoring. The choice must match the committee's authority and the boundary of professional practice.
Each row of the register has seven fields: event, evidence or source of the estimate, likelihood, impact, existing control and test result, owner, and review decision. For A, the supplier's statement that 'identity is checked' is insufficient. Test two similar records and show that the system stops a wrong dispatch and informs the user without exposing another person's data. For B, test transfer to the backup channel and compare open requests before and after the outage. For C, send a synthetic message during shift change and look for acknowledgment or a visible escalation. If a test fails, do not reduce the risk on a promised fix. The register is not static: a new interface or supplier should trigger review of the assumptions it changed rather than an automatic copy of the old score.
When likelihood values are judgmental and there is no reliable incident history, write a confidence note beside the number. For event C that might read: 'Likelihood 2, low confidence, because the trial did not include a duty-holder change.' The shift test then becomes a way to learn, not a ritual to confirm a predetermined decision. Likewise, 18 successes in 20 telephone attempts do not predict that 90% of all users can obtain help during a wide outage. The sample may have been collected in ordinary working hours, while the alternative channel could be overwhelmed when the digital route fails. Ask under what conditions the test ran and whether call volume exceeded staffing capacity. One complaint is a signal for investigation, not an incidence estimate. By distinguishing observation from assumption, the register helps the committee select a test that reduces the uncertainty that matters before expansion.
Control selection does not finish the committee's job. Give each risk an early signal and an outcome measure. For identity error, a signal might be mismatches stopped before dispatch, while the outcome is messages confirmed as misdirected. More stopped mismatches could mean that the control works, or that input quality is worsening; interpretation requires context. During channel outages, monitor requests still open at the end of the period rather than merely counting clicks on the alternative. For unreceived messages, measure dispatch-to-acknowledgment time and display missing values instead of silently excluding them. The indicator owner names who reviews an anomaly and when it reaches a decision holder. If the definition of 'received' changes, preserve both old and new definitions so that a measurement change cannot be mistaken for an improvement.
The case scores are 3×4=12, 4×2=8 and 2×5=10. We then assessed impact severity and detectability separately because a follow-up message without receipt may call for containment despite its raw score of 10.
Hold follow-up messaging until receipt and backup-duty tests pass; test rejection of dispatch to a wrong identity; improve the appointment alternative based on observed failures.
The one-to-five scale is an invented local decision aid, not a medical or legal standard. An untested control promise cannot lower risk, and a small synthetic test cannot guarantee wide-scale operation.
If a backup-duty test acknowledged all 20 synthetic messages with a documented receipt log, evidence of detecting C would improve, but its initial 2×5=10 score would not disappear. The committee could revisit its decision with real operating evidence rather than treating the test as an absolute guarantee.
Lesson summary
The case scores are 3×4=12, 4×2=8 and 2×5=10. We then assessed impact severity and detectability separately because a follow-up message without receipt may call for containment despite its raw score of 10. Hold follow-up messaging until receipt and backup-duty tests pass; test rejection of dispatch to a wrong identity; improve the appointment alternative based on observed failures. The one-to-five scale is an invented local decision aid, not a medical or legal standard. An untested control promise cannot lower risk, and a small synthetic test cannot guarantee wide-scale operation. If a backup-duty test acknowledged all 20 synthetic messages with a documented receipt log, evidence of detecting C would improve, but its initial 2×5=10 score would not disappear. The committee could revisit its decision with real operating evidence rather than treating the test as an absolute guarantee.