ISO Compliance Insights & Best Practices
CART
ISO Compliance Insights & Best Practices
A potassium of 7.4 sits in the laboratory system, correctly measured, quality-controlled and released. Nobody has told anyone. Four hours later a clinician opens the record for another reason and finds it. Nothing in that sequence was an analytical error, and preventing it is precisely what critical value reporting exists to do.
It is the failure mode with the most direct route to patient harm in a medical laboratory, and it is the control most often written down incompletely — thorough on what to notify, silent on what happens when the telephone rings out.
A critical value is a result indicating a risk of immediate harm and requiring urgent clinical action. It is not the same as a markedly abnormal result, and it is not the same as an alert value — a significantly abnormal result requiring prompt but not immediate attention. Laboratories that conflate the three end up notifying constantly, which is its own failure.
Clause 7.4.1.6 of ISO 15189 requires results falling within established critical intervals to be notified immediately, with records of the notification. Two words in that do the heavy lifting: established, meaning agreed and written down in advance, and notified, meaning it reached a person — not a system.
Critical value reporting starts with the list. It is agreed with the clinicians who use the service, approved by the laboratory director, published where requesters can see it, and reviewed at a defined interval. Three things separate a list that works from one that does not.
Partition it. The clinical threshold differs by population. Neonatal bilirubin, paediatric potassium, and results in patients on dialysis or chemotherapy all need their own limits, or the list generates false urgency in one group and misses real urgency in another.
Include qualitative results. Most lists are entirely numeric, which quietly omits some of the most urgent findings a laboratory produces: a positive blood culture, a significant finding in cerebrospinal fluid, a notifiable organism, an incompatible crossmatch, a critical histopathology finding.
Keep it short. This is the counter-intuitive one. Where every abnormal result is critical, notification becomes routine, the people receiving the calls stop treating them as urgent, and the genuinely dangerous result is lost among them. A long list feels safer and performs worse.
Confirm the result is real. A brief check costs a minute and prevents a clinical response to an artefact:
| Check | Why |
|---|---|
| Quality control for the run acceptable | A critical value from an uncontrolled run may be an instrument fault |
| Plausible against the patient’s previous results | A delta check failure has three explanations: the patient changed, the sample is not theirs, or the measurement is wrong |
| Sample integrity | Haemolysis for potassium, clotting for platelets — the two commonest false criticals, and both pre-analytical errors |
| Identity not in question | A critical result on the wrong patient produces treatment for someone who does not need it |
Confirmation must not delay notification beyond the time the laboratory has defined. Where it would, notify with the caveat and follow up — “this is a critical potassium, the sample looks haemolysed, we are repeating it now” is a useful call. Silence while you repeat it is not.
The obligation is discharged when a person able to act has received the result and confirmed what they heard. In order:
The read-back is the control. Critical results have been transposed, misheard and attached to the wrong patient in the few seconds between being spoken and being written down, and the read-back is what catches that.
A message left with someone who cannot act — a ward clerk, a voicemail, a colleague who happened to answer — is not notification. It may be a step towards it, and it should be recorded as one, but the clock keeps running.
This is the part most procedures omit, and it is precisely the situation that produces harm. Writing “notify the requester immediately” and stopping there leaves a biomedical scientist at 2 a.m. holding a dangerous result and no instructions.
Define the escalation explicitly, with times:
| Attempt | Contact | If no answer within |
|---|---|---|
| 1 | Requester or ward directly | A defined number of minutes → attempt 2 |
| 2 | Ward senior nurse, on-call team, or switchboard bleep | A defined number of minutes → attempt 3 |
| 3 | Duty consultant, site manager or the agreed emergency route | A defined number of minutes → attempt 4 |
| 4 | Laboratory director, and the escalation agreed with the organisation | Record as unnotified and escalate as an incident |
Two rules go with it. The laboratory does not stop trying because a shift ended — handover of an unnotified critical result is explicit, recorded, and to a named person. And a critical result that was never successfully notified is nonconforming work, and where harm may have followed, an incident to be escalated outside the laboratory.
A critical result on a patient at home, requested by a practice that closed at six, is the hardest version of this problem. The requester is unavailable and the patient is not somewhere anyone can reach them quickly.
The route has to be agreed in advance with users — the out-of-hours service, the emergency department, or a named on-call arrangement — and written into the service agreement. Working it out at 9 p.m. on a Friday, from a laboratory, with a name and a telephone number that no longer connects, is not a plan.
Every instance of critical value reporting is recorded with: patient identifiers and examination, the result, the time it became available, the time notification succeeded, who was notified and their role, whether a read-back was obtained, how many attempts it took, and who made them.
The measure that matters is the interval between availability and successful notification. Counting how many critical results were notified tells you the process ran. Timing them tells you whether it worked. Report monthly: the count, the proportion notified within the defined time, the proportion needing escalation, and — separately and prominently — any that were never notified at all.
Set the critical value reporting target at 100% and treat every miss as nonconforming work. Patterns by ward, by hour of day or by examination are almost always systemic and almost always fixable.
Critical value reporting also has to say what happens the second time. Decide in advance whether a repeat critical result on the same patient needs renewed notification. The common rule — notify on first occurrence and on significant change, but not on every subsequent result within a defined period — is sensible, and it should be written down rather than left to whoever is on shift.
Point-of-care testing inverts the problem. A critical result appears in front of the person treating the patient, so notification is immediate by definition, but the result still has to reach the patient’s permanent record. Where a device is not connected to the laboratory system, define how that happens and how the laboratory becomes aware of it — a critical result acted on correctly and recorded nowhere is a governance failure even though the patient came to no harm.
Our ISO 15189 Toolkit includes a critical and alert value notification procedure carrying the list template with partitioning, the pre-notification checks, the read-back requirement, a four-step escalation table, the out-of-hours route for community patients, and the notification record — alongside the results review and amended reports procedures the process depends on.
A result indicating a risk of immediate harm and requiring urgent clinical action. It is distinct from an alert value, which requires prompt but not immediate attention, and from a merely abnormal result.
The laboratory agrees it with the clinicians who use the service, and the laboratory director approves it. It is published to requesters and reviewed at a defined interval, and it should be partitioned by age and clinical group where the threshold differs.
No. Notification requires reaching a person who can act on the result and obtaining a read-back. A message left with someone who cannot act is a step towards notification, not notification, and the escalation clock keeps running.
The procedure should define a numbered escalation with times — ward, then senior nurse or on-call team, then duty consultant or site manager, then the laboratory director and the organisation’s own escalation. An unanswered telephone must never be where the process stops.
By the interval between the result becoming available and notification succeeding — not by how many notifications were made. Report the proportion notified within the defined time, the proportion requiring escalation, and any that were never notified.
Yes. Clause 7.4.1.6 of ISO 15189:2022 requires immediate notification of results falling within established critical intervals, and records of it. Accreditation bodies commonly examine the records rather than the procedure, because the records are where the process either worked or did not.
August 13, 2026
August 13, 2026
