POCT transcription errors: causes, cost and how to remove them

What a transcription error means in POCT

A transcription error is a result that changes between the analyser and the record. The analyser produced the right number; the number in the notes is different. In point-of-care testing (POCT) it happens in three ways.

• Screen to keyboard. The operator reads the value from the meter and types it into the electronic patient record (EPR), the GP system, the laboratory information system (LIS) or a paper form. This is the classic case and the most common.
• Printout to notes. The device prints a slip that is stuck in the notes, or copied from the slip into a chart. The slip fades, falls out or is filed against the wrong patient, and anything copied from it carries a second chance of error.
• USB or manual download. Results are pulled from the device to a spreadsheet on a memory stick and uploaded later. There is no re-keying, but there is no patient matching at the moment of testing either, so the wrong-patient and lost-result errors survive.

The common feature is a human step between the measurement and the record. Every step of that kind has a failure rate, and every one of them can be removed with a connected device. The wider picture is set out in What is POCT connectivity?.

The error types

Rounding belongs on the list too: a device reports 6.95, the field accepts one decimal place, and the operator types 6.9 or 7.0 without a rule for which. Individually small, these become a problem when a result sits close to a decision threshold.

Why transcription errors are under-reported

Transcription errors are under-reported for reasons that are structural, not personal.

• They are invisible when the wrong value is plausible. A transposed glucose is only found if someone compares the record with the device memory, and nobody does that routinely.
• The device memory is the only evidence, and it is short. Most meters hold a limited number of results; by the time a discrepancy is noticed the original may have been overwritten.
• Nobody owns the comparison. The laboratory does not see manually entered POCT results; the ward does not see the device log; the POCT coordinator sees neither unless connectivity exists.
• Reporting feels like self-incrimination. The person who finds the error is often the person who made it, and the incident form asks for a name.
• Harm is rarely traced back. A repeat test or a delayed decision is absorbed into the working day and never linked to the entry that caused it.

The consequence is that an organisation's incident system will usually show a handful of POCT transcription events a year, while the device logs, if anyone looked, would show a different picture. In England, patient safety events are now recorded through the Learn from Patient Safety Events (LFPSE) service, and an organisation that wants a true figure has to go looking for it rather than wait for reports to arrive.

What the published literature says

Laboratory medicine has studied its own errors for thirty years, and the consistent finding is that most errors occur outside the analytical step: in the pre-analytical phase (the wrong patient, the wrong sample, the wrong label) and the post-analytical phase (the wrong result reaching the wrong place, or not reaching it at all). The studies of a hospital stat laboratory by Plebani and Carraro, published in Clinical Chemistry in 1997 and repeated a decade later, and the review by Bonini and colleagues in the same journal in 2002, are the standard references. Their message is that analysers are rarely the weak point; the handling of identities and results around them is.

Studies that look specifically at manual entry of results, in laboratories and in point-of-care settings, report error rates that are not negligible, and the rate depends on who enters the result, how many results they enter, how the entry screen is designed and whether anyone checks. The one intervention that removes the error class altogether, rather than reducing it, is electronic transmission from the device to the record, because it removes the step in which the error occurs. This page does not quote a percentage, because the published figures vary widely with setting and method and a single number would mislead; if you need one for a business case, take it from your own audit, described below.

The regulatory documents point the same way. ISO 15189:2022 requires, in its post-examination clauses on reporting of results, that results are transmitted accurately and completely, and its Annex A applies the same requirements to point-of-care testing under the laboratory's responsibility. The MHRA guidance Management and use of IVD point of care test devices expects POCT results to be recorded in the patient record with enough detail to identify the patient, the operator, the device and the time, and describes data management and connectivity as the means of doing so reliably.

The cost of one error

The cost of a transcription error is paid four times.

1. Repeat testing. When a result looks wrong, the test is run again: another cartridge or strip, another sample, another ten minutes of an operator's time, and on a molecular or cardiac marker device a consumable that costs more than the labour.
2. Delayed or wrong treatment. A plausible wrong value does not get repeated; it gets acted on. The delay while a discrepancy is noticed, or the decision made on the strength of a wrong number, is the cost that matters clinically, and it is the one that never appears in a budget line.
3. Incident investigation. A reported error takes a coordinator, a ward manager and often a governance lead through a root cause analysis, a report to the POCT committee and an action plan. Several hours of senior time per incident is usual; more where harm is involved.
4. Litigation and regulatory exposure. A record that cannot show who ran a test, on which device, with what QC status, is difficult to defend. A transcription error found during a complaint or a claim turns a clinical question into a governance one. For an accredited service, a pattern of manual entry errors is a non-conformance against ISO 15189 that the assessor will follow up at the next visit.

Set against this, the cost of connecting the device is a subscription and an interface. The POCT ROI calculator puts your own repeat rates and staff time against the connection cost.

How electronic connectivity removes the step

Connectivity does not make people more careful. It removes the moment at which care was needed.

• Bidirectional worklist. The middleware sends the patient list or the order to the device. The operator picks the patient from the screen or scans a wristband; nothing is typed. The result returns already matched to that order. The message flows are described in HL7 vs POCT1-A2 vs ASTM.
• Positive patient identification. The device reads the patient identifier from a barcode, checks it against the list it holds or queries the patient index, and refuses to run a test on an unknown identifier if the site's policy says so. Wrong-patient entry becomes a rare event with an audit record rather than a routine risk.
• Result matched to order and sent with its context. The value, the unit, the flags, the reagent lot, the QC status, the device and the time travel together in one message. A missed flag is far less likely, because the flag travels as part of the result; validate at go-live that every required flag survives transmission and displays correctly.
• Operator recorded. The device knows who is logged in, and the middleware attaches that identity to the result. The audit trail answers who, when and on what without anyone writing it down.
• No lost results. Every result the device produced reaches the middleware, including ones that were never matched to a patient; those sit in a queue for a person to resolve, rather than vanishing.

Three workflows compared

What to audit this month

You do not need connectivity to find out what manual entry is costing you. One audit, one month, one device type.

1. Pick the highest-volume manually entered device: usually blood glucose meters on a ward, or HbA1c or INR in a practice.
2. Download the device memory for a defined period: a week for a busy ward, a month for a practice.
3. For every result in the device log, find the matching entry in the record. Note results that are missing, results that differ, and results in the record with no device log entry.
4. For each discrepancy, classify it against the table above: transposed, wrong patient, wrong unit, decimal, missed flag, lost, wrong time or operator.
5. Count repeat tests in the same period and ask, for each, whether the repeat happened because the first result was doubted.
6. Note how long the comparison took. That time is a cost of manual entry too.
7. Report the counts, not the names, to the POCT committee alongside the connection cost for that device type.

The audit is the evidence base for the business case, and it is the figure to use in place of any published percentage. Repeat it after connection and the difference is the return.

Where software helps

Removing the transcription step is the core job of POCT middleware. Catenix connects analysers over HL7 v2, POCT1-A2, ASTM E1394 and FHIR through an on-site edge gateway that stores and forwards when the network drops, sends worklists down to devices that support them, matches every result to a patient and an operator, holds unmatched results in a queue for resolution, and delivers the result with its units, flags, lot and QC status into the LIS, EPR or practice system over open standards. The protocols and gateway options are on Device connectivity, and patient matching, the unmatched queue and critical result handling are described under Results governance.

Questions people ask

What is a transcription error in point-of-care testing?

A transcription error is a difference between the result the analyser produced and the result in the patient record, introduced when a person copied it. In POCT it happens when a value is read from the device screen and typed into the record, copied from a printout, or transferred by file and matched to the patient later. Transposed digits, wrong patient, wrong unit and lost results are the common forms.

How common are transcription errors in POCT?

Published studies of manual result entry report rates that are not negligible, but they vary widely with setting, device and who does the entry, so no single figure applies to your organisation. The reliable way to know is a one-month comparison of a device's memory against the record. Organisations that do this usually find more discrepancies than their incident system had recorded.

Does a barcode scanner stop transcription errors?

It stops one type. Scanning the patient wristband or label into the device removes typed identifiers and most wrong-patient errors at the point of testing. It does nothing about the result itself if that is still read from the screen and typed into the record. The value, units and flags need to travel electronically from the device to the record as well.

Is a printout stuck in the notes a transcription error?

Not by itself, but it is a transcription risk. The slip can be filed against the wrong patient, fade, fall out, or be copied into a chart or an electronic record later, and each of those is a manual step with a failure rate. It also gives the record a result with no operator, QC status or device attached. Treat printouts as a temporary measure.

How does POCT connectivity stop transcription errors?

The device sends the result electronically, with the patient identifier, operator, device, lot, units, flags and time, into middleware that matches it to an order or a patient and forwards it to the record. Nobody reads a screen or types a number, so the step in which transcription errors occur no longer exists. Unmatched results are held for a person to resolve rather than lost.

Should POCT transcription errors be reported as incidents?

Yes. A wrong result in a record is a patient safety event whether or not harm followed, and reporting it is how the organisation learns its true rate. Report the near misses too. In England, use your organisation's route into the Learn from Patient Safety Events service. Report counts and error types to the POCT committee, and use them to decide which devices to connect first.

Sources and further reading

• Plebani M, Carraro P. Mistakes in a stat laboratory: types and frequency. Clinical Chemistry, 1997, volume 43, pages 1348 to 1351
• Carraro P, Plebani M. Errors in a stat laboratory: types and frequencies 10 years later. Clinical Chemistry, 2007, volume 53, pages 1338 to 1342
• Bonini P, Plebani M, Ceriotti F, Rubboli F. Errors in laboratory medicine. Clinical Chemistry, 2002, volume 48, pages 691 to 698
• ISO 15189:2022 Medical laboratories: requirements for quality and competence
• MHRA, Management and use of IVD point of care test devices
• NHS England, Learn from Patient Safety Events (LFPSE) service