POCT KPIs worth tracking: measures of quality, not just activity

Why most POCT services report activity but not quality

Ask a point-of-care testing (POCT) service for its numbers and you will usually get test volumes: how many glucose tests on the wards, how many HbA1c results in the practices, how many blood gases in the emergency department. Volume is easy to count because devices count it, and finance asks for it. It tells you the service is used. It does not tell you whether it is safe.

Quality data exists, but it is scattered. Quality control (QC) results sit in device memories or on paper logs. External quality assessment (EQA) reports arrive by email to whoever registered the scheme. Competency records live in a spreadsheet, or several. Incidents are in the organisation's incident system under whatever category the reporter chose. Nobody owns a definition of "QC pass rate" for the whole service, so nobody calculates it, and the POCT committee reviews volumes and anecdotes.

The standards expect more. ISO 15189:2022 clause 8.8.2 requires the laboratory to establish quality indicators, plan how each will be measured and interpreted, and review them periodically; Annex A applies this to POCT. The Medicines and Healthcare products Regulatory Agency (MHRA) guidance on the management and use of IVD point-of-care test devices expects the POCT committee to review the performance of the service, not only its size. An assessor who sees a committee pack made of volumes will ask what the committee actually oversees.

The set below is deliberately broad, because the failures in POCT are broad: the result that never reached the record, the operator whose competency lapsed, the device running an expired lot, the sample from the wrong patient. Choose the ten or twelve that fit your service, define them, and keep them stable for long enough to see a trend.

What makes a KPI usable

A key performance indicator (KPI) is only useful if two people would calculate the same number from the same data. Before adding one to the pack, write down five things.

1. Definition. One sentence that says what is being counted and what is excluded. "QC pass rate" needs to say whether repeated runs count, and whether a run rejected by a warning rule counts as a fail.
2. Numerator and denominator. Most POCT KPIs are rates, and the denominator is where arguments start. Per 1,000 results, per operator active in the period, per device scheduled hour: pick one and keep it.
3. Source and owner. Which system produces the number, and who is accountable for it and for acting on it. If the answer is "the coordinator, from six spreadsheets", the KPI will be late every quarter.
4. Segmentation. Always by site and device type, often by operator group. A service-wide average of twelve sites hides the one that is failing.
5. Direction of travel. Which way is good. Numeric targets come later, once you have four quarters of your own data; a target borrowed from another organisation with different devices and patients is a number, not a standard.

Resist the urge to measure everything. Twelve to sixteen indicators, stable from one quarter to the next, will produce discussion. Forty will produce a document nobody reads.

The KPI set

Sixteen indicators in six groups. Each has a definition, a calculation and a direction of travel. Where a per-volume rate is suggested, scale it to your own volumes so the numbers are readable.

Connectivity and safety: did the result reach the right record in time

The connectivity and safety groups measure the same journey from two ends. A POCT result has value only when it is in the patient's record, against the right patient, in time to be used. Connectivity is the mechanism, and the three connectivity KPIs describe how well it works: how much of the volume goes by interface at all, how much of that arrives with a patient the system cannot identify, and how much never arrives without someone intervening.

The two safety KPIs describe the outcome. Wrong-patient events are rare and under-reported, so the number is less useful than the trend and the time to detection; a service that detects a misidentified result in an hour, because the middleware flagged a mismatch, is in a different position from one that finds out at a complaint. Time to the record is the KPI clinicians care about, and it should be reported as a distribution, not an average, because the median result arriving in two minutes is no comfort to the patient whose result arrived the next day. Patient matching rules, critical result handling and unmatched result queues are described on our results governance page.

One warning: a rising electronic transmission rate can hide a rising unmatched queue. Report them side by side.

Quality and people: is the testing under control

QC pass rate is the indicator everyone starts with, and the one most often misread. A pass rate that never dips is not necessarily good news; it can mean the acceptance limits are too wide, or that failed runs are repeated until they pass and only the pass is recorded. Report it alongside Westgard rule violations per 1,000 QC runs, split by rule, because the rules tell you why QC is failing: a rising rate on the drift rules points to reagent or calibration shift, a rising rate on the random-error rule points to technique or sample handling.

Lockout events are a control working. A device that locked itself when QC failed, or refused an operator whose competency had lapsed, did its job, and the count should fall over time as the causes are addressed. Report the reasons, not just the count.

EQA acceptable returns need two numbers: performance and timeliness. In multi-site POCT, samples not run or results submitted after the deadline are as common as poor results, and each late submission is a missing data point about that site.

The two people KPIs are related but not the same. The first, operators with in-date competency, is a compliance measure per person. The second, tests run without in-date competency, is the patient-facing consequence, and the one an assessor will sample. With operator lockout in place the second should be structurally zero, which is the strongest argument for lockout.

Devices and cost: is the estate maintained and affordable

Device KPIs are the ones most often missing from a committee pack, because device data lives in service records rather than in the results stream. Uptime is worth measuring for the devices that matter clinically, such as blood gas analysers in emergency and critical care, and the reason for each period of downtime (fault, awaiting service, no consumables, no competent operator) is more useful than the percentage. Calibration and maintenance on time is a straightforward compliance measure that ISO 15189:2022 clause 6.4 expects records for; the KPI simply asks how often the schedule is met. Expired lots in use should be zero, and any non-zero month needs the affected results identified. Our device operations page describes how service, downtime and temperature records are kept per device.

Cost per test is the indicator finance will ask for, and the one whose definition matters most. Include QC and EQA material, staff time for QC and maintenance, and waste, or the number will flatter POCT against the laboratory. Consumable wastage explains the gap between purchased and reported: some of it is legitimate QC and verification use, some is expiry, some is failed tests and repeats. Splitting it by cause shows where money is going, and expiry in particular is a governance problem before it is a financial one. Our ROI calculator sets out the cost categories in more detail.

Cadence: who sees what and when

Not every KPI belongs in every meeting. A workable cadence:

• Daily or weekly, POCT coordinator and link nurses: unmatched results, transmission failures, lockout events, devices with expired lots, operators due to expire within 30 days. These are work queues, not statistics.
• Monthly, per site: QC pass rate and rule violations, tests without competency, calibration and maintenance due, EQA submissions outstanding. Sent to the site lead with the comparison to other sites.
• Quarterly, POCT committee: the full set, trended over at least four quarters, by site, with the open corrective actions from unacceptable EQA returns and incidents.
• Annually, management review: ISO 15189:2022 clause 8.9 expects the quality indicators, EQA performance, incidents and audit findings to be reviewed with resourcing decisions. Cost per test and wastage belong here.

Keep the definitions in a controlled document with the KPI's owner and source, and record any change of definition, because a KPI that changes definition breaks its own trend.

Where software helps

Almost every KPI above depends on data that exists only if it was captured at the moment of testing: which operator ran the test, which lot was loaded, whether the result reached the record, whether QC was run and what the rule said. A service running on paper QC logs and monthly spreadsheets can produce activity counts and little else. The reason most services report activity but not quality is not lack of interest; it is that the quality data was never captured in a form that can be counted.

Connected devices and middleware change that. When every result, QC run, lockout and lot change passes through one system with the operator identity attached, the indicators above are queries rather than projects, and the committee pack is generated for each site and for the whole network. Catenix dashboards and analytics studio produce these views from the captured data, per site and across a multi-site service, and let a coordinator build an indicator once and reuse it every quarter. Whatever software you use, ask it two questions: can it show, for one result, who ran it, on which device, with which lot, and when it reached the record; and can it count those things across the whole service without anyone typing them in.

Questions people ask

What KPIs should a POCT service track?

A balanced set covers six areas: connectivity (share of results transmitted electronically, unmatched results, transmission failures), quality (QC pass rate, Westgard violations, lockouts, EQA returns), people (operators with in-date competency, tests run without it), devices (uptime, maintenance on time, expired lots in use), safety (wrong-patient events, time to the record) and cost (cost per test, wastage). Twelve to sixteen indicators, defined in writing and reported by site, is enough.

What is a good QC pass rate for point-of-care testing?

There is no universal figure, and a borrowed target is not a standard. The pass rate depends on the device, the acceptance rules and how repeats are recorded. Establish your own baseline per device type over several quarters, report it alongside Westgard rule violations per 1,000 QC runs, and investigate both a falling rate and one that never dips, which can indicate limits that are too wide or failed runs going unrecorded.

How do you measure POCT connectivity performance?

Three indicators cover it: the share of patient results that reach the clinical record by interface rather than by manual entry, the share of transmitted results that could not be matched to a patient and needed manual resolution, and the number of transmissions that failed or needed intervention per 1,000 results. Report them by site and device, and always show the unmatched queue next to the transmission rate.

How often should POCT KPIs be reviewed?

Work-queue indicators such as unmatched results, lockouts and expiring competencies are reviewed daily or weekly by the POCT coordinator. Site-level quality indicators go to site leads monthly. The full set goes to the POCT committee quarterly, trended over at least four quarters, and the management review that ISO 15189:2022 clause 8.9 requires at planned intervals (annually in most UK services) covers the indicators together with EQA, incidents, audits and resourcing.

Should POCT KPIs have numeric targets?

Not at first. Start with a written definition, your own baseline and a direction of travel, and set targets once you have around four quarters of your own data per site and device type. Targets borrowed from another organisation with different devices, patients and volumes tend to be argued over rather than acted on. Some indicators, such as expired lots in use and tests run without competency, have an obvious target of zero.

What is the difference between activity metrics and quality indicators in POCT?

Activity metrics count what the service did: tests per site, per device, per month. Quality indicators measure whether it was done well and safely: whether results reached the right record, whether QC and EQA were acceptable, whether operators were competent, whether devices were maintained and lots in date. ISO 15189:2022 clause 8.8.2 expects a laboratory to define and review quality indicators; activity alone does not satisfy it.

Sources and further reading

• ISO 15189:2022 Medical laboratories: requirements for quality and competence (clauses 6.4, 8.8.2 and 8.9, Annex A)
• MHRA, Management and use of IVD point of care test devices
• CLSI QMS12, Developing and Using Quality Indicators for Laboratory Improvement
• IFCC Working Group on Laboratory Errors and Patient Safety, model of quality indicators for clinical laboratories