The Level 2 method: event logs, QC history, the subsystem map and test runs

Combine the analyser's event log, its QC and calibration history, the maintenance record and the operator's story into one hypothesis, place the fault in a subsystem, choose the cheapest test run that separates two suspects, and agree with the laboratory who decides what.

Where Level 1 ends

In Level 1 you learned the analyser's building blocks, how to work safely around blood and reagents, how the liquid path and the measuring principles work, the difference between calibration and QC, and the order in which to check the cheap causes. Level 2 starts when you have done all that and the fault is still there: the lamp is new, the probe is clean, the reagents are fresh and in date, the water unit is fine, and yet the analyser still stops, flags or gives results the laboratory does not trust.

At this level the usual mistake is not ignorance but impatience. An analyser is a chain of small subsystems that report on each other. A level sensor reports a 'short sample' that was caused by a leak upstream. A photometer reports a 'baseline error' that was caused by a bubble in the cuvette, a loose lamp connector or a power supply ripple. A motor reports a 'timeout' that was caused by a full database slowing the controller. Replace the part that reported the error and you have spent the laboratory's money and kept its problem. Your task now is to say which subsystem has really failed, prove it with a test, repair it within your authority, and give the laboratory the evidence it needs to decide when to restart patient testing.

Four sources of evidence, read together

Before you touch anything, collect four kinds of evidence, because each one alone misleads.

The event log tells you what the analyser noticed and when. Export or photograph it before any restart: some systems keep only a limited number of entries and overwrite the oldest. Look for the first event in a burst, not the last one on the screen, and check that the system clock is right; a wrong date scrambles the order and confuses anything that depends on elapsed time, such as reagent on-board stability.

The QC and calibration history tells you what the measurement did. Ask the laboratory to show the Levey-Jennings charts for the affected tests for the last weeks, the calibration dates and results, and any reagent or calibrator lot changes. A fault that appeared on the same day as a new reagent lot is a different problem from one that grew slowly over a month.

The maintenance record tells you what was changed and when: lamps, probes, syringes, tubing, electrodes, filters, software. Many faults start within days of a maintenance visit, including your own.

The operator's account tells you what nobody wrote down. Ask the people who run the analyser what they saw and heard, which tests or channels are affected, whether it happens on the first run of the day, after idle periods, on busy days or at a particular time, and what changed: a move, a power cut, a new water unit, a new cleaning solution, a new sample tube type.

• Log: first event of the burst, time, subsystem, frequency over weeks.
• QC: one test or many, random or systematic, sudden or gradual, linked to a lot change or not.
• Maintenance: what was touched recently, and by whom.
• Operators: when it happens, what it looks like, what changed.

The subsystem map and the test cycle

Every analyser runs a fixed cycle, and each stage needs a known group of subsystems. Placing the failure in the cycle narrows the suspects before you open a cover. Identification needs the barcode reader, the sample handler and the software. Aspiration needs level sensing, the probe, the syringe or pump, valves and the system liquid. Dispensing and mixing need the arm, the mixer and the cuvette or chamber. Incubation needs temperature control and time. Separation and washing, in immunoassay and in many cuvette-based chemistry systems, need wash pumps, needles, magnets and drains. Measurement needs the optics or the sensor, the electronics and the reference readings. Calculation needs the calibration data, the lot parameters and the software. Reporting needs the printer, the host computer and the LIS link.

Ask at which stage the fault appears. If identification fails, look at labels, the reader and the handler before the fluidics. If the analyser reports short samples, clots or bubbles, you are in aspiration. If every reading on one wavelength or one channel is wrong, you are in measurement. If the measurements are good but the reported numbers are wrong, look at calibration data, lot parameters and the units, not at hardware.

Also ask what the fault has in common. One test only points at its reagent, its calibration or its particular wavelength. All tests on one reagent probe point at that probe and its syringe. All tests at once point at something shared: temperature, water, power, the sample probe, a common reference. One analyser out of two identical ones in the same room points at the analyser; both at once point at the room, the water or the reagents.

Reproduce it with the cheapest test

A fault you cannot reproduce cannot be proven fixed. Look for the smallest experiment that makes it happen and gives a clear yes or no.

The analyser itself offers many tests. Service and maintenance menus usually include mechanical self-checks, a prime and a wash, a blank or background check, photometer readings, temperature readouts, valve and pump tests, sensor tests and a log of counters. Some analysers also trend their own health. The Sysmex XP series service manual, for example, describes a failure-prediction function, part of its remote support system, that tracks pressure, vacuum, drain time, the HGB blank and clogs, so that slow degradation shows before it stops the analyser. Use only the tools your access level and training allow, and follow the manufacturer's procedure.

Bench experiments are just as powerful and often cheaper. Run the same sample ten times in a row: a wide scatter means fluidics, optics or electronics; a steady shift means calibration, temperature or reagent. Run a sample of purified water or the system's blank solution: a result that should be near zero and is not points at contamination, carry-over or the measuring channel. Swap the order of samples: a high result that follows a very high sample is carry-over. Run the same control on two probes, two cuvette positions or two channels, if the analyser allows: if the error stays with one, you have found where it lives. Watch the analyser work with the covers safely open or through the windows, as far as the safety rules allow: many fluidic faults are visible, such as a drop on the outside of a probe, a bubble in a tube or a wash station that overflows.

Change one thing at a time

Swapping a suspect part with an identical good one is a strong test on analysers, because many contain repeated parts: identical valves, pumps, syringes, electrodes, chamber connectors. Do it with discipline. Change one thing, write it down, test, and put the part back if the test proves nothing. Swap only parts with the same part number and a revision the manufacturer allows together. Think before you move a power supply or a driver board into a circuit that may have a short: what killed the first part can kill the second. Isolating loads is safer. The HumaStar 300 service manual, for example, describes disconnecting the loads from the power supply board when fuses blow and reconnecting them one by one to find the defective branch.

Remember that many swaps and replacements require a check or calibration before the analyser is usable again: a probe position after an arm part, a photometer reference after a lamp, an electrode calibration after a sensor. A swap that leaves the analyser unchecked is not finished.

Who owns QC, results and the decision

An analyser lives inside a laboratory's quality system. The laboratory runs calibration and QC, investigates its own QC failures, decides whether results already reported must be reviewed, and decides when patient testing restarts. In many countries this follows ISO 15189 or national rules on internal quality control, such as the GOST R 53133 series in Russia or, in Kazakhstan, the standard for organising laboratory diagnostics approved by order ҚР ДСМ-257/2020; check the rules that apply where you work. The engineer's job is different: to show that the analyser works as the manufacturer specifies, to repair it, and to record what was found, what was done and what was verified.

Keep the two roles separate and say so in writing. You may say 'the photometer repeatability test passed after the lamp was replaced and the reference readings were updated'. You should not say 'the analyser is fine for patients'; that conclusion belongs to the laboratory after its calibration and QC. Equally, do not let the laboratory hand you its QC problem without its data. A request to 'fix the QC' comes with the charts, the lots and the calibration history, or you are guessing.

Within your authority the usual safe territory is: reading logs and running the tests your access level offers; cleaning, re-seating and adjusting by the manufacturer's procedure; replacing parts with the required checks afterwards. Outside it, and outside this course: service passwords or software you were not given, editing calibration factors or lot data to make QC pass, changing method parameters that belong to the laboratory, and bypassing any cover interlock, level or clot detection 'just for a test'.

Triage before you travel, and keep a fault diary

Many analyser faults can be narrowed by telephone or messenger. Ask for the exact message and error code, a photograph of the screen and of the log page, the affected tests, a picture of the relevant QC chart, and when it started. Ask about the room temperature, the water unit, recent maintenance and reagent lot changes. With that you can decide which parts and tools to bring and whether the laboratory should move its urgent work to a back-up analyser in the meantime, which is their decision based on your opinion.

On site, write as you go: the as-found state, the log excerpt, each test with its result, each change, counters and the as-left state. A good diary turns an intermittent fault into a pattern, lets a colleague carry on without repeating your work, and becomes part of the analyser's service history.

On-site checklist

• Export or photograph the event log before any restart.
• Check the system clock and find the first event of the burst.
• Get the QC charts, calibration history and lot changes for the affected tests.
• Read the maintenance record for recent changes.
• Map the failure to a stage of the test cycle and find what affected results share.
• Reproduce with the cheapest test: repeats, blanks, order, position swaps, observation.
• Change one thing at a time; same part number and allowed revision only.
• Record as-found, tests, changes, counters and as-left; hand over to the laboratory for calibration and QC.

Check yourself