Quality Assurance

External survey providers post unknown samples to the laboratory several times a year, and the graded return is one of the few objective checks that counts leaving the bench agree with peer laboratories using the same method. The technologist treats each survey vial exactly as a patient specimen: routine analyser, routine operator, routine review, no extra repeats that patient work would not get, and no phone call to a neighbouring site before submission. Unacceptable grades trigger a documented investigation covering method, reagents, operator and paperwork, followed by corrective action and a check that the fix held.

Point-of-care testing moves haematology away from the laboratory — handheld haemoglobin meters on dialysis and surgical units, INR devices in anticoagulation clinics, bedside platelet testing in theatre — but the laboratory usually keeps oversight of them. That means signing off operator training and recertification, confirming each device runs its liquid or electronic controls on schedule, tracking lot-to-lot performance, and comparing split specimens against the main analyser at defined intervals. A consistent gap between meter and core method is a calibration problem rather than a rounding quirk, and the device stays out of service until it is resolved and documented.

Control material is only useful if somebody reads it properly. Each shift the technologist compares control results against the assigned mean and standard deviation, applies the laboratory's rule set — 1-3s, 2-2s, R-4s, 4-1s, 10-x — and decides whether the run can be released. Calibration data is judged the same way: after a lamp change, major service or a new reagent lot, calibration is verified and every control level is repeated before patient testing resumes. Monthly summaries of mean, standard deviation and coefficient of variation show whether precision is holding over time or quietly drifting.

Two kinds of error appear on a control chart and they call for different responses. Random error is scatter — one wild point from an air bubble, a partly clotted control, an unstable pipette — and it widens the standard deviation without moving the mean. Systematic error moves the mean itself: an abrupt shift the day a reagent or calibrator lot changes, or a slow trend as a lamp ages, a control deteriorates or a temperature creeps. Reading the pattern across several runs, instead of reacting to a single flagged point, is what tells the technologist which one is in front of them.

Morphology cannot be reported off a badly stained film. Before calling a differential the technologist checks that the Romanowsky stain has produced the expected colour balance: orange-red erythrocyte cytoplasm, purple nuclear chromatin with visible parachromatin, and clearly tinted neutrophil and eosinophil granules. Films that are too blue point to buffer that is too alkaline; washed-out pink films point to acidic buffer or over-rinsing. Granular deposit lying above the cells means unfiltered or oxidised stain. Each fault has a documented remedy — adjust buffer pH, filter or replace the stain, change the timings — and affected films are restained rather than reported.

Every procedure used at the bench should be the version the laboratory has authorised, and it should be possible to prove that. Controlled documents carry a title, a version number, an effective date, an approval signature and a stated review interval. When a method changes, the revised version is approved and released, staff are trained and sign that they have read it, and the copy it replaces is withdrawn from the working area and archived for the retention period. Uncontrolled photocopies taped inside cupboards and personal notebooks are the usual failure, because they outlive the revision they were taken from.