Introduction
This policy is designed to highlight the processes Oil Analysis Laboratories use to ensure the reliability and accuracy of data we produce as a laboratory. Despite not carrying a formal ISO certification, our laboratory operates a full quality management system to ensure our daily lab processes generate as accurate data as would be expected of an ISO17025/9001 laboratory. The main focus of this is our ASTM proficiency testing which allows us to compare the data we produce to hundreds of laboratories across the world to achieve the highest quality of data.
Measure twice, report once
You may have heard of the phrase measure twice, cut once, well the same is in the approach we adopt as a laboratory in that if, in doubt, we take a second measurement. All laboratories occasionally produce a piece of data that appears erroneous even with a meticulous level of precision in following the standard operating procedure for the test. The confidence of repeatability adopted by most lasts is only 95%, but we aim to far exceed this target and as much as is possible the customer should never have to see a test has had a problem. The quality processes of the lab are to ensure these are addressed BEFORE the report ever leaves the lab.
In terms of retesting, we do not set a target on a maximum number of retests in the lab as is often seen in other commercial laboratories and there is no minimum hierarchy in the lab to request a retest e.g. you do not have to be a supervisor or diagnostician to request a retest and if something looks wrong we encourage you to recheck because for us failure is not an option.
Duplicate and Triplicate Testing
Certain tests in the lab are run in duplicate or even triplicate (see the lab method for details of what is required) including all wear metal analysis and any tests where the integrity of the sample can degrade over time to have to wait for a recheck e.g. Dissolved Gas Analysis. This addresses a large proportion of confirmatory testing in the lab.
Rechecking on instrument
Commonly retesting can be considered to doubling the work for lab operators and hence why it is discouraged in other laboratories, but a core number of tests can be retested whilst a sample is on the instrument with no additional prep work required from the technician. This aids in giving the benefit of fast results, whilst also maintaining confidence in data. Much of this is done by the LubeWear.com Laboratory Information Management System that reviews the data against historical data, sample information provided by customers and other tests to identify the most likely potential rechecks that will be requested using our Recheck AI system to predict what a diagnostician would recheck with a 98.9% accuracy so the recheck can be done there and then.
Quality control samples
Quality control standards can be split into primary e.g. purchased standards and secondary standards correlated to the primary standards. A secondary standard is to run a minimum of every 20 samples (as this is the minimum for ISO 17025), but typically every 12 or 10 samples, depending on the instrument design and row lengths of racks. Many tests will have a high and low QC too, to give a full range of testing in addition to midpoint calibrations – see specific method statements for details of QCs required. For tests that potentially take several days to perform e.g. RPVOT a standard is run weekly or every 10 samples, whichever occurs sooner.
If a quality control sample fails, the first step is the quality control is repeated automatically. Should this pass, the testing may continue, but repeated instances require a primary standard test and recalibration of a secondary standard.
If a quality control sample fails the confirmation, the testing on the instrument stops, and you will receive a notification on your company mobile and the instrument to alert you that this instrument has an issue. The samples from the last passed standard to the failed standard will be quarantined until the cause of the failure e.g. a blocked tube, insufficient sample etc has been identified and the entire batch since the last passed standard repeated before proceeding.
The Management quality meeting will review every single failed QC since the last meeting to identify potential corrective actions.
Better than the method of repeatability Quality control
Most labs set their quality control standards pass/fail criteria at the repeatability or even reproducibility of the test. This, with tests with quite wide ranges of repeatability/reproducibility, means quite large changes can be accepted in quality control standards and much later identification of problems. Hence, all QC threshold values shall be set at a maximum of 50% of the ASTM/ISO/IP method repeatability, or 50% of the diagnostic significance value, whichever is stricter so that even slight deviations, even within the method can be identified before data begins to be effected.
Proficiency testing.
Proficiency testing involves samples being sent to ours and hundreds of other laboratories around the world to establish the true average of the data, which is considered the correct answer. Many of you coming from other laboratories will be familiar with proficiency testing samples being treated separately as special samples with several weeks to perform the tests multiple times and potentially with specialist calibration sets. This is not the case with our laboratory. We believe the truest reflection of the accuracy of the lab is when the samples are treated like every other sample in the lab with the same turnaround pressures etc. Proficiency testing samples should receive the same high level of care as ordinary routine samples, but no specialist treatment as otherwise you are just assessing how well the lab performs under ideal rather than actual conditions. The target turnaround of all PT samples is the same working day of receipt.
Z-Scores
Z-scores are a measurement of how much bias a test has and is the key measurement in any proficiency testing scheme, with a value of 0 being the perfect score and +/- 2 being a pass for auditing purposes. In all cases the target OAL shall work to is +/- 1 for our Z scores and although +/-2 is a pass, the lab should treat scores exceeding 1 in either direction as a fail and the method should be reviewed to highlight how to bring the data back in line with the target 0 Z score.
Use of historical proficiency tests.
Once a PT sample has been tested, the sample shall be mixed and subsampled into ordinary lab bottles by a lab supervisor so these samples can be put through as blinded routine samples periodically to test the lab accuracy of data between PT samples.
The samples should also be introduced in load testing when the lab has a particularly busy day to confirm even under time pressures, we still deliver the highest level of accuracy.
Quality Record keeping
Wherever possible, all record keeping should be electronic with a target of zero paper, meaning that audit logs of test data, measurement, quality control values are automatically processed by the LIMS without requiring direct interaction from lab staff.
Normalisation.
The lab should use normalisation against quality control data to account for any small deviations from batch-to-batch testing to ensure consistent results.
