Start By Defining "Reliability" don't focus on "Accuracy"
Updated: Sep 9, 2022
To remind the Court, "thing 1" in 258(1)(c) is unconstitutional under Charter 11(d) (R v St-Onge) but is saved by section 1.
To remind the Court that 258(1)(c) is saved by section 1 because of Parliament's purpose which is weight consistent with scientific reliability.
To demonstrate to the Court that the SCC in St-Onge relied on the paper by Brian Hodgson. To bring that paper before the Court and make it an exhibit.
To ask the government scientist to concede the definition of "accuracy" in the Hodgson paper.
To ask the government scientist to concede the definition of "precision" in the Hodgson paper.
To ask the government scientist to concede the definition of "reliability" in the Hodgson paper, at least as it refers to instrumental reliability. The paper also contemplates other components of scientific reliability, i.e. good laboratory practice, see abstract.
To highlight the portion of the definition of "reliability" that contemplates "drift" ... "over time".
To demonstrate that such a definition requires empirical evidence of accuracy and precision at one time compared with accuracy and precision at a subsequent time.
To suggest that the ATC/CFS approach to determination of scientific reliability "at time of use" is not empirical - it is not an empirical approach and the approach is not supported by the scientific literature.
To suggest that the approach by the ATC/CFS of relying on a single data point control test, to determine reliability, renders any analysis by the instrument "qualitative" rather than "quantitative".
To connect such misuse with the principles of the Motherisk Inquiry Report.
To lay the groundwork for the argument that Criminal Code forensic use of what Parliament contemplates is a quantitative analysis instrument, in a manner that only results in a instrument qualitative analysis will result in an unlawful search under Charter section 8.
See pages 6-7 of the Motherisk Inquiry Report on confusion of qualitative and quantitative analysis as being unacceptable for a forensic purpose:
ELISA is widely used in both forensic and clinical toxicology as a screening test. It is intended to determine quickly if a sample is negative and merits no further testing or if it is a preliminary positive, in which case the sample must be tested again using another method (a confirmation test) to determine if the sample is in fact positive. e requirement to carry out a confirmation test on any preliminary positive results from immunoassay-based screens, such as ELISA, was highlighted in the literature and in all internationally recognized hair-test- ing standards well before 2005. Indeed, the Immunalysis kits that MDTL used included an explicit warning for the user about the preliminary nature of the ELISA results.
Until August 2010, MDTL reported its ELISA results to customers without a confirmation test. It reported its ELISA results – both qualitatively (as positive vs. negative) and quantitatively (in the form of a numerical drug concentration for positive samples). Such a practice is inconsistent with internationally recognized forensic standards and is unacceptable for two important reasons.
First, unlike confirmation techniques such as gas chromatography–mass spectrometry (GC-MS) or liquid chromatography–tandem mass spectrometry (LC-MS/MS), ELISA cannot identify the substances within a sample. Second, MDTL’s practice of using ELISA to quantify drug concentrations is simply unheard of in forensic toxicology lab- oratories. No forensic toxicology laboratory in the world uses ELISA testing the way MDTL did.
See pages 52-53 of the Motherisk Inquiry Report on number of calibrators required for quantitative v. qualitative analysis:
Sample cross-examination of a CFS scientist:
Q. Right. Now, you indicated in examination in-chief that you agreed with Brian Hodgson’s definitions of accuracy, precision and reliability. Did I get that right? A. I – that was October – I.... Q. Yes, all right. A. Can you just give it to me again? Q. Exhibit.... A. I could probably....
Q. Sure. Exhibit Number 2 at Tab 2 under – specifically under instrumentation. I’m just talking about instrumentation right now for these purposes, but Brian Hodgson in the article refers to accuracy as something in particular. There’s a definition of accuracy. I think you testified on the last day that you agreed with the – his definition of accuracy. A. Accuracy is the ability to measure a substance with a result that is as close to the true quantity of the substance as possible. For breath alcohol testing this means measuring the breath alcohol concentration as closely as possible to the true or actual concentration of alcohol that
is calculated to be in the breath at the time of sampling. Yes.
Q. Right? And precision? A. Is the ability to measure the same – that same substance repeatedly to produce the same result. For breath alcohol testing, this means repeated measurement of the same concentration of alcohol to produce results that are as close to each other as possible. Yes. Q. And.... A. So, that’s....
Q. And reliability. A. So, that’s the spread in the results. The accuracy is how close it is to the real result and the precision is the scatter of results, around that point. Q. Right. And reliability; his definition of reliability. Again, this is to do with instrumentation. A. Yes, that’s under that category. Refers to the ability of the instrument to perform over time without any significant drift in accuracy and precision. Q. All right. A. Yes. Q. He uses the words, ‘over time’. A. Yeah. Q. What does he mean – can you help us explain what he might mean by over time in that respect? A. Over a period of time. Over a number of days, months or years, depending on specifically what it is you’re referring to. Different devices will have different – not different reliabilities, but they’ll have ability to maintain certain parameters over time, however long that might be, like the screening device, the 68-10 by Draeger, has been shown by the manufacturer to be stable for six months. But
yet the Centre of Forensic Sciences says that the device should be accuracy checked every 15 days and the Alcohol Test Committee says it should be accuracy checked, or calibration checked, every 31 days. Q. So, if we are looking for empirical evidence, one way or another, whether it would support the Crown’s case or support the defence case, of changes in reliability, then we need to assess data at least at one point in time and at least at another point in time. A. Well, again, as I said earlier, the way that the instrument is assessed at the time of use is with the alcohol standard solution during a calibration check. Q. But we.... A. That’s what the Alcohol Test Committee and the Centre of Forensic Sciences, and in my opinion, the same thing, that is the procedure that is used to determine the accuracy and reliability of the instrument at the time of testing and whether it’s working properly. Q. It is the way that it is assessed, it is the procedure that is used, according to the Alcohol Test Committee and the C-F-S and your opinion. Did I get that right? A. Yes. Q. Where’s the scientific literature that backs up that approach? A. Well, it was long in place before I was a forensic toxicologist. I mean, that procedure and that language has been used for the 5000C and the 900 – 900A before that. And again, as I
explained, that calibration curve doesn’t shift, it’s just the response, and so what you’re checking again, is to see whether or not the response at the critical level of 100 is accurately reading within that range,
again, subject to the variability that can sometimes cause the result not to be exactly 100. And so, if it’s reading low, it means that the results at that point might be lower than they actually are. Q. I want to suggest to you, that inevitably, what you’re saying is that an Intoxilyzer 8000C that has been checked in this way is conducting a qualitative analysis, not a quantitative analysis. A. That's correct. Q. And that’s because you’re checking it, whether or not it is above or below a different level, a particular level of 100. A. Yes, because there’s no scientific reason why the calibration curve should change at the high end or the low end and not in the middle.
[Quaere: Where is the empirical study backing up the latter opinion?]
#accuracy #precision #reliability #Hodgson #drift #overtime #Motherisk #quantitativeanalysis