Free1: 9000 RADS

An Independent Review of the Intoxilyzer 9000

Part 1 - Residual mouth alcohol detection

From - Counterpoint Volume 2; Issue 2 - Article 3 (August 2017)

Jan Semenoff, BA, EMA
Forensic Criminalist

The opportunity to conduct an independent analysis and performance review of a new breath alcohol testing device is rare, particularly the higher-end, evidentiary-level units. Access to these technologies is stringently controlled by both their manufacturers and the police and government agencies that control them. Additionally, state agencies are often reluctant to publish the results of their official assessments and analysis of the devices [1].

When given the opportunity to perform such a review on a new Intoxilyzer 9000, I designed a series of experiments to quickly analyze the overall performance of the device. I attended the device’s location with colleague Tom Workman (1948-2019) to determine its suitability and reliability in a number of key areas, including:

Overall design and ease of use
Accuracy in determining in vitro [2] BrAC levels using a simulator
The ability of the device to determine the presence of Fresh Mouth Alcohol using a Residual Alcohol Detection System (RADS) or the so-called “slope detector”.
Reliability in reporting BrAC [3] readings that are highly specific to ethanol
The effect of Radio Frequency Interference on the device [4]

This article will provide a general overview of the operational characteristics of the Intoxilyzer 9000. We will additionally look at the apparent accuracy of the device using simulator readings, and examine the ability of the device to “flag” false positive reading caused by fresh mouth alcohol contamination. Parts two and three will examine the unit’s specificity towards ethanol detection and its ability to identify the presence of an interferent chemical, and the capacity of the device to detect Radio Frequency Interference.

[1] Regarding the Intoxilyzer 9000, only the Georgia Bureau of Investigation - Division of Forensic Sciences has publically released their sanctioned assessment on the device. You may wish to run the search parameters “Georgia replace Intoxilyzer 5000” or “Georgia Intoxilyzer 9000” into your favorite search engine… Other agencies have performed reviews and assessments on the Intoxilyzer 9000, some even going so far as to destroy their own raw data rather than allowing the information into the public domain.
[2] A Latin term used in medical and scientific literature, in vitro means “in the glass” and refers to biochemical testing done outside the normal biological setting, as in a test tube, or otherwise artificially in a lab setting. Using a Simulator to artificially recreate a biological Breath Alcohol Concentration (BrAC) is considered an in vitro test.
[3] I will use the term BrAC to denote a Breath Alcohol Concentration, and BAC to refer to a Blood Alcohol Concentration.
[4] These last two points will be addressed in future Counterpoint articles.

An important caveat:
This assessment on an individual Intoxilyzer 9000 was done in circumstances of complete access to the device, but under limited time constraints. Simply put, we did not have the time necessary to run exhaustive testing on the 9000 to generate the raw data necessary to ensure a proper statistical analysis. Over a ten-hour period, we were able to run about 60 individual tests on the 9000, and inspect its interior and component parts. We need further access to these devices to draw meaningful conclusions.

As an editorial position, Counterpoint, calls on the manufacturers of ALL breath test instruments, and all government agencies that operate and control them, to make these devices, used in criminal proceedings as evidentiary collection devices, available for independent review and analysis. Transparency regarding both the physical and software design, their manufacture, and the maintenance and operation of the devices is critical in maintaining a degree of openness and trust towards the numerical BrAC results generated.

The Intoxilyzer 9000 – An overview

We have examined the Intoxilyzer 9000 in two prior Counterpoint articles:
Article: Author :                       Reference:
The New Intoxilyzer 9000 Mark Thiessen             Archived
The Intoxilyzer 9000 & the Unknown Jan Semenoff                 Archived

The 9000 uses a Windows Mobile platform touchscreen interface for operator input, and control of the breath test sequence. Additionally, the unit we had access to offered a generic computer USB keyboard. Our unit also had an external laser printer that provided a printout of breath test results. Both the external keyboard and printer are optional – the unit can be operated without them using the onscreen touch keyboard, and an optional internal printer.

The 9000 can also incorporate an optional barcode reader or magnetic swipe scanner to allow automatic input of both operator certification and test-subject driver’s license information. The 9000 also can run portably on external battery power using a DC adapter (12V DC @ 8 Amps).

When first turned on, the unit takes about ten minutes to warm up, with the status of the warm-up procedure displayed on the color touchscreen.

Upon initiating the breath testing process, the operator is prompted to enter information: Name of test subject and operator, driver’s license number, occurrence number, etc. The unit then performs an internal diagnostic check, and begins the largely automated breath test sequence.

Figure 1 - The warm-up screen on the Intoxilyzer Model 9000, showing the progress of the internal circuitry warming.

Internal diagnostic check

Figure 2 - After successfully completing its internal diagnostics, the 9000 the PASS or FAIL status of the device.

I have no information on what specific internal measurements are taken and standards compared to in order to determine the pass or fail characteristics of this diagnostic, nor do I know what parameters are necessary for the diagnostic to either “pass” or “fail”. As such, I cannot comment on the overall reliability of the Internal Diagnostic Check algorithm. It was well established that the internal diagnostic checks on both the 5000, and to a lesser extent, the 8000 could be compromised by disconnecting various internal components and circuitry, and still generate an incorrect “DIAGNOSTIC PASS” message.

Air blanks

Similarly, the Air Blanks performed at the beginning and throughout the breath test sequence may well be “floating Zero” air blanks (See Counterpoint Volume 2, Issue 1, Article 6, “Best Practice in Breath Alcohol Testing, Part 1 – Environmental Conditions” for a discussion on “floating zero” air blanks).

Regardless, the unit seemed to perform as one would otherwise expect. We encountered no Ambient Fail errors (none that we did not try and generate, in any case), and we didn’t experience any diagnostic failures from this brand-new Model 9000.

Figure 3 - The air blanks performed by the 9000 may be floating air blanks.

Figure 4 - The effects of a floating air blank, effectively "masking" any potential contaminate read by the optical array.

The touchscreen

Figure 5 - The touchscreen on this particular 9000 was problematic, often non-responsive, or incorrectly activating nearby "buttons".

One thing that we discovered with the calibration and performance of the touchscreen interface was the difficulty in making contact with the desired points on the screen. One had to push very forcefully to get the touch screen to respond, or respond correctly. Its calibration was off, often enabling the function of a button beside the one pressed instead of the intended button pushed. It was very frustrating, and I think if this problem is systemic, and not the aberration of an individual instrument, will result in a lot of operator angst, particularly as we often had to re-enter an entire sequence of pre-test questions before the unit would accept a breath sample. The absolute inadequacy of the touchscreen severely impeded our ability to assess the 9000, as it slowed down the process to probably half the speed it could have been. If this is a systemic issue, I foresee units with damaged touch screens being returned for repair by frustrated operators (particularly after entering the same information over and over again, at three in the morning).

Breath sampling

Providing the actual breath sample was comparable to most other similar evidentiary instruments. The exhalation force to provide the breath sample, and the length of sample duration were about the same. Remember that breath alcohol testing devices typically use four parameters to determine the suitability of a breath sample:

The pressure by which the test subject exhales
The length of time of the exhalation
The volume of the exhaled sample, and
The “slope” of the readings obtained from one second to the next [5].

[5] I prefer the term Residual Alcohol Detection System (RADS) to the commonly used term “slope detector”, as I feel the latter implies an actual physical component within the breath testing device itself, rather than the mathematical algorithm that the systems actually employ. As such, I will use the term RADS to refer to what most people call the “slope detector”.

In vitro accuracy

Using a known alcohol concentration and a simulator, I was able to get a series of readings on the Intoxilyzer 9000 that closely corresponded to the anticipated results of the simulator solution. This was as expected.

Additionally, it does not seem possible to cover the exhaust port of the device in order to prevent exhaust escape. The Intoxilyzer models 5000, 5000EN, 8000 and the 9000 utilize a “flow through” design – the exhaled breath sample is not “captured” as it is on some other devices. It has been demonstrated in the past with the 5000 and 8000 models that any blockage of the breath sample’s exhaust port, intentional or otherwise, has the net effect of artificially raising the reported BrAC reading, with the over-reporting dependent upon the degree of blockage of the exhaled sample, and its level of contamination.

Figure 6 - The exhaust port of the 9000 is shielded from accidental or intentional blocking of the breath sample exhaust.

The design of the exhaust port prevents either accidental or intentional blockage of the port itself. The port is shielded by a plate with an exhaust hole itself, and blocking the port seems difficult, and unlikely to occur accidentally. Time will tell…

The Residual Alcohol Detection System

It is unknown and unreported in the available literature what specific algorithm is used by the Intoxilyzer 9000 to determine the suitability of a breath sample in terms of the “slope” detector, or RADS. It may be helpful to look at these systems as a means of review:

Fresh Mouth Alcohol, Residual Alcohol Detection, and Wait Periods

In general, RADS play a role in determining the presence of fresh-mouth alcohol. A subject who may have recently introduced alcohol into their mouth and respiratory tract by:

Vomiting
Burping
A condition such as Acid Reflux Disease, or GERD

will have an initial rapid rise in BAC that also falls off sharply as the false-high alcohol reading dissipates and is replaced by a “true” near-level slope.

Let’s pretend that the subject above has “micro-burped” (the polite hidden burp, as opposed to the loud burp that can't help but be noticed) immediately prior to, or while providing, a breath sample (as shown in red):

Figure 7 - Sample slope caused by the introduction of fresh mouth alcohol (Red) in comparison to the “true” BAC of the subject (Blue).

Ideally, the unit is able to determine the rise and subsequent fall in BrAC readings from one second to the next during the actual exhalation of the test subject. Any sudden or sharp drop in BrAC reading should be used to determine that the sample itself is contaminated. It is believed that the “true” BrAC level should not spike in this manner with the sudden drop in reading.

The Intoxilyzer 9000 has specific requirements in determining the suitability of a breath sample:

First, it requires a minimum flow rate of 0.15 litres per second, with a minimum breath time of five seconds.
The sample provided must be a minimum of 1.1 litres in volume.
The sample exhalation length must be a minimum of five-seconds, uninterrupted, in duration.
The IR source on the Intoxilyzer 9000 pulses at only 10 cycles per second (Hz). With four filters, a breath sample reading is obtained every 1/10 of a second (100 milliseconds) on each of the four-filtered points, for a total of 40 discrete pulses per second. As the pulses are analyzed, consecutive BrAC readings that do not differ by a pre-determined margin will indicate a level slope.

Once the four criteria (flow rate, volume, exhalation time, and slope) are met, a ZERO appears in front of the preliminary breath test results, indicating the sample obtained is suitable for analysis. The 9000 also displays a histogram of the breath test results that shows:

The subject's breath flow curve (profile of exhaled breath) and
The subject's flow rate in litres/second
The subject's BrAC curve (the peak BrAC, and a profile of the BrAC from second to second during exhalation)
The subject's exhalation duration in seconds

Figure 8 - The Histogram printed by the 9000 via the external printer, showing the exhalation and BrAC profiles of the sample capture.

The forensically acceptable standard of obtaining two readings within 0.02 grams/100ml of each another, coupled with a correctly conducted observation period before and between the two readings merely assists in obtaining suitable samples. The RADS adds only a certain degree of validity to the testing process. Many jurisdictions around the world do not obtain two readings, so the RADS become even more valuable to them. Unfortunately, they do not seem to warrant that degree of trust.

A problematic means of assessing the Residual Alcohol Detection Systems

I have often heard forensic criminalists and state crime lab chemists or technicians describe in court how they assess the reliability of RADS during routine annual maintenance. Typically, an alcohol-free subject (as in the technicians themselves) swish and spit out an alcohol-laden solution (often, simple mouthwash containing ethanol), then immediately provide a breath sample. There is a sudden spike in BrAC reading on the breath testing device that rapidly declines to a “true” BrAC reading of zero, setting off the RADS error message. See Figure 9:

Figure 9 - Laboratory testing of the RADS, with a Zero-BAC test subject (Blue) and an artificially inflated mouth alcohol contamination (Red).

This is, in my opinion, am improper way of assessing the reliability of the RADS. In the real world, under actual operating conditions, the test subject has probably consumed alcohol, perhaps even a considerable amount. Compare figures 1 and 2. The exhalation profiles are markedly different. The difference between the “true” baseline BrAC reading, and the falsely-elevated reported reading due to the alcohol contamination is often not enough for the “rise and fall” algorithm to identify the presence of contamination. Testing as per Figure 2, in my opinion, artificially creates the situation upon which the breath testing device easily passes this criterion, and in no way reflects the conditions experienced by the units in the field, as in Figure 1.

When I tested the Intoxilyzer 9000 on a test subject that had an actual BAC concentration (less than one standard drink), and provided some sort of very minor levels of oral contaminate (a drop of ethanol on the tip of the tongue that was swished and swirled to mix in the mouth), the 9000 often reported falsely-inflated readings, as shown in this chart:

Figure 10 - Data obtained showing the results of intentionally contaminated breath samples, and the response of the Residual Alcohol Detection System on the Intoxilyzer 9000.

I’m concerned about the last two results. The RADS seems to properly report Invalid Sample readings only at very high levels of contamination. I have often heard police breath test operators describe, in court, that they “saw a reading on the screen” before an error message was generated, and attempt to have that preliminary result entered as evidence of an actual measured BAC reading. We see two readings here that are well beyond the per se level, that were properly identified by the 9000 as Invalid Samples, yet a numerical result was also displayed. It must be stressed to qualified technicians during training that the preliminary results can never be relied upon, especially so when an error message is indicated.

A comparison to other breath test devices:

If this procedure to assess the RADS on the 9000 seems inappropriate, consider that I applied the same methodology to the following breath test devices:

Intoxilyzer 8000
DataMaster DMT

In ALL instances, the DataMaster DMT correctly identified the presence of fresh mouth alcohol contaminate and aborted the breath testing process without generating a numerical result.

To varying degrees, the Intoxilyzer 8000 provided, on more than one occasion, numerical results that were falsely elevated:

Figure 11 - A comparison of the Residual Alcohol Detection System's response on the Intoxilyzer 8000, and the DataMaster DMT.

This is also my general experience with the older Intoxilyzer 5000 and 5000EN models. I have generated falsely-elevated readings during training program demonstrations consistent with the 8000 readings above.

Obtaining proper samples & operational implications

Relying upon the pressure / time / volume / RADS to automatically determine the suitability of the sample is insufficient. It must still be the responsibility of the qualified technician to ensure that a suitable sample is properly obtained. Some subjects will be able to provide a breath sample that far exceeds the minimum 5-second requirement of the pressure-time circuit. The Model 5000, 8000 or 9000 sets minimum standards for a suitable sample, based on an average subject. The qualified operator is the one who must ensure that a given subject has provided their own unique suitable sample.

The necessity of a proper deprivation and wait / observation period

There is no override on the Model 5000, 8000 or 9000 as there are on some roadside screeners that are capable of manually drawing a breath sample into the test chamber. The Model 5000, 8000 and 9000 will continue to receive the sample as long as its parameters don’t fall outside the residual alcohol detection system’s threshold values. As long as the subject continues to provide exhaled breath sufficient to keep the pressure transducer open, the sample will be analyzed either 4, 10 or 30 times per second, per each instrument’s design. The RADS, coupled with an observation period of a reasonable length of time, may provide a degree of reliability in the breath testing results. But remember, an observation period is exactly that – observation. The operator should be paying attention with their eyes, ears, and in some cases their noses to detect the smell of the fresh burp, or unnoticed “micro-burp”.

Overall impression on the 9000 Residual Alcohol Detection System

As stated before, this assessment on the Intoxilyzer 9000 was done under circumstances of complete access, but was time limited. Simply put, we did not have the time necessary to run exhaustive testing on the device to generate the raw data necessary to make a proper statistical analysis. We need further inquiry to draw meaningful conclusions.

It has been my experience that, in general, the RADS can, and often are, fooled under a variety of circumstances, most notably, recent introduction of a small quantity of alcohol, similar to what would occur during a burp or “micro-burp”. This circumstance is precisely what the RADS was designed to detect, yet fails to do so. I have routinely observed the RADS fail to register mouth alcohol that is a few minutes old, often allowing the unit to register an abnormally high reading given a simple swish of alcohol, or even a small drop of alcohol on the tip of the tongue that is allowed to dissipate for a few minutes. I can only conclude that the RADS is merely an investigative aid, and is a highly inaccurate detector of mouth alcohol, with most breath test devices, with the notable exception of the DataMaster DMT.

Figure 12 - A graphical representation of a "true" breath alcohol concentration (blue) and a falsely contaminated reading (red).

What is concerning is the apparent inability of the Intoxilyzer 9000 to accurately determine this contamination. My testing shows, albeit with limited data, that the 9000 routinely gave false positive readings with minimum mouth alcohol contamination. In every instance but two where we contaminated a “true” reading with a minute quantity of ethanol, a falsely-inflated reading was obtained and reported as a true value. Our limited data shows that the true BrAC reading could be added to by as much as 0.014 to 0.068 without identifying the mouth alcohol contamination.

Final thoughts:

The only two instances that reported an INVALID sample error message was when extremely high levels of contaminate were introduced (adding .127 to .237 grams to the true amount). Even then, a numerical result was displayed as a preliminary reading that was subsequently reported as INVALID.

Simply put RADS or slope detectors, in general, are suspect at best, and in the Intoxilyzer 9000, do not provide a reliable means of identifying fresh mouth alcohol contamination. This points to the necessity of a properly observed and conducted deprivation and observation period prior to and between receiving evidentiary breath alcohol samples for forensic or court purposes, and for the continued use of replicate or duplicate breath alcohol testing that must fall within acceptable parameters of congruency.

Send me your questions or comments:

Comments and questions will be posted here with their responses:

For further study:

Gullberg, R. G., Common Legal Challenges and Responses in Forensic Breath Alcohol Determination, 2004, Indiana University Center for Studies of Law in Action, “Borkenstein Course”.
Gullberg, R. G., Breath Alcohol Measurement Variability Associated with Different Instrumentation and Protocols, Forensic Science International 131 (2003) 30-35
Gullberg, R.G., The Inadequacy of Instrumental “Mouth Alcohol” Detection Systems in Forensic Breath Alcohol Measurement, Northwest Association of Forensic Sciences, Oct., 2000.
Hlastala, M., Lam, W., and Nesci, J., The Slope Detector Does Not Always Detect the Presence of Mouth Alcohol, For the Defense, March 2006.
Jones, A. W., Concerning Accuracy and Precision of Breath-Alcohol Measurements, Clinical Chemistry, 33/10, 1701-1706 (1987)
Labianca, D. A., The Flawed Nature of the Calibration Factor in Breath-Alcohol Analysis, Journal of Chemical Education, Vol. 79, No. 10, October 2002
Simpson, G., Accuracy and Precision of Breath Alcohol Measurements for Subjects in the Absorptive State, Clinical Chemistry, 33/6, 753-756 (1987)
Sterling, Kari, The Rate of Dissipation of Mouth Alcohol in Alcohol Positive Subjects, The Journal of Forensic Science, 2011.