EVIDENCE LIBRARY
Hospital Alcohol Tests in DUI Cases: Serum Results, Uncertainty and Threshold Decisions
A 2017 study shows why hospital serum alcohol results near a DUI threshold require laboratory-specific uncertainty analysis—and why 19.74% is not a universal BAC deduction.
A hospital alcohol result close to a DUI decision threshold requires more than comparing two printed numbers. The specimen, analytical method and laboratory-specific measurement uncertainty matter. A 2017 emergency-laboratory study illustrates the problem: the authors estimated 19.74% expanded uncertainty for their enzymatic serum ethanol assay and identified eight driver results whose uncertainty intervals crossed their study’s 0.50 g/L decision threshold. That percentage is a finding about one laboratory’s procedure, not a universal deduction from every blood alcohol result.
This article examines the original study by Ustundağ and Huysal, Measurement uncertainty of blood ethanol concentration in drink-driving cases in an emergency laboratory. It addresses scientific interpretation in DUI cases; the study’s historical Turkish threshold does not establish the applicable law in another jurisdiction.
What did the laboratory actually test?
The investigators retrospectively reviewed 1,034 drivers tested between July and December 2016 at a hospital in Bursa, Turkey. Venous blood was collected in paired plain tubes and centrifuged to separate serum. The serum was analyzed with a Synchron Systems Ethanol assay on an Olympus AU400 analyzer. The method measured an alcohol-dehydrogenase reaction through changing absorbance at 340 nm.
Although the paper uses the expression “blood ethanol concentration,” its methods identify serum as the analyzed fraction. That distinction must survive the transition from a scientific article to an expert opinion. The study did not validate an interchangeable numerical relationship between its serum results and an individual driver’s forensic whole-blood result.
How was the uncertainty estimated?
The authors used a Nordtest approach combining within-laboratory reproducibility and a component associated with laboratory bias. They analyzed 368 internal quality-control results and external quality-assessment data. External targets were peer-group means, rather than an independent determination of the true concentration in each driver’s specimen.
| Reported item | Value | Interpretive limit |
|---|---|---|
| Driver records reviewed | 1,034 | Retrospective hospital population, not a random sample of all motorists |
| Driver results above 0.50 g/L | 181 | Classification under the study’s decision threshold |
| Internal control observations | 368 | Quality-control measurements, not additional drivers |
| Pooled within-laboratory reproducibility | 5.56% | One uncertainty component |
| Combined standard uncertainty | 9.87% | Estimate from the laboratory’s uncertainty model |
| Expanded uncertainty | 19.74%, coverage factor 2 | Approximately 95% coverage under the model’s assumptions |
| Results with intervals spanning the threshold | Eight drivers | Does not identify eight proven misclassifications |
These values describe a particular assay, quality-control history and calculation. They do not establish the uncertainty of a different hospital, a headspace gas-chromatography method, a breath instrument or a specimen tested years later.
What does an interval crossing a threshold mean?
An uncertainty interval spanning a decision threshold means the measurement, interpreted with its uncertainty model, does not support the same simple above-or-below conclusion as the displayed central value alone. It does not prove that the concentration was below the threshold. Nor does it establish how a court must apply that uncertainty.
The paper gives an example of a result of 0.51 g/L with an approximately 0.42–0.60 g/L interval. Its numerical presentation deserves care: the authors also describe 19.74% uncertainty at 0.50 g/L as approximately ±0.09 g/L, whereas direct multiplication gives ±0.0987 g/L, approximately ±0.10 g/L. A case-specific calculation should use the actual laboratory’s documented formula and rounding policy instead of borrowing that illustration.
The units also matter. A concentration of 0.50 g/L equals 0.050 g/100 mL. That unit conversion does not convert serum into whole blood, and it does not convert an alcohol concentration at specimen collection into a concentration at an earlier driving time.
Why the 19.74% estimate cannot be subtracted from every DUI result
Expanded uncertainty describes dispersion associated with a measurement under stated assumptions. It is not a measured error showing that a result is too high by that percentage. A symmetric uncertainty statement allows values on both sides of the reported concentration.
Whether a particular reporting or decision procedure uses a lower bound, an upper bound or another rule is a separate question. The first scientific task is to establish which uncertainty estimate belongs to the actual method and concentration range. Applying the Turkish hospital’s percentage to an unrelated forensic laboratory would replace case evidence with an unsupported assumption.
What the study did not establish
The authors expressly identify a major limitation: their uncertainty calculation concerned the analytical process and did not include preanalytical or postanalytical phases. It therefore does not quantify every possible contribution from collection, handling, storage, identification, reporting or later interpretation.
The investigation also did not measure on-road driving performance, validate an impairment threshold or reconstruct an individual driver’s earlier alcohol concentration. Its eight threshold-spanning results are an uncertainty observation, not a clinical finding that eight people were unimpaired or a demonstration that eight legal outcomes were wrong.
There is an internal numerical inconsistency worth recognizing when using the paper: it states that 181 of 1,034 results exceeded the threshold but elsewhere gives a positivity percentage of 7.8%. Those counts calculate to approximately 17.5%, not 7.8%. The reported counts can be described, but the incompatible percentage should not be repeated as if the figures agree.
Questions to resolve when reviewing a hospital alcohol result
| Question | Why it changes interpretation |
|---|---|
| Was the analyzed specimen serum, plasma or whole blood? | The specimen named in the methods controls the meaning of the result. |
| Which assay, instrument and units were used? | An enzymatic serum result and a forensic whole-blood chromatographic result require their own supporting validation. |
| What uncertainty estimate applies at this concentration? | Another laboratory’s percentage is not a substitute for the testing laboratory’s records. |
| Which sources of uncertainty were included? | An analytical estimate may omit collection, storage or reporting contributions. |
| When was the sample collected relative to driving? | Analytical uncertainty does not supply the assumptions needed for retrograde extrapolation. |
| Was any confirmatory testing performed? | The actual method and confirmation records determine the available analytical evidence. |
Frequently asked questions
Does this study prove hospital alcohol tests are unreliable?
No. It demonstrates why laboratory-specific uncertainty should accompany interpretation near a decision threshold. It does not invalidate every result from the studied assay or other hospital assays.
Does 19.74% uncertainty mean a 19.74% chance that the result is wrong?
No. It is a relative expanded measurement-uncertainty estimate. It is neither a probability of a false result nor a measured percentage of overstatement.
Can the study’s serum values be treated as whole-blood values?
No. The paper’s specimen-processing description identifies separated serum. Any proposed comparison with whole blood requires its own scientifically supported basis.
Did the authors establish that eight drivers were below the limit?
No. They identified eight results with intervals extending to both sides of the study’s threshold. That is different from establishing the actual concentration below it.
Does the uncertainty interval show the alcohol concentration at the time of driving?
No. The interval concerns the laboratory measurement. Earlier-time interpretation requires additional evidence and assumptions about timing and alcohol kinetics.
Does the paper determine the legal outcome of a DUI case?
No. It supplies scientific evidence about one analytical process. Applicable legal standards and the complete case evidence require separate evaluation.
For related analytical context, see our discussion of headspace GC-FID validation and BAC uncertainty.
Primary scientific source
Ustundağ Y, Huysal K. Measurement uncertainty of blood ethanol concentration in drink-driving cases in an emergency laboratory. Biochemia Medica. 2017;27(3):030708. doi:10.11613/BM.2017.030708. This is an original explanatory analysis of the supplied full paper; the tables above reorganize selected findings and interpretive questions rather than reproduce the publication’s tables.