EVIDENCE LIBRARY
Can a Drinking Denial Rule Out Blood Alcohol? A 201-Case Study
A 201-case poisoning study compared recorded drinking history with clinical blood ethanol results. Examine 2×2 counts, missing histories and forensic limits.
Quick answer: A patient’s report of recent drinking can be useful clinical evidence, but it is not a measurement of ethanol concentration. In a retrospective study of emergency-department patients treated for acute poisoning, reported alcohol ingestion agreed relatively well with a clinical blood alcohol classification among the 201 patients who had report data. Yet four patients who denied drinking had a result above the study’s 10 mg/dL threshold, 11 who reported drinking did not exceed that threshold, and 54 of 255 eligible analyzed records had no usable report. Those findings do not support treating either a denial or an admission as a substitute for a properly characterized specimen result.
Can self-reported drinking identify alcohol exposure in a poisoned patient?
Sometimes, within a defined clinical population. Woo and colleagues’ 2013 retrospective study compared a recorded history of alcohol ingestion with a contemporaneous clinical alcohol test among adults presenting with acute poisoning at a South Korean emergency department. In the 201 cases with a recorded answer, reported ingestion had 96.6% sensitivity and 86.7% specificity relative to the researchers’ test threshold. These are group-level classification statistics under the study’s conditions. They are neither a precision estimate for a person’s blood alcohol concentration nor validation for a courtroom reconstruction of drinking behavior.
The distinction matters because three propositions are often collapsed into one: a person reports drinking; ethanol was measured in a particular biological specimen above a threshold; and ethanol caused an observed behavior or adverse event. This study principally investigated the relationship between the first two. It did not experimentally evaluate impairment or demonstrate individual causation.
What was the study design?
The investigators reviewed emergency-department records from January 2004 through February 2008. Adults older than 18 years who arrived within 10 hours after acute poisoning were considered. The authors initially identified 358 presentations, then excluded records affected by specified missing information, transfers and lack of alcohol testing; 255 patients entered the principal alcohol-result analysis. Of those, 201 also had usable information about reported ingestion for the classification analysis. This was not a prospective random sample of drivers, a controlled alcohol challenge or a general-population screening study.
Most subjects—227 of 255 (89.0%)—had a recorded suicide-attempt intention in the underlying poisoning presentation. That is a description of case mix, not evidence that ethanol caused suicidal behavior. Drugs or other poisons, altered consciousness and differences in the timing of presentation can affect what the patient or another person can accurately report.
What did “self-report” mean in the records?
The recorded drinking history could come from the patient or a guardian or other informant, including a family member, friend or witness. Consequently, the paper’s label “self-report” does not establish that every answer was a first-person statement by the person tested. In an individual case, the source of the history, exact wording, record author and collection time must be established before interpreting an apparent admission or denial.
What alcohol measurement was used?
The investigators described sampling arterial or venous blood within approximately one hour of arrival and using the DRI Ethyl Alcohol Assay, a clinical enzymatic ethanol method. They classified results above 10 mg/dL as alcohol-positive for their study and results at or below the stated cutoff as the comparison group. The publication does not document a forensic headspace gas-chromatography confirmation and does not supply the procedural evidence necessary to treat a result as a particular jurisdiction’s evidentiary whole-blood BAC. The authors also did not stratify their conclusions by arterial versus venous sampling.
The 10 mg/dL research classification boundary is 0.01 g/dL, not a statutory driving threshold, and a test result on one specimen at one time cannot itself determine the exact time, volume or source of prior consumption. General analytical questions concerning method validation, controls and matrix are explained separately in the site’s blood alcohol testing methods guide.
What did the actual 201-case comparison show?
The paper’s published Table 3 provides a two-by-two breakdown by recorded drinking history and alcohol test classification. Combining the male and female cells yields the following observed counts; this is a direct arithmetic aggregation of that table, not a newly fitted statistical model.
| Recorded report | Test >10 mg/dL | Test not >10 mg/dL | Total |
|---|---|---|---|
| Reported alcohol ingestion | 114 | 11 | 125 |
| Denied alcohol ingestion | 4 | 72 | 76 |
| Total | 118 | 83 | 201 |
The most important discordant cell for a claimed denial is 4 of 76: these individuals had a recorded statement of no drinking, but their clinical alcohol measurement exceeded 10 mg/dL. The 11 opposite discordances likewise do not establish intentional misrepresentation. Someone may truthfully report earlier consumption when the sampled concentration later falls below a cutoff; history and measurement have different time frames. The study did not adjudicate the reason for each disagreement.
How do sensitivity, specificity and predictive value differ?
| Statistic | Observed calculation | Result | What it describes in this sample |
|---|---|---|---|
| Sensitivity | 114 / 118 | 96.6% | Positive reports among cases above the clinical ethanol cutoff |
| Specificity | 72 / 83 | 86.7% | Negative reports among cases not above the cutoff |
| Positive predictive value | 114 / 125 | 91.2% | Above-cutoff results among people recorded as reporting drinking |
| Negative predictive value | 72 / 76 | 94.7% | Not-above-cutoff results among people recorded as denying drinking |
These figures have different denominators. In particular, a 94.7% negative predictive value in the selected 201-case sample does not mean that any person’s denial is 94.7% accurate in a DUI investigation, workplace dispute, assault, or postmortem case. Predictive values depend on how frequently the condition occurs in the population tested, among other selection features. The researchers’ classification also used a low clinical cutoff and included possible third-party statements.
Why were the 54 missing histories important?
Fifty-four of the 255 analyzed patients—21.2%—lacked a usable drinking report and were excluded from the report-versus-test comparison. The investigators indicated that decreased mentation was common among these missing-report cases. That makes completeness especially important: the people least able to supply a history may differ materially from the people who can.
There is an additional source-table reconciliation issue. The paper’s prose assigns the missing histories between alcohol-positive and alcohol-negative groups in a way that does not reconcile with its published Table 3 and overall group totals. The 201-case contingency table above is internally consistent with the published sensitivity, specificity and predictive values; it should not be taken to resolve the article’s inconsistent missing-case description. An analyst should preserve the discrepancy rather than inventing a corrected allocation of the 54 omitted cases.
These limitations prevent a scientifically defensible assertion that the reported performance must generalize to unconscious patients, patients refusing interviews, police interviews, other hospitals or substantially different populations. Missingness is part of the evidence, not a detail that can be silently ignored.
Was the alcohol assay a direct measure of impairment?
No. A clinical laboratory ethanol result measures an analyte in a collected specimen using the laboratory’s stated procedure. The history question concerns prior ingestion, not whether a patient could drive safely, complete psychomotor tasks, remember an event accurately, or form a particular intention. The article reported neither a validated psychomotor battery nor an individual impairment model. Its population-level test-performance calculations cannot bridge that evidentiary gap.
Nor can the paper establish the extent to which a sedative drug, another poison, ethanol, injury, sleep loss or medical illness contributed to impaired consciousness in a particular presentation. Such claims require chronology, verified analytical findings, clinical records, alternate explanations and appropriately limited expert interpretation. The site’s pharmacokinetics guide explains why a concentration at one time cannot automatically reconstruct an earlier individual exposure.
What can one discordant clinical alcohol result mean?
Several competing interpretations exist. A reported exposure may have occurred long enough before collection that a result was below the research cutoff. A recorded history may have been incomplete, misunderstood or supplied by an observer without direct knowledge. The analyte method, reporting limit, specimen type and collection time might not match an assumption made about the record. Conversely, a denial accompanied by an above-cutoff result does not by itself identify when ethanol was consumed or prove a particular person’s intent.
This is why the first forensic question is not “Who was lying?” but “What exactly was asserted, by whom, at what time, against which measurement?” Preservation of the original interview and sampling chronology is essential. For practical distinctions involving collection site, tube, timing and handling, see the site’s specimen-collection framework.
What did the researchers observe across the entire 255-patient group?
In the full sample, 167 of 255 (65.5%) exceeded the clinical study’s 10 mg/dL threshold. The authors reported median (interquartile-range) alcohol measurements of 85.0 (10.0–173.5) mg/dL for the 97 men and 32.0 (4.0–137.5) mg/dL for the 158 women. These sample summaries describe a selected poisoned emergency-department population, not biological sex effects independently established by a controlled dosing design. Distributions could reflect case mix, exposures, sampling times and selection.
Similarly, the high proportion of poisoning presentations associated with suicide intention cannot be used to infer that measured alcohol caused an attempt. The investigators expressly did not assess that causal relationship. Conflating a subgroup percentage with a legal or clinical causal conclusion would exceed this study’s design.
How should a forensic expert review a claim based on drinking history?
A reproducible review distinguishes the provenance of the statement from the provenance of the measurement. For the statement, establish whether the patient or another informant spoke, the exact language, possible language or memory barriers, contemporaneous level of consciousness, who transcribed the answer, and when it was documented. For the laboratory result, identify the biological matrix, arterial or venous source, collection timestamp, method, units, threshold, laboratory acceptance criteria and availability of confirmatory analysis.
Next, compare like with like. A yes/no question about ever drinking before arrival is not a quantitative dose history. A concentration reported in mg/dL on a clinical assay is not automatically a forensic whole-blood determination suitable for statutory comparison. A negative answer does not rule out exposure, and a positive answer does not prove a concentration at the earlier time of an alleged incident.
Finally, separate exposure from effect. Before expressing an opinion about impaired performance, consciousness, decision-making or causation, look for independent, temporally relevant functional findings and plausible competing explanations. An expert can accurately report that the study found useful group-level agreement without claiming it validates a particular statement or an individual retrograde calculation.
Frequently asked questions
Can someone deny drinking and still have a measurable blood alcohol result?
Yes. Four of 76 people recorded as denying ingestion had clinical ethanol results above 10 mg/dL in this study. The discrepancy does not, without other evidence, reveal the reason for the denial or establish when alcohol was consumed.
Does an admission of drinking prove a positive BAC at hospital arrival?
No. Eleven of 125 people recorded as reporting ingestion did not exceed the study’s cutoff. Reporting earlier consumption and having ethanol above a particular threshold at a later sampling time are different propositions.
Did the study measure whole-blood ethanol by headspace GC?
It described an arterial or venous clinical blood sample tested using the DRI Ethyl Alcohol Assay. It did not document the forensic whole-blood headspace GC confirmation required to equate its method with every legal evidentiary program.
Was the self-report always the patient’s own statement?
No. The authors permitted reports from relatives, friends, guardians or witnesses. The identity and firsthand knowledge of the informant are material when interpreting an individual record.
Does the study establish that alcohol caused suicidal behavior or loss of consciousness?
No. Those causal effects were not experimentally isolated or proven in this retrospective review. Other ingested substances and clinical circumstances were common.
Can the reported 94.7% negative predictive value be used in a DUI case?
Not without independent validation in a substantially comparable setting. It describes a particular case mix, clinical cutoff, nonrandom availability of interview history and possible third-party accounts.
Why does it matter that 54 patients had no recorded drinking history?
Those patients were absent from the 201-case classification calculation. If missing reports cluster among people with altered consciousness or other important characteristics, results for the analyzable subset may not generalize to everyone encountered.
What is the scientifically supportable bottom line?
Reported drinking history may assist a clinical assessment; it is not an analytical substitute for measuring ethanol and does not establish an individual’s BAC, degree of impairment, intent or causal contribution to an event.
Primary research citation
Woo SH, Lee WJ, Jeong WJ, Kyong YY, Choi SM. Blood alcohol concentration and self-reported alcohol ingestion in acute poisoned patients who visited an emergency department. Scandinavian Journal of Trauma, Resuscitation and Emergency Medicine. 2013;21:24. doi:10.1186/1757-7241-21-24. Full original publisher study, including published Table 3. The article above is an independent forensic interpretation of that study’s reported evidence and limitations, not a determination in any individual case.