A crowd survey can describe alcohol exposure at a sporting event, but it cannot establish that a particular spectator was impaired or that alcohol caused an injury or confrontation. In a Swedish study of 4,420 football spectators, 46.8% had a positive breath-derived alcohol estimate and 8.9% reached the investigators’ high-concentration category. Those findings support population-level prevention questions. They do not identify an untested person’s concentration, behavior, or responsibility.
The 2017 primary study by Durbeej and colleagues illustrates three distinctions that matter in expert evidence: breath measurement versus collected blood, a subgroup average versus a crowd average, and statistical association versus individual causation.
What did the investigators actually measure?
This was a cross-sectional field survey at 12 Swedish Premier Football League matches in three arenas in Stockholm and Gothenburg. Data collection ran from April 17 through September 17, 2015. The researchers approached 6,132 spectators; 4,425 agreed to participate, and five under the minimum age were excluded, leaving 4,420 people aged 16 or older. The response rate was approximately 72%.
Recruiters approached every third person crossing an imaginary line. If that person belonged to a group, all group members were invited. Teams worked in supporter and other sections. This structured recruitment is stronger than simply asking whoever volunteers, but it still includes nonresponse and group recruitment that matter when judging representativeness.
Despite the paper’s use of “blood alcohol concentration” or BAC, investigators used a Dräger Alcotest 6820 breath analyzer. They did not collect venous whole blood, serum, plasma, urine, or an alcohol biomarker for this survey. Participants received water to clear the mouth, completed a brief interview, and then provided a breath sample. The paper reported breath-derived estimates in BAC-style percentage units. Those estimates should retain their measurement origin when described in a report or testimony.
No experimental alcohol dose was administered. Quantities, drinking rates, and absorption histories were not standardized. Pre-entry drinking was self-reported. A single survey reading therefore cannot supply an individual absorption curve, elimination rate, peak concentration, or earlier-time BAC.
The denominator changes the meaning of the average
Of the 4,420 participants, 2,068 had positive readings. The reported mean of 0.063% belonged to this positive-reading subgroup, rather than to all spectators. Likewise, the mean of 0.135% belonged to the 395 participants whose readings were at least 0.1%, the researchers’ definition of a high intoxication category.
Calling 0.063% “the average for every spectator” would silently remove the zero-reading participants from the denominator. Calling 0.135% the typical concentration of everyone drinking would substitute a selected higher-concentration subgroup for the broader drinking group. Both changes exaggerate what the numerical summaries mean.
The threshold of 0.1% was an operational category in this study. The survey did not validate it as a universal functional-impairment cutoff or establish the legal threshold applicable to another place, event, or proceeding.
Which associations were significant—and which comparisons were not?
In a multiple-regression analysis of 4,352 participants with complete predictor data, higher reported alcohol estimates were associated with male gender, younger age, attending a local derby, attending a weekend match, reporting drinking before entry, and being assessed in a supporter section. The gender association had p = 0.005; the other listed associations had p < 0.001. The model explained 42% of the observed outcome variance.
The positive-reading subgroup had a higher mean at local derbies than at non-derbies, 0.066% versus 0.061% (p = 0.011). Its supporter-section mean was 0.071%, compared with 0.056% in other sections (p < 0.001). These are group comparisons within a defined survey, not predictions of any named spectator’s reading.
Several comparisons were not statistically significant. The authors found no significant difference in participation response rates between supporter and other sections (p = 0.077). The mean among positive-reading spectators did not differ significantly across the three arenas overall (p = 0.189), and the mean within the at-least-0.1% subgroup did not differ significantly across arenas (p = 0.657). These null findings should not be replaced by a claim that every venue or supporter comparison showed a difference.
Why this is exposure evidence rather than an impairment examination
The investigators measured alcohol-related exposure and interviewed spectators. They did not administer a driving test, balance examination, cognitive battery, or validated individual impairment assessment. Nor did they pair each reading with a documented assault, fall, or collision and test whether alcohol caused that outcome.
Accordingly, membership in a higher-average group cannot prove a particular person’s impairment. Being young, male, seated in a supporter section, or attending a derby is not a substitute for that person’s chemical testing and observed performance. The regression associations also cannot establish that the event, section, or day of the week caused the higher concentration.
Timing and setting limit transfer to an actual case
Sampling began about an hour before kickoff and continued through halftime and roughly five minutes into the second half. It did not characterize the entire crowd at departure. Continued drinking could produce higher later readings, but the study did not measure those later values in the same people. It would be speculative to assign a universal increase.
The authors also identified the cross-sectional design, potentially biased interview responses, and limits on generalization beyond Swedish football. About 28% of those approached did not participate. Results from selected matches in two cities in 2015 cannot simply be treated as present-day prevalence for another country, sport, or venue.
How should the evidence be used?
For population planning, the study provides a reason to investigate alcohol availability, pre-entry drinking, and targeted prevention. It did not test whether a particular prevention program worked. For an individual forensic question, establish the actual specimen or breath method, collection time, instrument records, consumption chronology, behavior, and competing explanations before moving from concentration to impairment or causation.
Timing also matters when the disputed issue is a witness’s memory rather than crowd prevalence. Our discussion of drinking after an incident and eyewitness recall explains why exposure before observation, after observation, and during an interview pose different questions.
Primary scientific source
Durbeej N, Elgán TH, Jalling C, Gripenberg J. Alcohol intoxication at Swedish football matches: A study using biological sampling to assess blood alcohol concentration levels among spectators. PLOS ONE. 2017;12(11):e0188284. doi:10.1371/journal.pone.0188284.