Published version in Drug and Alcohol Dependence:
Are short AUDIT screeners effective in identifying unhealthy drinking of varying severity?
A prison population study
University College of Norwegian Correctional Service
Norwegian Institute of Public Health
Norwegian Centre for Addiction Research, University of Oslo,
Section for Clinical Addiction Research, Oslo University Hospital, Norway,
University College of Norwegian Correctional Service
University College of Norwegian Correctional Service
P.O. Box 1, 2001 Lillestrøm
Tel: +47 400 25 537 / +47 971 83 772
Conflict of interest: None
Background: Whether brief versions of the Alcohol Use Disorder Identification Test (AUDIT) can be
used as graded severity measures is largely unknown. We examined the performance of eight such
brief screeners in a prison population, and compared their effectiveness in detecting hazardous
drinking, harmful drinking, and possible alcohol dependence as classified by the full ten-item AUDIT.
Methods: The study sample included pre-prison drinkers who participated in the Norwegian Offender
Mental Health and Addiction (NorMA) study (n=758). We conducted receiver operating
characteristic curve (ROC) analyses and estimated the area under the curve (AUROC) to assess the
performance of AUDIT-C (three consumption items) and four-item versions that consisted of AUDIT-
C and one additional item.
Results: AUDIT-C performed very well in detecting unhealthy drinking of varying severity (AUROCs
of 0.933 or 0.935). Four-item versions performed even better. Of these, the well-established AUDIT-4
was superior in identifying harmful drinking (AUROC=0.969) and possible alcohol dependence
(AUROC=0.976). For AUDIT-C, the optimal cut-points in terms of the highest combined sensitivity
and specificity were ≥6 (hazardous drinking), ≥8 (harmful drinking) and ≥8 or ≥9 (possible
dependence). The corresponding cut-points on AUDIT-4 were ≥6, ≥9 and ≥10. The highest cut-point
whereby all cases of possible dependence were identified was ≥6 on AUDIT-C and ≥8 on AUDIT-4.
At these cut-points, almost all individuals with harmful drinking were also detected.
Conclusions: AUDIT-C and AUDIT-4 were both highly effective in detecting hazardous drinking,
harmful drinking and possible alcohol dependence. AUDIT-4 was superior, notably as a graded
Key words: AUDIT, short versions, screening, validation study, alcohol problems, prison
Hazardous drinking and alcohol problems are prevalent among individuals who enter prison (Fazel et
al., 2017; Newbury-Birch et al., 2016; Seal et al., 2018), which is unsurprising; alcohol is implicated in
various types of crime, notably violence (Evans et al., 2021; Graham and West, 2001; Rossow and
Bye, 2013), and pre-prison heavy drinking is predictive of recidivism (Dowden and Brown, 2002). In
addition to a range of adverse health and psychosocial outcomes (Babor et al., 2010; Rehm et al.,
2017), excessive alcohol use is associated with an increased suicide risk during incarceration (Fazel et
al., 2008) and an elevated post-release mortality rate (Chang et al., 2015). It is thus imperative to
identify those who have been drinking heavily, and to offer adequate interventions.
Many European countries screen individuals for alcohol problems upon entry into prison, but
validated tools are rarely used (WHO, 2019a, 2019b). The prison service in England and Wales stands
out in this respect, as universal screening based on the Alcohol Use Disorder Identification Test
(AUDIT) has been put into practice. The AUDIT has exhibited high validity across nations and
settings (Babor and Robaina, 2016; Meneses-Gaya et al., 2009) and been described as a gold standard
in detecting drinking behavior that is potentially or currently harmful to health. Studies of prison
populations also suggest high validity of this screening tool (Coulton et al., 2012; Thomas et al., 2014)
and that the reliability may be higher when the screening occurs a few weeks after incarceration rather
than upon entry into prison (Maggia et al., 2004).
The AUDIT was not merely designed to identify risky drinking behavior or alcohol problems,
but also to suggest interventions that vary according to the severity of the problems (Babor, et al.,
2001; Saunders et al., 1993). The ten items capture alcohol consumption (items 1-3), dependence
symptoms (items 4-6) and alcohol-related harm (items 7-10), and the total score ranges from 0 to 40.
Scores ≤7 denote a negative screen, that is, abstinence or low-risk drinking. The standard
categorization of positive screens and the recommended interventions are as follows (Babor and
Hazardous alcohol use (scores 8-15): Brief intervention
Harmful alcohol use (scores 16-19): Brief intervention, further monitoring, and diagnostic
Possible alcohol dependence (scores ≥20): Specialist treatment
The assessment of individuals who enter prison typically covers a broad range of domains, and
full versions of screening instruments may add undue burden to the intake procedure. Short versions of
the AUDIT may also be preferable for other reasons. Brady and co-workers (2002) found that many
items were perceived as “intrusive” or “prying into private lives” by those being screened. The
consumption items tended to be perceived as less invasive and easier to comprehend than those
capturing dependence symptoms and alcohol-related harm.
The three AUDIT consumption items are also essential to the instrument’s validity (Higgins-
Biddle and Babor, 2018), and they constitute the well-established AUDIT-C. This brief screener has
exhibited high effectiveness in detecting unhealthy drinking behavior as determined by external
validation criteria such as diagnostic assessment of alcohol use disorder (Kriston et al., 2019).
AUDIT-C has also shown to perform approximately as well as the full AUDIT in various contexts.
Other studies have examined the performance of abbreviated AUDIT screeners against the full
screening instrument. The bulk of this research focuses on the effectiveness of the AUDIT-C in
detecting a positive screen (AUDIT scores ≥8) (e.g. Morojele et al., 2017; Nehlin, et al., 2012;
Neumann et al., 2012; Seth et al., 2015), and high levels of accuracy have generally been reported.
However, when the screening solely identifies individuals with either hazardous drinking, harmful
drinking, or possible dependence, the results provide no cues with respect to likely treatment needs.
To our knowledge, only two studies have examined whether brief AUDIT screeners may work
as graded severity measures. One was based on a sample of Aboriginal Australians (Calabria et al.,
2014), the other was a general population study from Japan (Osaki et al., 2014). Both studies focused
on AUDIT-C, and both concluded that it performed well in identifying both hazardous, harmful, and
possible dependent drinking (as classified by the full AUDIT).
One may expect that four-item AUDIT screeners perform even better than the three-item
AUDIT-C. The AUDIT-4 includes one item about other people’s concern about one’s drinking (item
10) in addition to AUDIT-C, but validation studies indicate that it performs only slightly better than
AUDIT-C (Gual et al., 2002; Lee et al., 2018; Meneses‐Gaya et al., 2010; Wu et al., 2008). One may
thus ask whether other combinations of AUDIT-C and one additional item are preferable. No studies
seem to have addressed this question thus far.
A final issue of interest pertains to the effectiveness of brief AUDIT versions in screening
prison populations. Caviness and co-workers (2009) examined the accuracy of AUDIT-C and item 3
alone (i.e. the frequency of consuming 6+ units) in a sample of incarcerated females and their results
were encouraging – notably for AUDIT-C. However, a positive screen on the full AUDIT was the one
and only reference standard. No additional studies seem to have explored the validity of brief AUDIT
screeners among imprisoned persons. Similar studies of other groups with a high prevalence of alcohol
problems are also few and far between.
Against this backdrop, we examined the effectiveness of AUDIT-C and of four-item versions
that consisted of AUDIT-C and one additional item in identifying unhealthy drinking behavior of
varying severity (as classified by the full AUDIT) in a prison population. We compared the
performance of these brief screeners and examined cut-points on AUDIT-C and on the best
performing four-item version.
2. MATERIAL AND METHODS
Data stemmed from the Norwegian Offender Mental Health and Addiction (NorMA) study (Bukten et
al., 2015). The target sample comprised all incarcerated individuals in Norway at the time of the data
collection (June 2013- July 2014), and there were no pre-defined exclusion criteria.
The questionnaire was translated into four languages, but some individuals could not read any
of these. Other reasons for non-participation were temporarily absence from the prison unit and
preclusion of study eligibility by prison authorities for security reasons. About 40 percent of the prison
population responded (n=1499), and the sample was representative with respect to many demographic
variables (Bukten et al., 2015).
Participation was voluntary and based on written informed consent. It was explicitly pointed
out that refraining from participation was not associated with any sanctions and that the responses to
the self-report questionnaire were strictly confidential. The NorMA-study was approved by the
Norwegian Committee of Research Ethics. Details about data collection and ethics are reported
elsewhere (Bukten et al., 2015; Bukten et al., 2016).
2.1 Study sample
Eighty percent reported that they had consumed alcohol in the year before incarceration. From this
group, we extracted a subsample of 758 respondents whose current imprisonment had lasted 12
months or less, and who had responded to all the ten AUDIT items.
2.2.1 Alcohol Use Disorder Identification Test (AUDIT).
The original AUDIT consumption items are formulated in present tense without a specified time
frame, while all other items refer to the past year (Babor et al., 2001). However, a past-year reference
period is frequently used for all the items (e.g. Bush et al., 1998; Cook et al., 2005; Towers et al.,
2011). In our study, the AUDIT items were modified to assess the 12 months prior to imprisonment
Table 1 about here
A positive screen (scores ≥8), at least harmful drinking (scores ≥16), and possible alcohol dependence
(scores ≥20) were the classification standards. In addition to AUDIT-C (item 1+2+3), we constructed
all possible versions that consisted of the AUDIT-C and one additional item. We used the term
‘AUDIT-10’ rather than ‘the full AUDIT’ when describing our results.
We used data on gender, age, and current imprisonment length.
2.3 Statistical analyses
We conducted receiver operating characteristic (ROC) curve analysis (Hanley, 1989), and calculated
the area under the curve (AUROC) to examine the performance of the brief AUDIT screeners in
detecting unhealthy drinking behavior of varying severity (as classified by AUDIT-10). ROC curves
plot the true positive rate (sensitivity) against the false positive rate (1-specificity) for all cut-points,
and the AUROC ranges from 0.5 (no better than chance) to 1.0 (perfect match). The AUROC has been
considered fair for values between 0.7–0.8, good for values between 0.8–0.9 and excellent for values
above 0.90 (Cuparencu et al., 2020). Because we examined the performance of brief AUDIT screeners
against the full AUDIT, and thus relied on internal reference standards, the likelihood of obtaining
high AUROC-values was elevated.
We compared ROC curves using z-statistics for paired design (DeLong et al., 1988) to test
whether the four-item AUDIT screeners performed significantly better than the AUDIT-C, and to
identify the screener that most accurately identified possible alcohol dependence. Next, we examined
cut-points for each of the three categories of unhealthy drinking, and applied Youden’s (1950) J to
identify the highest combined level of sensitivity and specificity
-1). We also calculated the positive
predictive value (PPV) and the negative predictive value (NPV) of various cut-points.
In this study, sensitivity refers to the percentage of the positive cases on the AUDIT-10 that
were also identified as positive cases by the brief AUDIT screener. Specificity is the percentage of
negative cases on the AUDIT-10 that were correctly identified as such by the brief screener. PPV is
the percentage of the positive cases on the brief screener that were also positive cases on AUDIT-10,
while NPV is the percentage of the negative cases on the brief screener who were also identified as
such on AUDIT-10. Thus, the higher the PPV the lower is the occurrence of false positives, and the
higher the NPV the lower is the occurrence of false negatives. Unlike sensitivity and specificity, the
PPV and the NPV depend on the prevalence of “true” positive cases in the sample.
We used MedCalc Statistical Software 19.5.3 (ROC analyses) and SPSS 26 (cross-tabulations
2.3.1 Sensitivity analyses
Screening for health and psychosocial problems typically occurs shortly after imprisonment.
Therefore, we compared the percentage distribution across the four AUDIT-10 categories for
individuals reporting short-term (<3 months) and those reporting longer-term (3-12 months)
imprisonment. Next, we examined whether the main findings persisted when restricting the ROC
analyses to the former group.
3.1 Sample description
The vast majority (93%) of the respondents were males, 63% were 35 years old or younger, and 49%
had been imprisoned less than 3 months (Table 2). Moreover, two thirds had an AUDIT-10 positive
screen; 34% had hazardous drinking behavior (scores of 8-15), 10% had harmful drinking behavior
(scores of 16-19), and 22% were possibly alcohol dependent (scores ≥20).
Table 2 about here
3.2 Performance of the brief AUDIT screeners
The AUROCs for detecting each of the three categories of unhealthy drinking exceeded 0.90 for all
the brief versions of the AUDIT (Table 3). Moreover, the four-item versions performed significantly
better than the AUDIT-C, yet there was one exception; AUDIT-C + item 9 was not more effective
than AUDIT-C alone in detecting possible alcohol dependence (AUDIT-10 scores ≥ 20).
Table 3 about here
AUDIT-4 had the highest AUROC for identifying possible dependence. Additional analyses
showed that it differed significantly from AUDIT-C + item 4 (p=0.011), AUDIT-C + item 5
(p=0.002), AUDIT-C + item 6 (p<0.001), AUDIT-C + item 8 (p=0.003) and AUDIT-C + item 9
(p<0.001) in this respect. AUDIT-4 had also the highest AUROC for detecting at least harmful
drinking (AUDIT-10 scores ≥ 16). Its effectiveness in doing so exceeded that of AUDIT-C + item 5
(p=0.022), AUDIT-C + item 6 (p<0.001), AUDIT-C + item 8 (p=0.010) and AUDIT-C + item 9
(p=0.019). Hence, we selected AUDIT-4 for further analyses.
3.3 Cut-points on AUDIT-C and AUDIT-4
Table 4 provides the sensitivity, the specificity, Youden’s J, the positive predictive value and the
negative predictive value for selected cut-points on the AUDIT-C and the AUDIT-4 for each of the
three AUDIT-10 categories of unhealthy drinking.
Table 4 about here
Starting with AUDIT-C, a cut-point of ≥6 had the highest Youden’s J for detecting individuals
with an AUDIT-10 positive screen (sensitivity: 86%, specificity: 88%). The PPV was 94% (i.e. 6%
false positives) and the NVP was 77% (i.e. 23% false negatives). The highest J-value for at least
harmful drinking was reached at a cut-point of ≥8 (sensitivity: 85%, specificity: 85%), closely
followed by a cut-point of ≥7 (sensitivity: 94%, specificity: 76%). The former had higher PPV (74%)
than the latter (65%), while the NPVs were 96% and 92%, respectively. The optimal cut-point for
possible alcohol dependence was either ≥8 (sensitivity: 93%, specificity: 78%) or ≥9 (sensitivity: 80%,
specificity: 91%). The PPV was much lower at a cut-point of ≥8 (54%) than at a cut-point of ≥9
(71%), while the NPVs were almost equally high (98% and 94%, respectively). The percentages
scoring ≥6, ≥8 and ≥9 on AUDIT-C were 61%, 38% and 25%, respectively. Table 4 also shows that
the cut-point of ≥6 was the highest whereby all individuals with possible alcohol dependence were
identified. This cut-point also captured almost all (98%) individuals with at least harmful drinking.
Moving to AUDIT-4, the highest J-value for a positive screen was observed at a cut-point of
≥6 (sensitivity: 89%, specificity: 88%). The PPV and the NPV were 93% and 81%, respectively. The
optimal cut-point for at least harmful drinking was ≥9 (sensitivity: 89%, specificity: 92%), with a PPV
of 84% and a NPV of 94%. Regarding possible alcohol dependence, a cut-point of ≥10 yielded the
highest J-value (sensitivity: 96%, specificity: 82%). The PPV was 71% and the NPV was 99%. The
percentages scoring ≥6, ≥9 and ≥10 on AUDIT-4 were 63%, 34% and 30%, respectively. Moreover, a
cut-point of ≥8 was the highest to identify all individuals with possible alcohol dependence. It also
identified 97% of those with at least harmful drinking.
3.4 Sensitivity analyses
The percentage distribution across the four AUDIT-10 severity categories was not significantly
different for respondents who had been imprisoned <3 months and those who had spent 3-12 months
in prison (p=0.085). Moreover, all main findings were replicated when we restricted the ROC analyses
to the former group (n=371). Specifically, the AUROCs for the eight brief AUDIT screeners
consistently exceeded 0.90, almost all the four-item versions performed significantly better than
AUDIT-C, and AUDIT-4 had the highest AUROC for identifying at least harmful drinking and
possible alcohol dependence. It may be noted that a majority (58%) of the respondents in these
analyses had been imprisoned 30 days or less.
Our study included individuals who had consumed alcohol in the year before incarceration, and two
thirds had a positive screen on the full AUDIT. In this high-prevalence sample, AUDIT-C performed
very well in identifying both hazardous drinking, harmful drinking, and possible alcohol dependence
(as classified by the full AUDIT). Four-item versions that consisted of the AUDIT-C and one
additional item performed even better. Of these, AUDIT-4 was superior in detecting harmful drinking
and possible dependence.
We also identified optimal cut-points in terms of the highest combined sensitivity and
specificity (i.e. Youden’s J). Regarding AUDIT-C, they were ≥6 for a positive screen, ≥8 for at least
harmful drinking, and ≥8 or ≥9 for possible alcohol dependence. Thus, the cut-points for the two most
severe drinking categories were indistinctive. The corresponding cut-points on AUDIT-4 were ≥6
(positive screen), ≥9 (at least harmful drinking) and ≥10 (possible dependence). However, cut-points
with the highest Youden’s J are not necessarily the “best” choice, as we will discuss later.
4.1 Comparisons with other studies
Our study focused on issues that barely have been addressed in previous research, but two studies have
examined the effectiveness of AUDIT-C as a graded severity measure (Calabria et al., 2014; Osaki et
al., 2014). Both found that a cut-point of >6 captured (virtually) all individuals with the more severe
alcohol problems (i.e. scores >16 on the full AUDIT). This result was replicated in our study.
The cut-points with the highest Youden’ J for at least harmful drinking and for possible
dependence differed between the above-mentioned studies, and from those in our study. This may
reflect that the sample size and the proportion scoring high on the full AUDIT varied markedly
between the studies. Thus, the higher they are, the better the accuracy of the brief screener.
One may expect that four-item versions of an instrument perform better than a three-item
version because they are likely to capture more of the variance. Indeed, AUDIT-C and one additional
item performed better than AUDIT-C alone, which agrees with previous comparisons of the
effectiveness of AUDIT-C and AUDIT-4 (Gual et al., 2002; Lee et al., 2018; Meneses‐Gaya et al.,
2010; Wu et al., 2008). However, few – if any –studies have explored whether AUDIT-4 performs
better than other combinations of AUDIT-C and one additional item. Whether AUDIT-4 may work as
a graded severity measure is another issue that seems to have been overlooked.
Finally, only one previous study to our knowledge has examined the validity of brief AUDIT
screeners in a prison population (Caviness et al., 2009). It was restricted to females, and a positive
screen on the full AUDIT was the only reference standard. Moreover, two studies examined whether
brief AUDIT screeners may work as graded severity measures, but none of them employed samples
that were comparable to ours (Calabria et al., 2014; Osaki et al., 2014).
4.2 Choosing cut-points
Generally, the choice of cut-points depends on the screening purpose. If monitoring and crude
assessment of the nature and the scale of alcohol problems in a population are the main purposes,
relying on Youden’s J seems reasonable. One should consider other cut-points when the aim is to
identify individuals in likely need of professional help due to their harmful drinking behavior, as
It has been argued that sensitivity should be prioritized if the purpose is to detect individuals
with severe alcohol problems (Baggio and Iglesias 2020; RodrÍguez-Martos and Santamariña, 2007).
However, when it considered imperative to capture virtually all cases of harmful drinking and possible
dependence, some – or maybe many – will inevitably be false positives. It is equally obvious that the
larger the occurrence of false positives, the less effective is the screener.
Regarding AUDIT-C, a cut-point of >6 was the highest whereby all possible alcohol
dependent individuals and almost all (98 percent) with at least harmful drinking behavior were
detected. This cut-point also had the highest Youden’s J for identifying a positive screen. Six in ten
(61 percent) had scores of >6, of whom 49 percent were false positives with respect to harmful
drinking or possible dependence.
The results for AUDIT-4 showed that a cut-point of >8 was the highest that identified all cases
of possible alcohol dependence and almost all (97 percent) individuals with at least harmful drinking
behavior. Less than half (45 percent) of the respondents had scores in this range. Of these, 29 percent
were false positives regarding the two categories of unhealthy drinking.
Thus, for the purpose of identifying virtually all individuals with more severe alcohol
problems, AUDIT-4 was clearly preferable. In contrast to AUDIT-C, our results also suggested that
AUDIT-4 may work as a crude severity measure. Specifically, in addition to the cut-point of >8, a cut
point of >6 was optimal (cf. Youden’s J) in identifying an AUDIT positive screen. If high sensitivity is
considered less important for less severe alcohol problems, scores of 6 or 7 would thus be indicative of
A serious limitation of our study is the lack of external validation criteria. Rumpf and co-workers
(2002) examined the performance of the full and a short version of AUDIT against diagnostic
measures in general populations samples and found that the full version performed somewhat better in
detecting alcohol dependence. In prison populations and other groups with high prevalence of alcohol
problems, it is particularly important to use a diagnosis of alcohol use disorder as reference standard.
The original AUDIT items in our study were modified to assess the year before incarceration.
Especially for those who had been incarcerated quite a few months, the responses may have been
hampered by inaccurate recall – which is likely to strengthen a tendency to underreport one’s drinking
(Greenfield and Kerr, 2008). The “true” AUDIT scores may thus have been higher than those
observed, implying that the positive and the negative predictive values (cf. Table 4) are potentially
misleading. On the other hand, all main results were replicated when we restricted the ROC analyses
to individuals who had been imprisoned less than 3 months. Anyhow, it would have been
advantageous if the respondents had been recruited consecutively a couple of weeks after entry into
prison (cf. Maggia et al., 2004).
According to several validation studies of the AUDIT or its brief versions, the optimal cut-
points are higher for males than for females (e.g. DeMartini and Carey, 2012; Levola and Aalto, 2015;
Osaki et al., 2014; Reinert and Allen, 2007; Verhoog et al., 2020). The vast majority (93%) of the
respondents in our sample were males, and the number of females was too low to examine such
potential gender differences. Previous analyses of the full NorMA sample showed that the percentage
distribution across the four AUDIT categories was gender invariant (Pape et al., 2020), yet the “best”
cut-points on brief AUDIT screeners may still be gender specific. Thus, high AUDIT-C scores may be
more strongly associated with alcohol-related harm among females, reflecting gender differences in
alcohol metabolism and the quantity required to reach a high blood alcohol concentration (Thomasson,
2002). Hence, the cut-off scores that we suggested may not be recommendable for females.
Finally, it is possible that the responses to the brief AUDIT screeners would have been
different if they had not been embedded in the full AUDIT.
4.4 Implications and suggestions for further research
The prison setting may potentially offer a golden opportunity to detect and treat individuals with
harmful drinking behavior. However, according to MacAskill et al. (2011) “alcohol problems among
prisoners are under-detected, under-recorded and under-treated”. The screening may be restricted to a
yes/no-question (Parkes et al. 2011) or to few unvalidated items (Obstbaum et al., 2015). The failure to
use standardized assessment tools is staggering, but three- or four-item screeners are feasible and
probably more acceptable than more time-consuming alternatives. This, in turn, underscores the
potential importance of our study.
The results indicated that AUDIT-4 may work as a graded severity measure. However, further
assessment of individuals with a positive screen (scores >6) is important to make well-founded
decisions regarding adequate interventions. If limited resources or other obstacles preclude the
possibility to do so, individuals whose AUDIT-4 scores are indicative of harmful drinking or alcohol
dependence (scores >8) should be prioritized.
Our study suggested that AUDIT-C and AUDIT-4 are about equally useful as an initial filter
whereby those with a positive screen are singled out and asked the remaining AUDIT questions. This
is particularly feasible if the single-item scores are recorded electronically and added up successively.
If such a two-step strategy is implemented, about 60 percent of the pre-prison drinkers should be asked
all the ten AUDIT questions (cf. Table 4). The size of this group will be markedly smaller (48%) if
AUDIT-4 is applied and the selection of individuals for further assessment is restricted to those with at
least harmful drinking behavior (scores >8).
Finally, additional studies of the performance of brief AUDIT screeners in prison populations
should be conducted. Such studies should address the issue of gender specific cut-points and employ
external reference standards – including diagnostic assessment of alcohol use disorder. More
knowledge is also warranted about acceptable modes of screening for alcohol problems, as perceived
by those being screened as well as the prison staff. The extent to which short AUDIT versions may
work as graded severity measures should also be pursued, both in studies of incarcerated persons and
in other population groups.
AUDIT-C performed very well in detecting unhealthy pre-prison drinking behavior of varying
severity. AUDIT-C and one additional item performed even better, of which AUDIT-4 was superior in
detecting harmful drinking and possible dependence.
The NorMA-study was funded by the South-Eastern Norway Regional Health Authority and The
Norwegian Centre for Addiction Research (SERAF). This paper was supported by The University
College of Norwegian Correctional Service and The Norwegian Institute of Public Health. The authors
wish to thank all those who participated in the study, and the prison management and the helpful staff
members at the local prison units for assisting in collecting data.
Hilde Pape performed all the statistical analyses and wrote the first draft of the manuscript - mainly in
co-operation with Ingeborg Rossow. Anne Bukten designed the NorMA-study, and prepared the files
for statistical analyses. All authors were involved in the process of completing the paper.
Babor T., Caetano R., Casswell S., Edwards G., Giesbrecht N., Graham K. et al. 2010. Alcohol: No
Ordinary Commodity: Research and Public Policy. Second edition. Oxford: Oxford University
Babor, T. F., de la Fuente, J. R., Saunders, J., Grant, M. 2001. The Alcohol Use Disorders
Identification Test: Guidelines for use in Primary Health Care. Second Edition. Geneva;
World Health Organization, Department of Mental Health and Substance Dependence
Babor, T. F., Robaina, K. 2016. The Alcohol Use Disorders Identification Test (AUDIT): A review of
graded severity algorithms and national adaptations. Int. J. Alcohol Drug Res. 5(2), 17-24.
Baggio, S., Iglesias, K. 2020. Commentary on Campbell and Strickland (2019): Caution is needed
when using self-reported alcohol use disorder screening tools. Addict. Behav., 100, 106115.
Brady, M., Sibthorpe, B., Bailie, R., Ball, S., Sumnerdodd, P. 2002. The feasibility and acceptability
of introducing brief intervention for alcohol misuse in an urban Aboriginal medical service.
Drug Alcohol Rev. 21(4), 375-380.
Bukten, A., Lund, I. O., Rognli, E. B., Stavseth, M. R., Lobmaier, P. et al. 2015. The Norwegian
offender mental health and addiction study–design and implementation of a national survey
and prospective cohort study. Substance abuse: research and treatment, 9, SART. S23546.
Bukten, A., Stavseth, M. R., Skurtveit, S., Kunøe, N., Lobmaier, P. et al. 2016. Rusmiddelbruk og
helsesitusjon blant innsatte i norske fengsel. Resultater fra The Norwegian offender menatl
health and addiction study (NorMA). Oslo: Seraf.
Bush, K., Kivlahan, D. R., McDonell, M. B., Fihn, S. D., Bradley, K. A. 1998. The AUDIT alcohol
consumption questions (AUDIT-C): an effective brief screening test for problem drinking.
Arch. Intern. Med. 158(16), 1789-1795.
Calabria, B., Clifford, A., Shakeshaft, A. P., Conigrave, K. M., Simpson, L., et al. 2014. Identifying
Aboriginal-specific AUDIT-C and AUDIT-3 cutoff scores for at-risk, high-risk, and likely
dependent drinkers using measures of agreement with the 10-item Alcohol Use Disorders
Identification Test. Addict. Sci. Clin. Pract. 9(1), 17.
Caviness, C. M., Hatgis, C., Anderson, B. J., Rosengard, C., Kiene, S. M., et al. 2009. Three brief
alcohol screens for detecting hazardous drinking in incarcerated women. J. Stud. Alcohol
Drugs, 70(1), 50-54.
Chang, Z., Lichtenstein, P., Larsson, H., Fazel, S. 2015. Substance use disorders, psychiatric disorders,
and mortality after release from prison: a nationwide longitudinal cohort study. The Lancet
Psychiatry, 2(5), 422-430.
Cook, R. L., Chung, T., Kelly, T. M., Clark, D. B. 2005. Alcohol screening in young persons attending
a sexually transmitted disease clinic: Comparison of AUDIT, CRAFFT, and CAGE
instruments. J. Gen. Internal. Med. 20(1), 1-6.
Coulton, S., Newbury-Birch, D., Cassidy, P., Dale, V., Deluca, P., et al. 2012. Screening for alcohol
use in criminal justice settings: an exploratory study. Alcohol Alcohol. 47(4), 423-427.
Cuparencu, C., Rinnan, Å., Silvestre, M. P., Poppitt, S. D., Raben, A., et al. 2020. The anserine to
carnosine ratio: an excellent discriminator between white and red meats consumed by free-
living overweight participants of the PREVIEW study Eur. J. Nutr. 60: 179-192.
DeLong, E. R., DeLong, D. M., Clarke-Pearson, D. L. 1988. Comparing the areas under two or more
correlated receiver operating characteristic curves: a nonparametric approach. Biometrics,
DeMartini, K. S., Carey, K. B. 2012. Optimizing the use of the AUDIT for alcohol screening in
college students. Psychol. Assess. 24(4), 954.
Dowden, C., Brown, S. L. 2002. The role of substance abuse factors in predicting recidivism: A meta-
analysis. Psychol. Crime Law, 8(3), 243-264.
Evans, I., Heron, J., Murray, J., Hickman, M., Hammerton, G. 2021. The influence of alcohol
consumption on fighting, shoplifting and vandalism in young adults. Int. J. Environ. Res.
Public Health 18(7), 3509.
Fazel, S., Cartwright, J., Norman-Nott, A., Hawton, K. (2008). Suicide in prisoners: a systematic
review of risk factors. J. Clin. Psychiatry, 69(11), 1721-1731.
Fazel, S., Yoon, I. A., Hayes, A. J. 2017. Substance use disorders in prisoners: an updated systematic
review and meta‐regression analysis in recently incarcerated men and women. Addiction,
Graham, K., West, .P 2001. Alcohol and crime: Examining the link. In: Heather N., Peters T.J.,
StockwellT. (Eds) International Handbook of Alcohol Dependence and Problems (pp. 439–
470). New York; Wiley and Sons Ltd.
Greenfield, T. K., Kerr, W. C. 2008. Alcohol measurement methodology in epidemiology: recent
advances and opportunities. Addiction, 103(7), 1082-1099.
Gual, A., Segura, L., Contel, M., Heather, N., Colom, J. 2002. Audit-3 and audit-4: effectiveness of
two short forms of the alcohol use disorders identification test. Alcohol Alcohol. 37(6), 591-
Hanley, J. A. 1989. Receiver operating characteristic (ROC) methodology: the state of the art. Crit.
Rev. Diagn. Imaging 29(3), 307-335.
Higgins-Biddle, J. C., Babor, T. F. 2018. A review of the Alcohol Use Disorders Identification Test
(AUDIT), AUDIT-C, and USAUDIT for screening in the United States: Past issues and future
directions. Am. J. Drug Alcohol Abuse 44(6), 578-586.
Kriston, L., Hölzel, L., Weiser, A.-K., Berner, M. M., Härter, M. 2008. Meta-analysis: are 3 questions
enough to detect unhealthy alcohol use? Ann. Internn Med. 149(12), 879-888.
Lee, J. H., Kong, K. A., Lee, D. H., Choi, Y. H., Jung, K. Y. 2018. Validation and proposal for cut-off
values of an abbreviated version of the Alcohol Use Disorder Identification Test using the
Korean National Health and Nutrition Examination Survey. Clin. Exp.Emerg. Med. 5(2), 113-
Levola, J., Aalto, M. 2015. Screening for at‐risk drinking in a population reporting symptoms of
depression: a validation of the AUDIT, AUDIT‐C, and AUDIT‐3. Alcohol Clin. Exp. Res.
Maggia, B., Martin, S., Crouzet, C., Richard, P., Wagner, P., et al. 2004. Variation in audit (alcohol
used disorder identification test) scores within the first weeks of imprisonment. Alcohol
Alcohol. 39(3), 247-250.
MacAskill, S., Parkes, T., Brooks, O., Graham, L., McAuley, A., et al. (2011). Assessment of alcohol
problems using AUDIT in a prison setting: more than an'aye or no'question. BMC Public
Health, 11(1), 1-12.
Meneses-Gaya, C., Zuardi, A. W., Loureiro, S. R., Crippa, J. A. S. 2009. Alcohol Use Disorders
Identification Test (AUDIT): An updated systematic review of psychometric properties.
Psychol. Neurosci. 2, 83-97.
Meneses‐Gaya, C., Zuardi, A. W., Loureiro, S. R., Hallak, J. E., Trzesniak, C., et al. 2010. Is the full
version of the AUDIT really necessary? Study of the validity and internal construct of its
abbreviated versions. Alcohol Clin. Exp. Res. 34(8), 1417-1424.
Morojele, N. K., Nkosi, S., Kekwaletswe, C. T., Shuper, P. A., Manda, S. O., et al. 2017. Utility of
brief versions of the alcohol use disorders identification test (AUDIT) to identify excessive
drinking among patients in HIV care in South Africa. J. Stud. Alcohol Drugs, 78(1), 88-96.
Nehlin, C., Fredriksson, A., Jansson, L. 2012. Brief alcohol screening in a clinical psychiatric
population: Special attention needed. Drug Alcohol Rev. 31(4), 538-543.
Neumann, T., Linnen, H., Kip, M., Grittner, U., Weiβ-Gerlach, E., et al. 2012. Does the Alcohol Use
Disorders Identification Test–Consumption identify the same patient population as the full 10-
item Alcohol Use Disorders Identification Test? J. Subst. Abuse Treat. 43(1), 80-85.
Newbury-Birch, D., McGovern, R., Birch, J., O'Neill, G., Kaner, H., et al. 2016. A rapid systematic
review of what we know about alcohol use disorders and brief interventions in the criminal
justice system. Int. J. Prison Health, 12(1), 57-70.
Osaki, Y., Ino, A., Matsushita, S., Higuchi, S., Kondo, Y., et al. 2014. Reliability and validity of the
alcohol use disorders identification test—consumption in screening for adults with alcohol use
disorders and risky drinking in Japan. Asian Pac. J. Cancer Prev. 15(16), 6571-6574.
Obstbaum, Y., Tyni, S., Mattila, A. K., Vartiainen, H., et al. 2016. Not all substance dependence
problems are recognized as risks—Comparing a medical health study with prison
assessments. Eur. J. Crim. Policy Res. 22(1), 189-210.
Pape, H., Rossow, I., Bukten, A. 2020. Alcohol problems among prisoners: Subgroup variations,
concurrent drug problems, and treatment needs. Eur. Addict. Res. 1-10.
Parkes T, MacAskill S, Brooks O, Jepson R, Atherton I, McGhee S, et al. Prison health needs
assessment for alcohol problems. 2011, NHS Health Scotland, Available at:
Rehm, J., Gmel Sr, G. E., Gmel, G., Hasan, O. S., Imtiaz, S., et al. 2017. The relationship between
different dimensions of alcohol use and the burden of disease—an update. Addiction, 112(6),
Reinert, D. F., Allen, J. P. 2007. The alcohol use disorders identification test: an update of research
findings. Alcohol. Clin. Exp. Res., 31(2), 185-199.
RodrÍguez-Martos, A., Santamariña, E. 2007. Does the short form of the Alcohol Use Disorders
Identification Test (AUDIT-C) work at a trauma emergency department? Subst. Use Misuse
Rossow, I., Bye, E. K. 2013. The problem of alcohol-related violence: an epidemiological and public
health perspective. In: McMurran M. (Ed.) Alcohol-Related Violence: Prevention and
Treatment, pp. 3-18. Chichester: Wiley-Blackwell.
Rumpf, H. J., Hapke, U., Meyer, C., John, U. 2002. Screening for alcohol use disorders and at-risk
drinking in the general population: psychometric performance of three questionnaires. Alcohol
Alcohol, 37(3), 261-268.
Saunders, J. B., Aasland, O. G., Babor, T. F., De la Fuente, J. R., Grant, M. 1993. Development of the
alcohol use disorders identification test (AUDIT): WHO collaborative project on early
detection of persons with harmful alcohol consumption‐II. Addiction 88(6), 791-804.
Seal, D. W., Yancey, S., Reddy, M.,Kinner, S. A. 2018. Alcohol use among incarcerated individuals.
In A. Stuart and J. D. J. Rich (Eds.), Drug Use in Prisoners: Epidemiology, Implications, and
Policy Responses (pp. 57-70). New York: Oxford University Press.
Seth, P., Glenshaw, M., Sabatier, J. H., Adams, R., Du Preez, V., et al. 2015. AUDIT, AUDIT-C, and
AUDIT-3: drinking patterns and screening for harmful, hazardous and dependent drinking in
Katutura, Namibia. PloS one, 10(3), e0120850.
Thomas, E., Degenhardt, L., Alati, R., Kinner, S. 2014. Predictive validity of the AUDIT for
hazardous alcohol consumption in recently released prisoners. Drug Alcohol Depend., 134,
Thomasson H.R 2002. Gender differences in alcohol metabolism. In: Galanter M. (ed.) Recent
Developments in Alcoholism, vol 12. Boston; Springer. https://doi.org/10.1007/0-306-
Toner, P., Böhnke, J. R., Andersen, P., McCambridge, J. 2019. Alcohol screening and assessment
measures for young people: A systematic review and meta-analysis of validation studies. Drug
Alcohol Depend., 202, 39-49.
Towers, A., Stephens, C., Dulin, P., Kostick, M., Noone, J., et al. 2011. Estimating older hazardous
and binge drinking prevalence using AUDIT-C and AUDIT-3 thresholds specific to older
adults. Drug Alcohol Depend., 117(2-3), 211-218.
Verhoog, S., Dopmeijer, J. M., de Jonge, J. M., van der Heijde, C. M., Vonk, P., et al. 2020. The use
of the alcohol use disorders identification test–consumption as an indicator of hazardous
alcohol use among university students. Eur. Addict. Res. 26(1), 1-9.
WHO, 2019a. WHO. Health in Prisons: Fact Sheet for 38 European Countries. Copenhagen: WHO
Regional office for Europe; 2019. Available at:
WHO, 2019b. Screening for Harmful Use of Alcohol. Health in Prisons European Database. Available
Wu, S.-I., Huang, H.-C., Liu, S.-I., Huang, C.-R., Sun, F.-J., et al. 2008. Validation and comparison of
alcohol-screening instruments for identifying hazardous drinking in hospitalized patients in
Taiwan. Alcohol Alcohol. 43(5), 577-582.
Youden, W. J. 1950. Index for rating diagnostic tests. Cancer, 3(1), 32-35.
Table 1. The ten AUDIT items as formulated in the present study
1 How often did you have a drink containing alcohol in the year before incarceration?
2 How many drinks containing alcohol did you have on a typical day when you are drinking in
the year before incarceration
3 How often did you drink 6 or more units on one occasion in the year before incarceration?
Symptoms of dependence
4 How often in the year before incarceration were you not able to stop drinking once you had
5 How often in the year before incarceration did you fail to do what was normally expected from
you because of drinking?
6 How often in the year before incarceration did you need a first drink in the morning to get
yourself going after a heavy drinking session?
7How often in the year before incarceration did you have a feeling of guilt or remorse after
8 How often in the year before incarceration were you unable to remember what happened the
night before because you had been drinking?
9 Have you or someone else been injured as a result of your drinking?
10 Has a relative or friend or a doctor or another health worker been concerned about your
drinking, or suggested you to cut down?
Note: There were five response options for items 1-8 (scale range: 0-4), and three response options for items
9 and 10; “No” (scored 0), ‘Yes, but not in the year before incarceration’ (scored 2), and ‘Yes, in the year
before incarceration’ (scored 4).
Table 2. Descriptive statistics of the study sample1
Gender Males 92.6 (699)
Females 7.4 (56)
Age 17-25 years 26.4 (184)
26-35 years 37.0 (258)
≥ 36 years 37.6 (255)
Imprisonment length < 3 months 48.9 (371)
3-6 months 27.3 (207)
> 6 months123.7 (180)
Low-risk drinking (1-7) 33.9 (257)
Hazardous drinking (8-15) 33.6 (255)
Harmful drinking (16-19) 10.4 (79)
Possible dependence (≥ 20) 22.0 (167)
1 Those incarcerated >12 months were excluded
Table 3. AUROCs for detecting AUDIT-10 categories of unhealthy drinking
behavior, and differences between the AUDIT-C and the four-item screeners.
The screener with the highest observed AUROC appears in bold.
AUDIT-10 categories AUROC (95% CI) Comparison with
AUDIT-C (item 1+2+3) 0.935 (0.915-0.951) ---
AUDIT-4 (item 1+2+3+10)0.953 (0.935-0.967) p<0.001
AUDIT-C + item 4 0.950 (0.932-0.964)p<0.001
AUDIT-C + item 5 0.951 (0.937-0.964)p<0.001
AUDIT-C + item 6 0.946 (0.927-0.961)p<0.001
AUDIT-C + item 7 0.955 (0.938-0.969)p<0.001
AUDIT-C + item 8 0.953 (0.935-0.967)p<0.001
AUDIT-C + item 9 0.967 (0.952-0.979)p<0.001
At least harmful drinking2
AUDIT-C (item 1+2+3) 0.933 (0.913-0.950) ---
AUDIT-4 (item 1+2+3+10)0.969 (0.955-0.981) p<0.001
AUDIT-C + item 4 0.959 (0.942-0.972)p<0.001
AUDIT-C + item 5 0.956 (0.939-0.970)p<0.001
AUDIT-C + item 6 0.946 (0.927-0.961)p=0.001
AUDIT-C + item 7 0.961 (0.944-0.973)p<0.001
AUDIT-C + item 8 0.954 (0.937-0.968)p<0.001
AUDIT-C + item 9 0.954 (0.937-0.968)p=0.001
AUDIT-C (item 1+2+3) 0.935 (0.915-0.951) ---
AUDIT-4 (item 1+2+3+10)0.976 (0.963-0.986) p<0.001
AUDIT-C + item 4 0.961 (0.944-0.973) p<0.001
AUDIT-C + item 5 0.957 (0.940-0.970) p<0.001
AUDIT-C + item 6 0.948 (0.930-0.963) p=0.003
AUDIT-C + item 7 0.967 (0.951-0.978) p<0.001
AUDIT-C + item 8 0.958 (0.942-0.971) p<0.001
AUDIT-C + item 9 0.942 (0.923-0.958) p=0.281
1AUDIT-10 scores ≥8 2AUDIT-10 scores ≥16 3AUDIT-10 scores ≥20
Table 4. Sensitivity, specificity, positive predictive value (PPV) and negative predictive values (NPV) for selected cut-points on AUDIT-C (scale:1-12)
and AUDIT-4 (scale 1-16) for identifying unhealthy drinking behavior of varying severity (as classified by AUDIT-10). Optimal cut-points according to
Positive screen1≥4 78.8 95.8 54.5 0.50 80.4 87.0 81.5 98.4 51.4 0.50 79.0 94.3
≥5 69.0 91.2 74.3 0.66 87.4 81.3 71.8 94.4 72.4 0.67 86.9 86.9
≥6 60.9 86.2 88.3 0.75 93.5 76.7 63.3 89.4 87.6 0.77 93.3 80.9
≥7 47.0 69.9 97.7 0.68 98.3 62.4 51.1 75.9 97.3 0.73 98.2 67.4
≥8 37.6 56.9 100 0.57 100 54.3 44.5 67.3 100 0.67 100 61.0
At least harmful drinking2 ≥6 60.9 98.4 57.0 0.55 52.4 98.6 63.3 100 54.3 0.54 51.2 100
≥7 47.0 93.9 75.6 0.70 64.9 96.3 51.1 98.8 71.9 0.71 62.8 99.2
≥8 37.6 85.4 85.4 0.71 73.7 92.4 44.5 96.8 80.7 0.78 70.6 98.1
≥9 24.9 67.5 95.6 0.63 87.8 85.9 34.3 88.6 91.8 0.80 83.8 94.4
≥10 --- --- --- --- --- --- 29.6 80.9 95.1 0.76 88.8 91.2
≥11 --- --- --- --- --- --- 23.7 69.9 98.4 0.68 95.6 87.2
Possible dependence3 ≥6 60.9 100 50.1 0.51 36.1 100 --- --- --- --- --- ---
≥7 47.0 98.2 67.5 0.66 46.1 99.3 51.1 100 62.8 0.68 43.2 100
≥8 37.6 92.8 78.0 0.71 54.4 97.5 44.5 100 71.2 0.71 49.6 100
≥9 24.9 80.2 90.7 0.71 70.9 94.2 34.3 98.8 83.9 0.83 63.5 99.6
≥10 17.5 61.1 94.8 0.56 76.7 89.6 29.6 95.8 89.2 0.85 71.4 98.7
≥11 --- --- --- --- --- --- 23.7 88.0 94.4 0.82 81.7 96.5
≥12 --- --- --- --- --- --- 19.1 76.7 97.1 0.74 88.3 93.6
1AUDIT-10 scores ≥ 8 2AUDIT-10 scores ≥16 3AUDIT-10 scores ≥ 20
Youden’s J in bold. Cut-points with sensitivities or specificities below 50% are not displayed.