Study aims

This study forms part of a longer-term project to investigate the potential for randomised trials of brief motivation enhancement therapies for alcohol use disorders within acute psychiatric in-patient settings. In this study we wished to ascertain the prevalence and spread of severity of alcohol use disorders among psychiatric patients (including gender differences) and the extent to which these were associated with suicidality. Such data could be used to generate hypotheses about therapeutic techniques or potential obstacles that might influence the effectiveness of these interventions. The hypotheses under test were that alcohol use disorders would be particularly prevalent among psychiatric in-patients; that there would be a positive association between alcohol use disorders and in-patients with a history of suicidality; and that case-note review would indicate that alcohol use disorders were being underdetected.

An inclusive definition of suicidality was used. Individuals were classified as having a history of suicidality if there was a record in their multidisciplinary case notes that they had carried out any deliberate self-harm or had had thoughts or plans of self-harm or suicide.

Sample

Patients consecutively admitted to acute general psychiatry wards in two psychiatric hospitals were approached within 7 days of their admission and invited to take part in a confidential lifestyle survey. Potential participants were given verbal and written information about the study. Written, informed consent was obtained, including consent to review their case notes. Participants received a £5 shop voucher for their participation. Both voluntary patients and those detained under Sections 2 and 3 of the Mental Health Act 1983 were included, subject to their ability to complete the questionnaire in English. Transfers from other psychiatric units and those electively admitted to specialist units were excluded, as were those with serious physical illness and patients in medium secure or intensivecare wards.

A 6-month survey period was chosen because it was estimated that 200 subjects would be interviewed during this time, giving adequate statistical power: the overall population prevalence of alcohol use disorders in London is approximately 25% (Reference Singleton, Bumpstead and O'BrienSingleton et al, 2000). Participants were recruited from November 2001 to June 2002, excluding the Christmas/New Year period to avoid any seasonal bias towards excessive alcohol consumption.

Assessment tools

The first part of the survey instrument consisted of an anonymous self-completion lifestyle survey. It included demographic information, quality-of-life items (regularity of sleep pattern, use of leisure facilities, buying of lottery tickets) designed to be non-threatening and to maximise participation, a section on smoking, the Alcohol Use Disorders Identification Test (AUDIT; Reference Saunders, Aasland and BaborSaunders et al, 1993; Reference Babor, Higgins-Biddle and SaundersBabor et al, 2001) and a section on substance use.

The AUDIT is a ten-item self-completion screening instrument for alcohol use disorders that has been widely validated (Reference Reinert and AllenReinert & Allen, 2002) including in individuals with severe mental illness. Items are scored 0–4, giving a total range between 0 (no problems) and 40 (severe problems). The items cover three domains: excessive alcohol intake, dependence and problems related to drinking. To date there has been little research on the use of the AUDIT in the psychiatric population in the UK or the relationship between the AUDIT score and the risk of deliberate self-harm. The questions on smoking included items from the General Household Survey (Office of Population Censuses and Surveys, 1998). The substance use section was the Substance Abuse Assessment Questionnaire (Reference GhodseGhodse, 1995). In the second part of the survey the participants' multidisciplinary case notes for the period of the admission were reviewed and ICD–10 diagnoses (World Health Organization, 1992) were coded from clinical diagnoses made in the admission notes. Up to two diagnoses could be coded, allowing for primary and secondary/differential diagnoses. Subjects with and without a severe mental illness were compared. Serious mental illness was defined as a schizophreniform psychosis, obsessive-compulsive disorder, bipolar affective disorder or psychotic mood disorder.

The notes were reviewed to ascertain the duration of the psychiatric history and whether the subject was previously known to misuse alcohol or to have attended alcohol services. The quality of any alcohol and/or substance misuse history recorded was assessed according to predetermined criteria, including the recording of alcohol units, the results of any relevant physical investigations (gamma glutamyl transpeptidase, urine toxicology) and the use of any alcohol assessment instruments. These results are reported elsewhere (Reference Barnaby, Drummond and McCloudBarnaby et al, 2003).

Analysis and statistics

Two cut-offs for the AUDIT score were used: 8 or more indicates hazardous or harmful alcohol use (Reference Saunders, Aasland and BaborSaunders et al, 1993) and 16 or more represents a level of alcohol dependence requiring specific intervention and continued monitoring (Reference Singleton, Bumpstead and O'BrienSingleton et al, 2000; Reference Babor, Higgins-Biddle and SaundersBabor et al, 2001). Categorical data were analysed using χ² tests for association, pooling data with small numbers (e.g. ethnicity and diagnosis). Where continuous data were significantly skewed or there were significant between-group differences in variance (AUDIT score, age, duration of psychiatric history), non-parametric statistics were used. Logistic regression was performed using a forward stepwise model based on the Wald statistic to ascertain the relationship of the AUDIT score to suicidality. Data were analysed using the Statistical Package for the Social Sciences program, version 10.

Sample characteristics

During the study period there were 364 acute psychiatric admissions according to computerised records. Two hundred individuals (76% of those eligible) consented to take part in the study. Reasons for not taking part in the study are described in Table 1.

Table 1 Reasons for exclusion from study

	n	%
Included	200	54.9
Excluded
Ineligible	41	11.3
Not seen	60	16.5
Declined to take part	63	17.3
Total number of acute admissions	364	100
Reasons for ineligibility
Cognitive impairment	6	15
Language barrier	29	70
Transfer to a specialist unit	6	15
Reasons not seen
Participant on leave	2	3
Access refused by nursing staff	2	3
Participant transferred to another hospital	8	14
Participant self-discharged	10	17
Participant discharged	38	63

Subsequent analysis is based on the 200 individuals who consented to be involved with the study. Of those, 161 reported that they had drunk alcohol in the 1-year period covered by the AUDIT. Approximately equal numbers of men and women were admitted: 105 men (52.8%) and 94 women (47.2%). There were no significant differences between the proportions of men and women with a severe mental illness. Admission diagnoses were recorded on 189 participants. Thirty-nine received a secondary or differential diagnosis. Men were more likely to be diagnosed with a non-affective psychosis and women with a mood disorder. Substance and/or alcohol use disorders (ICD–10, F10.1–19.9) were recorded for 17 participants (4.7%), mostly alcohol-related (16 participants). There was no significant age difference between men and women (means of 39.6 and 42.6 years, respectively; t=–1.43, d.f.=190, P=0.19). Participants had been known to services for a mean of 8.3 years (range 0–43 years), with no significant gender differences. A quarter of participants had been known to services for a year or less, and half for less than 5 years. The demographic characteristics and psychiatric diagnoses of the sample are given in Table 2.

Table 2 Demographic characteristics and psychiatric diagnoses of the sample

Characteristic	Total		Men		Women		Significance
	n	%	n	%	n	%
Gender1	199	100	105	52.8	94	47.2
Ethnicity2
White British	129	64.5	68	64.8	60	63.8	NS
White Irish	12	6	8	7.6	4	4.3
Other White	16	8	6	5.7	10	10.6
Caribbean	11	5.5	9	8.6	2	2.1
African	8	4	3	2.9	5	5.3
Indian	6	3	2	1.9	4	4.3
Other categories	18	9	9	8.5	9	9.6
Accommodation
No fixed abode	16	8.0	11	10.6	5	5.6
Living alone3	97	50.5	53	50.4	43	45.7	NS
Employment status
Unemployed	59	29.5	39	37.1	19	25.3	χ2=6.9, d.f.=1, P=0.009
Invalid4	57	28.5	28	26.7	29	30.9	NS
Suicidality present5	103	52.5	55	52.4	48	51.1	NS
Diagnostic category
Psychosis F20—29	72	38.3	46	46.5	25	28.1	χ2=6.7, d.f.=1, P=0.01
Mood disorder F30—39	81	43.1	32	32.3	49	55.1	χ2=9.9, d.f.=1, P=0.002
All severe mental illness	147	78.2	77	77.8	69	77.5	NS
AUDIT score
≥ 8	97	48.5	56	53.3	41	43.6	NS
≥ 16	45	22.5	27	25.7	18	19.1	NS
Smoking status1
Regular smokers	125	63.1	75	72.1	50	53.2	NS
Sleep quality1,6
Regular sleep	73	36.7	39	37.1	34	36.2	NS

Prevalence of alcohol use disorders

Table 3 shows associations between participants' demographic characteristics and psychiatric diagnoses and their AUDIT scores. Ninety-seven participants (48.5%) had an AUDIT score of 8 or more, indicating hazardous or harmful alcohol use (‘AUDIT positive’). The AUDIT positive participants tended to be younger than the AUDIT negative ones (Mann–Whitney U test: z=–2.97, P=0.003), and were more likely to be White than from other ethnicities. Men and women were equally likely to be AUDIT positive. The AUDIT positive participants did not differ from the AUDIT negative ones on other significant variables: the proportion with psychosis (ICD–10, F20–29.9), any mood disorder (ICD–10, F30–39) or severe mental illness; the proportion of those with no fixed abode or living alone; the proportion who were unemployed or sick/disabled.

Table 3 Associations between participants' demographic characteristics and psychiatric diagnoses and their scores on the Alcohol Use Disorders IndentificationTest (AUDIT)¹

Characteristic	AUDIT score < 8 (n=103)		AUDIT score ≥ 8 (n=97)				AUDIT score < 16 (n=155)		AUDIT score ≥ 16 (n=45)
	n	%	n	%	χ2	P	n	%	n	%	χ2	P
Ethnicity
British/Irish/Other White	75	47.8	82	52.2	4.07	0.04	116	73.9	41	26.1	5.47	0.02
Other ethnic groups	28	65.1	15	34.9			39	90.7	4	9.3
Living status2
Living alone	44	45.3	53	54.6	2.93	0.09	67	69.1	30	30.9	6.89	0.01
No fixed abode	5	31.2	11	68.8	2.93	0.09	11	68.8	5	31.3	0.74	0.39
Employment status
Unemployed	27	45.8	32	54.2	3.21	0.73	41	69.5	18	30.5	3.08	0.08
Invalid3	25	43.9	32	56.1	1.86	0.17	43	75.4	14	24.6	0.19	0.66
Diagnostic category
Psychosis	40	55.6	32	44.4	0.83	0.36	63	87.5	9	12.5	6.36	0.01
Mood disorder	46	56.8	35	43.2	1.70	0.19	68	84.0	13	16.0	3.13	0.08
All severe mental illness	79	53.8	68	46.3	1.55	0.21	125	85.0	22	15.0	20.15	< 0.001
Smoking status4
Regular smoker	50	40.0	75	60.0	17.05	< 0.001	87	69.6	38	30.4	11.65	0.001
Not regular smoker	52	70.3	22	29.7			67	90.5	7	9.5
Illicit drug use
Lifetime use	44	40.0	66	60.0	13.14	< 0.001	79	71.8	31	28.2	5.09	0.02
Past 30 days	17	32.7	35	67.3	10.31	< 0.001	34	65.4	18	34.6	6.14	0.01
Interference with life5	17	29.8	40	70.2	15.08	< 0.001	35	61.4	22	38.6	12.4	< 0.001
Sleep quality
Sleep regular	44	59.5	30	40.5	2.98	0.08	66	89.2	8	17.8	9.20	0.002
Not regular	59	46.5	67	53.2			89	70.6	37	29.4

Forty-five participants (22.5%) had AUDIT scores of 16 or more, indicating significant alcohol dependence. Again, equal numbers of men and women scored at this level (27 men, 28 women; χ²=1.22, d.f.=1, P=0.27). They were almost exclusively White (British or Irish) rather than from other ethnic groups. Compared with other participants they were more likely to be living alone. They were less likely to be diagnosed with a schizophreniform psychosis or severe mental illness (ICD–10, F20–29) than those with lower scores.

Sleep quality, smoking and use of sedatives and illicit drugs

One hundred and fifty-seven participants (78.5%) had smoked cigarettes at least once in their lives and the majority were current regular smokers (125 participants, 62.5% of entire sample). Of those who stated that they smoked cigarettes regularly, the modal number of cigarettes smoked was 20 per day (range 2–60). Half of these individuals wanted help to enable them to stop smoking (50.3% of regular smokers). However, the majority (98 individuals, 79%) thought that it would be fairly or very difficult to give up.

The majority of the sample (63%) reported an irregular sleep pattern. This was strongly associated with illicit drug use and an AUDIT score of 16 or more. Participants were asked about their use of sedative drugs. These were defined as sedatives, sleeping pills and tranquillisers to calm you down or improve your sleep (e.g. librium, valium, temazepam, barbiturates). About half of all participants (107, 53.5%) had used sedative drugs in the past 30 days. In the majority of cases (83%) these were prescribed. However, 22% of the sample reported that they had used illicit sedatives.

Self-reported illicit drug use is detailed in Table 4. Fifty-seven participants (28.9%) felt that their substance use had interfered with work, social or home life in the previous year. Overall, men were more likely to have used drugs than women. Lifetime drug use was not associated with psychosis in this sample (χ²=0.572, d.f.=1, P=0.351). Individuals reporting lifetime drug use were younger than other participants (Mann–Whitney U test: z=–6.06, P<0.001) and had shorter psychiatric histories (Mann–Whitney U test: z=–1.99, P=0.045).

Table 4 Self-reported illicit drug use

Substance	Lifetime								Past 30 days
	Total		Men		Women				Total		Men		Women
	n	%	n	%	n	%	χ2	P	n	%	n	%	n	%
Cannabis	102	51.5	41	43.2	54	56.8	6.13	0.01	36	18.2	24	23.2	12	1.8
Cocaine	56	28.3	33	31.4	23	24.5	1.34	NS	22	11.1	14	13.3	8	8.5
Stimulants	52	26.3	35	34.0	17	18.1	6.39	0.01	5	2.5	4	3.9	1	1.1
Hallucinogens	52	26.3	36	35.0	16	17.0	8.13	0.004	2	1.0	2	1.9	0	0
Inhalants	38	19.3	26	25.5	12	12.8	5.07	0.04	1	0.5	1	0.9	0	0
Opiates	36	18.2	22	21.4	14	14.9	1.38	NS	20	10.1	12	11.4	8	8.5
Any illicit drug	110	57.9	66	64.1	44	46.8	5.94	0.02	52	27.4	34	33.3	18	19.1
Drug use interfering with life	57	28.9	40	38.8	17	18.1	10.2	< 0.001

Participants who had used drugs in the past 30 days were more likely to be men and were significantly younger than other participants (Mann–Whitney U test: z=–4.15, P<0.001), but did not differ from them in terms of diagnosis. Positive scores on the AUDIT were strongly associated with all measures of illicit drug use. Two-thirds of recent drug users scored 8 or more and one-third scored 16 or more on the AUDIT.

Relationship between alcohol use disorders and suicidality

According to admission case notes, 104 admissions (52.5%) were associated with suicidal ideation or deliberate self-harm (see Table 5). They were younger than other participants (mean of 38.8 compared with 43.9 years; Mann–Whitney U test: z=–2.08, P=0.038) and more likely to be White and unemployed. The AUDIT score was strongly associated with suicidality: 70% of those scoring 8 or more and 86.7% of those scoring 16 or more were so admitted (odds ratios=2.98 and 8.10, respectively). With respect to diagnosis, suicidality was positively associated with mood disorders and negatively with psychosis. Suicidality was associated with illicit drug use in the previous 30 days for the sample as a whole and for women specifically (χ²=5.15, d.f.=1, P=0.02, odds ratio=3.80, 95% CI 1.14–12.74), but not for men (χ²=0.75, d.f.=1, P=0.39).

Table 5 Associations with suicidality

Characteristic	No suicidality (n=94)		Suicidality (n=104)				Odds ratio	95% CI of odds ratio
	n	%	n 1	%	χ2	P
Gender
Male	49	46.7	55	53.3	0.32	NS	0.95	0.54—1.16
Female	45	47.9	48	52.1
Ethnicity
White	64	41.2	91	58.8	11.0	0.001	3.22	1.61—6.67
Other	30	70.0	13	30.0
Living status2
Living alone	42	43.3	54	56.7	1.03	NS	1.34	0.76—2.36
No fixed abode	5	31.3	11	68.8	1.78	NS	2.09	0.70—6.23
Employment
Unemployed	21	35.6	38	64.4	4.76	0.03	2.00	1.06—3.70
Invalid3	23	41.1	33	58.9	1.29	NS	1.40	0.77—2.69
Diagnostic category
Psychosis	50	70.4	21	30.6	13.9	<0.001	0.21	0.11—0.40
Mood disorder	30	37.5	50	62.5	5.71	0.02	2.05	1.13—3.70
All severe mental illness	82	56.6	63	43.4	20.8	<0.001	0.22	0.11—0.46
AUDIT score
≥ 8	32	30.0	63	70.0	13.93	<0.001	2.98	1.66—5.31
≥ 16	6	13.3	37	86.7	24.75	<0.001	8.10	3.23—20.3
Illicit drug use
Lifetime use	46	42.6	62	57.4	1.82	NS	1.48	0.84—2.60
Past 30 days	17	34.0	33	66.0	4.34	0.04	2.02	1.04—3.95
Drug use interfering with life4	17	30.4	39	69.6	8.66	0.003	2.65	1.37—5.12
Smoking status5
Regular smoker	52	42.3	71	57.7	3.20	NS	1.70	0.95—3.04
Not regular	41	55.4	33	44.5
Sleep quality
Regular	47	64.4	26	35.6	13.7	<0.001	0.33	0.18—0.61
Not regular	47	37.6	78	62.4

Logistic regression was carried out to ascertain which factors were independently associated with suicidality in this sample. In addition to gender, variables associated with alcohol use disorders according to AUDIT score or with suicidality at a level of P<0.05 were entered into the logistic regression. Age was entered as a continuous variable. The following were entered as dichotomous variables: audit score (8 or more v. less than 8), gender, employment status (unemployed/other groups), ethnicity (White/other groups), living alone, diagnostic category (severe mental illness/other diagnostic groups), lifetime use of illicit drugs, illicit drug use in the previous 30 days, sleep quality and whether a regular smoker. Significant predictors of suicidality are listed in Table 6. The regression indicates that the AUDIT score, irregular sleep pattern and not having a severe mental illness were significant predictors of suicidality (i.e. once controlled for age, gender and the other variables listed above). Similar results are obtained if the AUDIT score is entered as a continuous variable.

Table 6 Logistic regression: variables associated with suicidality in psychiatric in-patients¹

Variable	Wald statistic	d.f.	P	Exp(B)	95% CI of exp(B)
AUDIT score ≥8 v. <8	10.61	1	0.001	0.333	0.17—0.66
Diagnosis SMI v. others	15.44	1	<0.001	0.190	0.08—0.44
Sleep irregular v. regular	10.22	1	0.001	3.10	1.55—6.20
Constant	7.94	1	0.005	3.70

Comparison with other studies

This study adds to the growing evidence of the utility and acceptability of the AUDIT, even with acutely unwell psychiatric patients, at least when approached by independent clinical researchers. The overall prevalence of alcohol use disorders in our sample is considerably higher than found in the general population in the UK (Reference Singleton, Bumpstead and O'BrienSingleton et al, 2000). Hulse and colleagues (Reference Hulse, Saunders and RoydhouseHulse et al, 2000), in their study of psychiatric in-patients in Australia, reported that 49.1–72.4% of men and 29.2–44% of women were AUDIT positive, depending on the diagnosis (the overall prevalence was not reported).

Like Hulse and colleagues, we found that a smaller percentage of individuals with severe psychiatric illnesses were drinking at a significantly dependent level compared with those having other psychiatric diagnoses. However, our study is unique in finding equally high rates of alcohol use disorders in men and women regardless of diagnosis. The prevalence in women greatly exceeds that reported by Platt & Robinson (Reference Platt and Robinson1991). Women have been described as having higher rates of psychiatric comorbidity than men (Reference Brady and RandallBrady & Randall, 1999). Women experience more physical complications of alcohol use disorders than men, along with accelerated progression of these complications (‘telescoping’) (Reference Randall, Roberts and Del BocaRandall et al, 1999). A similar excess susceptibility to (or telescoping of) the psychiatric complications of alcohol use disorders might account for the high prevalence of alcohol use disorders found in the women of our sample (although other explanations are possible). Our study highlights the need to screen women for alcohol use disorders and to develop effective interventions in psychiatric settings for this group.

This study adds to the literature demonstrating that severity of alcohol misuse is related to suicidality. We are not aware of any other reports in which the AUDIT has been used to examine this relationship. The strength of this association rises with the AUDIT score, and remains when controlled for by confounders such as age, ethnicity, employment status and drug use. We found the degree of association between suicidality and alcohol use disorders striking and of some concern, given the evidence for the poor detection of alcohol use disorders in routine psychiatric practice (Reference Barnaby, Drummond and McCloudBarnaby et al, 2003). In comparison with other studies we did not find that regular smoking was associated with suicidality.

The significant association between reported poor sleep and suicidality replicates previous observations (Reference Tanskanen, Tuoilehto and ViinamakiTanskanen et al, 2001) and has significant implications for delivering effective interventions aimed at changing the individual's drinking behaviour. Drinking may be described by patients as a helpful way of coping with insomnia, especially when this is associated with other psychiatric symptoms. Our results would indicate that therapists should positively elicit such beliefs, examine alternative coping strategies for insomnia and raise the issue that hazardous drinking may actually exacerbate or maintain insomnia and increase the risk of subsequent suicidal or self-harming behaviour. Similarly there is an obligation to elicit from subjects any incidence of suicidal and self-harming thoughts or behaviours, the relationship of these to drinking episodes and potential risk-minimising strategies. Such therapeutic strategies call for sensitive training of study therapists and care in the running of trial interventions so that appropriate responses to suicidality in trial participants are in place.

Limitations of the study

The use of the AUDIT has been recommended for use in the psychiatric population, both in individuals presenting with self-harm (Reference Babor, Higgins-Biddle and SaundersBabor et al, 2001) and in those with psychotic illness (Reference Hulse, Saunders and RoydhouseHulse et al, 2000). However, the AUDIT has been criticised for lacking sensitivity compared with structured diagnostic interviews for alcohol use disorders (Reference Hearne, Connolly and SheehanHearne et al, 2002). We employed the same cut-off score of 8 for men and women, although it has been argued that a lower cut-off score is appropriate for women (Reference Reinert and AllenReinert & Allen, 2002). Some experts (Reference Hodgson, Abbasi and HodgsonHodgson et al, 2003) recently have recommended the Fast Alcohol Screening Test (FAST) to screen for alcohol use disorders in settings where time is limited, but this tool has yet to be validated as thoroughly as the AUDIT and we wished to compare our results with population means and results from studies in similar psychiatric settings.

There are some sources of potential bias within this study. In our sample, 17.9% of potential participants declined to take part and 8% were excluded because they were unable to complete the questionnaire in English. A written questionnaire may be less useful in multi-ethnic communities in which significant numbers of individuals have difficulty reading and communicating in English. The eventual sample size of 200 was relatively small, and we agree with Hulse and colleagues that it may be difficult to screen all psychiatric admissions for alcohol use disorders because of the short duration of many admissions to psychiatric units. Nevertheless, the sample was recruited from a large catchment area including inner-city as well as suburban areas and it is likely to be representative of general psychiatric practice in the UK. That the majority of participants did not have lengthy psychiatric histories strengthens the rationale for carrying out controlled trials of screening and brief alcohol interventions.

Concerns have been expressed about response bias to instruments designed to assess stigmatising conditions such as alcohol misuse (Reference Beich, Gannik and MalterudBeich et al, 2002). In an attempt to reduce this, we embedded the AUDIT within a longer ‘health and lifestyle questionnaire’. The AUDIT has been incorporated into a longer questionnaire without compromising the validity or reliability of the AUDIT itself (Reference Daeppen, Yersin and LandryDaeppen et al, 2000). In order to minimise the number of groups for analysis, we chose a broad definition of suicidality and may have included a larger group of individuals than those included under other definitions of parasuicide (Reference Hawton and van HeeringenHawton & van Heeringen, 2000: 51). As a result, we may have combined individuals with a variable level of risk of further self-harm and suicide. For the same reasons the questionnaire did not ask participants about behaviour that would have indicated their suicide risk prior to admission or at the time of assessment.

Although effort was made to make ICD–10 diagnoses from entries in the case notes, the study team were not blind to the responses obtained in the questionnaire, leading to the possibility of coding bias. Case notes may not precisely reflect the actual actions of mental health staff. The behaviour of the clinical staff caring for patients in hospital may have been influenced by the study taking place. However, data were collected by an independent study team, and clinical staff were not aware of the specific hypotheses under investigation.

Implications for clinical practice and further research

Alcohol misuse is implicated in a substantial proportion of suicides (Reference FosterFoster, 2001). Recently, mental health policies in Britain have been directed towards reducing suicide rates, although these have been focused on individuals with severe and persistent mental illness rather than alcohol use disorders. Differing explanations have been offered to account for the relationship between alcohol use disorders and self-harm or completed suicide, involving both social and neurobiological mechanisms: the expression of suicidality may be one of the routes by which individuals seek help for alcohol and substance use disorders; and alcohol may decrease protection against serotonin-mediated suicide impulses (Reference LovingerLovinger, 1997). The latter mechanism might mitigate the potential for psychological interventions alone (e.g. brief motivational enhancement therapy) to effect significant change in psychiatric patients with alcohol use disorders. Because medical staff may miss the opportunity to screen a proportion of psychiatric in-patients owing to the short duration of many psychiatric admissions, screening in community psychiatric settings should be explored further. The advantages of the AUDIT are that it does not require lengthy staff training and is quick to complete. The brevity and face validity of the AUDIT make it suitable for inclusion in assessments conducted by non-medical staff in community settings (Reference Daeppen, Yersin and LandryDaeppen et al, 2000).

Our findings can be used to generate hypotheses regarding the variable efficacy of brief interventions for alcohol use disorders in controlled trials carried out on the psychiatric population to date (Reference Baker, Lewin and ReichterBaker et al, 2002; Reference Hulse and TaitHulse & Tait, 2002). Gender mix, demographic and genetic factors, duration and nature of psychiatric problems, including suicidality and concomitant drug use (including prescription sedative drug use), may also influence outcome. We encourage researchers in this area to describe in detail their sample and the training provided to study therapists, especially responses to insomnia and suicide risk.

We conclude that the AUDIT is useful in the psychiatric population, especially in those presenting with suicidality. It could be included in assessments of suicide risk. Further research is warranted to examine this relationship in larger samples studied prospectively and in other settings (e.g. in community and liaison psychiatric settings). Interventions for alcohol use disorders in psychiatric populations should aim to reduce suicidal ideation and deliberate self-harm. Priority in the meantime should be given to screening those presenting with deliberate self-harm and suicidal ideation for alcohol use disorders.

CLINICAL IMPLICATIONS

1. ▪ The Alcohol Use Disorders Identification Test score is strongly associated with suicidality.

2. ▪ Screening for alcohol use disorders in psychiatric settings should be combined with detailed assessment of suicide risk and sleep disturbance.

3. ▪ Interventions for alcohol use disorders should monitor and aim to reduce the incidence of suicidal ideation and actual self-harm, in addition to addressing problem drinking.

LIMITATIONS

1. ▪ This was a small study carried out in only two hospitals in one London region.

2. ▪ Written questionnaires may exclude those for whom English is not their first language.

3. ▪ A broad definition of suicidality was used, grouping individuals together who may have differed in terms of risk.