The main positive finding of the featured review of brief alcohol interventions among hospital inpatients was a lower level of drinking six months after the intervention. Analysis of the four studies which summed to this finding suggests that in important respects they cannot be relied on as indicative of the impact of brief interventions and that therefore the combined finding too is questionable.
There is doubt over whether the data included in the review from two of these studies actually tested what is normally thought of as a brief intervention. One of them (1), conducted in Finland, recorded an unusually large impact when patients seriously injured (we can fairly assume) as a result of their heavy drinking were repeatedly counselled by a nurse and their doctors. Omitting this study meant the combined findings were no longer statistically significant. Another (4) which also contributed Based on the original paper. The relevant table (1.1) in the featured review seems unfinished and does not record the figures. substantial drinking reductions involved a home visit and was vulnerable to bias in its findings. In a third study (2), whether the brief intervention was responsible for drinking reductions was called in to question by the fact that simply handing patients an advice booklet was equally effective. In another (3), a positive contribution to the finding that brief interventions were effective would have been negative had another brief intervention variant tested in the study been selected for the analysis. In two of the studies (2 3) the degree of loss to follow-up or its imbalance across intervention and control groups raise concerns over the reliability of the findings or their generalisability to heavy drinking patients in general. Study-by-study details below.
1 Of the studies, by far the greatest effect was recorded by a study in Finland. Salient features of this study are that the patients: were fairly seriously injured; They had been admitted to the Department of Orthopaedics and Traumatology at the Helsinki University Central Hospital for at least 24 hours. in the week before their injury they had on average been drinking very heavily The brief intervention group had drunk on average 740g or 93 UK units, and most of the entire sample usually drank no more than three times a week. This implies a typical intake on a drinking day in the past week of at least 31 UK units. and usually in a 'binge' pattern; and they were counselled by a nurse while on the ward and then again a month later at an outpatient check-up by the same nurse, and usually several times by a doctor. The extent of the intervention cast doubt over whether it really could be termed 'brief', leading the reviewers to recalculate the figures excluding this study; the result was no statistically significant advantage for the remaining interventions. Beyond these features of the intervention and its context, is the methodological flaw that six months later those who had continued to drink heavily faced the choice of whether to admit to the same nurse how much they had ignored their (and the doctors') advice. Extrapolating only slightly from these facts, it seems likely that many of the patients had been injured as a result of fairly extreme drinking, and had this clear example of the dangers of their drinking repeatedly brought to their attention by medical staff, a potentially powerful intervention, especially when one of those staff was to check up on them six months later.
2 A second study was conducted in Scotland. The follow-up alcohol consumption amounts used in the featured review do not seem to be recorded in the study, which instead reported reductions in the amount drunk from the baseline assessment to six months later. These showed a greater reduction of about 10 UK units of alcohol in the drinking of patients allocated to a brief intervention from a mental health nurse, compared to patients who (apart from the study's baseline screening and assessment) were left to the wards' usual care. Patients whose admission was alcohol-related were excluded from the study, none of the wards were injury-specific, and drinking levels, though excessive, were much lower than in Finland. Impacts were less than in Finland, perhaps because patients had less reason to cut down, the admission was unlikely to be directly and clearly related to recent drinking, and the brief intervention was briefer and not reinforced by medical staff responsible for their care. However, impacts remained substantial. But here too there is doubt over whether a brief intervention was the cause, because almost the same extra reduction in drinking was seen among another group of patients who were simply handed a written guide to sensible drinking by the same nurse. The most likely explanation is that being identified as a risky drinker and professionally advised (as the offer of the booklet would probably have been interpreted) to consider cutting down, was sufficient to trigger such drinking reductions as there were going to be, and this was not improved on by more extended intervention. There was also the methodological concern that proportionately twice as many of the control group (29% v. 14%) could not be followed up. The study and the featured review say that an 'intention to treat' analysis was conducted, normally taken to mean that everyone who started the study was included in some way in the final results. In fact this was not the case and the results reflected only those who could be followed up. However, as study authors pointed out, on the assumption that drinkers who could not be retraced had done relatively poorly, the imbalance in the proportions followed up would have worked against the intervention.
3 The third study set in Australia reported a statistically significant extra reduction in alcohol consumption after brief counselling compared to merely assessing the patient; about six months later, a typical patient who had been counselled drank 13 UK units less a week than patients who had not been counselled. In this study, clearly the intervention was a typical brief intervention in style and duration. Here the main concern is the number of heavy drinkers who refused to join or were lost to the study: of the 273 identified heavy drinkers, 174 joined the study and 123 were able to be followed up. There can be no assurance that the minority of patients who contributed outcome data to the study were representative of all risky drinkers on the wards. Also, there were two intervention groups. The review chose to base its analysis on the one given a brief motivational interview, who six months later were drinking about 4 UK (31g alcohol) units less a week than control patients. Had they instead chosen the group allocated to skills-based counselling, drinking amounts at six months (thought not reductions from baseline) would have actually favoured the control group. In the groups selected for the review, of the control patients who started the study nearly a third could not be followed up and of the intervention patients a quarter. No assumptions were made about their drinking and the figures used in the review were based only on the followed up sample, figures which do not conform to the convention that everyone who started the study should be included in some way in the final results (an 'intention to treat' analysis) in order for the analysis to more adequately represent the impact of the intervention across the entire sample.
4 The last study conducted at a hospital in Manchester compared the later drinking of heavy-drinking patients who were simply screened versus two subsequent cohorts admitted to the same general medical wards who were both screened and counselled about their drinking. Screening on its own led to virtually no change in drinking but both counselling interventions (one session on the ward or that plus one later at the patient's home) led to substantial reductions. Unfortunately the screening procedures for the three cohorts differed, raising the issue of whether different types of patients were identified (though on the measured variables they were comparable), and the follow-up assessor probably knew whether the patient had been counselled or not, opening up the possibility of bias in the findings. One of the two interventions – and the one chosen to be represented in the featured review – involved a home visit so might not be considered truly a 'brief' intervention. However, the extra drinking reductions were about the same regardless of this extra intervention phase.
Where these studies have been cited above the same number is used. Each is considered in detail below. In summary, of the six studies known to Findings, one (4) involving mainly very heavily drinking women (the only study where most patients were women) found substantially greater drinking reductions after counselling than after screening only. Methodological issues cast some doubt over the validity of the of the findings or the implication that they represent an effect of brief intervention, perhaps the greatest being that during intervention phases patients were being treated by nurses newly trained in dealing with drink problems. Conceivably it was this rather than the brief interventions which led to the reductions in drinking. In another (6), one of six measures of drinking and related problems was reduced significantly more after brief advice than after screening only – a small extra reduction in average daily alcohol intake during a typical month. If anything the reduction was slightly greater after just five minutes of advice than after more extended counselling. Substantial loss to the study of positive-screen patients and the fact that some were from primary care mean the results may be an unreliable indication of the impact of brief advice on hospital patients as a whole, and had adjustment been made for multiple tests (allowing that is for the greater risk that one would reach significance purely by chance), the one in six which was found significant would no longer have been.
The remaining four studies found no significant extra reductions in consumption relative to just screening the patients and normal care on the ward (5 7 8), or in one study also handing them an alcohol advice booklet (2). Related problems were also unaffected, with the exception of study 5 which used an unvalidated measure and was vulnerable to 'regression to the mean', and in study 8 an isolated finding that more intervention patients felt their alcohol-related ill-health had improved, which might have failed to be significant if the study had adjusted for its multiple tests.
2 As previously analysed by Findings, over six months a study in Scotland published in 2007 recruited 215 adult patients from among the 2307 admitted as inpatients to 16 wards in a general hospital. Steps were taken Though there was no upper limit on how much patients could drink before being excluded from the study, none of the wards dealt specifically with substance dependence and (along with other criteria) patients with a history of drug or alcohol dependence or mental illness, or who had been admitted primarily for treatment of alcohol-related complaints, were excluded from the study. to exclude patients known to have serious drinking problems. Among the 819 not excluded for these or other reasons, screening tests identified 215 who had drunk excessively Operationalised as over 21 UK units for men and 14 for women, 168g and 112g alcohol respectively. over the past week. In two-week blocks (to reduce 'cross-contamination' between patients), they were allocated to one of three alcohol advice options.
About a third (the control group) were left to the wards' usual care. Another third met a mental health nurse who handed them a written guide Health Education Authority. That's the limit: a guide to sensible drinking. London: Health Education Authority; 1994. to sensible drinking. The same nurse engaged the remaining third in a roughly 20-minute discussion intended to bolster confidence in their abilities to control drinking and to lead them to the point where they set their own change goals.
Six months later 172 of the 215 were reinterviewed. Typically men in their 40s, before being admitted half had drunk at least 35 UK units 280g alcohol. of alcohol in the past week. Those allocated to either intervention had on average cut their weekly drinking by 14 or 15 UK units. 112–120g alcohol. Compared to assessment and normal care, both interventions had led to a further reduction of 10 units 80g alcohol. a week, a statistically significant difference highly unlikely to have occurred by chance. However, the interventions still left the patients drinking heavily; on average they still consumed perhaps Based on the average consumption in the full initial sample minus the reduction seen in those retained in the study in the intervention groups. about 30 units a week.
There are some concerns over the reliability of the findings, but none threaten the conclusion that the interventions led to equivalent drinking reductions relative to screening, research assessment and normal care only.
Accepting this, a key question becomes why such a well structured brief intervention, delivered by an apparently highly skilled interventionist, to patients in the relatively conducive (the very ill were excluded) environment of an inpatient ward, had no greater effect than handing them an alcohol advice booklet.
Since the same interventionist handed over the booklet, one possibility is that this entailed some discussion which, though presumably shorter, drew on the same skills and content as the longer intervention. Another is that this mistakenly targeted confidence in ability to cut down when this was not the decisive obstacle, Both active interventions led to increased confidence, particularly the discussion specifically designed to enhance confidence. Yet this (not quite statistically significant) extra boost in confidence had no effect on drinking itself, suggesting perhaps that ability to control drinking was not the main issue, rather, the resolve to do so. If this was the case, then an intervention more thoroughly focused on enhancing motivation might have had a greater impact. or perhaps muddied the water by asking patients to rehearse what for them were the benefits of drinking and by not giving clear advice to cut down.
Given other studies (detailed here), perhaps the most likely explanation is that being identified as a risky drinker and professionally advised (as the offer of the booklet would probably have been interpreted) to consider cutting down, was sufficient to trigger such drinking reductions as there were going to be.
Methodological concerns include the fact that proportionately twice as many of the control group (29% v. 14%) could not be followed up. The study say an 'intention to treat' analysis was conducted, normally taken to mean that everyone who started the study was included in some way in the final results. In fact this was not the case and the results reflected only those who could be followed up. However, as study authors pointed out, on the assumption that drinkers who could not be retraced had done relatively poorly, the imbalance in the proportions followed up would have worked against the intervention.
4 Published in 2003, a study at a hospital in Manchester compared the later drinking of heavy-drinking patients who were simply screened versus two subsequent cohorts admitted to the same general medical wards who were both screened and counselled about their drinking.
Men were included in the study if they drank more than 50 UK units 400g alcohol. a week and women 35 UK units. 280g alcohol. In the first phase of the study such patients were simply identified by the hospital's alcohol counsellor who interviewed every patient. Over the same period the counsellor trained ward nurses in making the same assessment and in understanding and responding to drinking problems.
In the next phase screening was done either by the ward nurses or the alcohol counsellor, who ensured all patients were screened. Positive screen patients received one session of counselling from the alcohol counsellor during their stay on the medical ward. Of this we are told little except that it was based on the stages of change model made famous by Prochaska and DiClemente. Patients were left with a specially prepared booklet to reinforce the counselling based on that used in study 5 which covered ways of coping with craving, cutting down drinking, the stages of change model, and recommended safe drinking limits.
In the final phase of the study all the screening was done by ward nurses as a routine part of the admission procedure. Positive screen patients were referred for counselling to the alcohol counsellor who offered two sessions of counselling, the first session during their hospital stay and the second a month later at the patient's home. This followed the same model as the single session in previous phase of the study.
In the first two phases all patients eligible for the study (80 and 45 respectively) agreed to join it but in the third (when another 45 patients were recruited) it is not known whether all admissions were screened, whether all positive screen patients were referred for counselling, and how many refused the offer. Of the 170 patients in the study, 10 died before they could be reassessed (eight of whom had not been counselled, nearly three times the death rate among counselled patients) and of the remaining 160, 133 were reassessed at home six months later by a research assistant. Nine in 10 were women and before the study they had been drinking on average 100 UK units 800g alcohol. and the typical patient was drinking around 70 UK units, 560g alcohol. very heavy drinking, especially for women.
Screening on its own led to virtually no change in drinking but both counselling interventions led to substantial and similar reductions. Expressed as medians – the amount exceeded by half the patients – there was 63–68% reduction in the counselled groups compared to 7% in the non-intervention group. Reductions in both counselled groups were statistically significant.
Differences in the screening procedures for the three cohorts raise the issue of whether a different selection of patients were identified. However, the authors explain that the same number of patients were recruited in phases two and three over a similar period, suggesting that the nurses were screening effectively, and on the measured variables the three groups were comparable. The follow-up assessor probably knew whether the patient had been counselled or not, opening up the possibility of bias in the findings. One of the two interventions – and the one chosen to be represented in the featured review – involved a home visit so might not be considered truly a 'brief' intervention. However, the extra drinking reductions were about the same regardless of this extra intervention phase. The choice of medians to reflect drinking levels is unusual and vulnerable to distortion. For example, if only the bottom 40% of patients reduced their drinking the median would be unchanged. However, the range statistics also given suggest that the medians reflected overall drinking reductions. Perhaps the biggest threat to the validity of the implications drawn from this study is that during the intervention phases patients were being treated by nurses newly trained in dealing with drink problems. Conceivably it was this rather than the brief interventions which led to the reductions in drinking.
5 Published in 1985, a Scottish study previously analysed by Findings screened 731 admissions to male medical wards in a hospital in Edinburgh, a procedure which included questions about alcohol intake and related problems. 161 problem drinkers met the criteria for inclusion. All but five agreed to join the study. The person conducting the intervention was not responsible for medical care on the ward.
Allocation was alternated by ward to either a control group who received no further intervention or to one hour of counselling. The study included patients whose problems may have dated back two years, but an average consumption of 10 units a day in the past week suggests current heavy drinking was common. There was no upper limit to drinking levels, but other criteria would have tended to exclude the least socially stable, isolated patients, and those so obviously in need of psychiatric help that a referral had already been made.
The result was two fairly evenly matched groups of whom 133 of the original 156 were re-interviewed a year later. One of the study's strengths is the low attrition rate, raising confidence that any benefits would generalise to the hospital's male patients as a whole and possibly to other hospitals. Low drop out was achieved partly by the seamless provision of screening, assessment and intervention by the same experienced nurse, virtually ensuring that all those eligible and agreeable would complete the intervention, though it did mean that the initial assessment In contrast, the follow-up interviewer did not know whether the subject had been counselled or not. was conducted by someone who knew what would follow.
Counselling was aimed mainly at achieving problem-free drinking. Though not identified as motivational in nature, the session deployed some elements used in motivational interviewing, including encouraging the patient to weigh up the drawbacks of their drinking and to make a commitment to changing it, and the interventionist was an alcohol specialist nurse known for her empathic style. The nurse left the patient with a specially prepared booklet to reinforce the counselling.
A year later across both this group and the normal care control group, average past-week alcohol consumption had fallen to half pre-intervention levels, in both cases a statistically highly significant reduction. However, in this respect there were negligible differences between the groups. Though the number of individuals who halved their consumption was higher in the intervention group (64% v. 48%), this narrowly missed reaching the conventional criterion of statistical significance.
Even if counselling did not further reduce consumption, it was followed by significantly greater reductions in alcohol-related problems (medical, social and dependence) and in levels of a chemical in the blood indicative of excessive drinking. However, the extra reduction in problems took the form of the (despite randomisation) higher pre-intervention level in the intervention group falling back to the same level as the control group, possibly a finding contaminated by the common tendency for atypical readings of this kind to revert to more normal figures over time, and the scale used to measure problems does not appear to have been a validated scale with proven reliability. Also, the blood tests used in the study are only loosely related to the amount drunk. There were also greater improvements in a composite characterisation of the proportion "definitely improved". Readmissions and (more so) lengthy readmissions (four people versus 11 in the control group) during the year were lower in the counselled group, but the report did not specify whether the differences were statistically significant. Given the small numbers, it is unlikely that they were.
In summary counselling probably did foster further remission in drink-related problems, but the study was unable to prove this convincingly, and there was no impact on the most sensitive measure of consumption – the patients' own accounts. Another issue is why consumption fell so steeply after just a brief assessment and no counselling. Perhaps before entering hospital these men were at an atypical peak in their alcohol consumption and later simply resumed more normal drinking. Maybe too (as other researchers suspect happens; 1 2) the focus on drinking in the assessment and in their agreeing to enter a study of "health and drinking habits" provoked some salutary reflections. An further explanation is the very human tendency to behave differently under observation: the patients knew there would be a follow-up interview and roughly when.
6 The British arm of a WHO study of brief alcohol interventions which started in 1982 foundered on the reluctance of GPs and patients to get involved. Only three health centres participated, forcing the study to seek further subjects in the medical wards of two general hospitals in Cardiff where implementation was easier. 961 patients were screened and 110 were recruited to the study and followed up, all men. Suspected dependent drinkers were excluded. Nearly half the excessive drinkers identified by the researchers refused to take part in the study and a third who did join could not be followed up.
Over two-thirds of the entirely male follow-up sample re-assessed about eight months later had been recruited in the hospitals, comprising 76 of the 110 patients. The results probably largely reflect their progress, but the results were not disaggregated so it is impossible to be sure. Even if they had been, the loss to follow-up of a third of the patients would cast doubt on the findings.
The patients were randomly assigned to either no intervention other than screening and research assessment, to five minutes of advice, or to this plus a further 15 minutes of counselling. Both advice groups were given written advice and the longer intervention was based on a self-help manual also given to the patient.
The sole statistically significant finding was a greater reduction in average daily alcohol intake during a typical month in the advice groups relative to the controls. Judging by the differences in average amounts drunk, this greater reduction in drinking was most apparent in the group given the shorter of the two interventions, five minutes of advice that their drinking might place them at risk. Before intervention the patients were drinking on average 8.5 centilitres of alcohol daily. The extra reduction relative to controls amounted to 2.54 centilitres of alcohol, the equivalent of two and half UK units or 20g alcohol. Overall the reduction amounted to 3.2 centilitres.
Supplementing brief advice with 15 minutes of counselling did not improve outcomes. On their own neither of the components of amount drunk (amount drunk on a drinking day and the number of these days) fell significantly more after advice, nor did related problems or levels of dependence. In both hospital and primary care settings, some patients did not see their drinking as excessive, and in practice nurses had to negotiate the intervention with the patient rather than simply delivering it.
7 A further UK study published in 1999 recruited inpatients admitted to four wards in a large Scottish general hospital. In total 1221 were admitted during the study, 998 interviewed, 225 admitted drinking more than the UK recommendations of the time (21 UK units for men, 14 for women), 153 had not previously been treated for alcohol problems and drank regularly rather than sporadically, and 150 agreed to join the study, about three quarters of whom were men. Typically the patients drank 33 UK units 264g alcohol. a week and (to judge from the problems they reported) perhaps a third were dependent.
In four-weekly cycles the wards implemented one of three interventions or none at all (the control condition) except for the screening and research assessments applied to all the patients. One intervention consisted simply of the researcher handing patients an official guide to sensible drinking; for another the same researcher advised the patient for 10–15 minutes using feedback from the assessment; and in the third she did both.
A year later another researcher unaware of the allocation re-assessed two-thirds of the patients, data from whom was included in the analyses. All the groups including the assessment-only control group (who cut their drinking by on average 15 UK units 120g alcohol. ) had substantially curbed their drinking, though to what degree this was an artefact due to the loss of patients to follow-up is unclear. Only the average reduction in the patients given advice plus booklet exceeded that in the control group, but then only by six UK units a week, and this extra reduction was not statistically significant. Neither were the differences in reductions in alcohol-related problems or levels of chemicals in the blood indicative of heavy drinking.
The authors concluded that none of the interventions was more effective than screening only. They speculated that the detailed inquiry into drinking of the screening procedure acted in itself as an intervention, the effect of which was to raise awareness of potentially harmful levels of alcohol consumption. Talking in detail with patients about the amount they drink when they are unwell and perhaps more sensitive to issues which relate to their health, seems, the authors thought, sufficient to encourage them to consider the implications of their drinking on their health, within an environment which provides more opportunity for reflection than when people are busy conducting their everyday lives.
8 In York 428 patients admitted to nine medical and orthopaedic wards of a general hospital from 1985 to 1987 were screened by ward staff and a researcher and identified as drinking at unsafe levels. According to a monthly timetable, patients were allocated to be taken by ward nurses to view an alcohol education audio-visual presentation and receive the booklet on which it was based, or to the ward's normal care, with the sole addition that ward consultants were informed about each patient's excessive drinking. On average patients in the study drank over 40 UK units 320g alcohol. a week.
Of the 411, 361 patients were re-interviewed a year later and 333 provided data, overwhelmingly men. At follow-up, about a third (31%) of the patients shown the presentation reported a reduction in their alcohol-related health problems, significantly more than in the control group (22%). However, there were no significant extra reductions in three other problem areas nor in average consumption levels, which had fallen modestly by just over a UK unit a day (this data was available from just 214 patients), and nor had more patients cut their consumption after the presentation than after screening only. At follow-up three quarters of both groups were still drinking at unsafe levels according to the study's initial criteria.
Of the 205 control patients, 42 recalled being advised about their drinking in the course of normal care while in hospital. They tended to be the heavier and more problematic drinkers. They had made greater improvements on average in their problem scores and proportion drinking to unsafe levels than the patients shown the presentation but, not significantly greater.
The authors concluded that the audio-visual presentation had no lasting effect on drinking. However, they speculated, the screening assessments may have done. Regarding the assessment interview, the researchers comment that "Certainly, those patients who felt affected by the interview reduced their mean consumption ... to significantly lower levels than those who felt unaffected by the interview".
Comment on these background notes Return to main entry