It is important to understand the implications of the substantial variation between the outcomes of the trials, and the decision Implied in the use of a random effects model for the analysis. to treat each as having its own impact due the features of that particular trial, rather than reflecting only chance variation from a generalisable impact of brief interventions. This means that anyone implementing a brief interventions programme cannot rely on these results as predictive of the outcomes they can expect. It will all depend on the nature of the caseload, context and interventions, among other factors. They cannot even be sure of having a positive impact, because it is unlikely that the studies included in the analysis are representative of all the studies which might have been done; presumably funders, researchers and clinicians had reasons for choosing these sites, caseloads and interventions, that went beyond sticking a pin in at random. In this situation, it will not be enough to set up a programme and assume that risk-reduction is being delivered; each programme will have to demonstrate this for itself.
There was a non-significant hint that longer interventions further reduced drinking (51g versus 23g for the shorter interventions). Though on average smaller, impacts from the shorter interventions were consistent across the studies, suggesting that the benefits are more predictable. However, the analysis of these trials also assumed each to have its own characteristic effect.
The verdict of generalisability to routine practice rests largely on the finding that impacts in the more real-world trials did not significantly differ from those of the more tightly controlled trials. But the analysis assessed the impact of different degrees of the rigour and control characteristic of efficacy trials, not whether the interventions remained effective in truly real-world conditions. Rather than setting a standard for what counted as a real-world trial, the analysts scored the trials on a continuum from highly controlled (0) to highly real-world (12), and then split the studies in to the top and bottom halves. Because scores were bunched towards the real-world end of the dimension, the cut off point was quite high – 9 out of 12 – suggesting that most of the trials were at least fairly clinically relevant according to the criteria used in the analysis. These included whether the interventions were delivered in typical practices, by practitioners for whom primary care is their main occupation, to routinely presenting patients, and in normal consultation times after the kind of preparatory training and support most practitioners are likely to be able to access.
Nevertheless, though not statistically significant, there was a difference favouring the less real-world trials (a reduction in alcohol consumptions of 45g a week compared to 33g). Most importantly, however, the degree of 'real-worldness' applied only to the brief intervention phase of the trial. Before this came the selection of sites and of patients at those sites willing to participate in the trials, and the crucial screening process without which the brief interventions could not have targeted appropriate patients, and which often supplied data for use in the interventions. Putting the whole procedure in to the frame, few if any can be considered to have been truly conducted in real-world conditions. The highest scoring, most real-world trial, was only able to recruit a quarter of the practices it approached (many said they had no time) and just over 1 in 10 contributed data to the analysis. The results cannot be assumed to be representative of what would happen in a normal practice less motivated or less well placed to get involved in and complete a brief intervention trial. Similarly for the next most realistic trial, which selected practices from among those affiliated to a special health promotion network, was able to gain data from under a quarter of the doctors it recruited, and was able to follow up only half the patients they saw. As the authors observed, only the most motivated practitioners contributed data to the study. Next was a US study conducted not at usual primary care practices, but at those affiliated to an academic medical centre. Screening and initial assessment were conducted by research assistants, half the time by phone rather than at the practices, and the results were used by the doctors when they conducted the interventions. The next ranking English-language report used research assistants to conduct screening. Next on the real world scale was a study which was able to recruit less than half the patients screened as eligible for intervention; most of the rest refused to participate in the study. The next study selected its practices on the basis of their expressed interest in alcohol research, screening and assessment was conducted by a researcher, the results were used in the intervention, and around a third of patients could not be followed up. In the last study, research assistants helped with screening, the results were used by the doctors conducting the interventions, just 31% of the identified heavy drinkers were included in the study and followed up, and one of the interventions was quite extensive and sophisticated; the other actually led to a smaller average drop in drinking among the men in the trial than no intervention.
Once patients were in the trials, further whittling down was known to have occurred or may have done, further reducing confidence in the applicability of the findings to all patients. In many cases it was unclear whether all the patients intended to receive or not receive the interventions were included the study, or whether these derived from selected subsamples. Also, about a quarter of patients who were included in the trials could not be followed up, and few trials tried to estimate what their outcomes might have been. The analysts believed that the latter problem was unlikely to have been great enough to invalidate their conclusions.
A related issue is that the studies and the featured analysis started at the point where patients were randomised to a brief intervention. However, the great majority of patients who might benefit never reach this point. In turn this means that even if brief intervention does work, it is unlikely to make the hoped-for health difference at the level of the population as a whole, the public health rationale behind the programmes. Most practices refuse screening or fail to implement it and when they do, it is rarely applied to more than a small minority of patients. Initiatives like practice visits and training, especially when combined with ongoing support, do modestly improve screening rate and intervention rates, but these remain low. As reported recently in Denmark, generally nurses and doctors are prepared to screen (if at all) only when this emerges naturally in the course of addressing the patient's complaint or because it is a logical component of a procedure applied to all patients in certain categories, such as those undergoing general health checks, new patients, and patients being monitored for chronic conditions which might be related to or aggravated by drinking.
Even when patients are screened, large numbers never reach the point of intervention because they do not score as risky drinkers, are unavailable, excluded by research criteria, or fail to participate, leading to an estimate that on average 1000 patients have to be screened to gain 12 months later just two or three who have stopped drinking excessively. The analysis which reached this conclusion had to rely on studies primarily designed to test the effectiveness of brief interventions, and which in the service of that objective deliberately excluded certain patients, required informed consent on the part of those participating in the trial, and sometimes divorced screening from brief intervention. Outside a research context more practices would implement screening and more patients might get through to the point of receiving a brief intervention.
The concerns apply no less to trials in Britain, where at least two studies have found that primary care brief intervention did reduce drinking. They demonstrated the approach's potential, but not necessarily that it would work in typical practices which themselves identified patients for intervention, and with patients not subject to the multiple selection gateways applied by the studies.
Perhaps importantly, in both patients were selected explicitly on the basis of excessive consumption and either no ceiling or a very high one was set before they were excluded. The result was a sample of on average clearly excessive drinkers (the men averaged over 60 UK units or 480g of alcohol a week). Most would have been towards the far end of the national distribution against which their drinking was compared during the interventions.
Other British effectiveness studies (see background notes to an earlier Findings analysis for citations) were either not reflective of primary care or inconclusive about the benefits of intervention. Feedback from staff and the sometimes very low rates of screening and intervention suggested lack of enthusiasm and/or of resources (such as skills, time and organisational support) for screening and intervention, but this may have been partly due to the burden of the associated research. Further UK studies have minimised this burden and found consequent increases in screening rates, but even in willing practices offered training and ongoing support, the results confirm that attempts at universal screening (and in respect of nurses, opportunistic screening too) result in only a small fraction of risky drinkers being advised about their drinking.
Most pertinent Findings is grateful to Professor Eileen Kaner of Newcastle University for this insight. are two English studies which asked practices to implement the same package either in the context of a brief intervention trial, or in the context of a trial geared to testing implementation processes which did not require the informed consent of patients or set exclusion criteria. Many more patients were screened in the latter, but still these amounted to just 2% of the patients seen by the nurses.
This was also the proportion of the adult Scottish population who attended primary care consultations for alcohol misuse in 2006/07, an indication of the extent of such activity entirely outside a research context. In England, patient records from 34 practices across Gateshead Primary Care Trust revealed that the drinking of certain categories of patients was very commonly documented; this was the case, Personal communication from Professor Eileen Kaner of Newcastle University, June 2009. for example, for 85% of diabetes patients and 72% with coronary heart disease.
In England selective screening and typical and promising intervention approaches have been codified in a protocol called How much is too much?, recommended in English guidelines for commissioning such work from GPs as an enhanced service. 'Enhanced' status means GPs are not required to undertake this work unless they have agreed to do so under contract to their local health authority, and authorities are not required to ensure its provision in their areas. As feared by England's national alcohol charity, in 2008 an audit of health service provision found that systematic screening by GPs was the exception and few patients were screened or offered brief advice, undermining the hoped-for public health benefits of a mass programme. The same year the system was reinforced by a new requirement for health commissioners to organise for GPs in their areas to screen adult patients newly registering with the practice, incentivised by a payment for each screening. These initiatives follow the commitment to selective screening and brief intervention in the 2004 English national alcohol strategy and resultant practice guidelines.
Scotland has similar practice recommendations and a national policy which prioritises screening and brief intervention in primary care, antenatal care, and accident and emergency departments, backed by a health service target for 2008/09–2010/11 to deliver 149,449 brief interventions supported (as in England) by dedicated funding. Set in the context of what was in any event 111,200 primary care consultations for alcohol misuse in a single year in 2006/07, this target of around 50,000 a year across all three priority areas does not seem ambitious.
In Scotland and in England, funding has been provided to train new and current medical staff in screening and brief intervention and other aspects of the management of alcohol problems.