According to the World Health Organization, there are an estimated two billion alcohol users, 1.3 billion smokers and 185 million users of illicit drugs in the world. In 2000, these substances contributed to 12.4% of deaths worldwide [1]. While illicit drug use constitutes the smallest proportion of psychoactive substance use, it is the category often perceived as most harmful, and consequently, significant resources are employed globally to prevent and reduce the use of illicit substances. In the United Kingdom (UK), the drugs considered to have the greatest negative consequences are opiates (predominantly heroin) and stimulants (predominantly cocaine and crack cocaine) and it is a reduction in the use of these problematic drugs which continues to be the major focus of the national drug strategy [2].

In many countries, including the UK, effective treatment is understood to play a central role in enabling individuals to adopt drug free lives [2, 3]. In recognition of the importance of treatment, the amount allocated by the Department of Health to local partnerships to support drug treatment has swelled to £385 million in 2006/07 [4]. In combination with other measures including an expanded workforce, reduced waiting times and the development of a criminal justice programme designed to divert drug users from crime into treatment, this additional funding saw 195,464 people accessing structured drug treatment between 1^st April 2006 and 31^st March 2007, compared to 85,000 in 1998 when the national drug strategy was launched [5], a rise of 130%. Increasing treatment uptake is, however, only one measure of success. Unplanned discharge, or 'drop out', is a well-recognised phenomenon of drug treatment, with people dropping out before participating fully in elements of treatment that will engender change [6].

In the United States of America (USA), longer durations of drug treatment are reported to be a consistent predictor of favourable treatment outcomes [7–10] and retaining drug users in treatment should be an important intermediate goal of the treatment system. In the UK, findings from the National Treatment Outcome Research Study, the only major UK-based, prospective cohort treatment outcomes study reported that clients who stayed in residential treatment programmes (programmes which typically see people with more severe problems at intake than those treated in the community) for 'critical times' had better outcomes at one-year follow up than those who were not retained [6]. Drawing on evidence from the USA and UK, the National Treatment Agency for Substance Misuse (NTA), the Government body overseeing the effectiveness of the drug treatment system in the UK, decided that individuals not retained in treatment for at least three months were unlikely to experience long-term benefits [11]. Accordingly, the NTA introduced a retention target in treatment plans for the first time in 2004/05, with 75% of individuals expected to be retained for 12 weeks or more during each annual reporting cycle [12]. Treatment provision, overseen by drug partnerships called Drug (and Alcohol) Action Teams (D [A]ATs), were clearly expected to change their practices in order to achieve this retention target but were obviously not expected to distort their treatment data accordingly. This risk was monitored by the NTA to expose any services and systems that manipulated their delivery in order to appear to achieve this performance indicator, in the absence of any real improvement [11].

To date, there has been little independent assessment of the impact of this government target on the number of drug users actually retained in treatment for 12 weeks. Furthermore, most of the existing evidence regarding factors associated with longer treatment episodes is derived within the USA, with less information available about factors that influence treatment retention in other countries. Factors associated with retention in the UK, where the majority of treatment is delivered via the National Health Service and charitable/voluntary organisations may be very different from those influencing treatment retention in the USA where patients pay for treatment and are discharged for non-payment [13]. In the UK, the factors that influence treatment retention and attrition remain poorly understood.

The North West of England (mid-year population of 6,846,249 in 2005) is a region containing both urban and rural and both deprived and affluent areas. It is also a region with a relatively high prevalence of problematic drug use. While the exact number of drug users living in the region remains unknown due to the covert nature of the behaviour, estimates suggest that the level of problematic drug use is as high as 52 per 1,000 of male population aged 15 to 44 years in some metropolitan centres [14]. Between 1^st April 2005 and 31st March 2006, 35503 North West residents were in contact with structured drug treatment services and these individuals accounted for almost 20% of the total number of people in contact with treatment services in England [15]. The North West region was therefore chosen as the setting for this study because it represents an area with a high prevalence of problematic drug use and treatment activity. The study set out to quantify the proportion of individuals retained in treatment in the North West of England for 12 weeks or more and to investigate the factors associated with treatment drop out, compared to retention, using established monitoring systems.

In 2005/06, 22330 triage assessments were recorded by the NDTMS database for the North West of England for drug users aged between 18 and 74. This figure included multiple triage assessments for the same individual when they had been seen by two different treatment services or had started a new period of treatment at the same service on multiple occasions. Following data preparation in the manner detailed above, 16626 individuals aged between 18 and 74 began a new episode of treatment in the North West of England in 2005/06. Data on treatment outcomes were unavailable for 272 (1.6%) individuals. Of the remaining 16354 people, 12230 (74.8%) were retained in treatment for 12 weeks or more and 1649 (10.1%) dropped out of treatment before 12 weeks. A further 2475 (15.1%) episodes resulted in other treatment outcomes. These other treatment episode outcomes were as follows: treatment completed in less than 12 weeks, client drug free (n = 275, 1.7%), treatment completed in less than 12 weeks, client not drug free (n = 326, 2.0%), treatment withdrawn, contract breached by client (n = 379, 2.3%), no appropriate treatment available (n = 106, 0.6%), treatment declined by client (n = 1, <0.1%), inappropriate referral (n = 26, 0.2%), client referred on (n = 555, 3.4%), client moved away from area (n = 103, 0.6%), client went to prison (n = 308, 1.9%), client died (n = 23, 0.1%), other outcome (n = 220, 1.3%), outcome not known (n = 153, 0.9%).

There was a significant interaction between age group and deprivation and the coefficients from the logistic regression model were used to explore this further. Figure 1 shows that, for those aged 18 to 24 years, the odds of drop out for deprivation quintile five (least deprived) were 1.93 (95% confidence interval 1.41 to 2.64) the odds of drop out for deprivation quintile one (most deprived). A similar, though smaller, effect was evident for the 25 to 34 age group. Conversely, for those aged 55 to 74 years, the odds of drop out for deprivation quintile five were 0.37 (95% confidence interval 0.18 to 0.73) the odds of drop out for those living in deprivation quintile one. A similar effect of smaller magnitude was observed for the 45 to 54 year age group. No interactions between age group and other explanatory variables were observed. Other explanatory variables (opiate use and recent treatment contact) were not significantly associated with treatment outcome following simultaneous adjustments for other variables. The goodness of fit test suggested that the final model was not a good fit to the data (Pearson's χ² = 322.16, P = 0.009).

Retention in drug treatment is considered by the NTA to be the best available measure of treatment effectiveness [4]. To encourage a reduction in drop out rates, treatment retention measures were included in local partnership treatment plans and Primary Care Trust local delivery plans for the first time in 2004/05 [12], with 75% of drug users expected to be retained in treatment for at least 12 weeks during each annual reporting period [4]. Rates of retention can be improved by better understanding factors associated with premature drop out. The aims of the present study were to assess whether the retention target was achieved in the North West of England in 2005/06 and to identify factors associated with drop out. While a number of factors associated with drop out are identified, results also highlight the complexity of factors associated with treatment outcomes.

Results show that, overall in the North West, the treatment systems achieved the Government's retention target in 2005/06, with 75% of people retained for 12 weeks or more. Contextual information provides a more valuable assessment of how services should be improved, and the analyses presented here shows that current services are failing to retain Asian drug users. Furthermore, demographic data from the 2001 census [20] shows that Asian people are underrepresented in drug treatment services while those from white, mixed, black and 'other' ethnicities are adequately represented. While people of Asian ethnicity account for 3.4% of the population of the North West [20] they accounted for just 1.9% of the drug treatment population (Table 3). This finding may reflect the lower prevalence of drug use among Asian people although a growing body of evidence indicates that drug use exists among black and minority ethnic (BME) groups and that it is increasing [21]. Qualitative research into the extent and nature of drug use among young Asians suggests that they were increasingly likely to use illicit drugs and that knowledge and use of heroin, and perhaps crack cocaine, indicated similar use to the general population of the UK [22].

An alternative explanation is that Asian drug users are not attracted by current treatment provision. A recent review of literature on drug use and service provision for BME communities in England reported a number of potential reasons for differential treatment outcomes among drug users from different ethnicities, over and above the possibility of differing drug profiles [21]. Such reasons included the ethnicity of drugs workers, communication problems for those unable to speak English and a perception among drug users that drug workers lacked awareness of their culture. No literature which evaluated outcomes of culturally-sensitive initiatives in the clinical drug treatment field in the UK could be identified by reviewers although they pointed to two USA-based studies which showed that culturally competent and culturally responsive treatment was often associated with greater treatment retention and longer treatment durations [23, 24]. With respect to drug use among young Asians, healthcare professionals must be sensitive to the cultural context in which their behaviours occur and in particular the potential conflict of being affiliated to two different and at times incongruent cultural groups – British youth culture and that of their ethnic origin [22]. Further work is required to identify perceived barriers to treatment among Asian drug users and the reasons why they are more likely to drop out than their white counterparts, once contact has been initiated.

With respect to other individual characteristics, studies in the USA have reported longer durations of treatment for older drug users [8, 10, 25] and for male drug users [10]. Evidence from a UK-based retrospective cohort study showed no difference between men and women in rates of drop out compared to drug free discharge but that younger drug users (aged 10 to 19 years) were more likely to drop out than their older counterparts [18]. Here, bivariate analysis showed that the odds of drop out exhibited a significant inverse linear relationship with age. However, the multivariate analysis showed that age was not significantly associated with drop out once adjusted for deprivation and that a significant interaction between deprivation and age existed. The effect of deprivation was not included in the retrospective cohort study just detailed so no assessment of the relationship between age and deprivation is available.

A wealth of data exists which shows the relationship between socio-economic status and health. In the North West of England, standardised mortality rates by deprivation show that the all-age rate for those in the most deprived quintile of population is one and three quarter times the rate for the most affluent quintile of the population [26]. However, results presented here show a complex relationship between deprivation and age and their effects upon drug treatment outcomes (Figure 1). For those aged between 18 and 24 years, people living in least deprived areas were almost twice as likely to drop out of treatment than those living in the most deprived areas. A similar effect, of smaller magnitude, was observed for those aged 25 to 34 years. For those in the oldest age group, (55 to 74 years) the converse was true and those living in the most deprived areas were less likely to drop out than those living in the least deprived areas. A similar, though smaller, effect was observed for those aged 45 to 55 years.

Encouraging treatment participation among younger drug users is a well-recognised problem [14]. Younger drug users are less likely to view their drug use as problematic or view themselves as dependent upon drugs [27] and young drug users living in affluent areas may be somewhat protected from the negative consequences of drug use from their affluent families. It is not clear why older drug users living in the most deprived environmental conditions were the least likely to drop out and this finding warrants further consideration. It should be noted however, that an ecological, rather than individual, measure of deprivation was used and that affluent individuals live in deprived areas and visa versa. It is conceivable that affluent areas have a lower prevalence of problematic drug use and lower levels of service provision and that younger drug users with less motivation to change are deterred from using services to which they have to travel. Travelling more than one mile to treatment has been shown in one USA-based study to reduce the likelihood of a person completing treatment by a half after the effects of demographic differences and the type of drugs used were controlled [28] and the possible impact of travelling to services requires further consideration in the UK.

Controlling for the effect of deprivation and other explanatory factors, significant differences in the odds of drop out were observed for different areas of the North West of England with the odds greater among residents of Cumbria and Lancashire and Greater Manchester than Cheshire and Merseyside. Further work is needed to understand what is different about these regions and what factors differentially affect drop out. It is possible that there are differences between these areas in the manner in which treatment is delivered, or differences in client satisfaction with the treatment process and it is important to consider the potential influence of the quality of treatment on drop out rates rather than simply emphasising the role of client characteristics [10]. At the service rather than regional level, previous UK-based work has already demonstrated the effect the service can make on the number of clients retained successfully in treatment for six months [29]. The time people wait to access treatment will also influence retention. Here, a person was deemed to be in continuous treatment if they reappeared in treatment within three weeks of leaving the first agency they were in contact with. Individuals who dropped out of service and did not reappear until more than three weeks had elapsed were recorded as having dropped out so geographical areas whose waiting times were longer than three weeks would have a greater proportion of people reported as dropping out. Unfortunately, no assessment of the impact of waiting times on treatment outcomes can be made in this study.

The only substance significantly associated with outcomes was alcohol, with alcohol users less likely to drop out of treatment than non-alcohol users. Following adjustment for other variables, opiate use was found not be significantly associated with treatment outcomes. The only large-scale longitudinal, prospective cohort study of treatment outcomes in the UK reported that there was no long-term reduction in drinking among patients of drug treatment programmes and authors concluded that poor drinking outcomes required urgent attention [30]. That alcohol use is considered as part of the overall drug profile of people contacting drug treatment is therefore encouraging, particularly in light of the role of alcohol in drug-related deaths [31, 32] and its exacerbation of chronic hepatitis C, currently the most significant infection affecting drug users who inject in the UK [33]. It is possible, however, that this finding is a feature of reporting. The NDTMS records up to three drugs for each person, with those identified as most problematic (used most frequently, associated with greatest health and criminal consequences) reported preferentially. Alcohol is generally perceived to be less problematic than illicit drugs and individuals whose drug profile does not include alcohol may represent drug users who use a larger range of illicit drugs – those with the most severe drug problems. Alcohol use may therefore act as a proxy measure of less severe drug problems. A recent investigation into the characteristics of drug users in contact with structured drug treatment reported that those whose drug profiles included alcohol as a supplementary drug were younger, were less likely to use heroin and were more likely to be referred into treatment via the criminal justice system [34].

Previous research in the North West of England showed that those referred from the criminal justice system were significantly more likely to drop out and less likely to complete treatment drug free than those referred from non-criminal justice sources [18]. Here the route of referral was not significantly associated with treatment drop out in the bivariate analysis. Most community sentences last longer than 12 weeks; the now defunct Drug Treatment and Testing Orders for example (orders that required the offender to attend regularly both court and treatment and provide a urine sample for drug testing) became available nationally in October 2000 and could last between six months and three years [35]. It is therefore conceivable that those referred from the criminal justice system were retained in treatment for 12 weeks due to fear of breaking their order but had poorer long-term outcomes because they lacked the intrinsic motivation for behaviour change. However, the identification of poorer outcomes for criminal justice referrals [18] relates to outcomes recorded between 1998 and 2001/02, while the present study was undertaken after the introduction of retention targets. If drug use is considered a chronic condition defined by relapse and re-presentation [18, 36], the present finding that a high proportion of both criminal justice and non-criminal justice referrals adhere to treatment for at least 12 weeks reflects positively on current policy and practice.

While this study investigated the effects of a variety of variables, the goodness of fit test suggests that other important covariates were missing from the model. This finding hints at the complexity of factors associated with treatment outcomes and confirms suggestions that drop out and retention is an index that capture the impact of many interrelated individual and process measures [7]. In particular, it was not possible to evaluate the effect of treatment modality on treatment retention because an individual's entire treatment episode could be comprised of treatment within different modalities, yet treatment modality is likely to have considerable bearing on treatment outcomes. An examination of the covariate patterns of the goodness of fit test showed discrepancies between the observed and expected values for deprivation and further work looking at the interaction between deprivation and other explanatory variables is needed. However, the reliability of the findings presented here is enhanced by treatment outcome data being unavailable for only 272 (1.6%) individuals. Large-scale prospective cohort studies of drug treatment outcomes are costly and time consuming and the robustness of findings are often questionable due to follow-up losses occurring differentially across comparison groups. The use of well-established monitoring systems provides a cost-effective means by which treatment outcomes can be assessed [18] and while case control studies are also prone to selection bias (with controls not adequately reflecting the exposure history of cases), here controls would have been cases if they had dropped out of treatment before 12 weeks. Recall bias, an accepted weakness of case control studies, did not affect results of this study because data were collected prospectively; exposure data recorded before treatment outcomes were ascertained.

Problems associated with the use of routine data (for example, inconsistencies in the manner in which information is recorded) are somewhat mitigated in this study by using nationally agreed definitions and reporting protocols, standardised coding framework and data validation checks [12]. Furthermore, the large sample size reduces the impact of slight non-systematic variations, increases the study power and reduces the likelihood of a Type II error. However, the study relies totally on self-reported data and the accuracy of such reports cannot be ascertained. The standardisation of the drug treatment and data reporting across England mean that the results from the North West of England can be generalised to other areas although results also show that the effect of factors specific to each region would require further consideration.

During the 2005/06 reporting period, 75% of clients were retained in drug treatment in the North West of England. Asian drug users were more likely to drop out of treatment than their white equivalents, and further work into why this is the case is needed in light of evidence to suggest the drug profiles of Asian drug users are similar to those of their white counterparts. Residents of Cumbria and Lancashire and Greater Manchester were more likely to drop out of treatment than people living in Cheshire and Merseyside, and work on the impact of treatment processes, treatment quality and aspects of client satisfaction would be welcomed. While overall, those living in the least deprived areas were more likely to drop out than those living in the most deprived areas, the effect of deprivation varied significantly according to the age of the individual. The effect of deprivation on treatment outcomes requires further consideration, preferably using an individual, rather than geographically aggregated measure of deprivation. These findings show that personal characteristics, process measures and situational factors all effect treatment outcomes and highlight the complexity of factors associated with treatment success. The large sample size, small proportion of missing data and the inclusion of data quality and standardisation measures increase the validity and reliability of these findings. Finally, the NTA are currently revising the retention target, which is due to change for the 2008/09 reporting cycle. To date, it is not clear how the retention target will change, or whether 12 weeks will continue to be the retention benchmark.