Advertisement

Effectiveness of inpatient withdrawal and residential rehabilitation interventions for alcohol use disorder: A national observational, cohort study in England

  • Brian Eastwood
    Affiliations
    Alcohol, Drugs and Tobacco Division, Health Improvement, Public Health England, 2nd Floor, Skipton House, 80 London Road, London SE1 6LH, United Kingdom

    Addictions Department, Institute of Psychiatry, Psychology and Neuroscience, King's College London, Box 48, DeCrespigny Park, Denmark Hill, London SE5 8AF, United Kingdom
    Search for articles by this author
  • Amy Peacock
    Affiliations
    National Drug and Alcohol Research Centre, University of New South Wales, Randwick, Sydney 2052, New South Wales, Australia

    School of Medicine (Psychology), University of Tasmania, Private Bag 30, Hobart 7001, Tasmania, Australia
    Search for articles by this author
  • Tim Millar
    Affiliations
    Centre for Mental Health and Safety, Division of Psychology and Mental Health, School of Health Sciences, Faculty of Biology, Medicine, and Health, 4th Floor, Block C, Ellen Wilkinson Building, The University of Manchester, Oxford Road, Manchester, M13 9PL, United Kingdom
    Search for articles by this author
  • Andrew Jones
    Affiliations
    Centre for Epidemiology, Division of Psychology and Mental Health, School of Health Sciences, Faculty of Biology, Medicine, and Health, 4th Floor, Block C, Ellen Wilkinson Building, The University of Manchester, Oxford Road, Manchester M13 9PL, United Kingdom
    Search for articles by this author
  • Jonathan Knight
    Affiliations
    Alcohol, Drugs and Tobacco Division, Health Improvement, Public Health England, 2nd Floor, Skipton House, 80 London Road, London SE1 6LH, United Kingdom
    Search for articles by this author
  • Patrick Horgan
    Affiliations
    Alcohol, Drugs and Tobacco Division, Health Improvement, Public Health England, 2nd Floor, Skipton House, 80 London Road, London SE1 6LH, United Kingdom
    Search for articles by this author
  • Tim Lowden
    Affiliations
    Alcohol, Drugs and Tobacco Division, Health Improvement, Public Health England, 2nd Floor, Skipton House, 80 London Road, London SE1 6LH, United Kingdom
    Search for articles by this author
  • Peter Willey
    Affiliations
    Alcohol, Drugs and Tobacco Division, Health Improvement, Public Health England, 2nd Floor, Skipton House, 80 London Road, London SE1 6LH, United Kingdom
    Search for articles by this author
  • John Marsden
    Correspondence
    Corresponding author at: Institute of Psychiatry, Psychology and Neuroscience, King's College London, Box 48, DeCrespigny Park, Denmark Hill, London SE5 8AF, United Kingdom.
    Affiliations
    Alcohol, Drugs and Tobacco Division, Health Improvement, Public Health England, 2nd Floor, Skipton House, 80 London Road, London SE1 6LH, United Kingdom

    Addictions Department, Institute of Psychiatry, Psychology and Neuroscience, King's College London, Box 48, DeCrespigny Park, Denmark Hill, London SE5 8AF, United Kingdom
    Search for articles by this author
Published:February 07, 2018DOI:https://doi.org/10.1016/j.jsat.2018.02.001

      Highlights

      • This is the first national study of inpatient withdrawal and residential rehabilitation for alcohol use disorder in England.
      • 59% successfully completed AUD treatment and did not represent for more treatment within 6 months;
      • Community-based treatment prior and subsequent to IW positively predicted favourable outcome
      • Community-based treatment subsequent to RR predicted favourable outcome
      • Provision of structured continuing care was associated with favourable outcome

      Abstract

      Background

      This was a national English observational cohort study to estimate the effectiveness of inpatient withdrawal (IW) and residential rehabilitation (RR) interventions for alcohol use disorder (AUD) using administrative data.

      Methods

      All adults commencing IW and/or RR intervention for AUD between April 1, 2014 and March 31, 2015 reported to the National Drug Treatment Monitoring System (n = 3812). The primary outcome was successful completion of treatment within 12 months of commencement, with no re-presentation (SCNR) in the subsequent six months, analysed by multi-level, mixed effects, multivariable logistic regression.

      Results

      The majority (70%, n = 2682) received IW in their index treatment journey; one-quarter (24%, n = 915) received RR; 6% (n = 215) received both. Of treatment leavers, 59% achieved the SCNR outcome (IW: 57%; RR: 64%; IW/RR: 57%). Positive outcome for IW was associated with older age, being employed, and receiving community-based treatment prior to and subsequent to IW. Patients with housing problems were less likely to achieving the outcome. Positive outcome for RR was associated with paid employment, self/family/peer referral, longer duration of RR treatment, and community-based treatment following discharge. Community-based treatment prior to entering RR, and receiving IW during the same treatment journey as RR, were associated with lower likelihood of SCNR.

      Conclusions

      In this first national effectiveness study of AUD in the English public treatment system for alcohol-use disorders, 59% of patients successfully completed treatment within 12 months and did not represent for more treatment within six months. Longer duration of treatment and provision of structured continuing care is associated with better treatment outcomes.

      Keywords

      1. Introduction

      Alcohol use is a leading risk factor for morbidity and mortality (
      • World Health Organisation
      Global status report on alcohol and health 2014.
      ). An estimated 3.6% of the global population aged 15–64 years meet criteria for alcohol use disorder each year (AUD;
      • American Psychiatric Association
      Diagnostic and statistical manual of mental disorders (DSM-5®).
      ), with relatively higher rates estimated for Europe (5.5%;
      • Rehm J.
      • Mathers C.
      • Popova S.
      • Thavorncharoensap M.
      • Teerawattananon Y.
      • Patra J.
      Global burden of disease and injury and economic cost attributable to alcohol use and alcohol-use disorders.
      ). Negative health, social and economic consequences are higher among the population with AUD (
      • Hasin D.S.
      • Stinson F.S.
      • Ogburn E.
      • Grant B.F.
      Prevalence, correlates, disability, and comorbidity of DSM-IV alcohol abuse and dependence in the United States: Results from the national epidemiologic survey on alcohol and related conditions.
      ;
      • Odlaug B.
      • Gual A.
      • DeCourcy J.
      • Perry R.
      • Pike J.
      • Heron L.
      • Rehm J.
      Alcohol dependence, co-occurring conditions and attributable burden.
      ). In Europe, it is estimated that AUDs are responsible for 60% of alcohol-related mortality (
      • Rehm J.
      • Shield K.D.
      • Gmel G.
      • Rehm M.X.
      • Frick U.
      Modeling the impact of alcohol dependence on mortality burden and the effect of available treatment interventions in the European Union.
      ). There are concerns that only a minority of people with AUD access treatment services (). For example, in England just 6% of those with AUD in England receive treatment (
      • National Institute for Health and Care Excellence
      Alcohol-use disorders: diagnosis, assessment and management of harmful drinking and alcohol dependence.
      ; ).
      The goal of AUD treatment is to help patients quit drinking or prevent harmful consumption, thereby reducing the health, social and economic harms (
      • Haber P.
      • Lintzeris N.
      • Proude E.
      • Lopatko O.
      Guidelines for the treatment of alcohol problems.
      ;
      • Rahhali N.
      • Millier A.
      • Briquet B.
      • Laramée P.
      • Aballéa S.
      • Toumi M.
      • Daeppen J.-B.
      Modelling the consequences of a reduction in alcohol consumption among patients with alcohol dependence based on real-life observational data.
      ). In the English public healthcare system, structured AUD treatment is mainly delivered by National Health Service or third-sector providers in the outpatient/community setting, offering psychosocial interventions (including motivational, cognitive behavioural, family/social network modalities and facilitation of access to 12-step groups) and pharmacotherapies (including acamprosate and naltrexone for approximately 6 months).
      This is complemented by a relatively small number of inpatient withdrawal (IW) and residential rehabilitation (RR) services. Patients are treated in the community or inpatient/residential setting based on a clinical assessment of problem severity and complexity; patient preference; and service availability (
      • National Institute for Health and Care Excellence
      Alcohol-use disorders: diagnosis, assessment and management of harmful drinking and alcohol dependence.
      ). There is some provision of detoxification management in the community over 7–10 days typically using benzodiazepines (
      • National Institute for Health and Care Excellence
      Alcohol-use disorders: diagnosis, assessment and management of harmful drinking and alcohol dependence.
      ).
      IW or RR are usually indicated for people with greater AUD severity (e.g. those drinking >30 standard drinks per typical drinking day), or instances of complexity due to unstable housing; comorbid psychiatric/physical conditions; or a history of seizures. IW is usually 5–7 nights in a controlled hospital environment with pharmacological interventions for medical management of withdrawal (
      • National Institute for Health and Care Excellence
      Alcohol-use disorders: diagnosis, assessment and management of harmful drinking and alcohol dependence.
      ). RR is usually a 6–12 weeks stay in a structured, residential facility which provides a phased, structured programme of psychosocial interventions. Detoxification support may be provided as needed. RR programmes usually follow an underlying therapeutic philosophy, including 12-step; therapeutic community; faith-based practice; cognitive behavioural therapy and social learning; personal and skills development; or an eclectic/integrated approach (
      • Moos R.H.
      • Moos B.S.
      • Andrassy J.M.
      Outcomes of four treatment approaches in community residential programs for patients with substance use disorders.
      ).
      Routine delivery of AUD treatment interventions is remarkably under-researched. Our group has previous reported reductions in offending associated with AUD treatment (
      • Willey H.
      • Eastwood B.
      • Gee I.L.
      • Marsden J.
      Is treatment for alcohol use disorder associated with reductions in criminal offending? A national data linkage cohort study in England.
      ), but there have been no national outcome studies. Addressing this gap is important because treatment outcomes in the clinic cannot be assumed to be the same as randomised controlled trials. AUD intervention trials are often designed to answer questions of efficacy; with participants selected on restricted characteristics (
      • Witkiewitz K.
      • Finney J.W.
      • Harris A.H.
      • Kivlahan D.R.
      • Kranzler H.R.
      Recommendations for the design and analysis of treatment trials for alcohol use disorders.
      ); using very detailed research assessment procedures (
      • Epstein E.E.
      • Drapkin M.L.
      • Yusko D.A.
      • Cook S.M.
      • McCrady B.S.
      • Jensen N.K.
      Is alcohol assessment therapeutic? Pretreatment change in drinking among alcohol-dependent women.
      ); and implemented with complex intervention exposures that are not routinely available in the healthcare system (
      • Allen J.
      • Mattson M.
      • Miller W.
      • Tonigan J.
      • Connors G.
      • Rychtarik R.
      • Litt M.
      Matching alcoholism treatments to client heterogeneity.
      ).
      The National Drug Treatment Monitoring System evaluates all public AUD treatment services in England (NDTMS;
      • Public Health England
      National drug treatment monitoring system (NDTMS): Adult drug treatment business definition.
      ). NDTMS has been in operation since 2005/06 and had an initial focus on services providing structured treatment and care for people with drug use disorders. All operational public alcohol and drug treatment services who deliver treatment interventions now report to the system, and ~98% of patients consent to the use of their administrative and clinical data for local treatment system needs assessment and national research (
      • Marsden J.
      • Eastwood B.
      • Bradbury C.
      • Dale-Perera A.
      • Farrell M.
      • Hammond P.
      • Group, N. D. T. M. S. O. S
      Effectiveness of community treatments for heroin and crack cocaine addiction in England: A prospective, in-treatment cohort study.
      ;
      • Marsden J.
      • Eastwood B.
      • Jones H.
      • Bradbury C.
      • Hickman M.
      • Knight J.
      • White M.
      Risk adjustment of heroin treatment outcomes for comparative performance assessment in England.
      ;
      • White M.
      • Burton R.
      • Darke S.
      • Eastwood B.
      • Knight J.
      • Millar T.
      • Marsden J.
      Fatal opioid poisoning: A counterfactual model to estimate the preventive effect of treatment for opioid use disorder in England.
      ;
      • Willey H.
      • Eastwood B.
      • Gee I.L.
      • Marsden J.
      Is treatment for alcohol use disorder associated with reductions in criminal offending? A national data linkage cohort study in England.
      ).
      In 2008/09, NDTMS was enhanced to monitor outcomes from all public treatment services for AUD. Elsewhere, we report on the effectiveness of community-based AUD interventions (
      • Peacock A.
      • Eastwood B.
      • Jones A.
      • Millar T.
      • Knight J.
      • Horgan P.
      • Marsden J.
      Effectiveness of psychosocial and pharmacological treatment for alcohol use disorder: A national observational, cohort study of community setting interventions in England.
      ). In this report, we estimate the clinical effectiveness of IW and RR interventions for AUD in the English public healthcare system.

      2. Materials and methods

      2.1 Design

      This was an observational, follow-up study of all individuals accessing publicly funded, IW and/or RR treatment for AUD in England. The study included all 152 upper-tier local authorities within England, and all specialist AUD services. The study is reported according to the STROBE and RECORD guidelines for cohort research (
      • Benchimol E.I.
      • Smeeth L.
      • Guttmann A.
      • Harron K.
      • Moher D.
      • Petersen I.
      • Committee R.W.
      The reporting of studies conducted using observational routinely-collected health data (RECORD) statement.
      ).

      2.2 Patient and treatment information

      NDTMS records were accessed on patient-demographic, behavioural, clinical and treatment outcome variables for each episode of treatment, including the dates of starting and finishing specific treatment interventions and the treatment exit date (
      • Public Health England
      Adult substance misuse statistics from the national drug treatment monitoring system (NDTMS): 1 April 2014 to 31 March 2015.
      ,
      • Public Health England
      National drug treatment monitoring system (NDTMS): Adult drug treatment business definition.
      ).
      Reflecting national reporting standards (
      • Public Health England
      National drug treatment monitoring system (NDTMS): Adult drug treatment business definition.
      ), individual treatment episodes were concatenated into ‘treatment journeys’, whereby multiple episodes (community-based or residential program) are subsumed under a single journey. AUD intervention episodes were allocated to the same journey if fewer than 21 days elapsed between the date of ending one treatment modality and the date of starting a subsequent one. In this way, a treatment journey for a patient could comprise a single intervention episode; concurrent episodes provided by more than one agency; or a continuing care package of consecutive episodes provided by one or more service providers.

      2.3 Study cohort

      The study population was adults (aged ≥18 years) who commenced IW and/or RR treatment for primary AUD between 1 April 2014 and 31 March 2015 (N = 3861). Patients were not included in the study cohort if they: (1) reported problematic use of other psychoactive substances at assessment; (2) had missing information on drinks per drinking day (DDD) at both triage and treatment admission; or (3) had missing information on clinical status at discharge were not considered for inclusion.
      Analyses were based on the patient's first treatment journey during the period (hereafter ‘index journey’). The observation period commenced from the date of starting IW or RR and ended: (1) six months after the date of discharge from the index journey, if discharge occurred within 12 months of starting IW or RR, or (2) 12 months after starting IW or RR if the patient was not yet discharged (the latter group was excluded from analysis of the primary outcome). Periods in community-based treatment subsequent but not prior to IW or RR contributed to the observation time, with discharge date adjusted accordingly. If the index journey involved progression from IW or RR, or vice versa, it was categorised as involving both.

      2.4 Outcome measure

      The study outcome measure is the English national outcome standard, defined as the proportion of the cohort that successful completed treatment within 12 months of commencement with no representation within six months (SCNR;
      • Public Health England
      National drug treatment monitoring system (NDTMS): Adult drug treatment business definition.
      ).
      The proportion of patients treated who complete treatment successfully has been used before in the AUD treatment literature (
      • Alterman A.I.
      • Langenbucher J.
      • Morrison R.L.
      State-level treatment outcome studies using administrative databases.
      ). This outcome may be associated with improvements in personal and social functioning (
      • Finigan M.
      Societal outcomes and cost savings of drug and alcohol treatment in the state of Oregon. Exploring the key components of drug courts.
      ), but it does not identify sustained benefit. This is important given the relapsing nature of AUD. In the present context, re-presentation for further AUD treatment within six months of discharge is taken to be an indicator of remission.
      Treatment journeys were categorised according to clinical assessment of the patient's discharge status, as: (1) successfully completed treatment within 12 months; (2) retained in the same treatment journey at 12 months from entry; or (3) withdrawn from treatment journey within 12 months of entry (unsuccessful transfer between agencies; treatment terminated due to incarceration; patient dropped out treatment died during treatment). Successful treatment was defined as the patient being discharged having: completed their care plan, with no AUD (and either abstinent or no heavy drinking), and no re-presentation to any service for further AUD treatment within six months of concluding their treatment journey (
      • Eastwood B.
      • Strang J.
      • Marsden J.
      Effectiveness of treatment for opioid use disorder: A national, five-year, prospective, observational study in England.
      ;
      • Peacock A.
      • Eastwood B.
      • Jones A.
      • Millar T.
      • Knight J.
      • Horgan P.
      • Marsden J.
      Effectiveness of psychosocial and pharmacological treatment for alcohol use disorder: A national observational, cohort study of community setting interventions in England.
      ;
      • Public Health England
      Public health outcomes framework: 2.15iii successful completion of alcohol treatment.
      ).

      2.5 Covariates

      Following our general evaluation approach (
      • Willey H.
      • Eastwood B.
      • Gee I.L.
      • Marsden J.
      Is treatment for alcohol use disorder associated with reductions in criminal offending? A national data linkage cohort study in England.
      ), the analysis included patient socio-demographics; indicators of clinical severity/case complexity; and summary measures of treatment journey exposure. Possible covariates were identified from reviews of predictors of treatment outcome (
      • Adamson S.J.
      • Sellman J.D.
      • Frampton C.M.
      Patient predictors of alcohol treatment outcome: A systematic review.
      ;
      • Brorson H.H.
      • Arnevik E.A.
      • Rand-Hendriksen K.
      • Duckert F.
      Drop-out from addiction treatment: A systematic review of risk factors.
      ). Patient-level measures were recorded at triage assessment at the first admission to the index journey. The following covariates were included:

      2.5.1 Socio-demographics

      Analyses were adjusted for: age (years), gender, ethnic origin, housing problems (homeless, in short-term hostel provision, or at risk of eviction in the past 28 days); and employment (whether in paid work in past 28 days). Analyses were also adjusted for an indicator of social deprivation, imputed by assigning patients to their electoral ward of residence, based on the partial postal code recorded by NDTMS, categorised according to the Department for Communities and Local Government ward-level Indices of Multiple Deprivation (IMD;
      • Department for Communities and Local Government
      The english index of multiple deprivation (IMD) 2015.
      ). If the partial postcode could be located in more than one electoral ward the median IMD score was assigned. The IMD score for the first treatment service address within the treatment journey was assigned if partial postcode was missing. Following UK local government reporting convention (
      • Public Health England
      Adult substance misuse statistics from the national drug treatment monitoring system (NDTMS): 1 April 2014 to 31 March 2015.
      ), IMD scores were grouped by quintile.

      2.5.2 Clinical characteristics

      Analyses were adjusted based on patients' self-report of the number of days on which they consumed alcohol; and the number of standard DDD in the previous 28 days. These were recorded via the Treatment Outcomes Profile (TOP;
      • Marsden J.
      • Farrell M.
      • Bradbury C.
      • Dale-Perera A.
      • Eastwood B.
      • Roxburgh M.
      • Taylor S.
      Development of the treatment outcomes profile.
      ). TOP is the national clinical outcome measure in NDTMS, administered as a face-to-face structured clinical interview, with timeline follow-back technique to maximise recall accuracy (
      • Sobell L.C.
      • Sobell M.B.
      • Leo G.I.
      • Cancilla A.
      Reliability of a timeline method: Assessing normal drinkers' reports of recent drinking and a comparative evaluation across several populations.
      ). Reflecting
      • National Institute for Health and Care Excellence
      Alcohol-use disorders: diagnosis, assessment and management of harmful drinking and alcohol dependence.
      guidelines, DDD were categorised as: abstinent (i.e. zero drinks in the past 28 days); low to high (1 to 15 DDD); high to extreme (16 to 30 DDD); and extreme (≥31 DDD). We also adjusted for source of treatment referral (health service; self/friend/family; criminal justice system), and whether the patient had previously received AUD treatment as recorded in NDTMS (i.e., from 2006 onwards).

      2.5.3 Treatment exposure

      Analyses were adjusted based on receipt within the index treatment journey involving: (1) community-based treatment that started prior to the RR or IW component; (2) community-based treatment that ended following cessation of RR/IW; and (3) recovery support (treatment agency provision or referral for: facilitated access to mutual aid; peer support involvement, family, parenting, support groups; housing, employment, education and training support; and complementary therapies). The total duration of treatment exposure, IW exposure, and RR exposure was recorded in days (IW) or weeks (RR) from the start of the index journey to discharge, right censored at 365 days and computed using the triage and discharge date for the specific intervention. Pharmacological intervention provided during residential rehabilitation was indicator coded (0,1).

      2.6 Statistical analysis

      In this national treatment population study, with a hierarchical design and participants grouped in treatment services, our statistical power concerns reflected control of clustering effects and minimising bias of model coefficients. We note that multi-level simulation studies have concluded that power is increased by adding groups rather than the number of cases per group. In the event, our sample well exceeded the minimum recommended for multi-level designs (e.g. 50 groups for random effects models;
      • Maas C.J.
      • Hox J.J.
      Sufficient sample sizes for multilevel modeling.
      ).
      Stata (Stata version 13.1; Stata Corporation, College Station, TX, USA) was used for analyses. All estimates were computed with associated 95% confidence intervals (CI). A multi-level approach to analysis was chosen given that individuals were nested within agencies. Multinomial logistic regression (command: mlogit) with robust standard errors was used to adjust for clustering of participants within treatment services and to identify correlates of IW and RR (i.e., “IW group”, “RR group”, “IW and RR group”). The analysis of SCNR excluded those individuals who did not leave treatment within 12 months of initiation.
      Missing data for employment, housing status, days of alcohol use and DDD from standard assessment procedures were replaced with data from the initial TOP assessment interview where available. Despite this, missing values were recorded for covariates ethnicity (n = 82), paid employment (n = 159), unstable housing (n = 46), and referral source (n = 19). With no evidence that data loss was not missing-at-random (
      • Little R.J.A.
      • Little D.B.
      Statistical analysis with missing data.
      ), we used a multiple imputation via chained equations procedure (command: mi: impute chained). An all-case multivariate logistic model was run to check on potential bias and loss of precision (
      • Sterne J.A.
      • White I.R.
      • Carlin J.B.
      • Spratt M.
      • Royston P.
      • Kenward M.G.
      • Carpenter J.R.
      Multiple imputation for missing data in epidemiological and clinical research: Potential and pitfalls.
      ). To achieve a relative efficiency above 98% (
      • Rubin D.B.
      Multiple imputation for non-response in surveys.
      ) and to ensure that reduction in power was <1% (
      • Graham J.W.
      • Olchowski A.E.
      • Gilreath T.D.
      How many imputations are really needed? Some practical clarifications of multiple imputation theory.
      ), 20 datasets of probabilistic values were created, each analysed separately, and then combined using Rubin's rules.
      Models were run separately for participants of each treatment setting (i.e., IW versus RR) for analysis of the SCNR outcome (command: meqrlogit with 7 integration points), with patient-level covariates were included as fixed effects. Simulation studies suggest a minimum of 10 or more events per covariate in logistic regression analyses; this criteria was satisfied in the current analyses (
      • Peduzzi P.
      • Concato J.
      • Kemper E.
      • Holford T.R.
      • Feinstein A.R.
      A simulation study of the number of events per variable in logistic regression analysis.
      ). For the purpose of these analyses, those participants who had received both IW and RR were included in analyses for the RR model, and a covariate identifying receipt of IW was included in the model. Models based on imputation are reported in-text; complete-case analyses are available in supplementary materials.

      3. Results

      3.1 Socio-demographic and clinical characteristics

      The majority of patients were male (65%) and the median age was 46 years (Table 1). Half (53%) had a history of structured AUD treatment prior to their index journey. Median drinking days at triage was 28 (i.e., daily in the past month), and the majority were classified as ‘High to Extreme’ (49%) or ‘Extreme’ (30%) based on DDD.
      Table 1Socio-demographic, clinical/treatment characteristics of the cohort (n = 3812).
      CharacteristicIW

      (n = 2682)
      RR

      (n = 915)
      Combined

      (IW + RR; n = 215)
      Total

      (n = 3812)
      Socio-demographic
      Recorded at triage assessment with reference to the preceding 28days where appropriate.
       No. (%) male1792 (66.8)569 (62.2)129 (60.0)2490 (65.3)
       Age (M, IQR)47 (40, 54)45 (37, 51)45 (37, 52)46 (39, 53)
       No. (%) Black or minority ethnic group309 (11.7)86 (9.9)22 (10.4)417 (11.2)
       No. (%) deprivation quintile
      1 (least deprived)508 (18.9)212 (23.2)33 (15.3)753 (19.8)
      2442 (16.5)219 (23.9)47 (21.9)708 (18.6)
      3494 (18.4)143 (15.6)54 (25.1)691 (18.1)
      4550 (20.5)168 (18.4)49 (22.8)767 (20.1)
      5 (most deprived)688 (25.7)173 (18.9)32 (14.9)893 (23.4)
       No. (%) with housing problems358 (13.5)196 (21.8)39 (18.2)593 (15.7)
       No. (%) in paid employment438 (17.1)160 (18.2)37 (17.5)635 (17.4)
      Clinical
      Recorded at triage assessment with reference to the preceding 28days where appropriate.
       Days alcohol use (M, IQR)28 (28, 28)28 (14, 28)28 (28, 28)28 (28, 28)
      No. (%) DDD group
      Recoded at treatment commencement with reference to the preceding 28days where appropriate.
       Abstinent113 (4.2)122 (13.3)11 (5.1)246 (6.5)
       Low-high408 (15.2)110 (12.0)28 (13.0)546 (14.3)
       High-extreme1340 (50.0)438 (47.9)103 (47.9)1881 (49.3)
       Extreme821 (30.6)245 (26.8)73 (34.0)1139 (29.9)
      No. (%) previous treatment for AUD1458 (54.4)442 (48.3)129 (60.0)2029 (53.2)
      No. (%) referral route
       Health service897 (33.5)409 (45.4)81 (37.7)1387 (36.6)
       Self/family member/friend33 (1.2)12 (1.3)5 (2.3)50 (1.3)
       Criminal justice1748 (8.6)479 (53.2)129 (60.0)2356 (62.1)
      Treatment exposure
       Total weeks in treatment (M, IQR)6 (2−20)13 (6–24)25 (13–41)9 (2−23)
       No. (%) prior structured community-based treatment1526 (56.9)359 (39.2)144 (67.0)2029 (53.2)
       Total days community-based treatment (M, IQR)62 (35–104)84 (50–133)83 (49–132)67 (38–111)
       No. (%) subsequent structured community-based treatment1627 (60.7)252 (27.5)84 (39.1)1963 (51.5)
       No. (%) recovery support1952 (69.6)453 (49.1)184 (87.6)2589 (65.7)
      IW, inpatient withdrawal; RR, residential rehabilitation; IW + RR, inpatient withdrawal and residential rehabilitation; M, median; IQR, inter-quartile range; DDD, drinks per drinking day; AUD, alcohol use disorder.
      a Recorded at triage assessment with reference to the preceding 28 days where appropriate.
      b Recoded at treatment commencement with reference to the preceding 28 days where appropriate.

      3.2 Treatment exposure and status

      Most patients (70%, n = 2682) received IW treatment (“IW group”); one-quarter (24%, n = 915) received RR treatment (“RR group”); and 6% (n = 215) received both (“Combined group”). Compared to the “IW group” (referent), patients in the “RR” and “Combined” groups were slightly younger and less likely to be male. The “RR group” were more likely to report: housing problems; self, family, or peer referral to treatment; and to report abstinence at triage (Table 2).
      Table 2Likelihood of IW and RR group classification: by socio-demographic and clinical/treatment characteristics (n = 3812).
      CovariateRR

      (n = 915)
      Combined

      (IW + RR; n = 215)
      Socio-demographic
      Recorded at triage assessment with reference to the preceding 28days as appropriate.
       Male0.82 (0.68, 0.99), p = 0.0370.74 (0.57, 0.97), p = 0.031
       Age0.98 (0.97, 0.99), p < 0.0010.98 (0.97, 0.99), p < 0.001
       Black or minority ethnic group0.84 (0.49, 1.42), p = 0.5070.88 (0.58, 1.33), p = 0.541
      Deprivation quintile
      Referent category: deprivation first quartile.
       21.19 (0.90, 1.57), p = 0.2291.64 (0.95, 2.81), p = 0.074
       30.69 (0.45, 1.07), p = 0.1001.68 (0.97, 2.92), p = 0.063
       40.73 (0.45, 1.20), p = 0.2151.37 (0.81, 2.33), p = 0.245
       5 (most deprived)0.60 (0.28, 1.28), p = 0.1860.72 (0.35, 1.45), p = 0.353
       Housing problems1.79 (1.24, 2.59), p = 0.0021.43 (0.86, 2.36), p = 0.164
       Paid employment1.08 (0.83, 1.40), p = 0.5641.03 (0.69, 1.53), p = 0.903
      Clinical
      Recorded at triage assessment with reference to the preceding 28days as appropriate.
       Days alcohol use0.92 (0.89, 0.95), p < 0.0011.02 (0.98, 1.06), p = 0.280
      DDD group
      Referent category: high-extreme, recoded at treatment commencement with reference to the preceding 28days were appropriate.
       Abstinent3.30 (1.81, 6.03), p < 0.0011.27 (0.63, 2.54), p = 0.506
       Low-high0.82 (0.59, 1.15), p = 0.2530.89 (0.60, 1.32), p = 0.568
       Extreme0.91 (0.70, 1.19), p = 0.5051.16 (0.87, 1.55), p = 0.325
      Previous treatment for AUD0.78 (0.59, 1.04), p = 0.0881.26 (0.88, 1.80), p = 0.209
      Referral route
      Referent category: health service.
       Self/family member/friend1.66 (1.12, 2.48), p = 0.0121.22 (0.83, 1.80), p = 0.309
       Criminal justice1.33 (0.59, 2.96), p = 0.4902.05 (0.75, 5.65), p = 0.164
      Treatment exposure
      Total weeks in treatment1.01 (1.00, 1.02), p = 0.0231.04 (1.03, 1.05), p < 0.001
      Prior structured community-based treatment0.49 (0.28, 0.85), p = 0.0111.54 (0.87, 2.71), p = 0.139
      Subsequent structured community-based treatment0.15 (0.05, 0.43), p = 0.0010.10 (0.04, 0.25), p < 0.001
      Recovery support0.68 (0.31, 1.47), p = 0.3242.47 (1.05, 5.81), p = 0.038
      IW, inpatient withdrawal; RR, residential rehabilitation; IW + RR, inpatient withdrawal and residential rehabilitation; DDD, drinks per drinking day; AUD, alcohol use disorder.
      Numbers in table are relative risks (95% confidence intervals) and p-values.
      Referent group in multinomial model is: Inpatient (n = 2682).
      a Recorded at triage assessment with reference to the preceding 28 days as appropriate.
      b Referent category: deprivation first quartile.
      c Referent category: high-extreme, recoded at treatment commencement with reference to the preceding 28 days were appropriate.
      d Referent category: health service.
      The cohort accessed 171 specialist IW and RR treatment services (median of 7 clients per agency, IQR 2–23). Median duration of treatment exposure for the “Combined group” (25 weeks; IQR 13–41) was greater than that of the “IW group” (6 weeks, IQR 2–20), with the “RR group” reporting a median of 13 weeks (IQR 6–24) in treatment.
      Half (53%) of the cohort received structured community-based treatment prior to the earliest IW or RR admission within their index treatment journey (Table 1). The relative risk of prior community-based treatment was lower for the “RR group” relative to the “IW group”. A similar proportion (52%) received community-based treatment subsequent to IW or RR cessation, although delivery was less likely for the ‘RR’ and ‘Combined’ groups. Two-thirds (66%) received recovery support within their treatment journey.
      Three-fifths of the sample (60%) successfully completed their index treatment journey within 12 months of admission to IW or RR; nearly one-fifth (17%) were unsuccessfully referred to an agency, one-sixth (15%) had an unplanned discharge, and less than one-tenth (7%) were still in their index treatment journey at the end of observation (Table 3).
      Table 3Treatment status and outcome at 6-month follow-up (n = 3812).
      Status/outcomeIW

      (n = 2682)
      RR

      (n = 915)
      Combined

      (IW + RR; n = 215)
      Total

      (n = 3812)
      Treatment status
       No. (%) successfully completed1560 (58.2)595 (65.0)123 (57.2)2278 (59.8)
       No. (%) unsuccessful agency transfer504 (18.8)97 (10.6)34 (15.8)635 (16.7)
       No. (%) still in treatment207 (7.7)47 (5.1)29 (13.5)283 (7.4)
       No. (%) dropped out369 (13.8)175 (19.1)27 (12.6)571 (15.0)
       No. (%) prison terminated treatment1 (0)0 (0)0 (0)1 (0)
       No. (%) died in treatment41 (1.5)1 (0.1)2 (0.9)44 (1.2)
      Outcome (excluding those still in treatment; n = 3529)
       No. (%) SCNR outcome1417 (57.3)554 (63.8)105 (56.5)2076 (58.8)
      SCNR, successful completion of treatment and no representation within 6 months; IW, inpatient withdrawal; RR, residential rehabilitation; IW + RR, inpatient withdrawal and residential rehabilitation.
      Relative to the “IW group”, the “RR group” were less likely to be recorded as having an unsuccessful transfer between agencies (RR 0.51, 95%CI 0.27, 0.96, p = 0.037). Further, the “RR group” had a lower risk ratio (RR 0.60, 95%CI 0.36, 0.98, p = 0.041), and the “Combined group” had a higher risk ratio (RR 1.78, 95%CI 1.07, 2.96, p = 0.028), of being retained in the index treatment journey 12 months post-admission.

      3.3 SCNR outcome

      Three-fifths (59%) of those who left treatment within 12 months of commencement (i.e., excluding those still in treatment) achieved SCNR (IW: 57%; RR: 64%; Combined: 57%).
      A multi-level, mixed effects, multivariable logistic model for the ‘IW group’ showed that being older, engaged in paid employment, and receiving community-based treatment prior and subsequent to IW were associated with greater likelihood of the SCNR outcome (Table 4). Having a housing problem was a negative predictor of the SCNR outcome. Notably, the odds of the SCNR outcome were 59% higher for those who received community-based treatment prior to IW, and 47% higher for those who received community-based treatment subsequent to IW. No such association was evident for provision of recovery support throughout the treatment journey.
      Table 4Multi-level, mixed effects, multivariable logistic model of SCNR outcome for inpatient (n = 2475) and residential rehabilitation (n = 1054) samples.
      CovariateIW

      (n = 2475)
      RR/Combined

      (n = 1054)
      Socio-demographic
      Recorded at triage assessment with reference to the preceding 28days as appropriate.
       Male0.93 (0.77, 1.13), p = 0.4840.84 (0.62, 1.15), p = 0.277
       Age1.01 (1.01, 1.02), p = 0.0021.00 (0.99, 1.01), p = 0.989
       Black or minority ethnic group0.90 (0.67, 1.21), p = 0.4861.51 (0.92, 2.49), p = 0.103
      Deprivation quintile
      Referent category: deprivation first quartile.
       21.11 (0.81, 1.51), p = 0.5121.16 (0.76, 1.78), p = 0.491
       30.99 (0.73, 1.34), p = 0.9291.02 (0.64, 1.64), p = 0.928
       40.96 (0.71, 1.30), p = 0.7850.78 (0.49, 1.24), p = 0.293
       5 (most deprived)1.03 (0.76, 1.41), p = 0.8300.87 (0.54, 1.40), p = 0.555
       Housing problems0.67 (0.51, 0.88), p = 0.0040.95 (0.65, 1.37), p = 0.771
       Paid employment1.32 (1.03, 1.70), p = 0.0281.52 (1.02, 2.26), p = 0.038
      Clinical
      Recorded at triage assessment with reference to the preceding 28days as appropriate.
      DDD group
      Referent category: high-extreme, recoded at treatment commencement with reference to the preceding 28days were appropriate.
       Abstinent1.29 (0.78, 2.16), p = 0.3240.69 (0.42, 1.13), p = 0.139
       Low-high1.07 (0.82, 1.40), p = 0.6151.10 (0.69, 1.74), p = 0.700
       Extreme0.90 (0.73, 1.10), p = 0.3010.95 (0.67, 1.34), p = 0.749
      Previous treatment for AUD0.90 (0.75, 1.08), p = 0.2530.82 (0.61, 1.10), p = 0.177
      Referral source
      Referent category: health service.
       Self/family member/friend1.22 (0.97, 1.54), p = 0.0901.89 (1.37, 2.60), p < 0.001
       Criminal justice1.23 (0.52, 2.89), p = 0.6360.76 (0.26, 2.24), p = 0.616
      Treatment exposure
       Days IW exposure1.00 (1.00, 1.00), p = 0.290
       Weeks RR exposure1.09 (1.07, 1.11), p < 0.001
       Received IW0.60 (0.38, 0.94), p = 0.026
       Any prescribing in RR1.18 (0.80, 1.73), p = 0.405
       Prior structured community-based treatment1.59 (1.17, 2.17), p = 0.0030.67 (0.45, 0.99), p = 0.047
       Subsequent structured community-based treatment1.47 (1.11, 1.96), p = 0.0081.69 (1.10, 2.60), p = 0.016
       Recovery support1.06 (0.82, 1.37), p = 0.6810.69 (0.48, 1.00), p = 0.047
      Model parameters
       Treatment clinic (ICC)
      Range for imputed models reported here.
      0.17, 0.180.09, 0.10
       Constant0.41 (0.23, 0.74)0.76 (0.32, 1.78)
       Wald
      Range for imputed models reported here.
      104.34, 106.4299.61, 104.26
       LR
      Range for imputed models reported here.
      171.69, 176.1313.38, 15.33
      SCNR, successful completion of treatment and no representation within 6 months; IW, inpatient withdrawal; RR, residential rehabilitation; IW + RR, inpatient withdrawal and residential rehabilitation; DDD, drinks per drinking day; AUD, alcohol use disorder; ICC, intra-class correlation; LR, Likelihood-Ratio Test.
      Numbers in table are adjusted odds ratios (95% confidence intervals) with p-values. These outputs are based on multiple imputation.
      a Recorded at triage assessment with reference to the preceding 28 days as appropriate.
      b Referent category: deprivation first quartile.
      c Referent category: high-extreme, recoded at treatment commencement with reference to the preceding 28 days were appropriate.
      d Referent category: health service.
      e Range for imputed models reported here.
      For the RR analysis, a multi-level, mixed effects, multivariable logistic model (i.e., ‘RR group’ and ‘Combined group’) showed that being engaged in paid employment, self/family/peer referral to treatment, a longer duration of RR treatment exposure, and receipt of community-based treatment subsequent to RR were associated with greater likelihood of achieving the SCNR outcome. Receiving community-based treatment prior to RR, and IW in the same treatment journey as RR, were associated with a lower likelihood of this outcome (Table 4). Notably, the odds of SCNR were 69% higher for those who received structured outpatient intervention after RR; no such association was evident for provision of recovery support.

      3.4 Sensitivity analyses

      Models for IW and RR groups were repeated using complete-case data (Table S1) and were comparable. We calculated the E-value for a common outcome (where the outcome >15%). The E-value is an estimate of the minimum strength of association that an unmeasured confounder would need to account for a treatment-outcome association, conditional on the included covariates (
      • VanderWeele T.J.
      • Ding P.
      Sensitivity analysis in observational research: Introducing the E-value.
      ).
      For the IW model, the magnitude of the observed AOR for prior structured community-based treatment (1.59) and for subsequent structured community-based treatment (1.47) was relatively protected from influence of unmeasured confounding. The E-value indicated that these treatment-outcome associations could be explained away by an unmeasured confounder, but only one that that was associated with both the treatment and the outcome with a risk ratio of 1.83 and 1.72, respectively. Weaker unmeasured confounding could not do so, although the confidence intervals could be moved towards the null with smaller risk ratios (1.38 and 1.29, respectively). For the RR model, the observed AORs of 0.67 for prior community-based treatment and 1.69 for subsequent community-based treatment could be explained away by E-values of 1.74 and 1.92, respectively and with relative modest impact of the confidence intervals towards the null (1.08 and 1.28, respectively).

      4. Discussion

      The aim of this study was to estimate the clinical effectiveness of publicly-funded IW and RR treatment for AUD in England. In this national cohort, categorisation of patients who entered a form of accommodation-based treatment revealed that the majority received treatment in an IW setting; less than one-third entered a RR facility. IW is recommended where individuals are at risk of alcohol withdrawal seizures or delirium tremens, and thus require medically-assisted alcohol withdrawal and 24 h assessment and monitoring (
      • National Institute for Health and Care Excellence
      Alcohol-use disorders: diagnosis, assessment and management of harmful drinking and alcohol dependence.
      ). RR is typically recommended where the individual may not have stable housing and/or may require intensive treatment longer-term (
      • National Institute for Health and Care Excellence
      Alcohol-use disorders: diagnosis, assessment and management of harmful drinking and alcohol dependence.
      ). Whilst pharmacological intervention may be offered in some RR facilities, abstinence on treatment commencement can be a requirement. Indeed, the current study showed that those who received RR were more likely to be homeless or at housing risk, had a longer duration of treatment exposure, and were more likely to report abstinence at triage, relative to those who received IW.
      Notably, half the sample had received outpatient treatment prior to IW/RR, with greater prevalence among IW patients. Outpatient-based community-assisted withdrawal programs with psychosocial intervention and a medication regime where necessary are recommended as the first-line response to AUD (
      • National Institute for Health and Care Excellence
      Alcohol-use disorders: diagnosis, assessment and management of harmful drinking and alcohol dependence.
      ). Half the sample received structured outpatient treatment after discharge from accommodation-based treatment, and two-thirds received some form of recovery support, showing general support for recommendations regarding engagement in continuing care in outpatient settings (
      • National Institute for Health and Care Excellence
      Alcohol-use disorders: diagnosis, assessment and management of harmful drinking and alcohol dependence.
      ).
      Over half of those who left treatment (planned or unplanned) achieved SCNR, with a higher rate among those who received RR versus IW. The present study is not able to shed light on the relative efficacy of treatment here given that participants were not randomly allocated to settings. However, IW is often considered the first step in the treatment journey, focused purely on withdrawal, whilst RR is often focused on maintenance of abstinence (
      • National Institute for Health and Care Excellence
      Alcohol-use disorders: diagnosis, assessment and management of harmful drinking and alcohol dependence.
      ). Further, IW is typically a week or so in duration, and RR several months. Longer duration of treatment, as opposed to treatment intensity, is associated with better short and long-term outcomes, including abstinence, severity of dependence and broader psychosocial wellbeing (
      • Moos R.H.
      • Moos B.S.
      Long-term influence of duration and intensity of treatment on previously untreated individuals with alcohol use disorders.
      ).
      The analysis of correlates of SCNR for those who received IW and RR revealed that longer duration of exposure to IW or RR predicted greater likelihood of SCNR after covariate control. Previous AUD treatment exposure was associated with lower odds of SCNR, aligning with current understanding of AUD as a chronic, remitting condition (
      • Grahn R.
      • Chassler D.
      • Lundgren L.
      Repeated addiction treatment use in Sweden: A national register database study.
      ). Receipt of community-based treatment following discharge from IW or RR, what we would term here ‘continuing care’, was associated with greater odds of SCNR across both subsamples. Continuing care has been defined as treatment which follows a period of more intensive care, typically IW/RR or intensive outpatient care, intended to maintain progress and provide support for continued engagement in other recovery activities (
      • McKay J.R.
      Continuing care research: What we've learned and where we're going.
      ). Whilst there is considerable variability in patient response, there is evidence to support a positive effect of continuing care in enhancing positive short- and long-term treatment outcomes (
      • McKay J.R.
      The role of continuing care in outpatient alcohol treatment programs.
      ,
      • McKay J.R.
      Continuing care research: What we've learned and where we're going.
      ). Indeed, several studies have shown increased rates of abstinence over time with provision of structured outpatient psychosocial treatment following RR and IW for problematic alcohol and other drug use (
      • Gossop M.
      • Harris J.
      • Best D.
      • Man L.-H.
      • Manning V.
      • Marshall J.
      • Strang J.
      Is attendance at alcoholics anonymous meetings after inpatient treatment related to improved outcomes? A 6-month follow-up study.
      ;
      • Gossop M.
      • Stewart D.
      • Marsden J.
      Attendance at narcotics anonymous and alcoholics anonymous meetings, frequency of attendance and substance use outcomes after residential treatment for drug dependence: A 5-year follow-up study.
      ;
      • Kim J.W.
      • Choi Y.S.
      • Shin K.C.
      • Kim O.H.
      • Lee D.Y.
      • Jung M.H.
      • Choi I.G.
      The effectiveness of continuing group psychotherapy for outpatients with alcohol dependence: 77-month outcomes.
      ;
      • Ouimette P.C.
      • Moos R.H.
      • Finney J.W.
      Influence of outpatient treatment and 12-step group involvement on one-year substance abuse treatment outcomes.
      ).
      We did not observe a significant association between provision of recovery support and SCNR for IW or RR subsamples. Recovery support can be delivered and recorded outside structured treatment, and comprises activities targeted at relapse prevention (e.g., periodic contact between provider and client regarding recovery progress) and supports for broader functioning (e.g., employment, housing, parenting, family). It should be noted that non-structured recovery support is not captured within NDTMS, nor is services offered outside of reporting agencies, so rates of engagement in recovery support may be under-ascertained.

      4.1 Strengths and limitations

      Past research exploring outcomes following AUD treatment within IW and RR settings is typically restricted to prospective cohort studies, focusing on achievement of abstinence, as well as retention and successful completion of treatment (
      • Gossop M.
      • Harris J.
      • Best D.
      • Man L.-H.
      • Manning V.
      • Marshall J.
      • Strang J.
      Is attendance at alcoholics anonymous meetings after inpatient treatment related to improved outcomes? A 6-month follow-up study.
      ). To the best of our knowledge, this is the first study of IW and RR treatment outcomes, as conceptualised, using a national monitoring system.
      We acknowledge several study limitations. Firstly, the sample was restricted to those who reported only problematic alcohol use. Whilst existing research indicates high comorbidity between AUD and other substance disorders (
      • Stinson F.S.
      • Grant B.F.
      • Dawson D.A.
      • Ruan W.J.
      • Huang B.
      • Saha T.
      Comorbidity between DSM-IV alcohol and specific drug use disorders in the United States: Results from the national epidemiologic survey on alcohol and related conditions.
      ), annual 2014/15 NDTMS data (
      • Public Health England
      Adult substance misuse statistics from the national drug treatment monitoring system (NDTMS): 1 April 2014 to 31 March 2015.
      ) indicated that alcohol only patients formed 30% of the total treatment population, and patients with comorbid alcohol and non-opiate problematic use only 10%. Second, NDTMS lacks research measures of self-efficacy, personality traits and specific co-existing mental health conditions, so there is relatively limited opportunity for confounder control and evaluation of outcome mediation, and these are recognised as potential unmeasured confounders (
      • Adamson S.J.
      • Sellman J.D.
      • Frampton C.M.
      Patient predictors of alcohol treatment outcome: A systematic review.
      ). However, we included an analysis of the strength of unmeasured confounding (using the E-value) and found that this was not likely to represent a threat to the adjusted models; although the assessment of unmeasured confounding for the RR model did point to some imprecision in the confidence intervals for the intervention-outcome associated identified.
      Third, we were not able to link treatment records to national deaths registry or the national prisons system, meaning that we could not censor data for barriers to treatment re-presentation, and privately-funded treatment may not be captured in NDTMS. Given that only 1.2% of the sample ceased treatment within 12 months of admission due to mortality, we do not anticipate that the rate of SCNR would decrease substantially accounting for these factors.
      Fourth, SCNR outcome is a proxy for remission. It does not include people who have relapsed but have, for whatever reason, not re-presented to treatment. This sub-group are detectable by research studies with research follow-up, but are not detectable by data registries such as NDTMS. Fifth, whilst our definition of SCNR matched national reporting standards (
      • Public Health England
      National drug treatment monitoring system (NDTMS): Adult drug treatment business definition.
      ), further research using longer period to identify relapse (≥12 months) could evaluate benefit in the longer-term. Finally, it should be emphasised that this study does not comprise a comparison of treatment outcomes for IW versus RR given that patients were not randomised to setting.

      5. Conclusion

      In this first national effectiveness study of AUD in the English public treatment system for alcohol-use disorders, 59% of patients successfully completed treatment within 12 months and did not represent for more treatment within six months. Greater likelihood of SCNR with longer treatment exposure aligns with current literature suggesting that the duration of treatment may be critical in determining outcomes. Further, provision of continuing care in the form of structured outpatient intervention was associated with greater likelihood, and previous history of AUD treatment was associated with lower likelihood, of successful completion without re-presentation. Taken together, these findings reinforce current conception of AUD as a chronic condition, whereby continued provision of support over time may delay the time until relapse.

      Contributors

      The design and statistical analysis plan for this study was developed and implemented by A.P., J.M. and B.E. Data validation was implemented by B.E, P.H., T.L and P.W. The analysis was implemented by A.P., B.E and J.M. A.P. wrote the first draft of the manuscript, with J.M. and B.E. drafting further versions. All authors reviewed and approved the final manuscript.

      Conflict of interests

      B.E. is enrolled at a part-time PhD programme at King's College London. He is employed full-time at Public Health England with the Evidence Application Team, Alcohol, Drugs and Tobacco Division, Public Health England.
      A.P. is supported by a National Health and Medical Research Council Early Career Fellowship (#APP1109366). A.P. works at the National Drug and Alcohol Research Centre, which is funded by the Australian Government as part of the National Drug Strategy. She has received untied educational grants from Mundipharma and Seqirus for post-marketing surveillance of pharmaceutical opioid formulations.
      J.M. works in an integrated university and National Health Service academic health sciences centre (Institute of Psychiatry, Psychology and Neuroscience [IoPPN], King's College London and King's Health Partners). He is supported by research grants from the Department of Health, Institute for Health Research (NIHR), and the NIHR Biomedical Research Centre for Mental Health at South London and Maudsley NHS Mental Health Foundation Trust (SLaM MHFT) and has part-time employment as Senior Academic Advisor for the Alcohol, Drugs and Tobacco Division, Health and Wellbeing Directorate, Public Health England. He declares untied educational grant funding from the pharmaceutical industry at IoPPN and SLaM MHFT for a study of psychological interventions in opioid maintenance (2010–2016; Indivior PLC via Action on Addiction). In the past three years he has received honoraria from Merck Serono in 2015 (clinical oncology medicine) and from Indivior via PCM Scientific in relation to the Improving Outcomes in Treatment of Opioid Dependence conference (co-chair, 2015; 2016; chair: 2017). He holds no stocks in any company.
      All other authors have no disclosures in relation to this article.

      Role of funding source

      The study was supported by the Alcohol, Drugs and Tobacco Division, Health Improvement Directorate, Public Health England. The contents of this article do not necessarily reflect the views or stated position of PHE.

      Acknowledgements

      We thank Craig White and Jez Stannard (Public Health England) for their comments on the analytic approach.

      Appendix A. Supplementary data

      References

        • Adamson S.J.
        • Sellman J.D.
        • Frampton C.M.
        Patient predictors of alcohol treatment outcome: A systematic review.
        Journal of Substance Abuse Treatment. 2009; 36: 75-86
        • Allen J.
        • Mattson M.
        • Miller W.
        • Tonigan J.
        • Connors G.
        • Rychtarik R.
        • Litt M.
        Matching alcoholism treatments to client heterogeneity.
        Journal of Studies on Alcohol. 1997; 58: 7-29
        • Alterman A.I.
        • Langenbucher J.
        • Morrison R.L.
        State-level treatment outcome studies using administrative databases.
        Evaluation Review. 2001; 25: 162-183
        • American Psychiatric Association
        Diagnostic and statistical manual of mental disorders (DSM-5®).
        American Psychiatric Association, 2013
        • Benchimol E.I.
        • Smeeth L.
        • Guttmann A.
        • Harron K.
        • Moher D.
        • Petersen I.
        • Committee R.W.
        The reporting of studies conducted using observational routinely-collected health data (RECORD) statement.
        PLoS Medicine. 2015; 12e1001885
        • Brorson H.H.
        • Arnevik E.A.
        • Rand-Hendriksen K.
        • Duckert F.
        Drop-out from addiction treatment: A systematic review of risk factors.
        Clinical Psychology Review. 2013; 33: 1010-1024
        • Department for Communities and Local Government
        The english index of multiple deprivation (IMD) 2015.
        (Retrieved from:)
        • Eastwood B.
        • Strang J.
        • Marsden J.
        Effectiveness of treatment for opioid use disorder: A national, five-year, prospective, observational study in England.
        Drug and Alcohol Dependence. 2017; 176: 139-147
        • Epstein E.E.
        • Drapkin M.L.
        • Yusko D.A.
        • Cook S.M.
        • McCrady B.S.
        • Jensen N.K.
        Is alcohol assessment therapeutic? Pretreatment change in drinking among alcohol-dependent women.
        Journal of Studies on Alcohol. 2005; 66: 369-378
        • Finigan M.
        Societal outcomes and cost savings of drug and alcohol treatment in the state of Oregon. Exploring the key components of drug courts.
        Office of Alcohol and Drug Abuse Programs, Portland1996
        • Gossop M.
        • Harris J.
        • Best D.
        • Man L.-H.
        • Manning V.
        • Marshall J.
        • Strang J.
        Is attendance at alcoholics anonymous meetings after inpatient treatment related to improved outcomes? A 6-month follow-up study.
        Alcohol and Alcoholism. 2003; 38: 421-426
        • Gossop M.
        • Stewart D.
        • Marsden J.
        Attendance at narcotics anonymous and alcoholics anonymous meetings, frequency of attendance and substance use outcomes after residential treatment for drug dependence: A 5-year follow-up study.
        Addiction. 2008; 103: 119-125
        • Graham J.W.
        • Olchowski A.E.
        • Gilreath T.D.
        How many imputations are really needed? Some practical clarifications of multiple imputation theory.
        Prevention Science. 2007; 8: 206-213
        • Grahn R.
        • Chassler D.
        • Lundgren L.
        Repeated addiction treatment use in Sweden: A national register database study.
        Substance Use & Misuse. 2014; 49: 1764-1773
        • Haber P.
        • Lintzeris N.
        • Proude E.
        • Lopatko O.
        Guidelines for the treatment of alcohol problems.
        Australian Government Department of Health and Ageing, Sydney2009
        • Hasin D.S.
        • Stinson F.S.
        • Ogburn E.
        • Grant B.F.
        Prevalence, correlates, disability, and comorbidity of DSM-IV alcohol abuse and dependence in the United States: Results from the national epidemiologic survey on alcohol and related conditions.
        Archives of General Psychiatry. 2007; 64: 830-842
        • Kim J.W.
        • Choi Y.S.
        • Shin K.C.
        • Kim O.H.
        • Lee D.Y.
        • Jung M.H.
        • Choi I.G.
        The effectiveness of continuing group psychotherapy for outpatients with alcohol dependence: 77-month outcomes.
        Alcoholism: Clinical and Experimental Research. 2012; 36: 686-692
        • Little R.J.A.
        • Little D.B.
        Statistical analysis with missing data.
        2nd ed. Wiley, New York2002
        • Maas C.J.
        • Hox J.J.
        Sufficient sample sizes for multilevel modeling.
        Methodology. 2005; 1: 86-92
        • Marsden J.
        • Eastwood B.
        • Bradbury C.
        • Dale-Perera A.
        • Farrell M.
        • Hammond P.
        • Group, N. D. T. M. S. O. S
        Effectiveness of community treatments for heroin and crack cocaine addiction in England: A prospective, in-treatment cohort study.
        The Lancet. 2009; 374: 1262-1270
        • Marsden J.
        • Eastwood B.
        • Jones H.
        • Bradbury C.
        • Hickman M.
        • Knight J.
        • White M.
        Risk adjustment of heroin treatment outcomes for comparative performance assessment in England.
        Addiction. 2012; 107: 2161-2172
        • Marsden J.
        • Farrell M.
        • Bradbury C.
        • Dale-Perera A.
        • Eastwood B.
        • Roxburgh M.
        • Taylor S.
        Development of the treatment outcomes profile.
        Addiction. 2008; 103: 1450-1460
        • McKay J.R.
        The role of continuing care in outpatient alcohol treatment programs.
        in: Alcoholism. Springer, 2002: 357-372
        • McKay J.R.
        Continuing care research: What we've learned and where we're going.
        Journal of Substance Abuse Treatment. 2009; 36: 131-145
        • Moos R.H.
        • Moos B.S.
        Long-term influence of duration and intensity of treatment on previously untreated individuals with alcohol use disorders.
        Addiction. 2003; 98: 325-338
        • Moos R.H.
        • Moos B.S.
        • Andrassy J.M.
        Outcomes of four treatment approaches in community residential programs for patients with substance use disorders.
        Psychiatric Services. 1999; 50: 1577-1583
        • National Institute for Health and Care Excellence
        Alcohol-use disorders: diagnosis, assessment and management of harmful drinking and alcohol dependence.
        (Retrieved from:)
        • Odlaug B.
        • Gual A.
        • DeCourcy J.
        • Perry R.
        • Pike J.
        • Heron L.
        • Rehm J.
        Alcohol dependence, co-occurring conditions and attributable burden.
        Alcohol and Alcoholism. 2016; 51: 201-209
        • Ouimette P.C.
        • Moos R.H.
        • Finney J.W.
        Influence of outpatient treatment and 12-step group involvement on one-year substance abuse treatment outcomes.
        Journal of Studies on Alcohol. 1998; 59: 513-522
        • Peacock A.
        • Eastwood B.
        • Jones A.
        • Millar T.
        • Knight J.
        • Horgan P.
        • Marsden J.
        Effectiveness of psychosocial and pharmacological treatment for alcohol use disorder: A national observational, cohort study of community setting interventions in England.
        2018 (under review)
        • Peduzzi P.
        • Concato J.
        • Kemper E.
        • Holford T.R.
        • Feinstein A.R.
        A simulation study of the number of events per variable in logistic regression analysis.
        Journal of Clinical Epidemiology. 1996; 49: 1373-1379
        • Public Health England
        Adult substance misuse statistics from the national drug treatment monitoring system (NDTMS): 1 April 2014 to 31 March 2015.
        Public Health England, London2015
        • Public Health England
        National drug treatment monitoring system (NDTMS): Adult drug treatment business definition.
        Public Health England, London2015
        • Public Health England
        Public health outcomes framework: 2.15iii successful completion of alcohol treatment.
        Public Health England, London2016
        • Rahhali N.
        • Millier A.
        • Briquet B.
        • Laramée P.
        • Aballéa S.
        • Toumi M.
        • Daeppen J.-B.
        Modelling the consequences of a reduction in alcohol consumption among patients with alcohol dependence based on real-life observational data.
        BMC Public Health. 2015; 15: 1
        • Rehm J.
        • Mathers C.
        • Popova S.
        • Thavorncharoensap M.
        • Teerawattananon Y.
        • Patra J.
        Global burden of disease and injury and economic cost attributable to alcohol use and alcohol-use disorders.
        Lancet. 2009; 373: 2223-2233
        • Rehm J.
        • Shield K.D.
        • Gmel G.
        • Rehm M.X.
        • Frick U.
        Modeling the impact of alcohol dependence on mortality burden and the effect of available treatment interventions in the European Union.
        European Neuropsychopharmacology. 2013; 23: 89-97
        • Rubin D.B.
        Multiple imputation for non-response in surveys.
        John Wiley and Sons, New York1987
        • Sobell L.C.
        • Sobell M.B.
        • Leo G.I.
        • Cancilla A.
        Reliability of a timeline method: Assessing normal drinkers' reports of recent drinking and a comparative evaluation across several populations.
        British Journal of Addiction. 1988; 83: 393-402
        • Sterne J.A.
        • White I.R.
        • Carlin J.B.
        • Spratt M.
        • Royston P.
        • Kenward M.G.
        • Carpenter J.R.
        Multiple imputation for missing data in epidemiological and clinical research: Potential and pitfalls.
        British Medical Journal. 2009; 338b2393
        • Stinson F.S.
        • Grant B.F.
        • Dawson D.A.
        • Ruan W.J.
        • Huang B.
        • Saha T.
        Comorbidity between DSM-IV alcohol and specific drug use disorders in the United States: Results from the national epidemiologic survey on alcohol and related conditions.
        Drug and Alcohol Dependence. 2005; 80: 105-116
        • UK Home Office
        The government's alcohol strategy.
        (Retrieved from)
        • United Nations
        Sustainable development goals.
        (Retrieved from)
        • VanderWeele T.J.
        • Ding P.
        Sensitivity analysis in observational research: Introducing the E-value.
        Annals of Internal Medicine. 2017; 167: 268-274
        • White M.
        • Burton R.
        • Darke S.
        • Eastwood B.
        • Knight J.
        • Millar T.
        • Marsden J.
        Fatal opioid poisoning: A counterfactual model to estimate the preventive effect of treatment for opioid use disorder in England.
        Addiction. 2015; 110: 1321-1329
        • Willey H.
        • Eastwood B.
        • Gee I.L.
        • Marsden J.
        Is treatment for alcohol use disorder associated with reductions in criminal offending? A national data linkage cohort study in England.
        Drug and Alcohol Dependence. 2016; 161: 67-76
        • Witkiewitz K.
        • Finney J.W.
        • Harris A.H.
        • Kivlahan D.R.
        • Kranzler H.R.
        Recommendations for the design and analysis of treatment trials for alcohol use disorders.
        Alcoholism: Clinical and Experimental Research. 2015; 39: 1557-1570
        • World Health Organisation
        Global status report on alcohol and health 2014.
        World Health Organisation, Geneva2014