BACKGROUND: Globally alcohol consumption is a leading risk factor for premature death and disability and is associated with crime, social and economic consequences. Local communities may be able to play a role in addressing alcohol-related issues in their area.

OBJECTIVES: To evaluate the effectiveness and cost-benefit of an asset-based community development approach to reducing alcohol-related harm and understand the context and factors that enable or hinder its implementation.

DESIGN: A mixed-methods evaluation. Area-level quasi-experimental trial analysed using four different evaluation methods (a stepped-wedge design where each area was a control until it entered the intervention, comparison to matched local/national controls and comparison to synthetic controls), alongside process and economic evaluations.

SETTING: Ten local authorities in Greater Manchester, England.

PARTICIPANTS: The outcomes evaluation was analysed at an area level. Ninety-three lay persons representing nineareas completed questionnaires, with 12 follow-up interviews in five areas; 20 stakeholders representing ten areas were interviewed at baseline, with 17 follow-up interviews in eight areas and 26 members of the public from two areas attended focus groups.

INTERVENTIONS: Professionals in a co-ordinator role recruited and supported lay volunteers who were trained to become alcohol health champions. The champion's role was to provide informal, brief alcohol advice to the local population and take action to strengthen restrictions on alcohol availability.

MAIN OUTCOME MEASURES: Numbers of alcohol-related hospital admissions, accident and emergency attendances, ambulance call-outs, street-level crime and antisocial behaviour in the intervention areas (area size: 1600-5500 residents). Set-up and running costs were collected alongside process evaluation data exploring barriers and facilitators.

DATA SOURCES: Routinely collected quantitative data on outcome measures aggregated at the intervention area and matched control and synthetic control areas. Data from policy documents, licensing registers, meeting notes, invoices, time/cost diaries, training registers, questionnaires, interviews, reflective diaries and focus groups.

RESULTS: The intervention rolled out in nine out of ten areas, seven of which ran for a full 12 months. Areas with better-established infrastructure at baseline were able to train more champions. In total, 123 alcohol health champions were trained (95 lay volunteers and 28 professionals): lay volunteers self-reported positive impact. Champions engaged in brief advice conversations more readily than taking action on alcohol availability. There were no consistent differences in the health and crime area-level indicators between intervention areas and controls, as confirmed by using three different analysis methods for evaluating natural experiments. The intervention was not found to be cost-beneficial.

LIMITATIONS: Although the sequential roll-out order of the intervention was randomised, the selection of the intervention areas was not. Self-reported impact may have been subject to social desirability bias due to the project's high profile.

CONCLUSIONS: There was no measurable impact on health and crime outcomes. Possible explanations include too few volunteers trained, volunteers being unwilling to get involved in licensing decisions, or that the intervention has no direct impact on the selected outcomes.

FUTURE WORK: Future similar interventions should use a coproduced community outcomes framework. Other natural experiment evaluations should use methodological triangulation to strengthen inferences about effectiveness.