Background Drug-related deaths are increasing in many countries and death rates in Scotland are among the highest in Europe. Better understanding of the clustering of health conditions and social experiences, and the wider social and political factors surrounding drug-related deaths may improve the design of effective policies and interventions. Systems science methods provide an opportunity to better understand the interactions between multiple factors to support intervention development. We aimed to: (1) develop a broader understanding of the complex system relating to drug-related deaths in Scotland; and (2) identify potential action ideas for interventions.

Methods A systems-informed intervention development study following the 6 Steps in Quality Intervention Development (6SQuID) framework. We designed and facilitated a series of co-production workshops using systems science tools: critical system heuristics, variable elicitation, system mapping, multiple perspective diagrams, priority setting, and 6 cohering questions.

We linked Public Health Scotland’s National Drug-Related Deaths Database, Prescribing Information System, and Scottish Morbidity Records for inpatient stays in acute and psychiatric hospitals. The linked dataset held information relating to 6,608 drug-related deaths in Scotland between 2009 and 2018. We estimated a co-occurrence network based on binary indicators of presence/absence of a health/social condition or prescription using gaussian graphical models. We used network analysis to study the structure of the system map and linked data: Louvain community detection (which factors were in more closely connected subsystems) and degree centrality (the number of connections to other factors in the network).

Results The final system map contained 98 factors and the linked dataset had 1,367 co-occurring variables. Analysis identified sub-systems that were common across the system map and linked data (mental health and proximal causes of death), and others that were unique to one data source only (stigmatising attitudes and social relationships in system map, vein damage and assault in linked data). The most central factor in the system map was prevalence of stigmatising norms (degree centrality 18); and in the linked data assault with sharp object (14). Workshops identified three priority intervention areas: workforce development, service navigation, and community connections, with stigma as a leverage point relevant across the wider system.

Conclusion Future policy and practice could benefit from considering the mechanisms through which interventions around stigma, workforce development, service navigation and community connections may interact and influence each other to prevent drug deaths.