Complex system structures around drug-related deaths in Scotland: a mixed method network analysis approach to study subsystems in population data and system maps

Complex system structures around drug-related deaths in Scotland: a mixed method network analysis approach to study subsystems in population data and system maps | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Complex system structures around drug-related deaths in Scotland: a mixed method network analysis approach to study subsystems in population data and system maps Mark McCann, Rosemary Seaman, Srebrenka Letina, Tessa Parkes, and 7 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-8767159/v1 This work is licensed under a CC BY 4.0 License Status: Under Review Version 1 posted 8 You are reading this latest preprint version Abstract Background Drug-related deaths are rising in many countries. Understanding connections between health conditions, social experiences, and broader political factors may enhance death prevention efforts. Systems science methods can help identify areas for effective interventions. This study aimed to understand the complex system relating to drug-related deaths in Scotland. Methods We used gaussian graphical models to identify co-occurring variables in linked datasets: Public Health Scotland’s National Drug-Related Deaths Database, Prescribing Information System, and Scottish Morbidity Records for inpatient and day case stays in acute and psychiatric hospitals (n = 6,608). Preliminary findings were integrated into co-production workshops. We conducted a systems-informed intervention development study using the 6SQuID Intervention Development framework. We facilitated co-production workshops using soft systems methods and system mapping. System map factors were coded according to the social ecological model of health. Network metrics and Louvain community detection were applied to the system map and linked data graphical model. We applied a mixed method visual, text, and numeric approach to integrate findings from both data sources. Results Stigma, mental health and poverty were central factors in the system map; while population data found living arrangements and assault as central. Linked data analysis found 78 subsystems; 58 related to distinct conditions, eight to co-occurring conditions, and 12 to substance use. System map analysis found eight subsystems including direct causes of death, life experiences, stigmatising attitudes, treatment services and public perspectives. All subsystems contained factors across multiple social ecological levels, but no single system traversed all levels. Assault and alcohol treatment and harms were distinct subsystems in the linked data but less prominent in the system map; while frailty and housing were common features of both data sources. Conclusions Integrating systems science and social ecological perspectives with network analysis provides novel insights into complex health issues, with practical implications for designing interventions. Future policy and practice should consider how to align actions more closely to system-level outcomes alongside individual and clinical outcomes. Increased attention to social, community and living environments would give Scottish drug death policy more comprehensive coverage of the drug death system. Drug-related deaths co-occurring conditions comorbidity/multimorbidity syndemics co-production methods systems science data linkage network analysis systems mapping Figures Figure 1 Figure 2 Introduction Drug-related deaths (DRDs) are a growing public health concern globally. Scotland has the highest drug-related death (DRD) rate in Europe, with over 1,000 deaths annually since 2018, despite slight reductions in 2022 and 2024. ( 1 ). Multiple factors contribute to DRDs, including multimorbidity, polypharmacy, interaction with health and social services, and broader social and political factors. Evidence shows rising multimorbidity among people who use drugs, significantly increasing DRD risk. In Scotland, 70% of individuals who experienced a DRD in 2017 had at least one medical condition recorded in the six months prior to death, up from 46% in 2009. Common conditions included depression (45%), respiratory conditions (27%), and blood-borne viruses (19%) ( 2 ). Social factors may also contribute to risk of a DRD. Over half of DRDs occur among individuals living in the 20% most deprived areas of Scotland ( 2 ). Increases in DRD between 1980 and 2012 were highest in the most deprived areas ( 3 ). Other social factors associated with DRD include experiencing domestic violence (43% of women having a DRD) and living alone (52% of all DRDs) ( 2 ). Around half of those who experienced a DRD in Scotland had contact with drug treatment services in the six months prior to death. Many also interacted with healthcare (e.g., mental health services, hospital stays) and social care (e.g., housing, social work) services during this period ( 2 ). Closer inspection of patterns of co-occurring treatment could illuminate common multimorbidity patterns and implications for death prevention. There is comprehensive, biennial reporting on the prevalence of health conditions, social experiences, and services used among DRD decedents in Scotland ( 4 ). This reporting uses Public Health Scotland’s National Drug-related Deaths Database (NDRDD), which collects detailed information on DRDs. These Official Statistics draw on a range of administrative data sources to provide information on the prevalence of single conditions among people who have experienced a DRD. However, the standard reporting framework has limited scope to explore multimorbidity or integrate expertise and insights of those with lived/living experience of substance use and those working in the sector. A detailed focus on co-occurring health conditions, social experiences, and service use, combined with experiential evidence may help to support preventive strategies. Aim We aimed to develop a systems-informed understanding of factors relating to DRDs in Scotland drawing upon multiple data sources and integrating experiential and statistical evidence. To meet this aim, we developed system maps through cross-sectoral workshops and conducted analysis of linked administrative data. Research Questions RQ1: What are the most central (connected) factors within the DRD system? RQ2: What are the key subsystems of related factors within the DRD system? RQ3: What common and unique perspectives on system structure are obtained from administrative data and system maps? Better understanding of how causes of DRD are connected across levels of influence could help inform more effective actions across the system. System mapping provides an effective way to develop understanding of complex causal processes interacting at multiple levels ( 5 ) and could help to improve decision making around DRD prevention. Syndemic theory helps explain how certain health conditions co-occur and, combined with social influences, lead to amplified adverse outcomes differently from effects in isolation ( 6 ). Syndemic theory has been used to understand outcomes among people who use drugs ( 7 ), and systems approaches can play an important part in the identification of co-occurring health and social factors which may underpin emergent patterns of syndemic ill health and vulnerability ( 8 ). Network analysis methods are suitable for exploring the structural features of relationships, such as describing how commonly factors co-occur and identifying groups of interconnected factors. Instead of focusing on the prevalence or pairwise associations of variables, network analysis provides insight into the broader system structures that encompass individual experiences related to wellbeing, health, and mortality ( 9 , 10 ). By studying the connections between health and social conditions, network analysis can identify factors that may form a syndemic, helping to understand the broader social structures related to DRDs and identify effective leverage points for intervention ( 11 ). Methods and data This paper reports on data from a project aimed at defining, identifying, and prioritising action areas to prevent DRDs in Scotland. We analysed administrative data and conducted stakeholder co-production workshops. This paper covers steps 1 and 2 of the 6 Steps in Quality Intervention Development (6SQuID) process: defining and understanding the problem and its causes, and identifying which causal or contextual factors are modifiable ( 12 ). The project used a multi method design with two concurrent work streams - linked data analysis and system workshops - with two data integration points. The first integration point involved preliminary linked data analysis forming part of system workshop activities which in turn informed the development of the system map. The second integration point involved holistic triangulation of the findings from the network analysis of data from each work stream. In multi method and mixed method studies, convergent triangulation is commonly applied to enhance validity or generalisability of findings (e.g. quantitatively assessing an idea based on qualitative data, or assessing consensus across datasets). An alternative analytical approach is holistic triangulation, where multiple methods are applied in order to obtain a more comprehensive understanding of the topic under study (e.g. using multiple data sources to extend or develop a theory) ( 13 ). In this study, we applied holistic triangulation to understand the overlapping and unique perspectives provided from two data sources, applying methods that follow a different paradigm from more common quantitative and qualitative health and social science methods. For the remainder of this paper, we consider DRDs and the collective set of factors and causal influences at all levels - from individual to political - as the System of Interest. We view a system as comprising subsystems , or subsets of closely related factors, these are indicated in quotation marks. We use the term factor when referring to aspects of the broader system discussed in workshops (e.g. poverty or positive future aspirations), and variable to refer to aspects of quantitative data included in the study (e.g. Venlafaxine in prescription data). Factors and variables are indicated in italics. Data source 1: linked administrative data The National Drug-Related Deaths Database (NDRDD) holds data on most individuals whose death was recorded as a DRD in Scotland since 2009 (excluding suicides and cases with recording issues). Data collection coordinators in NHS Health Boards gather information on suspected drug deaths from pathology and toxicology reports, police sudden death reports, and local health and social work systems (e.g. contacting General Practitioners or social workers to obtain information on recent health/treatment). This information is entered into the NDRDD and, if the death matches the inclusion criteria, linked to NHS administrative health records covering community-dispensed prescriptions (Prescribing Information System, PIS), general and acute hospital admissions (Scottish Morbidity Records, SMR01), and psychiatric hospital admissions (SMR04) within six months prior to death. For our research, we accessed the NDRDD for 6,608 individuals who died from a DRD in Scotland between 2009 and 2018. Using NHS unique patient identifiers (Community Health Index or CHI numbers), we linked information from PIS, SMR01, and SMR04. Our analysis focused on binary indicators of conditions noted in the NDRDD, primary and secondary International Classification of Disease 10 (ICD-10) codes in SMR, or dispensed as prescriptions in PIS (prescriptions are free under the NHS in Scotland, a record appears in PIS whenever a medicine is dispensed). If an individual’s unique identifier was absent in one dataset, their record was set to zero for all conditions in that database. For example, individuals present in the NDRDD but not in the psychiatric admissions SMR04 dataset were assumed not to have any conditions recorded in SMR04, contributing zeroes for estimating the co-occurrence of those conditions with others in the data. We excluded variables where fewer than 20 individuals had the condition. This reduced model estimation time and was unlikely to bias results as there were few potential cases of co-occurrence. The final dataset included 1,371 variables for 6,608 individuals. Preparatory analysis: condition co-occurrence network We estimated a gaussian graphical model representing the correlations between the 1,371 conditions using the R package GGMnonreg ( 14 ). This produced a square matrix where each cell represented the partial correlation between each pair of conditions in the dataset, conditioned on other variables. The estimation used Pearson correlation which returns the Phi coefficient for binary variables. Coefficients that replicated in less than 95% of bootstrap replications, as well as all negative correlations, were fixed at zero. The retained correlations indicated pairs of conditions that co-occurred more frequently than expected due to chance. Removing factors with no co-occurrence resulted in 1,229 variables for the next stage of analysis. We excluded negative correlations for two reasons. First, causal inference methodology suggests that if two variables independently affect the likelihood of appearing in a study sample, analysis of sample data will reveal a negative correlation, even without a causal association in the wider population (collider bias) ( 15 ). It is plausible that many conditions are independently associated with risk of a DRD and thus appear in this sample, making negative coefficients inaccurate. Second, our primary focus was on co-occurrence and connections most relevant to DRDs, so excluding negative or absent correlations aligns with this purpose. As the primary aim was to understand the overall structure of co-occurrence, analysis did not use demographic variables such as gender, age at death, or area deprivation measures. The final matrix was converted to a weighted network for the analysis described below. Data source 2: co-produced system map of drug-related deaths We drew upon systems science methods to build a collective understanding of the individual, service-related, health, social and political factors surrounding DRDs. We recruited a diverse group of co-production partners, representing multiple perspectives relevant to the DRD system. Using Critical System Heuristics (CSH) ( 16 ), we identified co-production partners who could provide the following role: sources of motivation (whose interests should be served?); power (who is the decision maker/who is in a position to change the measure of improvement?); knowledge (who should be considered an expert? What counts as relevant knowledge?); and legitimation (who is witness to the interests of those affected but not involved? Who can argue the case of those not able to speak for themselves?). We mapped relevant roles onto CSH domains (Appendix 1) and identified individuals representing each role to invite to the workshops. Twenty-eight people participated in the workshop series and all provided written, informed consent. Participants were offered travel reimbursement and received no other reimbursement if attending as a salaried professional. Participants who were unsalaried such as those attending due to relevant lived/living experience or as an affected family member received honoraria to acknowledge their time. Workshop 1 (February 2022) focused on understanding the problem and its causes. Four groups worked through a series of activities: a rich picture activity ( 17 ) to visualise and discuss the various perspectives held by the participants and generate ideas on the salient factors in the DRDs system; a variable elicitation activity to generate lists of relevant factors; drawing maps where each factor was placed within a circle (node) and the perceived causal associations between factors drawn as arrows (ties) ( 5 ). Discussions were not guided around themes (e.g. individual, community, political influences), participants were encouraged to think broadly about factors. We combined the four maps by amalgamating similar concepts differently phrased into single nodes. At workshop 2 (September 2022), participants were shown preliminary analysis of the NDRDD (see Appendix 2) and then reviewed the combined system map to make further changes and additions. There was one further meeting of the author team to review and finalise the map. At workshop 3 (November 2022), participants were presented with the system map and used it to develop action ideas and identify relevant roles and organisations to take actions forward (these not reported in the current paper). No further modifications to the map were requested at the workshop. Network analysis of the system map and co-occurrence network We used the R package igraph ( 18 ), online platform Kumu.io, and yEd ( 19 ) software to visualise and analyse the linked data network and system map. We applied network analysis to calculate network metrics for each node, including degree metrics which measured how many connections each node had to other factors. In the linked data network, degree represents how many other conditions co-occurred with the condition at levels greater than expected due to chance. In the system map, degree represents how many other factors were deemed by workshop participants to be causally connected to the factor. We additionally calculated indegree (how many factors influenced this node) and outdegree (how many factors this node affected). Complex systems comprise interacting factors at multiple levels of influence from the individual to societal, each factor was categorised according to its level of influence. The classification was informed by the authors’ understanding of the levels of the social ecological model of health ( 20 ), inductively coded by MMcC and KS independently, and consensus reached after discussion (see Appendix 5). We applied a multilayer visualisation algorithm to the coded system map using the R package ggraph ( 21 ), and Appendix 5 shows the distribution of factors across the subsystems and levels of influence. To answer RQ1, we calculated betweenness centrality. Low betweenness centrality gives an indication of which factors sit on the periphery of a network with few connections to central nodes and far from all other nodes, while high betweenness centrality indicates factors occupy a central location (those that are ‘closer’ to other nodes in terms of how many ‘steps’ along ties in the network are needed to reach the node). The magnitude of centrality measures scales according to the size of the network in question: they must be interpreted relative to the other nodes in the network and are not comparable between networks. In the context of the system map, high centrality provides a measure of variables which act as potential mediators on many causal pathways in the system of interest. We do not infer causal direction from the correlational data in the administrative data network. The values of the networks metrics were not used to draw inference over points to intervene ( 22 ); network metrics provide numerical starting points for analysis of the map structures, similar to how cardinal directions orient a reader to a geographical map. To answer RQ2, we used the Louvain community detection algorithm to detect sub-systems representing groups of factors that were closely related. The Louvain algorithm optimises network modularity; it returns the number of subsystems that maximises the ratio of within-subsystem to between-subsystem connections. The authors looked at the factors included in each subsystem and applied a label to describe the common theme or factor among the variables. In the system map, KS initially coded the subsystems and this was cross validated by MMcC and RS. In the linked data network, MMcC initially coded the subsystems, these were cross-validated by SL, CM, JS, JH and AB and the final subsystem labels agreed by consensus. Consensus meetings also involved initial interpretation of the subsystems and implications for DRD prevention. The subsystems were subsequently grouped into three higher-level categories described below. Kumu was used to detect feedback loops in the system map. Summary findings were circulated to workshop delegates and feedback on presentations about the study also informed our collective interpretation. To answer RQ3, we conducted a holistic multi-method integration ( 13 ) to qualitatively assess the conceptual overlap and differences in the factors, variables and subsystems arising from the networks relating to each data source. This was achieved by iterating over the following steps: matching the thematic description of subsystems in one data source with the subsystem descriptions in the other; checking if individual nodes in one data source related to a subsystem theme in the other, revisiting workshop field notes to obtain context on discussion of nodes & connections; and visually scanning the linked data network to follow the thread on connections arising from thematic comparisons. This approach allowed us to combine insights from the most comprehensive source of quantitative data on DRDs in Scotland with the diverse perspectives and insights obtained from systems mapping. Results Co-produced system map The final system map contained 98 factors with 225 connections between them (a network density of 0.02 i.e. 2% of the potential connections between all factors), including eight sub-systems. Factors fell across nine levels of influence within a social ecological framework. Figure 1 shows the nodes and sub-systems with vertical layers representing levels of influence, from wider contextual factors to political, organisational and individual factors, with the ultimate outcome of DRD in the lowest layer. The node colour shows which nodes belong to the same subsystem of connected factors. A zoomable, interactive version of the map can be downloaded from Github and opened using yEd. To the top left, the red nodes represent the subsystem around public and workforce stigmatising attitudes which falls across the wider context, policy, organisational and interpersonal levels of the system. At the bottom of the image the grey nodes represent the subsystem around the proximal causes of death. Factors in this subsystem fall across the “thoughts and emotions”, “behavioural”, and “physical” levels of the system. Subsystems contained factors that traversed upstream and downstream levels of influence within the social ecological model, rather than containing factors at a single level (see appendix Fig. 4 for the distribution of factors across levels and subsystems). Table 1 shows the subsystems identified in the map, along with the factors with the highest degree centrality in each sub-system. Data on all factors and the subsystems are available on github ( 23 ). Appendix 5 gives a detailed description of how to interpret the subsystems, factors and metrics. The largest subsystem contained 22 factors relating to the proximal causes of death. The two most connected factors in the systems map - mental health and prevalence of stigmatising norms – each appeared in their own subsystem of closely related factors. Table 1 Subsystems returned from community detection algorithms applied to the system map, and five factors within each subsystem with highest degree Subsystem description and factors Social ecological levels within subsystem Five central factors Degree Between- ness Drug death and proximal causes (Grey nodes in Fig. 1 ) Physical, behavioural, cognitive-emotional Control and regularity of substance use 10 0.11 22 factors Initiation of drug use 8 0.07 Physical health 8 0.02 Continued use of drugs 7 0.06 Drug-related death 7 0 Life experiences and mental wellbeing Behavioural, cognitive -emotional, interpersonal, service interactions, policy, wider context Mental health and wellbeing 17 0.09 (Purple nodes in Fig. 1 ) Poverty 14 0.03 16 factors Criminalisation of drugs 12 0.03 Positive future aspirations 11 0.03 Contact with criminal justice system 9 0.04 Public and workforce stigmatising attitudes Interpersonal, service interaction, organisational, wider context Prevalence of stigmatising norms around drugs 18 0.14 (Red nodes in Fig. 1 ) Quality of treatment planning 6 0.01 16 factors Service Quality 6 0.02 Workforce development 4 0 Poor treatment of PWUD 4 0.04 Social influences on drug harms Physical, behavioural, interpersonal, service interaction, organisational, wider context Peer relationships 6 0 (Brown nodes in Fig. 1 ) Drug type 6 0.02 13 factors Exposure to drugs 5 0.04 Public versus private drug taking 5 0 Knowledge of drugs and risk among general public 4 0 Experience of services Cognitive-emotional, service interaction, organisational Attending services 9 0.01 (Blue nodes in Fig. 1 ) Retention in services 6 0.06 11 factors Collaboration between services 5 0 Missed appointments 5 0.01 Internalised stigma 4 0.01 Community influences on wellbeing Interpersonal, community Trusting community relationships 12 0.03 (Orange nodes in Fig. 1 ) Trusting family relationships 7 0.05 8 factors Recovery capital 5 0.02 Practical support 3 0 Community hubs 1 0 Public perspectives on substance use Organisational, policy, wider context Policy making environment: health focussed vs criminal justice 11 0.09 (Green nodes in Fig. 1 ) Availability of services 5 0.01 7 factors Harm reduction interventions 4 0 Relative value of professional versus peer evidence 3 0.07 Lived experience representation in policy 3 0 Safe physical environments Cognitive-emotional, policy, wider context Feeling safe 6 0.01 (Pink nodes in Fig. 1 ) Homelessness 6 0.01 5 factors Housing policy 3 0 Quality of housing 3 0 Risk of crime victimisation 3 0 Feedback loop analysis The system map included feedback loops which described how drug use initiation may lead to drug use for self-medication or functional reasons which could increase continued drug use. This increases exposure to a greater range of drug types which is then related to continued use. This corresponds to the concept of entry into a wider drug market which in the context of novel and high toxicity drugs entering the supply introduces higher risk of fatality. A further loop related to stigmatising reporting of substance use in the media which may drive negative perceptions and poor treatment of people who use drugs. This increases the prevalence of stigmatising norms that loops back into further stigmatising media reports. Linked administrative data co-occurrence network The linked data network contained 1,229 variables and 1,950 connections (density = 0.003) grouped into 78 communities which we describe as subsystems below. Table 2 lists the labels and the number of variables in each subsystem in three groups: eight related to co-occurring conditions, 12 with conditions related to substance use, and 32 related to distinct conditions or forms of treatment. There were a further 24 subsystems containing only two variables reported in Appendix Table 3. Table 2 Description of subsystems and variables per subsystem identified by Louvain community detection on co-occurrence network for linked hospital records, prescription and drug deaths database Co-occurring conditions and treatment Subsystem Number of variables Subsystem Number of variables Respiratory infection, depression, pain 82 Mixed allergy, asthma medications 42 Assault and self-harm 77 Antidepressants and OCD 4 Mixed stomach, skin, antibiotics 47 Skin cream and ibuprofen 2 Mixed common prescriptions: skin, stomach, sinuses 42 Musculoskeletal/Skin 2 Substance use related conditions and treatment Subsystem Number of variables Subsystem Number of variables Mixed NDRDD group 110 Alcohol, fungal infection & stomach prescriptions 42 Accidental and intentional self-poisoning 70 Alcohol dependence and gastritis 41 Liver cirrhosis, malnourishment 63 Alcohol treatment and detoxification 32 Phlebitis, hepatitis, mental health, substance use 56 Suboxone 4 Benzos and pain 56 Buprenorphine 2 Methadone and common prescriptions 44 Fall and drug use noted at hospital 2 Distinct conditions and medications Subsystem Number of variables Subsystem Number of variables Mental health inpatient treatment 87 Nausea tablets 2 Respiratory disease 43 Dermol shower cream 2 Heart disease 38 Magnesium sulphate paste 2 Morphine, constipation, haemorrhoid treatment 38 Levetiracetam - Epilepsy 2 Epilepsy 37 Naproxen painkiller 2 Skin cream 35 Proctosedyl - haemorrhoids 2 Diabetes 35 Nicotine patches 2 Pregabalin, Duloxetine, Neuropathic pain 31 Blood clot prevention 2 Asthma inhalers 5 Fucidin antibiotic cream 2 Cardiac arrest 5 Ovarian cysts 2 Thyroid condition 4 Deviated septum 2 Venlafaxine antidepressant 3 Dental issues 2 Orchitis 3 Obesity 2 Antidepressant Dosulepin 3 No known conditions 2 Employment 3 Migraine 2 Child custody 3 Learning disability, attention deficit hyperactivity disorder 2 Anti-inflammatory medication 2 Death in prison 2 Figure 2 is a network visualisation of 24 subsystems listed in Table 2 , excluding the subsystems that did not have connections to other subsystems. The largest community contained 110 factors from the NDRDD itself, the data “spine” to which other data sources were linked. The NDRDD contains specific variables often reflecting very closely related information that was not available in SMR or PIS. For example, it was common for a toxicology report to find multiple substances and these appeared as connected within the NDRDD cluster. NDRDD variables on route of administration (oral versus injecting) were highly correlated with the toxicology reports for the respective substances, but not with other health conditions. Despite the distinct dataset effect, many variables in this subsystem connected to other subsystems e.g. toxicology variables in this subsystem co-occurred with related prescription variables located in other subsystems (see the Suboxone – NDRDD connection in Fig. 2 ). The preliminary analysis of the NDRDD database before linkage to other datasets (see Appendix 2) identified further structures nested within this subsystem, some overlapping with the main analysis (e.g. mental health, substance use treatment), and others containing unique information in the NDRDD around living arrangements and criminal justice involvement. Table 3 gives information on the two variables within each subsystem with the highest number of connections for the 15 largest subsystems. Full information on all variables is available at the github repository. Table 3 Subsystem name, number of variables, and two most highly connected variables and degree (number of connections to other variables) for the 10 largest subsystems in the linked data co-occurrence network Subsystem Number of factors ICD code where relevant variable description and (data source) degree Mixed NDRDD variables 110 Living in own home (NDRDD) 14 Currently prescribed methadone (NDRDD) 13 Mental health 87 Mental health services (NDRDD) 7 Venlafaxine, anti-depressant (PIS) 7 Respiratory infection, depression, pain 82 Propranolol, anti-anxiety (PIS) 7 Chlorhexidine mouthwash (PIS) 7 Assault and self-harm 77 X999 assault by sharp object in unspecified place (SMR) 14 S099_unspecified head injury (SMR) 13 Accidental and intentional self-poisoning 70 X619 self-poisoning by antiepileptic drug in unspecified location (SMR) 12 X610 self-poisoning by antiepileptic drug at home (SMR) 11 Liver cirrhosis 63 Ensure plus milkshake (PIS) 9 N179 acute renal failure (SMR) 8 Phlebitis, hepatitis, mental health, substance use 56 I802 phlebitis and thrombophlebitis of other deep vessels of lower extremities (SMR) 12 Hepatitis C (NDRDD) 10 Benzodiazepines and pain 56 Diazepam (NDRDD) 9 Gabapentin, analgesic/anticonvulsant (PIS) 9 Mixed stomach, skin, antibiotics 47 Flucloxicillin, antibiotic (PIS) 7 Simple linctus, cough syrup (PIS) 7 Methadone and common prescriptions 44 Methadone (PIS) 7 Paracetamol (PIS) 6 Single condition subsystems There were 58 subsystems related to distinct health conditions and common forms of treatment, including: depression; diabetes; heart disease; chronic pain; and thyroid issues. These characterise the range of health conditions that occurred among people who experienced a DRD, although this analysis does not tell us whether these are more prevalent among those that died than among the wider population. The community detection method distinguished between various aspects of mental health, for example, depression treated with prescription medications appeared in a different subsystem than inpatient psychiatric hospital stays. Co-occurring condition subsystems Eight subsystems related to multiple issues, including: assault and self-harm; respiratory infections, depression and pain; and stomach issues, skin conditions and infections. While this gives a further characterisation of some of common multimorbid conditions among the population, it also highlights potential points of intervention such as at the point of emergency room attendance after assault victimisation. Substance use related subsystems There were 12 subsystems that were directly related to substance use, including the large cluster of NDRDD variables, treatments such as Opi oid Agonist Therapy(OAT) , wound care and alcohol treatment, and also distinct co-occurrences around emergency room treatment. The cooccurrence of substance use with a fall is also notable as falling at ground level and receiving treatment could suggest high physical frailty or high levels of intoxication. This highlights a potential preventive intervention point in emergency room settings. The 10 most highly connected variables (and data source) in the co-occurrence network were, in descending order based on number of connections: living arrangements (NDRDD); assault by sharp object (SMR); methadone (OAT) (NDRDD); not currently using OAT (NDRDD); salbutamol (an asthma medication) (PIS); head injury (SMR), Phlebitis of lower leg (SMR), self-poisoning (SMR), assault by bodily force (SMR), and heroin use (NDRDD). Unlike the system map factors, the number of connections does not relate to the potential central causal role of the factors in the system. Instead, this gives an indication of how commonly each variable is related to other health and social factors. This suggests that these factors are important features of multimorbidity and potential aspects of syndemic ill health. While some factors are artefacts of the data (there were many living arrangements variables that are correlated with each other, and many ICD codes would be recorded at the same time in relation to an assault), these factors give a broad insight into factors that may predispose individuals to risk of poor health outcomes and that may be commonly recognised by health and social care providers. Common and unique perspectives from system maps and linked data While there were several overlapping themes between the factors discussed in workshops that produced the system map and the subsystems identified in the linked data, there were substantial differences in emphasis between the two. Mental health and wellbeing was a central node in the system map and there were six subsystems in the linked data related to mental illness and its treatment, suggesting that this was a prominent feature of the clinical datasets and the collective causal understanding of the wider system. Pain was mentioned as a peripheral factor in the system map, mediating the connection between physical health and drug use as a coping mechanism, while in the linked data there were several subsystems describing different aspects of pain and pain medication. Heart and lung health was mentioned in the system map as factors mediating the connection between physical health and drug death, and seven subsystems in the linked data related to heart and lung health. Linked data subsystems also uncovered connections which were not a feature of the system map, i.e. between: mental health and pain; mental health, respiratory health and pain, and the multiple patterns of co-occurrence visible in Fig. 2 . Poor treatment of people who use drugs was outlined as a key aspect of the stigma subsystem, while the linked data subsystem around assault gives insight into one aspect of negative social experiences and treatment. The linked data subsystem containing the substance use and trip and fall variables – indicative of intoxication and potential frailty – corresponded with frailty as one of the proximal factors relating to death in the system map. Re-inspection of the linked data network found that fall ICD codes also appeared as a distinct, unconnected variable (unspecified fall), part of the liver cirrhosis subsystem (fall at home), and many variables in the assault subsystem (fall in road, down stairs, from a height etc.). Lastly, there were several linked data subsystems relating to alcohol dependence, treatment and harm, but discussion of alcohol was largely absent from the workshops and did not feature as a relevant factor in the system map. This was likely due to the framing of the workshop activities with DRDs as the system of interest, rather than substance use more broadly, and the professional roles and lived experience of the workshop delegates being aligned with drug use and treatment. The two data sources also provided several distinct perspectives: the linked data gave an overview of detailed clinical characteristics of the sample beyond the scope of the system map, while the system map covered many aspects of the workforce as well as social, political, and economic factors that were not captured in the linked data. Discussion Applying a mixed method approach that integrates systems thinking, data linkage and network analysis, this study provides detail on the interconnections between factors that contribute to DRDs in Scotland, how these factors are structured, and a novel perspective on the levels of influence through which complex structures have their effects. The process of developing the systems map and co-occurrence network informed action areas for DRD prevention plans among the research team and workshop delegates, and informed wider policy discussions. The maps are openly available for re-use in systems-informed approaches to drug death prevention. Our first research question aimed to identify central factors in the DRDs system. We found stigma, mental wellbeing, and poverty in the system map, and living arrangements and assault within the linked data. Both findings point towards the important facets of the social and political environment, rather than solely on aspects of individual health or biological considerations. Targeting improvements around these areas may leverage wider change, potentially producing positive change or mitigating the negative influence of wider factors affecting risk of death. Interventions on these factors must appropriately account for the wider system structures surrounding them. Our second research question aimed to explore the substructures within the DRDs complex system. The two data sources provided broad coverage of influences at multiple levels, from the individual to political. Subsystems within the linked data correspond with common clinical experience of treatment and prescribing, and provide a comprehensive summary of the range of co-occurring health and social factors among the population who experienced a DRD. The system map drew upon a wealth of expertise and multiple perspectives and provides a broad view of the wider social and political system. The key factors were in broad agreement with wider literature on the bio, psycho, social, and political aspects of drug related harms ( 24 ), and provide a new way of contextualising the connections between the diverse areas of personal and public life that affect wellbeing thereby serving as potent points of intervention. Ingram et al. modelled the network of drug use symptoms that may be risk factors for experiencing a self-reported drugs overdose ( 25 ) finding that drug tolerance and withdrawal symptoms were central factors for risk of experiencing a self-reported drug overdose. This survey-based analysis aligns with our system map subsystem around proximal influences on DRD. Our findings also suggest avenues to explore the potential occurrence of syndemic ill health. Previous work has pointed towards worse health outcomes driven by the co-occurrence of substance use and violence, co-occurring physical and mental health issues ( 26 ), and other combinations of syndemic ill health. The subsystems identified here described related areas of health risk, as well as additional issues of co-occurrence. Future research could identify whether the condition clusters outlined in this analysis are associated with risk of poorer outcomes, which would require new data linkages with information on living individuals in addition to the data reported in this study. In addition to the comprehensive, holistic perspective on DRDs that system mapping methodologies provide ( 5 ), network analysis provides ways to explore the underlying structure of relationships and interactions that comprise the wider system. Community detection approaches to explore subsystems, in combination with a social ecological classification of factors in the system map, facilitates the development of important conceptual insights with implications for future research, policy and practice. The social ecological model and “upstream” metaphor are two common features of public health discourse around health improvement and addressing inequalities ( 4 ). These are heuristic rather than analytical tools, used to encourage the consideration of health across multiple levels ( 20 ). A complex systems framing emphasises the mutual interdependence between elements across all levels of a social system ( 5 ) and emphasises that points of greatest leverage may be found at multiple levels. Complexity framings have been used to argue against public health interventions on the basis that appropriate leverage points are too difficult to identify ( 27 ). A key contribution of our approach is to delineate subsystems of related factors that allow for tractable action planning compared to considering the system as a whole, while also going beyond a heuristic consideration of levels of influence, independent of the key structures and interactions that influence outcomes. For example, viewing the whole system map, there are many long causal chains (over 16,000) outlining processes linking stigmatising norms to death. These complex, multiple pathways which may operate over different time frames make it difficult to draw inference on the statistical association between stigma and death and thus poses problems when evaluating the effect of stigma reduction interventions on death rates. Viewing the map in terms of underlying structures, the stigma subsystem comprised factors ranging from the wider environment through to individuals’ experience of services, and the death subsystem considered physical health, drug toxicity and knowledge of drug effects. The substructures uncovered by community detection suggest that stigma interventions could be more effectively evaluated in terms of outcomes related to organisational level of the social ecological model (e.g. interactions with services), rather than in terms of death rates. Considering DRDs in terms of structurally coupled subsystems facilitates a polycontextual approach ( 28 ), meaning that the targets of action, ways of working, and data collection should vary according to the context of the subsystem, rather than applying a single perspective across all factors (e.g. statistical associations between various factors in the system and DRD). A subsystem approach may pose a challenge for health research and practice. It is simpler to focus on health at the individual level and within the body rather than evaluate change across a complex social system. Additionally, it may an expectation of public health bodies and research funders to take an individual focus. The end result is that research effort is directed towards “short causal chain” interventions, such as pharmacological interventions affecting individual causal processes, or upstream interventions with short causal pathways (e.g. Minimum Unit Pricing > purchasing behaviour > physical health) ( 29 ), that can be less prone to dilution due to multiple pathways operating over different time scales, at the expense of approaches which may more comprehensively effect system level change but which require longer time frames and more holistic forms of evaluation. Network methods used to summarise substructures could improve how systems mapping approaches inform the identification of sets of relevant indicators for change and outcome assessment at multiple levels, helping to restructure “models of evidence” ( 30 ) in systemic, rather than individual, terms. Previous system mapping exercises applied to health have often taken the approach of predefining the subsystems or domains as starting points for a system mapping exercise. For example, Finegood et. al applied network analysis to a reduced form of the Foresight obesity map to show the strength of connections between the eight clusters, including food production, physical activity, individual psychology etc. ( 31 ) While providing a helpful viewpoint of the connections between these clusters of factors, the interpretation is constrained by the starting point of categorising factors into domains. By comparison, we started with an open view of the system, going from rich pictures to various sub-systems, with community detection returning cross-disciplinary and cross-level sets of factors. We propose that system mapping activities would benefit from avoiding discipline- or theme-directed factor elicitation and instead utilise more open data collection methods, followed by analytical approaches that can uncover emergent structures. The maps provide tools to inform future policy and intervention development. For example, anyone proposing an intervention targeting one node on the system map could navigate upstream and downstream from their target node to think through potential points of constraint, resistance, or support for the actions on the target node. Policy makers focusing on reducing substance use stigma could consider the ways in which the factors in the stigma subsystem may have an influence at a national level, the extent to which local investment in community development may augment stigma reduction, or how external influences such as media representation may counteract intended effects of a policy. Visual navigation of the maps - moving upstream, downstream or around loops from any focal issue - provides a “thinking tool” to navigate any intended intervention or action plan from multiple levels and perspectives. Such an approach can enhance the quality of decision making and the identification of the most effective set of actions. The maps can also aid evaluation, particularly at policy level. For example considering the Scottish Government national drugs mission plan ( 32 ), against the system maps we find some alignment between the six cross-cutting priorities, but with notable gaps and differences in perspective. For example the mission’s focus on ‘lived and living experience at the heart’ (Priority 1) was reflected in subsystems around public perspectives on substance use, as well as stigmatising attitudes. While the mission focused on developing ways for lived experience involvement, the system map additionally highlighted issues of value and relative status for those involved with organisations. Equalities and human rights (Priority 2) did not appear as a prominent set of factors, although housing, living standards and poor treatment were evident as important factors in the system map and linked data. Tackling stigma (Priority 3) was a prominent feature of our analysis: a key insight from our analysis was to uncover the various levels at which stigma may have its effect. In addition to taking action at the level of service provider and media organisations, our findings point to the importance of tackling public perceptions and poor treatment in terms of interpersonal and community experiences. Subsystems and factors related to safe physical environments and social and community influences were not as clearly represented in the national mission. This may reflect a tendency for policy and action planning to focus either on the individual level factors (e.g. naloxone) or on upstream levels (e.g. change in national or local funding), at the expense of paying attention to the intermediate level factors which bridge micro and macro-level processes. The integration of two data sources (RQ3) uncovered a difference in focus in relation to alcohol dependence and treatment which is noteworthy from a methodological point of view: preliminary analysis of the NDRDD was presented at workshops, and alcohol treatment and detoxification was a central subsystem in this analysis (see Appendix Fig. 1). Despite this data-informed aspect of the workshop, alcohol was not strongly reflected in the subsequent development of the system map. To the authors’ knowledge, this is the first integration of linked data co-occurrence networks into a system mapping process. While there are ‘scripts’ or formal processes to document ideas from the group (e.g. Nominal group technique), no such scripts exist for data integration; the presentation of the networks involved open discussion and freehand annotation of the printed networks. Future research could develop new formal process to structure how co-occurrence networks – or other sources of data - are discussed in groups, and to document in what ways the quantitative data is represented, or purposefully omitted, from workshop final outputs. Strengths and limitations This study provides a wealth of insight on the structures and features of the complex system surrounding DRDs. The combination of linked administrative data and a coproduction systems thinking model provides a rigorous framework for understanding health outcomes in new ways. We combined these methods with the 6SQuID framework, providing a robust output for Step 1: understanding the problem and its causes, and Step 2: identifying which causal or contextual factors are modifiable, with the greatest scope for change, and who would benefit most. The workshop participants identified three action areas and started to consider specific activities in each area, partially supporting Step 3: deciding on the mechanisms of change. These can serve as a starting point for future intervention development. There are some limitations to the study. Firstly, while the system map represents the collective view of a diverse stakeholder group, it may overlook some key perspectives, offer little in-depth information in specific areas, and may not account for changes in the wider DRDs system since the time of the workshops. We would encourage readers to access the supplementary data and modify and expand them so that they are better suited to specific context and intended use. Additionally, when linking data, it was assumed that an individual who was absent from a database did not have any of the conditions noted in that database e.g. those not in SMR04 (Psychiatric inpatient) database were assumed not to have any of the psychiatric diagnoses noted in SMR04. This means that the prevalence, and thus co-occurrence for some conditions, is likely to have been underestimated. However, this limitation is partially mitigated by the inclusion of the prescription information. Community dispensing of prescriptions for mental health conditions are noted in the data and distinct mental health subsystems appeared e.g. depression treated with prescriptions and depressive disorder noted in inpatient records. An issue that remains is that those who avoid all contact with health services are under-represented in this study design. As our dataset focussed only on those who have died, there is the potential for collider bias in our dataset. If two factors are independently associated with a higher risk of death, they will appear negatively correlated in a dataset of those who have died, even if there is no correlation in the full population ( 15 ). This limitation is partly overcome as we focussed on identifying co-occurrence and removed all negative associations, rather than interpreting negative correlations as casual. However, collider bias may have the potential to underestimate the correlation between factors identified here. Further research to compare co-occurrence in living populations could investigate this possibility and better estimate the prevalence of co-occurrence and magnitude of causal associations between co-occurrence and health outcomes. Another point to note is that the use of community detection methods has some limitations. The Louvain algorithm allocates every variable into exactly one distinct community. While the most highly correlated factors reliably appear in the same community, less clearly connected factors may be assigned to one community or another at random. Running the analysis with different random number seeds found differing numbers of subsystems. While these tended to have broadly similar groupings of variables that were comparable across replications, the findings should be considered as providing broad signals around the potential structure and behaviour of highly complex systems, rather than providing deductive inference around co-occurrence patterns or causal processes. This reinforces the importance of verifying whether the subsystems and structures found in this analysis occur in other datasets, such as living individuals, and discerning which patterns of co-occurrence can be safely ruled out as being risk factors and which require further attention as potential areas of syndemic ill health. Finally, our selected methods could not account for the timing or sequence of events: our analysis represents any co-occurrence and does not consider trajectories in experience of diseases, prescribing, or social factors. The NDRDD data relates to six months prior to death, while other linked data extends further back in the life course. Future research should consider pathways via connections that occur within certain time frames, study changes over time considering age, period or cohort effects, or sequence analysis. Such research could strengthen understanding of how co-occurrence relates to elevated risk of death and shed light on further points for preventive intervention. Conclusions This study, and the open source system artefacts describing the findings from the project, provide a novel representation of the complex system surrounding a key priority for public health and health inequalities. These outputs can provide a starting point for future research and action to prevent such deaths at local intervention and national policy level. While the systems approach provided a wide range of potential intervention points, a strategic approach to DRDs requires further evidence, funding, political will to take coordinated, committed action across the wider social system. Declarations Declarations and acknowledgements For the purpose of open access, the authors have applied a Creative Commons Attribution (CC BY) licence to any Author Accepted Manuscript version arising from this submission. We are very grateful to all the participants for their contribution and insights throughout the workshop series. Many thanks to those providing feedback on the study after presentations at the Society for Social Medicine, International Network for Social Network Analysis, and Emergency Medicine at the deep end meetings. The authors would like to acknowledge the support of: the eDRIS Team (Public Health Scotland) for their involvement in obtaining approvals, provisioning, and linking data and the use of the secure analytical platform within the National Safe Haven; National Records of Scotland for data linkage; Scottish Exchange of Data (Scottish Government) for allowing us to access the data; the Scottish Drugs Forum for inviting lived experience participants; and the Drugs Research Network for Scotland for help coordinating the workshop series. We would like to acknowledge the support of the following colleagues who helped to run the workshops: Jessica Greenhalgh, Julie Riddell, Danilo Falzon, Catriona Connell, Hazel Booth, Wendy Masterson. Ethics and consent: The Public Benefit and Privacy Panel for Health and Social Care approved access to the linked data (Ref: 1920-0196), with endorsement from the University of Glasgow College of Medicine, Veterinary and Life Sciences ethics committee. Ethical approval for the systems workshops was obtained from the University of Stirling General University Ethics Panel (GUEP 19 20 861) and participants consented to take part. Consent for publication: Not applicable Availability of data and materials: Individual level linked data are not publicly available; contact researchdata.scot to make a data access request. Co-occurrence matrices, system maps and analytical scripts are available at github.com/MRC-CSO-SPHSU/NDRDD-linkage. Funding: This project was funded by the Scottish Government Chief Scientist Office (HIPS/19/32). MMcC, RS, SL and KS were part of the Relationships and Health programme in the MRC/CSO Social and Public health Sciences Unit, funded by the Medical Research Council (MC_UU_00022/3) and the Scottish Government Chief Scientist Office (SPHSU18). Author contributions: Conceptualization: MMcC, TP, LB, CM, JS, JH, AB, AC, KS. Data curation: MMcC, SL, RS. Formal analysis: MMcC, SL, KS, RS. Funding acquisition: MMcC, TP, LB, CM, JS, JH, AB, AC, KS. Investigation: MMcC, RS, SL, TP, LB, CM, JS, JH, AB, AC, KS, Jessica Greenhalgh, Julie Riddell, Danilo Falzon, Catriona Connell, Hazel Booth, Wendy Masterson. Methodology: MMcC, SL, TP, LB, CM, JS, JH, AB, AC, KS. Project administration: KS, Jessica Greenhalgh. Resources: Electronic Data Research and Innovation Service. Software: MMcC, SL, RS. Supervision: KS. Validation: MMcC, SL, TP, LB, CM, JS, JH, AB, AC, KS. Visualization: MMcC, SL, KS, RS. Writing original draft: MMcC, RS. Writing review and editing: SL, TP, LB, CM, JS, JH, AB, AC, KS. References Scotland NRo. Drug-related deaths in Scotland, 2024. 2025. Scotland PH. National drug related death database (Scotland). Analysis of deaths occurring in 2017 and 2018.; 2022. Brown D, Allik M, Dundas R, Leyland AH. All-cause and cause-specific mortality in Scotland 1981–2011 by age, sex and deprivation: a population-based study. Eur J Pub Health. 2019;29(4):647–55. Scotland NRo. Drug-related deaths in Scotland in 2022. 2022. Barbrook-Johnson P, Penn AS. Systems Mapping: How to build and use causal models of systems. Springer Nature; 2022. Mendenhall E, Newfield T, Tsai AC. Syndemic theory, methods, and data. Social Science & Medicine (1982). 2022;295:114656. Coid J, Zhang Y, Bebbington P, Ullrich S, De Stavola B, Bhui K, et al. A syndemic of psychiatric morbidity, substance misuse, violence, and poor physical health among young Scottish men with reduced life expectancy. SSM-population Health. 2021;15:100858. Rod MH, Rod NH, Russo F, Klinker CD, Reis R, Stronks K. Promoting the health of vulnerable populations: three steps towards a systems-based re-orientation of public health intervention research. Health Place. 2023;80:102984. Schofield J, Papathomas M, Macnamara C, McCann M, Ardestani BM, Skivington K et al. Contextualising multimorbidity in people who use drugs: analysis of drug-death decedents in Scotland. Ir J Psychol Med. 2025:1–9. Lee J, Bainter S, Carrico A, Glynn T, Rogers B, Albright C, et al. Connecting the dots: a comparison of network analysis and exploratory factor analysis to examine psychosocial syndemic indicators among HIV-negative sexual minority men. J Behav Med. 2020;43:1026–40. Meadows D. Places to Intervene in a System. Whole Earth. 1997;91(1):78–84. Wight D, Wimbush E, Jepson R, Doi L. Six steps in quality intervention development (6SQuID). J Epidemiol Community Health. 2016;70(5):520–5. Turner SF, Cardinal LB, Burton RM. Research design for mixed methods: A triangulation-based framework and roadmap. Organizational Res methods. 2017;20(2):243–67. Williams DR. GGMnonreg: Non-Regularized Gaussian Graphical Models in R. J Open Source Softw. 2021;6(67):3308. Banack HR, Mayeda ER, Naimi AI, Fox MP, Whitcomb BW. Collider Stratification Bias I: Principles and Structure. Am J Epidemiol. 2024;193(2):238–40. Ulrich W, Reynolds M. Critical systems heuristics. Systems approaches to managing change: A practical guide. Springer; 2010. pp. 243–92. Barbrook-Johnson P, Penn AS. Rich Pictures. Systems Mapping: How to build and use causal models of systems: Springer; 2022. pp. 21–32. Csardi G, Nepusz T. The igraph software. Complex syst. 2006;1695:1–9. GmbH y. yEd 2024 [Available from: https://www.yworks.com/products/yed Dahlgren G, Whitehead M. The Dahlgren-Whitehead model of health determinants: 30 years on and still chasing rainbows. Public Health. 2021;199:20–4. Pedersen TL, Pedersen M, LazyData T, Rcpp I, Rcpp L. Package ‘ggraph’. Retrieved January. 2017;1:2018. Crielaard L, Quax R, Sawyer AD, Vasconcelos VV, Nicolaou M, Stronks K, et al. Using network analysis to identify leverage points based on causal loop diagrams leads to false inference. Sci Rep. 2023;13(1):21046. McCann M. Github page for network analysis of Scotland's Drug-related death system 2025 [Available from: github.com/MRC-CSO-SPHSU/NDRDD-linkage. Rhodes T. Risk environments and drug harms: a social science for harm reduction approach. Elsevier; 2009. pp. 193–201. Ingram PF, Bailey AJ, Finn PR. Applying network analysis to investigate substance use symptoms associated with drug overdose. Drug Alcohol Depend. 2022;234:109408. Tsai AC. Syndemics: a theory in search of data or data in search of a theory? Soc Sci Med. 2018;206:117–22. Petticrew M, Katikireddi SV, Knai C, Cassidy R, Hessari NM, Thomas J, et al. Nothing can be done until everything is done’: the use of complexity arguments by food, beverage, alcohol and gambling industries. J Epidemiol Community Health. 2017;71(11):1078–83. Meyer S, Gibson B, Ward P. Niklas Luhmann: Social Systems Theory and the Translation of Public Health Research. In: Collyer F, editor. The Palgrave Handbook of Social Theory in Health, Illness and Medicine. London: Palgrave Macmillan UK; 2015. pp. 340–54. Beeston C, Robinson M, Giles L, Dickie E, Ford J, MacPherson M, et al. Evaluation of minimum unit pricing of alcohol: a mixed method natural experiment in Scotland. Int J Environ Res Public Health. 2020;17(10):3394. Rutter H, Savona N, Glonti K, Bibby J, Cummins S, Finegood DT, et al. The need for a complex systems model of evidence for public health. lancet. 2017;390(10112):2602–4. Finegood DT, Merth TD, Rutter H. Implications of the foresight obesity system map for solutions to childhood obesity. Obesity. 2010;18(S1):S13–6. Government S. National Drugs Mission Plan: 2022–2026. 2022. Additional Declarations No competing interests reported. Supplementary Files AppendixnetworkanalysisofDRDsystem.docx Cite Share Download PDF Status: Under Review Version 1 posted Reviews received at journal 23 Feb, 2026 Reviewers agreed at journal 19 Feb, 2026 Reviewers agreed at journal 18 Feb, 2026 Reviewers invited by journal 16 Feb, 2026 Editor invited by journal 06 Feb, 2026 Editor assigned by journal 03 Feb, 2026 Submission checks completed at journal 03 Feb, 2026 First submitted to journal 02 Feb, 2026 You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-8767159","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":593576259,"identity":"38edcb5b-6618-4223-8b03-574d2f5c2937","order_by":0,"name":"Mark McCann","email":"data:image/png;base64,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","orcid":"","institution":"University of Glasgow","correspondingAuthor":true,"prefix":"","firstName":"Mark","middleName":"","lastName":"McCann","suffix":""},{"id":593576261,"identity":"27d74236-7035-4b9f-bb5e-d2d74d0779bf","order_by":1,"name":"Rosemary Seaman","email":"","orcid":"","institution":"University of Glasgow","correspondingAuthor":false,"prefix":"","firstName":"Rosemary","middleName":"","lastName":"Seaman","suffix":""},{"id":593576263,"identity":"06b58c1a-8e50-42de-8449-ca31654722c0","order_by":2,"name":"Srebrenka Letina","email":"","orcid":"","institution":"University of Glasgow","correspondingAuthor":false,"prefix":"","firstName":"Srebrenka","middleName":"","lastName":"Letina","suffix":""},{"id":593576270,"identity":"4f2af293-162d-4a11-85ea-bc0867d6f539","order_by":3,"name":"Tessa Parkes","email":"","orcid":"","institution":"University of Stirling","correspondingAuthor":false,"prefix":"","firstName":"Tessa","middleName":"","lastName":"Parkes","suffix":""},{"id":593576275,"identity":"b3ec04f3-9674-45d0-bb8c-392c54df2e3c","order_by":4,"name":"Lee Barnsdale","email":"","orcid":"","institution":"Public Health Scotland","correspondingAuthor":false,"prefix":"","firstName":"Lee","middleName":"","lastName":"Barnsdale","suffix":""},{"id":593576280,"identity":"8081bf8a-0c63-4bf6-93a8-e184d3050ac9","order_by":5,"name":"Catriona Matheson","email":"","orcid":"","institution":"University of Stirling","correspondingAuthor":false,"prefix":"","firstName":"Catriona","middleName":"","lastName":"Matheson","suffix":""},{"id":593576286,"identity":"269fa3ba-fa73-41a5-85a4-c77c819704e7","order_by":6,"name":"Joe Schofield","email":"","orcid":"","institution":"University of St Andrews","correspondingAuthor":false,"prefix":"","firstName":"Joe","middleName":"","lastName":"Schofield","suffix":""},{"id":593576287,"identity":"e5b86967-0c56-4d2b-99c1-6edf79f3c1e3","order_by":7,"name":"Jeremy Hilton","email":"","orcid":"","institution":"University College London","correspondingAuthor":false,"prefix":"","firstName":"Jeremy","middleName":"","lastName":"Hilton","suffix":""},{"id":593576288,"identity":"d8a5d6cf-6367-42b9-9423-a55ba047a922","order_by":8,"name":"Alexander Baldacchino","email":"","orcid":"","institution":"University of St Andrews","correspondingAuthor":false,"prefix":"","firstName":"Alexander","middleName":"","lastName":"Baldacchino","suffix":""},{"id":593576292,"identity":"1c7f3e66-970d-4651-a65d-c8ba75519af9","order_by":9,"name":"Adrian Crofton","email":"","orcid":"","institution":"Torry Medical Practice","correspondingAuthor":false,"prefix":"","firstName":"Adrian","middleName":"","lastName":"Crofton","suffix":""},{"id":593576293,"identity":"2ba49b3e-883b-48c5-b384-c8e280651877","order_by":10,"name":"Kathryn Skivington","email":"","orcid":"","institution":"University of Glasgow","correspondingAuthor":false,"prefix":"","firstName":"Kathryn","middleName":"","lastName":"Skivington","suffix":""}],"badges":[],"createdAt":"2026-02-02 16:15:24","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-8767159/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-8767159/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":103504633,"identity":"77ab0dfc-dd3d-4b23-9782-09aec3fb07ed","added_by":"auto","created_at":"2026-02-26 13:20:51","extension":"jpeg","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":902552,"visible":true,"origin":"","legend":"\u003cp\u003eSystem map of connections between 98 factors relating to drug-related deaths in Scotland. Node colour represent seven subsystems of closely connected factors. Vertical position represents system levels, from wider context (highest) to death (lowest). Line colour represents within level (orange) and cross level (grey) connections. Bottom left inset box shows subsystems as shaded areas (large version in Appendix 5).\u003c/p\u003e","description":"","filename":"floatimage1.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-8767159/v1/992b1212c6ccf279f6260518.jpeg"},{"id":103166715,"identity":"8ef0e54a-d13f-4971-85bf-4682385895e1","added_by":"auto","created_at":"2026-02-22 12:41:36","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":277091,"visible":true,"origin":"","legend":"\u003cp\u003eNetwork showing the connected subsystems from louvain community detection. Node size and red shading indicate number of variables in the subsystem, line width is proportional to the number of connections to other subsystems. Subsystems with no connection to others have been removed.\u003c/p\u003e","description":"","filename":"floatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-8767159/v1/b7a610a455e86183517bbaee.png"},{"id":103510802,"identity":"51d89e38-fc08-4a76-8d4d-d08c7b58dfe3","added_by":"auto","created_at":"2026-02-26 14:07:18","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":2542223,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8767159/v1/13a01dce-2932-4474-bfcd-bd69805f8f6e.pdf"},{"id":103166716,"identity":"728046ea-6da8-4d10-b1d2-319922116229","added_by":"auto","created_at":"2026-02-22 12:41:36","extension":"docx","order_by":0,"title":"","display":"","copyAsset":false,"role":"supplement","size":1667013,"visible":true,"origin":"","legend":"","description":"","filename":"AppendixnetworkanalysisofDRDsystem.docx","url":"https://assets-eu.researchsquare.com/files/rs-8767159/v1/95aa5237c83eccd21d005746.docx"}],"financialInterests":"No competing interests reported.","formattedTitle":"Complex system structures around drug-related deaths in Scotland: a mixed method network analysis approach to study subsystems in population data and system maps","fulltext":[{"header":"Introduction","content":"\u003cp\u003eDrug-related deaths (DRDs) are a growing public health concern globally. Scotland has the highest drug-related death (DRD) rate in Europe, with over 1,000 deaths annually since 2018, despite slight reductions in 2022 and 2024. (\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e). Multiple factors contribute to DRDs, including multimorbidity, polypharmacy, interaction with health and social services, and broader social and political factors. Evidence shows rising multimorbidity among people who use drugs, significantly increasing DRD risk. In Scotland, 70% of individuals who experienced a DRD in 2017 had at least one medical condition recorded in the six months prior to death, up from 46% in 2009. Common conditions included depression (45%), respiratory conditions (27%), and blood-borne viruses (19%) (\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eSocial factors may also contribute to risk of a DRD. Over half of DRDs occur among individuals living in the 20% most deprived areas of Scotland (\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e). Increases in DRD between 1980 and 2012 were highest in the most deprived areas (\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e). Other social factors associated with DRD include experiencing domestic violence (43% of women having a DRD) and living alone (52% of all DRDs) (\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e). Around half of those who experienced a DRD in Scotland had contact with drug treatment services in the six months prior to death. Many also interacted with healthcare (e.g., mental health services, hospital stays) and social care (e.g., housing, social work) services during this period (\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e). Closer inspection of patterns of co-occurring treatment could illuminate common multimorbidity patterns and implications for death prevention.\u003c/p\u003e \u003cp\u003eThere is comprehensive, biennial reporting on the prevalence of health conditions, social experiences, and services used among DRD decedents in Scotland (\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e). This reporting uses Public Health Scotland\u0026rsquo;s National Drug-related Deaths Database (NDRDD), which collects detailed information on DRDs. These Official Statistics draw on a range of administrative data sources to provide information on the prevalence of single conditions among people who have experienced a DRD. However, the standard reporting framework has limited scope to explore multimorbidity or integrate expertise and insights of those with lived/living experience of substance use and those working in the sector. A detailed focus on co-occurring health conditions, social experiences, and service use, combined with experiential evidence may help to support preventive strategies.\u003c/p\u003e\n\u003ch3\u003eAim\u003c/h3\u003e\n\u003cp\u003eWe aimed to develop a systems-informed understanding of factors relating to DRDs in Scotland drawing upon multiple data sources and integrating experiential and statistical evidence. To meet this aim, we developed system maps through cross-sectoral workshops and conducted analysis of linked administrative data.\u003c/p\u003e \u003cp\u003eResearch Questions\u003c/p\u003e \u003cp\u003eRQ1: What are the most central (connected) factors within the DRD system?\u003c/p\u003e \u003cp\u003eRQ2: What are the key subsystems of related factors within the DRD system?\u003c/p\u003e \u003cp\u003eRQ3: What common and unique perspectives on system structure are obtained from administrative data and system maps?\u003c/p\u003e \u003cp\u003eBetter understanding of how causes of DRD are connected across levels of influence could help inform more effective actions across the system. System mapping provides an effective way to develop understanding of complex causal processes interacting at multiple levels (\u003cspan class=\"CitationRef\"\u003e5\u003c/span\u003e) and could help to improve decision making around DRD prevention. Syndemic theory helps explain how certain health conditions co-occur and, combined with social influences, lead to amplified adverse outcomes differently from effects in isolation (\u003cspan class=\"CitationRef\"\u003e6\u003c/span\u003e). Syndemic theory has been used to understand outcomes among people who use drugs (\u003cspan class=\"CitationRef\"\u003e7\u003c/span\u003e), and systems approaches can play an important part in the identification of co-occurring health and social factors which may underpin emergent patterns of syndemic ill health and vulnerability (\u003cspan class=\"CitationRef\"\u003e8\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eNetwork analysis methods are suitable for exploring the structural features of relationships, such as describing how commonly factors co-occur and identifying groups of interconnected factors. Instead of focusing on the prevalence or pairwise associations of variables, network analysis provides insight into the broader system structures that encompass individual experiences related to wellbeing, health, and mortality (\u003cspan class=\"CitationRef\"\u003e9\u003c/span\u003e, \u003cspan class=\"CitationRef\"\u003e10\u003c/span\u003e). By studying the connections between health and social conditions, network analysis can identify factors that may form a syndemic, helping to understand the broader social structures related to DRDs and identify effective leverage points for intervention (\u003cspan class=\"CitationRef\"\u003e11\u003c/span\u003e).\u003c/p\u003e "},{"header":"Methods and data","content":"\u003cp\u003eThis paper reports on data from a project aimed at defining, identifying, and prioritising action areas to prevent DRDs in Scotland. We analysed administrative data and conducted stakeholder co-production workshops. This paper covers steps 1 and 2 of the 6 Steps in Quality Intervention Development (6SQuID) process: defining and understanding the problem and its causes, and identifying which causal or contextual factors are modifiable (\u003cspan class=\"CitationRef\"\u003e12\u003c/span\u003e). The project used a multi method design with two concurrent work streams - linked data analysis and system workshops - with two data integration points. The first integration point involved preliminary linked data analysis forming part of system workshop activities which in turn informed the development of the system map. The second integration point involved holistic triangulation of the findings from the network analysis of data from each work stream. In multi method and mixed method studies, convergent triangulation is commonly applied to enhance validity or generalisability of findings (e.g. quantitatively assessing an idea based on qualitative data, or assessing consensus across datasets). An alternative analytical approach is holistic triangulation, where multiple methods are applied in order to obtain a more comprehensive understanding of the topic under study (e.g. using multiple data sources to extend or develop a theory) (\u003cspan class=\"CitationRef\"\u003e13\u003c/span\u003e). In this study, we applied holistic triangulation to understand the overlapping and unique perspectives provided from two data sources, applying methods that follow a different paradigm from more common quantitative and qualitative health and social science methods.\u003c/p\u003e\u003cp\u003eFor the remainder of this paper, we consider DRDs and the collective set of factors and causal influences at all levels - from individual to political - as the System of Interest. We view a system as comprising \u003cem\u003esubsystems\u003c/em\u003e, or subsets of closely related factors, these are indicated in quotation marks. We use the term factor when referring to aspects of the broader system discussed in workshops (e.g. poverty or positive future aspirations), and variable to refer to aspects of quantitative data included in the study (e.g. Venlafaxine in prescription data). Factors and variables are indicated in italics.\u003c/p\u003e\n\u003ch3\u003eData source 1: linked administrative data\u003c/h3\u003e\n\u003cp\u003eThe National Drug-Related Deaths Database (NDRDD) holds data on most individuals whose death was recorded as a DRD in Scotland since 2009 (excluding suicides and cases with recording issues). Data collection coordinators in NHS Health Boards gather information on suspected drug deaths from pathology and toxicology reports, police sudden death reports, and local health and social work systems (e.g. contacting General Practitioners or social workers to obtain information on recent health/treatment). This information is entered into the NDRDD and, if the death matches the inclusion criteria, linked to NHS administrative health records covering community-dispensed prescriptions (Prescribing Information System, PIS), general and acute hospital admissions (Scottish Morbidity Records, SMR01), and psychiatric hospital admissions (SMR04) within six months prior to death.\u003c/p\u003e \u003cp\u003eFor our research, we accessed the NDRDD for 6,608 individuals who died from a DRD in Scotland between 2009 and 2018. Using NHS unique patient identifiers (Community Health Index or CHI numbers), we linked information from PIS, SMR01, and SMR04. Our analysis focused on binary indicators of conditions noted in the NDRDD, primary and secondary International Classification of Disease 10 (ICD-10) codes in SMR, or dispensed as prescriptions in PIS (prescriptions are free under the NHS in Scotland, a record appears in PIS whenever a medicine is dispensed). If an individual\u0026rsquo;s unique identifier was absent in one dataset, their record was set to zero for all conditions in that database. For example, individuals present in the NDRDD but not in the psychiatric admissions SMR04 dataset were assumed not to have any conditions recorded in SMR04, contributing zeroes for estimating the co-occurrence of those conditions with others in the data. We excluded variables where fewer than 20 individuals had the condition. This reduced model estimation time and was unlikely to bias results as there were few potential cases of co-occurrence. The final dataset included 1,371 variables for 6,608 individuals.\u003c/p\u003e\n\u003ch3\u003ePreparatory analysis: condition co-occurrence network\u003c/h3\u003e\n\u003cp\u003eWe estimated a gaussian graphical model representing the correlations between the 1,371 conditions using the R package GGMnonreg (\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e). This produced a square matrix where each cell represented the partial correlation between each pair of conditions in the dataset, conditioned on other variables. The estimation used Pearson correlation which returns the Phi coefficient for binary variables. Coefficients that replicated in less than 95% of bootstrap replications, as well as all negative correlations, were fixed at zero. The retained correlations indicated pairs of conditions that co-occurred more frequently than expected due to chance. Removing factors with no co-occurrence resulted in 1,229 variables for the next stage of analysis. We excluded negative correlations for two reasons. First, causal inference methodology suggests that if two variables independently affect the likelihood of appearing in a study sample, analysis of sample data will reveal a negative correlation, even without a causal association in the wider population (collider bias) (\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e). It is plausible that many conditions are independently associated with risk of a DRD and thus appear in this sample, making negative coefficients inaccurate. Second, our primary focus was on co-occurrence and connections most relevant to DRDs, so excluding negative or absent correlations aligns with this purpose. As the primary aim was to understand the overall structure of co-occurrence, analysis did not use demographic variables such as gender, age at death, or area deprivation measures. The final matrix was converted to a weighted network for the analysis described below.\u003c/p\u003e\n\u003ch3\u003eData source 2: co-produced system map of drug-related deaths\u003c/h3\u003e\n\u003cp\u003eWe drew upon systems science methods to build a collective understanding of the individual, service-related, health, social and political factors surrounding DRDs. We recruited a diverse group of co-production partners, representing multiple perspectives relevant to the DRD system. Using Critical System Heuristics (CSH) (\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e), we identified co-production partners who could provide the following role: sources of motivation (whose interests should be served?); power (who is the decision maker/who is in a position to change the measure of improvement?); knowledge (who should be considered an expert? What counts as relevant knowledge?); and legitimation (who is witness to the interests of those affected but not involved? Who can argue the case of those not able to speak for themselves?). We mapped relevant roles onto CSH domains (Appendix 1) and identified individuals representing each role to invite to the workshops. Twenty-eight people participated in the workshop series and all provided written, informed consent. Participants were offered travel reimbursement and received no other reimbursement if attending as a salaried professional. Participants who were unsalaried such as those attending due to relevant lived/living experience or as an affected family member received honoraria to acknowledge their time.\u003c/p\u003e \u003cp\u003eWorkshop 1 (February 2022) focused on understanding the problem and its causes. Four groups worked through a series of activities: a rich picture activity (\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e) to visualise and discuss the various perspectives held by the participants and generate ideas on the salient factors in the DRDs system; a variable elicitation activity to generate lists of relevant factors; drawing maps where each factor was placed within a circle (node) and the perceived causal associations between factors drawn as arrows (ties) (\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e). Discussions were not guided around themes (e.g. individual, community, political influences), participants were encouraged to think broadly about factors. We combined the four maps by amalgamating similar concepts differently phrased into single nodes. At workshop 2 (September 2022), participants were shown preliminary analysis of the NDRDD (see Appendix 2) and then reviewed the combined system map to make further changes and additions. There was one further meeting of the author team to review and finalise the map. At workshop 3 (November 2022), participants were presented with the system map and used it to develop action ideas and identify relevant roles and organisations to take actions forward (these not reported in the current paper). No further modifications to the map were requested at the workshop.\u003c/p\u003e\n\u003ch3\u003eNetwork analysis of the system map and co-occurrence network\u003c/h3\u003e\n\u003cp\u003eWe used the R package igraph (\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e), online platform Kumu.io, and yEd (\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e) software to visualise and analyse the linked data network and system map. We applied network analysis to calculate network metrics for each node, including degree metrics which measured how many connections each node had to other factors. In the linked data network, degree represents how many other conditions co-occurred with the condition at levels greater than expected due to chance. In the system map, degree represents how many other factors were deemed by workshop participants to be causally connected to the factor. We additionally calculated indegree (how many factors influenced this node) and outdegree (how many factors this node affected).\u003c/p\u003e \u003cp\u003eComplex systems comprise interacting factors at multiple levels of influence from the individual to societal, each factor was categorised according to its level of influence. The classification was informed by the authors\u0026rsquo; understanding of the levels of the social ecological model of health (\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e), inductively coded by MMcC and KS independently, and consensus reached after discussion (see Appendix 5). We applied a multilayer visualisation algorithm to the coded system map using the R package ggraph (\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e), and Appendix 5 shows the distribution of factors across the subsystems and levels of influence.\u003c/p\u003e \u003cp\u003eTo answer RQ1, we calculated betweenness centrality. Low betweenness centrality gives an indication of which factors sit on the periphery of a network with few connections to central nodes and far from all other nodes, while high betweenness centrality indicates factors occupy a central location (those that are \u0026lsquo;closer\u0026rsquo; to other nodes in terms of how many \u0026lsquo;steps\u0026rsquo; along ties in the network are needed to reach the node). The magnitude of centrality measures scales according to the size of the network in question: they must be interpreted relative to the other nodes in the network and are not comparable between networks. In the context of the system map, high centrality provides a measure of variables which act as potential mediators on many causal pathways in the system of interest. We do not infer causal direction from the correlational data in the administrative data network. The values of the networks metrics were not used to draw inference over points to intervene (\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e); network metrics provide numerical starting points for analysis of the map structures, similar to how cardinal directions orient a reader to a geographical map.\u003c/p\u003e \u003cp\u003eTo answer RQ2, we used the Louvain community detection algorithm to detect sub-systems representing groups of factors that were closely related. The Louvain algorithm optimises network modularity; it returns the number of subsystems that maximises the ratio of within-subsystem to between-subsystem connections. The authors looked at the factors included in each subsystem and applied a label to describe the common theme or factor among the variables. In the system map, KS initially coded the subsystems and this was cross validated by MMcC and RS. In the linked data network, MMcC initially coded the subsystems, these were cross-validated by SL, CM, JS, JH and AB and the final subsystem labels agreed by consensus. Consensus meetings also involved initial interpretation of the subsystems and implications for DRD prevention. The subsystems were subsequently grouped into three higher-level categories described below. Kumu was used to detect feedback loops in the system map. Summary findings were circulated to workshop delegates and feedback on presentations about the study also informed our collective interpretation.\u003c/p\u003e \u003cp\u003eTo answer RQ3, we conducted a holistic multi-method integration (\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e) to qualitatively assess the conceptual overlap and differences in the factors, variables and subsystems arising from the networks relating to each data source. This was achieved by iterating over the following steps: matching the thematic description of subsystems in one data source with the subsystem descriptions in the other; checking if individual nodes in one data source related to a subsystem theme in the other, revisiting workshop field notes to obtain context on discussion of nodes \u0026amp; connections; and visually scanning the linked data network to follow the thread on connections arising from thematic comparisons. This approach allowed us to combine insights from the most comprehensive source of quantitative data on DRDs in Scotland with the diverse perspectives and insights obtained from systems mapping.\u003c/p\u003e"},{"header":"Results","content":"\u003cdiv id=\"Sec9\" class=\"Section2\"\u003e \u003ch2\u003eCo-produced system map\u003c/h2\u003e \u003cp\u003eThe final system map contained 98 factors with 225 connections between them (a network density of 0.02 i.e. 2% of the potential connections between all factors), including eight sub-systems. Factors fell across nine levels of influence within a social ecological framework. Figure\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e shows the nodes and sub-systems with vertical layers representing levels of influence, from wider contextual factors to political, organisational and individual factors, with the ultimate outcome of DRD in the lowest layer. The node colour shows which nodes belong to the same subsystem of connected factors. A zoomable, interactive version of the map can be downloaded from Github and opened using yEd. To the top left, the red nodes represent the subsystem around public and workforce stigmatising attitudes which falls across the wider context, policy, organisational and interpersonal levels of the system. At the bottom of the image the grey nodes represent the subsystem around the proximal causes of death. Factors in this subsystem fall across the \u0026ldquo;thoughts and emotions\u0026rdquo;, \u0026ldquo;behavioural\u0026rdquo;, and \u0026ldquo;physical\u0026rdquo; levels of the system. Subsystems contained factors that traversed upstream and downstream levels of influence within the social ecological model, rather than containing factors at a single level (see appendix Fig.\u0026nbsp;4 for the distribution of factors across levels and subsystems).\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eTable\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e shows the subsystems identified in the map, along with the factors with the highest degree centrality in each sub-system. Data on all factors and the subsystems are available on github (\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e). Appendix 5 gives a detailed description of how to interpret the subsystems, factors and metrics. The largest subsystem contained 22 factors relating to the proximal causes of death. The two most connected factors in the systems map - \u003cem\u003emental health\u003c/em\u003e and \u003cem\u003eprevalence of stigmatising norms\u003c/em\u003e \u0026ndash; each appeared in their own subsystem of closely related factors.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab1\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eSubsystems returned from community detection algorithms applied to the system map, and five factors within each subsystem with highest degree\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"5\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eSubsystem description and factors\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eSocial ecological levels within subsystem\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eFive central factors\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eDegree\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003eBetween-\u003c/p\u003e \u003cp\u003eness\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eDrug death and proximal causes\u003c/p\u003e \u003cp\u003e(Grey nodes in Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003ePhysical, behavioural,\u003c/p\u003e \u003cp\u003ecognitive-emotional\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eControl and regularity of substance use\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e10\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.11\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e22 factors\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eInitiation of drug use\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e8\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.07\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003ePhysical health\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e8\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.02\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eContinued use of drugs\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e7\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.06\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eDrug-related death\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e7\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eLife experiences and mental wellbeing\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\" morerows=\"3\" rowspan=\"4\"\u003e \u003cp\u003eBehavioural,\u003c/p\u003e \u003cp\u003ecognitive -emotional,\u003c/p\u003e \u003cp\u003einterpersonal,\u003c/p\u003e \u003cp\u003eservice interactions,\u003c/p\u003e \u003cp\u003epolicy,\u003c/p\u003e \u003cp\u003ewider context\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eMental health and wellbeing\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e17\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.09\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e(Purple nodes in Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003ePoverty\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e14\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.03\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e16 factors\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eCriminalisation of drugs\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e12\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.03\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003ePositive future aspirations\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e11\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.03\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eContact with criminal justice system\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e9\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.04\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePublic and workforce stigmatising attitudes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\" morerows=\"2\" rowspan=\"3\"\u003e \u003cp\u003eInterpersonal,\u003c/p\u003e \u003cp\u003eservice interaction,\u003c/p\u003e \u003cp\u003eorganisational,\u003c/p\u003e \u003cp\u003ewider context\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003ePrevalence of stigmatising norms around drugs\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e18\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.14\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e(Red nodes in Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eQuality of treatment planning\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e6\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.01\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e16 factors\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eService Quality\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e6\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.02\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eWorkforce development\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e4\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003ePoor treatment of PWUD\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e4\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.04\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eSocial influences on drug harms\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\" morerows=\"3\" rowspan=\"4\"\u003e \u003cp\u003ePhysical,\u003c/p\u003e \u003cp\u003ebehavioural,\u003c/p\u003e \u003cp\u003einterpersonal,\u003c/p\u003e \u003cp\u003eservice interaction,\u003c/p\u003e \u003cp\u003eorganisational,\u003c/p\u003e \u003cp\u003ewider context\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003ePeer relationships\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e6\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e(Brown nodes in Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eDrug type\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e6\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.02\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e13 factors\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eExposure to drugs\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e5\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.04\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003ePublic versus private drug taking\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e5\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eKnowledge of drugs and risk among general public\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e4\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eExperience of services\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\" morerows=\"3\" rowspan=\"4\"\u003e \u003cp\u003eCognitive-emotional,\u003c/p\u003e \u003cp\u003eservice interaction,\u003c/p\u003e \u003cp\u003eorganisational\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eAttending services\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e9\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.01\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e(Blue nodes in Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eRetention in services\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e6\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.06\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e11 factors\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eCollaboration between services\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e5\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eMissed appointments\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e5\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.01\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eInternalised stigma\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e4\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.01\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCommunity influences on wellbeing\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\" morerows=\"3\" rowspan=\"4\"\u003e \u003cp\u003eInterpersonal,\u003c/p\u003e \u003cp\u003ecommunity\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eTrusting community relationships\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e12\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.03\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e(Orange nodes in Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eTrusting family relationships\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e7\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.05\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e8 factors\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eRecovery capital\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e5\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.02\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003ePractical support\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e3\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eCommunity hubs\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePublic perspectives on substance use\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\" morerows=\"3\" rowspan=\"4\"\u003e \u003cp\u003eOrganisational,\u003c/p\u003e \u003cp\u003epolicy,\u003c/p\u003e \u003cp\u003ewider context\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003ePolicy making environment: health focussed vs criminal justice\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e11\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.09\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e(Green nodes in Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eAvailability of services\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e5\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.01\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e7 factors\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eHarm reduction interventions\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e4\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eRelative value of professional versus peer evidence\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e3\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.07\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eLived experience representation in policy\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e3\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eSafe physical environments\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\" morerows=\"3\" rowspan=\"4\"\u003e \u003cp\u003eCognitive-emotional,\u003c/p\u003e \u003cp\u003epolicy,\u003c/p\u003e \u003cp\u003ewider context\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eFeeling safe\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e6\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.01\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e(Pink nodes in Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eHomelessness\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e6\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.01\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e5 factors\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eHousing policy\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e3\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eQuality of housing\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e3\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eRisk of crime victimisation\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e3\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003c/div\u003e\n\u003ch3\u003eFeedback loop analysis\u003c/h3\u003e\n\u003cp\u003eThe system map included feedback loops which described how drug use initiation may lead to drug use for self-medication or functional reasons which could increase continued drug use. This increases exposure to a greater range of drug types which is then related to continued use. This corresponds to the concept of entry into a wider drug market which in the context of novel and high toxicity drugs entering the supply introduces higher risk of fatality. A further loop related to stigmatising reporting of substance use in the media which may drive negative perceptions and poor treatment of people who use drugs. This increases the prevalence of stigmatising norms that loops back into further stigmatising media reports.\u003c/p\u003e \u003cdiv id=\"Sec11\" class=\"Section2\"\u003e \u003ch2\u003eLinked administrative data co-occurrence network\u003c/h2\u003e \u003cp\u003eThe linked data network contained 1,229 variables and 1,950 connections (density\u0026thinsp;=\u0026thinsp;0.003) grouped into 78 communities which we describe as subsystems below. Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e lists the labels and the number of variables in each subsystem in three groups: eight related to co-occurring conditions, 12 with conditions related to substance use, and 32 related to distinct conditions or forms of treatment. There were a further 24 subsystems containing only two variables reported in Appendix Table\u0026nbsp;3.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab2\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 2\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eDescription of subsystems and variables per subsystem identified by Louvain community detection on co-occurrence network for linked hospital records, prescription and drug deaths database\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"5\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colspan=\"2\" nameend=\"c2\" namest=\"c1\"\u003e \u003cp\u003eCo-occurring conditions and treatment\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colspan=\"3\" nameend=\"c5\" namest=\"c3\"\u003e\u0026nbsp;\u003c/th\u003e \u003c/tr\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eSubsystem\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colspan=\"2\" nameend=\"c3\" namest=\"c2\"\u003e \u003cp\u003eNumber of variables\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eSubsystem\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003eNumber of variables\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eRespiratory infection, depression, pain\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c3\" namest=\"c2\"\u003e \u003cp\u003e82\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eMixed allergy, asthma medications\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e42\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAssault and self-harm\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c3\" namest=\"c2\"\u003e \u003cp\u003e77\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eAntidepressants and OCD\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e4\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMixed stomach, skin, antibiotics\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c3\" namest=\"c2\"\u003e \u003cp\u003e47\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eSkin cream and ibuprofen\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e2\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMixed common prescriptions: skin, stomach, sinuses\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c3\" namest=\"c2\"\u003e \u003cp\u003e42\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eMusculoskeletal/Skin\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e2\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eSubstance use related conditions and treatment\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c3\" namest=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eSubsystem\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c3\" namest=\"c2\"\u003e \u003cp\u003e\u003cb\u003eNumber of variables\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e\u003cb\u003eSubsystem\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e\u003cb\u003eNumber of variables\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMixed NDRDD group\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c3\" namest=\"c2\"\u003e \u003cp\u003e110\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eAlcohol, fungal infection \u0026amp; stomach prescriptions\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e42\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAccidental and intentional self-poisoning\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c3\" namest=\"c2\"\u003e \u003cp\u003e70\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eAlcohol dependence and gastritis\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e41\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eLiver cirrhosis, malnourishment\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c3\" namest=\"c2\"\u003e \u003cp\u003e63\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eAlcohol treatment and detoxification\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e32\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePhlebitis, hepatitis, mental health, substance use\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c3\" namest=\"c2\"\u003e \u003cp\u003e56\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eSuboxone\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e4\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBenzos and pain\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c3\" namest=\"c2\"\u003e \u003cp\u003e56\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eBuprenorphine\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e2\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMethadone and common prescriptions\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c3\" namest=\"c2\"\u003e \u003cp\u003e44\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eFall and drug use noted at hospital\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e2\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c2\" namest=\"c1\"\u003e \u003cp\u003e\u003cb\u003eDistinct conditions and medications\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"3\" nameend=\"c5\" namest=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eSubsystem\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c3\" namest=\"c2\"\u003e \u003cp\u003e\u003cb\u003eNumber of variables\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e\u003cb\u003eSubsystem\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e\u003cb\u003eNumber of variables\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMental health inpatient treatment\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c3\" namest=\"c2\"\u003e \u003cp\u003e87\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eNausea tablets\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e2\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eRespiratory disease\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c3\" namest=\"c2\"\u003e \u003cp\u003e43\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eDermol shower cream\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e2\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eHeart disease\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c3\" namest=\"c2\"\u003e \u003cp\u003e38\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eMagnesium sulphate paste\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e2\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMorphine, constipation, haemorrhoid treatment\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c3\" namest=\"c2\"\u003e \u003cp\u003e38\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eLevetiracetam - Epilepsy\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e2\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eEpilepsy\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c3\" namest=\"c2\"\u003e \u003cp\u003e37\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eNaproxen painkiller\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e2\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eSkin cream\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c3\" namest=\"c2\"\u003e \u003cp\u003e35\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eProctosedyl - haemorrhoids\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e2\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eDiabetes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c3\" namest=\"c2\"\u003e \u003cp\u003e35\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eNicotine patches\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e2\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePregabalin, Duloxetine, Neuropathic pain\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c3\" namest=\"c2\"\u003e \u003cp\u003e31\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eBlood clot prevention\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e2\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAsthma inhalers\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c3\" namest=\"c2\"\u003e \u003cp\u003e5\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eFucidin antibiotic cream\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e2\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCardiac arrest\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c3\" namest=\"c2\"\u003e \u003cp\u003e5\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eOvarian cysts\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e2\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eThyroid condition\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c3\" namest=\"c2\"\u003e \u003cp\u003e4\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eDeviated septum\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e2\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eVenlafaxine antidepressant\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c3\" namest=\"c2\"\u003e \u003cp\u003e3\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eDental issues\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e2\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eOrchitis\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c3\" namest=\"c2\"\u003e \u003cp\u003e3\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eObesity\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e2\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAntidepressant Dosulepin\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c3\" namest=\"c2\"\u003e \u003cp\u003e3\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eNo known conditions\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e2\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eEmployment\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c3\" namest=\"c2\"\u003e \u003cp\u003e3\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eMigraine\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e2\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eChild custody\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c3\" namest=\"c2\"\u003e \u003cp\u003e3\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eLearning disability, attention deficit hyperactivity disorder\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e2\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAnti-inflammatory medication\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c3\" namest=\"c2\"\u003e \u003cp\u003e2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eDeath in prison\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e2\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eFigure \u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e is a network visualisation of 24 subsystems listed in Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e, excluding the subsystems that did not have connections to other subsystems. The largest community contained 110 factors from the NDRDD itself, the data \u0026ldquo;spine\u0026rdquo; to which other data sources were linked. The NDRDD contains specific variables often reflecting very closely related information that was not available in SMR or PIS. For example, it was common for a toxicology report to find multiple substances and these appeared as connected within the NDRDD cluster. NDRDD variables on route of administration (oral versus injecting) were highly correlated with the toxicology reports for the respective substances, but not with other health conditions. Despite the distinct dataset effect, many variables in this subsystem connected to other subsystems e.g. toxicology variables in this subsystem co-occurred with related prescription variables located in other subsystems (see the Suboxone \u0026ndash; NDRDD connection in Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e). The preliminary analysis of the NDRDD database before linkage to other datasets (see Appendix 2) identified further structures nested within this subsystem, some overlapping with the main analysis (e.g. mental health, substance use treatment), and others containing unique information in the NDRDD around living arrangements and criminal justice involvement.\u003c/p\u003e \u003cp\u003eTable\u0026nbsp;\u003cspan refid=\"Tab3\" class=\"InternalRef\"\u003e3\u003c/span\u003e gives information on the two variables within each subsystem with the highest number of connections for the 15 largest subsystems. Full information on all variables is available at the github repository.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab3\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 3\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eSubsystem name, number of variables, and two most highly connected variables and degree (number of connections to other variables) for the 10 largest subsystems in the linked data co-occurrence network\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"4\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cem\u003eSubsystem\u003c/em\u003e\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003e\u003cem\u003eNumber of factors\u003c/em\u003e\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003e\u003cem\u003eICD code where relevant\u003c/em\u003e\u003c/p\u003e \u003cp\u003e\u003cem\u003evariable description and (data source)\u003c/em\u003e\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003e\u003cem\u003edegree\u003c/em\u003e\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMixed NDRDD variables\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e\u003cem\u003e110\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eLiving in own home (NDRDD)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e14\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eCurrently prescribed methadone (NDRDD)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e13\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMental health\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e87\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eMental health services (NDRDD)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e7\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eVenlafaxine, anti-depressant (PIS)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e7\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eRespiratory infection, depression, pain\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e\u003cem\u003e82\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003ePropranolol, anti-anxiety (PIS)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e\u003cem\u003e7\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eChlorhexidine mouthwash (PIS)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e\u003cem\u003e7\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAssault and self-harm\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e\u003cem\u003e77\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eX999 assault by sharp object in unspecified place (SMR)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e14\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eS099_unspecified head injury (SMR)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e13\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAccidental and intentional self-poisoning\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e70\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eX619 self-poisoning by antiepileptic drug in unspecified location (SMR)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e12\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eX610 self-poisoning by antiepileptic drug at home (SMR)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e11\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eLiver cirrhosis\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e63\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eEnsure plus milkshake (PIS)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e9\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eN179 acute renal failure (SMR)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e8\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePhlebitis, hepatitis, mental health, substance use\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e\u003cem\u003e56\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eI802 phlebitis and thrombophlebitis of other deep vessels of lower extremities (SMR)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e\u003cem\u003e12\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eHepatitis C (NDRDD)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e10\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBenzodiazepines and pain\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e56\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eDiazepam (NDRDD)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e\u003cem\u003e9\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eGabapentin, analgesic/anticonvulsant (PIS)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e\u003cem\u003e9\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMixed stomach, skin, antibiotics\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e47\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eFlucloxicillin, antibiotic (PIS)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e\u003cem\u003e7\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eSimple linctus, cough syrup (PIS)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e\u003cem\u003e7\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMethadone and common prescriptions\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e\u003cem\u003e44\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eMethadone (PIS)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e\u003cem\u003e7\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eParacetamol (PIS)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e\u003cem\u003e6\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec12\" class=\"Section2\"\u003e \u003ch2\u003eSingle condition subsystems\u003c/h2\u003e \u003cp\u003eThere were 58 subsystems related to distinct health conditions and common forms of treatment, including: depression; diabetes; heart disease; chronic pain; and thyroid issues. These characterise the range of health conditions that occurred among people who experienced a DRD, although this analysis does not tell us whether these are more prevalent among those that died than among the wider population. The community detection method distinguished between various aspects of mental health, for example, depression treated with prescription medications appeared in a different subsystem than inpatient psychiatric hospital stays.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec13\" class=\"Section2\"\u003e \u003ch2\u003eCo-occurring condition subsystems\u003c/h2\u003e \u003cp\u003eEight subsystems related to multiple issues, including: assault and self-harm; respiratory infections, depression and pain; and stomach issues, skin conditions and infections. While this gives a further characterisation of some of common multimorbid conditions among the population, it also highlights potential points of intervention such as at the point of emergency room attendance after assault victimisation.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec14\" class=\"Section2\"\u003e \u003ch2\u003eSubstance use related subsystems\u003c/h2\u003e \u003cp\u003eThere were 12 subsystems that were directly related to substance use, including the large cluster of NDRDD variables, treatments such as \u003cem\u003eOpi\u003c/em\u003eoid \u003cem\u003eAgonist Therapy(OAT)\u003c/em\u003e, wound care and alcohol treatment, and also distinct co-occurrences around emergency room treatment. The cooccurrence of substance use with a fall is also notable as falling at ground level and receiving treatment could suggest high physical frailty or high levels of intoxication. This highlights a potential preventive intervention point in emergency room settings.\u003c/p\u003e \u003cp\u003eThe 10 most highly connected variables (and data source) in the co-occurrence network were, in descending order based on number of connections: \u003cem\u003eliving arrangements\u003c/em\u003e (NDRDD); \u003cem\u003eassault by sharp object\u003c/em\u003e (SMR); \u003cem\u003emethadone\u003c/em\u003e (OAT) (NDRDD); \u003cem\u003enot currently using OAT\u003c/em\u003e (NDRDD); \u003cem\u003esalbutamol\u003c/em\u003e (an asthma medication) (PIS); \u003cem\u003ehead injury\u003c/em\u003e (SMR), \u003cem\u003ePhlebitis of lower leg\u003c/em\u003e (SMR), \u003cem\u003eself-poisoning\u003c/em\u003e (SMR), \u003cem\u003eassault by bodily force\u003c/em\u003e (SMR), and \u003cem\u003eheroin use\u003c/em\u003e (NDRDD). Unlike the system map factors, the number of connections does not relate to the potential central causal role of the factors in the system. Instead, this gives an indication of how commonly each variable is related to other health and social factors. This suggests that these factors are important features of multimorbidity and potential aspects of syndemic ill health. While some factors are artefacts of the data (there were many living arrangements variables that are correlated with each other, and many ICD codes would be recorded at the same time in relation to an assault), these factors give a broad insight into factors that may predispose individuals to risk of poor health outcomes and that may be commonly recognised by health and social care providers.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec15\" class=\"Section2\"\u003e \u003ch2\u003eCommon and unique perspectives from system maps and linked data\u003c/h2\u003e \u003cp\u003eWhile there were several overlapping themes between the factors discussed in workshops that produced the system map and the subsystems identified in the linked data, there were substantial differences in emphasis between the two. Mental health and wellbeing was a central node in the system map and there were six subsystems in the linked data related to mental illness and its treatment, suggesting that this was a prominent feature of the clinical datasets and the collective causal understanding of the wider system. Pain was mentioned as a peripheral factor in the system map, mediating the connection between physical health and drug use as a coping mechanism, while in the linked data there were several subsystems describing different aspects of pain and pain medication. Heart and lung health was mentioned in the system map as factors mediating the connection between physical health and drug death, and seven subsystems in the linked data related to heart and lung health.\u003c/p\u003e \u003cp\u003eLinked data subsystems also uncovered connections which were not a feature of the system map, i.e. between: mental health and pain; mental health, respiratory health and pain, and the multiple patterns of co-occurrence visible in Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e. Poor treatment of people who use drugs was outlined as a key aspect of the stigma subsystem, while the linked data subsystem around assault gives insight into one aspect of negative social experiences and treatment. The linked data subsystem containing the \u003cem\u003esubstance use\u003c/em\u003e and \u003cem\u003etrip and fall\u003c/em\u003e variables \u0026ndash; indicative of intoxication and potential frailty \u0026ndash; corresponded with frailty as one of the proximal factors relating to death in the system map. Re-inspection of the linked data network found that \u003cem\u003efall\u003c/em\u003e ICD codes also appeared as a distinct, unconnected variable (unspecified fall), part of the liver cirrhosis subsystem (fall at home), and many variables in the assault subsystem (fall in road, down stairs, from a height etc.). Lastly, there were several linked data subsystems relating to alcohol dependence, treatment and harm, but discussion of alcohol was largely absent from the workshops and did not feature as a relevant factor in the system map. This was likely due to the framing of the workshop activities with DRDs as the system of interest, rather than substance use more broadly, and the professional roles and lived experience of the workshop delegates being aligned with drug use and treatment. The two data sources also provided several distinct perspectives: the linked data gave an overview of detailed clinical characteristics of the sample beyond the scope of the system map, while the system map covered many aspects of the workforce as well as social, political, and economic factors that were not captured in the linked data.\u003c/p\u003e \u003c/div\u003e"},{"header":"Discussion","content":"\u003cp\u003eApplying a mixed method approach that integrates systems thinking, data linkage and network analysis, this study provides detail on the interconnections between factors that contribute to DRDs in Scotland, how these factors are structured, and a novel perspective on the levels of influence through which complex structures have their effects. The process of developing the systems map and co-occurrence network informed action areas for DRD prevention plans among the research team and workshop delegates, and informed wider policy discussions. The maps are openly available for re-use in systems-informed approaches to drug death prevention.\u003c/p\u003e \u003cp\u003eOur first research question aimed to identify central factors in the DRDs system. We found stigma, mental wellbeing, and poverty in the system map, and living arrangements and assault within the linked data. Both findings point towards the important facets of the social and political environment, rather than solely on aspects of individual health or biological considerations. Targeting improvements around these areas may leverage wider change, potentially producing positive change or mitigating the negative influence of wider factors affecting risk of death. Interventions on these factors must appropriately account for the wider system structures surrounding them.\u003c/p\u003e \u003cp\u003eOur second research question aimed to explore the substructures within the DRDs complex system. The two data sources provided broad coverage of influences at multiple levels, from the individual to political. Subsystems within the linked data correspond with common clinical experience of treatment and prescribing, and provide a comprehensive summary of the range of co-occurring health and social factors among the population who experienced a DRD. The system map drew upon a wealth of expertise and multiple perspectives and provides a broad view of the wider social and political system. The key factors were in broad agreement with wider literature on the bio, psycho, social, and political aspects of drug related harms (\u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e), and provide a new way of contextualising the connections between the diverse areas of personal and public life that affect wellbeing thereby serving as potent points of intervention. Ingram et al. modelled the network of drug use symptoms that may be risk factors for experiencing a self-reported drugs overdose (\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e) finding that drug tolerance and withdrawal symptoms were central factors for risk of experiencing a self-reported drug overdose. This survey-based analysis aligns with our system map subsystem around proximal influences on DRD.\u003c/p\u003e \u003cp\u003eOur findings also suggest avenues to explore the potential occurrence of syndemic ill health. Previous work has pointed towards worse health outcomes driven by the co-occurrence of substance use and violence, co-occurring physical and mental health issues (\u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e), and other combinations of syndemic ill health. The subsystems identified here described related areas of health risk, as well as additional issues of co-occurrence. Future research could identify whether the condition clusters outlined in this analysis are associated with risk of poorer outcomes, which would require new data linkages with information on living individuals in addition to the data reported in this study.\u003c/p\u003e \u003cp\u003eIn addition to the comprehensive, holistic perspective on DRDs that system mapping methodologies provide (\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e), network analysis provides ways to explore the underlying structure of relationships and interactions that comprise the wider system. Community detection approaches to explore subsystems, in combination with a social ecological classification of factors in the system map, facilitates the development of important conceptual insights with implications for future research, policy and practice. The social ecological model and \u0026ldquo;upstream\u0026rdquo; metaphor are two common features of public health discourse around health improvement and addressing inequalities (\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e). These are heuristic rather than analytical tools, used to encourage the consideration of health across multiple levels (\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e). A complex systems framing emphasises the mutual interdependence between elements across all levels of a social system (\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e) and emphasises that points of greatest leverage may be found at multiple levels. Complexity framings have been used to argue against public health interventions on the basis that appropriate leverage points are too difficult to identify (\u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e). A key contribution of our approach is to delineate subsystems of related factors that allow for tractable action planning compared to considering the system as a whole, while also going beyond a heuristic consideration of levels of influence, independent of the key structures and interactions that influence outcomes. For example, viewing the whole system map, there are many long causal chains (over 16,000) outlining processes linking stigmatising norms to death. These complex, multiple pathways which may operate over different time frames make it difficult to draw inference on the statistical association between stigma and death and thus poses problems when evaluating the effect of stigma reduction interventions on death rates.\u003c/p\u003e \u003cp\u003eViewing the map in terms of underlying structures, the stigma subsystem comprised factors ranging from the wider environment through to individuals\u0026rsquo; experience of services, and the death subsystem considered physical health, drug toxicity and knowledge of drug effects. The substructures uncovered by community detection suggest that stigma interventions could be more effectively evaluated in terms of outcomes related to organisational level of the social ecological model (e.g. interactions with services), rather than in terms of death rates. Considering DRDs in terms of structurally coupled subsystems facilitates a polycontextual approach (\u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e28\u003c/span\u003e), meaning that the targets of action, ways of working, and data collection should vary according to the context of the subsystem, rather than applying a single perspective across all factors (e.g. statistical associations between various factors in the system and DRD).\u003c/p\u003e \u003cp\u003eA subsystem approach may pose a challenge for health research and practice. It is simpler to focus on health at the individual level and within the body rather than evaluate change across a complex social system. Additionally, it may an expectation of public health bodies and research funders to take an individual focus. The end result is that research effort is directed towards \u0026ldquo;short causal chain\u0026rdquo; interventions, such as pharmacological interventions affecting individual causal processes, or upstream interventions with short causal pathways (e.g. Minimum Unit Pricing\u0026thinsp;\u0026gt;\u0026thinsp;purchasing behaviour\u0026thinsp;\u0026gt;\u0026thinsp;physical health) (\u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e29\u003c/span\u003e), that can be less prone to dilution due to multiple pathways operating over different time scales, at the expense of approaches which may more comprehensively effect system level change but which require longer time frames and more holistic forms of evaluation. Network methods used to summarise substructures could improve how systems mapping approaches inform the identification of sets of relevant indicators for change and outcome assessment at multiple levels, helping to restructure \u0026ldquo;models of evidence\u0026rdquo; (\u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e30\u003c/span\u003e) in systemic, rather than individual, terms.\u003c/p\u003e \u003cp\u003ePrevious system mapping exercises applied to health have often taken the approach of predefining the subsystems or domains as starting points for a system mapping exercise. For example, Finegood et. al applied network analysis to a reduced form of the Foresight obesity map to show the strength of connections between the eight clusters, including food production, physical activity, individual psychology etc. (\u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e31\u003c/span\u003e) While providing a helpful viewpoint of the connections between these clusters of factors, the interpretation is constrained by the starting point of categorising factors into domains. By comparison, we started with an open view of the system, going from rich pictures to various sub-systems, with community detection returning cross-disciplinary and cross-level sets of factors. We propose that system mapping activities would benefit from avoiding discipline- or theme-directed factor elicitation and instead utilise more open data collection methods, followed by analytical approaches that can uncover emergent structures.\u003c/p\u003e \u003cp\u003eThe maps provide tools to inform future policy and intervention development. For example, anyone proposing an intervention targeting one node on the system map could navigate upstream and downstream from their target node to think through potential points of constraint, resistance, or support for the actions on the target node. Policy makers focusing on reducing substance use stigma could consider the ways in which the factors in the stigma subsystem may have an influence at a national level, the extent to which local investment in community development may augment stigma reduction, or how external influences such as media representation may counteract intended effects of a policy. Visual navigation of the maps - moving upstream, downstream or around loops from any focal issue - provides a \u0026ldquo;thinking tool\u0026rdquo; to navigate any intended intervention or action plan from multiple levels and perspectives. Such an approach can enhance the quality of decision making and the identification of the most effective set of actions.\u003c/p\u003e \u003cp\u003eThe maps can also aid evaluation, particularly at policy level. For example considering the Scottish Government national drugs mission plan (\u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e32\u003c/span\u003e), against the system maps we find some alignment between the six cross-cutting priorities, but with notable gaps and differences in perspective. For example the mission\u0026rsquo;s focus on \u0026lsquo;lived and living experience at the heart\u0026rsquo; (Priority 1) was reflected in subsystems around public perspectives on substance use, as well as stigmatising attitudes. While the mission focused on developing ways for lived experience involvement, the system map additionally highlighted issues of value and relative status for those involved with organisations. Equalities and human rights (Priority 2) did not appear as a prominent set of factors, although housing, living standards and poor treatment were evident as important factors in the system map and linked data. Tackling stigma (Priority 3) was a prominent feature of our analysis: a key insight from our analysis was to uncover the various levels at which stigma may have its effect. In addition to taking action at the level of service provider and media organisations, our findings point to the importance of tackling public perceptions and poor treatment in terms of interpersonal and community experiences. Subsystems and factors related to safe physical environments and social and community influences were not as clearly represented in the national mission. This may reflect a tendency for policy and action planning to focus either on the individual level factors (e.g. naloxone) or on upstream levels (e.g. change in national or local funding), at the expense of paying attention to the intermediate level factors which bridge micro and macro-level processes.\u003c/p\u003e \u003cp\u003eThe integration of two data sources (RQ3) uncovered a difference in focus in relation to alcohol dependence and treatment which is noteworthy from a methodological point of view: preliminary analysis of the NDRDD was presented at workshops, and alcohol treatment and detoxification was a central subsystem in this analysis (see Appendix Fig.\u0026nbsp;1). Despite this data-informed aspect of the workshop, alcohol was not strongly reflected in the subsequent development of the system map. To the authors\u0026rsquo; knowledge, this is the first integration of linked data co-occurrence networks into a system mapping process. While there are \u0026lsquo;scripts\u0026rsquo; or formal processes to document ideas from the group (e.g. Nominal group technique), no such scripts exist for data integration; the presentation of the networks involved open discussion and freehand annotation of the printed networks. Future research could develop new formal process to structure how co-occurrence networks \u0026ndash; or other sources of data - are discussed in groups, and to document in what ways the quantitative data is represented, or purposefully omitted, from workshop final outputs.\u003c/p\u003e \u003cdiv id=\"Sec17\" class=\"Section2\"\u003e \u003ch2\u003eStrengths and limitations\u003c/h2\u003e \u003cp\u003eThis study provides a wealth of insight on the structures and features of the complex system surrounding DRDs. The combination of linked administrative data and a coproduction systems thinking model provides a rigorous framework for understanding health outcomes in new ways. We combined these methods with the 6SQuID framework, providing a robust output for Step 1: understanding the problem and its causes, and Step 2: identifying which causal or contextual factors are modifiable, with the greatest scope for change, and who would benefit most. The workshop participants identified three action areas and started to consider specific activities in each area, partially supporting Step 3: deciding on the mechanisms of change. These can serve as a starting point for future intervention development.\u003c/p\u003e \u003cp\u003eThere are some limitations to the study. Firstly, while the system map represents the collective view of a diverse stakeholder group, it may overlook some key perspectives, offer little in-depth information in specific areas, and may not account for changes in the wider DRDs system since the time of the workshops. We would encourage readers to access the supplementary data and modify and expand them so that they are better suited to specific context and intended use. Additionally, when linking data, it was assumed that an individual who was absent from a database did not have any of the conditions noted in that database e.g. those not in SMR04 (Psychiatric inpatient) database were assumed not to have any of the psychiatric diagnoses noted in SMR04. This means that the prevalence, and thus co-occurrence for some conditions, is likely to have been underestimated. However, this limitation is partially mitigated by the inclusion of the prescription information. Community dispensing of prescriptions for mental health conditions are noted in the data and distinct mental health subsystems appeared e.g. depression treated with prescriptions and depressive disorder noted in inpatient records. An issue that remains is that those who avoid all contact with health services are under-represented in this study design.\u003c/p\u003e \u003cp\u003eAs our dataset focussed only on those who have died, there is the potential for collider bias in our dataset. If two factors are independently associated with a higher risk of death, they will appear negatively correlated in a dataset of those who have died, even if there is no correlation in the full population (\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e). This limitation is partly overcome as we focussed on identifying co-occurrence and removed all negative associations, rather than interpreting negative correlations as casual. However, collider bias may have the potential to underestimate the correlation between factors identified here. Further research to compare co-occurrence in living populations could investigate this possibility and better estimate the prevalence of co-occurrence and magnitude of causal associations between co-occurrence and health outcomes.\u003c/p\u003e \u003cp\u003eAnother point to note is that the use of community detection methods has some limitations. The Louvain algorithm allocates every variable into exactly one distinct community. While the most highly correlated factors reliably appear in the same community, less clearly connected factors may be assigned to one community or another at random. Running the analysis with different random number seeds found differing numbers of subsystems. While these tended to have broadly similar groupings of variables that were comparable across replications, the findings should be considered as providing broad signals around the potential structure and behaviour of highly complex systems, rather than providing deductive inference around co-occurrence patterns or causal processes. This reinforces the importance of verifying whether the subsystems and structures found in this analysis occur in other datasets, such as living individuals, and discerning which patterns of co-occurrence can be safely ruled out as being risk factors and which require further attention as potential areas of syndemic ill health. Finally, our selected methods could not account for the timing or sequence of events: our analysis represents any co-occurrence and does not consider trajectories in experience of diseases, prescribing, or social factors. The NDRDD data relates to six months prior to death, while other linked data extends further back in the life course. Future research should consider pathways via connections that occur within certain time frames, study changes over time considering age, period or cohort effects, or sequence analysis. Such research could strengthen understanding of how co-occurrence relates to elevated risk of death and shed light on further points for preventive intervention.\u003c/p\u003e \u003c/div\u003e"},{"header":"Conclusions","content":"\u003cp\u003eThis study, and the open source system artefacts describing the findings from the project, provide a novel representation of the complex system surrounding a key priority for public health and health inequalities. These outputs can provide a starting point for future research and action to prevent such deaths at local intervention and national policy level. While the systems approach provided a wide range of potential intervention points, a strategic approach to DRDs requires further evidence, funding, political will to take coordinated, committed action across the wider social system.\u0026nbsp;\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eDeclarations and acknowledgements\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eFor the purpose of open access, the authors have applied a Creative Commons Attribution (CC BY) licence to any Author Accepted Manuscript version arising from this submission.\u003c/p\u003e\n\u003cp\u003eWe are very grateful to all the participants for their contribution and insights throughout the workshop series. Many thanks to those providing feedback on the study after presentations at the Society for Social Medicine, International Network for Social Network Analysis, and Emergency Medicine at the deep end meetings.\u003c/p\u003e\n\u003cp\u003eThe authors would like to acknowledge the support of: the eDRIS Team (Public Health Scotland) for their involvement in obtaining approvals, provisioning, and linking data and the use of the secure analytical platform within the National Safe Haven; National Records of Scotland for data linkage; Scottish Exchange of Data (Scottish Government) for allowing us to access the data; the Scottish Drugs Forum for inviting lived experience participants; and the Drugs Research Network for Scotland for help coordinating the workshop series.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eWe would like to acknowledge the support of the following colleagues who helped to run the workshops: Jessica Greenhalgh, Julie Riddell, Danilo Falzon, Catriona Connell, Hazel Booth, Wendy Masterson.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eEthics and consent: The Public Benefit and Privacy Panel for Health and Social Care approved access to the linked data (Ref: 1920-0196), with endorsement from the University of Glasgow College of Medicine, Veterinary and Life Sciences ethics committee. Ethical approval for the systems workshops was obtained from the University of Stirling General University Ethics Panel (GUEP 19 20 861) and participants consented to take part.\u003c/p\u003e\n\u003cp\u003eConsent for publication: Not applicable\u003c/p\u003e\n\u003cp\u003eAvailability of data and materials: Individual level linked data are not publicly available; contact researchdata.scot to make a data access request. Co-occurrence matrices, system maps and analytical scripts are available at github.com/MRC-CSO-SPHSU/NDRDD-linkage.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eFunding: This project was funded by the Scottish Government Chief Scientist Office (HIPS/19/32). MMcC, RS, SL and KS were part of the Relationships and Health programme in the MRC/CSO Social and Public health Sciences Unit, funded by the Medical Research Council (MC_UU_00022/3) and the Scottish Government Chief Scientist Office (SPHSU18).\u003c/p\u003e\n\u003cp\u003eAuthor contributions:\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eConceptualization: MMcC, TP, LB, CM, JS, JH, AB, AC, KS.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eData curation: MMcC, SL, RS.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eFormal analysis: MMcC, SL, KS, RS.\u003c/p\u003e\n\u003cp\u003eFunding acquisition: MMcC, TP, LB, CM, JS, JH, AB, AC, KS.\u003c/p\u003e\n\u003cp\u003eInvestigation: MMcC, RS, SL, TP, LB, CM, JS, JH, AB, AC, KS, Jessica Greenhalgh, Julie Riddell, Danilo Falzon, Catriona Connell, Hazel Booth, Wendy Masterson.\u003c/p\u003e\n\u003cp\u003eMethodology: MMcC, SL, TP, LB, CM, JS, JH, AB, AC, KS.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eProject administration: KS, Jessica Greenhalgh.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eResources: Electronic Data Research and Innovation Service.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eSoftware: MMcC, SL, RS.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eSupervision: KS.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eValidation: MMcC, SL, TP, LB, CM, JS, JH, AB, AC, KS.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eVisualization:\u0026nbsp;MMcC, SL, KS, RS.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eWriting original draft: MMcC, RS.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eWriting review and editing: SL, TP, LB, CM, JS, JH, AB, AC, KS.\u0026nbsp;\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eScotland NRo. Drug-related deaths in Scotland, 2024. 2025.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eScotland PH. National drug related death database (Scotland). Analysis of deaths occurring in 2017 and 2018.; 2022.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBrown D, Allik M, Dundas R, Leyland AH. All-cause and cause-specific mortality in Scotland 1981\u0026ndash;2011 by age, sex and deprivation: a population-based study. Eur J Pub Health. 2019;29(4):647\u0026ndash;55.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eScotland NRo. Drug-related deaths in Scotland in 2022. 2022.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBarbrook-Johnson P, Penn AS. Systems Mapping: How to build and use causal models of systems. Springer Nature; 2022.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMendenhall E, Newfield T, Tsai AC. Syndemic theory, methods, and data. Social Science \u0026amp; Medicine (1982). 2022;295:114656.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCoid J, Zhang Y, Bebbington P, Ullrich S, De Stavola B, Bhui K, et al. A syndemic of psychiatric morbidity, substance misuse, violence, and poor physical health among young Scottish men with reduced life expectancy. SSM-population Health. 2021;15:100858.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRod MH, Rod NH, Russo F, Klinker CD, Reis R, Stronks K. Promoting the health of vulnerable populations: three steps towards a systems-based re-orientation of public health intervention research. Health Place. 2023;80:102984.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSchofield J, Papathomas M, Macnamara C, McCann M, Ardestani BM, Skivington K et al. Contextualising multimorbidity in people who use drugs: analysis of drug-death decedents in Scotland. Ir J Psychol Med. 2025:1\u0026ndash;9.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLee J, Bainter S, Carrico A, Glynn T, Rogers B, Albright C, et al. Connecting the dots: a comparison of network analysis and exploratory factor analysis to examine psychosocial syndemic indicators among HIV-negative sexual minority men. J Behav Med. 2020;43:1026\u0026ndash;40.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMeadows D. Places to Intervene in a System. Whole Earth. 1997;91(1):78\u0026ndash;84.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWight D, Wimbush E, Jepson R, Doi L. Six steps in quality intervention development (6SQuID). J Epidemiol Community Health. 2016;70(5):520\u0026ndash;5.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTurner SF, Cardinal LB, Burton RM. Research design for mixed methods: A triangulation-based framework and roadmap. Organizational Res methods. 2017;20(2):243\u0026ndash;67.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWilliams DR. GGMnonreg: Non-Regularized Gaussian Graphical Models in R. J Open Source Softw. 2021;6(67):3308.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBanack HR, Mayeda ER, Naimi AI, Fox MP, Whitcomb BW. Collider Stratification Bias I: Principles and Structure. Am J Epidemiol. 2024;193(2):238\u0026ndash;40.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eUlrich W, Reynolds M. Critical systems heuristics. Systems approaches to managing change: A practical guide. Springer; 2010. pp. 243\u0026ndash;92.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBarbrook-Johnson P, Penn AS. Rich Pictures. Systems Mapping: How to build and use causal models of systems: Springer; 2022. pp. 21\u0026ndash;32.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCsardi G, Nepusz T. The igraph software. Complex syst. 2006;1695:1\u0026ndash;9.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGmbH y. yEd 2024 [Available from: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.yworks.com/products/yed\u003c/span\u003e\u003cspan address=\"https://www.yworks.com/products/yed\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDahlgren G, Whitehead M. The Dahlgren-Whitehead model of health determinants: 30 years on and still chasing rainbows. Public Health. 2021;199:20\u0026ndash;4.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePedersen TL, Pedersen M, LazyData T, Rcpp I, Rcpp L. Package \u0026lsquo;ggraph\u0026rsquo;. Retrieved January. 2017;1:2018.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCrielaard L, Quax R, Sawyer AD, Vasconcelos VV, Nicolaou M, Stronks K, et al. Using network analysis to identify leverage points based on causal loop diagrams leads to false inference. Sci Rep. 2023;13(1):21046.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMcCann M. Github page for network analysis of Scotland's Drug-related death system 2025 [Available from: github.com/MRC-CSO-SPHSU/NDRDD-linkage.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRhodes T. Risk environments and drug harms: a social science for harm reduction approach. Elsevier; 2009. pp. 193\u0026ndash;201.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eIngram PF, Bailey AJ, Finn PR. Applying network analysis to investigate substance use symptoms associated with drug overdose. Drug Alcohol Depend. 2022;234:109408.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTsai AC. Syndemics: a theory in search of data or data in search of a theory? Soc Sci Med. 2018;206:117\u0026ndash;22.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePetticrew M, Katikireddi SV, Knai C, Cassidy R, Hessari NM, Thomas J, et al. Nothing can be done until everything is done\u0026rsquo;: the use of complexity arguments by food, beverage, alcohol and gambling industries. J Epidemiol Community Health. 2017;71(11):1078\u0026ndash;83.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMeyer S, Gibson B, Ward P. Niklas Luhmann: Social Systems Theory and the Translation of Public Health Research. In: Collyer F, editor. The Palgrave Handbook of Social Theory in Health, Illness and Medicine. London: Palgrave Macmillan UK; 2015. pp. 340\u0026ndash;54.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBeeston C, Robinson M, Giles L, Dickie E, Ford J, MacPherson M, et al. Evaluation of minimum unit pricing of alcohol: a mixed method natural experiment in Scotland. Int J Environ Res Public Health. 2020;17(10):3394.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRutter H, Savona N, Glonti K, Bibby J, Cummins S, Finegood DT, et al. The need for a complex systems model of evidence for public health. lancet. 2017;390(10112):2602\u0026ndash;4.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFinegood DT, Merth TD, Rutter H. Implications of the foresight obesity system map for solutions to childhood obesity. Obesity. 2010;18(S1):S13\u0026ndash;6.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGovernment S. National Drugs Mission Plan: 2022\u0026ndash;2026. 2022.\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"bmc-public-health","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"pubh","sideBox":"Learn more about [BMC Public Health](http://bmcpublichealth.biomedcentral.com/)","snPcode":"","submissionUrl":"https://www.editorialmanager.com/pubh/default.aspx","title":"BMC Public Health","twitterHandle":"@BMC_series","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"em","reportingPortfolio":"BMC Series","inReviewEnabled":true,"inReviewRevisionsEnabled":true},"keywords":"Drug-related deaths, co-occurring conditions, comorbidity/multimorbidity, syndemics, co-production methods, systems science, data linkage, network analysis, systems mapping","lastPublishedDoi":"10.21203/rs.3.rs-8767159/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-8767159/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003e\u003cstrong\u003eBackground\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eDrug-related deaths are rising in many countries. Understanding connections between health conditions, social experiences, and broader political factors may enhance death prevention efforts. Systems science methods can help identify areas for effective interventions. This study aimed to understand the complex system relating to drug-related deaths in Scotland.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eMethods\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eWe used gaussian graphical models to identify co-occurring variables in linked datasets: Public Health Scotland’s National Drug-Related Deaths Database, Prescribing Information System, and Scottish Morbidity Records for inpatient and day case stays in acute and psychiatric hospitals (n = 6,608). Preliminary findings were integrated into co-production workshops.\u003c/p\u003e\n\u003cp\u003eWe conducted a systems-informed intervention development study using the 6SQuID Intervention Development framework. We facilitated co-production workshops using soft systems methods and system mapping.\u003c/p\u003e\n\u003cp\u003eSystem map factors were coded according to the social ecological model of health. Network metrics and Louvain community detection were applied to the system map and linked data graphical model. We applied a mixed method visual, text, and numeric approach to integrate findings from both data sources.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eResults\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eStigma, mental health and poverty were central factors in the system map; while population data found living arrangements and assault as central. Linked data analysis found 78 subsystems; 58 related to distinct conditions, eight to co-occurring conditions, and 12 to substance use. System map analysis found eight subsystems including direct causes of death, life experiences, stigmatising attitudes, treatment services and public perspectives. All subsystems contained factors across multiple social ecological levels, but no single system traversed all levels. Assault and alcohol treatment and harms were distinct subsystems in the linked data but less prominent in the system map; while frailty and housing were common features of both data sources.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConclusions\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eIntegrating systems science and social ecological perspectives with network analysis provides novel insights into complex health issues, with practical implications for designing interventions. Future policy and practice should consider how to align actions more closely to system-level outcomes alongside individual and clinical outcomes. Increased attention to social, community and living environments would give Scottish drug death policy more comprehensive coverage of the drug death system.\u003c/p\u003e","manuscriptTitle":"Complex system structures around drug-related deaths in Scotland: a mixed method network analysis approach to study subsystems in population data and system maps","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-02-22 12:41:31","doi":"10.21203/rs.3.rs-8767159/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"editorInvitedReview","content":"","date":"2026-02-23T16:09:13+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"105349586211132347500806837704822013198","date":"2026-02-19T11:10:45+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"206356661939256322406092419669436465957","date":"2026-02-18T15:18:55+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2026-02-16T15:00:19+00:00","index":"","fulltext":""},{"type":"editorInvited","content":"","date":"2026-02-06T16:10:01+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2026-02-04T01:47:03+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2026-02-04T01:46:15+00:00","index":"","fulltext":""},{"type":"submitted","content":"BMC Public Health","date":"2026-02-02T15:53:40+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"bmc-public-health","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"pubh","sideBox":"Learn more about [BMC Public Health](http://bmcpublichealth.biomedcentral.com/)","snPcode":"","submissionUrl":"https://www.editorialmanager.com/pubh/default.aspx","title":"BMC Public Health","twitterHandle":"@BMC_series","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"em","reportingPortfolio":"BMC Series","inReviewEnabled":true,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"acef497b-17ac-4c8a-9eee-f96bba98fcd3","owner":[],"postedDate":"February 22nd, 2026","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"under-review","subjectAreas":[],"tags":[],"updatedAt":"2026-02-22T12:41:31+00:00","versionOfRecord":[],"versionCreatedAt":"2026-02-22 12:41:31","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-8767159","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-8767159","identity":"rs-8767159","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}