MONARCH: Map of Need Aggregated Research Study
The MONARCH project represents the UK's first consolidated dataset tracking how military veterans and the Armed Forces Community access charity services. Developed at Northumbria University, this report presents the platform's development, capabilities, evaluation findings, and future direction with the overarching aim of enabling early intervention and preventing veterans reaching crisis point.
Executive Summary
Purpose & Context
The UK's first consolidated dataset tracking veteran access to Armed Forces charity services, enabling tailored service provision and early intervention before crisis.
Platform Development
Casework data from 10 charities consolidated through an eight-phase process, accounting for assistance from a further 173 organisations via almonisation.
Platform Capabilities
Interactive SAS Viya dashboard enabling geospatial analysis, demographic filtering, and needs mapping at local, regional, and national levels.
Evaluation Results
Three-month prototype evaluation (July–September 2025) with 17 stakeholders. Mean SUS score of 74.12 which is within the 'acceptable' usability range.
Impact & Future Direction
A transformative shift from reactive to proactive veteran support, aligned with the NHS 10-year plan's emphasis on early intervention and crisis prevention.
The MONARCH Team
Principal Investigators
- Primary Investigator: Professor Matt Kiernan
- Co-Investigators: Professor Marco Tomietto & Professor Gavin Oxburgh
Research Team
- Senior Research Assistant: Dr Amy Johnson
- Research Assistant: Andrew McGill
- Charity Relationship Manager: Meri Mayhew
Visiting Scholars
- Dr Michael Rodrigues
- Dr Giuseppe Serra M.D.
- Dr Federico Turoldo M.D.
Funding & Support
The MONARCH project was funded by The Armed Forces Covenant Fund Trust and Forces in Mind Trust. Technical support was provided by Analytium (SAS Viya platform) and Northumbria University across all eight development phases.
Acknowledgements & Participating Organisations
We extend sincere thanks to the 10 Armed Forces charities who generously shared their data whilst maintaining the highest standards of data protection and veteran confidentiality.
We are also grateful to the 29 focus group participants from NHS England, NHS Wales, Armed Forces charities, and local government (October–December 2023), and to the 37 organisations who participated in the platform evaluation (July–September 2025). Special recognition goes to the Data Strategy Working Group whose expertise enabled harmonisation of complex datasets.
Introduction: Background & Context
UK military veterans present complex, overlapping needs spanning physical health (hypertension, ischaemic heart disease, type 2 diabetes), mental health (PTSD, depression), and social deprivation (homelessness, food insecurity). Despite NHS and charity sector support, research consistently highlights poor help-seeking behaviour — many veterans only seek support at the point of crisis.
The Reactive Problem
Both the NHS and Armed Forces charity sector have historically operated a reactive model whereby intervention only takes place when veterans self-declare a need. The NHS 10-year plan now reflects a shift towards proactive, preventative approaches.
The Data Challenge
A wide range of agencies collect data on veterans, but limited connectivity between systems has hampered effective linkage. The process of almonisation, where multiple charities support the same individual, created significant risks of double counting, resulting in unreliable datasets and potential over-estimation of need.
The MONARCH Project: Aims & Objectives
MONARCH was developed to map Armed Forces charity provision across the UK at local, regional, and national levels — something never previously achieved. The project had two key objectives:
1
National Aggregated Platform
Create a national, aggregated, and dynamic visual data platform of Armed Forces charity data providing appropriate information to service providers.
2
Robust Research Dataset
Create a methodologically robust dataset for scientific investigation, encompassing data design and different coding schemas to support secondary analysis and policy development.
Full ethical clearance was obtained from the Health and Life Sciences Ethics Department, Northumbria University (reference number: 2055) before the project commenced.
Platform Development: Eight-Phase Process
The MONARCH platform's development was separated into eight distinct phases within three broad categories (see image on left): (i) development; (ii) the platform itself; and (iii) evaluation. These are summarised across four key sections below.
Each phase was built upon the last, ensuring the resulting platform was technically robust, ethically sound, and practically useful for policymakers and service providers across the UK.
Phase 1: Focus Groups with Key Policymakers
Four focus groups (between October - December 2023) included 29 representatives from NHS England, NHS Wales, Armed Forces charities, and local government. Braun and Clarke's thematic analysis identified three main themes:
1: Overview of Services and Needs
Sub-themes:
Identifying needs; Resource planning
Summary:
Participants emphasised evidencing complex needs, supporting resource planning, and preventing crisis through proactive identification of risk factors.
2: Locations
Sub-themes:
Geospatial analysis
Summary:
Geospatial analysis at local, regional, and national levels deemed essential to understand geographic variation in needs and service provision.
3: Dataset
Sub-themes:
Pre-existing datasets; Autonomy and Accessibility
Summary:
Participants requested an interactive online platform compatible with existing datasets, with user guidance to prevent misinterpretation and ensure GDPR compliance.
Despite sustained efforts, focus groups with policymakers from Scotland and UK Central Government could not be organised due to scheduling constraints.
Phases 2 & 3: Data Sharing, Anonymisation & Cleaning
Data Sharing
Ten charities contributed data through formal data-sharing agreements with full GDPR and UK Data Protection Act compliance. Legally approved agreements were signed between Northumbria University and each charity. The aggregated dataset is managed via SAS Viya, with each charity retaining access to their own data.
Secure Hashing Algorithm (SHA)
To address double counting caused by almonisation, a SHA was applied to each case using: (i) gender; (ii) date of birth, and; (iii) military service number to generate a unique 64-digit anonymous ID. A cryptographic 'salt' string ensured data could not be reverse engineered.
Results showed the SHA identified unique cases with significantly outperforming previous approaches. Data preparation took approximately six months due to variation in charity data collection practices.
Phase 4: Data Almonisation, Aggregation & Harmonisation
A Data Strategy Working Group refined 280 different recorded needs into 51 distinct categories, mapped to 11 All Party Parliamentary Group complex need criteria and WHO's three health elements.
Single Need (37%)
Single contact for a single need — not returning suggests the need was most probably met.
Multiple Needs (23%)
Single contact for multiple needs — likely resolved as the veteran did not return.
Single Need – Multiple Contacts (33%)
Multiple contacts for the same need — suggests support received was insufficient and issues remain ongoing.
Multiple Needs – Multiple Contacts (7%)
Multiple contacts with new needs each time — suggests an underlying problem not being adequately addressed.
Key finding: 40% of veterans had repeated contact with a charity, suggesting sustained or recurring needs. 70% presented with more than one type of need. Younger non-officer veterans were disproportionately represented among those with complex and repeated needs.
Veteran Needs: Data Analysis
Key Demographics (42,264 veterans analysed)
- 89% male, average age 60 years
- 64% receiving other benefits
- 65% of needs were social (financial, housing, employment)
- 34% contacted charities for physical health needs
- 0.6% for mental health needs
The highest complexity was observed among veterans who were homeless or in custody. Veterans who were married with family support demonstrated lower service contact, suggesting a protective effect of social stability.
Phase 5: Data Visualisation
Built in SAS Viya (already used by MoD and HM Government departments), the platform presents census data, veteran demographics, and needs analysis with dynamic filtering at local, regional, and national levels. Examples can be found below:
Opening Page
Central navigation hub with links to About MONARCH, How to Use, and Troubleshooting sections.
Please click on the image to enlarge it
Summary Page
Displays headline figures: 78,586 total individuals including 55,376 veterans, 18,937 family members, 1,279 serving, and 2,994 not recorded.
Please click on the image to enlarge it
Census & Demographics
Interactive census data with demographic breakdowns including age, gender, rank, service branch, benefits receipt, and marital status.
Please click on the image to enlarge it
Needs Data
Needs mapped to APPG complex need criteria and WHO health determinants, filterable by geography and year to identify trends.
Please click on the image to enlarge it
Platform Capabilities
Early Intervention
Identify risk factors and emerging needs before veterans reach crisis point.
Geographic Analysis
Understand geographic variations in veteran needs and service provision at local, ICB, regional, and national levels.
Evidence-Based Planning
Support resource allocation, service planning, and funding bids with robust, multi-source data.
Advanced Analytics
Machine learning models and AI predictive tools already developed to identify clusters of need and risk factors for charity access.
Secondary Data Analysis: Risk Factors for Charity Access
The MONARCH platform has the capacity to scale beyond previous data management and visualisation functions. The aggregated dataset provides a foundation for further secondary analyses and the integration of additional analytical features to support policy decision-making. The research team has already created machine learning models for the identification of specific clusters of health and social care needs among veterans (outlined previously) and AI predictive models for identifying the risk factors leading veterans to access military charities. A summary of the latter paper is provided below, however, please also refer to Appendix B for the full published paper.
Summary of Previously Published Paper
Key factors were identified which were associated with veterans seeking support from the Armed Forces charity sector with the focus on identifying those most at risk of needing support at crisis. This case control study compared a randomised sample of 838 veterans from the MONARCH dataset to 838 controls who reported never seeking support from Armed Forces charities, identified through a previous study focusing on veteran food insecurity. Risk factors were identified via regression analysis and machine learning models. The key risk factors of those more likely to seek support were as follows:
3 x
Working age veterans
Those under 66 years old
17.5 x
Non-officers
At time of discharge
12 x
Living alone
Veterans living alone
5 x
Rented accommodation
Veterans residing in rented accommodation
1.4 x
Female veterans
Increased the likelihood
Having dependants was protective for veterans living with others, but became a significant risk factor for veterans living alone, underscoring the importance of social context in interpreting data.
Identifying and understanding factors such as these, early intervention and targeted approaches could be developed to prevent veterans reaching crisis level. In this vein, the MONARCH platform provides a scalable solution for secondary data analyses which could support policy makers in allocating health and social care resources and providing tailored solutions for health and social care provision in public health.
Evaluation of the MONARCH Platform
The prototype was disseminated to 66 stakeholders across 38 organisations for three months (July–September 2025). A convergent mixed-methods evaluation using the System Usability Scale (SUS) and a qualitative survey was completed by 17 respondents.
SUS Results
Scores ranged from 32.5 – 97.5 (Mean = 74.12, SD = 17.52). The mean falls within the acceptable usability range (≥70).
- 65% — Acceptable usability (11 respondents)
- 17% — High marginal usability (3 respondents)
- 12% — Low marginal usability (2 respondents)
- 6% — Significant concerns (1 respondent)
Useability Notes
The overwhelming majority of users viewed the platform as having acceptable or high marginal usability. Those reporting lower scores cited limited platform exposure, the need for more definitions, and accessibility challenges related to neurodiversity.
Respondent demographics
Details of all respondent demographics are outlined in the table below: *Three participants reported multiple health conditions. Demographic Response Respondents Age 20-30 years 31-40 years 41-50 years 51+ years Prefer not to say/did not answer 2 (11.76%) 2 (11.76%) 4 (23.53%) 7 (41.18%) 2 (11.76%) Gender Male Female Prefer not to say/ did not answer 9 (52.94%) 6 (35.29%) 2 (11.76%) Job role Executive/Director Civil Servant Research Charity Sector Management Prefer not to say/ did not answer 5 (29.41%) 3 (17.65%) 1 (5.88%) 1 (5.88%) 5 (29.41%) 2 (11.76%) Length of time in role Below 5 years 6-10 years 10+ years Prefer not to say/ did not answer 7 (41.18%) 6 (35.29%) 3 (17.65%) 1 (5.88%) Health conditions* ________________________ Dyslexia Dyspraxia ADHD Autism Colour Vision Deficiency 2 (66.67%) 1 (33.33%) 1 (33.33%) 1 (33.33%) 1 (33.33%) * Some participants (N=3) reported multiple health conditions.
Qualitative Evaluation: Three Key Main Themes
1
1: Credibility
Sub-themes: Data integrity; Lack of clarity
Summary: Users praised multi-source data integration but raised concerns about missing NHS data and unclear terminology affecting confidence for non-specialists.
2
2: Platform Experience & Functionality
Sub-themes: Usability & accessibility; Confidence using the platform
Summary: Positive feedback on quick data access and filtering, but issues included red/white colour scheme impacting colour vision deficiency users, screen reader incompatibility, and information density.
3
3: Key Strengths
Sub-themes: Analytical capability
Summary: Strong analytical capability for geospatial analysis, trend identification, and evidence generation for service planning and funding bids.
Overview & Reflections
Platform Development
MONARCH addresses a longstanding gap in the evidence base. The SHA proved more accurate than previous approaches, and harmonisation of 280 needs into 51 categories ensured the dataset is internally consistent and policy relevant.
Secondary Data Analyses
Machine learning and predictive modelling identified clear risk profiles — specifically younger, non-officer veterans living alone — providing a robust foundation for proactive, targeted intervention strategies.
Platform Evaluation
The mean SUS score of 74.12 denotes acceptable usability. Qualitative feedback affirmed the platform's analytical capability and geospatial functionality as strong assets for policymakers and commissioners.
Limitations & Future Development
The dataset captures only veterans who self-identified and sought charity support, potentially underrepresenting the most marginalised. Absence of NHS data limits comprehensiveness, though its inclusion was never the primary aim.
1
Data Coverage & Representativeness
The dataset only captures veterans who self-identified and sought charity support, potentially underrepresenting the most marginalised or those who never engaged with services.
2
Geographic Coverage
Coverage is uneven across UK regions, reflecting the distribution of participating charities rather than the true geographic spread of veteran need.
3
Platform Accessibility
Access is currently limited to approved stakeholders. Broader rollout will require investment in training, data literacy support, and secure access infrastructure.
4
Interpretability & Data Literacy
Users require sufficient data literacy to interpret outputs accurately. Misinterpretation of geospatial or demographic data could lead to poorly targeted commissioning decisions.
5
Scalability & Sustainability
Long-term sustainability depends on securing ongoing funding, maintaining charity partnerships, and integrating new data sources such as NHS referral data.
Overall Conclusion
The MONARCH platform represents a significant step forward in the evidence infrastructure supporting UK veteran services. The foundations laid — technically, analytically, and through cross-sector collaboration — provide a strong platform on which to develop further.
Broader Significance
MONARCH demonstrates that aggregating sensitive casework data across independent organisations, protecting individual privacy, and producing a scientifically rigorous and practically useful platform is achievable — offering a transferable framework for other complex populations.
Next Steps
Addressing limitations around data coverage, accessibility, and interpretability will be essential to realising the platform's full potential as a tool for preventing veteran crisis and improving service delivery at scale across the entire support ecosystem.
Northern Hub for Veterans' & Military Families' Research
Funded by: The Armed Forces Covenant Fund Trust, and Forces in Mind Trust
Appendix A
Published Paper: Validating a Secure Hashing Algorithm
Tomietto, M., McGill, A., & Kierman, M.D. (2023). Implementing an electronic public health record for policy planning in the UK military sector: Validation of a secure hashing algorithm. Heliyon, 9(6).
Study Aim
Test the sensitivity and specificity of a Secure Hashing Algorithm (SHA) to generate a unique anonymous identifier for data linkage across different organisations in the veterans' population.
Datasets Used
- Army Benevolent Fund (ABF) 2021: 2,622 records
- Army Benevolent Fund (ABF) 2022: 2,513 records
- Royal Naval Benevolent Trust (RNBT): 26,684 records
Key Results
- SHA identified more unique cases than the military service number gold standard across all three datasets
- 100% sensitivity across all datasets
- 94.7–99.3% specificity depending on dataset
- Area under ROC curve: 98.9–99.8%
- 1,771 records successfully linked across ABF 2021 and 2022 datasets
Conclusion
The SHA improved uniqueness of records, ensured data anonymity, and enabled enhanced data linkage — providing the foundation for big data management and precision public health strategies for the veteran population.
Appendix B
Published Paper: Nationwide Registry via Machine Learning
Serra, G., Tomietto, M., McGill, A. and Kierman, M. (2025). Developing a nationwide registry of UK veterans seeking help from sector charities — a machine learning approach to stratification. European Journal of Public Health, 35(1), pp.5–10.
Analysing data from five Armed Forces charities, this study examined 42,264 veterans and 113,521 needs. A K-means clustering approach returned four subgroups of use — identical to those created using a priori knowledge — confirming the analytical process.
Single Access, Single Need — 37%
Need likely met; veteran did not return.
Single Access, Multiple Needs — 23%
More complex but likely resolved at single contact.
Multiple Access, Repeated Needs — 33%
Ongoing issues; intervention effectiveness not optimal.
Multiple Access, All New Needs — 7%
Underlying problem not addressed; highest complexity subgroup.
Key points: The MONARCH dataset is the first comprehensive nationwide registry of UK military charity data. Double counting was resolved using the SHA. Younger, non-officer veterans may be more at risk of complex needs.
Appendix C
Published Paper: AI-Powered Predictive Model for Early Intervention
Serra, G., Turoldo, F., Tomietto, M., McGill, A. and Kierman, M.D. (2025). Improving early intervention: identifying risk factors for UK military veterans that access military charities — a case-control study and an AI-powered predictive model. European Journal of Public Health, 35(5), pp.867–872.
Study Design
Case-control study comparing 838 veterans who accessed charities in 2022 (cases) with 838 veterans who reported never accessing charities (controls). Logistic regression and random forest algorithms were used to identify risk factors.
Random Forest Findings
Living condition was the most important predictor (mean decrease in accuracy: 30.22). Model accuracy on training set: 81.8%; test set: 65.8%. All key variables confirmed as significant by the Boruta algorithm.
Multivariate Risk Factors
- Working age (<66 years): OR 2.99
- Non-officer rank: OR 17.53
- Living alone: OR 11.95
- Having dependants (living alone): OR 9.20 (risk factor)
- Having dependants (living with others): OR 0.09 (protective)
Conclusion
Predictive models could enable more efficient resource allocation and targeted preventive strategies, allowing proactive support for veterans before they reach crisis — a practical implementation of Precision Public Health.