The Challenge
SLaM serves 1.3 million local residents, and its records cover more than 400,000 patients. Turning that information into day-to-day decisions was hard:
- Information overload. Records grow faster than anyone can read them, and standard EHR screens show tables and static text that hide trends across a patient’s history or a team’s caseload.
- The story lives in free text. Mental health assessments rely heavily on what patients report, so symptoms, medications and social context are mostly written in notes and letters rather than structured fields.
- Physical health slips through. People with severe mental illness live 10–20 years less on average, partly because of medication side effects, yet tracking annual physical health checks for tens of thousands of patients by hand is close to impossible.
- Manual reporting eats time. Reviewing a caseload for priorities, workload or CQC compliance took at least 20 minutes each time.
The Solution
VIEWER sits on top of SLaM’s existing data infrastructure, including CogStack, entirely within the Trust’s firewall. Every day it takes in more than 1.7 million documents and turns them into dashboards clinicians can act on.
- Structured and free-text data in one view. NLP picks out symptoms, medications and interventions from clinical notes and combines them with lab results, prescriptions and demographics.
- Four views for four jobs. Population health, clinical pathways, caseload management and individual patient charts, each linking straight back to the EHR so staff can act.
- Evidence on screen. Every NLP finding comes with the text snippet it came from, so clinicians can check the context for themselves.
- Governance built in. Clinical users see identifiable data, while managers and analysts see a de-identified version, and every action is logged.
Results & Impact
In a pharmacist’s check of three completed medication reviews, VIEWER found every antipsychotic, antidepressant and mood stabiliser the clinician had identified, plus four more, three of which were treatments considered but not started. Completed physical health checks rose within two months of a six-person team adopting VIEWER, and caseload reviews that took 20 minutes now take a single filter. VIEWER also helped the Trust meet CQC requirements for its annual inspection, and SLaM and the Maudsley Charity have committed four years of funding to run it as a business-as-usual service.
Why CogStack
VIEWER shows what becomes possible once clinical text is searchable: the same CogStack foundation that supports research at SLaM now drives everyday care. It is open source, runs inside the Trust’s own firewall and was designed with clinicians from the start, so it fits how teams actually work.
Based on Wang T et al., “VIEWER: an extensible visual analytics framework for enhancing mental healthcare”, arXiv:2411.07247 (2024). Results come from an observational evaluation at a single NHS trust and may vary across organisations.
