Diagnosis in psychosis – does it matter?

When treating psychosis, determining the exact diagnosis can take time, especially during early stages of illness. This is why mental health services have adopted a ‘clinical staging model’ for diagnosis. However, in practice, many patients are left with an unspecified diagnosis of psychosis for long periods, often more than a year. 

This research project investigates whether diagnostic delays prevent patients from receiving targeted treatments – such as lithium for bipolar disorder, clozapine for schizophrenia, or diagnosis-specific psychological interventions – ultimately leading to worse health outcomes and greater use of services.

Aims of this research:

1. Map diagnostic patterns from first contact with South London and Maudsley (SLaM) NHS Foundation Trust services, tracking how often ‘unspecified psychosis’ evolves into specific diagnoses such as schizophrenia
2. Use natural language processing (NLP) to identify symptom clusters associated with specific psychosis diagnoses from clinical records
3. Test whether diagnosis matters by examining if patients whose NLP symptom profiles match their clinical diagnosis receive better-matched treatments and achieve better outcomes
4. Examine treatment patterns to see whether receiving evidence-based care depends more on having a formal diagnosis or on symptom presentation alone
5. Focus on manic symptoms to determine if people experiencing these are more likely to receive potentially harmful high-dose antipsychotics and benzodiazepines, especially when mood stabilizers are not prescribed
6. Explore depression by analysing relationships between NLP-identified depression symptoms, clinical depression diagnoses, treatment patterns and outcomes including suicide
7. Address inequalities, particularly examining whether Black service users – who are known to receive clozapine, psychological therapies, and bipolar diagnoses less frequently – are disproportionately affected

How the research will be carried out

Researchers will use the Clinical Record Interactive Search (CRIS) system to analyse pseudonymised patient data from SLaM services looking specifically at records on clinical diagnosing, service use and outcomes. Teams of clinicians and data analysts will conduct retrospective cohort studies within SLaM's secure data environment, comparing outcomes across different patient groups based on their diagnostic patterns.

Involving patients and the public 

We will engage patient and public involvement members through groups in the ARC, SLaM, and the psychosis clinical academic group, seeking input on the project and feedback on findings throughout the project.

Our collaborators 

We are working closely with psychiatrists, data analysts, service users, and service providers in South London and Maudsley NHS Foundation Trust (SLaM) and King’s Health Partners.

Potential benefits

Clinical reluctance to diagnose conditions like schizophrenia in the early years of psychosis may inadvertently delay crucial treatments, and potentially expose patients to higher doses of less-specific medications with greater side effects. If this research confirms that specific diagnoses improve outcomes, findings will inform clinical recommendations within SLaM and more broadly. Addressing these potential problems could lead to improved patient care and outcomes, while reducing costs and burden on the healthcare system.

This project is expected to complete in November 2026.