Projects per year
Methods: Data were obtained from pseudonymised EHRs of 1,835 people with FEP in South London. Five psychosis symptom clusters (positive, negative, disorganisation, mania and depression) were identified using CRIS‐CODE NLP software. Data on subsequent hospital admission and the number of unique antipsychotics prescribed in the 5 years following presentation were obtained. Their relationship with psychosis symptom clusters was analysed using multivariable negative binomial regression with age, gender and ethnicity as covariates.
Results: Mania (IRR 6.37, 95% CI 3.24 to 12.5) and positive symptoms (3.87, 1.98 to 7.58) were more strongly associated with increased hospital admission than disorganisation (1.68, 0.46 to 6.15), depression (1.55, 0.71 to 3.39) or negative symptoms (1.51, 0.51 to 4.48). Negative symptoms were more strongly associated with increased antipsychotic treatment failure (4.49, 2.35 to 8.59) than mania (3.65, 2.42 to 5.52), positive symptoms (3.35, 2.22 to 5.06), disorganisation (2.64, 1.23 to 5.68) or depression (2.18, 1.36 to 3.49).
Conclusion: Increased mania and positive symptom burden predict greater hospitalisation than negative symptoms. However, negative symptoms are more strongly associated with increased antipsychotic treatment failure. These findings illustrate that although mania and positive symptoms result in increased hospitalisation, negative symptoms are less responsive to antipsychotic treatment.
|Number of pages||1|
|Journal||Early Intervention in Psychiatry|
|Publication status||Published - 17 Sept 2018|
FingerprintDive into the research topics of 'Negative Symptoms in First Episode Psychosis (FEP) Predict Antipsychotic Treatment Failure: A Natural Language Processing (NLP) Electronic Health Record Study'. Together they form a unique fingerprint.
- 2 Finished
Linking electronic health records with passive smartphone activity data to predict outcomes in psychotic disorders
14/02/2018 → 13/02/2021
Symptom dimensions in first episode psychosis: predicting clinical outcomes using natural language processing
3/10/2016 → 2/10/2018