Search

Background Using novel data mining methods such as natural language processing (NLP) on electronic health records (EHRs) for screening and detecting individuals at risk for psychosis. Method The study included all patients receiving a first index diagnosis of nonorganic and nonpsychotic mental disorder within the South London and Maudsley (SLaM) NHS Foundation Trust between January 1, 2008, and July 28, 2018. Least Absolute Shrinkage and Selection Operator (LASSO)-regularized Cox regression was used to refine and externally validate a refined version of a five-item individualized, transdiagnostic, clinically based risk calculator previously developed (Harrell's C = 0.79) and piloted for implementation. The refined version included 14 additional NLP-predictors: tearfulness, poor appetite, weight loss, insomnia, cannabis, cocaine, guilt, irritability, delusions, hopelessness, disturbed sleep, poor insight, agitation, and paranoia. Results A total of 92 151 patients with a first index diagnosis of nonorganic and nonpsychotic mental disorder within the SLaM Trust were included in the derivation (n = 28 297) or external validation (n = 63 854) data sets. Mean age was 33.6 years, 50.7% were women, and 67.0% were of white race/ethnicity. Mean follow-up was 1590 days. The overall 6-year risk of psychosis in secondary mental health care was 3.4 (95% CI, 3.3–3.6). External validation indicated strong performance on unseen data (Harrell's C 0.85, 95% CI 0.84–0.86), an increase of 0.06 from the original model. Conclusions Using NLP on EHRs can considerably enhance the prognostic accuracy of psychosis risk calculators. This can help identify patients at risk of psychosis who require assessment and specialized care, facilitating earlier detection and potentially improving patient outcomes. [ABSTRACT FROM AUTHOR]

Background: Gender differences in the clinical presentation of first episode psychosis (FEP) and its relationship to illicit substance use are inconclusive and underexplored. We conducted an electronic health record (EHR) study applying natural language processing (NLP) techniques to investigate gender differences in psychiatric symptoms when controlling for age, ethnicity, and illicit substance use. Methods: Data were extracted from EHRs of 3,340 people presenting to the South London and Maudsley (SLaM) NHS Trust between April 2007 and March 2017 with FEP. Logistic regression was used to examine gender differences in presentation of 42 psychiatric symptoms, grouped into positive, negative, depressive, mania and disorganisation symptoms. We controlled for age of onset, ethnicity and illicit substance use (cocaine, amphetamine, MDMA or cannabis) and adjusted p-values for multiple comparisons. Results: Patients were predominantly male (62%). Eight symptoms were more prevalent in males (poverty of thought, negative symptoms, social withdrawal, poverty of speech, aggression, grandiosity, paranoia and agitation), and thirteen in females (tearfulness, low energy, reduced appetite, low mood, pressured speech, mood instability, flight of ideas, guilt, mutism, insomnia, poor concentration, tangentiality and elation), before adjustment for age, ethnicity and substance use. Male patients were significantly more likely to experience negative symptoms than females (e.g. poverty of thought, OR 1.85, 95% CI 1.33 to 2.62); female patients showed increased likelihood of depressive and manic symptoms (e.g. tearfulness, 0.30, 0.26 to 0.35). Male patients were significantly more likely to misuse amphetamines, cannabis and cocaine (e.g. cannabis, OR 3.18, 2.75 to 3.70). All significant differences survived controls for age and ethnicity. After adjustment for illicit substance use gender differences in aggression, agitation, paranoia and grandiosity became insignificant (p > 0.05). However, adjustment for illicit substance use resulted in increased strength of gender associations with negative, manic and depression symptoms. Discussion: There are clear gender differences in the clinical presentation of FEP which are modified by exposure to illicit substances. These findings highlight a need to better understand the impact of gender on clinical presentation and treatment outcomes in psychosis, and to ensure that clinicians are aware of how gender differences in presentation could be modified by illicit substance use. [ABSTRACT FROM AUTHOR]

Background: The biological, clinical and social factors which underpin the aetiology of psychotic disorders are known to overlap between different ICD-10/DSM-5 diagnostic categories. A transdiagnostic approach to investigate clinical phenotype may enable a better understanding of pathophysiology at individual patient level. We applied natural language processing (NLP) tools to electronic health record (EHR) data from patients presenting with an ICD-10 diagnosis of unipolar depression to determine if symptoms at diagnosis could predict subsequent onset of a bipolar or psychotic disorder. Methods: Data were obtained from 20,582 adults presenting with unipolar depression (ICD-10 F32 or F33, excluding F32.3 and F33.3) to the South London and Maudsley (SLaM) NHS Foundation Trust between April 2006 and March 2018. Natural language processing (NLP) techniques were used to extract data on 21 mood and affective symptoms from free text clinical assessments documented in the period -3/+3 months from the date of the diagnosis of unipolar depression. We obtained descriptive analyses of demographics and symptom prevalence. Symptoms were categorised into four groups: 1. Depressive (low mood, anhedonia, feelings of guilt, hopelessness, helplessness, psychomotor retardation, worthlessness, tearfulness, low energy), 2. Manic (elation, grandiosity, pressured speech, flight of ideas), 3. Biological symptoms (insomnia, disturbed sleep, low appetite, weight loss, poor concentration) and 4. Emotional/behavioural symptoms (mood instability, agitation, irritability). The symptom network structure was estimated using the Enhanced Least Absolute Shrinkage and Selection Operation procedure. We assessed network stability via a case-dropping bootstrapping procedure. We investigated associations between each of the four symptom groups and clinical outcomes using multivariable Cox regression to predict five-year risk of bipolar disorder (ICD-10 F30/F31) or a psychotic disorder (ICD-10 F2*). Results: Of all patients presenting with unipolar depression, 19,569 (95.1%) had at least one documented depressive symptom, 16,199 (78.7%) had at least one biological symptom, 10,006 (48.6%) had at least one emotional/ behavioural symptom, and 1,372 (6.67%) had at least one manic symptom. Patients with at least one manic symptom were significantly more likely to be male (OR: 1.25 (95% CI 1.12 - 1.40), p < 0.001) and less likely to be of Black (OR: 0.80 (0.68 - 0.93), p = 0.004) or Other ethnicity (OR: 0.78 (0.66 - 0.91), p = 0.003). Elation was the most commonly reported manic symptom (3.17%). Network analysis revealed that the presence of manic symptoms was associated with co-occurrence of agitation, irritability and mood instability. Agitation was the most central symptom in terms of strength, betweenness and expected influence. The resulting network remained stable after dropping up to 33% of cases from the sample. 1,861 (9.04%) patients who initially presented with unipolar depression subsequently developed a mania/bipolar disorder or psychotic disorder within 5 years. The presence of at least one manic (HR: 1.71, 1.50 – 1.97), biological (HR: 1.33, 1.16 – 1.53) or emotional (HR: 1.91, 1.73 – 2.13) symptom was associated with significantly increased risk of onset of a bipolar or psychotic disorder. Discussion: We found that patients with unipolar depression have a heterogenous clinical phenotype with a significant proportion going on to develop a bipolar or psychotic disorder within 5 years. Symptoms extracted from the EHR using NLP were predictive of subsequent onset of a bipolar or psychotic disorder. A transdiagnostic approach to defining clinical phenotype may help to better predict subsequent clinical outcomes. [ABSTRACT FROM AUTHOR]