Research into suicide prevention has been hampered by methodological limitations such as low sample size and recall bias. Recently, Natural Language Processing (NLP) strategies have been used with Electronic Health Records to increase information extraction from free text notes as well as structured fields concerning suicidality and this allows access to much larger cohorts than previously possible. This paper presents two novel NLP approaches – a rule-based approach to classify the presence of suicide ideation and a hybrid machine learning and rule-based approach to identify suicide attempts in a psychiatric clinical database. Good performance of the two classifiers in the evaluation study suggest they can be used to accurately detect mentions of suicide ideation and attempt within free-text documents in this psychiatric database. The novelty of the two approaches lies in the malleability of each classifier if a need to refine performance, or meet alternate classification requirements arises. The algorithms can also be adapted to fit infrastructures of other clinical datasets given sufficient clinical recording practice knowledge, without dependency on medical codes or additional data extraction of known risk factors to predict suicidal behaviour.

ObjectivesTo investigate recorded poor insight in relation to mental health and service use outcomes in a cohort with first-episode psychosis.DesignWe developed a natural language processing algorithm to ascertain statements of poor or diminished insight and tested this in a cohort of patients with first-episode psychosis.SettingThe clinical record text at the South London and Maudsley National Health Service Trust in the UK was used.ParticipantsWe applied the algorithm to characterise a cohort of 2026 patients with first-episode psychosis attending an early intervention service.Primary and secondary outcome measuresRecorded poor insight within 1 month of registration was investigated in relation to (1) incidence of psychiatric hospitalisation, (2) odds of legally enforced hospitalisation, (3) number of days spent as a mental health inpatient and (4) number of different antipsychotic agents prescribed; outcomes were measured over varying follow-up periods from 12 months to 60 months, adjusting for a range of sociodemographic and clinical covariates.ResultsRecorded poor insight, present in 46% of the sample, was positively associated with ages 16-35, bipolar disorder and history of cannabis use and negatively associated with White ethnicity and depression. It was significantly associated with higher levels of all four outcomes over all five follow-up periods.ConclusionsRecorded poor insight in people with recent onset psychosis predicted subsequent legally enforced hospitalisations and higher number of hospital admissions, number of unique antipsychotics prescribed and days spent hospitalised. Improving insight might benefit patients’ course of illness as well as reduce mental health service use.