Understanding social and clinical associations with unemployment for people with schizophrenia and bipolar disorders: large-scale health records study.

PURPOSE: People with severe mental illness (SMI) experience high levels of unemployment. We aimed to better understand the associations between clinical, social, and demographic inequality indicators and unemployment. METHODS: Data were extracted from de-identified health records of people with SMI in contact with secondary mental health services in south London, UK. A Natural Language Processing text-mining application was applied to extract information on unemployment in the health records. Multivariable logistic regression was used to assess associations with unemployment, in people with SMI. RESULTS: Records from 19,768 service users were used for analysis, 84.9% (n = 16,778) had experienced unemployment. In fully adjusted models, Black Caribbean and Black African service users were more likely to experience unemployment compared with White British service users (Black Caribbean: aOR 1.62, 95% CI 1.45-1.80; Black African: 1.32, 1.15-1.51). Although men were more likely to have experienced unemployment relative to women in unadjusted models (OR 1.36, 95% CI 1.26-1.47), differences were no longer apparent in the fully adjusted models (aOR 1.05, 95% CI 0.97-1.15). The presence of a non-affective (compared to affective) diagnosis (1.24, 1.13-1.35), comorbid substance use (2.02, 1.76-2.33), previous inpatient admissions (4.18, 3.71-4.70), longer inpatient stays (78 + days: 7.78, 6.34-9.54), and compulsory admissions (3.45, 3.04-3.92) were associated with unemployment, in fully adjusted models. CONCLUSION: People with SMI experience high levels of unemployment, and we found that unemployment was associated with several clinical and social factors. Interventions to address low employment may need to also address these broader inequalities.

Improving our understanding of the social determinants of mental health: a data linkage study of mental health records and the 2011 UK census.

OBJECTIVES: To address the lack of individual-level socioeconomic information in electronic healthcare records, we linked the 2011 census of England and Wales to patient records from a large mental healthcare provider. This paper describes the linkage process and methods for mitigating bias due to non-matching. SETTING: South London and Maudsley NHS Foundation Trust (SLaM), a mental healthcare provider in Southeast London. DESIGN: Clinical records from SLaM were supplied to the Office of National Statistics for linkage to the census through a deterministic matching algorithm. We examined clinical (International Classification of Disease-10 diagnosis, history of hospitalisation, frequency of service contact) and socio-demographic (age, gender, ethnicity, deprivation) information recorded in Clinical Record Interactive Search (CRIS) as predictors of linkage success with the 2011 census. To assess and adjust for potential biases caused by non-matching, we evaluated inverse probability weighting for mortality associations. PARTICIPANTS: Individuals of all ages in contact with SLaM up until December 2019 (N=459 374). OUTCOME MEASURES: Likelihood of mental health records' linkage to census. RESULTS: 220 864 (50.4%) records from CRIS linked to the 2011 census. Young adults (prevalence ratio (PR) 0.80, 95% CI 0.80 to 0.81), individuals living in more deprived areas (PR 0.78, 95% CI 0.78 to 0.79) and minority ethnic groups (eg, Black African, PR 0.67, 0.66 to 0.68) were less likely to match to census. After implementing inverse probability weighting, we observed little change in the strength of association between clinical/demographic characteristics and mortality (eg, presence of any psychiatric disorder: unweighted PR 2.66, 95% CI 2.52 to 2.80; weighted PR 2.70, 95% CI 2.56 to 2.84). CONCLUSIONS: Lower response rates to the 2011 census among people with psychiatric disorders may have contributed to lower match rates, a potential concern as the census informs service planning and allocation of resources. Due to its size and unique characteristics, the linked data set will enable novel investigations into the relationship between socioeconomic factors and psychiatric disorders.

Text mining occupations from the mental health electronic health record: a natural language processing approach using records from the Clinical Record Interactive Search (CRIS) platform in south London, UK.

OBJECTIVES: We set out to develop, evaluate and implement a novel application using natural language processing to text mine occupations from the free-text of psychiatric clinical notes. DESIGN: Development and validation of a natural language processing application using General Architecture for Text Engineering software to extract occupations from de-identified clinical records. SETTING AND PARTICIPANTS: Electronic health records from a large secondary mental healthcare provider in south London, accessed through the Clinical Record Interactive Search platform. The text mining application was run over the free-text fields in the electronic health records of 341 720 patients (all aged ≥16 years). OUTCOMES: Precision and recall estimates of the application performance; occupation retrieval using the application compared with structured fields; most common patient occupations; and analysis of key sociodemographic and clinical indicators for occupation recording. RESULTS: Using the structured fields alone, only 14% of patients had occupation recorded. By implementing the text mining application in addition to the structured fields, occupations were identified in 57% of patients. The application performed on gold-standard human-annotated clinical text at a precision level of 0.79 and recall level of 0.77. The most common patient occupations recorded were 'student' and 'unemployed'. Patients with more service contact were more likely to have an occupation recorded, as were patients of a male gender, older age and those living in areas of lower deprivation. CONCLUSION: This is the first time a natural language processing application has been used to successfully derive patient-level occupations from the free-text of electronic mental health records, performing with good levels of precision and recall, and applied at scale. This may be used to inform clinical studies relating to the broader social determinants of health using electronic health records.