Olubanke Dzahini

Identification of Adverse Drug Events from Free Text Electronic Patient Records and Information in a Large Mental Health Case Register.

Objectives Electronic healthcare records (EHRs) are a rich source of information, with huge potential for secondary research use. The aim of this study was to develop an application to identify instances of Adverse Drug Events (ADEs) from free text psychiatric EHRs. Methods We used the GATE Natural Language Processing (NLP) software to mine instances of ADEs from free text content within the Clinical Record Interactive Search (CRIS) system, a de-identified psychiatric case register developed at the South London and Maudsley NHS Foundation Trust, UK. The tool was built around a set of four movement disorders (extrapyramidal side effects [EPSEs]) related to antipsychotic therapy and rules were then generalised such that the tool could be applied to additional ADEs. We report the frequencies of recorded EPSEs in patients diagnosed with a Severe Mental Illness (SMI) and then report performance in identifying eight other unrelated ADEs. Results The tool identified EPSEs with >0.85 precision and >0.86 recall during testing. Akathisia was found to be the most prevalent EPSE overall and occurred in the Asian ethnic group with a frequency of 8.13%. The tool performed well when applied to most of the non-EPSEs but least well when applied to rare conditions such as myocarditis, a condition that appears frequently in the text as a side effect warning to patients. Conclusions The developed tool allows us to accurately identify instances of a potential ADE from psychiatric EHRs. As such, we were able to study the prevalence of ADEs within subgroups of patients stratified by SMI diagnosis, gender, age and ethnicity. In addition we demonstrated the generalisability of the application to other ADE types by producing a high precision rate on a non-EPSE related set of ADE containing documents. Availability The application can be found at http://git.brc.iop.kcl.ac.uk/rmallah/dystoniaml.

ADEPt, a semantically-enriched pipeline for extracting adverse drug events from free-text electronic health records

Adverse drug events (ADEs) are unintended responses to medical treatment. They can greatly affect a patient’s quality of life and present a substantial burden on healthcare. Although Electronic health records (EHRs) document a wealth of information relating to ADEs, they are frequently stored in the unstructured or semi-structured free-text narrative requiring Natural Language Processing (NLP) techniques to mine the relevant information. Here we present a rule-based ADE detection and classification pipeline built and tested on a large Psychiatric corpus comprising 264k patients using the de-identified EHRs of four UK-based psychiatric hospitals. The pipeline uses characteristics specific to Psychiatric EHRs to guide the annotation process, and distinguishes: a) the temporal value associated with the ADE mention (whether it is historical or present), b) the categorical value of the ADE (whether it is assertive, hypothetical, retrospective or a general discussion) and c) the implicit contextual value where the status of the ADE is deduced from surrounding indicators, rather than explicitly stated. We manually created the rulebase in collaboration with clinicians and pharmacists by studying ADE mentions in various types of clinical notes. We evaluated the open-source Adverse Drug Event annotation Pipeline (ADEPt) using 19 ADEs specific to antipsychotics and antidepressants medication. The ADEs chosen vary in severity, regularity and persistence. The average F-measure and accuracy achieved by our tool across all tested ADEs were 0.83 and 0.83 respectively. In addition to annotation power, the ADEPT pipeline presents an improvement to the state of the art context-discerning algorithm, ConText.

Antipsychotic use and unexpected death: a hospital-based case-control study.

To examine the risk of unexpected death in patients prescribed an antipsychotic. Unexpected death was defined as death occurring within 7 days of the onset of acute symptoms.

Variation in dose and plasma level of lamotrigine in patients discharged from a mental health trust

Background: The objectives of this study were to investigate the dose of lamotrigine when prescribed with an enzyme inhibitor or enzyme inducer in patients discharged from a mental health trust and to determine the corresponding lamotrigine plasma concentrations and the factors that may affect these. Methods: All patients discharged on lamotrigine between October 2007 and September 2012 were identified using the pharmacy dispensing database. We recorded demographic details, lamotrigine dose and plasma levels and coprescribed medication. Results: During the designated period, 187 patients were discharged on lamotrigine of whom 117 had their plasma levels recorded. The mean lamotrigine daily dose was 226.1 mg (range 12.5–800 mg) and the mean plasma level 5.9 mg/l (range 0.8–18.1 mg/l). Gender, ethnicity, diagnosis and smoking status had no significant effect on dose or plasma levels. Patients taking an enzyme-inducing drug (n = 6) had significantly lower plasma levels [mean (SD) 3.40 (1.54) mg/l] than those not taking enzyme inducers [n = 111; 6.03 (3.13) mg/l; p = 0.043]. Patients taking an enzyme-inhibiting drug (n = 23) had significantly higher levels [7.47 (3.99) mg/l] than those not taking an inhibitor [n = 94; 5.52 (2.75) mg/l; p = 0.035]. No significant difference was found between the doses of lamotrigine in patients taking an enzyme inhibitor and those not taking one (p = 0.376). No significant difference was found between the doses of lamotrigine in patients taking an enzyme-inducing drug and those not taking any (p = 0.574). Conclusions: Current dosing recommendations indicate that lamotrigine doses should be halved in individuals taking enzyme inhibitors and doubled in those on enzyme inducers. In our survey these recommendations were rarely followed with the consequence that patients received too high or too low a dose of lamotrigine, respectively.

Aripiprazole long-acting injection – a mirror image study of its effects on hospitalisation at one year:

Second generation antipsychotic long-acting injections have a greater cost than older depots. Their cost-effectiveness has yet to be established. We conducted a non-interventional, observational, follow-up of patients prescribed aripiprazole long-acting injection in two centres using a mirror image method. Data were available for 160 patients consecutively prescribed aripiprazole long-acting injection, of whom 30 were not included in the analysis (21 forensic patients, five incomplete data and four lost to follow-up). Of the 130 patients, 66 (51%) remained on aripiprazole long-acting injection at one year. The mean number of bed days in the year following aripiprazole long-acting injection initiation reduced to 22.82/patient (standard deviation [SD]=55.07) from 30.09/patient/year (SD=30.40) over the three years before initiation (p<0.001). The mean number of admissions fell from 0.71/patient/year (SD=0.55) to 0.45/patient/year (SD=0.93) over the same period (p<0.001). The median number of bed days in the three years before aripiprazole long-acting injection was 21.67/year; in the year following it was zero. Outcomes were not statistically better in those who remained on aripiprazole long-acting injection at one year compared with those who discontinued. The prescribing of aripiprazole long-acting injection reduces average bed days and admissions compared with prior treatments. The reduction in bed days is of a magnitude that renders aripiprazole long-acting injection broadly cost-neutral.

Haloperidol decanoate long-acting injection (HDLAI): Results of a 1-year mirror-image study

Background: We sought to determine clinical outcomes of the prescribing of haloperidol decanoate long-acting injection (HDLAI) at 1 year. Method: A 1-year mirror-image study of 84 inpatients initiated on HDLAI. Admissions and bed days in the year preceding HDLAI were compared with the year after initiation. Predictors for discontinuation were evaluated. Results: At 1 year, 33% of patients remained on treatment. Patients starting HDLAI because of nonadherence were more likely to stop treatment [relative risk (RR) 1.72; 95% confidence interval (CI) 1.01, 2.91; p = 0.044] whilst patients with a longer duration of illness were more likely to remain on treatment (RR 0.88; 95% CI 0.78, 1.00; p = 0.050). In the bed days cohort overall, (n = 65), there was a significant reduction in mean hospital admissions (1.4/patient/year to 0.6/patient/year; p = 0.0001) but not bed days (55.6/patient to 45.0/patient; p = 0.07) in the year following HDLAI initiation compared with the year before. Continuers had a significant reduction in mean bed days (53.1 to 4.0; p = 0.0002) and hospital admissions (1.5 to 0.2; p = 0.0001). Discontinuers demonstrated a significant reduction in hospital admissions (1.5 to 0.8; p = 0.0001) but not bed days (56.7 to 64.5; p = 0.83). Conclusion: HDLAI was associated with a high treatment discontinuation rate. Hospital admissions fell in the year after HDLAI but there was no change in bed days. Our study suggests that patients with a longer duration of illness and patients initiated on HDLAI for reasons other than poor adherence may benefit from HDLAI initiation.

Author Correction: Knowledge graph prediction of unknown adverse drug reactions and validation in electronic health records

A correction to this article has been published and is linked from the HTML and PDF versions of this paper. The error has been fixed in the paper.

Associations between illness beliefs, medication beliefs, anticoagulation-related quality of life, and INR control: Insights from the Switching Study

Anticoagulation control with vitamin‐K antagonists (VKAs) in patients with atrial fibrillation (AF) or venous thromboembolism (VTE) can be measured using time in therapeutic range (TTR), where TTR >65% is considered good and low TTR may be associated with low adherence.

Knowledge graph prediction of unknown adverse drug reactions and validation in electronic health records

Unknown adverse reactions to drugs available on the market present a significant health risk and limit accurate judgement of the cost/benefit trade-off for medications. Machine learning has the potential to predict unknown adverse reactions from current knowledge. We constructed a knowledge graph containing four types of node: drugs, protein targets, indications and adverse reactions. Using this graph, we developed a machine learning algorithm based on a simple enrichment test and first demonstrated this method performs extremely well at classifying known causes of adverse reactions (AUC 0.92). A cross validation scheme in which 10% of drug-adverse reaction edges were systematically deleted per fold showed that the method correctly predicts 68% of the deleted edges on average. Next, a subset of adverse reactions that could be reliably detected in anonymised electronic health records from South London and Maudsley NHS Foundation Trust were used to validate predictions from the model that are not currently known in public databases. High-confidence predictions were validated in electronic records significantly more frequently than random models, and outperformed standard methods (logistic regression, decision trees and support vector machines). This approach has the potential to improve patient safety by predicting adverse reactions that were not observed during randomised trials.