Browsing by Author "Josephson, Colin Bruce"

Now showing 1 - 5 of 5
  • Thumbnail Image
    ItemOpen Access
    A systematic approach to using regression modelling and ‘big data’ to derive a meaningful clinical decision rule for epilepsy
    (2018-08-22) Josephson, Colin Bruce; Wiebe, Samuel; Jetté, Nathalie; Sajobi, Tolulope T.; Marshall, Deborah A.
    Introduction: clinical decision rules (CDRs) have been developed in a number of medical fields resulting in improved patient outcomes, quality of care, and health economics. Aims: to identify all CDRs developed for epilepsy and to derive one that guides the prescription of the antiepileptic drug (AED), levetiracetam, according to its risk of a psychiatric adverse effect. Methods: a systematic review and meta-analysis was first performed to determine the state of the literature with respect to CDRs in epilepsy. The Health Improvement Network (THIN) electronic medical records register was used to identify patients with epilepsy by employing a modified validated case definition with a 5-year washout. Analyses were restricted to patients receiving AED monotherapy and the association between levetiracetam use and psychiatric adverse effects was explored Cox proportional hazards regression with timevarying covariates. Finally, logistic regression with parameter regularisation and k=5 fold cross validation was used to derive the CDR that predicts the development of psychiatric adverse effects following levetiracetam prescription. Results: the systematic review identified four epilepsy-specific CDRs, none of which guided AED prescription. A total of 9595 presumed incident cases of epilepsy (85.7 cases per 100,000 persons) were identified in THIN. Both carbamazepine (hazard ratio [HR]: 0.84, 95% confidence interval [95% CI]: 0.73– 0.97; p = 0.02) and lamotrigine (HR: 0.83, 95% CI: 0.70–0.99; p = 0.03) were associated with reduced hazards of a psychiatric sign, symptom, or disorder iii compared to no AED treatment. Levetiracetam was not associated with psychiatric adverse effects but the analyses were underpowered (n=202; 3%). All patients receiving levetiracetam (1173/7400; 16%) were included for CDR derivation. Prediction variables were incorporated into multiple logistic regression models with parameter regularisation. Odds of reporting a psychiatric complaint were elevated for females and those with a pre-exposure history of depression, anxiety, recreational drug use, or higher social deprivation. The prediction model performed well (area under the curve [AUC] 0.68; 95% confidence interval 0.58- 0.79 after stratified k=5 fold cross-validation). Using a cut-off threshold 0.1, the CDR had a specificity of 83%. Conclusion: If externally validated and properly implemented, this CDR could be used to guide prescription in clinical practice.
  • Thumbnail Image
    ItemEmbargo
    Developing Novel Supervised Learning Model Evaluation Metrics Based on Case Difficulty
    (2024-01-05) Kwon, Hyunjin; Lee, Joon; Josephson, Colin Bruce; Greenberg, Matthew
    Performance evaluation is an essential step in the development of machine learning models. The performance of a model informs the direction of its development and provides diverse knowledge to researchers. Most common ways to assess a model’s performance are based on counting the numbers of correct and incorrect predictions the model makes. However, this conventional approach to evaluation is limited in that it does not consider the differences in prediction difficulty between individual cases. Although metrics for quantifying the prediction difficulty of individual cases exist, their usefulness is hindered by the fact that they cannot be applied universally across all types of data; that is, each metric requires specific data conditions be met for its use, which can be a significant obstacle when dealing with real-world data characterized by diversity and complexity. Therefore, this thesis proposes new metrics for calculating case difficulty that perform well across diverse datasets. These new case difficulty metrics were developed using neural networks based on varying definitions of prediction difficulty. In addition, the metrics were validated using various datasets and compared with existing metrics from the literature. New performance evaluation metrics incorporating case difficulty to reward correct predictions of high-difficulty cases and penalize incorrect predictions of low-difficulty cases were also developed. A comparison of these case difficulty-based performance metrics with conventional performance metrics revealed that the novel evaluation metrics could provide a more detailed explanation and deeper understanding of model performance. We anticipate that researchers will be able to calculate case difficulty in diverse datasets under various data conditions with our proposed metrics and use these values to enhance their studies. Moreover, our newly developed evaluation metrics considering case difficulty could serve as an additional source of insight for the evaluation of classification model performance.
  • Thumbnail Image
    ItemOpen Access
    Predicting the Side Effects of Antiseizure Medications Using Machine Learning Models
    (2024-01-02) Lin, Chantelle Qing Yang; Josephson, Colin Bruce; Sajobi, Tolulope; Klein, Karl Martin; Forkert, Nils Daniel
    With over 20 anti-seizure medications (ASMs), identifying the ideal drug is often imprecise and time-consuming. Developing predictive models to expedite optimal drug selection is challenging due to the minimal differences in efficacy among adult patients with epilepsy. However, side-effects vary considerably between medications, and are one of the main reasons for discontinuation of ASM treatment. The aim was to (1) assess the prognostic utility of high- dimensional data such as genetic features with clinical features to predict ASM discontinuation, and (2) determine the optimal regression/machine learning model for predicting ASM discontinuation. This retrospective cohort study included 4,853 exposures to any ASM, and 624 patients exposed to valproic acid (VPA) from the RAISE-GENIC study during the years 2006-2020. The predicted outcome was defined as ASM discontinuation due to any side-effect reported by the patient. Clinical features included age of onset, patient age, sex, comorbidities, seizure type, EEG variables, and imaging variables. Network analysis of mRNA expression data from VPA-exposed neurons derived from control induced pluripotent stem cells (iPSCs) was leveraged to extract exome sequencing and genome-wide single nucleotide polymorphism data. Features were selected for model inclusion based on relevance as determined by the ReliefF algorithm. Penalized logistic regression, support vector machine, random forest, and k-nearest neighbor models were trained on the normalized bootstrapped dataset and model quality was assessed using stratified 10-fold cross validation. Models with only clinical and combined clinical and genetic features were compared by quantitative as well as visual discrimination and calibration metrics. The results showed that the best performing model was the penalized logistic regression using the VPA dataset with genetic and clinical features. The accuracy was 0.75 [95% confidence interval 0.74-0.76], area under the receiver operating characteristic curve was 0.66 [0.66-0.67], Brier score was 0.20 [0.19-0.21], sensitivity was 0.42 [0.41-0.42], and specificity 0.82 [0.82-0.83]. Machine learning using clinical and genetic features can moderately predict treatment-ending side- effects to VPA with moderate performance, discrimination, and calibration. If these results can be validated and improved upon, decision tools can be incorporated into clinical routines, simplifying drug prescriptions, saving time, and improving patient quality of life.
  • Thumbnail Image
    ItemEmbargo
    Risk Factors for Dementia Development, Frailty, and Mortality in Older Adults with Epilepsy: A Population-Based Analysis
    (2020-08-28) Subota, Ann Ana; Holroyd-Leduc, Jayna M.; Jetté, Nathalie J.; Josephson, Colin Bruce; McMillan, Jacqueline M.
    As the global population ages, more individuals will develop and live with epilepsy and dementia. Previous literature suggests older adults with epilepsy are more likely to develop dementia, yet little is understood about the impact of dementia and frailty in older adults with epilepsy. This thesis aimed to address these knowledge gaps by examining the risk factors for incident dementia development and the role of dementia and frailty on mortality in older adults with incident epilepsy. A cohort of 1048 older adults with incident epilepsy aged 65 years or older were identified using a large population-based primary care dataset from the United Kingdom. The odds of having dementia were 7.39 [95% CI 5.21-10.50] times higher in older adults with incident epilepsy compared to older adults without epilepsy (p-value < 0.001). In the final multivariable logistic model, only the Charlson comorbidity index score at baseline incident epilepsy diagnosis was significantly associated with an increased odds of incident dementia [OR: 1.34, 95% CI 1.14-1.56, p-value < 0.001]. In a multivariate Cox model, age [HR: 1.06, 95% CI 1.02-1.11, p-value < 0.002], baseline dementia [HR: 2.68, 95% CI 1.66-4.33, p-value < 0.001] and baseline e-frailty index score [HR: 7.64, 95% CI 1.21-48.19, p-value = 0.030] were significantly associated with a higher hazard of death among older adults with incident epilepsy. The presence of dementia and the degree of frailty experienced both significantly increase the hazard of death in older adults with incident epilepsy. e-Frailty index scores could be utilized more widely in epilepsy clinics in all older adults with epilepsy to identify individuals who may be at greater risk of dying; additional supports and interventions could be provided to reduce mortality in older adults with epilepsy.
  • Thumbnail Image
    ItemOpen Access
    Using Machine Learning Towards Decision Support for Refractory Epilepsy Cases
    (2023-01-25) Farhoudi, Bijan; Maurer, Frank; Wiebe, Samuel; Federico, Paolo; Josephson, Colin Bruce
    Between 0.5% to 1.0% of people in North America suffer from epilepsy, and around 30% of patients are drug-resistant. Some drug-resistant patients are candidates for surgery and up to 60% to 70% of patients who undergo surgery become seizure-free. Finding a magnetic resonance imaging (MRI) abnormality on pre-operative imaging increases the chance of surgical success. However, up to 30% to 40% of pre-operative MRIs have no clear lesion in people with drug-resistant epilepsy, and only up to 40% to 50% of non-lesional MRI cases become seizure-free after surgery. The focus of this work was to design decision support tools to help clinicians evaluate patients for surgery. As the first step, we investigated the possibility of segregating MRIs with abnormality from MRIs without any abnormality using Deep Learning models. Such models would help clinicians when they examine MRIs to find an abnormality. Considering the value of predicting surgery results, in our next step, we explored the possibility of predicting the outcome of surgery using MRI and Deep Learning. Our results indicate that both lesional and non-lesional MRIs of patients with epilepsy contain signals that Deep Learning models can harness to predict the operative success., Finally, we explored the possibility of finding an abnormality in MRIs that were reported by radiologists as non-lesional by using Deep Learning.