Jenna Reps

Luis H. John, Jan A. Kors, Jenna M. Reps et al.

Background: Existing clinical prediction models often represent patient data using features that ignore the semantic relationships between clinical concepts. This study integrates domain-specific semantic information by mapping the SNOMED medical term hierarchy into a low-dimensional hyperbolic space using Poincaré embeddings, with the aim of improving lung cancer onset prediction. Methods: Using a retrospective cohort from the Optum EHR dataset, we derived a clinical knowledge graph from the SNOMED taxonomy and generated Poincaré embeddings via Riemannian stochastic gradient descent. These embeddings were then incorporated into two deep learning architectures, a ResNet and a Transformer model. Models were evaluated for discrimination (area under the receiver operating characteristic curve) and calibration (average absolute difference between observed and predicted probabilities) performance. Results: Incorporating pre-trained Poincaré embeddings resulted in modest and consistent improvements in discrimination performance compared to baseline models using randomly initialized Euclidean embeddings. ResNet models, particularly those using a 10-dimensional Poincaré embedding, showed enhanced calibration, whereas Transformer models maintained stable calibration across configurations. Discussion: Embedding clinical knowledge graphs into hyperbolic space and integrating these representations into deep learning models can improve lung cancer onset prediction by preserving the hierarchical structure of clinical terminologies used for prediction. This approach demonstrates a feasible method for combining data-driven feature extraction with established clinical knowledge.

Luis H. John, Chungsoo Kim, Jan A. Kors et al.

Background: Conventional prediction methods such as logistic regression and gradient boosting have been widely utilized for disease onset prediction for their reliability and interpretability. Deep learning methods promise enhanced prediction performance by extracting complex patterns from clinical data, but face challenges like data sparsity and high dimensionality. Methods: This study compares conventional and deep learning approaches to predict lung cancer, dementia, and bipolar disorder using observational data from eleven databases from North America, Europe, and Asia. Models were developed using logistic regression, gradient boosting, ResNet, and Transformer, and validated both internally and externally across the data sources. Discrimination performance was assessed using AUROC, and calibration was evaluated using Eavg. Findings: Across 11 datasets, conventional methods generally outperformed deep learning methods in terms of discrimination performance, particularly during external validation, highlighting their better transportability. Learning curves suggest that deep learning models require substantially larger datasets to reach the same performance levels as conventional methods. Calibration performance was also better for conventional methods, with ResNet showing the poorest calibration. Interpretation: Despite the potential of deep learning models to capture complex patterns in structured observational healthcare data, conventional models remain highly competitive for disease onset prediction, especially in scenarios involving smaller datasets and if lengthy training times need to be avoided. The study underscores the need for future research focused on optimizing deep learning models to handle the sparsity, high dimensionality, and heterogeneity inherent in healthcare datasets, and find new strategies to exploit the full capabilities of deep learning methods.

Uwe Aickelin, Jenna Marie Reps, Peer-Olaf Siebers et al.

In this paper, we present a case study demonstrating how dynamic and uncertain criteria can be incorporated into a multicriteria analysis with the help of discrete event simulation. The simulation guided multicriteria analysis can include both monetary and non-monetary criteria that are static or dynamic, whereas standard multi criteria analysis only deals with static criteria and cost benefit analysis only deals with static monetary criteria. The dynamic and uncertain criteria are incorporated by using simulation to explore how the decision options perform. The results of the simulation are then fed into the multicriteria analysis. By enabling the incorporation of dynamic and uncertain criteria, the dynamic multiple criteria analysis was able to take a unique perspective of the problem. The highest ranked option returned by the dynamic multicriteria analysis differed from the other decision aid techniques.

Luis H. John, Jan A. Kors, Jenna M. Reps et al.

Objective: Provide guidance on sample size considerations for developing predictive models by empirically establishing the adequate sample size, which balances the competing objectives of improving model performance and reducing model complexity as well as computational requirements. Materials and Methods: We empirically assess the effect of sample size on prediction performance and model complexity by generating learning curves for 81 prediction problems (23 outcomes predicted in a depression cohort, 58 outcomes predicted in a hypertension cohort) in three large observational health databases, requiring training of 17,248 prediction models. The adequate sample size was defined as the sample size for which the performance of a model equalled the maximum model performance minus a small threshold value. Results: The adequate sample size achieves a median reduction of the number of observations of 9.5%, 37.3%, 58.5%, and 78.5% for the thresholds of 0.001, 0.005, 0.01, and 0.02, respectively. The median reduction of the number of predictors in the models was 8.6%, 32.2%, 48.2%, and 68.3% for the thresholds of 0.001, 0.005, 0.01, and 0.02, respectively. Discussion: Based on our results a conservative, yet significant, reduction in sample size and model complexity can be estimated for future prediction work. Though, if a researcher is willing to generate a learning curve a much larger reduction of the model complexity may be possible as suggested by a large outcome-dependent variability. Conclusion: Our results suggest that in most cases only a fraction of the available data was sufficient to produce a model close to the performance of one developed on the full data set, but with a substantially reduced model complexity.

Jenna Marie Reps, Jonathan M. Garibaldi, Uwe Aickelin et al.

Big longitudinal observational medical data potentially hold a wealth of information and have been recognised as potential sources for gaining new drug safety knowledge. Unfortunately there are many complexities and underlying issues when analysing longitudinal observational data. Due to these complexities, existing methods for large-scale detection of negative side effects using observational data all tend to have issues distinguishing between association and causality. New methods that can better discriminate causal and non-causal relationships need to be developed to fully utilise the data. In this paper we propose using a set of causality considerations developed by the epidemiologist Bradford Hill as a basis for engineering features that enable the application of supervised learning for the problem of detecting negative side effects. The Bradford Hill considerations look at various perspectives of a drug and outcome relationship to determine whether it shows causal traits. We taught a classifier to find patterns within these perspectives and it learned to discriminate between association and causality. The novelty of this research is the combination of supervised learning and Bradford Hill's causality considerations to automate the Bradford Hill's causality assessment. We evaluated the framework on a drug safety gold standard know as the observational medical outcomes partnership's nonspecified association reference set. The methodology obtained excellent discriminate ability with area under the curves ranging between 0.792-0.940 (existing method optimal: 0.73) and a mean average precision of 0.640 (existing method optimal: 0.141). The proposed features can be calculated efficiently and be readily updated, making the framework suitable for big observational data.

Jenna M. Reps, Uwe Aickelin, Richard B. Hubbard

Purpose: To develop a framework for identifying and incorporating candidate confounding interaction terms into a regularised cox regression analysis to refine adverse drug reaction signals obtained via longitudinal observational data. Methods: We considered six drug families that are commonly associated with myocardial infarction in observational healthcare data, but where the causal relationship ground truth is known (adverse drug reaction or not). We applied emergent pattern mining to find itemsets of drugs and medical events that are associated with the development of myocardial infarction. These are the candidate confounding interaction terms. We then implemented a cohort study design using regularised cox regression that incorporated and accounted for the candidate confounding interaction terms. Results The methodology was able to account for signals generated due to confounding and a cox regression with elastic net regularisation correctly ranked the drug families known to be true adverse drug reactions above those.

Jenna Reps, Zhaoyang Guo, Haoyue Zhu et al.

Big longitudinal observational databases present the opportunity to extract new knowledge in a cost effective manner. Unfortunately, the ability of these databases to be used for causal inference is limited due to the passive way in which the data are collected resulting in various forms of bias. In this paper we investigate a method that can overcome these limitations and determine causal contrast set rules efficiently from big data. In particular, we present a new methodology for the purpose of identifying risk factors that increase a patients likelihood of experiencing the known rare side effect of renal failure after ingesting aminosalicylates. The results show that the methodology was able to identify previously researched risk factors such as being prescribed diuretics and highlighted that patients with a higher than average risk of renal failure may be even more susceptible to experiencing it as a side effect after ingesting aminosalicylates.

Jenna Reps, Uwe Aickelin, Jonathan Garibaldi et al.

Mobile advertising is a billion pound industry that is rapidly expanding. The success of an advert is measured based on how users interact with it. In this paper we investigate whether the application of unsupervised learning and association rule mining could be used to enable personalised targeting of mobile adverts with the aim of increasing the interaction rate. Over May and June 2014 we recorded advert interactions such as tapping the advert or watching the whole advert video along with the set of apps a user has installed at the time of the interaction. Based on the apps that the users have installed we applied k-means clustering to profile the users into one of ten classes. Due to the large number of apps considered we implemented dimension reduction to reduced the app feature space by mapping the apps to their iTunes category and clustered users based on the percentage of their apps that correspond to each iTunes app category. The clustering was externally validated by investigating differences between the way the ten profiles interact with the various adverts genres (lifestyle, finance and entertainment adverts). In addition association rule mining was performed to find whether the time of the day that the advert is served and the number of apps a user has installed makes certain profiles more likely to interact with the advert genres. The results showed there were clear differences in the way the profiles interact with the different advert genres and the results of this paper suggest that mobile advert targeting would improve the frequency that users interact with an advert.

Jenna M. Reps, Uwe Aickelin, Jiangang Ma et al.

Side effects of prescribed medications are a common occurrence. Electronic healthcare databases present the opportunity to identify new side effects efficiently but currently the methods are limited due to confounding (i.e. when an association between two variables is identified due to them both being associated to a third variable). In this paper we propose a proof of concept method that learns common associations and uses this knowledge to automatically refine side effect signals (i.e. exposure-outcome associations) by removing instances of the exposure-outcome associations that are caused by confounding. This leaves the signal instances that are most likely to correspond to true side effect occurrences. We then calculate a novel measure termed the confounding-adjusted risk value, a more accurate absolute risk value of a patient experiencing the outcome within 60 days of the exposure. Tentative results suggest that the method works. For the four signals (i.e. exposure-outcome associations) investigated we are able to correctly filter the majority of exposure-outcome instances that were unlikely to correspond to true side effects. The method is likely to improve when tuning the association rule mining parameters for specific health outcomes. This paper shows that it may be possible to filter signals at a patient level based on association rules learned from considering patients' medical histories. However, additional work is required to develop a way to automate the tuning of the method's parameters.

Jenna M. Reps, Uwe Aickelin, Jonathan M. Garibaldi

In previous work, a novel supervised framework implementing a binary classifier was presented that obtained excellent results for side effect discovery. Interestingly, unique side effects were identified when different binary classifiers were used within the framework, prompting the investigation of applying a multiple classifier system. In this paper we investigate tuning a side effect multiple classifying system using genetic algorithms. The results of this research show that the novel framework implementing a multiple classifying system trained using genetic algorithms can obtain a higher partial area under the receiver operating characteristic curve than implementing a single classifier. Furthermore, the framework is able to detect side effects efficiently and obtains a low false positive rate.

Jenna Reps, Jonathan M. Garibaldi, Uwe Aickelin et al.

The pharmaceutical industry is plagued by the problem of side effects that can occur anytime a prescribed medication is ingested. There has been a recent interest in using the vast quantities of medical data available in longitudinal observational databases to identify causal relationships between drugs and medical events. Unfortunately the majority of existing post marketing surveillance algorithms measure how dependant or associated an event is on the presence of a drug rather than measuring causality. In this paper we investigate potential attributes that can be used in causal inference to identify side effects based on the Bradford-Hill causality criteria. Potential attributes are developed by considering five of the causality criteria and feature selection is applied to identify the most suitable of these attributes for detecting side effects. We found that attributes based on the specificity criterion may improve side effect signalling algorithms but the experiment and dosage criteria attributes investigated in this paper did not offer sufficient additional information.

Jenna M. Reps, Jonathan M. Garibaldi, Uwe Aickelin et al.

Background: Children are frequently prescribed medication off-label, meaning there has not been sufficient testing of the medication to determine its safety or effectiveness. The main reason this safety knowledge is lacking is due to ethical restrictions that prevent children from being included in the majority of clinical trials. Objective: The objective of this paper is to investigate whether an ensemble of simple study designs can be implemented to signal acutely occurring side effects effectively within the paediatric population by using historical longitudinal data. The majority of pharmacovigilance techniques are unsupervised, but this research presents a supervised framework. Methods: Multiple measures of association are calculated for each drug and medical event pair and these are used as features that are fed into a classiffier to determine the likelihood of the drug and medical event pair corresponding to an adverse drug reaction. The classiffier is trained using known adverse drug reactions or known non-adverse drug reaction relationships. Results: The novel ensemble framework obtained a false positive rate of 0:149, a sensitivity of 0:547 and a specificity of 0:851 when implemented on a reference set of drug and medical event pairs. The novel framework consistently outperformed each individual simple study design. Conclusion: This research shows that it is possible to exploit the mechanism of causality and presents a framework for signalling adverse drug reactions effectively.

Jenna Reps, Jonathan M. Garibaldi, Uwe Aickelin et al.

The electronic healthcare databases are starting to become more readily available and are thought to have excellent potential for generating adverse drug reaction signals. The Health Improvement Network (THIN) database is an electronic healthcare database containing medical information on over 11 million patients that has excellent potential for detecting ADRs. In this paper we apply four existing electronic healthcare database signal detecting algorithms (MUTARA, HUNT, Temporal Pattern Discovery and modified ROR) on the THIN database for a selection of drugs from six chosen drug families. This is the first comparison of ADR signalling algorithms that includes MUTARA and HUNT and enabled us to set a benchmark for the adverse drug reaction signalling ability of the THIN database. The drugs were selectively chosen to enable a comparison with previous work and for variety. It was found that no algorithm was generally superior and the algorithms' natural thresholds act at variable stringencies. Furthermore, none of the algorithms perform well at detecting rare ADRs.

Jenna Reps, Jonathan M. Garibaldi, Uwe Aickelin et al.

Longitudinal observational databases have become a recent interest in the post marketing drug surveillance community due to their ability of presenting a new perspective for detecting negative side effects. Algorithms mining longitudinal observation databases are not restricted by many of the limitations associated with the more conventional methods that have been developed for spontaneous reporting system databases. In this paper we investigate the robustness of four recently developed algorithms that mine longitudinal observational databases by applying them to The Health Improvement Network (THIN) for six drugs with well document known negative side effects. Our results show that none of the existing algorithms was able to consistently identify known adverse drug reactions above events related to the cause of the drug and no algorithm was superior.

Jenna Reps, Jonathan M. Garibaldi, Uwe Aickelin et al.

The wealth of computerised medical information becoming readily available presents the opportunity to examine patterns of illnesses, therapies and responses. These patterns may be able to predict illnesses that a patient is likely to develop, allowing the implementation of preventative actions. In this paper sequential rule mining is applied to a General Practice database to find rules involving a patients age, gender and medical history. By incorporating these rules into current health-care a patient can be highlighted as susceptible to a future illness based on past or current illnesses, gender and year of birth. This knowledge has the ability to greatly improve health-care and reduce health-care costs.

Feng Gu, Jan Feyereisl, Robert Oates et al.

Theoretical analyses of the Dendritic Cell Algorithm (DCA) have yielded several criticisms about its underlying structure and operation. As a result, several alterations and fixes have been suggested in the literature to correct for these findings. A contribution of this work is to investigate the effects of replacing the classification stage of the DCA (which is known to be flawed) with a traditional machine learning technique. This work goes on to question the merits of those unique properties of the DCA that are yet to be thoroughly analysed. If none of these properties can be found to have a benefit over traditional approaches, then "fixing" the DCA is arguably less efficient than simply creating a new algorithm. This work examines the dynamic filtering property of the DCA and questions the utility of this unique feature for the anomaly detection problem. It is found that this feature, while advantageous for noisy, time-ordered classification, is not as useful as a traditional static filter for processing a synthetic dataset. It is concluded that there are still unique features of the DCA left to investigate. Areas that may be of benefit to the Artificial Immune Systems community are suggested.

Jenna Reps, Jan Feyereisl, Jonathan M. Garibaldi et al.

Data-mining techniques have frequently been developed for Spontaneous reporting databases. These techniques aim to find adverse drug events accurately and efficiently. Spontaneous reporting databases are prone to missing information, under reporting and incorrect entries. This often results in a detection lag or prevents the detection of some adverse drug events. These limitations do not occur in electronic health-care databases. In this paper, existing methods developed for spontaneous reporting databases are implemented on both a spontaneous reporting database and a general practice electronic health-care database and compared. The results suggests that the application of existing methods to the general practice database may help find signals that have gone undetected when using the spontaneous reporting system database. In addition the general practice database provides far more supplementary information, that if incorporated in analysis could provide a wealth of information for identifying adverse events more accurately.