Machine Learning in Drug Safety
Expert-defined terms from the Postgraduate Certificate in AI Applications in Pharmacovigilance course at Stanmore School of Business. Free to read, free to share, paired with a globally recognised certification pathway.
Absolute Risk Reduction refers to the difference in the risk of a specific adver… #
The concept of Absolute Risk Reduction is closely related to relative risk and odds ratio, which are also used to assess the risk of adverse events. For example, if a clinical trial shows that a new drug has an Absolute Risk Reduction of 0.05 Compared to a placebo, it means that the new drug reduces the risk of a specific adverse event by 5%.
Adverse Drug Reaction refers to an unwanted and potentially harmful effect of a… #
Adverse Drug Reactions can be further categorized into type A and type B reactions, depending on their predictability and severity. Type A reactions are predictable and dose-dependent, while type B reactions are unpredictable and not dose-dependent. For instance, an allergic reaction to a drug is an example of a type B Adverse Drug Reaction.
Adverse Event Reporting is the process of documenting and submitting reports of… #
Adverse Event Reporting is an essential aspect of pharmacovigilance, as it helps to identify potential safety issues and inform regulatory decisions. For example, healthcare professionals and patients can report adverse events using MedWatch or EudraVigilance, which are online reporting systems.
Artificial Intelligence refers to the development of computer systems that can p… #
In the context of pharmacovigilance, Artificial Intelligence can be used to analyze large datasets, identify patterns, and predict potential safety issues. For instance, machine learning algorithms can be used to analyze electronic health records and identify potential adverse events.
Bayesian Methods are a type of statistical approach that uses Bayes' theorem to… #
In pharmacovigilance, Bayesian Methods can be used to analyze adverse event reports and estimate the probability of a causal relationship between a drug and an adverse event. For example, Bayesian Methods can be used to analyze spontaneous reports of adverse events and estimate the probability of a causal relationship.
Bias refers to any systematic error or distortion in the collection, analysis, o… #
In pharmacovigilance, bias can occur in various forms, such as selection bias, information bias, and confounding bias. For instance, selection bias can occur when the sample population is not representative of the target population, which can lead to incorrect estimates of adverse event rates.
Causal Relationship refers to a cause #
and-effect relationship between a drug and an adverse event, which can be established through various types of evidence, including epidemiological studies and clinical trials. In pharmacovigilance, establishing a causal relationship is crucial for making regulatory decisions and informing clinical practice. For example, a case-control study can be used to establish a causal relationship between a drug and an adverse event.
Confounding Variable refers to a variable that can affect the outcome of a study… #
In pharmacovigilance, confounding variables can lead to bias and confounding, which can distort the results of a study. For instance, age and sex can be confounding variables in a study of adverse event rates, as they can affect both the exposure to a drug and the risk of an adverse event.
Data Mining refers to the process of automatically discovering patterns and rela… #
In pharmacovigilance, data mining can be used to identify potential safety issues and inform regulatory decisions. For example, data mining can be used to analyze electronic health records and identify potential adverse events.
Deep Learning refers to a type of machine learning that uses neural netwo… #
In pharmacovigilance, deep learning can be used to analyze large datasets, such as electronic health records and social media data, to identify potential safety issues. For instance, deep learning can be used to analyze medical images to identify potential adverse events.
Drug #
Drug Interaction refers to a potential interaction between two or more drugs that can lead to adverse events or alter the efficacy of one or both drugs. In pharmacovigilance, drug-drug interactions are a critical safety concern, as they can lead to serious adverse events. For example, a proton pump inhibitor can interact with a blood thinner to increase the risk of bleeding.
Electronic Health Records refer to digital versions of a patient's medical histo… #
In pharmacovigilance, electronic health records can be used to identify potential adverse events and inform regulatory decisions. For instance, electronic health records can be used to analyze adverse event reports and identify potential safety issues.
Epidemiology refers to the study of the distribution and determinants of health #
related events, diseases, or health-related characteristics among populations. In pharmacovigilance, epidemiology is used to study the incidence and prevalence of adverse events and to identify potential safety issues. For example, an epidemiological study can be used to estimate the incidence of adverse events associated with a new drug.
Evidence #
Based Medicine refers to the use of current best evidence in making decisions about the care of individual patients, which integrates clinical expertise and patient values with the best available research evidence. In pharmacovigilance, evidence-based medicine is critical for making regulatory decisions and informing clinical practice. For instance, systematic reviews and meta-analyses can be used to inform evidence-based medicine.
FDA refers to the Food and Drug Administration, which is a regulatory agency res… #
In pharmacovigilance, the FDA plays a critical role in regulating the safety of drugs and medical devices. For example, the FDA can require post-marketing surveillance studies to monitor the safety of a new drug.
Genomics refers to the study of the structure, function, and evolution of genome… #
In pharmacovigilance, genomics can be used to identify potential genetic biomarkers that can predict the risk of adverse events. For instance, genetic testing can be used to identify patients who are at risk of adverse events associated with a particular drug.
Health Informatics refers to the intersection of healthcare and information tech… #
In pharmacovigilance, health informatics is critical for collecting, analyzing, and reporting adverse event data. For example, electronic health records and adverse event reporting systems rely on health informatics to function effectively.
Imbalanced Data refers to a type of dataset where the number of instances of one… #
In pharmacovigilance, imbalanced data can occur when the number of adverse event reports is significantly smaller than the number of non-adverse event reports. For instance, oversampling and undersampling can be used to address imbalanced data.
K-Fold Cross-Validation refers to a type of cross-validation that involve… #
In pharmacovigilance, k-fold cross-validation can be used to evaluate the performance of a model that predicts adverse events. For example, k-fold cross-validation can be used to evaluate the performance of a random forest model that predicts adverse events.
Label Noise refers to a type of noise that occurs when the labels or anno… #
In pharmacovigilance, label noise can occur when adverse event reports are misclassified or incomplete. For instance, active learning can be used to address label noise by selecting the most informative samples for annotation.
Machine Learning refers to a type of artificial intelligence that involve… #
In pharmacovigilance, machine learning can be used to analyze large datasets, identify patterns, and predict potential safety issues. For example, supervised learning can be used to train a model that predicts adverse events.
Medication Error refers to a preventable adverse event that occurs when a patien… #
In pharmacovigilance, medication errors are a critical safety concern, as they can lead to serious adverse events. For instance, a barcoding system can be used to prevent medication errors by ensuring that the correct medication is administered to the correct patient.
Meta #
Analysis refers to a type of statistical analysis that combines the results of multiple studies to draw a conclusion about a particular research question, which can be used to estimate the incidence and prevalence of adverse events. In pharmacovigilance, meta-analysis can be used to estimate the incidence of adverse events associated with a new drug. For example, a meta-analysis can be used to estimate the incidence of adverse events associated with a new anticoagulant.
Natural Language Processing refers to a type of artificial intelligence t… #
In pharmacovigilance, natural language processing can be used to analyze electronic health records and identify potential adverse events. For instance, named entity recognition can be used to extract relevant information from adverse event reports.
Neural Network refers to a type of machine learning model that is inspire… #
In pharmacovigilance, neural networks can be used to analyze large datasets, identify patterns, and predict potential safety issues. For example, a convolutional neural network can be used to analyze medical images and identify potential adverse events.
Off #
Label Use refers to the use of a drug for a purpose or in a manner that is not approved by the regulatory authorities, which can increase the risk of adverse events. In pharmacovigilance, off-label use is a critical safety concern, as it can lead to serious adverse events. For instance, a black box warning can be used to warn healthcare professionals and patients about the risks of off-label use.
Overfitting refers to a type of bias that occurs when a machine learni… #
In pharmacovigilance, overfitting can occur when a model is trained on a small dataset or when the model is too complex. For example, regularization can be used to prevent overfitting by adding a penalty term to the loss function.
Pharmacogenomics refers to the study of the relationship between genetic variati… #
In pharmacovigilance, pharmacogenomics can be used to identify potential genetic biomarkers that can predict the risk of adverse events.
Predictive Modeling refers to the use of statistical models to predict the likel… #
In pharmacovigilance, predictive modeling can be used to predict the risk of adverse events associated with a new drug. For instance, a logistic regression model can be used to predict the risk of adverse events associated with a new anticoagulant.
Random Forest refers to a type of machine learning model that consists of… #
In pharmacovigilance, random forest can be used to predict the risk of adverse events associated with a new drug. For example, a random forest model can be used to predict the risk of adverse events associated with a new antibiotic.
Regression Analysis refers to a type of statistical analysis that involves model… #
In pharmacovigilance, regression analysis can be used to estimate the incidence of adverse events associated with a new drug. For instance, a linear regression model can be used to estimate the incidence of adverse events associated with a new anticoagulant.
Risk Management refers to the process of identifying, assessing, and mitigating… #
In pharmacovigilance, risk management is critical for ensuring the safe use of drugs and minimizing the risk of adverse events. For example, a risk management plan can be used to mitigate the risks associated with a new drug.
Root Cause Analysis refers to a type of analysis that involves identifying the u… #
In pharmacovigilance, root cause analysis can be used to identify the underlying causes of adverse events associated with a new drug. For instance, a fishbone diagram can be used to identify the underlying causes of adverse events.
Safety Signal refers to a potential safety concern that is identified through th… #
In pharmacovigilance, safety signals are critical for identifying potential safety issues and minimizing the risk of adverse events. For example, a safety signal can be used to inform the development of a risk management plan.
Sentinel Initiative refers to a type of active surveillance that involves… #
In pharmacovigilance, the Sentinel Initiative is a critical tool for monitoring the safety of drugs and minimizing the risk of adverse events. For instance, the Sentinel Initiative can be used to monitor the safety of a new anticoagulant.
Signal Detection refers to the process of identifying potential safety concerns… #
In pharmacovigilance, signal detection is critical for identifying potential safety issues and minimizing the risk of adverse events. For example, a signal detection algorithm can be used to identify potential safety concerns associated with a new drug.
Spontaneous Report refers to a type of adverse event report that is submitted vo… #
In pharmacovigilance, spontaneous reports are critical for identifying potential safety concerns and minimizing the risk of adverse events. For example, a spontaneous report can be used to inform the development of a risk management plan.
Supervised Learning refers to a type of machine learning that involves tr… #
In pharmacovigilance, supervised learning can be used to predict the risk of adverse events associated with a new drug. For instance, a supervised learning model can be used to predict the risk of adverse events associated with a new anticoagulant.
Systematic Review refers to a type of review that involves the use of a systemat… #
In pharmacovigilance, systematic reviews are critical for informing regulatory decisions and ensuring the safe use of drugs. For example, a systematic review can be used to estimate the incidence of adverse events associated with a new antibiotic.
Underreporting refers to the phenomenon where the number of adverse event report… #
In pharmacovigilance, underreporting is a critical concern, as it can lead to incorrect estimates of adverse event rates. For instance, active surveillance can be used to address underreporting by actively seeking out adverse event reports.
Unsupervised Learning refers to a type of machine learning that involves… #
In pharmacovigilance, unsupervised learning can be used to identify potential safety issues and inform regulatory decisions. For example, an unsupervised learning model can be used to identify clusters of adverse event reports.
Validation refers to the process of evaluating the performance of a machine l… #
In pharmacovigilance, validation is critical for ensuring that the results of a machine learning model are accurate and reliable. For instance, cross-validation can be used to evaluate the performance of a machine learning model.
Variability refers to the phenomenon where the results of a study or analysis ca… #
In pharmacovigilance, variability is a critical concern, as it can lead to incorrect estimates of adverse event rates. For example, sensitivity analysis can be used to address variability by evaluating the robustness of the results to different assumptions and scenarios.
Web Mining refers to the process of automatically discovering and extracting use… #
In pharmacovigilance, web mining can be used to collect adverse event reports and identify potential safety concerns. For instance, web scraping can be used to collect adverse event reports from online forums and social media platforms.