Abstract
Pharmaco-cybernetics is an upcoming interdisciplinary field that supports our use of medicines and drugs through the combined use of computational technologies and techniques with human-computer-environment interactions to reduce or prevent drug-related problems. The advent of pharmaco-cybernetics has led to the development of various software, tools, and Internet applications that can be used by healthcare practitioners to deliver optimum pharmaceutical care and health-related outcomes. Patients are becoming more informed through health information on the Internet, which empowers them to better participate in the management of their own conditions. Focusing on patients with cancer, this chapter describes the use of a pharmaco-cybernetics approach to identify clinically relevant predictors of two debilitating adverse drug reactions, which are a cause of patient safety – chemotherapy-induced nausea and vomiting and febrile neutropenia. The early identification of such clinical predictors enables clinicians to prevent or reduce the occurrence of adverse drug reactions in cancer patients undergoing chemotherapy through appropriate management strategies. The computational methods used in this approach involve two unsupervised machine-learning techniques – principal component and multiple correspondence analyses. Using two case examples, this chapter shows the potential of machine-learning techniques for identifying patients who are at greater risks of these adverse drug reactions, thus enhancing patient safety. This chapter also aims to increase the awareness among healthcare professionals and clinician-scientists about the usefulness of such techniques in clinical patient populations, so that these can be considered as part of clinical care pathways to enhance patient safety and effectively manage cancer patients on chemotherapy.
TopIntroduction
Pharmaceutical care is critical for improving medication management in patients, particularly those with chronic diseases. The practice of pharmaceutical care is the basis of clinical pharmacy and it involves identifying, preventing and solving drug-related problems with regards to patients' drug therapies (American Society of Hospital Pharmacists, 1993; Westerlund, Almarsdóttir, & Melander, 1999). Simply put, it helps patients make the best use of their medications. Drug-related problems exist in many forms, but essentially, they are events or circumstances involving drug therapies that can actually or potentially interfere with desired health outcomes (American Society of Hospital Pharmacists, 1993; van Mil, Westerlund, Hersberger, & Schaefer, 2004). Drug-related problems impact medication safety, and as a consequence, patient safety as well. The Pharmaceutical Care Network Europe Foundation categorizes drug-related problems in terms of problems and causes (Pharmaceutical Care Network Europe, 2010), while the American Society of Health-System Pharmacists classifies drug-related problems into eight main categories - see Figure 1. Drug-related problems can result due to a lack of knowledge or misinterpretation of drug information, which can compromise patients’ safety and quality of life if not treated effectively and appropriately.
Figure 1. List of drug-related problems
Cancer is a highly prevalent health problem with increasing incidence worldwide. In 2007, one in 8 deaths was due to cancer (Garcia et al., 2007). The global burden of cancer is also expected to grow to 27 million new cancer cases and 17.5 million cancer deaths by 2050 (Garcia et al., 2007). The rapid growth of informatics technologies and the World Wide Web in the last decade has enabled the development of many applications that can assist clinicians in delivering optimum pharmaceutical care and health-related outcomes. Patients with cancer are becoming more well informed through online health information that is readily accessible, and they can better participate in the management of their own conditions through knowledge obtained from the Internet. However, despite the advancement of informatics technologies, little has been done in clinical oncology practice to leverage on the use of these technologies to improve the pharmaceutical care of patients with cancer.
Key Terms in this Chapter
Chemotherapy-Induced Nausea And Vomiting: An adverse drug reaction characterised by nausea and vomiting experienced by patients with cancer receiving emetogenic chemotherapies. Acute nausea and vomiting usually lasts up to 24 hours after chemotherapy; while delayed nausea and vomiting usually occurs after 24 hours and lasts up to 5 to 7 days.
Pharmaco-Informatics: A field of pharmacy which involves the use of informatics and internet technologies to target drug-related problems. It is focused on the acquisition, dissemination, storage, analysis and use of medication-related knowledge and data within the continuum of healthcare systems.
Multiple Correspondence Analysis: A multivariate technique within unsupervised machine learning for exploring data. It is similar to PCA but is catered towards discretized categorical data and usually involves frequency tables or counts.
Febrile Neutropenia: An adverse drug reaction that is usually a serious complication of chemotherapy, particularly in cancer patients who receive myelosuppressive chemotherapies. This condition is characterised by a fever and a significant reduction in their white blood cells.
Unsupervised Machine Learning: A computational method that typically involves a model that is generated to fit observations. Unlike supervised learning, there is no a priori output, therefore characteristics about the data set can be described from the observations without any predefined target.
Adverse Drug Reaction: A condition experienced by patients characterised by injury or harm associated with the use of drugs/medications (e.g. chemotherapies) at normal dosages during normal use.
Pharmaceutical Care: A concept involving identifying, solving and preventing potential or actual drug-related problems with regards to a patient's drug therapy.
Supervised Machine Learning: A computational method that is able to predict the output value of an unknown object based on the training of a set of training examples (input and output data).
Drug-Related Problem: An event or circumstance which involves drug therapies that can actually or potentially interfere with the desired health outcome for patients.
Pharmaco-Cybernetics: An upcoming interdisciplinary field that describes the science of supporting medicines or drugs use through the application of computational technologies and techniques, combined with the principles of human-computer-environment interaction, to reduce or prevent drug-related problems. Also known as cybernetic pharmacy.
Machine Learning: A field in artificial intelligence which designs computational algorithms and techniques capable of inducing knowledge from data. Generally, there are two main types: supervised and unsupervised learning.
Principal Component Analysis: A multivariate projection technique within unsupervised machine learning that can investigate relationships among multiple variables and explain the causes of variance in a data set. This technique linearly transforms an original set of variables into a substantially smaller set of uncorrelated variables called principal components, which represent most of the information in the original data set.