Modeling medication prescriptions and adherence
Date
2015
Authors
Diemert, Simon
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
In modern medicine medication prescriptions are one of the primary forms of intervention. Prescriptions are effectively a specification for how to consume a particular kind of substance. A prescription’s complexity can range from a simple verbal orders to intricate sequences of actions that vary depending on a number of different conditions. The adoption of Health Information Technology (HIT) has given rise to electronic prescribing (e-prescribing). Much of the functionality available e-prescribing systems requires that prescriptions be entered in a discrete and well-structured format. Unfortunately, these systems typically only capture a subset of the concepts of prescribing, while the rest of the concepts are entered as free text. The result is poor data quality which affects the quality of the decision support provided by the HIT system.
Advances in consumer technology have also allowed for the ability to measure how patients take medications. Such technologies are aimed at addressing the concern of medication adherence, where a patient does not adhere to their prescribed medication regimes. Non-adherence can lead to serious medical problems, and has been identified by numerous authorities and academics as a factor effecting the health care outcomes of both individuals and populations.
This thesis attempts to address parts of both e-Prescribing data quality and medication adherence. A Domain Specific Language (DSL) for authoring prescriptions was created. The goal of the DSL is to provide a structured means of creating medication prescriptions that is similar to the natural language expression of prescriptions used by many clinicians. The textual input from the DSL is transformed into a graph model. Once in graph form, the prescription can be manipulated using a computational technique called graph transformation that reduces prescription to a core model. The core model describes the entire prescription in a series of Atomic Prescribed Medication Actions (APMAs). This core model can be thought of as a plan for the prescription. When the core model is combined with the patient’s actual “execution” of the prescription (collected from medication adherence measuring device) the degree of medication adherence can be computed. Using these graph models and transformation systems provides formal definition for the semantics of the prescribing DSL used as input. Such formality provides a means of reasoning about prescriptions in a mathematical way that can then be used to provide feedback and support for clinical and patient users.
Supervisors: Dr. Jens Weber and Dr. Morgan Price, Department of Computer Science
Description
SENG498 Honours Thesis
Keywords
Software Engineering Program