Jenna Reps

BSc Mathematics (University of Bath)

MSc Mathematical Biology (University of Bath)

Research

It is becoming more apparent that in the future we will have personalised healthcare.  An example of personalised healthcare in the context of drug prescriptions is a doctor determining the optimal treatment for a patient based on the patients attributes (such as age, gender, medical history or genetics).  This can be done by identifying the general associations between patient’s attributes and outcomes after being prescribed a drug.  With this knowledge we can help identify and reduce the occurrence of negative side effects and ensure the patients have the best prognosis possible.

One research area for this type of personalised healthcare is the application of data mining techniques to medical databases to find associations between certain attributes (dosage, patient age, patient gender, patient medical history, etc...) and the occurrence of a negative side effect.  The majority of existing methods aiming to determine negative side effects are applied to spontaneous reporting system (SRS) databases.  These SRS databases consist of a collection of voluntary reports corresponding to suspected cases of a negative side effect and contain information about the drugs being taken, the side effect/s, the patient’s age and the patient’s gender.  Unfortunately, the majority of SRS entries contain missing information and the database is known to contain duplicated or incorrect entries.  The database is also limited by under-reporting as people may not report less serious side effects or very rare side effects may go undetected and will therefore not be reported.  As a consequence it is difficult to mine these databases to find patient attributes that increase a patient’s chance of having a negative side effect.

Longitudinal observation databases (LODs) are a more recently available type of medical database that often contains a wealth of information on the patients.  Interest in mining this type of database for negative side effects has attracted recent attention due to the LODs not suffering from many of the limitations of the SRS databases.  My current research interest is applying association and sequential pattern mining algorithms to a specific LOD database known as the health improvement network (THIN) database to detect associations between patient attributes and negative side effects that can be used for personalised healthcare.

Teaching & Supervision

• My supervisors are Jon Garibaldi and Uwe Aickelin

Intelligent Modelling and Analysis Research Group, School of Computer Science, The University of Nottingham, Jubilee Campus, Nottingham, NG8 1BB, UK.