An informatics based approach to respiratory healthcare.

By 2005 one person in every five UK households suffered with asthma. Research has shown that episodes of poor air quality can have a negative effect on respiratory health and is a growing concern for the asthmatic. To better inform clinical staff and patients to the contribution of poor air quality on patient health, this thesis defines an IT architecture that can be used by systems to identify environmental predictors leading to a decline in respiratory health of an individual patient.
Personal environmental predictors of asthma exacerbation are identified by validating the delay between environmental predictors and decline in respiratory health. The concept is demonstrated using prototype software, and indicates that the analytical methods provide a mechanism to
produce an early warning of impending asthma exacerbation due to poor air quality. The author has introduced the term enviromedics to describe this new field of research.
Pattern recognition techniques are used to analyse patient-specific environments, and extract meaningful health predictors from the large quantities of data involved (often in the region of '/o million data points).
This research proposes a suitable architecture that defines processes and techniques that enable the validation of patient-specific environmental predictors of respiratory decline. The design of the architecture was validated by implementing prototype applications that demonstrate, through hospital admissions data and personal lung function monitoring, that air quality can be used as a
predictor of patient-specific health. The refined techniques developed during the research (such as Feature Detection Analysis) were also validated by the application prototypes.
This thesis makes several contributions to knowledge, including: the process architecture; Feature Detection Analysis (FDA) that automates the detection of trend reversals within time series data; validation of the delay characteristic using a Self-organising Map (SOM) that is used as an unsupervised method of pattern recognition; Frequency, Boundary and Cluster Analysis (FBCA), an additional technique developed by this research to refine the SOM.

A thesis submitted to Middlesex University in partial fulfilment of the requirements for the degree of Doctor of Philosophy.