City Research Online

Abstract

This thesis examines the role of mathematical models in the management of patients with chronic thyroid disorders.

The major benefit arises from the ability of a model to integrate large quantities of biochemical and clinical data into a simple statement of expectation and uncertainty about the current state of a patient and likely developments. The immediate application of modelling is to the assessment and selection of tests or treatment to be applied to the individual patient. However, the provision of data in a form suitable for modelling introduces a data base which may be used for longer-term objectives. A sizable "routine" data-base can provide an important source of new information for modelling and allow a realistic assessment of tests and treatments over periods of time or groups of patients which would not otherwise be possible.

On the basis of a comprehensive systems analysis, an approach to more effective patient care is shown to involve enhancement of the role of the laboratory. Rather than act as a simple, positive, "instrument-at-a-distance", the laboratory should take a more active role in transforming raw clinical and biochemical data into useful information for clinical decision making. In such a system the physician remains the decision maker but the laboratory acts as an intelligent archivist and integrator of data.

The results of a pilot study are examined to determine the acceptability and usefulness of collecting a limited amount of clinical data. An attempt is made to tackle directly the immediate problem of selecting tests in thyroid disease. Although benefits can be shown from the use of chemical data it becomes clear that this is of limited value. The major constraint appears to be the absence of time series data.

The benefit of long-term feedback from routine requests collected for modelling purposes is examined and an attempt is made to show that it provides an accessible and intelligible, though limited, data-base on thyroid patients. An important point is that the data represent the population of patients seen by the laboratory, not by the physician who tends to act as a "filter". This means however that the data can only be a starting point for epidemiological surveys.

The clinical application of existing comprehensive models of thyroid function is examined highlighting the need for more limited model structures appropriate for the specific clinical problem. Data from twelve thyrotoxic patients are used to identify a reduced clinical model. The distinct components of thyroid, pituitary and "peripheral" tissue are identified with varying degrees of success. A simple model of the thyroidal response can be identified even with a small number of measures with discrepancies between prediction and data of the order of the measurement error. The model appears to be a particulary effective method of drawing attention to unusal or anomalous results . Problems remain, however, with pituitary modelling and the extension of identification to sequential estimation. These dificulties are discussed and to demonstrate the feasibility of clinical and laboratory application the thyroidal model has been implemented upon a micro-computer. This model based approach offers aids to both the operation of the clinical chemistry laboratory and the physician during decision-making required in routine patient management.

Publication Type:	Thesis (Doctoral)
Subjects:	Q Science T Technology
Departments:	School of Science & Technology > Computer Science School of Science & Technology > School of Science & Technology Doctoral Theses Doctoral Theses

Export

Downloads