Use this URL to cite or link to this record in EThOS: https://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.772963
Title: The application of predictive statistical modelling in the investigation of suspected classical Myeloproliferative Neoplasms
Author: Cullen, Fiona Louise
ISNI:       0000 0004 7960 4172
Awarding Body: University of York
Current Institution: University of York
Date of Award: 2018
Availability of Full Text:
Access from EThOS:
Access from Institution:
Abstract:
The Myeloproliferative Neoplasms are a clinically heterogenous group of bone marrow haemopoietic disorders that result in the overproduction of myeloid blood cells. The identification of recurrent gene mutations has aided positive identification of disease in a substantial proportion of patients, however, a significant number of individuals with classical MPNs do not have a detectable aberration. The clinical and laboratory presentation of these disorders shows significant overlap with features associated with reactive conditions, which, in patients without detectable genetic mutations, can lead to ambiguity in their diagnosis. Traditionally, the decision to investigate an individual for a suspected classical MPN has been based upon thresholds in blood count parameters. In this work, current working practices have been audited to ascertain the extent to which diagnostic guidelines are adhered toe. We demonstrate that a significant proportion of referrals for the investigation of suspected classical MPNs do not meet these criteria. Furthermore, this work objectively assesses the diagnostic sensitivity and specificity of current guidelines in the identification of patients with classical MPNs. The use of predictive statistical modelling is a contemporary approach to the identification of individuals with increased likelihood of suffering from a classical MPN. In this work, several predictive modelling methods were applied to a data set of laboratory and basic demographic information taken from a series of patients investigated for suspected classical MPNs. This work shows that predictive statistical modelling can reproducibly identify those patients who are likely to have a classical MPN from those who do not. These models offer increased specificity and sensitivity compared with the use of published investigatory and diagnostic guidelines. Predictive statistical modelling also offers the ability to triage those patients who are likely to have classical MPNs prior to further investigation, resulting in potentially significant cost savings to both clinical and laboratory services.
Supervisor: Crouch, Simon ; Roman, Eve ; Kelly, Richard Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID: uk.bl.ethos.772963  DOI: Not available
Share: