White Rose University Consortium logo
University of Leeds logo University of Sheffield logo York University logo

Computer-assisted approaches to the collection of quality of life data in oncology

Smith , Adam Barnett (2004) Computer-assisted approaches to the collection of quality of life data in oncology. PhD thesis, University of Leeds.

[img] Text
uk_bl_ethos_406883.pdf

Download (22Mb)

Abstract

The assessment of cancer patients' quality of life (QOL) has been increasing in both importance and relevance in recent years, and is becoming more integrated into clinical practice. This has been greatly facilitated by the development of standard QOL instruments. However, the standard questionnaires may overlook certain aspects of QOL or focus on areas which do not present a problem to patients. The aims of this thesis were to increase the relevance of QOL instruments to patients by developing systems that allow patients to select relevant domains from questionnaires and secondly, to minimise patient burden by reducing the number of questions presented to patients. Initially, a computer-assisted version of the EORTC QLQ-C30 was compared with a standard electronic version of the questionnaire. Patients completed both forms on the same day. The results demonstrated that although patients completed the computer-assisted questionnaire more quickly, there was poor exact agreement, between the two forms. However, general agreement was good (i. e. > 70%) for all symptom scales, but not for the majority of the functioning scales. In addition, patients tended to report higher levels of symptoms and poorer functioning on the standard questionnaire. Studies were then developed and conducted using Factor and Rasch analyses on a series of standard questionnaires, namely the HADS, the EORTC QLQ-C30, and the FACT-G, in order to assess their structure and the performance of each item. The results from HADS scale demonstrated a two-factor structurecorresponding to anxiety and depression, and an overall psychological distress measure. In addition to confirming this structure, the Rasch analysis identified one misfitting item for each of the full HADS-scale and two subscales. For the EORTC QLQ-C30 the results demonstrated a four-factor structure corresponding to a physical functioning factor, a factor covering social and role functioning, and including pain and fatigue symptoms, a third factor covering the emotional and cognitive functioning domains, and finally a factor covering the remaining symptoms. The Rasch analysis demonstrated good fit for all items of the Emotional Functioning, and Fatigue scales, and only one misfitting item from the Physical Functioning scale. The results for the FACT-G demonstrated four factors corresponding to the four FACT-G subscales, although all subscales contained at least two misfitting items. The misfitting items from the HADS were systematically removed from the HADS and its subscales, and the screening efficacy of the scales re-evaluated against psychiatric interview data (PSE/SCAN). The results demonstrated no loss in screening efficacy when these items were removed. In the final study scores from the corresponding scales of the EORTC QLQC30 and FACT-G were converted to log-odds (logit) scores and agreement between the scales was calculated. The results demonstrated high levels of agreement between three of the scales, namely Physical and Emotional Functioning and overall quality of life, and good levels of agreement for the other two scales (Role and Social Functioning). In conclusion, the utility of Rasch models in identifying items for removal from instruments in order to reduce patient burden was demonstrated in this thesis. This work provides a foundation for the subsequent development of computer-adaptive questionnaires.

Item Type: Thesis (PhD)
Academic Units: The University of Leeds > Faculty of Medicine and Health (Leeds) > School of Medicine (Leeds)
Depositing User: Ethos Import
Date Deposited: 22 Mar 2010 12:24
Last Modified: 07 Mar 2014 10:21
URI: http://etheses.whiterose.ac.uk/id/eprint/732

Actions (repository staff only: login required)