AbstractBackground: Factor analysis (FA) is a statistical technique that analyses underlying relationships in a data set. It is mainly used in the development and validation of questionnaires. Since valid outcome measures have become important in palliative care, criteria to judge the quality of FA are needed
Aims and objectives: To systematically review the current use and reporting of either exploratory or confirmatory FA and recommendations for FA, in order to design a checklist for reporting FA in the field of palliative care. Objectives were to describe and analyse sample characteristics of included studies, to appraise reporting of FA and to develop a succinct outcome score. Variables that influenced the quality of reporting were assessed.
Methods: In this systematic literature review, a hand search identified all studies using factor analytical techniques published in Palliative Medicine, Journal of Pain & Symptom Management, Journal of Palliative Medicine, and the Journal of Palliative Care (inception – September 2011). Included were studies of any design, except reviews, reporting the results of one or several exploratory or confirmatory factor analysis (EFA and CFA). Studies were scored using data extraction forms and one fifth was independently assessed by a second
reviewer. Percentages were examined for each reporting variable. Differences between reporting practices of EFA and CFA and different journals were assessed by nonparametric tests. Two outcome scores were calculated based on sums of variables that were scored as ‘reported’.
Results: In 8720 articles screened, 191 factor analytical studies were identified (166 EFA and 42 CFA). The majority of studies used cross-sectional designs (74.3%), factor analysed one questionnaire (87.4%) and used non-random sampling procedures (80.1%). Variables such as stage of disease and diagnosis indicated that not only palliative patients were included in original studies with some being at the beginning of disease. For EFA studies, purpose of factor analysis, factor extraction model, rotation, factor retention criteria and quality of reporting were examined. Each variable had ca. 20% of missing values, indicating
paucity of reporting. A similar picture was observed for CFA. For EFA, the outcome score showed a mean of 8.1 (SD = 2.28), indicating that the majority of articles reported less than 10 variables defined to be essential in EFA. In CFA, the mean was 10.4 (SD = 5.17), showing that CFA studies on average scored higher than EFA studies and almost reported the 11 essential variables. Hypothesis tests indicated that FA was better reported when EFA or CFA was a part of the aim. Comparison to similar systematic reviews of other disciplines showed that problems with reporting of factor analysis in palliative care are comparable to other research fields.
Conclusions and implications: To enhance reporting practices in palliative care, a checklist was developed that can guide researchers, authors and editors when evaluating or calculating a factor analysis. The finding that reporting practices were mainly poor with only one fifth of articles having high reporting quality shows the need for guidance. However, the checklist needs to be further evaluated with experts in a Delphi study.
|Date of Award||2012|
|Supervisor||Richard Siegert (Supervisor) & Irene Higginson (Supervisor)|