Multivariate methods to identify cancer-related symptom clusters

Res Nurs Health. 2009 Jun;32(3):345-60. doi: 10.1002/nur.20323.

Abstract

Multivariate methods are required to assess the interrelationships among multiple, concurrent symptoms. We examined the conceptual and contextual appropriateness of commonly used multivariate methods for cancer symptom cluster identification. From 178 publications identified in an online database search of Medline, CINAHL, and PsycINFO, limited to articles published in English, 10 years prior to March 2007, 13 cross-sectional studies met the inclusion criteria. Conceptually, common factor analysis (FA) and hierarchical cluster analysis (HCA) are appropriate for symptom cluster identification, not principal component analysis. As a basis for new directions in symptom management, FA methods are more appropriate than HCA. Principal axis factoring or maximum likelihood factoring, the scree plot, oblique rotation, and clinical interpretation are recommended approaches to symptom cluster identification.

Publication types

  • Research Support, Non-U.S. Gov't
  • Review

MeSH terms

  • Cluster Analysis*
  • Cross-Sectional Studies*
  • Data Interpretation, Statistical*
  • Decision Support Techniques
  • Factor Analysis, Statistical
  • Guidelines as Topic
  • Humans
  • Likelihood Functions
  • Multivariate Analysis*
  • Neoplasms / complications*
  • Neoplasms / nursing
  • Nursing Assessment
  • Nursing Research / methods*
  • Principal Component Analysis
  • Regression Analysis
  • Reproducibility of Results
  • Research Design