Statistical methods for developing index measures of compassion

Abstract

Compassion is a complex, multi-dimensional concept. As a result, measuring compassion is a complex task. The measurements can be subjective, varying by person and context. Often, the questions used to create measures of compassion are indirect or incomplete due to their multi-dimensional aspects. Quantifying compassion utilizes statistical tools not commonly used in the field of epidemiology.

This paper reviews two families of statistical methods (primarily Principal Components Analysis and Factor Analysis) that can be used to meet these challenges. The review will cover the central concepts of each method and reference sources with data, annotated output, and code using several popular products (R, SPSS, STATA, SAS). Several currently used compassion indices are discussed, using a simple conceptual framework designed to illustrate typical challenges in the process of index development. The framework is composed of three components: the caregiver, the environment, and the recipient, with the environment broadly defined to include management policies, the type of health care facility, and other factors external to the caregiver and patient.

Summary discussions include the challenges associated with these methods and those associated with integrating results from these analyses into the statistical and mathematical modeling processes.