The thresholds of change model: An approach to analyzing stages of change data

Abstract

Stage models are prominent in research describing health behavior change. Since stage models often propose that different factors have varying influences on membership in the different stage, statistical methods that can estimate the thresholds that separate the stages and the relative value of variables in influencing these thresholds are useful. This article describes use of a "thresholds of change" model for analyzing the thresholds separating stages and specifically for examining the effects of explanatory variables on these thresholds using a generalization of an ordinal logistic (or probit) regression model. Data from a skin cancer prevention study (N=3,185) in which participants were grouped into three stages for sunscreen use (precontemplation, contemplation, and action) are used to illustrate the Thresholds of Change Model. For this example, two thresholds exist: a contemplation (between precontemplation and contemplation) and an action threshold (between contemplation and action). Variables examined include gender, skin type, perceived susceptibility to sunburn, worry about skin cancer, and sun protection self-efficacy. We examine models that assume that the effects of these variables are the same across thresholds, and then allow the effects of these variables to vary across thresholds. Results indicate that perceived susceptibility has an equal effect on both thresholds, but that worry and self-efficacy have differential effects: worry exerts a greater influence on the contemplation threshold, whereas self-efficacy has a significantly stronger effect on the action threshold. Gender also has a stronger effect on the action threshold; males were less likely to be classified in the action stage than females. This analytic approach has broad applications to many types of stage data.

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 38 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.