The importance of separating between- and within- person effects as a basis for theory-based intervention

Abstract

Background. If theories and interventions are to improve health outcomes then they must apply to change within the individual. This cannot be examined in between person analysis. We apply within- and between- person methods to demand/resource models of occupational stress in nurses. Method. 100 nurses were assessed over 2 shifts with real-time measurement of demand, control and reward (predictors) and negative affect (NA) and positive affect (PA) (outcomes) every 90 minutes. Multilevel modelling was used to test the between- and within-person effects, and examine differences in within-person effects over individuals (random effects). Findings. NA: Demand was associated with increased NA within but not between people (between beta 1.0 (SE 0.8), within 2.7 (0.2)). Control and reward were associated with decreased NA both between and within (control between -4.5 (1.8), within -2.4 (0.5): reward between -2.6 (1.1); within -1.5 (0.4)). PA: Demand was associated with decreased PA within (-0.7 (0.3)), but not between people (-0.3 (1.6)); however a substantial minority of participants displayed within-person increases in PA with increasing demand. Control related positively both within (2.8(0.5)); and between (6.4(3.0)) as did reward ((2.6(0.4), 9.8(1.6)). There were substantial random effects in both NA and PA analyses. Discussion. Between and within person analyses provided important and different information. Between person analyses failed to show the relationship between demand and PA and obscured the high within person variability of all the relationships. Within person analyses could not show the importance of absolute levels of control and reward.