On Nurses’ Learning from Errors at Work

Abstract

In this chapter, we aim at synthesizing several of our studies that investigated processes, practices, and conditions of nurses’ learning at work related to the occurrence of errors. Dealing with errors in a learning-oriented way has been subject to intensive debate in health care. Researchers have argued that errors can be important sources of professional learning, if they are analyzed properly, and thereby help reduce the probability of their re-occurrence. The chapter starts with a discussion of nursing as a profession with high demands for continuous learning at work and professional development. In this context, we highlight the potential contribution of learning from errors. Next, we summarize and integrate findings from two expert-interview studies and two surveys, respectively, which investigated typical errors in hospital and elder care nursing as well as available evidence on a model of nurses’ engagement in error-related learning activities. In the conclusion, we offer suggestions on how nurses’ learning from errors may be supported on the individual, interpersonal, and organizational levels. Moreover, we elaborate on consequences for nurses’ initial vocational education and continuing professional development.

Appendix

MASEM is a recent development in the field of model-based meta-analysis aiming at the integration of findings from structural equation models (Cheung & Chan, 2005). For synthesizing the findings from Bauer and Mulder (2013) and Leicher et al. (2013), we used the two-stage structural equation modeling (TSSEM) approach to MASEM (Cheung & Chan, 2005). This approach proceeds in two steps. In stage one, the homogeneity of the correlation matrices of the individual studies’ variables is tested using multiple group SEM (i.e., with studies as groups). If a model imposing equality constraints on the correlation matrices across studies fits the data well, as judged by SEM fit indices, then there is sufficient evidence of homogeneity. In that case, a pooled correlation matrix can be estimated across the studies. In stage two, this pooled correlation matrix is used to fit the substantive model.

We used the metaSEM package version 0.9–1 (Cheung, 2015) in the R statistical environment version 3.1.2 (R Core Team, 2014) to fit the MASEM. In stage one, a fixed effects model was used because only two studies entered the analysis.

The results of stage one indicated sufficient homogeneity of the two studies’ correlation matrices (χ²(105) = 213, p < 0.001; RMSEA = 0.069, CFI = 0.970, TLI = 0.940, SRMR = 0.059). In the stage two analysis we specified the model as depicted in Fig. 7.1. To avoid clutter, the measurement part of the model is omitted in the figure. ESLA is a second-order factor comprising the social learning activities general cause analysis, specific cause analysis and development of new strategies as first order factors (cf. Bauer & Mulder, 2013). Fit indices from the stage two analysis indicated acceptable fit for the mediation model (see Fig. 7.1). First order factor loadings range between 0.53 and 0.99, second order factor loadings between 0.67 and 0.74.