Patient safety and satisfaction drivers in emergency departments re-visited – an empirical analysis using structural equation modeling

Abstract

How can emergency department (ED) decision makers contribute to increase patient satisfaction rates? This question has been thoroughly investigated in many hospital departments but not so much in the ED, which has led to a number of untested hypotheses. Maximising value-added activities seen from a patient’s perspective has become an essential outcome in health care, meaning that the untested hypotheses are in need of quantitative testing. This study proposes an integrated framework in which four latent constructs reflecting principal aspects of patient care are tested. The four constructs are entitled safety and satisfaction, waiting time, information delivery, and infrastructure accordingly. As an empirical foundation, a recently published comprehensive survey in 11 Danish EDs is analysed in depth using structural equation modeling (SEM). Consulting the proposed framework, ED decision makers are provided with information of where to launch high-impact initiatives to enhance current satisfaction levels.

Appendices

Appendix A1

Selected measures from the survey instrument were examined for an initial unrotated solution by calculating the ‘Kaiser-Meyer-Olkin (KMO)’ measure of sampling accuracy and ‘Bartlett’s Test of Sphericity’ (Miller et al, 2005). These two tests advise the researcher whether or not factor analysis is an appropriate technique for categorising the underlying data set. The method of extraction used was the Principal Components Analysis with varimax rotation. A gross list of 14 items listed in the third column of Table 4 were analysed on a sample of 685 eligible patients. KMO equalled 0.767 that is in range of the acceptable boundaries {0.6; 1} (Tabachnick & Fidell, 2013). Bartlett’s Test of Sphericity tests the null hypothesis that the correlation matrix is an identity matrix, thus rejection is wanted. This test displayed a satisfactory P-value (P<0.001), since this value must be significant at α=0.05 (Nunnally & Bernstein, 1994). Determination of the optimal number of clusters can be estimated by Kaiser’s criterion, which relies on eigenvalues>1 (Kaiser, 1958). Kaiser’s criterion revealed a total of four clusters. It is important to acknowledge that Kaiser’s criterion is criticised for its unilateral threshold value for the eigenvalues. This means that an item with eigenvalue 1.01 is included, whereas another item with eigenvalue 0.99 is excluded. Hence, the scree plot was consulted for confirmation purposes. Again, four clusters were found to be the optimum number of clusters.

Appendix A2

Table A2

Table A2 Categorisation of actual waiting times