Implementation science-guided rollout of a standalone radiology clinical decision support system in a tertiary emergency department: findings from a pragmatic controlled before-after study

Abstract

Radiological clinical decision support systems (CDSS) promote value-based healthcare and guide appropriate imaging; however, acceptance and uptake remain inconsistent. Suboptimal implementation can limit the effectiveness of CDSS, widening the research-practice gap. This study examined whether applying implementation science principles for standalone CDSS implementation influenced imaging appropriateness in the emergency department (ED). A pragmatic controlled before-and-after study was conducted. A purposive sample of junior ED physicians was recruited from the largest acute tertiary hospital in Singapore. Implementation science principles from a process model were incorporated in the implementation of the standalone CDSS. Appropriateness rates of X-ray cervical and lumbar spine were examined before and after “Do nothing” and availability of CDSS, respectively with reference to institutional imaging referral guidelines. Radiologically significant positive rates of imaging performed were also explored. Wilcoxon signed-rank tests, Mann-Whitney U tests and multilevel logistic regression models were performed. A total of 17 and 23 ED physicians participated in the control and intervention groups, respectively, yielding 593 and 755 imaging procedures for each group. Among physicians in the intervention group, CDSS adoption was 47.8% ( n  = 11/23). There were no statistically significant within-group and between-group changes in appropriateness rates. Patient age was strongly associated with appropriateness rate with each additional year of age increasing the appropriateness rate by 8% (OR = 1.08, 95% CI: 1.07–1.09, p  < 0.001). Multilevel logistic regression revealed a strong association between appropriateness of imaging and positive radiological findings. Patients with appropriate imaging had higher odds of positive findings compared to inappropriate imaging patients (OR = 21.55, 95% CI: 2.88–161.3, p  < 0.05), although the estimate was imprecise due to sparse data. Availability of CDSS in the ED did not significantly improve appropriateness rates of cervical and lumbar spine X-rays, suggesting limited practice change from implementing standalone CDSS. There were underlying issues, weaknesses, and gaps in the implementation strategies that constrained the potential of CDSS and warranted attention to optimise its effect on appropriate imaging. Not applicable.