Impact of four training conditions on physician use of a web-based clinical decision support system.

Abstract

Training has been identified as an important barrier to implementation of clinical decision support systems (CDSSs), but little is known about the effectiveness of different training approaches. Using an observational retrospective cohort design, we examined the impact of four training conditions on physician use of a CDSS: (1) computer lab training with individualized follow-up (CL-FU) (n=40), (2) computer lab training without follow-up (CL) (n=177), (3) lecture demonstration (LD) (n=16), or (4) no training (NT) (n=134). Odds ratios of any use and ongoing use under training conditions were compared to no training over a 2-year follow-up period. CL-FU was associated with the highest percent of active users and odds for any use (90.0%, odds ratio (OR)=10.2, 95% confidence interval (CI): 3.2-32.9) and ongoing use (60.0%, OR=6.1 95% CI: 2.6-13.7), followed by CL (any use=81.4%, OR=5.3, CI: 2.9-9.6; ongoing use=28.8%, OR=1.7, 95% CI: 1.0-3.0). LD was not superior to no training (any use=47%, ongoing use=22.4%). Training format may have differential effects on initial and long-term follow-up of CDSSs use by physicians.