Methods: We conducted a cluster randomized controlled trial (RCT) of immediate feedback on 16 wards of a tertiary care hospital in Toronto, Canada. At baseline, all wards recieved monthly feedback on HHC. Units were randomized within pairs determined by baseline HHC over 1 year. Audits without feedback were performed 2X/week on both control and intervention wards. Additionally, 2 feedback audits were performed weekly on intervention wards. During feedback audits, the auditor provided feedback on HH events and missed HH opportunities. After 10 weeks, feedback was initiated on the control units, and stopped on the intervention units (i.e. crossed-over). A comparison of non-feedback audits between the intervention and control units was the primary outcome. Feedback audits were not included because we anticipated that there would be an increase HHC during feedback audits due to increased observation bias (i.e. Hawthorne effect). If immediate feedback resulted in sustained behavior change, improvements in HHC should be seen during the non-feedback audits on the intervention wards.
Results: Baseline HHC was 65% and 67% for the intervention and control wards. Overall, 14527 HH opportunities were observed, 9215 during the baseline period, 3161 during the intervention period, and 2151 during the cross-over period. HHC on the intervention wards did not change between the baseline and intervention phase (65% vs. 65%) while HHC on the control wards increased 2% (67% to 69%). When both feedback and non-feedback audits are included, a 6% increase was observed on the intervention wards vs. 2% on the control wards. This difference occurred because of a 6% increase in HHC on the intervention unit during the feedback vs. non-feedback audits (71% vs. 65%). Results from the crossover period were similar, with no substantial benefit to feedback seen but with persistent evidence of observation bias.
Conclusion: We failed to identify any increase in HHC following the initiation of immediate feedback in a cluster RCT. Compliance was 6% higher during feedback vs. non-feedback audits, likely due to an increase in observation bias.
M. P. Muller,
S. Junaid, None