Quality of chest compressions (CC) is an important determinant of resuscitation outcome for cardiac arrest patients.To characterize the quality of CC performed by hospital personnel, evaluate for predictors of CC performance, and determine the effects of audiovisual feedback on CC performance.Seven hundred and fifty four individuals participated in a CPR quality improvement challenge at 30 US hospitals. Participants performed 2min of CC on a manikin with an accelerometer-based system for measuring both rate (CC/min) and depth (in.) of CC (AED Plus:ZOLL Medical). Real-time audiovisual feedback was disabled. A subset of participants performed a second trial of CC with the audiovisual feedback prompts activated.Mean depth of CC was below AHA minimum guidelines (<1.5in.) for 34% (1.30±0.14in.) and above maximum guidelines (>2.0in.) for 12% of participants (2.20±0.22in.). Depth of CC was greater for male vs. female (p<0.001) and younger vs. older (p=0.009) but did not differ between ACLS, BCLS, and non-certified participants (p=0.6). Predictors of CC depth included CC rate (r part =−0.34, p<0.0001), gender (r part =0.13, p=0.001), and age (r part =−0.09, p=0.02).Mean depth of CC increased, mean rate decreased, and variance in CC depth and rate declined when feedback was used (p≤0.0001 vs. without feedback). The percentage of CC performed within AHA guidelines (1.5–2in.) improved from 15 to 78% with feedback.The quality of CC performed by personnel at US hospitals as judged by their performance on a manikin is often suboptimal. Quality of CC can be improved with use of CPR feedback technologies.