Previous observational measures of healthcare worker (HCW) hand-hygiene behaviour (HHB) fail to provide adequate standard operating procedures (SOPs), accounts of inter-rater agreement testing or evidence of sensitivity to change. This study reports the development of an observational tool in a way that addresses these deficiencies. Observational categories were developed systematically, guided by a clinical guideline, previous measures and pilot hand-hygiene behaviour observations (HHOs). The measure, a simpler version of the Geneva tool, consists of HHOs (before and after low-risk, high-risk or unobserved contact), HHBs (soap, alcohol hand rub, no action, unknown), and type of HCW. Inter-observer agreement for each category was assessed by observation of 298 HHOs and HHBs by two independent observers on acute elderly and intensive care units. Raw agreement (%) and Kappa were 77% and 0.68 for HHB; 83% and 0.77 for HHO; and 90% and 0.77 for HCW. Inter-observer agreement for overall compliance of a group of HCWs was assessed by observation of 1191 HHOs and HHBs by two pairs of independent observers. Overall agreement was good (intraclass correlation coefficient = 0.79). Sensitivity to change was examined by autoregressive time-series modelling of longitudinal observations for 8 months on the intensive therapy unit during an Acinetobacter baumannii outbreak and subsequent strengthening of infection control measures. Sensitivity to change was demonstrated by a rise in compliance from 80 to 98% with an odds ratio of increased compliance of 7.00 (95% confidence interval: 4.02-12.2) P < 0.001.