A Model to Estimate Hand Hygiene Compliance Using Administrative Data and Electronic Counter Devices

Background: The availability of a reliable but simple and low cost method to measure hand hygiene (HH) compliance prospectively in health care services is essential to allow for the assessment of quality of care, to promote performance improvement, to improve infrastructure design, and to evaluate interventions aiming to increase compliance.

Objective: 1) To correlate an easy and simple indirect method of measuring HH and with the direct observation method (DO) (reference standard) during routine care in an NICU. 2) To build a model to estimate HH compliance without the use of DO.

Methods:  Measurement through DO of HH among health care workers and visitors during 225 periods of 1 hour. Simultaneous measurement of HH using electronic counter devices (EC) installed inside the alcohol gel and soap dispensers. Assessment of correlation between the 2 methods using dispersion diagram and regression. A linear multivariate model was built to predict HH opportunities using the DO as dependent variable and the following independent variables: number of beds occupied, number of nurses, number of other health professionals, and number of visitors. Then, correlation between DO and EC, and the coefficient of each independent variable was used to develop a method to estimate HH compliance prospectively not needing DO.

Results: 7,324 HH opportunities and 3,677 HH actions were observed during the 255 hours (from December 2008 to March 2009). The EC documented a total of 4,898 product utilizations for HH for the same period. The regression of the dispersion diagram was a linear curve: #DO = 0.73 x #EC + 0. Correlation was highly significant (Pearson 0.910; Spearman 0.911; p = 0.000). The multivariate model arrived to the following result: total of HH opportunities i.e. 7,324 = 3.11 + (0.21 x # beds occupied) + (0.37 x # nurses) + (0.28 x # other health professionals) + (0,14 x # visitors). Time consumption was 1 hour for DO compared to 3 minutes for EC in each period.

Conclusions:  This model allowed us to estimate opportunities of HH using routine administrative data (beds occupied, nurses, other HCW and visitors). Due to the strong correlation observed, we may replace DO with EC and apply the number obtained to estimate the HH compliance over time by the following formula: % of compliance = # HH (through EC) / # of estimated opportunities. An unexpected finding was an EC count higher than that obtained through DO that was possibly due to human factors while measuring HH through DO. We developed a tool to estimate HH compliance prospectively with a high internal validity. Other facilities may use the same method to build their own compliance estimates as an alternative to DO. From time to time there may be a need to repeat the method to adjust the coefficients due to internal changes.