Observational Bias Within Hospital-Wide Hand Hygiene Program

Abstract

Background: Hand hygiene (HH) is critical to prevent hospital-acquired infections. Running a successful HH program requires valid and accurate HH data to monitor the status and progress of HH improvement efforts. HH data are frequently subject to variable forms of bias, for which considerations must be made to enhance the validity of HH data. Objective: We assessed the extent to which observers may be prone to report more favorable HH rates when observing healthcare workers from the same professional group versus members of other job categories. Methods: We analyzed HH data from 48,543 electronically collected observations conducted by frontline healthcare workers in a 793-bed acute-care hospital from January 1, 2019, through July 31, 2019. All auditors received training on HH observations and proper use of the data collection application. Compliance data were sorted into peer versus nonpeer observations by profession. We compared HH compliance rates for members of each professional group when monitoring peers versus nonpeers. We further stratified results by ancillary professions (central transport, unit associates, food services, pharmacy, phlebotomy, rehabilitation services, and respiratory therapy) versus nonancillary professions (doctors, nurses, physician assistants, patient care assistants). Results: Of 12,488 ancillary observations, 7,184 (57.5%) were peer observations and 36,055 were nonancillary observations, of which 15,942 (44.2%) were peer observations. The percentage of peer-to-peer observations versus nonpeer observations varied by profession, ranging from 96% of central transport workers and 91% of environmental services observations to 21% of patient care assistants and 34% of physician’s assistants. Average compliance rates for peer versus nonpeer observations in ancillary groups were 98% (95% CI, 98.7%–99.2%) versus 83% (95% CI, 82.5%–84.5%). Average compliance rates nonancillary groups were 92% (95% CI, 92.0%–92.8%) for peers versus 88% (95% CI, 87.8%–88.7%) for nonpeers (Table 1). Conclusions: We documented a propensity for some categories of healthcare workers to record discrepant rates of HH compliance when observing members of the same peer group versus others. This effect was more pronounced amongst ancillary versus nonancillary services. This study adds to the literature of potential sources of bias in HH monitoring programs. Operational changes in HH program data collection may be warranted to try to mitigate these biases such as increasing the frequency of validation exercises conducted by nonaffiliated observers, weighting peer versus nonpeer observations differently, or switching to automated electronic monitoring systems.

Funding: None

Disclosures: None