• Nabeel Imam, Tim Spelman, Sandra A Johnson, and Leon J Worth.
• Victorian Healthcare Associated Infection Surveillance System (VICNISS) Coordinating Centre, The Peter Doherty Institute for Infection and Immunity, Melbourne, Victoria, Australia.
• Qual Manag Health Care. 2019 Jan 1; 28 (1): 39-44.

BackgroundStaphylococcus aureus bloodstream infection (SAB) in health care settings contributes significantly to mortality, and improved processes are associated with reduced burden of infection. In Australia, health care-associated SAB (HA-SAB) rates are reported as a health care performance indicator, but standardized methods for analyzing longitudinal data are not applied. Our objective was to evaluate the utility of statistical process control chart methodology for reporting HA-SAB and flagging higher than expected rates.MethodsA real-world test data set was defined as HA-SAB surveillance data collected by 155 Australian health care facilities between June 1, 2015, and June 30, 2017. This included 788 HA-SAB events, corresponding to an overall rate of 0.7 HA-SAB events per 10 000 occupied bed-days. The u-chart was selected as an appropriate tool, given the need for reporting natural units (HA-SAB rates) to a range of stakeholders. Facility-level data were plotted as u-charts, applying warning and control limits (2- and 3-SD thresholds, respectively).ResultsSixty-eight of the 155 participating facilities (43.9%) observed at least 1 HA-SAB event during the studied period. Using the traditional method of Poisson modeling, 56 of these 68 facilities demonstrated overdispersion with variance-to-mean ratio spanning 1.03 to 42.82. Modeling by negative binomial (NB) distribution was therefore applied to enhance functionality.ConclusionThe u-chart is an accessible method for monitoring HA-SAB, interpretable by a range of stakeholders. We demonstrate the benefit of NB modeling to account for overdispersion, providing an effective tool to avoid inappropriate flags while maintaining early detection of out-of-control systems throughout a wide range of health care settings.

16 keywords...

hide…