Use of optimised dual statistical process control charts for early detection of surgical site infection outbreaks

Surgical site infections (SSI) are common healthcare-associated infections resulting in substantial morbidity, mortality and hospital costs.1–4 However, no standard algorithm for SSI surveillance or outbreak detection exists, and traditional surveillance techniques may fail to provide timely identification of important SSI rate increases.5 6 We previously showed that standard Shewhart and exponentially weighted moving average statistical process control (SPC) charts have potential to provide early detection of SSI outbreaks.7 We then performed a large-scale empirical optimisation study and determined that simultaneous use of two moving average (MA) SPC charts in this application was most effective in identifying clinically important increases in SSI rates, or SSI clusters, that occurred in our network of community hospitals.8 The objective of the current analysis was to evaluate the performance of this optimised combination of control charts when applied to known SSI outbreaks. We retrospectively applied an optimised pair of MA SPC charts8 to all 30 SSI outbreaks previously identified and investigated from 2007 to 2015 in the Duke Infection Control Outreach Network (DICON), a network of more than 50 community hospitals (online supplementary table).9 We used procedure-specific SSI data from either the entire network or the single outbreak hospital to calculate chart baselines, or expected SSI rates. The baseline window was the time period used to estimate the expected SSI rate on a rolling basis. The lag was the offset (in months) between each evaluated time point and the corresponding baseline window. The MA span was the number of monthly SSI rates that were averaged, weighted by their respective sample sizes, to calculate …