Search

Background Patient safety events result from failures in complex health care delivery processes. To ensure safety, teams must implement ways to identify events that occur in a nonrandom fashion and respond in a timely manner. To illustrate this, one children's hospital's experience with an outbreak of unplanned extubations (UEs) in the neonatal ICU (NICU) is described. Methods The quality improvement team measured UEs using three complementary data streams. Interventions to decrease the rate of UE were tested with success. Three statistical process control (SPC) charts (u-chart, g-chart, and an exponentially weighted moving average [EWMA] chart) were used for real-time monitoring. Results From July 2015 to May 2016, the UE rate was stable at 1.1 UE/100 ventilator days. In early June 2016, a cluster of UEs, including four events within one week, was observed. Two of three SPC charts showed special cause variation, although at different time points. The EWMA chart alerted the team more than two weeks earlier than the u-chart. Within days of discovering the outbreak, the team identified that the hospital had replaced the tape used to secure endotracheal tubes with a nearly identical product. After multiple tape products were tested over the next month, the team selected one that returned the system to a state of stability. Conclusion Ongoing monitoring using SPC charts allowed early detection and rapid mitigation of an outbreak of UEs in the NICU. This highlights the importance of continuous monitoring using tools such as SPC charts that can alert teams to both improvement and worsening of processes.

Background The key driver diagram (KDD) is an important tool used by improvement teams to guide and frame their work. Methods to build a KDD when little relevant literature or reliable local data exist are poorly described. This article describes the process used in our neonatal ICU (NICU) to build a KDD to decrease unplanned extubations (UE) in chronically ventilated infants. Methods Twenty-seven factors hypothesized to be associated with UE in our NICU were identified. An expert panel of 33 staff members completed three rounds of a modified Delphi process administered through an online interface. After the third round, panel members provided suggestions for interventions to target all factors meeting criteria for consensus. These qualitative data were analyzed by inductive thematic analysis. A follow-up survey to all panel members was used to assess the feasibility of this process for future use. Results After three Delphi rounds, 14 factors met consensus and eight main interventions were identified through thematic analysis. These data were used to build a KDD for testing. All participants who completed the follow-up survey (20/20) stated willingness to participate in this process in the future and 18/20 (90%) stated they would be "more willing" or "much more willing" to support interventions developed using this process. Conclusion A novel mixed-methods approach was used to generate a KDD combining a Delphi process with thematic analysis. This approach provides improvement teams a rigorous and reproducible method to understand local context, generate consensus KDDs, and improve local buy-in for improvement interventions.