Methods

The number of CAUTIs per year was used as an outcome measure, and the device utilization rate was used as a process measure, which is reported as number of IUC days per 100 patient days. To establish normative values for both metrics, the available historical data were reviewed, which amounted to 7 years (137,051 patient days). The investigation revealed a historical CAUTI rate of 2 CAUTIs per year (1.67 CAUTIs per 10,000 patient days) and a historical device utilization rate of 2.94. In FY2021, the CAUTI rate was 4 CAUTIs per year (2.15 CAUTIs per 10,000 patient days) and the device utilization rate was 3.21. The average device utilization rate was 3.99 for months in which a CAUTI was identified, and 2.73 for months in which no CAUTI was identified. We used these values as upper and lower control limits respectively on a chart (Fig. 1), which was updated and reviewed monthly. To ensure that the intervention was not leading to patient harm, all patients who required transfer to an acute-care hospital were logged and reviewed for complications relating to IUC management.

Fig. 1.

The chart used to monitor the success of the intervention. The upper control limit was set at 3.99 IUC-days per 100 patient days. The Lower control limit was set at 2.73 IUC days per 100 patient days. Mean is calculated for preintervention months and postintervention months. Note. CAUTIs, catheter-associated urinary tract infections; IUC, indwelling urinary catheter; PDSA, plan–do–study–act.

A quality improvement (QI) team was formed with members from key stakeholder groups. Overall, 10 clinical staff were interviewed; 9 interviewees identified the lack of a standardized process for IUC removal as a significant barrier. Accordingly, the QI team created a standardized protocol outlining criteria to initiate a void trial, with IUC removal occurring on hospital day 1 (Supplementary Fig. 1 online). The protocol was initially designed as a “nurse-driven protocol,” allowing nurses the autonomy to remove an IUC without a physician order, thereby streamlining the process. Members of the QI team attended nursing staff meetings for 2 consecutive weeks and answered questions pertaining to the protocol. Additionally, the protocol was presented to clinicians, and printed versions were placed in all clinical areas. Patients with spinal cord injuries were initially excluded from the protocol due to the intricacies of managing a neurogenic bladder.

After implementation, plan–do–study–act (PDSA) cycles were planned quarterly or sooner if the device utilization rate trended above the upper control limit. In PDSA cycle 1, after implementation of the protocol, the device utilization rate decreased and remained below the upper control limit throughout the first quarter, with zero reported CAUTIs. After interviewing stakeholders, it became apparent that having a nurse-driven protocol was inefficient because providers were still being contacted for an order prior to removal of the IUC. As a result of this feedback, the protocol was modified to be physician driven. Although the protocol was not being executed exactly as planned, it still appeared to be successful at reducing the device utilization rate. This improvement prompted the QI team to expand the scope of the project to include patients with spinal cord injuries.

In PDSA cycle 2, stakeholder interviews revealed that providers were forgetting to use the protocol on admission, which was leading to minor delays in initiating the void trial. To address this problem, the protocol was embedded in the standard history and physical (H&P) note template that is used for all admissions. This change forced providers to interact with it on the day of admission.

In PDSA cycle 3, the feedback from stakeholders was positive, and the protocol was being effectively used with most patients who met the inclusion criteria. The decision was made to discontinue scheduled PDSA cycles and to reconvene only if the device utilization rate trended above the upper control limit.

All data were analyzed using Microsoft Excel (Microsoft, Redmond, WA). Comparisons were done using t tests when able. When data were heavily weighted with zero, no P values were calculated. P < .05 was considered statistically significant.