Improving Timely Discharges Through Nurse-Initiated Conditional Discharge Orders

The study took place in a pediatric tertiary care academic children’s hospital with 166 acute care beds. All pediatric patients admitted to a hospital medicine teaching service (intervention team) were targeted from July 2018 through March 2020. An additional hospital medicine team (control team) was used as the control group. The control team had a census cap of 2 additional patients. All clinical diagnoses were used in this study without restriction.

Hospitalists, hospitalist fellows, pediatric residents, family medicine residents, and medical students cared for all patients admitted to the hospital medicine teams and patients were admitted to multiple units throughout the hospital.

As the discharge process has been shown to be a target area for patient flow improvement,¹¹  we elected to use QI methodology to identify opportunities to increase discharge efficiency. A multidisciplinary committee was formed, including 2 hospital medicine physicians, a hospital medicine fellow, nurse coordinators, nurse unit managers, chief residents, hospital-wide patient flow committee members, and statisticians. A fishbone diagram was used to evaluate the process and a key driver diagram was developed (Fig 2). Knowledge, engagement, culture, and workflow were deemed the most important factors to address. It was determined that targeting education and culture change to effectively reinforce utilization of the EMR NICD order could improve patient flow.

In choosing this intervention, we considered the existing medical staff and nursing workflow, discharge process, and information technology infrastructure. The NICD requires anticipatory and preparatory work to be effective. Current resources already in place for such work included multidisciplinary discharge rounds for each care team after traditional work rounds. These structured sit-down rounds were led by the team senior resident and/or attending and included case management, social work, nutrition, and an appointment scheduler. In addition, the NICD order was an existing order available in the EMR. Utilizing bundled interventions to maximize education and awareness would improve the intended outcomes of increasing morning discharges. The NICD process allowed any patient who met all clinical criteria stated in the NICD order to be discharged at a provided time (usually before 10:00 am), unless otherwise directed by the medical team. If a patient did not meet the criteria stated in the NICD order, the bedside nurse would not proceed with the discharge and the medical team would round on the patient in the usual fashion. Lastly, all patients who met criteria to proceed as a NICD were seen prior to work rounds by the attending physician to have a billable encounter on the day of discharge. An algorithm was created to help illustrate this process as a resident and attending physician work tool (Fig 3).

The first bundled intervention included NICD process and EMR order education. A written tool containing clinically appropriate examples for NICD criteria and the logistics behind nursing, hospitalist, and resident workflow was developed and disseminated. The written tool included examples of common admitting diagnoses, such as asthma, gastroenteritis, and neonatal fever and examples of criteria that would be appropriate in a NICD order (Fig 1). The written tool also served as a “cheat sheet” with tips to optimize discharge workflow, including a discussion of conditional discharge eligibility on rounds, guidance for the actual EMR order with reference to the sample diagnoses as described, and expectations regarding the timing of discharge documentation and care coordination (eg, teams encouraged to complete these elements of the discharge process at least 24 hours in advance of discharge). Nurse coordinators disseminated and posted this information at nursing stations and the written tool was also posted in resident workrooms. Additionally, weekly e-mails containing this educational material and reminders to use NICDs were provided to faculty.

Nurse and resident educational modules were developed in our institutional training center to ensure appropriate education. The modules focused on the logistics of developing appropriate clinical, social, and system NICD criteria as well as the process and expectations for each multidisciplinary role. These modules were completed over the course of 1 month. A resident champion interested in patient flow research joined the team to help address resident concerns and resistance to the use of NICDs. Resident surveys were distributed and the results were analyzed.

Modifications to the process were made over time as necessary based on survey results and questions were answered. Additionally, weekly resident reminders and a NICD workflow algorithm were distributed by the resident champion to increase utilization. The weekly reminders emphasized the importance of NICDs in improving patient flow and included links to written educational tools. The algorithm offered a stepwise approach for residents to ensure a smooth NICD process: (1) identifying patients as candidates for NICDs, (2) creating the order in the EMR, and (3) completing the discharge workflow (Fig 3). The algorithm also provided guidance on what to do when criteria were not met or if there was a clinical event that prevented the patient from being discharged as planned.

The primary outcome measure was the percentage of morning NICDs (6:00 am–12:00 pm) per 24 hours. These data were pulled from the EMR monthly and calculated by dividing the number of NICDs from 6:00 am to 12:00 pm by the number of total discharges. The goal was to increase morning discharges by 5%. The secondary outcome measure was length of stay (LOS) for both the intervention team and the control team. This information was derived from the EMR using admission time and discharge time to calculate LOS in days to the 10^th decimal.

The primary process measure was the percentage of total NICD orders written for the intervention team and control team. This measure was pulled from the EMR every month and calculated by dividing the number of NICD orders written on each specific team by the total number of discharges for that team. This was evaluated using P-charts.

Additional process measures included rate of compliance with NICD education module completion by both resident and nursing staff. Lastly, a 15-question survey was disseminated to pediatric residents to elicit resident feedback regarding potential barriers to NICD order use as well as knowledge of use.

The balancing measure was evaluation of 30-day readmission data of patients who were discharged with a NICD order compared with those discharged in standard fashion.

The primary outcome and process measures were analyzed using statistical process control in Excel QI Macros and were represented by line charts. Special cause variation was defined as 6 or more consecutive points above the mean or points above the upper and lower control limits. As a secondary outcome measure, LOS was evaluated by comparing the intervention team LOS in days before and after the first PDSA using the paired t test. Similarly, LOS in the control team before and after the first PDSA was evaluated for comparison.

The readmission balancing measure was evaluated by χ² statistical test comparing 30- day readmission rates for discharges using the NICD order process versus standard discharges.

The primary outcome measure of percentage of morning NICDs revealed a baseline measure of 6% and eventual special cause variation in the intervention team (Fig 4), with a mean shift after PDSA 2.

The secondary outcome measure of LOS showed no significant difference in those that were discharged pre and postintervention for both control and intervention groups. Although the LOS for 649 discharges in the intervention group decreased by 0.62 days (4.5 days preintervention and 3.88 days postintervention) it did not reach statistical significance (P = .104) The LOS for 902 discharges in the control group increased by 0.16 days (3.59 days preintervention and 3.75 days postintervention (P = .522).

Baseline data revealed that NICD order percentage for the intervention team was 47% and the control team was 37% of discharges. Special cause variation in the NICD order process measure was observed for the intervention team after the first PDSA cycle with 6 points above the mean (Fig 5), but not for the control team. The significance was lost in November 2019, however, a reduction in variation from the mean continued to be observed.

The second process measure included 100% education module compliance for residents and nurses. Resident survey responses were collected and reviewed by the resident champion. The balancing measure of 30-day readmissions showed no significance in those that were discharged using a NICD order (6.01%) and those with the standard discharge process (5.06%, P = .20). There was no difference in readmissions between those with a NICD order and those with standard discharge.

Through multidisciplinary education, written tool dissemination, resident engagement, and weekly reminders to use NICD orders, there was significant improvement in morning discharge efficiency. Our study echoes the use of discharge criteria, education,^6,7  and improvement in physician workflow⁵  to create more timely discharges. Our study focused on early discharge preparation and a NICD order with specific criteria, helping to discharge patients safely and as early as medically ready. We found that we could effectively increase morning discharges by noon⁹  with the addition of a NICD order. This tool facilitates an early morning discharge without formal rounding by the resident team, allowing the supervising physician to have a brief, billable encounter prior to rounds. Further, increasing morning discharges likely allows earlier admissions from the ED, PICU transfers to the floor, and direct and surgical admissions.

There was no difference in 30-day readmissions, suggesting that conditional discharges were completed without adversely affecting clinical representation. Although not statistically significant, LOS was lower with the use of NICD orders, ensuring that discharges were expedited and not delayed until the following morning. This suggests that NICDs are a safe and effective way to improve patient flow in a large academic center.

We found that a resident champion and evaluation of resident concerns and feedback were crucial for improving patient flow and morning efficiency. Directly addressing these concerns and those of attending physicians and nurses helped to maintain buy-in to the value of the NICD order process. As a matter of system level thinking, it was vital to impress upon learners the importance of patient flow for the safety of patients and improved quality care, even if it meant more admissions and discharges per day for a given resident team.

Although there was an initial signal and shift in the percentage of NICD orders, the loss of signal may be because of high patient census and level of acuity associated with winter season diagnoses as well as holiday mixed coverage on the hospital medicine teams. High census creates a constant push for efficient and safe discharges. It is possible that because of this push, NICD orders were not written preemptively with an increased workload, but the discharge happened as soon as possible without the NICD process. Sustainability may have been further hampered by competing priorities and disincentives to efficient discharges given that a team will often receive a new admission immediately upon discharging a patient.

Barriers to NICD usage included concerns that patients may be prematurely discharged before all criteria are met. We had 2 of these patient encounters. One was discharged before their medications were delivered to bedside and education completed by the medical team and another before they received their final dose of medication. We learned that new nursing staff required continued NICD education. Additionally, team members were unsure what to do when conditions were not met as expected in the morning. To mitigate provider apprehension, we learned that providers could benefit from a NICD algorithm to help with decision making (Fig 3). Lastly, because of our large faculty pool, retaining consistent utilization of NICD orders was challenging and required frequent education and reminders.

This was a single center study and therefore, the results may have been impacted by institutional culture and practice. This study is best applicable to large pediatric academic centers working with resident teams and may not be generalizable to smaller hospitals. We did not correlate the percentage of NICD orders with the current hospital census directly, thereby missing the opportunity to potentially explain why discharge efficiency improved despite losing the effect of the percentage of NICD orders. Additionally, readmission data were collected from our institution alone. Patients may have sought care at other local and regional medical centers.

Future cycles include focus on sustainability via EMR enhancements to promote NICD discussions during rounds. For example, a note template including a NICD order prompt may increase process adoption. We intend to add NICD orders into admission order sets to encourage early and consistent use of this tool. The implementation of a NICD prompt and marker on physician sign-out may facilitate the use of NICD orders with handoffs. Given our observations, focusing on barriers to sustainability will be key to future study. Lastly, system level impact may also be an area for future study, such as the impact of morning discharge efficiency on ED boarder metrics and ED to bedtimes.