A Multimodal Interdisciplinary QI Intervention Is Associated with Reduction in After Hours Inpatient Endoscopy Cases

Abstract

Background and Aims:

Increasing demand for inpatient endoscopic services results in performing more non-emergent endoscopic cases after-hours, which poses risks to patient safety and negatively impacts patient and provider satisfaction. This study sought to quantify the existing state using quality improvement (QI) methodology, design targeted interventions, and determine their effectiveness.

Methods:

We conducted an existing state evaluation through a process map, time-series study, and caseload analysis from 7/2017–12/2018. Using end-of-workday (EOW) as a proxy for patient/provider dissatisfaction and risk for patient safety events, we performed a prospective evaluation of a staged interdisciplinary multimodal intervention aimed to decrease the proportion of days with EOW after 7PM, decrease the proportion of cases begun after 5PM, and decrease EOW variability. The post-intervention period was 6/2019–2/2020.

Results:

Based on existing state analyses, we implemented a series of targeted interventions: (1) provider workflow tips, (2) expedited transport for select patients, (3) pathway to reschedule appropriate cases to outpatient endoscopy, and (4) increased staffing for high caseload days through resource pooling. The proportion of days with EOW after 7PM decreased from 42.4% to 29.3% (caseload-adjusted odds ratio of 0.39, p< 0.001). Despite increased caseload, cases begun after 5PM decreased from 17.5% to 14.2% (OR 0.75, p = 0.009). EOW SD decreased from 2:20 hours to 1:36 hours.

Conclusions:

The multimodal intervention reduced days with EOW after 7PM and the proportion of cases begun after 5PM, despite increased caseload. This study shows how applying research methods to implement QI interventions successfully decreases late inpatient endoscopic cases.

Introduction

The increase in breadth of conditions that can be diagnosed or managed endoscopically has fueled demand for endoscopic procedures in the inpatient setting, where patients are of higher acuity¹ and require same- or next day scheduling. At our institution, there is an increased burden of after-hour cases and days with late end-of-workday (EOW), when reduced staffing necessitate additional transitions in care. This is problematic as timeliness and efficiency are two of the six key dimensions in providing quality care per the National Academy of Medicine². As case volume increases, adaptation of the workflow processes is necessary to accommodate and optimize efficiency. Moreover, performing urgent, but nonemergent, inpatient endoscopic cases after hours pose risks to patient safety, negatively impacts patient and provider satisfaction, and potentially prolongs length of stay. Therefore, improving capacity for completing endoscopic cases during regular working hours by enhancing efficiency and appropriately allocating resources is a high priority to meet the demands of the hospitalized patient population.

There is a small but growing body of literature evaluating endoscopic unit efficiency, which provides metrics for outpatient endoscopy^3–6 and uses quality improvement (QI) tools like process maps, time series analyses, and quantification of delays to design site-specific interventions^7–14. Yet, there is scant literature on evaluating dedicated inpatient endoscopy settings, in which workflow is especially complex due to the need for rapid multidisciplinary care coordination and on-demand scheduling in high-risk patients. We provide a successful example of reproducible and rigorous QI methodology that can be adapted to enhance efficiency at other institutions.

The first objective of this initiative was to quantify the existing state and find targetable opportunities to improve workflow using QI tools. The second objective was to design, implement, and evaluate interventions aimed to improve the state of inpatient endoscopy. Using EOW as a proxy for patient/provider dissatisfaction and risk for patient safety events, we performed a prospective evaluation of a staged interdisciplinary multimodal intervention aimed to improve EOW.

Methods

This study followed the framework of a multidisciplinary QI project, with the key steps and methods outlined in table 1. We utilized multiple data sources (electronic data systems, stakeholder interviews, observation) to measure the existing state, analyze for root causes by evaluating delays, design a series of interventions, and measure their impact through a pre-post performance analysis.

Table 1:

Key steps in a QI project and related components for evaluating endoscopy workflow

Step	Method	Output
Define the problem and goal	Conduct stakeholder interviews	Project charter with problem statement, target outcomes, and team of stakeholders
Define the Process	Observe the process and key workers	Swimlane diagram
Measure the existing state	Analyze Epic and CWS data for caseload	Baseline performance analysis of caseload, EOW, and cycle time
Analyze for root causes	Observe and record frequent assignable delays Analyze CWS timestamp data	Pareto chart of delay causes Run chart Time series study
Identify targets and propose solutions	Use an impact effort matrix to compare interventions	List of interventions and implementation plans
Measure impact of interventions	Collect and analyze data	Updated performance analysis compared to baseline
Monitor for continued improvement	Continue to gather and analyze data	Monthly updates CWS datasets for continued performance analysis

Study setting and population

We conducted this study at the Hospital of the University of Pennsylvania, an urban tertiary-care academic hospital. All inpatient endoscopic procedures, excluding bedside cases are performed in two dedicated operating rooms (ORs) within the hospital surgical ORs. Only select high-risk outpatient procedures are performed in this setting; most outpatient cases are done in a separate outpatient endoscopy center. All endoscopic procedures are performed under monitored anesthesia care or general anesthesia. General OR resources are used for transport, preoperative assessment in a defined area (PRA), post-procedure recovery (PACU), anesthesia, and personnel after 7PM.

The endoscopy team is composed of the following personnel: a nurse anesthetist or anesthesiology resident present in the OR, a supervising anesthesiologist available within the OR suite, an attending gastroenterologist, an endoscopy registered nurse (RN), and an endoscopy technician. General gastroenterology and advanced endoscopy fellows participate in most procedures. There is one dedicated team assigned to the endoscopy unit per weekday from 7:30AM to 5PM that performs procedures consecutively. If additional personnel become available, a second team performs procedures simultaneously, where the second endoscopist is the GI attending on the hospital luminal consultation service. Endoscopic procedures performed after 5PM are performed by the on-call GI attending, and other team members are assigned as available from endoscopic and OR staff.

Ward patients are transported to the 42-bed PRA by the OR transport team. The PRA RN performs preoperative preparation and the gastroenterologist and anesthesiologist both perform a required history and physical and obtain written consent. The patient is transported to the OR when these tasks have been completed and the OR is ready to receive the patient. For patients from the intensive care unit or emergency department, or after the PRA closes at 7PM, patients come directly to the OR where the above activities are performed by the endoscopy team. After the procedure, the patient is transported to the PACU for postprocedural monitoring. The resource changes at 5PM (gastroenterologist and anesthesiologist) and 7PM (nurses, technicians) lengthen wait times for patients in queue. The on-call GI team stays to complete all scheduled cases for the day; cases are infrequently postponed until the next day based on staff availability, patient health factors, or patient preference.

Study design for existing state analyses

We conducted a retrospective analysis of all weekday inpatient endoscopy cases at HUP from 7/2017 through 9/2018 through the electronic medical record (EMR), Epic¹⁵. A case was defined as a single set of procedures for one patient in one day and in one setting. We evaluated the number of non-emergent cases performed and the EOW, defined by the endoscope-out time for the day’s last non-emergent case. Cases were considered emergent and excluded if started outside of work hours and scheduled within the prior 4 hours.

For all weekday inpatient cases from 10/2018–12/2018, we conducted a retrospective time series study to establish existing process times and identify sources of variability. Our hospital uses Clinical Workflow Suite (CWS)¹⁶ to track patient progression through care processes and hospital locations, with icons that update staff with readiness for the next step. Each step is time-stamped. Room turnover time is a key process measure in endoscopy⁶; we sought to categorize the time in the OR suite divided into pre-procedure preparation, procedure, post-procedure recovery, and empty.

We assessed workflow via direct observations to build a detailed swimlane diagram that followed a patient through the different areas (hospital room, PRA, OR, PACU), and divided tasks into categories of patient-pre-OR, empty OR, patient-in-OR preparation, procedure, and post-procedure. We conducted stakeholder interviews with representatives of all team members to understand their workflow and collect perceived common causes of delays. Study design of intervention evaluation

We implemented a series of targeted interventions to address the identified delays from 12/2018–5/2019 including: (S1) provider education and EMR prompts with workflow tips, (S2) expedited transport for select patients, (S3) pathway to offload appropriate cases to outpatient endoscopy, and (S4) increased staffing for high caseload days through resource pooling (Table 2). We used a statistical process control chart to assess the impact of post-intervention outcomes, including daily EOW and the number of cases performed after hours. Independent and dependent variables

Table 2:

Interventions implemented with identified root causes of delay they were designed to address.

Stage	Delay	Identified root cause	Implemented intervention
Stage 1	Results of required pregnancy tests	Standard urine test requires awaiting sample production. Must be coordinated with transport as cannot be processed in PRA. Times out after 48 hours, so may need to be repeated for cases postponed a few hours.	Switch to serum test to be drawn with AM lab draw the morning of the procedure
Stage 1	Transport for patients on telemetry	Patients on telemetry require specially-trained transports who do not start in time for the first case	EMR prompt for fellows to avoid patients on telemetry on the first case
Stage 1	Intermittent need for extra staff to position patients	High BMI, who also require a specific stretcher for the case. Earlier notice prompts coordination of staff	EMR prompt for fellows to alert OR team in the nightly case communication email
Stage 1	PRA RN awaiting endoscopy consent before signaling for transport to OR	Endoscopist not consistently signaling to PRA RN (through verbal and EMR methods) when consent is complete. Large pool of endoscopists who may not be aware of workflow.	Fellow in communication with PRA to alert charge RN directly.
Stage 2	Empty OR for direct-to-room transport	For direct-to-room transport, time waiting for transport from a hospital room is shifted to empty OR instead of PRA.	Offer expedited transport request for this subpopulation
Stage 3	Variable caseload often higher than existing state can accommodate	Some patients kept in the hospital for inpatient procedures when outpatient procedure scheduling is not available during the weekend	Reserve 3 outpatient endoscopy slots that fellows can schedule with clinically appropriate patients during the weekend
Stage 4	Variable caseload often higher than standard existing state can accommodate	Second endoscopist and OR available to perform simultaneous cases, but cannot predict need for other staffing resources when caseload not known until the prior afternoon	Increase pool of staffing resources to include non-endo OR staff (already in hospital for emergent cases), to be assigned the prior afternoon if caseload above 7

Our primary aim was to decrease the proportion of days with EOW after 7PM, with the EOW as the dependent variable. One secondary aim was to decrease the EOW variability, with the dependent variable being standard deviation of EOW. The primary independent variable for both aims is date. While daily caseload is a key outcome measure for outpatient endoscopy units⁶, caseload for our inpatient endoscopy unit was primarily an independent variable in the multivariate logistic regression in assessing EOW.

A secondary aim was to decrease the number and proportion of cases begun after 5PM. The dependent variable per case was the scope-in time. The independent variables were the date and daily caseload.

An added measure to assess the impact of S3’s intervention was to calculate the intervention’s impact on length of stay. For each patient that qualified for rescheduling to outpatient endoscopy, we documented the discharge date and compared it to what would have been the earliest discharge date based on the next available date of inpatient endoscopy. Statistical analysis

For a pre-post analysis, the pre-period (S0) was the retrospective existing state data 7/2017–12/2018 and post-intervention (S4) was 5/2019–2/2020. We categorized procedures as general versus advanced based on CPT codes (supplemental table 1); small bowel enteroscopies and EGDs with dilation were not categorized. Our data did not allow differentiation when more than one procedure was done during the case, therefore procedure time for individual procedures is not reported.

To evaluate the proportion of days with EOW after 7PM and proportion of cases before 5PM, we used chi-square test and logistic regression adjusted for caseload. All analyses including summary statistics were performed using STATA version 13.0 (College Station, TX)¹⁷.

This project was reviewed and determined to qualify as Quality Improvement by the University of Pennsylvania’s Institutional Review Board.

Results

Baseline state

During the 315 weekdays over the 15-month baseline period, the total number of cases was 2220, with a mean daily caseload of 7 (SD 2) and wide range from 2 to 17. 400 cases (18%) were started after 5PM. The mean EOW was 5:46PM (SD 2:23 hours: minutes), with a maximum of 1AM. EOW was after 5PM for 66% of days (206/315). There were no statistical differences in caseload or EOW by weekday due to wide overlapping ranges. Mean procedure time for advanced cases (0:46) was significantly longer than general cases (0:28) (p < 0.001).

Process and root cause analyses

The swimlane process map detailed the steps, providers, and different areas involved in completing an endoscopic case (supplemental figure 1). This tool was used in stakeholder conversations and interviews.

The time series study included 28 weekdays of OR consecutive cases. There were a total 176 procedures of which 45 were same-day add-on cases, and 49 were advanced cases. The total time from the patient leaving their hospital room to leaving the OR was 2:38 and 1:53, with and without PRA respectively. The mean times of each process step for patients who went to the PRA are in figure 1. The mean total OR cycle time, defined as the time of a patient’s OR arrival to the next patient’s OR arrival, was 1:33 and 1:56 with and without PRA respectively, where the latter included awaiting patient transport and arrival. The mean time spent on endoscopy per case was 0:26 (IQR 0:12–0:36), which was 26% of the total OR time (from 7:30AM to last case endoscope-out). The OR was unoccupied for 40% of total OR time, with an average time between cases of 0:34 minutes (SD 0:22).

Figure 1:

Processes involved in an endoscopic procedure. A) Time series study with processes for patients that went to the pre-op area (PRA) prior to the OR. Dashed lines indicate OR processes. B) Pie chart showing total accumulated OR time.

Targeting the empty OR time, observations of categorized delays and stakeholder interviews resulted in identifying the most frequent assignable, avoidable causes of delay: pending pregnancy tests, transportation delays due to patients on telemetry or patients needing to come directly to the OR, high patient BMI requiring more staff to move the patient, and missed communication of endoscopist consent completion in the PRA. Other assignable causes (like misfunctioning intravenous lines and patient’s bathroom needs) were deemed impractical to intervene upon systematically within the scope of this initiative.

Intervention Development

Our root cause analysis, with consideration of intervention feasibility and stakeholder buy-in, drove intervention efforts (Table 2).

Stage 1 involved addressing specific assignable causes that caused delays at higher frequencies. For required pregnancy tests, we switched from urine to serum specimens, where serum studies can be systematically performed in the morning, eliminating wait time, and are of similar cost. When scheduling the next day’s case list, a new standardized method in the EMR alerted fellows to patients with telemetry and/or on high BMI. We worked with PRA staff to improve signaling for endoscopist consent completion.

Stage 2 targeted direct-to-OR transport, in which the OR must be otherwise ready and empty for the entire transport time, causing a systematic delay. Transport to the OR took a mean of 0:42 and our unit averaged 0.89 direct-to-room cases per day. The intervention, in coordination with OR transport, was to offer expedited transport requests for this subpopulation for up to 2 cases a day.

Stage 3 was to convert inpatient cases to outpatient when appropriate. We reserved 3 weekly outpatient slots in our ambulatory endoscopy center that could be filled with low-risk inpatients to be discharged over the weekend. Under these scenarios, the patient avoided hospitalization days awaiting inpatient endoscopy and the Monday inpatient caseload was reduced.

Stage 4 was expanding the pool of staffing resources to include non-endoscopy OR staff; when the anticipated caseload for the next day was greater than 7 cases, OR staff was assigned to staff a second OR and complete cases simultaneously.

Post-intervention analysis

Caseload and case-mix:

Between S0 and S4, the number of days were 360 and 204 respectively and the mean daily caseload increased from 7.09 to 7.83 (p < 0.001). Advanced endoscopic cases continued to have a significantly longer mean procedure time than general cases with no difference between the time periods (general: 0:28 vs. 0:27, p=0.75; advanced: 0:46 for both p=0.91). There was a non-significant increase in the proportion of advanced cases from 26.8% to 27.2% (p = 0.07) and non-statistically significant decrease in proportion of advanced cases being completed before 5PM (94.0% to 92.8%, p = 0.22).

Late EOW and after-hour cases:

The proportion of days with EOW after 7PM decreased from 42.4% to 29.3%, with an unadjusted odds ratio of 0.63 (p = 0.045) and caseload-adjusted odds ratio of 0.39 (p< 0.001). Figure 2a shows a trending increase in the number of cases completed per month, with the mean number of cases being completed before 5PM per month in S4 (mean 144.4, SD 13.6) comparable to the total monthly caseload in S0 (mean 150.0, SD 14.8). There was also a trend for increasing days with EOW after 7PM during S0; that trend was reversed in S4 (figure 2b). Despite an increasing daily caseload, the percentage of cases begun > 5PM decreased from 17.5% to 14.2% (OR 0.75, p = 0.002). EOW SD decreased from 02:20 to 1:36 (figure 3).

Figure 2:

Trend in proportion of cases before 5PM and EOW before 7PM. Intervention period is marked with dotted lines. A) Number of cases per month done before and after 5PM. B) Percentage of days per quarter with EOW before 7PM, with fit lines from linear regression for pre- and post-intervention quarters.

Figure 3:

Statistical Process Control chart of daily end of workday (EOW). Hours after midnight are represented at numbers above 24. The circled numbers indicate the stages of intervention. The dashed lines indicate 3 SD above and below the mean for each stage.

The mean number of cases completed per day before 5PM was 6.7 in S0 and improved to 7.6 in the S4 (p < 0.001). There were fewer outliers in S4 for very late days even at higher caseloads (Figure 4a). When comparing days with similar caseloads, there was a narrowing of interquartile ranges for caseloads of 6 through 8 (Figure 4b).

Figure 4:

EOW per number of cases. A) Each scatter point represents a single workday. B) Box plot comparing baseline to post-intervention based on caseload.

Shift of lower acuity cases to the outpatient setting:

In the 10 months of intervention stage 3, 20 patients were scheduled for outpatient reserved slots, of which 18 were completed on the date scheduled (90%) and the other two were rescheduled. Comparison of true and expected discharge dates yielded a total difference of 38 hospital days during this period, which extrapolates to 46 hospital days per calendar year.

Discussion

There is great need for established methods in developing efficient inpatient endoscopic units. The efficiency of endoscopy units has been a topic of study since at least 2002⁴, interest increasing with the growing endoscopic caseload. Analysis methods vary; some groups look at eliminating aberrations from the intended process by focusing on frequent causes of delays, like the unavailability of team members^7,8, while others aimed to improve elements of the process itself, like nurse handoffs and documentation processes¹¹ and timing of intravenous line placement⁹. In this study, multiple points of intervention allowed for addressing both types of issues causing late EOWs. To our knowledge, this is the first in-depth study demonstrating the use of multiple QI methods to evaluate and improve endoscopy in an inpatient setting.

Due to inefficient existing state workflow, the OR was empty for a higher percentage of time than performing endoscopic procedures. The swimlane diagram and time series analysis allowed us to identify root causes and address specific causes for empty OR time. Stage 1 interventions targeted assignable causes with low-investment, quickly implementable solutions. Stage 2 targeted the systematic delay for direct-to-room cases, which was only 13.9% of cases but impacted 89% of days. Inpatient transport was noted as a cause of delays in another tertiary institution¹⁴, suggesting other sites may share this issue. Stage 3 offered a formal system to discharge clinically appropriate patients to scheduled outpatient endoscopy instead of waiting through a weekend for an inpatient procedure. The small number of cases rescheduled resulted in a significant number of hospital days saved. Inpatient status is associated with inadequate bowel preparation¹⁸, which leads to longer length of stay and costs¹⁹, offering another reason to offer expedited outpatient endoscopy. Finally, the stage 4 intervention addressing caseload variability required a structural change that allowed flexibility in staffing resources. The solution took advantage of OR resources that remained otherwise available for emergent surgeries.

Our analysis revealed other intervention targets. A large structural intervention would be to incorporate the perioperative assessments into the endoscopy suite, where processes may be better tailored for short endoscopic cases instead of the current pan-OR PRA. Our data is being evaluated to support this larger intervention.

Compared to reported procedure lengths for ambulatory endoscopy units^12,13 our data may be more representative of inpatient caseload, where there is a high proportion of advanced cases and likely more complicated cases due to procedure and patient factors. Still, this study concurs with prior studies that the procedure time is not the most frequent cause of delay and better targets for intervention are related to before and after the procedure itself^12,13.

A major strength of our study was in the multitude of ways in which we were able to analyze the existing state. Electronic data from Epic and CWS allowed the collection of detailed timestamps in order to do the time series analysis and track changes in caseload. Our data collection methods are now automated and set up for further monitoring of caseload and EOW. Another strength was the multidisciplinary nature of our intervention design. Inpatient endoscopy involves many divisions and interventions required consensus and multidisciplinary stakeholder buy-in among non-GI hospital groups, including peri-operative services and the transport team. Our ability to continuously collect, update, and present data to institutional leaders allowed for increasing stakeholder buy-in. This subsequently led to further investment in resource allocation, such as increased staffing, which cumulatively led to incremental improvements in our process metrics. Our study also has several limitations. Due to the nature of this QI project, we can only assess pre- and post- periods in a quasi-experimental study design, and adjust for simultaneous changes in caseload. The COVID-19 pandemic caused multiple process changes, which terminated the post-intervention period, though our data suggest it was sufficiently long to identify significant impacts of interventions and sustainability. Additionally, it reflects the endoscopy unit structure of an urban academic hospital that may not translate to other settings.

Objective data can be used to champion system modifications to enhance proficiency and quality of inpatient endoscopic services. We outline here the QI tools used to evaluate our inpatient endoscopy process and causes of late cases, and study the impact of our multimodal intervention. Despite increases in caseload and advanced procedures, our multimodal intervention successfully decreased the proportion of days with EOW after 7PM, decreased the proportion of cases begun after 5PM, and reduced EOW variability. This study demonstrates how QI methods can be used to design and implement interventions that successfully decrease late inpatient endoscopic procedures and increase the capacity to match growing demand for inpatient endoscopy.

Supplementary Material

1

Supplemental figure 1: Swimlane diagram of the workflow of how a patient gets an inpatient endoscopic procedure. Swimlane rows indicate different locations involved. Columns indicate stages of the process. The blue background follows the patient’s location through the process. The colors of each process indicate the team member(s) managing the task, as listed in the key.

What You Need to Know.

Title:

A Multimodal Interdisciplinary QI Intervention Is Associated with Reduction in After Hours Inpatient Endoscopy Cases

Background:

There is great need for established methods in developing efficient inpatient endoscopic units, especially with increasing demand for inpatient endoscopic services.

Findings:

Our multimodal intervention successful decreased the proportion of days with end-of-workday after 7PM, decreased the proportion of cases begun after 5PM and reduced endo-of-workday variability, despite increased caseload.

Implications for patient care:

We provide a successful example of reproducible and rigorous QI methodology that can be adapted to implement interventions and enhance inpatient endoscopy efficiency.