Investing in a New Role to Increase Timely Morning Discharges in the Inpatient Setting

Delays in patient discharge from pediatric hospitals are a common and costly problem.^1,2 Discharge delays from acute care floors have a negative downstream effect on emergency department (ED), intensive care unit (ICU), and elective admission throughput. This increased strain on hospital capacity may worsen patient outcomes.³ 

Delays in discharge are multifactorial, including staff workflow, hospital culture, practice variability, lack of planning, and social factors. However, past studies suggest delays in physician orders and discharge paperwork are leading causes.⁴ Past quality interventions have shown success with high team engagement in discharge planning, establishment of standardized discharge criteria, use of checklists, changes to electronic medical records (EMRs), and adjustment in staff workflow.^4–9 

Our project SMART aim was to increase the number of discharges before noon for a hospital medicine service from 19% to 30% over a 6-month period and sustain the increase for 6 months.

This study took place from January to December 2023 at Children’s National Hospital in Washington, DC, a tertiary care academic children’s hospital with 180 acute care beds. Of the 5 hospital medicine teams, 1 team, accounting for 40% of all hospital medicine discharges, was targeted for the intervention. This team consists of an attending physician, a second-year resident, and medical students. This team has a cap of 14 to 16 patients, depending on hospital census, and consists of patients with noncomplex admissions (Supplemental Figure 1). All patients admitted to this team were included in the intervention without clinical diagnosis restriction. The primary metric of discharge before noon was chosen to align with an ongoing organizational goal at our institution, as well as cited literature, to improve throughput and patient flow.¹⁰ 

Quality improvement (QI) methodologies were used to identify areas to increase discharge efficiency. Residents collected daily data for 4 weeks, categorizing why patients were not discharged before noon on the intervention team. A Pareto chart was created, demonstrating that most discharges after noon were due to clinical condition, late consultation recommendations, awaiting patient/family education, medication delays, and transportation delays (Supplemental Figure 2). A Gemba walk was performed for both physician and nursing workflow to identify additional barriers to discharge. A multidisciplinary team including residents, hospital medicine fellows, attending physicians, nurses, and case managers created a key driver diagram (Figure 1).

To address the primary drivers of physician workflow and hospital culture in a time of high hospital census, it was determined that additional resources were needed to improve patient flow. The first intervention consisted of creating a new role entitled patient flow provider (PFP) (Supplemental Figure 3). The PFP role focused on assisting the team with discharge tasks while the remainder of the team continued with other patient care activities. One advanced practice provider (APP) was hired for this position and began January 10, 2023. The PFP was initially hired for 0.5 full-time equivalent (FTE) and scheduled for 6 hours, 3.5 days per week. On July 24, 2023, the PFP role was expanded to 1.0 FTE, which included 8 hours Monday through Friday.

The second intervention was focused on optimizing the PFP workflow. Attention was paid to safe patient handoffs, clear delineation of tasks, closed-loop communication, interdisciplinary collaboration, and equity. For example, the team performed rounds first with all Spanish-speaking patients with an in-person Spanish interpreter. The resultant PFP workflow diagram was reviewed with hospital staff and posted in workrooms and electronically on February 14, 2023 (Supplemental Figure 4).

The third intervention focused on leveraging technology to improve discharge efficiency. Previous workflow included providers writing free-text discharge instructions for each patient and attaching standardized educational material. Addition of standard discharge instruction dot phrases (late April) allowed for quicker EMR discharge completion and ensured PFP clinical accuracy. Discharge delays in pharmacy prescribing were identified for uninsured patients, which required medications to be sent by the attending physician to qualify for hospital discount medications. Partnership with information technology and pharmacy expanded prescribing options to residents and all licensed independent practitioners, including the PFP in early May.

To address hospital culture, the study team disseminated information highlighting the impact providers have on patient flow and best practices for efficient discharges. A PowerPoint was distributed to all inpatient physician, nursing, and case management staff, which was accompanied by leadership review at unit and departmental meetings. This increased the visibility of the PFP role and reinforced early discharge planning to help ensure sustainability.

The primary outcome measure was percentage of patients discharged before noon. The secondary outcome measure was hospital length of stay (LOS) for patients on this hospital medicine team. LOS was defined as time from inpatient registration to time when departed from the electronic health system system, which generally reflects the time the patient spent on the inpatient floor. The process measure was difference in percentage of patients discharged before noon on days with the PFP present compared with days without the PFP present. The balancing measure was number of patients discharged between 7-day readmission.

Data collection was performed using a Power BI dashboard linked to the EMR data warehouse (Cerner HealtheIntent). Demographic and clinical data included race and ethnicity, language, sex, weekday of discharge, and daily census of the team. Race and ethnicity, language, and sex were chosen based on which of the REL-plus variables were available within our baseline dataset.¹¹ Each of these variables has also been highlighted as important to consider when developing an equitable discharge process.¹² The social constructs of race and ethnicity were self-reported in the EMR as 2 separate categories. However, 89% of patients who identified as Hispanic ethnicity selected “other race” for race. Given recent guidance from the US Office of Management and Budget on the reporting of race and ethnicity data, race and ethnicity were combined into 1 variable to more accurately represent patients who identify as Hispanic. The combined race and ethnicity variable included non-Hispanic white (white), non-Hispanic Black (Black), Hispanic, other race and ethnicity, and missing race and ethnicity. Daily census data were categorized as high census (>10 patients) vs normal census (≤10 patients). Data were analyzed biweekly using Microsoft Excel QI Macros. STATA was used for the regression analysis.

Statistical process control (SPC) charts utilizing Provost and Murray’s rules were used to analyze the QI interventions.¹³ A p-chart was used for the primary outcome measure of patients discharged before noon. A p-chart with rational subgrouping to compare patients discharged before noon on days with PFP vs without PFP present was used for the process measure. Additionally, p-charts with rational subgrouping were also used to evaluate the equity of the intervention by race and ethnicity, language, and sex. An Xbar and s chart was created for LOS. A g-chart was used for our balancing measure of 7-day readmissions because they were rare events.

SPC charts were analyzed for special cause variation, and the centerline and control limits were revised when special cause variation was demonstrated and sustained for 12 or more data points.¹³ Due to the sample size of the race and ethnicity, language, and sex subgroups, these analyses were performed monthly rather than biweekly.

A multivariate logistic regression was performed for the primary outcome using a dependent variable of discharge before noon, an independent variable of PFP presence on the day of discharge, and control variables of sex, race and ethnicity, language, daily census, fixed effects for day of the week and month, and an interaction term for PFP presence and daily census. An equity analysis repeated this regression including an interaction term first for race and ethnicity and PFP presence, then for language and PFP presence, and finally for sex and PFP presence.

This study was reviewed by the institutional review board and deemed to be a QI study. Patient identifiers were collected within the Power BI Dashboard stored on the hospital network and subsequently removed when data were aggregated and analyzed.

A total of 4148 patients were discharged during the study period: 2006 patients during the baseline period and 2142 patients during the intervention period. During the intervention period, 1222 patients were discharged on days when the PFP was present. Fifty-seven percent of patients identified as male. Fifteen percent identified as white, 43% identified as Black, 28% identified as Hispanic, 9% identified as other or multiple races, and 5% did not have race or ethnicity documented. Twenty-two percent of patients used a language other than English (LOE). The average team census was 12.3 patients (±2.4 patients).

Overall, 27% of patients discharged before noon. The p-chart for discharge before noon demonstrated special cause variation after the implementation of the PFP role and was sustained over 12 months. This resulted in a centerline shift from baseline of 19% to 34% discharges before noon after implementation of the PFP role (Figure 2).

Average LOS remained stable from 2.74 days (±1.98 days) during the baseline period to 2.54 days (±1.87 days) during the intervention period. An Xbar and s chart for LOS showed no sustained special cause variation throughout the study period (Supplemental Figure 7). There were 7 points of special cause variation during the baseline period and 1 point of special cause variation during the intervention period. There was a cluster of 5 points of special cause variation with increased LOS between September 24 and December 2, 2022, possibly related to high hospital census during a respiratory viral surge, but no sustained change.

On days without the PFP present, an average of 22% of discharges occurred before noon, similar to the baseline of 19%. On days with the PFP present, 43% of discharges occurred before noon (Figure 3).

The p-chart with rational subgrouping comparing discharge before noon based on race and ethnicity demonstrated an initial difference, with white patients’ discharge before noon at 27%, Black patients’ at 17%, and Hispanic patients’ at 16%. All groups’ discharge before noon improved by 14% to 17% after the interventions, with white patients’ discharge before noon at 41%, Black patients’ at 31%, and Hispanic patients’ at 33% (Figure 4).

The p-chart with rational subgrouping comparing discharge before noon based on language demonstrated an initial difference, with English-speaking patients’ discharge before noon at 20% compared with 13% for patients who used an LOE. Both groups demonstrated improvement after the interventions, with discharge before noon at 35% for English-speaking patients and 30% for patients who used an LOE (Supplemental Figure 5).

The p-chart with rational subgrouping comparing discharge before noon based on sex demonstrated a similar initial discharge before noon: 18% for female patients and 19% for male patients. Both groups demonstrated improvement after the interventions, with discharge before noon at 31% for female patients and 36% for male patients (Supplemental Figure 6).

A multivariate logistic regression demonstrated that discharge before noon was more likely when the PFP was present (odds ratio [OR], 2.11; 95% CI, 1.42–3.12; P < .001) (Table 1). In addition, discharge before noon was more likely when the patient was male (OR, 1.21; 95% CI, 1.04–1.41; P = .014). Discharge before noon was less likely when the team was in high census (OR, 0.75; 95% CI, 0.59–0.95; P = .015), the patient used an LOE (OR, 0.60; 95% CI, 0.47–0.77; P < .001), or the patient identified as Black (OR, 0.55; 95% CI, 0.44–0.68; P < .001). The equity analysis including interaction terms for race and ethnicity and PFP presence (Supplemental Table 1), language and PFP presence (Supplemental Table 2), and sex and PFP presence (Supplemental Table 3) did not find any significant interactions, indicating that the PFP increased discharge before noon approximately equally for all groups.

Seven-day readmissions were rare events occurring on average every 184 patients. The g-chart of the balancing measure demonstrated no significant change in the number of patients between 7-day readmissions after the interventions (Supplemental Figure 8).

In this QI initiative, discharges before noon on a hospital medicine team increased from 19% to 34%, exceeding the initial target. The largest driver of improvement appeared to be the addition of the PFP. LOS did not change throughout the study period. Rational subgrouping and the regression analysis demonstrated an initial disparity with Black patients and patients who used an LOE discharging before noon less often than white and English-speaking patients, which persisted despite the interventions of this study. No changes in 7-day readmission rate were observed over the 12-month study period.

Past studies of incorporation of APPs in the inpatient setting demonstrated reductions in LOS and improvement in discharge efficiency.^14,15 This study adds to the evidence supporting the thoughtful use of APPs in the inpatient pediatric setting. At our hospital, an organizational goal relating to postpandemic financial recovery created a willingness to invest in resources to improve inpatient throughput. The position was approved because of this hospital leadership buy-in.

In comparison with past QI initiatives to improve discharge efficiency through changes in physician workflow or training, this project focused on the addition of a new team member. Sustainment of increased discharges before noon was seen when the PFP transitioned to full time but did not significantly increase, likely representing a point of saturation. Although no decrease in LOS was observed, it is overall reassuring that LOS did not increase, as would be expected if discharges were being delayed until the following day to improve the primary metric of discharge before noon.¹⁶ Although an additional salaried position may not seem reproducible in all settings, an internal cost analysis showed favorable return in hospital savings. Future studies to quantify cost savings not only in hospital LOS but also in ED, pediatric ICU (PICU), and operating room throughput may help strengthen the case for upfront investment in additional staff in addition to traditional workflow optimization.

This study took steps to integrate QI and health equity principles by examining existing disparities in discharge efficiency based on racial, ethnic, language, and sex groups and tracking these throughout the study. Past studies have shown that patients who are from communities with lower household income, are uninsured or publicly insured, and identify as Black or Hispanic have longer average LOS.^17–20 In this study, we found that Black patients and patients who used an LOE discharged before noon less often than white and English-speaking patients. Although many QI studies are designed to target broad and heterogeneous patient populations equally, these interventions frequently have the unintended consequence of benefitting certain subgroups disproportionately, with the potential to improve, maintain, or worsen systemic inequities.¹¹ A health equity approach recognizes that different groups may require different resources to achieve the desired health outcome.¹² Although our study did not specifically aim to reduce inequities, several interventions—including performing rounds with an interpreter for Spanish-speaking patients first and improving early access to discounted medications for the uninsured—were done to try to improve equity. Future work is needed to engage historically marginalized patient populations to identify key drivers specific to these populations and design discharge processes that are efficient, patient-centered, and equitable.^11,21 

Lessons learned during this project include the need for robust orientation and mentorship for the new APP role. Because the inpatient pediatric space was new for the provider, there was an initial learning curve that is reflected in the early weeks of implementation. Development of standardized discharge return precautions and familiarity with commonly prescribed medications improved APP efficiency. Buy-in from hospital medicine attendings was also crucial to ensure continued supportive supervision and mentorship for the new role. Additionally, 1 attending mentor was assigned to the PFP to act as longitudinal support and address concerns.

Another important consideration was the effect of the new clinical provider on resident learning and autonomy. The project had a resident champion to ensure open communication during implementation. Informal feedback was overwhelmingly supportive, with residents reporting more time to attend educational conferences, engage in teaching, and interact with patients at the bedside. Further medical education research is needed as more APPs are incorporated within inpatient teaching hospitals.

There are several limitations to this study. This quality initiative took place in a single hospital on a single inpatient hospital medicine service. Because this service excluded children with medical complexity or multiple subspeciality involvement, this may limit the generalizability of the results. The sample size was adequate to perform biweekly analysis in the primary outcome and to observe sustained special cause variation in this outcome. Due to smaller sample sizes in the subgroups of our equity analyses, monthly analysis was required and thus was slower to identify change.

This QI project demonstrated improvement and sustainability of discharge before noon. Future work to address other barriers to efficient discharge, including nursing workflow and transportation access, may provide additional benefit. Future work should also explore patient-centered measures such as discharge upon clinical readiness, patient-rated readiness for discharge, and patient satisfaction with discharge. Additional hospital throughput measures, such as admission wait times, PICU transfer times, and bed turnover times, could be useful in understanding broader impacts of early discharges and potential return on investment. Other work has demonstrated improvement in discharge efficiency with integration of diagnosis-specific discharge criteria into the EMR, but this is not yet globally available for all diagnoses and is not integrated into the EMR at our institution.⁵ Future work should also engage historically marginalized patient populations to ensure equitable timely discharges in both qualitative and QI studies.

The addition of the PFP on a high-volume, low-acuity pediatric hospital medicine team increased timely morning discharges without increasing LOS or readmissions. Although percentages of Black patients, Hispanic patients, and patients using an LOE discharged before noon increased, the initial identified disparity compared with white and English-speaking patients persisted. Future work is needed to better understand how to address these ongoing disparities as well as examine the financial impacts of increased staff investment. Such a model should be considered in similar tertiary care, high-volume academic medical centers.

The authors acknowledge the contribution of Julia Arons, ARNP, to the success of this quality improvement effort by taking on the role of patient flow provider.

APP

advanced practice provider

ED

emergency department

EMR

electronic medical record

FTE

full-time equivalent

ICU

intensive care unit

LOE

language other than English

LOS

length of stay

OR

odds ratio

PFP

patient flow provider

PICU

pediatric ICU

QI

quality improvement

SPC

statistical process control