Enhancing patient flow through standardised discharge pathways for neurology and medicine services

Abstract

Background and objectives

Poor discharge planning impairs hospital throughput, adds to the financial strain on health systems and diminishes patient and provider satisfaction. We developed consensus-based discharge criteria coupled with a standardised discharge pathway for four presenting diagnoses and tracked their effect on discharge timing and length of stay (LOS).

Methods

Medical readiness for discharge criteria for patients diagnosed with transient ischaemic attack, seizure, demyelinating disease or syncope were generated by expert consensus at our institution. A standardised discharge pathway was developed for eligible patients based on discussions with stakeholders and staff. Discharge timing and readmissions were tracked for 6 months pre-intervention and 12 months post-intervention (divided into 6 months of implementation and post-implementation periods). The primary outcome was a discharge time of ≤2 hours for 60% of patients during the implementation period. Secondary outcomes included reduced time to discharge (TTD) and LOS compared with the pre-intervention period.

Results

318 total patient visits were included across the baseline, implementation and post-implementation periods. Median TTD improved from 171 min at baseline to 88 and 92 min, respectively, during the implementation and post-implementation periods. Median LOS similarly decreased from 94 hours to 35 and 30 hours, respectively. All primary and secondary outcomes were achieved during the implementation period and sustained post-implementation. The rate of emergency department visits and hospital readmissions within 30 days remained low (~1.5%) post-intervention. Additionally, most providers reported that the intervention improved clinical workflow.

Conclusions

This standardised discharge framework improved discharge efficiency for patients with four common diagnoses during an 18-month quality improvement study. The framework and its implementation are highly scalable, and similar systems-level approaches should be considered by hospitals to improve throughput.

WHAT IS ALREADY KNOWN ON THIS TOPIC

• Poor discharge planning impedes hospital throughput and increases the financial burden on health systems.

WHAT THIS STUDY ADDS

• This study demonstrates that a standardised discharge framework that incorporates diagnosis-specific and consensus-based criteria can substantially reduce time to discharge and hospital length of stay, as well as help to identify barriers to discharge early.

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

• This study presents a highly scalable framework for other health systems to adopt. It provides a practical model for creating durable, system-level solutions that can be expanded to other diagnoses and care teams to improve hospital-wide throughput.

Introduction

Hospitals throughout the country face problems caused by a lack of organised discharge pathways for patients. In addition to insufficient bed capacity and inefficient discharges, poor delineation of the steps required for appropriate discharges impacts patient flow and hospital throughput, that is, the efficiency by which patients move through the hospital system from admission to discharge, encompassing all aspects of patient care and utilisation of hospital resources. This adverse effect has been evidenced by studies showing that lack of standardisation of admission and discharge pathways, system-level barriers and provider decision-making all contribute to prolonged hospital stays.¹,⁴ These deficiencies consequently increase financial burdens on health systems and diminish patient and provider satisfaction.⁵,⁸ To address these concerns, the Institute for Healthcare Improvement (IHI) has published articles (eg, Rutherford et al⁹) that focus on improving patient flow and discharge predictability and has recently promoted a conceptual change from arbitrary time-based discharge criteria that are not supported by studies^{10 11} to consensus-based discharge criteria. These consensus-based discharge criteria should be established for common disease presentations by the appropriate experts in the health system and incorporate agreed-upon prerequisites for discharge readiness. While prior studies have created consensus-based discharge criteria that are disease-specific and situation-specific,¹²,¹⁴ institutional differences and varying access to follow-up studies and care require that consensus-based discharge criteria be institution-specific. Nevertheless, such initiatives can help interdisciplinary care teams predict the timing of patient discharges and identify discharge barriers in a timely fashion. Prior studies such as El-Eid et al¹⁵ have demonstrated that a standardised framework to address discharge barriers can produce effective and durable solutions. Coupling a standardised discharge pathway, where discharge barriers can be identified early and addressed at a systems level, with consensus-based discharge criteria that determine when a patient is medically ready for discharge, enables hospitals to enhance patient flow and improve throughput by coordinating admissions through the emergency department, postoperative care, operating rooms, outside transfers, as well as care team staffing and postdischarge care. For example, this approach can ensure timely coordination of discharges to ensure that newly admitted patients arrive quickly in the correct ward with the appropriate nursing expertise.

This article describes the implementation of consensus-based criteria for determining medical readiness for discharge (MRD), that is, completion of all diagnostic and therapeutic interventions necessary during the admission as well as stabilisation of the patient’s health, along with a standardised discharge pathway for four common diagnoses to improve discharge timing and length of stay (LOS) at a busy academic tertiary care hospital. While the disease-specific discharge criteria may be unique to our hospital (or others with similar available resources), we believe the framework of coupling these MRD criteria with a standardised discharge pathway that includes central tracking of MRD criteria completion and notification to all hospital care team members can be implemented by many other hospital systems. This would improve coordination of patient throughput and help earlier identify barriers to discharge.

Methods

Setting

This quality improvement study was conducted from 1 May 2022 to 31 October 2023 at the Ronald Reagan Medical Center at the University of California, Los Angeles, a 446-bed hospital with 24 077 admissions during the 2022–2023 fiscal year. The study duration was divided into: (1) the pre-intervention baseline from 1 May 2022 to 31 October 2022; (2) the study implementation period from 1 November 2022 to 30 April 2023 and (3) the post-implementation period from 1 May 2023 to 31 October 2023.

Establishing consensus-based discharge criteria

MRD criteria were established by experts in the departments of neurology and medicine for four common presenting diagnoses encountered on a weekly basis by these inpatient services. These diagnoses were transient ischaemic attack (TIA), seizures, demyelinating disease and syncope. The MRD criteria for these diagnoses are provided in figure 1, as generated from expert discussions and community standards.

Figure 1. Consensus medical readiness for discharge criteria created in collaboration with appropriate experts for each diagnosis selected, including transient ischaemic attack, seizure, demyelinating disease and syncope. TCD, transcranial doppler.

Discharge planning coordination

The standardised discharge pathway was implemented for each patient presenting to the emergency department or admitted to the observation or inpatient neurology units with a qualifying diagnosis. A multidisciplinary care team was involved in all aspects of the discharge pathway and consisted of emergency medicine physicians, medicine hospitalists, critical care providers, neurology attendings/fellows/residents, medical students, advanced practice providers, charge/staff nurses, nursing unit directors, pharmacists, case managers, social workers, physical/occupational/speech therapists, diagnostic study representatives, as well as quality officers and members of the leadership and logistics teams. However, the care team structures varied across units. For example, the observation unit was led by medicine hospitalists and nurse practitioners, while the inpatient neurology team was led by an attending physician and a neurology senior resident physician. Creating a uniform discharge pathway among all stakeholders required meetings to delineate expectations regarding each stakeholder’s role and to define the steps necessary to meet these expectations. Nevertheless, each stakeholder was given flexibility in meeting the unique needs of each patient, with the pathways serving as standard guidelines. Discharge planning checklists were created for each patient and then tracked throughout their hospital stay. These included, for example, the expectation that providers send discharge medication orders to the pharmacy within 24 hours prior to discharge to avoid delays. Transport staff were also included within this discharge pathway framework. Several meetings were held among the global group to ensure successful coordination among all roles and across service lines. This process additionally identified commonly encountered discharge barriers that occurred across several inpatient units, due to poorly delineated roles, and proposed solutions to mitigate such barriers.

Communication between providers and staff was essential as patients proceeded through the discharge pathway. Once the workup was completed and MRD criteria were met, the managing provider placed the discharge orders and sent a secure message within the electronic medical record (EMR) to members of the multidisciplinary care team who were involved in the discharge process, including the discharge pharmacist, the unit charge nurse and the unit case manager. The multidisciplinary care teams existed even prior to the pre-intervention period of this study. After the team members were notified, discharge orders were released with the expectation that patients would be discharged within 2 hours. Time stamps for all phases of hospitalisation (admission orders, discharge orders, time of actual discharge, etc) were already recorded in the EMR as part of the basic clinical workflow. These time stamps were manually extracted by personnel who were not involved in the clinical care of the patient nor the study outcomes or analyses. Barriers to discharge were also documented in the EMR. For example, delays in inpatient studies could be extracted in an unbiased manner using the time the study order was placed and the time the study result was signed. Other barriers, such as locating accepting post-hospital care facilities, could be extracted from the EMR through interdisciplinary notes and consult orders for case management/social work services. When the 2-hour discharge goal was not met, a review among the team members was performed to identify/review case-specific barriers and brainstorm how to address such barriers in the future. This discussion among the team members typically occurred during interdisciplinary rounds each day, and changes were executed subsequently. An IHI waste tool was also used to assist in identifying discharge barriers in real time that were shared among team members. To support timely discharges, stakeholders were rewarded with modest gift cards prior to the pre-intervention period and during all periods of the study. Additionally, discharge data were shared periodically with each unit.

Inclusion criteria

Only adult patients with a primary presenting diagnosis of TIA, seizure, demyelinating disease or syncope were included in this study. Patients whose initial qualifying presenting diagnosis was later changed to another diagnosis during the admission were removed from the study. There were no other exclusion criteria.

Study design

Overall, there were four main drivers throughout this project: (1) to obtain buy-in from all stakeholders and create a global discharge pathway to be applied across all involved service lines and units; (2) to develop consensus MRD criteria for the four common diagnoses that were the focus of this study; (3) to identify discharge barriers early and (4) develop solutions to address these. We hypothesised that using MRD criteria would reduce time to discharge (TTD, time from when discharge orders were actually released to completion of the discharge), decrease LOS and minimise provider variability in discharge decision-making. We focused on TTD and LOS since the former has been used in prior studies¹⁵ and the latter is a common metric tracked nationally.¹⁶ To this end, the primary outcome of this study was a TTD of 2 hours or less for at least 60% of patients with a qualifying diagnosis who met MRD criteria. To this end, TTD was tracked for all enrolled patients across the emergency department and inpatient observation and neurology units. The secondary outcomes were a significant decrease (single-tail Mann-Whitney U test, α=0.05) in the TTD and LOS during the implementation period compared with the pre-intervention period. TTD and LOS during the post-implementation period were also compared with their pre-intervention baseline. Note that non-parametric tests were used because Kolmogorov-Smirnov tests demonstrated that the data were not normally distributed. The rate of emergency department visits and readmissions within 30 days of discharge (balancing measures) was tracked to ensure these remained <5% during the implementation and post-implementation periods. Lastly, to evaluate provider satisfaction with the discharge pathway as well as any unintended burden on daily workflow, anonymous electronic surveys were sent to providers after the post-implementation period.

Results

A total of 130 patient visits were included during the study implementation period, of which 31 were for TIA, 24 for demyelinating disease, 28 for seizures and 47 for syncope. There were 52 and 138 patient visits during the pre-intervention and post-implementation periods, respectively, although two patient visits during the post-implementation period were discarded due to missing data regarding discharge timing; thus, 136 were analysed. The median baseline pre-intervention time between discharge orders and actual discharge was 171 min, while during the implementation period it was 88 min, and post-implementation it was 92 min. This was equivalent to a 49% reduction in median discharge time during the implementation period compared with the pre-intervention period across all four diagnoses. Pre-intervention median LOS was 94 hours, while during the implementation period, it was reduced to 35 hours and further declined to 30 hours post-implementation.

The primary outcome of the study was achieved, with >60% of patients discharged within 2 hours during the study implementation period (figure 2). This was maintained post-implementation (figure 2). During the implementation period, the emergency department and observation unit achieved the 2-hour goal for 95% and 84% of discharges, respectively. However, the high nursing turnover on the inpatient neurology service resulted in achieving the 2-hour goal for only 42% of discharges. In particular, one of the two neurological wards had several newly hired nurses who were unfamiliar with the documentation expectations and delayed discharge times in order to complete their documentation. Once this was realised, the unit director addressed this with the nursing staff to avoid subsequent unnecessary delays in discharges. Subsequently, in the post-implementation period, the number of discharges from the inpatient neurology service achieving the 2-hour goal increased to 55%.

Figure 2. Histograms of TTD for patients during the (A) pre-intervention, (B) implementation and (C) post-implementation periods. The y-axis on the right corresponds to cumulative distribution across discharge times (dashed line). The primary outcome of the study intervention was achieved, as >60% of discharges in the implementation period (as well as the post-implementation period) took place within 2 hours. This was not the case during the pre-intervention period. TTD, time to discharge.

The secondary outcomes of the study were also achieved. As shown in figure 3, there was a profound reduction in the TTD and LOS between the implementation and pre-intervention periods, as well as between the post-implementation and pre-intervention periods (p<10⁻⁹ for all comparisons). Note that groundwork for the intervention pathway was under development 6 weeks prior to the implementation period start date, possibly contributing to a slight decrease in TTD at the end of the pre-intervention period.

Figure 3. TTD (A) and LOS (B) for each patient (dots) during the pre-intervention, implementation and post-implementation periods. To improve visualisation of the data, extreme outliers outside the visible range are represented by crosses at the top of each plot. The black line corresponds to the 5-point moving median (non-overlapping and includes extreme outliers). The secondary outcomes of the study were achieved as the TTD and LOS were decreased during the implementation period (as well as during the post-implementation period) compared with the pre-intervention baseline (one-tailed Wilcoxon rank sum, for all comparisons ****p<10⁻⁹). LOS, length of stay; TTD, time to discharge.

Among the four diagnoses, syncope had the highest proportion of discharges within 2 hours (see figure 4). Again, most of these patients were typically housed in the emergency department or observation unit where discharge expediency is stressed and is well understood by nursing staff. The diagnoses of demyelinating diseases and seizures were associated with the longest discharge times as the patients were typically admitted to the inpatient neurology service.

Figure 4. TTD (min) for each diagnosis. The box plot demonstrates the median and quartile values (25%, 75%) with 5% and 95% whiskers. Values for each patient are overlaid. TIA, transient ischaemic attack; TTD, time to discharge.

When the 2-hour discharge goal was missed, hospital stays were often prolonged by 2 days or more. As noted in prior studies,^{6 17 18} this typically occurred when the discharge pathway steps were not followed. For example, discharges were delayed when medications were not sent to the outpatient pharmacy with a reasonable lead time. This barrier alone could cause a delay of one to 2 days if an insurance authorisation was needed or if there was a temporary medication shortage.¹⁹ This was a reason why the discharge pathway included prescriptions being sent to the pharmacy at least 24 hours prior to discharge. Another example was delayed requests for further diagnostic testing by consulting services on the planned day of discharge, which could delay discharge by several hours.

Emergency department visits and readmissions were tracked for patients with a qualifying diagnosis during the pre-intervention, implementation and post-implementation periods. One patient (2.0%) was readmitted within 30 days of prior discharge during the pre-intervention period. This patient initially presented with a seizure episode after purging in the setting of bulimia nervosa and re-presented within 2 weeks with gastrointestinal upset after being prescribed antibiotics by an outside facility for a presumed infection at the site of a prior intravenous placement. Two patients in the implementation period (1.6%) and two in the post-implementation period (1.5%) had at least one emergency department visit or readmission within 30 days of prior discharge. One patient with post-traumatic epilepsy re-presented three times to the emergency department during the implementation period, all for higher seizure burden or clustering, and was directly admitted for surgical evaluation during one of these presentations. Another patient re-presented once to the emergency department during the implementation period for TIA symptoms (similar to their original presentation), but the patient had already been started on appropriate therapy and a repeat workup was unrevealing. One patient with epilepsy was readmitted during the post-implementation period after a subsequent seizure but had just started on an increased dose of antiseizure medication as recommended during the initial admission and repeat workup was unrevealing. Another patient with a history of an intracranial tumour complicated by epilepsy was readmitted during the post-implementation period for higher seizure burden, but subsequent workup revealed that these episodes were non-epileptic. Overall, none of the repeat presentations/admissions during the implementation or post-implementation periods represented a failure of discharge planning during the previous presentation. Moreover, the readmission rate was not significantly different among the pre-intervention, implementation and post-implementation periods based on χ² testing (p=0.81).

A satisfaction survey sent to 27 providers after the post-implementation period had a response rate of 63%. When asked whether it became easier to meet the goal of discharge within 2 hours after implementing the study intervention, 58% agreed or strongly agreed (figure 5A). Likewise, 76% of responders felt the discharge pathway helped them organise discharges (figure 5B).

Figure 5. Responses to two questions (A and B) from an anonymous electronic survey sent to providers after the post-implementation period.

Discussion

During this 18-month pilot study, the implementation of a standardised discharge pathway with consensus-based discharge criteria for four common diseases seen by neurology and internal medicine providers improved TTD and LOS substantially and was well received by providers. These improvements persisted even in the post-implementation period, suggesting a durable effect. Moreover, the rates for subsequent emergency department visits and readmissions remained low. Overall, implementation of such pathways can be highly scalable and should be considered by hospitals to improve patient throughput as well as patient/provider satisfaction.

Standardisation of discharge criteria across providers was a guiding principle for this study. The study used the same inpatient-to-outpatient transitional model for neurology patients that had been in place prior to the study and did not change after the study terminated. This transitional model included (1) communication with patients within 48 hours postdischarge via Health Insurance Portability and Accountability Act (HIPAA)-compliant electronic messaging and (2) follow-up with a transitional care provider in the clinic within 7–14 days postdischarge based on readmission risk (LACE index²⁰). Any outstanding diagnostic results were reviewed and addressed during this transition clinic visit. Again, all patients, regardless of whether or not they were discharged through the standardised pathway with MRD criteria, received the same postdischarge care and follow-up. This was also true for patients with syncope who were followed by the medicine team during hospitalisation. As noted in the results, only two patients in the implementation period and two patients in the post-implementation period re-presented to the emergency department within 30 days of discharge, and all of these were for clinically unavoidable reasons rather than a failure of discharge planning during the previous presentation.

To ensure optimisation in implementing the discharge pathway with consensus-based discharge criteria, the interdisciplinary team iteratively reviewed performance and identified common discharge barriers. Subsequent steps were taken to mitigate these barriers in the future. In this manner, along with updated expectations among the interdisciplinary team, the implementation was constantly adapted to overcome hurdles in discharge and reduced the LOS across all service lines. For example, when timely recommendations by other consulting teams for further inpatient diagnostic workup were identified as a barrier to discharge, the discharge pathway was updated to include notifying consult services of patients’ upcoming discharges at least 24 hours before to avoid attempting these studies on the day of the intended discharge.

Provider variability in clinical decision-making and the idiosyncratic nature of provider-specific discharge criteria have been previously demonstrated to negatively impact patient flow throughout healthcare systems. At our medical centre, the time attending physicians spent on service varied significantly, as some were on service for only a few weeks per year, whereas others were on service more regularly. Additionally, this is compounded by the high turnover of managing providers for the neurology services and observation unit, based on scheduling restraints. Some attending physicians are on service for 24 hours per day for 14 days straight, while others attend strictly during daytime hours with an alternative provider covering overnight after a sign-out. Thus, the process of creating consensus-based discharge criteria for all providers required strong collaboration among many of the specialists in our health system who were experts in managing the four diagnoses included in this study (ie, TIA, seizures, demyelinating disorders and syncope). Neurology specialists, fellows and residents, as well as medicine hospitalists, were also involved in vetting these criteria. The resulting framework sought to minimise the impact of provider variability regarding discharge decisions and create consistency with patient care and appropriately expedite discharges.

Certain obstacles related to the timely completion of diagnostic studies, such as echocardiograms and radiology studies, were identified as frequent barriers to discharge during this study and required system-wide changes. For example, it was discovered that an echocardiogram ordered Tuesday through Thursday as an inpatient typically took 24 hours to complete, whereas if it was ordered Friday through Monday, it could take as long as 48–72 hours to complete. Technician availability, especially during weekends, was determined to be the limiting factor, and this was reported to hospital leadership to ensure support for this diagnostic study 7 days per week. However, this occurred after the conclusion of the study period and was not a substantial driver in the reduction in LOS observed during the study. Completion of radiology studies was also identified as a discharge barrier, leading to the creation of a dashboard to track turnaround times. It was revealed that providers throughout the hospital were designating studies as urgent for routine matters, which placed additional burden on radiologists who were already dealing with a high volume of studies and had limited capacity to deal with studies that were necessary for discharge. Educational sessions regarding ordering of echocardiograms and radiology studies were provided for multiple care teams throughout the study; however, these educational sessions had little impact on ordering behaviour for diagnostic scans during the study, as evidenced by a largely unchanged proportion of incorrectly ordered diagnostic studies in the pre-intervention, implementation and post-implementation periods. Moreover, turn-around times of these studies were largely unchanged throughout the study periods. Thus, while obstacles were identified during the study, they were not adequately addressed prior to the end of the study and did not significantly contribute to a reduction in LOS during the study.

Transportation of patients to home was another frequent reason for delayed discharge, with patients waiting hours for family members or needing to make alternative arrangements. As such, the discharge pathway was updated to incorporate early planning for transportation home to ensure adequate arrangements were made prior to the day of discharge. Additionally, a patient discharge lounge was used to avoid delays and improve throughput and bed availability.

One limitation to this study is its narrow scope, as only four diagnoses were included, with an emphasis on neurologic conditions. In fact, the demyelinating disease discharge pathway during this study showed anomalously low volume of presentations and thus may be affected by insufficient sampling. Nevertheless, we expect that the implementation is generalisable to a larger number of diagnoses and participating services, and we plan to scale it up and expand in the future. In addition, while the complex barriers noted previously all impact patient flow, the nuanced solutions to address them at our institution may not be generalisable or feasible to others and were not specifically built into the discharge pathway design. Hospital flow oversight may also be institution-specific, but a standardised discharge pathway, nevertheless, allows a system to better predict bed availability at any point in time and plan for future contingencies. It should be noted that as part of the implementation of this standardised discharge pathway, hospital leadership was also involved in managing bed access. Another limitation was the need for a time-consuming manual review of patient charts to extract time stamps for placed discharge orders and discharge completion, along with deviations from the standardised discharge pathway. In the future, we plan to create an automated process to extract discharge metrics from the EMR and track discharge criteria using automated checklists, as well as notify all members of the interdisciplinary team when an anticipated discharge is nearing. There was also inter-unit variability in adhering to the standardised discharge pathways and completing discharges within 2 hours of when the orders were placed. While this was discussed at the unit director level to mitigate inconsistencies, it suggests that additional education would be useful for providers/interdisciplinary team members in clarifying expectations and leadership support.

Modest gift cards and encouragement from leadership regarding discharge timing were provided during all periods of the study (including the pre-intervention period), but may have been more emphasised during the implementation period. It is possible that this could have contributed to better adherence to the standardised discharge pathway and shorter TTD and LOS during the implementation period. However, the primary and secondary outcomes were maintained during the post-implementation period, suggesting a durable effect of the intervention itself. In addition, incentives such as gift cards have become more common in quality improvement initiatives and could be implemented generally by other hospital systems. When resources are limited and monetary items (eg, gift cards) are not feasible, non-monetary options (shout-outs or other tokens of appreciation) can be used. Another limitation is that it is unclear if certain aspects of the standardised discharge pathway (eg, utilisation of the discharge lounge or transportation resources) played significant roles in reducing TTD and LOS. Nevertheless, all of these resources were available prior to the study (including the pre-intervention period). However, it is possible that awareness of discharge timeliness during the implementation period led to increased utilisation of these available resources. It should be noted that the roll-out of the standardised discharge pathway occurred a few weeks prior to the November 2022 implementation of the consensus MRD criteria, but a corresponding early decrease in TTD and LOS is not clearly seen in figure 3 that would have indicated a specific effect of the standardised discharge pathway. Another potential limitation was the imbalanced pre-intervention and implementation/post-implementation periods (51 patients over 52 patient visits vs 126/134 patients over 130/136 visits, respectively). However, the study was designed such that each period was 6 months long, and so this imbalance appears to have been by chance. Regardless, the analysis used non-parametric testing that is robust to outliers. We do not expect patients to be aware of this quality improvement study, and although providers could have relayed to patients what further diagnostics were pending prior to their discharge, this is largely standard practice regardless of an MRD criteria checklist. Since the patient satisfaction surveys are sent to randomised patients, there was no way of tracking whether or not they had been involved in the standardised discharge pathway with MRD criteria, and therefore, the effect of this intervention on patient satisfaction could not be adequately measured. Providers’ knowledge that the initiative was being implemented could have influenced discharge timeliness (eg, pushing for study results, increased communication to stakeholders even prior to signing discharge orders, etc), although once the discharge orders were placed, their ability to affect TTD was minimal.

Conclusion

This study, involving multiple specialty services and multidisciplinary team members at a single tertiary care academic hospital, demonstrates that implementation of an integrated and standardised discharge pathway with disease-specific consensus-based discharge criteria can improve discharge efficiency, shorten LOS and maintain provider satisfaction without sacrificing patient care. Given the consistent results among diverse structures of the neurology and medicine services, it is likely that this strategy can be effective more generally for other disease processes and service lines.

Data availability statement

Data are available upon reasonable request.