Revitalizing Transfers: Creating an EHR-based admission and lateral transfer decision support tool to identify patients for transfer to non-Parnassus sites

PROJECT LEAD: Timothy Judson, MD, MPH. Medical Director of Care Delivery Transformation, Office of Population Health. Medical co-Director, Adult Care Transitions, UCSF. Physician Informaticist, Center for Digital Health Innovation. Assistant Professor, Division of Hospital Medicine.

EXECUTIVE SPONSOR(S):

Adrienne Green, MD, SFHM, FACP. Chief Medical Officer, Adult Services. VP Regulatory and Medical Affairs. Professor of Clinical Medicine, Division of Hospital Medicine.

Michelle Mourad MD, Vice Chair of Clinical Affairs and Value, Department of Medicine. Clinical Informatics Lead, Center for Digital Health Innovation. Professor, Division of Hospital Medicine.

Brad Monash, MD, Vice Chair of Clinical Affairs, Division of Hospital Medicine. Professor of Clinical Medicine and Pediatrics

ABSTRACT: It is currently challenging to determine which medicine patients may be eligible for admission or transfer to non-Parnassus sites of care (e.g. Mount Zion, Saint Mary’s Medical Center (SMMC)). Admitting physicians must manually check clinical and social inclusion criteria and bed availability for each site, an inefficient process that may lead to suboptimal use of available non-Parnassus beds. The process of identifying medicine patients appropriate for lateral transfers is similarly tedious and because it is dependent on physician referral, it fails to capture a large proportion of potentially eligible patients. We propose creating 1) an EHR-based, admission clinical decision support tool to guide admitting physicians in determining the optimal site of admission for medicine patients, and 2) a workbench report to identify admitted patients appropriate for lateral transfer to a non-Parnassus site of care. The combination of these tools will reduce the time needed to identify and transfer appropriate patients, increasing use of unfilled, staffed beds at non-Parnassus sites. This will increase tertiary/quaternary bed availability at Parnassus, creating an estimated $1.79M/year in contribution margin by better matching patient acuity to the appropriate site of care.  This will also improve patient and provider experience of the admission and transfer process.

TEAM

• Toni Workman-Braden, DNP, RN, Patient Care Director of Capacity Management, Integrated Transfer Center, Patient Capacity Management Center
• Maggie Jones MD, Associate Professor, Division of Hospital Medicine, Director of Swing Service and ED triage
• Rashmi Manjunath MD, Assistant Professor, Division of Hospital Medicine, Director of Mount Zion Medicine Service
• Raman Khanna MD Associate Professor of Medicine, Division of Hospital Medicine, Medical Director of Inpatient Informatics at UCSF Medical Center
• Jill Reynolds, Technical Program Manager, Patient Capacity Management Center

PROBLEM: This proposal is aimed at solving two related problems that contribute to capacity constraints at Parnassus, where inpatient capacity is currently over 90%.

First, when a medicine patient is being admitted from the emergency department, the Medical Officer on Duty (MOD) must decide where the patient should be admitted. There are currently three sites where hospital medicine patients may be admitted (Parnassus, Mount Zion, SMMC) with two additional sites (Hospital at Home and Chinese Hospital) being considered. The patient inclusion and exclusion criteria for each site is different (an example in Supplementary Materials), as are conditions necessary for acceptance, including bed availability and nurse staffing. Triaging physicians refer to frequently changing protocols and criteria to determine candidacy and manually confirm availability and acceptance at the receiving facility. This process of chart reviewing and seeking approval from multiple parties often takes 2+ hours from start to finish. This inefficient process may cause delays and lead to suboptimal bed usage, as some eligible patients are not identified and get admitted to Parnassus beds before they can be admitted to other sites.

Second, when Mount Zion beds are available, we aim to transfer lower acuity medicine patients from Parnassus. These patients typically have longer than usual length of stay (average 15.8 days), so transferring them can create significant capacity at Parnassus. Potentially eligible patients are currently identified based on referral from primary teams, and manually tracked on an APeX list by the Flow MD and MZ MD. This process is suboptimal for multiple reasons. First, referrals often do not yield enough candidates to fill available beds. When leadership has instead resorted to chart reviewing patients on the medicine census—akin to finding needles in a haystack of 180-200 patients – they have consistently identified more than double the number of potential MZ transfer candidates that were identified by referral.

TARGET: The goals of building these tools and implementing new workflows around medicine admissions and lateral transfer would be to:

1. Increase tertiary/quaternary bed availability at Parnassus by increasing use of unfilled, staffed beds at non-Parnassus sites. Specifically, we aim to:
   1. Increase monthly average daily census at MZ by 1.0 in FY24 compared to FY23
   2. Reduce unfilled bed capacity at all non-Parnassus sites by 50% in FY24 compared to FY23
2. Improve the provider experience by decreasing the amount of time that physicians take to identify and transfer patients to non-Parnassus sites
3. Improve patient experience by decreasing ED boarding times and improving transfer efficiency

GAPS: The process of determining which medicine patients may be eligible for admission or transfer to non-Parnassus sites is currently manual and referral dependent. The UCSF Patient Capacity Management Center (PCMC) has developed tools to solve similar problems. For example, they have developed a report that uses objective criteria to determine whether a patient is appropriate for the discharge lounge, and have begun to identify surgical patients appropriate for lateral transfers from Parnassus to Mission Bay.Prior to the creation of the PCMC, there was no central group with visibility into all admitted UCSF patients that could be responsible for maintaining and overseeing such a tool.

INTERVENTION: We propose to work with the UCSF Patient Capacity Management Center to create two related tools and associated workflows to improve the process of admitting and transferring patients to non-Parnassus sites.

1. Admission Clinical Decision Support (CDS) Tool: To better identify adult medicine patients on admission who are eligible for care at non-Parnassus sites (e.g. Mount Zion, SMMC, Hospital at Home), we propose an EHR-based, automated clinical decision support tool. This APeX tool would incorporate data from multiple sources to determine whether a patient meets inclusion and exclusion criteria for admission to a non-Parnassus site of care. This tool would capture data from discrete fields in APeX (e.g. insurance, home address, diagnosis, anticipated level of care) and via a short provider click-through form (e.g. need for telemetry or in-person specialty consultation). The tool will then incorporate real-time bed availability and staffing data, overseen by the Patient Capacity Management Center, to determine the recommended site of care. The MOD would use this tool prior to admitting the patient. If the tool recommends a non-Parnassus site of care, MOD would confirm assent from the patient and begin the admission process.
2. Medicine Lateral Transfer Identification Report: We propose building a workbench report to improve the process of identifying potential patients for lateral transfer to Mount Zion (and Hospital at Home in the future). This would include APeX metrics that help determine eligibility for transfers, such as current consults, expected discharge date, case management disposition, telemetry use, and AMPAC score (a measure of mobility). The Flow Physician on Duty (and/or MZ MD) would use this list as a starting point for identifying eligible patients for transfer. The PCMC team would also view the list in real time to better manage patient flow. The list will also be able to be filtered by bed type so that if, for example, there is not hemodialysis capacity available at Mount Zion, patients needing hemodialysis can be filtered out.

These tools would be first built for use with Hospital Medicine patients, given the immediate need. The tools could then be replicated for other services. If successful at UCSF, the tools could be shared with affiliates, and adapted to be used for transferring T/Q patients from regional affiliates to UCSF.

We have considered possible barriers to project success. First, some data may need to be manually entered by the admitting physician if it is not available in a discrete field (e.g. overall clinical stability). However, that entry would be brief, and could ultimately be streamlined by adding new fields to APeX (e.g. a “clinical stability” field). Second, bed availability data will be key to project success. We believe the UCSF Patient Capacity Management Center will be able to maintain this information in near-real time, but this may be more challenging for non-APeX sites like St. Mary’s and may continue to require confirmation via Voalte. Third, providers must know to use the tool for it to be successful. Fortunately, change management would be simplified because only specific attending hospitalist roles hold the MOD pager, and the MZ hospitalist is primarily responsible for identifying lateral transfers (see workflow diagram figure 1). There is already a robust onboarding process for each of these roles, and service leadership would ensure these changes are reflected in future onboarding.

PROPOSED EHR MODIFICATIONS: Described above in intervention section.

COST: Based on available data, we believe the average cumulative number of staffed, unfilled non-Parnassus beds is at least 2-4 per day. For example, the average daily census for hospital medicine patients at Mount Zion in 2022 was 14.2, which is 79% of the technical MD cap of 18 beds (though this may not account for MD/RN staffing changes that may increase or decrease this cap). This cap recently increased to 30 patients, which may increase the potential impact of this tool going forward. However even with that higher cap, the March ADC was 21.1. Similarly, median bed utilization at SMMC was 78% over the past 9 months.

Financial Benefit: Increase T/Q revenue. Using both the admission and transfer tools, we aim to increase tertiary/quaternary bed availability at Parnassus by increasing use of unfilled, staffed beds at non-Parnassus sites. Each time a low-acuity Parnassus patient is transferred to a non-Parnassus site, this Parnassus bed can be filled with another patient. Realistically, these open Parnassus beds may be filled with any type of adult patient. For that reason, in our calculations we use the contribution margin for all adult patients, not just T/Q patients. Specifically, we aim to:

1. Increase monthly average daily census at MZ by 1.0
2. Reduce the number of non-Parnassus unfilled, staffed beds by 50%

*This calculation assumes backfill with a mix of all adult patients (not just T/Q patients)

Other benefits of this project would include:

1. Benefit Hospital at Home: These tools would be integral to the success of future capacity-building programs such as Hospital at Home, which rely on proactive patient identification for both admission and lateral transfer. These tools could decrease the need for additional FTE to be hired for the purpose of patient identification in Hospital at Home. 
2. Decreased physician transfer/admission work, (resource-intensive workflows visualized in tan boxes in Figure 1, with plan to replace with much more efficient tool in green) resulting in:
   1. Increased clinical productivity and billable activities (e.g. admission hospitalist/MOD can conduct History and Physical’s instead of spending time on triage). This benefit is difficult to quantify in the absence of a dedicated time study, so we did not include this in the value calculation above.
   2. Increased provider satisfaction (e.g. inc. time for teaching instead of coordinating transfers)
   3. Improved patient outcomes, safety and experience (e.g. by allowing MZ hospitalists to focus on discharge education rather than identifying transfers)
   4. Potential to decrease ED boarding: While there are other confounding factors, more efficient use of non-Parnassus beds could decrease ED boarding, ED LWBS rates and PACU boarding.
   5. Empower the PCMC: This tool would increase transparency of clinical acuity and admission characteristics, allowing the PCMC to prioritize which non-Parnassus beds to fill first, and giving them more data regarding the number of patients potentially eligible for lateral transfer.

SUSTAINABILITY: These tools would require some upkeep by service directors and the APeX team to reflect changes in clinical criteria for admission to non-Parnassus sites. We have support from Hospital Medicine service leaders, PCMC leadership, Department of Medicine leadership, and the Chief Medical Officer of Adult Services, so are confident the tools would become part of standard work for admissions and transfers.

BUDGET

• APeX modifications: $25k
• Technical project management support: $5k
• Subject matter expert salary support: $20k

SUPPORTING DOCUMENTS

Figure 1a: Admission Current and Future State Workflow (SEE ATTACHED)

Figure 1b: Lateral Transfer Current and Future State Workflow (SEE ATTACHED)

Mount Zion Medicine Exclusion Criteria

• TCU level of care
• Actual or anticipated LOS >8 weeks** (only applies to lateral transfers)
• Patients requiring high flow nasal cannula
• Patients too unstable for transfer
• Patient requiring in person consultant evaluation (primarily telemed consults)
• Patients needing administration of chemotherapy (oral or IV)
• Patients with significant expected specialty involvement, specifically
  • Active co-morbid or primary surgical disease
  • Active co-morbid or primary neurologic or neurosurgical disease
  • Active hematologic disease
  • Active oncologic disease
  • Severe cardiac disease (i.e., active/unstable CAD, decompensated heart failure, severe pHTN)
  • Decompensated liver disease or liver transplant patient
  • Pregnancy
  • Hemodialysis (not an exclusion criterion, but patient must be cleared by MZ Nephrology prior to admit/transfer. Those who may need emergent/urgent dialysis for decompensated renal disease should not be transferred)
  • Peritoneal dialysis (relative exclusion, only possible with lateral transfers + coordination with Nephrology/nursing)
  • 5150 or very active comorbid or primary psychiatric disease
    
    

CW Transfer Tool Round 1 Reviewer Comments/Questions and Responses

1. Your proposal would be strengthened if you can comment on how you would ensure that providers who are tasked with transferring patients to non-Parnassus beds would know to utilize your tool, and how they would utilize your tool. Outlining a proposed workflow is a great starting point for this key behavior change that would be required for your project to be successful.

• We have included current state and proposed future state workflow diagrams in our revised proposal. The green boxes in both the admission and lateral transfer workflows indicate where the use of the digital tool would replace the time consuming process (tan box) in the current state.
• For the lateral transfer tool, change management would be straight-forward, because the MZ hospitalist already references an Apex list to try to find patients. If the lateral transfer tool were built, the hospitalist would just use that tool instead.
• The admission CDS tool would be a new process for the MOD, but only 3 hospitalist roles currently hold MOD, and each receive orientation and onboarding via email/wiki pages. We feel confident that given the engagement of service leadership in this proposal, we would be able to achieve universal use of these tools.

1. It would be helpful to know if identification of patients who would qualify for non-Parnassus beds is a key contributor to unfilled non-Parnassus beds. Prior work and health systems kaizens have been conducted to optimize this transfer process, and utilizing these prior learnings will also strengthen your proposal. In other words, how do we know that solving the problem of more efficient patient identification will result in increased transfers to SMMC and/or Mt. Zion? There are other gaps/barriers in this complex process as well, and how will your project team also address these barriers?

• There are certainly several barriers to optimizing the transfer process. Based on discussions with our team and past improvement efforts, we would group these barriers into 2 categories:

1) Bed availability including staffing challenges often limits volume of possible non-Parnassus admissions/lateral transfers. Bed availability would not be impacted by this proposal.

2) Patient factors may lead to available beds not being filled. Patient factors Include identifying appropriate patients, getting their assent, confirming they have appropriate insurance, etc. This proposal would increase the size of the pool of potential patients, which should result in more beds being filled, even without directly addressing issues like assent or insurance.

• It is challenging to measure missed opportunities for transfers. However, the best evidence available comes from recent experience. When the MZ hospitalist team recently increased its cap to 30, there were not enough patients on the APeX list of potential lateral transfers, so service leadership manually chart reviewed all patients on the medicine census. They identified more than double the number of patients who were on the transfer list, supporting the hypothesis that we are missing potentially appropriate lateral transfers. Despite the Mount Zion medicine cap being raised to 30 in February, March ADC was ~21.1, attributed in part to difficulty identifying enough patients.
• The Adult Capacity Management Dashboard consistently shows that the bed occupancy rates at MZ and SMMC (~75-80% max) are much lower than those at Parnassus (>90%). This imbalance suggests an opportunity to better match patient acuity to bed availability.
• We feel that these tools would be immediately valuable in our current state. In addition, we want to emphasize that improving these workflows would be essential if we continue to expand our non-Parnassus capacity (e.g. Chinese hospital, Hospital at Home). Without such tools and revised workflows, it would get increasingly complex and resource intensive to identify patients for the correct disposition, and we would fail to identify potentially appropriate patients.

3.     We would strongly suggest connecting with Jahan Fahimi, Brandon Scott, and Kim Berry to discuss your cost savings / increased revenue projections. Many of the patients who will benefit from filled non-Parnassus beds are already patients in the ED who are awaiting admission and thus there may not be a true increase in health system revenue.

We connected with Brandon Scott, Kim Berry, Jahan Fahimi and Cat Lau. They agreed with our approach. We did make the following adjustments to our value model based on their feedback:

• Clarified that we are referring to increased contribution margin (I had mistakenly written revenue in prior version)
• Used Kim’s figure for contribution margin per bed day instead of the prior figure (~10% difference)
• Adjusted our specific aims: We believe that the bulk of the benefit from this tool in the short term will come from increasing lateral transfers to MZ, so we revised our goal to increase monthly ADC at MZ by 1.0 (from 0.5). Our calculation assumes that we will fill an additional cumulative 1.5 beds per day accounting for all non-Parnassus sites. We would therefore estimate that 0.5 of this ADC would be in increase at SMMC, or other non-Parnassus sites opened in the future.