Brief
Current surgical workflows result in lost time for optimization and planning due to incomplete information at the time of patient presentation. We developed a learning decision support algorithm that identifies medical, socioeconomic, and health system factors to help the preoperative anesthesia and surgical screening (PASS) clinic efficiently communicate patient surgical risk and risk-mitigation strategies to Urology, Orthopedic, and Spine Surgery Clinics to support earlier decisions for/against ambulatory surgery center (ASC) care. This tool will result in three primary benefits:
- Identify patients who are eligible for ASC with enhanced planning and scheduling.
- For patients who are not eligible, facilitate early evaluation by PASS clinic for optimization that may lead ASC eligibility.
- Identify patient clusters (volume, characteristics, surgery type) that, while not currently supported in an ASC, can strategically be considered for transition off of the inpatient platform.
Context
A simplified surgical workflow contains four stages. First, a patient is referred from a clinic to Duke Surgery. Second, the patient visits an outpatient surgical clinic where they are evaluated for surgery. There, or soon after, a surgical case is created, and the surgery is scheduled. Third, a preoperative anesthesia and surgical screening (PASS) clinic with perioperative enhancement teams (POET) communicates with the patient and helps prepare them for surgery. The patient might be referred to other Duke clinics for specialized optimization. Finally, the patient arrives for surgery, receives anesthesia support, and undergoes surgery in the Operating Room (OR). After surgery, the patient recovers under the observation of professional care providers and is discharged with a follow-up call or visit within two weeks.
The OR care may be delivered in one of two facility types: the Ambulatory Surgery Center (ASC) or hospital. Generically, ASCs are designed for lower-risk patients, while hospitals are designed for higher-risk patients. The ASC is designed for patients who already have mitigated risk, are likely to have high-quality outcomes, and can be moved in and out of the operating room as quickly as possible without increasing their risk. Note that a patient may be an immediate ASC candidate at the time of case booking or, through optimization by the PASS clinic, may become eligible for the ASC to undergo surgery. ASC patients are typically discharged in less than 24 hours following admission and face lower outpatient charges than they would by staying longer in an inpatient hospital.
Undergoing high-risk surgery or any surgery as a high-risk patient may result in complications with significant ramifications for long-term health. To reduce such complications, preoperative optimization is often necessary. It may reduce the risk of mortality, return to the Emergency Department (ED), stay at the Intensive Care Unit (ICU), readmission rate, longer-than-expected lengths of hospitalization, extra clinic visits, and the cost of care.1,2,3
The PASS clinic at Duke University Hospital is the ideal team and physical location to intervene with optimization strategies. The PASS clinic has a unique perioperative and population health program that serves over 90% of all planned surgical and procedural volumes (50,000/year), and 15-20% of PASS patients require additional optimization to be ready for surgery. This PASS-POET model is a differentiator for Duke. The “one pipeline to the OR” model that the PASS clinic facilitates a unique approach to standardized, evidence-based, collaborative, multidisciplinary preoperative assessment and preparation.
The Problem
Despite the PASS-POET model, surgery workflow delays and misdirection occur due to incomplete or obscure information at the time of patient presentation in the surgical outpatient clinic. Missing patient information hinders early identification of patient surgical needs of patients and results in increasing the rates of poor outcomes.1,2 Our experience has taught us that about 15% of patients miss out on optimization due to poor anticipation of surgical needs.
Poor optimization is visible when high-risk patients show up at an ASC, and low-risk patients with low-risk surgeries are operated on at a hospital. At the time of problem identification, the largest ASC was 80% full, and the second largest was 50-60% full. We believe two problems cause the mismatch. First, a lot of information needed for a complete assessment of ASC versus Hospitalization readiness is not complete or clear, when the patient first arrives or even while being seen in the PASS clinic. Second, incentives for sending patients to the right place and taking sufficient time to screen patients are not aligned. While the PASS team wants to spend time mitigating the risk to the patient as much as possible, surgeons, administrators, and patients are motivated by financial and psychological incentives to move patients toward the operating room as soon as possible, especially given that ORs are critical drivers of health system revenue.
Our Solution
Our primary goal is to create an iterative perioperative learning platform with complete and transparent data to drive efficient preoperative planning and patient optimization. We expect to define criteria for admission to an ASC or hospital and measure the achievement of our goal by monitoring how well we meet those criteria during surgery workflow. Ultimately, we hope to share these learnings across the health system.
First, we developed eligibility criteria for ASC using PDFs of ASC exclusion criteria provided by physician leadership. Then, we developed a prototype platform to screen patients. Using the developed criteria, this platform screened individual patients and placed them into one of three categories: Red Light, Yellow Light, and Green Light. Patients who failed the criteria would be placed in the Red Light and scheduled for hospital surgery. They would require more time with the PASS clinicians. Patients who may become eligible for ASC with a few checks or optimization would be placed in the Yellow Light. Patients who passed ASC criteria would be placed in the Green Light and only receive a nurse’s screening call. Then, they would be expedited for ASC surgery.
This measurement platform added a degree of discernment for patients that preoperative care teams may better optimize for surgery. For example, a patient who is an obvious Green Light for ASC care is unlikely to require several in-person visits at the PASS clinic. The only data entry required was a medical record number, and then the critical information was presented on one straightforward screen.
Plans (in Dec 2023)
Our next step is to test the solution at the PASS clinic and among ASC surgeons. We anticipate observing a higher percentage of ASC volume among surgeries. Consequently, we expect a lower average post-operation length of stay and lower readmission rates.
The tool will improve as the development and efficiency of large language models (LLMs) increase, because the current tool is limited to extracting structured data. LLMs will allow the extraction of deeper insights about the patient’s comorbid, imaging, and social history that are found in narrative clinical notes.
Finally, the tool will synergize with the Pythion models (see Diffusion & Scaling update) by helping physicians evaluate the Yellow-Light patients. There are many patients for whom, even with full use of clinical notes, the correct destination of ASC or hospital needs to be clarified. These machine learning models will be able to review the entire patient’s chart, accurately organize the otherwise unclear cases into high- medium- and low-risk likelihoods, and save physicians’ time.
Anticipated Impact
Integrating a learning decision support tool into the surgical landscape at Duke, where approximately 5,000-6,000 surgeries result in same-day discharges on inpatient platforms, has profound implications for both the immediate and distant future of patient care and system efficiency. In the short term, the tool’s introduction promises to revolutionize how surgeries are planned and executed. While the ASCs are often utilized within 70% of their capacity, rooms still need to be used. Such under-utilization, juxtaposed against the backdrop of ASCs being stewards of efficient, cost-effective, and patient-centered care, underscores the critical need for streamlining patient selection based on comorbidities and other determinants of perioperative outcomes. By facilitating upstream visibility into a patient’s ASC eligibility, this tool addresses the identified gap. This real-time, data-informed insight will empower surgeons to plan more confidently, thus optimizing their utilization. Patients who are not immediately eligible can be directed to Duke North or the PASS clinic, thereby enhancing planning efficiency and precision. Additionally, perioperative teams can now optimize such patients at an early stage so they may become ASC-eligible.
In the long term, successfully integrating this tool could reshape the surgical ecosystem. The immediate outcome would be an uptick in the volume of surgeries at ASCs, which directly enhances their operational efficiency. Simultaneously, this would liberate more block time at Duke North, catering to surgeries requiring an inpatient framework. Such ripple effects underscore the tool’s potential to streamline operations and inform future strategic endeavors. As the platform amasses data, it can pinpoint patient and surgical clusters conducive to same-day discharges. This invaluable data trove can pave the way for infrastructural developments and innovations, possibly steering these clusters into an ASC environment. In essence, this is not just a tool. It is a compass directing the future of surgical care at Duke North, making it safer and more efficient.
References
- Jerath A, Austin PC, Ko DT, Wijeysundera HC, Fremes S, McCormack D, et al. Socioeconomic Status and Days Alive and Out of Hospital after Major Elective Noncardiac Surgery: A Population-based Cohort Study. Anesthesiology. 2020;132(4):713-22.
- Qi AC, Peacock K, Luke AA, Barker A, Olsen MA, Joynt Maddox KE. Associations Between Social Risk Factors and Surgical Site Infections After Colectomy and Abdominal Hysterectomy. JAMA Netw Open. 2019;2(10):e1912339.
- Ryan McGinn 1, Yonathan Agung 2, Alexa L Grudzinski 1et al. Perioperative Cost of Frailty after Major, Elective Noncardiac Surgery: A Population-based Cohort Study. Anesthesiology. 2023; Aug 1;139(2):143-152.
