The Importance of Incorporating Patient Throughput in Crisis Standards of Care Simulations

Numerous publications have sought to further our understanding of how crisis standards of care (CSOC) strategies might perform, with specific attention to excessive deaths or exacerbating existing social disparities, during the coronavirus disease (COVID-19) pandemic. We write to raise an important concern with many CSOC studies to date, that simulate patient cohorts by synchronizing patients’ presentation to a single time point, rather than the reality where patients present continually over time. ^1–4 This collapsed model may not accurately reflect patient throughput and dynamic resource strain, which would preclude identifying those patients affected by CSOC policies. Understanding how CSOC might perform remains important as areas within New Hampshire, Arizona, New Mexico, Idaho, Alaska, and Maryland have activated their own crisis standards of care protocols.

Methods

All intubated COVID-19 patients at a single New York City (NY) health care system during the first surge (March 1, 2020 to June 30, 2020) were included in a simulation requiring CSOC activation once 95% of pre-pandemic ventilators were utilized and lasted 2 weeks (crisis period), consistent with a prior simulated length of ventilator rationing utilizing patient throughput under the New York State Ventilator Allocation Guidelines (NY ⁵ ) CSOC. ⁶ Patient charts were reviewed to determine whether NY, Maryland (MD), ⁷ Pittsburgh (PA), ⁸ Saskatchewan Canada (SAC), ⁹ and California (CA) ¹⁰ CSOC criteria were satisfied (Table 1, Supplemental Methods) and whether patient ventilator usage occurred during the 2-week crisis period. NY, MD, SAC, and CA** CSOC use exclusionary criteria to preclude patients from receiving a ventilator under CSOC. Subsequently, NY and SAC only use a Sequential Organ Failure Assessment (SOFA) score for triage, whereas MD, PA, and CA all integrate graded comorbidities, in addition to a SOFA score to generate an overall triage score for ventilator allocation. PA, CA, and SAC each make occupational accommodations to partially prioritize whether the patient is an essential, critical, or occupation related to health care, respectively. PA uses the Area Deprivation Index (ADI, State scores 8-10) to favorably adjust an overall priority score for a ventilator, whereas other CSOC use the Social Vulnerability Index (SVI, scores ≥ 0.75); both were derived from the patient’s address. Categorical data were analyzed with the chi-squared test (95% confidence interval). This study was exempt by the New York University Langone Institutional Review Board.

Table 1. Crisis standards of care exclusion and triage score modifiers grouped by theme

See Supplemental Methods for specific criteria and how operationalized. Sequential Organ Failure Assessment (SOFA) score. Fixed triage score modifiers can deprioritize (comorbidities) or increase the prioritization (special patient populations) of an individual for a scarce resource.

* As expressed in the original CSOC publication.

** California CSOC does not have an explicit exclusion category but provides to only this group if excess supply is present.

Results

In total, 911 patients were included in the cohort, of which 573 were involved during the crisis period. Table 2 depicts the total affected, excluded, comorbidities modifying triage score, and the occupation or social vulnerability adjustment by each CSOC. NY, MD, PA, SAC, and CA would have excluded 1, 3, 0, 93, and 3 patients, respectively, for the entire cohort, except 0, 0, 0, 45, and 2, respectively, during the specific crisis period. MD, PA, and CA would have modified 44, 88, and 106 individuals’ triage scores, respectively, due to comorbidities in the entire cohort but only 17, 46, and 43, respectively, during the specific crisis period. The crisis period statistically affected MD (P = 0.04) and CA (P = 0.0056), with a trend seen for PA and SAC CSOC.

Table 2. Prevalence of patients satisfying exclusion or triage score modifying criteria for 5 Crisis Standards of Care strategies during the crisis period vs entire cohort

Discussion

When studying how resource allocation under CSOC might perform, any simulation that synchronizes all admissions rather than incorporating real patient throughput and dynamic resource strain would inadvertently include patients whose outcomes would be associated with normal or contingency standards of care instead of those truly secondary to CSOC. This failure to identify which patients would actually be affected by CSOC guidelines would likely distort valuable CSOC objectives, such as maximizing life-years saved without exacerbating existing social disparities. While MD and CA were the only CSOC strategies reaching statistical significance for cohort differences when including patient throughput, we believe the PA and SAC trends toward significance represent a limitation of our cohort size. We remain concerned with CSOC features, as expressed in NY, MD, and SAC CSOC, that categorically exclude patients regardless of resource strain because rationing might not occur dichotomously but across a continuum over time. While PA has replaced a list of objective comorbidity criteria with a physician’s broad assessment of expected death within 5 years or 1 year, despite successful treatment of acute illness, objective criteria achieve impartiality, preserve fairness toward patients, promote transparency with the community, maintain reproducibility between providers, eliminate the encroachment of bias or prejudices by decision makers, and may reduce provider distress about resource allocation decisions.

A limitation was that this study was not a full simulation designed to determine excessive deaths per CSOC; however, our objective was rather to demonstrate the differences in cohort compositions that would likely affect key CSOC outcomes. A 2-week crisis period was determined by the length of time ventilators required to be rationed during a separate NY-CSOC simulation in this patient cohort. ⁶ Other CSOCs might require a shorter or longer duration of rationing that would affect those patients included in a crisis period cohort. In conclusion, we highlight the dynamic nature of the crisis period and encourage future CSOC studies to incorporate dynamic patient throughput to correctly capture patients who would be affected by CSOC policies.