key: cord-0901047-6goj7ilh
authors: Lyon, Martha E; Bajkov, Andrew; Haugrud, Diane; Kyle, Barry D; Wu, Fang; Lyon, Andrew W
title: COVID-19 Pandemic Planning: Simulation models to predict biochemistry test capacity for patient surges
date: 2020-11-28
journal: J Appl Lab Med
DOI: 10.1093/jalm/jfaa231
sha: fa3f0d9b4b0544ed63c196fd44a2148076c03d3d
doc_id: 901047
cord_uid: 6goj7ilh

BACKGROUND: Patient surges beyond hospital capacity during the initial phase of the COVID-19 pandemic emphasized a need for clinical laboratories to prepare test processes to support future patient care. The objective of this study was to determine if current instrumentation in local hospital laboratories can accommodate the anticipated workload from COVID-19 infected patients in hospitals and a proposed field hospital in addition to testing for non-infected patients. METHODS: Simulation models predicted instrument throughput and turn-around-time for chemistry, ion-selective-electrode and immunoassay tests using vendor-developed software with different workload scenarios. The expanded workload included tests from anticipated COVID patients in two local hospitals and a proposed field hospital with a COVID-specific test menu in addition to the pre-pandemic workload. RESULTS: Instrumentation throughput and turn-around time at each site was predicted. With additional COVID-patient beds in each hospital the maximum throughput was approached with no impact on turnaround time. Addition of the field hospital workload led to significantly increased test turnaround times at each site. CONCLUSIONS: Simulation models depicted the analytic capacity and turn-around times for laboratory tests at each site and identified the laboratory best suited for field hospital laboratory support during the pandemic.

Healthcare professionals rely on the timely delivery of clinical laboratory test results to triage patients and make decisions that affect treatment. This vital role has been accentuated during the COVID-19 pandemic as critical care bed numbers are expanded and field hospitals are established to support patient care. In this study, simulation models were used to predict if current instrumentation in local hospital laboratories could accommodate the anticipated workload. Simulation models depicted the analytic capacity and turn-around times for laboratory tests at each site and identified the laboratory best suited for field hospital laboratory support during the pandemic.

The rapid emergence of coronavirus disease 2019 (COVID-19) has had a profound effect on the delivery of clinical laboratory medicine. Many organizations have created surge capacity planning committees to organize and optimize local healthcare resources in response to the COVID pandemic. Bed number has traditionally been the metric of hospital capacity, although it does not directly capture the complexity of a healthcare system organization (1) . Modern capacity planning has advanced to anticipate increased demand on hospital laboratories using the "Four-S" interdisciplinary approach which accounts for relationships and dependencies among staff, stuff, structures and systems (2) . Computer simulation models have aided in projecting hospital wide surge capacity during extended periods of operational stress (3, 4) . Unfortunately, during a crisis there is seldom time to develop appropriate computer models to assist in surge planning.

The increasing demand for timely, accurate and precise methods to detect the presence of SARS-CoV-2 viral RNA has strained operations in many microbiology laboratories. Concurrent expectations for core clinical laboratories to support increasing numbers of COVID patients, in addition to non-COVID patients, have contributed to the stress on healthcare systems. For pandemic planning, laboratory test capacity (i.e. maximum number of tests per hour or day) is an important and dynamic metric to evaluate, complicated by varying parameters such as reaction time for each assay, analyzer processing (throughput) speed, automated track throughput, and laboratory staff dependent manual processes, etc. To this end, many clinical laboratory vendors have developed software to simulate and predict throughput, turn-around-time (TAT) and capacity throughout the workday with different analyzer configurations and volumes of tests. The objective of this study was to use a pre-existing simulation model developed by a manufacturer to predict the throughput and TATs for chemistry and immunoassay instruments in two local hospital laboratories with scenarios of increasing workload derived during the COVID pandemic (up to 914 additional inpatients) in addition to the pre-pandemic workload.

Saskatchewan is a 250-bed tertiary care facility consisting of emergency medicine, medical units, critical care, regional kidney health and provincial transplant programs. The SPH biochemistry service supports hospitalized patients however eighty percent of the test workload is derived from out-patients that attend clinics as well as analysis of outreach community-patient tests ordered by family medicine. Functionally, this translates into an approximate pre-COVID workload of 2,900 patient specimens and 21,000 automated chemistry, ion selective electrode (ISE) and immunoassay tests per day. The SPH laboratory instrumentation consists of a single automated track that links a Roche® c8100 pre-analytic module with two groups of ISEs, model c702 and c502 spectrophotometric chemistry analyzers, an e801 model immunoassay analyzer, and a single p701 model refrigeration unit. Royal University hospital (RUH) and the adjoining Jim Pattison Children's Hospital (JPCH) together represent a 606 bed trauma center for the province that provides acute care services, maternal and child services, neurosurgery and cardiovascular surgery. The biochemistry service for these hospitals is centralized at RUH where the pre-COVID workload is 1,200 patient specimens with 8,000 automated chemistry, ISE and immunoassay tests per day. The RUH instrumentation consists of a single automated track that links a Roche® c8100 pre-analytic module with two pairs of ion selective electrodes, model c502 spectrophotometric chemistry analyzers and a model e801 immunoassay analyzer.

De-identified patient data from 2016 was extracted from the laboratory information system (LIS) for a representative 24-hour period for the following parameters:

site, specimen accession number, test code(s) ordered per specimen, specimen and container type, time of collection and arrival in the laboratory. Extracted data was increased by a 15% volume, proportionately distributed throughout the day, to reflect the pattern of the current workload. Statistical and Calculation Methods: For each simulation condition, the hourly predicted 90 th centile TATs and the daily (24 hr) mean of the hourly predicted 90 th percentile TATs were calculated. TAT was determined from when the specimen was introduced onto the pre-analytic module to when the analysis was complete and the result ready for reporting.

The change in this index was determined by subtracting the simulation condition daily mean from the baseline condition daily mean. The mean changes (in minutes) were compared by ANOVA (n=24). Confidence intervals for the mean changes in 90 th percentile TATs for each simulation condition were also determined.

During the initial phase of the COVID-19 pandemic, laboratory leaders were asked to plan how to deliver services with expanded populations of patients in the local hospitals and in a field hospital. We approached the vendor of our automated laboratory systems to ask if the proprietary software used to simulate laboratory workflow for marketing could be adapted to predict workflow with expanded COVID patient populations during the pandemic. In a vendor-academic partnership, we simulated module capacities at both the RUH and SPH sites for chemistry, immunoassay and ISE for a 24-hour period for each of the six simulation conditions, Table 1 .

Royal University Hospital: Figure 1 

SPH showed a slightly different daily pattern of throughput consumption compared to RUH. Laboratory workload at the SPH site is largely derived from community-based patients collected throughout the daytime hours. Figure 2 panel A shows the incremental rise for chemistry module throughput consumed for each simulation condition throughout the 24hr time period, similar to that seen at the RUH. For the baseline condition, capacities peaked at 10am (~60%), and progressively declined until 2pm, where another small peak (~20%) was observed at 4pm, and then generally remained low for the remainder of the day. This pattern was consistent for the simulation conditions although the peaks increased approximately 2-3% respectively as bed occupancy increased up to 100%. Predictably, the condition simulating maximal COVID occupancy at SPH and the field hospital generated an 10am peak of ~75% followed by a 4pm peak of ~50%.

The simulation conditions for immunoassays (figure 2 panel B) depict a similar pattern to the chemistry module capacity ( figure 2 panel A) . Two peaks of IA capacity consumption were noted. The initial peak magnitude was observed at 10am and the second occurred between 4 to 6pm. The baseline condition had an initial 10am peak of ~65% throughput capacity consumed and a second peak of 50% at 6pm. IA throughput did not exceed the maximum module capacity. The final condition with field hospital workload generated an 11am peak of ~90% and a second peak of ~85% capacity consumption.

The anticipated ISE workload (figure 2 panel C) demonstrated a major peak between 9 to 10am and it generally decreased throughout the day for all stress conditions.

For the baseline condition, the 9am peak consumed ~15% ISE module capacity. Each COVID stress condition increased the workload and required more throughput capacity, to a maximum of 20% with addition of the field hospital workload. Unlike RUH, the capacity consumed for the chemistry, IA and ISE modules at SPH resembled a more bimodal distribution, with the major increase in volumes occurring between 8am and noon and a subsequent smaller peak occurring between 4 to 6pm.

The effect of various simulation conditions on the hourly 90 th centile TAT is shown in Figure 3 for RUH and SPH. Simulations indicated that only the addition of the field hospital workload influenced the anticipated 90 th percentile for TAT at each site. Tables   2 and 3 illustrate the statistically significant mean changes in the 90 th centile TAT over a 24 hour period at each site. Maximal occupancy of field hospital COVID beds started to lengthen hourly TATs by 7am which remained lengthened until midnight at RUH ( Figure   3 panel A) . A similar lengthened TAT pattern was observed at SPH that resolved by 6pm ( Figure 3 panel B) . As test volumes increased with the scenario of maximum patient numbers, TATs were predicted to significantly lengthen causing tubes to remain in the instrument buffer and be analyzed in the subsequent hour. The simulations predicted that both hospital sites could not accommodate the workload from the field hospital in addition to their expanded COVID beds and the pre-pandemic workload without extending

TATs. It is anticipated that the 90% percentile TATs will remain constant until the maximum throughput of the analytic system is exceeded. When this occurs, we speculate that the 90% TAT will likely begin to increase in a non-linear manner. However, an investigation of the relationship between the specific saturation point and the 90% TAT was not conducted. Simulation studies also prompted a re-assessment of our scheduled instrument daily maintenance times. Our current practice, at both hospitals, is to perform maintenance of one analyzer line during the middle of the day (noon to 3pm) which translates into peak workload time and the second line maintenance is scheduled for midnight to 3am. Workload efficiencies can be gained if instrument maintenance were instead conducted during "off peak" times. Such changes in maintenance times however will impact technologist/operator shift scheduling.

Simulation modeling and its forecasts are not without limitations. The usefulness of simulations will be dependent upon parameters, such as physician ordering patterns for management of COVID as well as non-COVID patients during peak pandemic that were estimated for this study. This study assumed that existing laboratory staff levels would be sufficient to accommodate the additional workload. A major limitation of this study was the use of proprietary vendor-provided simulation software that has not been described in peer-reviewed publications and had a limited evaluation under baseline conditions for this study. At this time, local emergency medicine, hospital admissions and laboratory workload during the COVID-19 pandemic has not exceeded capacity or exceeded the baseline workload described in this study. It remains a limitation that the impact of elevated workload has not been assessed. In spite of these caveats, our simulation findings have been helpful to identify instrumentation capacity issues that could affect our ability to delivery timely test results during a pandemic surge in our organization.

We anticipate that currently employed simulation software from many different vendors could facilitate pandemic planning of laboratory resources.

In conclusion, the simulation models provided an in-depth perspective of the 

Understanding surge capacity: essential elements

Health systems 'surge capacity': state of the art and priorities for future research

Epidemiologic modeling with FluSurge for pandemic (H1N1) 2009 outbreak

Health service resource needs for pandemic influenza in developing countries: a linked transmission dynamics, interventions and resource demand model

An innovative approach to functionality testing of analyzers in the clinical laboratory

Modular Analyzer Series Evaluated under Routine-like Conditions at 14 Sites in