Pith. sign in

REVIEW 4 major objections 5 minor 36 references

Ambulance Allocation for Patient-Centered Care

T0 review · 4 major / 5 minor · reviewed 2026-08-07 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read A two-stage allocation model shows 15–25% diversion-capable fleets achieve up to 80% of possible ED diversions.

desk verdict A solid OR modeling paper whose headline 80% and 3.4–8.6x numbers are model-based estimates built on a machine-generated eligibility pool; the comparative dispatch insights are the real contribution. read the letter →

arxiv 2505.23560 v2 pith:2RB2CY7C submitted 2025-05-29 math.OC

Eric G. Stratman , Justin J. Boutilier , Laura A. Albert This is my paper
classification math.OC MSC 90B8090C1190B2290C15
keywords emergencymedicalservicesambulanceallocationEDdiversionpatient-centeredcaretwo-stagestochasticoptimizationmultipleresponsedispatchsecondaryassignmentM/G/d/dqueue
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

Emergency medical services traditionally send nearly every 911 patient to an emergency department, even when a lower-acuity setting would do. This paper argues that patient-centered care does not require converting an entire ambulance fleet: a two-stage optimization model that allocates and dispatches two types of units finds that equipping 15–25% of a fleet with diversion-capable ambulances captures roughly 80% of the achievable ED diversions while an $M/G/d/d$ queueing constraint keeps ambulances available. The same model shows that adaptive dispatch—sending multiple units up front or activating a secondary unit after on-scene assessment—raises diversion rates by 3.4–8.6 times relative to conventional single-unit dispatch when diversion-capable units are scarce. If these numbers hold, modest equipment investments plus dispatch coordination can unlock most of the modeled benefit of patient-centered EMS.

What carries the argument

The carrying object is a two-stage mixed-integer optimization with recourse. The first stage chooses where to station traditional and diversion-capable units and makes initial assignments from phone-screening information; the second stage, after on-scene assessment reveals the true condition, chooses treatment actions and whether to dispatch a secondary unit. Availability is enforced by an $M/G/d/d$ queueing chance constraint that limits the blocking probability at each station, and the uncertainty parameters linking the stages come from two machine-learning classifiers: a Diversion Eligibility Classification model that labels which patients could be diverted and a Dispatcher Screening Prediction model that estimates how accurately phone screening reproduces those labels.

What would settle it

Take a random sample of patients that the model labels diversion-eligible and have a blinded clinical panel independently judge whether treat-in-place or alternative-destination care was appropriate; if the panel's eligible share differs materially from the 16% target used to set the classification thresholds, the 80% diversion figure and the 15–25% fleet guidance would have to be re-estimated.

Watch

Extended reading notes

Core claim

The paper's central claim is that a mixed-integer optimization model of patient-centered EMS, with two unit types and two decision stages, can identify allocation and dispatch policies that divert most eligible patients without sacrificing ambulance availability. In simulations across urban, rural, mixed, and suburban regions, roughly 80% of the modeled diversion potential is reached with only 15–25% of the fleet diversion-capable, and the paper's Full Dispatch strategy (multiple response plus secondary assignment) achieves 3.4–8.6 times more diversions than single-unit dispatch at low capability levels. The model also shows that when recourse dispatch is available, screening accuracy has little effect on diversion rates; screening becomes important only in single-response systems.

Load-bearing premise

The machine-learning labels that mark a patient as eligible for diversion are trained on patients who were actually diverted or refused transport in a system with almost no diversion options, and the threshold is tuned so that exactly 16% of patients are labeled eligible; if that eligibility pool is wrong, every diversion percentage in the paper shifts.

Editorial extensions

If this is right

  • Agencies can equip a minority of their fleet—15–25%—and still capture most of the modeled diversion benefit, leaving the rest as conventional ED-transport units.
  • Dispatch coordination is the strongest lever when diversion-capable units are scarce: full adaptive dispatch multiplies diversions 3.4–8.6 times over single-unit response.
  • When agencies cannot implement recourse dispatch, improving phone-screening accuracy becomes the main alternative lever, raising urban single-response diversion rates from 33.0% to 58.3% under perfect information in the paper's experiments.
  • Since the optimal strategy shifts with fleet composition, agencies face an explicit trade-off between upgrading more units with simple operations and upgrading fewer units with advanced dispatching protocols.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • The paper leaves untested whether the 15–25% threshold is robust to a different definition of clinical eligibility; if a prospective clinician audit relabels the eligibility pool, the equipment and dispatch results should be recomputed before adoption.
  • A testable extension would run the same allocation model with eligibility labels from a blinded clinical panel rather than from historical diversion behavior; matching the 80% and multiplier results would separate the model's mechanism from its training labels.
  • Because the simulations fall back to ED transport when recommended units are busy, real-world diversion gains may be smaller in systems with heavier concurrent demand; the $\alpha=0.05$ availability constraint bounds this effect but does not eliminate it.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 5 minor

Summary. The paper proposes a two-stage mixed-integer optimization model for allocating ambulance units in a patient-centered EMS system with two unit types (traditional and diversion-capable) and three treatment pathways (ED, AD, TIP). The first stage makes initial dispatch decisions from dispatcher-screening information; the second stage chooses treatment actions after the patient's true condition is observed, with recourse through secondary assignment. A chance constraint based on an M/G/d/d queueing model enforces a minimum ambulance-availability level. Dispatcher accuracy and diversion eligibility are estimated from NEMSIS data with a random-forest pipeline, and the model is evaluated by discrete-event simulation in four Ontario regions. The headline findings are that 80% of potential ED diversions can be achieved with 15-25% of the fleet diversion-capable, and that adaptive dispatch improves diversion rates by 3.4-8.6 times over conventional single-unit dispatch.

Significance. If the underlying eligibility labels and service-time assumptions were reliable, the paper would provide directly actionable guidance on a real operational question and would extend the EMS facility-location literature by combining multiple response, secondary assignment, and an M/G/d/d availability constraint in one mixed-integer model. The formulation is careful: the constraints are linear, the ραd capacity transformation is standard, and 699 of 714 instances (97.9%) solved to an optimality gap below 0.01, which is a genuine computational achievement. The paper's main vulnerability is empirical: every headline percentage is measured against a machine-learned, threshold-calibrated eligibility pool rather than a clinically validated ground truth, and the simulation draws true patient conditions from the same estimated pθθ matrix used in the optimization objective. In addition, the service-time specification in the availability constraint appears to omit travel time. With those issues addressed—or with the claims explicitly reframed as conditional on the estimated labels—the paper would be a solid methodological and policy contribution.

major comments (4)
  1. [Section 4.2 and Table 1] The DEC labels define the denominator for every diversion rate in the paper, but the positive class consists of patients who were currently diverted, treated in place, or declined transport, and the classification threshold is calibrated so that exactly 16% of patients are labeled diversion-eligible, matching a 2013 DHHS estimate. In a system where nearly all patients are transported to the ED, current diversion and refusal behavior is at best a biased proxy for clinical appropriateness. The claims that agencies can achieve 80% of possible diversions with 15-25% of the fleet, and that adaptive dispatch improves diversion rates by 3.4-8.6 times, are therefore all measured against an unvalidated machine-generated denominator. The paper's own caveat that the ML framework is 'not intended to be deployed for real-time clinical triage' does not resolve the issue, because the same labels are used as ground truth for scoring results. Please validate the eligibility labels against an external clinical standard (for example, clinician chart review or ET3-type criteria) or, failing that, re-frame the results as a sensitivity analysis over plausible eligibility pools and report how the fleet-composition and dispatch-strategy conclusions shift.
  2. [Sections 4.2, 4.3, and 5.5] The simulation samples actual patient conditions from the same pθθ matrix that appears in the optimization objective, and the DSP 'dispatcher accuracy' is measured as the agreement of a random forest with the DEC labels, not as human dispatcher performance. This is not a logical circularity in the sense that the model is internally coherent, but it means that the policy experiments in Section 6 do not test the model against an independently realized world; they test how well the model optimizes and simulates under its own estimates. At minimum, the paper should state this coupling explicitly and should add an external validation step for pθθ (for example, hold-out regions, prospective data, or clinical adjudication of a sample of DEC labels) before the fleet-composition recommendations are presented as empirical guidance.
  3. [Section 3.3.2, Eqs. (23)-(24)] The service-time specification appears to omit travel time for dispatched units. The text in Section 5.3 says that the qijkmθ~ parameters exclude travel and that the rij terms are travel times, but in Eq. (23) rij appears only as the waiting time contributed by a secondary unit, and in Eq. (24) τhat contains no rij term at all. If this reading is correct, constraint (26) understates the true busy time of EMS units, and the α=0.05 availability guarantee may not hold in the simulation, which does include travel times as separate exponential durations. Please either include rij in both τ and τhat for initial and secondary units, or clarify that q includes travel, and then report the availability levels actually achieved in the simulation.
  4. [Section 5.5 and Section 6] The simulation results report only diversion rates; they do not report realized ambulance availability, mean response time, or the frequency with which the fallback protocol is invoked. Because the availability constraint is central to the feasibility of the recommended allocations, the paper should report these quantities for each configuration so that readers can verify that the 95% availability target is actually met in the dynamic simulation and that the diversion gains are not accompanied by degraded operational performance.
minor comments (5)
  1. [Abstract, Section 6.1, Section 7] The abstract and conclusion state that agencies can equip 15-25% of the fleet, but Section 6.1 reports 28% for urban, 25% for rural, 25% for mixed, and 16% for suburban; please harmonize the reported range or state the per-region percentages explicitly in the abstract and conclusion.
  2. [Section 3.3.1] The phrase 'the two parameters represent' contains a typo ('twod' appears in 'the twod parameters'); please correct it.
  3. [Eq. (22)] The telescoping sum on the right-hand side implicitly requires ρα0 = 0; please state this definition before Eq. (22).
  4. [Table 1] The column header 'Actual Condition (Θ)' should read 'Actual Condition (Θhat)' to be consistent with the notation introduced in Section 3.2.
  5. [Sections 4.2 and 4.3] Please report standard classification performance metrics (AUC, calibration, sensitivity/specificity at the chosen thresholds) for both the DEC and DSP models; currently only feature-importance plots are shown, which do not quantify predictive reliability.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the optimization and simulation are self-contained given externally estimated inputs; the DEC/DSP eligibility pipeline is a validity caveat, not a circular reduction.

full rationale

The central claims are derived from solving a two-stage mixed-integer program whose parameters come from external data (NEMSIS) and an external benchmark (DHHS 16% estimate). The p_theta_thetatilde matrix is estimated by comparing DSP predictions to DEC labels, then used both in the objective (10) and in the simulation. Reusing the same fitted matrix in the evaluation is an internal-consistency choice, not a forced equivalence: the reported 80% and 3.4-8.6x figures are outputs of the optimization and discrete-event simulation, not algebraically equal to p_theta_thetatilde or to the 16% threshold. The DEC threshold is tuned to match the DHHS proportion, so the size of the eligible pool is an input assumption rather than a predicted result of the model. The paper's own caveat that the ML framework is 'not intended to be deployed for real-time clinical triage' flags a real external-validity limitation, but concerns about whether historical diversion/refusal behavior is a good proxy for clinically appropriate non-ED care are correctness risks, not circularity under the quote-and-reduction standard. The self-citations (Stratman et al. 2022, 2023; Boutilier and Chan 2022) provide background and standard queueing results (Erlang-B type blocking probabilities) and are not load-bearing in a way that reduces the paper's claims to its own inputs.

Assumptions & free parameters 6 free parameters · 5 assumptions · 0 invented entities

The central quantitative claims rest on a chain of estimated inputs: a threshold-tuned eligibility classifier, a fitted dispatcher-accuracy matrix, point-estimate service times, and coverage-model-derived fleet sizes. None of these is validated against independent clinical or operational data, so the model's absolute outputs inherit that uncertainty.

free parameters (6)
  • DEC classification threshold = 0.65 urban, 0.689 rural
    Tuned so that 16% of patients are labeled diversion-eligible, matching a 2013 DHHS estimate; sets the size of the potential-diversion pool.
  • Dispatcher accuracy matrix p_theta_thetatilde = Table 1 (urban), Table B.1 (rural)
    Random forest DSP model compared to DEC labels; used in the objective, availability constraint, and simulation ground truth.
  • Service times q_ijkm_thetatilde = ED 49 min, AD 43 min, TIP 45 min, support 5 min
    Point estimates from NEMSIS or community paramedicine literature; support time explicitly described as a conservative assumption.
  • Fleet size per region = 25 urban, 24 mixed, 38 suburban, 8 rural
    Chosen by a coverage optimization model to achieve 95% availability; actual fleet sizes for the regions are unknown.
  • Availability target alpha = 0.05
    Policy choice for the blocking probability in the M/G/d/d constraint; results depend on it.
  • Demand scaling ratio = 157:1
    Ratio of total EMS responses to cardiac arrests from NEMSIS, used to scale cardiac-arrest demand into EMS call volumes in each region.
assumptions (5)
  • standard math M/G/d/d blocking probability depends only on offered load rho for given d and alpha; rho_alpha_d values from Boutilier and Chan (2022) are valid
    Invoked in Section 3.3.1 (Eq. 19-22) to replace the service-level guarantee with a linear capacity constraint.
  • domain assumption Arrival processes at stations are Poisson and service times are independent with known means
    M/G/d/d queue model requires Poisson arrivals; Section 3.3.1 states this, citing Kim and Whitt (2014) for empirical consistency.
  • domain assumption Dispatcher screening accuracy estimated on NEMSIS data transfers to the four Ontario study regions
    p_theta_thetatilde from Section 4 is applied to Hamilton, Muskoka, Halton, and Simcoe in Section 5.4 without region-specific validation.
  • domain assumption Historical diversion or transport refusal behavior identifies patients who are clinically eligible for alternative pathways
    DEC labels in Section 4.2 treat currently diverted patients as ground truth for eligibility, despite the absence of alternative pathways in most of the data.
  • domain assumption Hierarchical treatment pathway structure: TIP-eligible patients may also use AD or ED, and AD-eligible patients may use ED
    Defines the action sets M' and M'' in Section 5.4 and underlies the diversion-counting logic in the objective.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Ambulance Allocation for Patient-Centered Care." pith.science (2026). https://pith.science/paper/2RB2CY7C

@misc{pith2026250523560,
  author       = {Pith},
  title        = {Pith review of: Ambulance Allocation for Patient-Centered Care},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/2RB2CY7C}},
  note         = {Machine review of arXiv:2505.23560}
}
read the original abstract

Emergency Medical Services (EMS) in the United States and similar systems typically utilize a single treatment pathway, transporting all patients to emergency departments (EDs), regardless of their actual care needs or preferences. Recent policy reforms have sought to introduce alternative treatment pathways to divert lower acuity patients from the ED, but operationalizing these options has proven difficult. This paper proposes a patient-centered EMS (PC-EMS) ambulance allocation model that supports multiple care pathways by aligning EMS responses with individual patient needs. We develop a two-stage mixed-integer optimization framework that incorporates multiple dispatch and secondary assignment strategies which enable dynamic resource deployment. The model maximizes appropriate ED diversions while maintaining ambulance availability using a queueing-based availability constraint. We leverage national EMS data and machine learning to estimate dispatcher accuracy and diversion potential. Simulations across diverse geographic regions suggest that agencies can achieve up to 80% of possible ED diversions by equipping only 15 to 25% of their fleet with diversion capable units. Adaptive dispatch strategies improve diversion rates by 3.4 to 8.6 times compared to conventional single unit dispatch. These results provide actionable guidance for PC-EMS implementation by quantifying the trade off between equipment investment and operational coordination. Using the allocation model, agencies can strategically choose between upgrading fewer units with advanced dispatching protocols versus larger fleet investments with simpler operations. This approach offers flexible pathways suited to different organizational capabilities and implementation readiness.

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

36 extracted references · 35 canonical work pages

  1. [1]

    , " * write output.state after.block = add.period write newline

    ENTRY address author booktitle chapter doi edition editor eid howpublished institution isbn issn journal key month note number organization pages publisher school series title type url volume year label extra.label sort.label short.list INTEGERS output.state before.all mid.sentence after.sentence after.block FUNCTION init.state.consts #0 'before.all := #1...

    work page

  2. [2]

    write newline

    " write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 global.max substring 't := if while FUNCTION word.in "" FUNCTION format.date year ...

    work page

  3. [3]

    , " * write output.state after.block = add.period write newline

    ENTRY address author booktitle chapter doi edition editor eid howpublished institution isbn issn journal key month note number organization pages publisher school series title type url volume year label extra.label sort.label short.list INTEGERS output.state before.all mid.sentence after.sentence after.block FUNCTION init.state.consts #0 'before.all := #1...

  4. [4]

    Transportation Research Part E: Logistics and Transportation Review 103:143--157

    Ansari S, Yoon S, Albert LA (2017) An approximate hypercube model for public service systems with co-located servers and multiple response. Transportation Research Part E: Logistics and Transportation Review 103:143--157

    work page 2017

  5. [5]

    Manufacturing & Service Operations Management 11(4):674--693

    Argon NT, Ziya S (2009) Priority assignment under imperfect information on customer type identities. Manufacturing & Service Operations Management 11(4):674--693

    work page 2009

  6. [6]

    Transportation Science 23(4):277--287

    Batta R, Dolan JM, Krishnamuthy NN (1989) The maximal expected covering location problem: Revisited . Transportation Science 23(4):277--287

    work page 1989

  7. [7]

    Issue brief (Commonwealth Fund) (435):1--5

    Billings J, Parikh N, Mijanovich T (2000) Emergency department use in New York City : a survey of Bronx patients. Issue brief (Commonwealth Fund) (435):1--5

    work page 2000

  8. [8]

    Scandinavian Journal of Trauma, Resuscitation and Emergency Medicine 26:94

    Bohm K, Kurland L (2018) The accuracy of medical dispatch - a systematic review. Scandinavian Journal of Trauma, Resuscitation and Emergency Medicine 26:94

    work page 2018

Show all 36 references
  1. [9]

    Computers & Operations Research 123:104974

    Boujemaa R, Jebali A, Hammami S, Ruiz A (2020) Multi-period stochastic programming models for two-tiered emergency medical service system. Computers & Operations Research 123:104974

    2020

  2. [10]

    Manufacturing & Service Operations Management

    Boutilier JJ, Chan TCY (2022) Drone network design for cardiac arrest response. Manufacturing & Service Operations Management

    2022

  3. [11]

    Program Information , U.S

    Centers for Medicare & Medicaid Services (2022) Emergency Triage , Treat , and Transport ( ET3 ) Model . Program Information , U.S. Department of Health and Human Services, Baltimore, MD

    2022

  4. [12]

    Health Services Research 53(5):3455--3470

    Dainty KN, Seaton MB, Drennan IR, Morrison LJ (2018) Home visit‐based community paramedicine and its potential role in improving patient‐centered primary care: a grounded theory study and framework. Health Services Research 53(5):3455--3470

    2018

  5. [13]

    Transportation Science 17(1):48--70

    Daskin MS (1983) A maximum expected covering location model: Formulation , properties and heuristic solution. Transportation Science 17(1):48--70

    1983

  6. [14]

    White Paper , Department of Health and Human Services, Washington, DC

    Department of Health and Human Services (2013) Innovation Opportunities for Emergency Medical Services . White Paper , Department of Health and Human Services, Washington, DC

    2013

  7. [15]

    Scandinavian Journal of Trauma, Resuscitation and Emergency Medicine 26(1):8

    Eastwood K, Morgans A, Stoelwinder J, Smith K (2018) Patient and case characteristics associated with ‘no paramedic treatment’ for low-acuity cases referred for emergency ambulance dispatch following a secondary telephone triage: a retrospective cohort study. Scandinavian Jour...

    2018

  8. [16]

    Scandinavian Journal of Trauma, Resuscitation and Emergency Medicine 28(1):17

    Kang DY, Cho KJ, Kwon O, Kwon Jm, Jeon KH, Park H, Lee Y, Park J, Oh BH (2020) Artificial intelligence algorithm to predict the need for critical care in prehospital emergency medical services. Scandinavian Journal of Trauma, Resuscitation and Emergency Medicine 28(1):17

    2020

  9. [17]

    Kim SH, Whitt W (2014) Are call center and hospital arrivals well modeled by nonhomogeneous Poisson processes? Manufacturing & Service Operations Management 16(3):464--480

    2014

  10. [18]

    Operations Research 23(4):614--627

    Kolesar P, Walker W, Hausner J (1975) Determining the relation between fire engine travel times and travel distances in New York City . Operations Research 23(4):614--627

    1975

  11. [19]

    BMC Health Services Research 17:1--10

    Lake R, Georgiou A, Li J, Li L, Byrne M, Robinson M, Westbrook JI (2017) The quality, safety and governance of telephone triage and advice services–an overview of evidence from systematic reviews. BMC Health Services Research 17:1--10

    2017

  12. [20]

    Location Science 6(1-4):355--368

    Mandell MB (1998) Covering models for two-tiered emergency medical services systems. Location Science 6(1-4):355--368

    1998

  13. [21]

    Papers in Regional Science 71(3):217--241

    Marianov V, ReVelle C (1992) A probabilistic fire-protection siting model with joint vehicle reliability requirements. Papers in Regional Science 71(3):217--241

    1992

  14. [22]

    Journal of Regional Science 38(3):401--424

    Marianov V, Serra D (1998) Probabilistic maximal covering location—allocation models for congested systems. Journal of Regional Science 38(3):401--424

    1998

  15. [23]

    European Journal of Operational Research 135(1):195--208

    Marianov V, Serra D (2001) Hierarchical location-allocation models for congested systems. European Journal of Operational Research 135(1):195--208

    2001

  16. [24]

    Annals of Operations Research 111(1):35--50

    Marianov V, Serra D (2002) Location–allocation of multiple-server service centers with constrained queues or waiting times. Annals of Operations Research 111(1):35--50

    2002

  17. [25]

    IIE Transactions 41(8):730--741

    McLay LA (2009) A maximum expected covering location model with two types of servers. IIE Transactions 41(8):730--741

    2009

  18. [26]

    IIE Transactions 45(1):1--24

    McLay LA, Mayorga ME (2013) A model for optimally dispatching ambulances to emergency calls with classification errors in patient priorities. IIE Transactions 45(1):1--24

    2013

  19. [27]

    Technical report, U.S

    National Highway Traffic Safety Administration (2021) National Emergency Medical Services Information System ( NEMSIS ). Technical report, U.S. Department of Transportation, Washington, DC

    2021

  20. [28]

    Operations Research Perspectives 8:100202

    Nelas J, Dias J (2021) Locating emergency vehicles: Modelling the substitutability of resources and the impact of delays in the arrival of assistance. Operations Research Perspectives 8:100202

    2021

  21. [29]

    Health Care Management Science 12(1):67--79

    Restrepo M, Henderson SG, Topaloglu H (2009) Erlang loss models for the static deployment of ambulances. Health Care Management Science 12(1):67--79

    2009

  22. [30]

    Prehospital Emergency Care 7(4):434--439

    Shah MN, Bishop P, Lerner EB, Czapranski T, Davis EA (2003) Derivation of emergency medical services dispatch codes associated with low-acuity patients. Prehospital Emergency Care 7(4):434--439

    2003

  23. [31]

    Journal of the American Geriatrics Society 66(11):2213--2220

    Shah MN, Hollander MM, Jones CM, Caprio TV, Conwell Y, Cushman JT, DuGoff EH, Kind AJ, Lohmeier M, Mi R (2018) Improving the ED ‐to‐home transition: the community paramedic–delivered care transitions intervention—preliminary findings. Journal of the American Geriatrics Society...

    2018

  24. [32]

    Technical report, Ottawa, ON

    Statistics Canada (2020) The Open Database of Healthcare Facilities ( ODHF ). Technical report, Ottawa, ON

    2020

  25. [33]

    Proceedings of the 2022 IISE Annual Conference (Seattle, WA)

    Stratman EG, Boutilier JJ, Albert LA (2022) Patient assessment and response dynamics in emergency medical service systems. Proceedings of the 2022 IISE Annual Conference (Seattle, WA)

    2022

  26. [34]

    International Series in Operations Research & Management Science 213--250

    Stratman EG, Boutilier JJ, Albert LA (2023) Uncertainty in Facility Location Models for Emergency Medical Services . International Series in Operations Research & Management Science 213--250

    2023

  27. [35]

    Transportation Research Part E: Logistics and Transportation Review 133:101807

    Yoon S, Albert LA (2020) A dynamic ambulance routing model with multiple response. Transportation Research Part E: Logistics and Transportation Review 133:101807

    2020

  28. [36]

    Transportation Research Part E: Logistics and Transportation Review 152:102405

    Yoon S, Albert LA (2021) Dynamic dispatch policies for emergency response with multiple types of vehicles. Transportation Research Part E: Logistics and Transportation Review 152:102405

    2021

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.