Highlights of research accomplishments  

I love to collaborate, and almost all of my research has involved graduate students and sometimes other researchers. With the caveat that these accomplishments are not solely mine, here are some highlights: 


 Mark Van Oyen’s education in stochastic modeling, control, and estimation and OR inspired him to develop methods to replace many static and inflexible operations with nearly optimal dynamically monitored and controlled operations. This vision has shaped his leadership in agile/flexible operations, next generation systems for healthcare, and monitoring and control of chronic diseases. Two seminal papers used dynamic “queueing control” as an alternative to traditional ad hoc approaches to polling systems:

  • [Q-CTRL-1] Heuristic scheduling of parallel heterogeneous queues with set-ups, Man Sci 1996
  • [Q-CTRL-2] Stochastic scheduling of parallel queues with set-up costs, Queueing Sys 1995.

In seminal contributions to the science of operational flexibility, Van Oyen addressed labor flexibility (cross-training, economy of scope, and policies for worker coordination) and production lines, cellular manufacturing, plants, and supply chains.He generated the notions of Structural Flexibility (SF) and capability flexibility as predictive indices to guide the design of resilience in production. Among others, see

  • [FLEX-1] Agile workforce evaluation: a framework for cross-training and coordination, IIE Trans 2004
  • [FLEX-2] Structural flexibility: A new perspective on the design of mfg. and service operations, Man Sci 2005
  • [FLEX-3] Compensating for dynamic supply disruptions: Backup flexibility design, OR 2016.  

With many collaborators, he created state of the art methods bridged operations to individual patient needs. He devised new solutions for many key areas of health system operations: Emergency Department (ED) redesign; “in-advance” appointment scheduling in multiple contexts; system-wide patient flow and operations eng.; clinical research units OM; coordinated care for surgery; online/self-serve appointment systems; ward/unit shift design plus assignment; design of skills-based nursing teams and staffing; and surgical case duration prediction, start time setting, and scheduling.

Surmounting the decades old barriers to network models of stochastic hospital patient flow, his 1st place award winning paper [HOSP] tractably modeled time-correlated patient flow itineraries through hospital and increased useable capacity while placing more patients in the proper bed unit. His emergency department (ED) “virtual streaming” paradigm based on new triage information may reduce waits by around 20%, and won the INFORMS Pierskalla Best Paper Award [EMRG-1]. His paper, [EMRG-2], on patient complexity assessment at triage cut adverse health outcomes 15%, winning Best Paper awards from MSOM journal for 2013-2015 and the 2016 MSOM Service Op’s S.I.G.  Van Oyen is a world leader in advancing methods to achieve targets for the access delay from request date to the appointment date, with different target for each tier as indicated by urgency. In this nascent area, he generated four papers treating Clinical Research Units, integrated outpatient care, and coordinated surgical care. He advanced personalized medical decision making with the operational issues of the timing of office visits and diagnostic tests (monitoring) in addition to predicting disease progression [MDM-1]. This research won the 2012 Doing Good with Good OR 1st Prize and a 1st prize for POMS best healthcare OM paper. Stochastic control methods jointly optimizing medical control of risk factors and the timing of costly diagnostics in [MDM-2] were key to his student earning 2nd place in the INFORMS Dantzig dissertation competition. More recently, we developed the first contextual multi-armed bandit online learning method for joint personalized optimization of treatment/medication and the nested decision of the dosage for blood pressure control (see Keyvanshokooh, E., M. Zhalechian, C. Shi, M.P. Van Oyen, and P. Kazemian, Contextual learning with online convex optimization: Theory and application to chronic diseases. Minor Revision at Management Science 2022. Available at SSRN: https://ssrn.com/abstract=3501316).

  • [EMRG-1] Patient streaming as a mechanism for improving responsiveness in emergency departments, OR 2012
  • [EMRG-2] Complexity-augmented triage: A tool for improving patient safety and operational efficiency, MSOM 2014
  • [HOSP] Design and optimization methods for elective hospital admissions, OR 2014
  • [MDM-1] Dynamic forecasting and control algorithms of glaucoma progression for clinician decision support, OR 2015 
  • [MDM-2] Dynamic monitoring and control of irreversible chronic diseases with application to glaucoma, POM 2019

At the time of writing this paper, the concept of using a “chain” structure of multi-functionality to increase production system performance was still relatively new. Anecdotally, Toyota first identified the power of chaining, but General Motors was the first to publish on it. Both GM and Ford developed the concept and exploited it in multiple ways, including investing billions in mixed-model manufacturing plants. The extant research at that time focused on a set of plants in parallel as in the automotive plant application; however, this paper was the first serious treatment of the opposite extreme – a linear, serial production line (whether manufacturing or service sector). This paper made significant claims that break with conventional practice and teaching. It considered the common practice of using worker cross-training to add the skill required at the bottleneck station to one or more workers to increase the bottleneck capacity, which increases its capacity. Note that we speak of human cross-training, but the essential feature is multi-purpose resources, which can include multi-product factories designed with flexible workstations or multi-functional machines or robots. Conventional wisdom suggests investment at the bottleneck station (with the lowest capacity). This is called “cherry picking”, because it invests resources focusing only on the station with the greatest need for additional capacity. Provided each worker (assumed to be a specialist), can be cost-effectively cross-trained with the skill to work on an adjacent task, this strategy termed “chaining” intentionally cross-training flexibility at every station. This paper helped establish the chain design as being an “ideal” paradigm, because it has a high level of flexibility while frugally limiting the number of additional skills beyond a traditional system of specialized workers. In settings with significant variability (i.e., demand or processing time fluctuations), the study suggested that chaining strategies are usually preferable to cherry picking. We focus on settings with a queue at each workstation to hold waiting work/jobs in process (WIP), so any worker with more than one skill must follow some resource allocation policy that choses which skill to apply at any time. The skill chain allows it to perform very well across a variety of simple, implementable WIP-based worker coordination policies. The flexibility created by skill chaining is so great that the throughput for a given WIP level breaks the near-universal pattern of marginally diminishing returns in the number of skills added due to high marginal benefit from the addition of the final chain-completing skill. While not always feasible in practice, chaining has been of great interest, and developments like U-shaped lines helped facilitate their adoption. This paper was a finalist for the INFORMS Manufacturing and Service Operations Management (MSOM) Best Student Paper Award.  Dr. Van Oyen later broadened the space of canonical flexible structures/designs that could be quantified to increase production flexibility and performance. The next paper [FLEX] is one of those efforts.

  • Hopp, W.J., E. Tekin, and M.P. Van Oyen, Benefits of skill chaining in serial production lines with cross-trained workers, Management Science, 50:1, (2004) 83-98. https://doi.org/10.1287/mnsc.1030.0166  

Unlike priority-based emergency care, most medical services are advance-scheduled appointments that do little to provide earlier care to relatively more urgent patients. A simple First-Come-First-Served scheduling logic (with deviations based on preferences) pervades healthcare. Dr. Van Oyen is the global leader in developing potent, generalizable new methods to use capacity efficiently while meeting patient type-specific targets for patient access delay/wait. He achieved a next-generation scheduling innovation through novel methods to transform these stochastic optimization problems into mixed integer optimization models that reserve appointment slots over time in a calendar. More urgent patient types receive an earlier appointment date, and the optimization models respect the limited clinician time and working hours of a department, thereby accommodating high patient throughput. The idea of creating stratified tiers of waiting time to begin treatment was developed after the Michigan Clinical Research Unit (MCRU) asked Van Oyen to help reduce excessive overtime expenses. MCRU has been a hub for conducting large numbers of clinical research trials, mostly using their own facility at that time. Many of their trials required the participant to make multiple visits over weeks or months with tight requirements on the timing of the carefully controlled treatments (frequently an infusion). The central decisions were (1) determining which of a number of proposed trials could be accepted without excessive conflicts with existing patients/participants needing precise timetables for future visits, and (2) computing a calendar with pre-allocated, optimized appointment slots guiding when new trial enrollees could schedule appointments to initiate treatment. Optimization of clinical research nurse staffing assignments was also a part of the optimization model. The optimization ensured that future participants in each trial would obtain their first visit within a promised number of days from enrollment to first visit (i.e., maximum access delay) based upon the trial’s justifiable urgency level. He was the lead PI of a 3-year NSF grant of $420,000 entitled “Stochastic Modeling and Optimization of Longitudinal Health Care Coordination” that funded this work. The key paper won a 2nd  Place Best Paper Award from the Production and Operations Management Society. Plans for a homegrown system were laid, but the switch to the Epic hospital management system prevented it. We filed the invention disclosure “CApacity Planning Tools and INformatics (CAPTAIN) Decision Support System for Phase I Trials Performance Sites,” Sept. 14, 2011. IR# 5152. 

                  This innovation was further developed in two projects with Mayo Clinic.  The first was on the computation of physician scheduling templates across a department to provide slots organized by patient type/tier so that a patient in a more urgent tier has a shorter limit on their wait for their out-patient visit. The “integrated care” system of Mayo Clinic was modeled, capturing the network effects of various departments with rapid referrals between them. This prevents changes in one department from unintentionally hurting another. Simulation showed the ability to serve 31% more patients while reducing urgent patient waits by 3.2 weeks at the cost of adding only 6 days of wait for non-urgent patients.   See  See Deglise-Hawkinson, J, David L. Kaufman, B.J. Roessler, and M.P. Van Oyen, Access Planning and Resource Coordination for Clinical Research Operations, IISE Transactions, 52:8, 2020. https://doi.org/10.1080/24725854.2019.1675202 doi.org/10.1080/24725854.2019.1675202  (NIHMS1543900 – PMID)

                  The innovation also allowed us to address a surgical service at the Mayo Clinic. Patients experienced an access delay of so many weeks for a clinic visit consultation that many gave up and went to other providers. The problem was ineffective patient scheduling that wasted physician time. The goal was to reduce the time of visit request until the time of surgery (i.e., access delay), and to have more urgent patient types wait less than those less urgent. All patients first have a clinic visit to confirm the need for surgery. The surgery must be with the same physician to achieve coordination of care. The new system created 4 tiers of access delay limits at 3, 10, 20, and 40 days from request for a visit until the day of surgery. Due to the technical difficulty, this was the first system having this with this capability, to the best of our knowledge. The system simulation suggested the solution would achieve the desired targets with far fewer wasted physician appointment slots, while giving greatly reduced waits for surgery to each urgency tier. After our proof of concept paper, Mayo Clinic engineers revised the method to be simpler and implemented that decision support system in-house to great success. Our first paper employed ML, heuristics, and some optimization, but in a later paper, Van Oyen’s group developed a methodologically more powerful and general optimization approach for such problems.  

  1. Kazemian, P., M.Y. Sir, M.P. Van Oyen, J. Lovely, D. Larson, K. Pasupathy, “Coordinating Clinic and Surgery Appointments to Meet Access Service Levels for Elective Surgery,” of Biomedical Informatics 66, 105-115, 2017. dx.doi.org/10.1016/j.jbi.2016.11.007 PMID: 27993748
  2. Keyvanshokooh, E., P. Kazemian, M. Fattahy, M.P. Van Oyen, Coordinated and Priority- based Surgical Care: An Integrated Distributionally Robust Stochastic Optimization Approach, Production and Operations Management (POM) 31:4, 2022, 1-11. https://doi.org/10.1111/poms.13628