Innovative Applications of O.R.
A statistical Markov chain approximation of transient hospital inpatient inventory

https://doi.org/10.1016/j.ejor.2010.06.021Get rights and content

Abstract

Inventory levels are critical to the operations, management, and capacity decisions of inventory systems but can be difficult to model in heterogeneous, non-stationary throughput systems. The inpatient hospital is a complicated throughput system and, like most inventory systems, hospitals dynamically make managerial decisions based on short term subjective demand predictions. Specifically, short term hospital staffing, resource capacity, and finance decisions are made according to hospital inpatient inventory predictions. Inpatient inventory systems have non-stationary patient arrival and service processes. Previously developed models present poor inventory predictions due to model subjectivity, high model complexity, solely expected value predictions, and assumed stationary arrival and service processes. Also, no models present statistical testing for model significance and quality-of-fit. This paper presents a Markov chain probability model that uses maximum likelihood regression to predict the expectations and discrete distributions of transient inpatient inventories. The approach has a foundation in throughput theory, has low model complexity, and provides statistical significance and quality-of-fit tests unique to this Markov chain. The Markov chain is shown to have superior predictability over Seasonal ARIMA models.

Section snippets

Introduction and motivation

Inventory predictions are critical to the operations, management, and capacity of throughput systems. Throughput system complexities including heterogeneous entity types, non-stationary arrival and service processes, and limited production visibility inhibit the application of historical inventory prediction models. Examples of such systems include complex manufacturing systems, organ banks, and blood banks. Hospital inpatient (IP) units are prime examples of very complex throughput systems as

Literature review

The review of the literature generally categorizes relevant transient inventory prediction literature as either a throughput model (e.g. Markov chains, queuing theory, computer simulation) or an empirical statistical model (e.g. time series, regression, summary statistics). Some throughput models include non-stationary arrival and service processes. All throughput models require explicit prior knowledge of the arrival and service process, rates, and non-stationary behavior. No empirical

DTCM model formulation and statistical fitting

This section presents a DTMC probability formulation that captures the short term effects of inventory fluctuations in an IP throughput system. The formulation incorporates hourly seasonality (non-stationarity) in both the arrival and service rate. Because the service rate is assumed unknown by hour of the day, the probability model is combined with maximum likelihood regression to best estimate hourly service rates. Statistical significance and model validation tests specific to this

DTMC validation and model performance in an IP unit

In this section, we explore the validation and performance of the DTMC described in Section 3 using data from two IP units. We consider an intensive care unit (ICU) and a telemetry unit (TELE) from one of Banner Health’s hospitals in Arizona. In this particular hospital, the TELE is significantly larger, receives higher patient volume, and typically has longer treatment times than the ICU. Descriptions of the hospital’s ICU and TELE units are presented in Table 3.

The data used are the IP hourly

Conclusion and future work

This paper presents a DTMC for generically predicting transient inventory in complex, non-stationary, heterogeneous inventory systems. The system is motivated by the hospital IP unit. Unlike much of the IP inventory modeling literature, this formulation has a throughput theory foundation (not strictly empirical) and does not have high complexity. Unlike much of the throughput modeling literature, this research captures the transient inventory in a non-stationary arrival and service process

Acknowledgements

This research is support by Banner Health, Phoenix, AZ. The authors would like to thank Twila Burdick, VP of Organizational Performance, and Management Engineering at Banner Health.

References (37)

  • G.M. Frankfurter et al.

    Management control of blood through a short-term supply-demand forecast

    Management Science

    (1974)
  • P.R. Harper

    A framework for operational modeling of hospital resources

    Health Care Management Science

    (2002)
  • P.R. Harper et al.

    Modelling for the planning and management of bed capacities in hospitals

    Journal of the Operational Research Society

    (2002)
  • G. Hsu et al.

    Transient solutions for denumerable-state Markov processes

    Applied Probability Trust

    (1994)
  • M.W. Isken

    Modeling and analysis of occupancy data: A healthcare capacity planning application

    International Journal of Information Technology and Decision Making

    (2002)
  • M.W. Isken et al.

    Data mining to support simulation modeling of patient flow in hospitals

    Journal of Medical Systems

    (2002)
  • E.P.C. Kao et al.

    Incorporating exogenous factors in adaptive forecasting of hospital census

    Management Science

    (1978)
  • K. Katsaliaki et al.

    Using simulation to improve the blood supply chain

    Journal of the Operational Research Society

    (2007)
  • Cited by (38)

    • Modeling COVID-19 hospital admissions and occupancy in the Netherlands

      2023, European Journal of Operational Research
      Citation Excerpt :

      The studies (Broyles, Cochran, & Montgomery, 2010; Kortbeek, Braaksma, Burger, Bakker, & Boucherie, 2015) mainly focus on hourly seasonality; in Kortbeek et al. (2015) the impact of the master surgical schedule is taken into account, leading to crucially different dynamics in patient admissions. Broyles et al. (2010) uses a discrete-time Markov chain to model patient inventory and then applies maximum likelihood regression. In (Littig & Isken, 2007) a predictive occupancy database is used, in which the data of each patient within the hospital is registered; our predictions are based on aggregated data.

    • Analysis and optimization of replenishment process for robotic dispensing system in a central fill pharmacy

      2021, Computers and Industrial Engineering
      Citation Excerpt :

      A discrete-time Markov Chain has a finite, or countable, state space, while continuous-time Markov Chains are constantly monitored and updated through time (Pinsky & Karlin, 2010). Broyles, Cochran, and Montgomery (2010) presents a discrete-time Markov Chain probability model that uses maximum likelihood regression to show that the predictability of the Markov Chain outperforms other modeling methods with short prediction lead times. Other discrete-time Markov Chains model complex inventory systems and accurately represent variations in the system (Abboud, 2001; He & Jiang, 2018).

    • Strategic room type allocation for nursing wards through Markov chain modeling

      2019, Artificial Intelligence in Medicine
      Citation Excerpt :

      Bartolomeo et al. [17] applied a Discrete-Time Markov Chain (DTMC) model to assess the re-admission probability of patients. Further, Broyles et al. [18] applied a DTMC to predict the number of inpatients, demonstrating how their model attains superior predictability compared to a seasonal Autoregressive Integrated Moving Average model. Concerning CTMCs, both He et al. [19] and Shao et al. [20] developed a model to assess and identify bottlenecks with regard to the colonoscopy screening process and surgical operations in an emergency department, respectively.

    • Optimization of hospital ward resources with patient relocation using Markov chain modeling

      2017, European Journal of Operational Research
      Citation Excerpt :

      A range of studies, using Markov chains to model patient flow, have already been conducted (Bartolomeo, Trerotoli, Moretti, & Serio, 2008; Broyles, Cochran, & Montgomery, 2010; Deng, Deng, Deng, Liu, & Liu, 2015; Xie, Li, Swartz, & Dong, 2013; Zhang et al., 2013). As relevant examples, Broyles et al. (2010) predicts distribution and expected admissions, and Bartolomeo et al. (2008) determines the probabilities of readmission for two different patient categories. However, none of these exploit the advantages of Markov chains, to model patient relocation, and subsequently use these models to optimize the system.

    • Flexible nurse staffing based on hourly bed census predictions

      2015, International Journal of Production Economics
      Citation Excerpt :

      Workload encountered in nursing wards depends heavily on patient arrivals and lengths of stay, both of which are inherently variable. Predicting workloads and staffing nurses accordingly are essential for guaranteeing quality of care in a cost effective manner (Broyles et al., 2010; de Véricourt and Jennings, 2011). Accurate workload predictions require that the dynamics of surrounding departments are considered, given that many patient arrivals at the inpatient care facility originate from the operating theater and the emergency department.

    View all citing articles on Scopus

    The views expressed in this article are those of the authors and do not reflect the official policy of the United States Air Force, Department of Defense, or the United States Government.

    View full text