Healthcare Decision-Support by Hybrid Simulation – Mobile Stroke Units

Problem

To understand both the medical and economic impact of new healthcare technologies, they must be evaluated before the design and development phase begins. Therefore, decision support system approached through hybrid simulation is applied in the case of Mobile Stroke Units (MSUs).

Stroke causes severe disability, produces high costs for care and rehab, and its incidences are increasing due to an ageing population. Thrombosis causes most strokes and if possible, should be treated with thrombolysis (not in case of haemorrhagic strokes and after 4.5 hours have been elapsed). Currently, the process of transportation and internal hospital administration causes the patient to lose valuable time. 

MSUs have been suggested as a possible improvement. MSUs start with diagnostics and therapy steps at a stroke occurrence location. The purpose is to reduce the alarm-to-therapy-decision-time in order to prevent severe disabilities of people and high costs.

Objectives

The goal is to assess the medical and economic impact of MSUs in comparison to the current situation and optimize the geographic distribution. Main metrics are the impact on the alarm-to-therapy-decision-time, costs, and other relevant output parameters. A metropolitan area (Berlin) and a rural region in Germany are used for comparison since there are few hospitals, and they are different structures.

MSU Simulation

Figure 1: MSU scenario animation screenshot

Implementation in AnyLogic

The project has been realized with AnyLogic; that is well suited to develop models using all necessary simulation paradigms and enables the developer to add own Java code to model unique and custom problem solutions. AnyLogic also allows the creation of illustrative animations in order to enhance communication with domain experts.

Figure 1 shows a screenshot of the MSU scenario animation. The MSUs are located on predefined positions and can be sent by the dispatcher when a stroke case occurs. If no free vehicle is available, a usual rescue service will be sent to bring the patient to the next hospital. Wrong decisions and unnecessary MSU missions (that do occur in reality) are modeled as well as cost values for the various actions and interventions. Stroke patients are monitored until 10 years after affection to enable long term cost analyses.

Excel tables with varied parameters are used as input files for different years and are loaded during simulation runs. To evaluate multiple scenarios of healthcare decision-support a library has been developed. Different modules (e.g., demography, error injector) can be connected in order to add certain characteristics to a dedicated scenario. A positioning tool allows defining home locations for MSUs and hospitals on a map by clicking on the area before running a scenario. In the selected papers below, further modeling and implementation features can be found.

Modeling Approach

To perform a simulation and particularly within the scope of healthcare decision-support, macro-simulation as well as micro-simulation approaches are necessary to increase model accuracy. This is where one can benefit from multi-method and hybrid simulation paradigms. System dynamics is used for modeling at a distant perspective, while discrete-event and agent-based techniques are appropriate for detailed modeling at individual levels. Patients are represented by agents, and medical workflows are embedded in system dynamics models which represent aspects as demography, economy, and epidemiology. The impact of medical technologies is represented by quantitative parameters.

MSU Models

Figure 2: Agent-Based, Discrete-Event and System Dynamics Model examples

Outcomes

One significant result from this case study is that MSUs do not automatically lead to more patients with thrombolysis as a treatment. However, those who are treated with thrombolysis receive care earlier, resulting in reduced probability for severe disabilities. This is a clear medical benefit. Research also shows that a broad distribution (e.g., uniform distribution) of MSUs on a map leads to better results in contrast to centralization on few locations (e.g., station, hospital). Furthermore, the simulation has shown that, in rural regions in Germany with only few hospitals, such vehicles are not profitable, as there are not many affections per year, and most of the people live close to an urban center with a hospital able to treat stroke. The above case study results may differ in countries with less specialized hospitals.

Conclusions

AnyLogic allows developing detailed models for healthcare decision-support. The configurable MSU simulation model helped the modelers to answer important questions about the impact of medical and health economics. Regulatory agencies, companies, researchers and other decision-makers can use the results to optimize an MSU roll-out in any country and to improve the stroke diagnostics and treatment in the future. Long-term monitoring of stroke patients allows the comparison of saved costs to the additional costs of MSUs, leading to a basis for an investment decision. This work has been conducted by Prospective HTA (funded by the German Government) in conjunction with doctors, engineers, and health economists, from both industry and academia.

The simulation of large-scaled complex systems (e.g., in healthcare, automotive, industry, and energy) is a major field of work at the University of Erlangen-Nuremberg Computer Networks and Communication Systems Group. Please contact model developers, if you have further questions.

Selected Publications

  1. Djanatliev A. and German R. “Prospective Healthcare Decision-Making by Combined System Dynamics, Discrete-Event and Agent-Based Simulation”. Proceedings of the 2013 Winter Simulation Conference, Editors: R. Pasupathy, S.H. Kim, A. Tolk, R. Hill, M. Kuhl. Washington D.C./USA. December 8th – 11th, 2013, p. 270-281.
  2. Djanatliev A., Kolominsky-Rabas P., Hofmann B. M., Aisenbrey A., German R. “System Dynamics and Agent-Based Simulation for Prospective Health Technology Assessments”. Simulation and Modeling Methodologies, Technologies and Applications - Advances in Intelligent Systems and Computing. Edited by Obaidat, Mohammad; Filipe, Joaquim; Kacprzyk, Janusz; Pina, Nuno (Eds.). Volume 256, 2014, pp 85-96.
  3. Djanatliev A., German R. “Large Scale Healthcare Modeling by Hybrid Simulation Techniques using AnyLogic”. Proceedings of the 6th International ICST Conference on Simulation Tools and Techniques. Cannes/French Riviera. March 5th – 7th, 2013, p. 248-257.

Similar case studies

More case studies