Systems thinking in a health care world

How modeling and simulating the complex, dynamic environment of Neonatal Intensive Care Units can lead to more sustainable solutions.

Written by Wim Rill, MSc

Introduction

A systems thinking approach in Neonatal Intensive Care Units (NICUs) would preferably involve considering the entire ecosystem surrounding the care of premature or critically ill newborns. This includes variables related to medical care coordination, parental support, technology integration, environmental factors and follow-up care. By considering these interconnected elements, a systems thinking approach in NICUs can lead to better health outcomes for the infants, reduced stress for families, and more efficient use of healthcare resources.

In a survey at the NICU of Isala all alarm frequency data between 1 February and 31 March 2024 was retrospectively collected. A total of 678,212 alarms were recorded. Despite their intended purpose of prompting timely responses and clinical actions, a substantial proportion (72–99%) of these alarms are deemed non-actionable (Leenen et al., 2024). The alarm pressure can be considered high and might in some cases be too high and lead to alarm fatigue. Alarm pressure and alarm fatigue can be seen as two of the variables in this complex, dynamic system of NICUs. But, how can we know all the variables in this system and how are they connected?

Linear versus systems thinking approach

A NICU is a highly specialized environment dedicated to the care of critically ill newborns. The research group IT Innovations in Healthcare of Windesheim University of Applied Sciences started in cooperation with Isala Hospital in Zwolle the project digital TWins fOr Simulating neOnatal Monitoring tEchnologies (TWOSOME). In this replica of the physical, dynamic NICU setting, data from patients, devices and their interactions can be simulated, and artificial intelligence can make predictions about the impact of changes (Hakvoort, 2023). The focus was on the immediate medical care of the newborns with advanced monitoring technologies which are intended to alert the health care workers to potential issues that may require immediate clinical intervention. The monitoring systems continuously assess parameters such as heart rate, respiratory rate, blood oxygen saturation and blood pressure and give an alarm if a certain threshold is reached.

We often seek straightforward solutions without considering the broader system and how different variables interact. However, this can lead to unintended consequences and sometimes make the problem worse over time. This concept is often referred to as “linear thinking” versus “systems thinking.” Linear thinking focuses on direct cause-and-effect relationships, while systems thinking considers the complex inter-dependencies within a system. The TWOSOME research can be considered as a linear approach. It was mainly based on the assumption that if we can reduce the number of alarms (potential) alarm fatigue will be reduced as well. This reductionist approach leaves a lot of questions unanswered, such as: To what extent does alarm pressure influence alarm fatigue? What other variables influence alarm fatigue and to what extent? Is every health care worker influenced by the alarm pressure to the same amount? Will the same amount of alarm pressure have the same influence on a health care worker every day? To what extent has the mental, physical and emotional state of the health care worker an influence on to what extent the alarm pressure leads to alarm fatigue? To what extent is the group of health care workers of importance? If we want to be able to answer these questions and be able to make more accurate predictions for every single newborn a systems thinking approach is needed.

Causal Loop Diagram

By adopting a systems thinking approach, we can better anticipate on the effects of our actions and develop more sustainable solutions. During the TWOSOME project a system dynamics[1] analysis was made in cooperation with seven NICU health care workers. The first step was to model a Causal Loop Diagram (CLD). The CLD (figure 1 and 2) shows a mental model of the system.

Figure 1: Causal Loop Diagram NICU Isala GMB Health Care Workers (partly English)

The CLD gives an overview of the important variables mentioned by the group in relation to alarm fatigue and how they correlate. The CLD does not give answers to the amount of influence the variables have on each other. For that the CLD needs to be translated in a Stock Flow Diagram (SFD). A SFD can be used as a simulation model. A first draft is presented in figure 3.

When the correlation between two variables is presented with a plus it means that they both develop in the same direction. For example if the ‘Reaction time health care worker’ increases the ‘Alarm pressure’ will increase as well, and if the ‘Reaction time health care worker’ decreases the ‘Alarm pressure’ will also decrease. If the correlation between two variables is presented with a minus the variable develops in the opposite direction. For example if ‘Alarm fatigue health care worker’ increases the ‘Load capacity health care worker’ will decrease, and if ‘Alarm fatigue health care worker’ decreases the ‘Load capacity health care worker’ will increase. The CLD in figure 1 is presented to give an overview of all the variables and their interactions the group has found so far. In total eight feedback loops[2] can be identified.

Figure 2: Four positive feedback loops around the variable ‘Alarm fatigue health care worker’.

In figure 2 the central part of the CLD with the main variables is presented. In this part of the CLD four reinforcing (=positive) feedback loops can be identified. One of the positive feedback loops can be explained as follows: More ‘Alarm fatigue health care worker’ will lead to less ‘Load capacity of the health care worker’ and less ‘Net capacity health care workers’. The ‘Reaction time health care worker’ will increase and so will the ‘Alarm pressure’ (because less alarms will be followed up) and over time (dotted line) the ‘Alarm fatigue health care worker’ will increase with will lead to less ‘Load capacity health care worker’ and so on. This means that if a health care worker becomes alarm fatigue it will get worse over time if nothing else changes for the better. Only if variables in the positive loop are changed in the good direction the positive loop can lose strength. ‘Alarm fatigue health care worker’ can directly be decreased by increasing for example the ‘Knowledge of patient through alarms’ and/or ’Ability to estimate the value of alarm’. By increasing the load capacity of the health care worker on a physical, mental or emotional level the overall ‘Load capacity health care worker’ will increase thus decreasing the strength of the positive feedback loop. This might however not be of value equally for every health care worker. A personal approach is needed.

Stock Flow Diagram

Figure 3: First draft of a Stock Flow Diagram of the CLD NICU Isala GMB health Care Workers.

In figure 3 a first draft of the SFD of the CLD NICU Isala GMB health Care Workers (figure 1) is presented. The SFD contains many stocks[3] (like ‘Alarm fatigue health care worker’).

The variable ‘Alarm pressure’ is influenced by six stocks with the alarms of Elektrocardiogram — hartfrequency — ECG, Noninvasive bloodpressure — NiBD, Saturation — SpO2, Invasive bloodpressure — ABP, Respiration — Resp and Temperature — Temp. There is always a number of alarms for every infant. Each of the stocks has an inflow and outflow. For example the number of alarms from Elektrocardiogram — hartfrequency — ECG has an inflow from the ECG’s Red, Blue and Yellow. The inflow can be reduced by Reduction ECG — adapting lower limit hart frequency.

Every connection will be translated in a mathematical equation. The SFD can then be used as a simulation model. By use of this simulation model one can get a better and more quantitative answer to questions asked before (see: Linear versus systems thinking approach). The monitoring systems continuously assess parameters. The number of alarms will have an impact on the alarm pressure. On the basis of (empirical) research the mathematical equations can be formulated. The ‘Reaction time of the health care worker’ will have an impact of the number of alarms. The longer it takes before action is taken the longer the alarms will go off. This will likely differ per parameter.

When all connections are formulated in mathematical equations simulation can take place. What if for example will happen in the system if we are able to reduce the alarm fatigue with 10%. Will this have a lasting impact?

References

Hakvoort, G. (2023). Projectvoorstel KIEM 2023. Indieningsronde februari 2023.

Leenen, J.P.L., Kemmink, C.A.M.C., Dam-Vervloet, L., Mulder-de Tollenaer, S.M., Blanken, M.O. (2024). Multidimensional analysis of alarm management on the NICU: evaluating alarm frequencies, responsiveness, and nurse perceptions in a single-centre prospective observational study.

System Dynamics Society. (2024). https://systemdynamics.org/what-is-system-dynamics/ Downloaded 31–03–2025.

[1] System Dynamics is a computer-aided approach for strategy and policy design. The approach provides methods and tools to model and analyze dynamic systems. Model results can be used to communicate essential findings to help everyone understand the system’s behavior. It uses simulation modeling based on feedback systems theory that complements systems thinking approaches (System Dynamics Society, 2024a).

[2] A feedback loop exists when information resulting from some action travels through a system and eventually returns in some form to its point of origin, potentially influencing future action. If the tendency in the loop is to reinforce the initial action, the loop is called a positive or reinforcing feedback loop; if the tendency is to oppose the initial action, the loop is called a negative or balancing feedback loop (System Dynamics Society, 2024).

[3] Stocks (levels) and the flows (rates) that affect them are essential components of system structure. A map of causal influences and feedback loops is not enough to determine the dynamic behavior of a system. A constant inflow yields a linearly rising stock; a linearly rising inflow yields a stock rising along a parabolic path, and so on. Stocks (accumulations, state variables) are the memory of a dynamic system and are the sources of its disequilibrium and dynamic behavior (System Dynamics Society, 2024b). Some examples: 1) the amount of water in a bathtub. By opening the valve water flows in. The level of water will increase, but just slowly. By opening the sink the water level will decrease, also just slowly. 2) the money on your bank account. If you get your salary it will increase. If you pay for a dinner it will decrease. 3) the level of CO2 in the sky. If we burn fossil fuel it will increase. CO2 will decrease by photosynthesis.

Acknowledgments

This work is part of the TWOSOME-project, a collaboration between the Isala hospital, the software company Little Rocket, Windesheim University of Applied Sciences, and has been made possible through a grant from SIA (KIEM.K23.01.135). Special thanks goes out to Lida Dam-Vervloet for her invaluable input and to Paul Hiemstra and Gido Hakvoort for providing feedback on drafts of this article.