Introduction of computation in the healthcare field has opened the gate for continuous amazing discoveries in modern medicine and improved diagnostic and therapeutic approaches. Predictive computational modelling is used in different fields, such as marketing, weather forecasting, and resource management. While there is a great need for accurate and prompt intervention to treat many diseases, healthcare providers are reluctant in using computational models in their daily work to manage their patients. That is, however, understandable because they are worried about the accuracy of such models, especially if the prediction may suggest shifting the patient’s therapeutic regime one way or another. In this article, I aim to explore the types of computational models, and how a clinician can build a successful model, which they may use as an irreplaceable tool in their daily patient care routine.
What are computational models?
A model is built to simulate how a system works, like building a model airplane or a car. So, using the computer to build a model that simulates the functions of a system (how it works) is generally what a computational model is. The model’s concept is built on a scientific basis, which may include mathematical equations, statistical analysis, or biomedical rules obtained from scientific literature or a combination of different scientific disciplines. A model does not have to be an exact replica of the system it aims to simulate, it just needs to simulate simply how it works and produce the outcomes in a manner like the original system or close enough to provide a useful understanding for the user about the process in order to take a proactive approach in anticipation to the outcome of the simulation. A model aircraft or a model car does not have to have an AC or a top of the line sound system to simulate the damage of a car crash or how air turbulence affects the stability of an aircraft.
Why build models?
Computational models have many appealing benefits for healthcare providers and health researchers. Computational models can be used to perform In-silico experiments. When a model is valid and well-designed, it can perform thousands of simulations in a short time and at negligible cost. For example, a model simulating the response of cancer cells to treatment can run many simulations in minutes costing only the hardware, compared to performing actual wet-labs experiments that require funding and hundreds of hours to perform the same experiments. Computational simulations can help in testing medications in the safe – virtual – environment; it can also bring different understandings into the causative factors of different physiological processes or triggering factors in disease conditions.
As health professionals, how to build our model?
As healthcare professionals, we have a common preconception that this computational model is beyond our understanding or capability. That notion is inaccurate because many of the successful health-related computational models are built by teams that involve health professionals who do not have any programming skills or technical expertise.
What is the proper design method?
There are different methods to design computational models. Conventional methods are mathematical, statistical, or agent-based models. It is essential to choose the proper approach carefully to simulate the system or the condition of interest.
Mathematical models are designed based on complex mathematical equations. The equations are used to estimate and calculate the parameters of the factors involved in the simulation. Those models require a professional mathematician in the team to create the appropriate equations.
Statistical models use statistical methods such as regression analysis or ordinary differential equation to determine the most relevant factors involved in the system to include them in the model. Statistically based models require a large sample size during the factors’ analysis and later during the training of the model. Statistical-based methods can be used to create models that analyse x-rays, or radiological images, to diagnose bone fractures or brain tumours.
Agent-based models (ABM) treat the active components in a system as agents. Each agent has a life-span, rules of interactions with other agents, and specific attributes. ABM’s require to have the system or the process to be simulated well-understood in the literature. It is better to build the model using the most accurate parameter values from the literature and have a professional on the subject who works closely with the team – or a part of it – in order to have the proper model’s design. ABM has been used to simulate many systems that range from cellular interactions and wound healing to the simulation of disease spread in a community.
How to start?
Like any research project, a thorough literature review is required to understand the process or the system in interest. The process must be generally explained well enough to build a sound model. Not all the details are required nor are expected to be available in the literature. If we want to simulate the digestion of a material in the stomach, for example, we will find some missing values such as the amount of enzymes or acidic secretions required to digest a gram of the material. This is when we need to go and start building the equations to estimate that number. If the process is not well-explained in the literature, the design step is going to be difficult, and the model will be criticised for being a technical experiment rather than a health directed solution.
After having the process mapped with all the steps and values, select the proper design method according to the resources available. A healthcare professional who is an expert on the system or process being simulated is an important member of the team. Such an individual can advise the team and provide the much-required feedback to ensure the model’s design credibility.
A healthcare professional who is an expert on the system or process being simulated is an important member of the team.
The model was built, now what?
The computational model, like any other software tool, requires much testing to ensure that the model is appropriately running. It may be a good idea to perform usability tests to make it more acceptable for healthcare providers.
The most important step comes next: validation. The team must carefully determine the proper approach to validating the model. In many cases, where the model is simulating a disease condition or a response to treatment, there must be a proper sample of actual cases available. For example, if we aim to simulate the effect of a drug on reducing the inflammation, we must have recorded cases of individuals who received that drug (control) and compare the outcome to the virtual case where the virtual patient is matching the criteria of the control. By comparing the outcome values of the simulation, e.g., inflammatory cells, body temperature to the reported values from the case, and statistically comparing the results using the proper tests, the team can decide whether the model is valid or not. The validation process must be performed and well documented. The model must also go through the other steps for any tool, meaning: reliability, sensitivity, and specificity.
Finally, like any software, the model must be updated and tested regularly. Adding or removing factors in simulations, testing new drugs, or other required changes due to new information available in the literature helps to keep the model trust-worthy by the clinicians and users alike.
Computational models and simulations are being used in all industries but are not well utilised in healthcare. We should move in and take this solution and create our models instead of waiting for companies to sell us packaged solutions that will need many customisations. In this article, we simply reviewed the general concepts and requirements for health professionals to build their model. Health professionals – it is time about time to build our own computational models, and its that simple.
References available on request.