Amaury Jeandrain
Which AI Method for Clinical Supply?
[EPISODE 3] - Mathematical Optimization
Imagine a world where every decision is the best possible one, leading to optimal outcomes every time. This is what Mathematical Optimization aims at.
What is Mathematical Optimization?
Mathematical Optimization has not received the widespread attention that Generative AI has.
The first half of the term can be scary. It might even remind you of your maths classes at school, and seem as complex. The second half of this concept is probably one of the most overused current buzzwords in a professional setting.
In any case, let’s try to understand what Mathematical Optimization means, what value it brings to clinical supply chain management, and how it differs from other AI techniques.
Mathematical Optimization is, in itself, a powerful AI technique. It involves finding the best solution (maximum or minimum) to a problem, or the best way to reach a certain objective, within a defined set of constraints. The solution itself is often a set of decisions.
Mathematical Optimization applications in clinical supply chain management
To illustrate, and simplify, Mathematical Optimization could be used in clinical supply chain to solve the following problem:
|
You want to define an optimal supply chain plan to supply your clinical trial without risks, but you want to do it with a minimal cost.
|
But the applications are endless. Let’s look at another potential application:
|
You want to identify the optimal sizes of the shippers/shipping boxes used to supply sites’s demand efficiently, while minimizing costs, inventories and workload for your team.
|
Commons uses of Mathematical Optimization
Naturally, Mathematical Optimization is often used for:
1. Resource Allocation, Scheduling and Planning
AI systems often rely on optimization algorithms to:
- allocate resources efficiently (e.g. split a batch between trials),
- schedule tasks (e.g. optimize the use of a bioreactor),
- optimize operations (when to send a shipment, and what would be its content), as illustrated by example 1.
2. Decision-Making
AI systems use optimization algorithms to make decisions in complex environments, as illustrated by example 2.
The tradeoff we need to accept with Mathematical Optimization lies in the fact that they require complex models and very extensive calculation power. Therefore, they tend to be used in very complex, expensive or critical (e.g. risky) environments. Is clinical supply chain management not well described by these three words?
GenAI, Simulations, and Mathematical Optimization
Let’s compare mathematical optimization to the two previous techniques we analyzed (GenAI and Simulations).
Simulations and GenAI can both be considered as predictive analytics. Predictive analytics describe what is most likely to happen. Simulations, per nature, will simulate a process, predicting its future state. GenAI (and Machine Learning, more specifically) on the other hand, will use past data to predict potential futures.
Mathematical Optimization are considered as prescriptive analytics. Prescriptive analytics show you how to make it happen the way you want it. Let’s illustrate the difference with another example:
|
You want to make the right decision and your objective is to know how much drug to produce while reducing the costs of your clinical supply chain management and ensuring no drug shortages happen. |
Simulations
Without an optimization algorithm, simulations will show you what will happen in your clinical supply chain in the future, should every decision be set in stone. It will show you whether you have risk, show you your waste, demand projections, and more.
But if you wanted to use simulations alone to make decisions, you would need to test scenarios (e.g. What would happen if I produced 500 kits, what if it were 400 kits?). After multiple scenarios, you might find an amount that works (hint: but not the optimal amount).
Gen AI and ML
GenAI and Machine Learning could support you in analyzing the trial and supply chain designs and, based on past observations, inform you on whether certain factors are likely to increase or decrease the amount of drug you need for your clinical trial. But it will never be able to give you an accurate recommendation of the number of kits to produce.
Mathematical optimization
The very clear answer from an optimization model to this program will be “you need to produce exactly X amount of drug to minimise your costs and avoid risks”.
While an optimization model will quantify the best possible solution, a well-trained AI could indicate which solutions might be leaning towards the objective, without ever being able to accurately quantify it or provide all the decisions needed to get there.
Nonetheless, the highest value is extracted when these models are used together.
Due to the uncertainty of clinical trial supply chains, simulations are highly recommended to consider the variability of the patient demand. Simulated results can then be an input for optimization, which will decide how much drug is needed to cover that demand without risks. GenAI can support by identifying aspects of the trial design that are driving costs or risks, for example, and recommend the user certain scenarios to test.
Other AI methods for clinical supply chain management
In the next episodes of this series, we will analyze how other AI methods can support the clinical trial supply chain. Look forward to our analysis on the following topics in the coming weeks:
AI in Clinical Supply Chain Episode 4: Machine learning
AI in Clinical Supply Chain Episode 5: Other methods
Want to receive the next article directly in your mailbox?
If you missed it, you can find the first episode of this series here:
AI in Clinical Supply Chain Episode 1: Generative AI
AI in Clinical Supply Chain Episode 2: Simulations
AI at N-SIDE
As an optimization company, N-SIDE has been successfully using AI for more than 20 years in over 12.000 clinical trials. Optimization, simulations and machine learning algorithms are native in our solutions.
Wish to learn more about how N-SIDE uses AI to reduce drug waste and risks in the clinical supply chain, while saving time and money? Book a meeting with one of our experts.
