Connect with us


Solving the medical device supply chain crisis with probabilistic AI

By Neha Puri, supply chain and logistics lead at Faculty.



Picture taken from a drone of shipping containers in a city depot warehouse.

The pandemic sparked a crisis in the medical device supply chain, one that is still eroding the confidence of stakeholders and customers today. Something we all hoped would last only a matter of months, a new, perpetual state of uncertainty has exposed the fragility that lay beneath.

This vulnerability has led to poor quality forecasts which have diminished trust, leading to frustrated suppliers and operational teams. Most importantly, it has jeopardised those in need of medical supplies.

The ‘’just in time’’ model, which is common across supply chains, has created risks for the healthcare industry. This model is largely built for times of surety, which means that poor forecasting is unexposed; fewer shocks mean forecasts have been able to rely on stability, without the precision needed to counter volatility.

Fewer fluctuations and less real time analysis mean that inaccuracies in forecasting are compounded over time, without medical device companies even realising.

To state the obvious, inaccurate forecasts make it much harder to plan and budget, leading to inefficiencies and excess costs. No business wants this.

To put the most positive spin on the last two years of turbulence, at least the fact that many executive teams have lost confidence in their predictive forecasting means that they have recognised that an overhaul is needed.

So despite intense short to medium-term pain from the pandemic and the broader supply chain issues, the medical device companies who have been exposed are in a far better position to set themselves up to be more competitive in the long term.

Probabilistic forecasting as power

Luckily, there is a solution. Obviously the key to better supply chain management is better forecasting. Undoubtedly even improvements in manual processes and reporting mechanisms would be beneficial. Some medical device companies are even already using embedded AI forecasting to account for volatility. But companies can do even better than this with newer, more advanced tools.

But how do these advanced tools differ from what operations teams are already using? This is where an approach known as probabilistic forecasting is making a difference. Probabilistic forecasting makes this possible by projecting a range of possible outcomes which consider supply chain volatility – instead of simply relying on historical data.

A worker going into a clean room to repair technology where medical equipment is rapidly being made to stem the spread of the coronavirus.

Rather than providing a singular forecast like most traditional machine learning methods, probabilistic forecasting provides a range of likely values. The range of forecasts can be used to consider different possible scenarios and determine plans to brace for the worst.

The lack of confidence in single-figure forecasts has damaged the ability of operational teams to become resilient in crises, a fact particularly highlighted by the pandemic. Probabilistic forecasting builds resilience by mapping out multiple outcomes and therefore highlighting possible risks.

In turn, decision-making confidence grows, as predicting medical device needs within an estimated ‘boundary’ increases the chances of the best decision being made.

Value-chain resilience is defined as the ability to recover quickly from supply chain shocks and is integral to confidence in the supply chain. The commonly used just in time model can weaken this by resulting in stock-out. This is critical in the medical supply chain as stock-out can mean running out of beds or ventilators, and negatively impact patient care.

Knowing the full range of possible outcomes, rather than just the most likely outcome, allows medical leaders to make principled, data-driven decisions and improve value-chain resilience. The range of forecasts mitigates the risk of stock-outs as demand calculations become more reliable.

Simplifying risk appetite

Using multiple forecasts from probabilistic forecasting, medical device companies can decide their own risk appetite.

This method has proven successful in the wider medical supply chain, from demand forecasting to machine downtime and staff absence.

Even the UK’s National Health Service faced a crisis as poor forecasts led to insufficient supplies of medical devices such as ventilators and personal protective equipment.

The NHS implemented a probabilistic forecasting model via the Early Warning System (EWS), which provided critical and unparalleled insight into hospital demand.

The EWS allowed NHS leaders to foresee waves of COVID hospitalisations, meaning patients could be moved from beds and care prioritised for those who needed it.

This saved time, reduced inefficiencies, and saved thousands of lives amidst a state of crisis. The same principles can be applied to forecasting demand for medical devices.

One major advantage of better forecasting is that accurate risk boundaries can be discussed across the supply chain ecosystem. Rather than optimising locally in their silos, teams can make decisions knowing the effect the decision will have on the wider organisation.

The teams can discuss the risks highlighted in their forecasts and can work with the wider business to decide the overall risk appetite in the business. This can improve trust across the supply chain and help inform holistic decision-making.

Building confidence amid uncertainty

The ability of probabilistic forecasting to recognise uncertainty can help restore stakeholders’ confidence in medical supply chains. Uncertainty can be quantified early through forecasting, so predictions can be made confidently about demand, even in volatile circumstances.

However, the success of probabilistic forecasting is dependent upon the human element: the AI has to be embedded fully and teams need to understand the insights they are given. Organisations need the tools to take advantage of the probabilistic forecasts.

This requires changing the data visualisations that operational teams look at daily to explicitly highlight uncertainty, as well as educating teams on how to use the estimates.

This is critical, as the integration of AI into teams strengthens planning and reduces the risk of stock out or excess inventory.

Better forecasts mean executive strategy can be framed around reliable data, as well as helping organisations to better react to supply chain shocks.

Medical supply chains remain especially volatile due to the pandemic as well as wider supply chain problems, and while that may not change immediately, probabilistic forecasting provides accurate new insight for medical device supply chain planning. These new tools can be embedded easily in existing systems and are quick to value.

They could be just the overhaul that executive teams are looking for.

Neha Puri, supply chain and logistics lead at Faculty.

Continue Reading
Click to comment

Leave a Reply

Your email address will not be published. Required fields are marked *

Trending stories