The Defense Logistics Agency (DLA) recently awarded a $100,000 contract to Charles River Analytics for the creation of a means to assemble information from global trade networks and pharmaceutical drug data and present it in a way that would allow quick identification of problems in the supply chain.
PRESCRIPTION, or Probabilistic Reasoning on Supply Chain Readiness of International Pharmaceuticals using Trade Information, is a program that would assist analysts in understanding the risks and vulnerabilities inherent in the global supply chain, be it political turmoil, natural disasters or shortages of raw materials. It’s complex data, on the whole, but PRESCRIPTION aims to visualize it through a simple dashboard.
“Any time there’s an event, it can affect drug supplies; hurricanes, fires, volcanic eruptions, and politics have all created shortages in the past,” Sean Guarino, principal scientist and director for Human-Centered Intelligent Systems at Charles River Analytics, told Homeland Preparedness News. “During the past year, COVID-19 has certainly been that event and we’re learning firsthand how these disasters can impact drug and other supply chains. For example, the pandemic caused a shortage of propofol (the drug used for sedation when you’re placed on a ventilator) and other drugs, as well as hand sanitizer, masks and other PPE.”
The whole pharmaceutical supply chain could therefore benefit from PRESCRIPTION’s chronicling, although Charles River Analytics will first focus on drugs needed by military service personnel. Even vaccines, as they often utilize ingredients from particular locations, could be compromised by stalled, necessary equipment and supply production from critical centers like China or India.
“We believe PRESCRIPTION would work well for vaccine manufacturing because common vaccines, such as annual flu vaccines, are quite well documented, which allows us to experiment and model with that data,” Guarino said. “Even if data is not publicly available, we can customize the PRESCRIPTION solution to let companies enter in their own data.”
Charles River Analytics is one of four awarded a contract in this area from DLA. It will utilize its own probabilistic programming language, Figaro, and a library of inference and machine learning algorithms to capture and determine missing constructs within dynamic probabilistic networks. Its team will be assisted by Dr. Alexis Bateman, principal at SustainChain and director at MIT Sustainable Supply Chains.
Active and non-active ingredients in drugs will be put under the microscope, allowing analysts to determine what type of events could occur upstream, as Guarino put it. This is where the probability part comes in. PRESCRIPTION would allow visualization of what might happen if India ceased shipping products beyond its borders, for example. This also puts PRESCRIPTION a step above other such efforts, since many tend to focus on the supply chain of the end product, not the synthesis of drugs.
“Immediately after COVID-19 hit, this problem became very evident,” Guarino said. “Workers were unable to go to work to produce the ingredients, which ultimately had an impact on the production of drugs like propofol. For example, let’s say you have a drug that’s 99 percent made in the United States. You might think you wouldn’t care much about an earthquake in India. But the underlying ingredients that get shipped in from India might be necessary to produce the drug. So even though you’re making the drug in the U.S., that earthquake matters very much to you, because you no longer have enough of a key ingredient to continue manufacturing.”
Missing information is another potential problem for analysts. However, PRESCRIPTION proposes to rise above potential disruptions in any one data stream because its probabilistic programming gives it a bit of reasoning capability. Guarino described it as a sort of thinking – it can provide insights into vulnerabilities even in the face of incomplete data, although it may be less confident in its conclusions as a result, or provide alternate options.
If companies enter their own data, as proposed by Guarino, that uncertainty goes down and results become more precise.
“For example, Pfizer could enter all its ingredients and sources; they’d have every detail about their contracts and backup contracts that PRESCRIPTION would need to reason even more effectively and deliver extremely accurate results,” Guarino said.
With all this in mind, Charles River is considering how PRESCRIPTION could be utilized by hospitals and other pharmaceutical suppliers to track and forecast the effects of problems on their personal supply. This could allow them to better guarantee the stock in place for their patients when in need.
“Our work on PRESCRIPTION has strong commercial potential as a product that can help to analyze and predict risks and vulnerable points in the supply chain for a drug,” Guarino said. “We think it’ll be beneficial to the pharmaceutical community as a whole: Large-scale drug producers do some analyses on their own, but they could benefit from our inference. Medium-scale producers don’t have anything like PRESCRIPTION and would really be able to benefit by designing their synthesis process from the ground up to make it less and less vulnerable to events.”