Learning from a Crisis

Roughly 10 years ago, Junghee Lee was a Ph.D. student when he happened upon a newspaper article about recurring shortages of essential drugs in the U.S. The problem had persisted since the early 2000s, compromising patients’ care quality and forcing physicians to make wrenching rationing decisions. Even with his background studying supply chains and healthcare operations, Lee found himself stumped by the problem – and therefore deeply curious. 

The scale and the impact of the problem also shook Lee, who now teaches business analytics at Mendoza College of Business as an assistant professor. He decided to dig deeper, writing his first paper, “Alleviating Drug Shortages: The Role of Mandated Reporting Induced Operational Transparency,” published in Management Science, on the impact of President Obama’s 2011 executive order requiring greater transparency from drug manufacturers about looming shortages.

But Lee’s concern about the issue didn’t diminish. He uncovered a troubling pattern: “The same companies experienced drug shortages again and again and again,” Lee said. Individual drug manufacturing plants experienced an average of 33 shortages per plant between 2015 and 2020.

For Lee, this raised a pressing question, though it was one that had never before been studied in the research literature: Are these individual drug manufacturing plants actually learning from their own past shortage recoveries, or are they simply repeating the same mistakes?

That question is at the heart of Lee’s new research paper, “Learning in Recovery from Disruption: Empirical Evidence from the U.S. Drug Shortages,” forthcoming in Management Science and co-authored with Hyun Seok Lee and In Joon Noh of Korea University, and Bradley Staats of the University of North Carolina.

The study’s findings reveal that drug factories did indeed get faster at fixing shortages over time, but only under certain circumstances. Learning was mostly focused on problems the plants themselves caused, not on external shocks. What’s more, learning remained completely contained within each plant and failed to translate into process improvements across the firm’s other factories.

Patterns lurking in shortages

Why do drug shortages happen in the first place? The factors at play are usually not unpreventable external forces such as labor strikes, wars or natural disasters. More often, the reasons behind a shortage should have been preventable.

“The largest chunk of drug shortages happened simply because the firm had some kind of operational issue,” Lee said.

Such problems account for roughly two-thirds of all shortages. Common examples include bacterial contamination, sterility failures and deviations from standards set by the Food and Drug Administration.

One major reason these operational failures persist, Lee explained, is the lack of economic incentive to prevent them. That’s why shortages are far more common for generic drugs than branded ones: Generics are a lower-margin enterprise.

“The profit from these shortage-experiencing drugs is very small,” Lee said, “so firms and researchers tend to ignore them.”

To better understand what facilitates manufacturers’ learning about how to overcome shortages, Lee and his colleagues created a dataset that drew from multiple sources, including the University of Utah Drug Information Service, which collects drug shortage start and end dates, as well as the reason behind the shortage.

The researchers also conducted interviews with executives from five global pharmaceutical firms to add explanatory context to the patterns uncovered by the findings. Their final sample searched for patterns among nearly 5,000 shortages involving more than 1,500 drugs produced at over 100 facilities between 2015 and 2020.

Concentrated learning

The study produced four major findings. Some of them Lee found reassuring; some of them, less so.

Drug shortage recovery is an irregular, poorly defined process and far more chaotic than the more structured tasks typically studied in Operations Management. Even so, the researchers’ data revealed clear evidence of learning. Plants that had been through more recoveries got meaningfully faster at resolving future ones.

However, that learning was driven almost entirely by shortages rooted in internal failures. When the breakdown was caused by the plant itself, workers and managers were accountable, motivated to understand what went wrong and positioned to do something about it. When the cause was external, on the other hand, the data showed no evidence of learning at all. Behavioral factors played a role here, too. One executive interviewed for the study described how plants simply felt less responsible for problems that originated outside their walls, and therefore invested less effort in understanding them.

Notably, the breadth of a plant’s prior experience with shortages mattered. Plants that had dealt with a wider variety of key ingredients in their manufacturing processes and shortage types learned more from each subsequent recovery than plants with a narrower range of experience.

But the finding that surprised Lee the most concerned the inertness of learning and improvements post-shortage. Even as a given firm’s plant updated its processes after it suffered a shortage, the firm’s other plants showed no improvements.

“That seems like it would be the low-hanging fruit,” Lee said. He speculates that either plants do share knowledge, but the sharing is a largely meaningless practice — “maybe somebody writes it, but nobody reads it,” he said — or they simply don’t share at all, too consumed by the pressures of daily operations.

Public-private partnership

This is precisely where Lee suggested pharmaceutical companies should begin to tackle the still-pervasive drug-shortage issue: by investing in cross-facility communication.

In the study, he and his colleagues recommended regular cross-site meetings, dedicated task forces to document best practices and centralized digital repositories accessible to engineers and managers across facilities. He emphasized that technology can accelerate this effort. In interviews, one senior executive described how their firm built an AI-based system to enable operators across multiple sites to efficiently access information for resolving operational issues.

Lee envisions a future where years of accumulated plant documentation — hundreds of folders worth of institutional knowledge — get fed into a system like ChatGPT, enabling frontline workers to type in a problem and get actionable, contextually informed guidance.

While it’s critical that firms take action where they can, Lee also acknowledges that the government has a role to play, too.

“We clearly show that for external issues, the manufacturers and the industry by themselves haven’t figured this out,” Lee said. “When it comes to these external shocks, policymakers should intervene.”

For instance, policymakers could help manufacturers diversify their supplier base or loosen regulations around who is permitted to manufacture active pharmaceutical ingredients to help hedge against the possibility of specific-ingredient shortages. The study also recommends targeted measures such as subsidies for dual sourcing of important ingredients, mandated contingency planning for critical drugs and financial incentives for supplier diversification.

Lee believes that targeted government subsidies could help to ameliorate this seemingly intractable problem. After all, underlying the various learning and process issues bound up in drug shortages is a financial one, Lee said.

“I believe this is fundamentally a money problem,” he said. “If the government were ready to spend some money on the lower-profit drugs, things will get better very quickly, I think.”

For many generic drugs experiencing repeated shortages, margins are simply too thin for manufacturers to invest in the kind of resilience-building infrastructure that would prevent future crises. Subsidizing those drugs could free up resources for companies to do what they demonstrably can do when they have the accountability, the experience and the incentive to do it.