Sharing the wealth: Providing necessary medical supplies to underserved countries

Doing Good with Good OR: Award Winning Research by Can Zhang

Every year, INFORMS holds a student paper competition, Doing Good with Good OR. This year’s award was given to Can Zhang from Georgia Institute of Technology. Here are the highlights from his research paper, “Truthful Mechanims for Medical Surplus Product Allocation.”

Even though American news commonly has stories about the high costs of insurance and medical bills, it is rare that you hear about the financial efficiency of the medical institutions themselves. How consciencious are hospitals with their use (or nonuse) and disposal of medical equipment? Just think of how many pairs of gloves a single doctor or nurse uses during just one day (even if only for sanitary purposes). Unfortunately, healthcare institutions in the US tend to throw nearly 6 million tons of unused medical supplies straight into the trash every year. Some of these materials are unused left-overs from surgeries, while others are highly expensive medical machinery.

Examples of commonly wasted medical supplies. Images from “A Catalogue of Waste” by Propublica.

At the opposite end of the spectrum, developing countries are in stark need of both basic and advanced medical supplies. This shortage has dire consequences for millions of people. In 2016, the World Health Organization (WHO) recorded nearly 6 million children under the age of five died — more than half of these lives could have been saved with access to “simple, affordable interventions.” Luckily, special organizations called Medical Surplus Recovery Organizations (MSROs) have stepped in to solve both problems by collecting surplus medical supplies from wealthy countries and redistributing them to others in need of them.

Hospital beds before and after replacement by MedShare, an example of an MSRO.

However, these MSROs face their own efficiency dilemma while considering redistribution. To assist in their distribution of supplies, the distributor asks recipients to rank or score their “wishlist.” However, the MSRO has to carefully design this preferential mechanism, sort through the resulting data, and decided what information to make publicly available.

There are many complicating factors when determining which supplies to send to whom:

• The inventory for the distributer is usually supplied through donations, which are random and unreliable.
• A need for one hospital is not a need for every hospital. Often medical equipment is highly specialized, and either a hospital needs a unit or already has a functioning unit. For example, I have a functioning toaster in my kitchen, but what I really need is a crock pot. If you give me another toaster, it’s as good as garbage to me. (This happens a lot in the medical supply scenario: WHO estimates that 70% or more of donated supplies were inappropriate.)
• The needs of hospitals are difficult to determine, even by asking them directly. Just like every Christmas when I have a hard time determining a list of things that I need or want, hospitals can’t always provide a complete or perfect list of supplies that they need.
• If the mechanism for recipients to rank or score the items on their wishlist is not carefully designed, the recipient may be tempted to misreport (i.e., lie) in order to increase their chances of receiving a particular item. For example, this is the case when the inventory information is shared.
• Due to severely limited budgets, the distributors ship supplies in bulk or bundles to each recipient after enough of the desired supplies is available. This complicates the timing of when to serve each recipient.

All of these factors combine to create a resource allocation problem that is unique to MSROs.

First, the authors developed a truthful mechanism for ascertaining the preferences of recipients. “Truthful” here means that recipients cannot somehow game the system by misreporting or lying about their preferences. From the recipient’s perspective, in order to maximize the chance of getting their desired supplies, they must report them honestly. Only truthful mechanisms allow for the MSRO to distribute supplies fairly.

Furthermore, the authors were able to prove that in order to have a truthful response by recipients, the distributor must not reveal the available inventory. To be precise, if the inventory information is shared, then the only way to choose supplies for a recipient fairly is to do so randomly, which defeats the whole purpose of collecting preferences. Making their inventory publicly visible is an easy mistake for an MSRO to make in the attempt for transparency, however it eliminates an MSRO’s fairness. The authors also proved that having recipients score their preferred items rather than just rank them (scoring requires more information) does not provide any better allocation. These results suggest immediate (and easy) ways for all MSROs to significantly improve their operations.

Consider again the overall problem of maximizing the amount of useful supplies delivered to the recipients. The authors propose a specific mechanism for choosing which recipient to serve and compare it to randomly selecting one. Their mechanism certainly improves the performance compared to random choosing. In addition, using past data it’s possible to find the optimal allocation in hindsight, what the authors call an optimal clairvoyant solution. Even though it would be unreasonable to expect to solve for such a solution for the (unpredictable) future, this does provide an absolute upper bound for comparing the new metric. Impressively, the author’s mechanism performs nearly as well as the clairvoyant solution, closing 75% of the gap between random choice and the best possible solution!

Again, the practical results of these results are profound. There already exist several medical surplus recovery organizations that are trying to do great good — to improve the quality of lives in developing countries by solving a dire scarcity of medical supplies. With some clever analysis of fairness, truthfulness, and optimization, the authors are able to offer strategies for improving their operations to vastly improve their services. In doing so, the research team has empowered the MSRO and aided their efforts to save the lives of thousands of people.