In this second interview in the series on the Zambian supply chain pilot, A Humourless Lot talks with Prashant Yadav, professor of supply chain management at the MIT-Zaragoza Logistics Program.
AHL: Could you tell us a bit more about your role in the project?
PY: I had conducted research on the medicines supply chain in Zambia in 2006 funded by the UK DFID which highlighted deficiencies in the system. After conducting the study to diagnose the supply chain problems, one of my specific mandates from DFID and the World Bank was to come up with four of five options that could possibly solve the issues that were identified in the earlier reports. A second task was to give input on measurement and the metrics to measure success vs. failure: what indicators to use and how to measure them in such a way that we could draw scientifically valid conclusions. We wanted to integrate monitoring and evaluation into the project from its earliest stages.
AHL: You say you came up with four or five options, but only two were in the end tested. What were the others and why weren’t they incorporated into the pilot?
PY: One option that was brought up by some stakeholders but did not make it was to transport the commodities directly from the central warehouse to the facilities, using a fleet of smaller vehicles. One key issue with this option was that it was difficult to quantify costs in advance, and we believed that direct distribution to clinics would become very expensive from a transport cost standpoint. Also, it would not be technically feasible in some areas that are hard to reach.
A second option involved regional medical stores that each would service a large chunk of the country and supersede the district stores. The issue with this model was that it would become too big to pilot: to be able to make any analysis, we would need to include a number of regional warehouses that would in the end encompass a very large part of the country. We agreed that this is something we can pursue at a later stage using a simulation model.
A third and final option that was not selected for the pilot was to outsource transport to the facilities. This was dropped just due to practical aspects: we found out that we could probably only find transporters on the high-frequency routes, and many facilities are not located anywhere near those routes.
AHL: How about measurement? Can you tell a bit more about your analysis of the results?
PY: Demand for some of the 25 tracer drugs that were analyzed under this pilot was not very stable. We had thought that the results for commodities with stable demand will show that the cross-dock model [where supplies were pre-packed at the central level – MK] performs better, and the results for those items were clearly statistically significant. However, to our surprise, even for the items with large variations in demand, the results for the cross-docking model were still significantly better than either the original situation or the first model. Seasonality in demand, time of conducting the data collection, quantifying the outcomes all made the analysis fairly challenging. However, working together with Jed Friedman at the World Bank’s research group and Jérémie Gallien, a colleague at MIT, we found some robust ways to quantify the impact.
Another issue is that both models presuppose that there are no stock-outs at the central stores. We don’t know how robust the models will be if stock-outs would occur at the central store.
AHL: How about the future? What is happening next?
PY: We are now involved in the progressive roll-out of the model to the whole of Zambia. Together with the government of Zambia and cooperating partners, we are discussing various options to scale up the cross-docking model. We are also thinking about how to handle clinics that are cut-off during the rainy season under the cross-docking model.
We are also exploring some options with primary health center kits. This is an alternative model, more push-based than the normal fulfilment models and fairly rigid; consequently, it sometimes leads to more excess and wastage than might be necessary, but is robust to events such as the clinic not placing an order. We are looking at possibilities for customised kits, e.g. on a regional basis, depending on the specific circumstances and needs.
Finally, we are still collecting data and updating our evaluations. We will also use the data to feed a parameterised model that should enable us to simulate other solutions. One example is the regional-store model that was discarded at the outset for practical purposes; once we have sufficient data, which I expect to be the case in a few more months, we should be able to simulate what would happen if regional stores were to be implemented.
AHL: How well could this model be ‘exported’ to other countries?
PY: I can think of two or three countries that have a similar setup as Zambia and suffer from similar issues. They would be good candidates for a similar model, but spatial distribution of facilities could make a big difference and could in the end mean that the model would be less effective there. Even where a direct verison of this may not be applicable, there are learnings which can improve the distribution systems in many other countries. We are in discussion with several countries and large donors on how to disseminate this to a wide group of public health specialists.
AHL: Finally, what was your experience with the cooperation with so many and various partners?
PY: From my perspective, things went quite well. The collaboration offered something unique to each of the partners. For instance, for MIT/Zaragoza it offered the opportunity to use our academic knowledge for a practical improvement in the lives of Zambians. Similarly, for the World Bank and USAID it offered the possibility to show that these two organisations, who have not always cooperated smoothly, could partner closely and productively.
Likewise, each of the partners brought an important aspect into the project: the World Bank delivered funding, and impact evaluation knowledge; the USAID/Deliver project brought local presence that delivered some economies of scope; MIT/Zaragoza contributed academic knowledge; and so on.
Of course things did not always work smoothly. For instance, it was hard work to convince everyone in the joint team that it was acceptable to do a quasi-randomized trial instead of agreeing with the stakeholders on what is the one ‘best’ solution and then implementing it. Similarly, there were discussions about the profile and reporting structure of the commodity planners stationed in the districts vis-à-vis the district pharmacists who were already present in some of them.
However, in the end we were able to overcome all our differences of opinion, and I think the result shows how well we were able to work as a group.