Maisha Meds believes it is possible to use systematically collected data from across the medicine supply chain to detect shortages in countries before they start to affect patients. In 2020, our organisation built an initial prototype of just such a system that, while imperfect, could inform future attempts to design early warning systems for the medicine supply chain. The attached paper describes the methodology, results, caveats and implications of this attempt.

Introduction and background

The COVID-19 pandemic has triggered a number of health commodity supply chain challenges that have led to healthcare commodity shortages and price hikes. Maisha Meds has internally published a paper that aims to describe how we have used a combination of international customs trade data and retailer point of sale (POS) data to develop and test a prototype of a medicine shortage flagging tool, to identify potential essential medicine shortages across sub-Saharan Africa.

Maisha Meds used export data from India, import data from Kenya and Point of Sale data from Kenyan pharmacies to identify anomalies in the supply of medicines, view price changes and look for leading indicators of potential national level medicine shortages across sub-Saharan Africa.


Analysis 1: Using an ARIMA forecast model Maisha Meds identified 116 out of 525 finished pharmaceutical products that experienced a significant drop in exports from India to all of Sub-Saharan Africa ex. South Africa after Feb 2020. This was an increase from 34 / 525 products experiencing a significant drop in exports indicative of a shortage from October 2019 – Feb 2020, before Covid-19’s impact on supply chains.

Wholesaler interviews identified 39 molecules as being in shortage in at least one country, of these Maisha Meds had sufficient data to analyse 37. The Maisha Meds ARIMA forecast model was able to flag 18/37 molecules as at risk of shortage to a 95% confidence interval. In the instances where there was sufficient API data the ARIMA model successfully flagged 11/13 products as at risk of shortage that were confirmed by local wholesalers (or 15/18 including the products highlighted as in shortage not included on the original list of tracer molecules).

Analysis 2: Maisha Meds also conducted an analysis to link historic drops in Indian exports to major price increases in Kenyan pharmacy prices, in an effort to prove a link between change in exports and change in prices for patients in Africa. In 8/15 molecules, Maisha Meds was able to identify a statistically significant change in the value of exports from India to Kenya, using an 80% confidence interval, prior to a trend break price increase in Kenya. Looking at the products that exhibited a relationship between exports and price increase in Kenya, a negative correlation between price and export value was observed with a lag of 4-5 months in half of the cases, in particular for products with a higher import dependence on India. This would indicate that with refinement and improved data inputs it might be possible to identify a shortage before it affects medicine supply in a country.


As an early stage prototype to show what might be possible in building a medicine shortage flagging tool, this analysis had a large number of constraints including resource availability, issues in the accuracy and breadth of the data cleaning algorithm, lack of differentiation of medicines by form, lack of consistent methodology, incomplete data inputs, lack of granularity on smaller products and difficulties analysing individual country trends using trade data.

With these caveats in mind, Maisha Meds believes that the results were encouraging. From 37 shortages identified by wholesalers the model was able to flag 18 as being at risk of shortage based on a 95% confidence interval. Furthermore, the analysis was able to show examples of a link between interruptions in supply from India and an increase in pharmacy prices in Kenya on average 4-5 months later, for a small number of products. This would indicate that with increased resources, and additional data inputs of sufficient quality, it should be possible to design an early warning system that would actually give countries enough time to act on a potential shortage. However this would require improvements in data sharing and capture at a global level that is unlikely to come about without significant political will to intervene.


Methodology, results, discussion and caveats of this study are all covered in depth in the attached document “Post Covid-19 health commodity shortage tracking_V5.pdf”.