This article was also published on Medium.
As stated by our CEO and co-founder, Chakri Gottemukkala, o9 Solutions is taking on the challenge to enable integrated planning across the healthcare supply chains. In order to tackle this situation, we believe we need to be able to model the following key planning topics, as they all play a role, from the number of cases down to the drugs and equipment required.
The following points represent our current understanding of the planning challenges we see. We are sharing those freely as we look for any contributions that could improve and help with deploying expertise and systems to address the current situation.
Forecasting infected cases
Today, many models are currently being worked out publicly, from the CDC and other health institutions to forecast the number of future cases at the national and state level. The classic SIR (Susceptible, Infected, Recovered) model is currently the starting point with many open source forecasting, machine learning and AI models being used to augment the SIR models. The challenge, we see, is to be able to express a more granular model of the number of cases that are potentially infected within a county or a city. More refined models using various causal data like population demographics by spatial dimension, could help improve visibility into the number of cases to appear on a daily basis and within a certain geographic perimeter. This is key to be able to plan the next level.
Example of epidemic calculator available here
An Augmented forecasting model, developed by IHME
Forecasting patients and needs
Having the number of cases infected to appear in a state or at Federal level on a given day, is one thing. The next step is to derive from it the number of patients that will turn up at a hospital or at a level of a city, in severe and critical state and for how long those patients will typically (on average) stay in those conditions. It is not enough to use a percentage of severe cases (usually 20% from infected) and critical (usually 5% from infected). Different lead times need to be applied in the model to take into account the time it takes for a possible patient to go from one stage to the next. For this, we are proposing the following model:
This model is currently deterministic as the number of days are averages and the different percentages are applied to the whole population. In reality, we understand that the number of days will be vary in a stochastic way and the percentages will vary by various drivers like age and comorbidities. Nevertheless, at an aggregated level, we believe it gives a better indication in terms of time and the length of stay in ICU or in normal hospitalization conditions.
Creating the supply model
Having created a “Demand” for Patient care per day, per Severe or Critical status and by state, county or hospital, it then becomes clear that we need to create a “Supply Model” to provide care for the patients. We need to develop a list of critical resources e.g. beds, doctors, respirators and of critical drugs supplies like PPE, consumables for respirators, . Such a supply model could be described as below:
Managing scarce & shared resources
Having modeled the beds, the ventilators, the doctors and nurses as resources in our Supply model, it becomes then possible to understand the amount of each will be required, day by day, to meet the increasing demand of patients in severe and critical conditions. We understand that even if more beds and more ventilators can be purchased, the qualified ER doctors and nurses are also absolutely required to operate those equipment and provide for the patients. This can become a clear bottleneck in the capacity of a hospital or city to provide for the increasing number of patients. The same can be applied to critical drugs that can become scarce and requires rationing and allocation. Having a clear view on the future need of resources and the coordination required between these will be key to treating all patients.
Creating the network
Now that the demand of drugs, consumables or equipment is known, it needs to be projected across a distribution network from the suppliers to the different distribution centers to the hospitals. Inventory on hand, inventory targets and existing orders need to be taken into account to plan the future purchasing requirements.
The different demands should be pooled together for a given supplier and a collaboration framework should allow for transparency in the status of the orders, and where the stock is expected to be delivered. In case the requested volumes cannot be delivered in full, it will become important to allocate the scarce supplies according to clear and transparent rules.
At o9 Solutions, we believe that our expertise in tackling supply chain problems and our platform can support those requirements and could bring much needed visibility, collaboration and planning capabilities in a time when such decisions need to be made.
Nevertheless, it might be that somebody has developed a more sophisticated model and is looking to deploy it at scale. Or you have access to data sources that could improve the forecasting or planning of the situation. If so, please don’t hesitate to reach out to us, so we can collaborate and bring our share of the collective effort in the fight against COVID-19.
Get free industry updates
Each quarter, we'll send you a newsletter with the latest industry news and o9 knowledge. Don’t miss out!
About the author
Tanguy CailletExecutive Vice President - Growth Markets & Global Partnerships
Tanguy Caillet is Senior Vice President Global Industry Solutions at o9 Solutions. In his role at o9, Tanguy is a global lead across the retail, CPG, process, and discrete manufacturing verticals. Throughout his 20+ year career, Tanguy has advised on supply chain transformation and planning processes, and IT solutions for global industry leaders.