Second Order Consequences

9/10 times the patient is responsible for securing the blood for their transfusion, which leads to three possible scenarios:

📈

LSTM for Multivariate Time Series Forecasting

The first major component of the algorithm leverages a Long Short-Term Memory Neural Network to forecast each blood bank's collection and utilization for the next week. It then compares the forecasted inventory to the network baseline to determine if the bank will be in a shortage or surplus.

💻

Implementation of the Hungarian Algorithm

The second component of the algorithm is my implementation of the Hungarian Algorithm which takes in the map of shortage and surplus blood banks. The Hungarian Algorithm determines how to fulfill all the shortages while minimizing the total distance that the blood must travel.

⏳

Transportation of Expiring Units

Blood is perishable and therefore its shelf life must be taken into account. Circulate's algorithm includes a component to determine whether the expiring blood in a surplus blood bank has a higher probability of being used in the shortage blood bank (therefore it is transported) or not.

Expanding on Phase 2 | Flying to India for On-the-Ground Validation

Pivoting Circulate's Focus and Kickstarting a Partnership with the Governing Body for Blood

Toward the end of June 2022, the lack of perspective sitting in Canada was now becoming crucial as it was time to talk to blood banks and get validation on whether or not they would use our algorithm. On top of that, my main priority was to get the required data to train the machine learning component of the algorithm and kickstart relationships for a pilot in the future. In the summer of 2022, I travelled to India for 3 weeks to understand the problem from an on-the-ground perspective, which provided me with crucial insights that altered the way I framed the problem as well as the final solution.

I spent weeks talking to blood banks and hospitals which resulted in a variety of insights, one of the key ones being that most blood banks in India actually had delivery vehicles and round-the-clock staff to deliver blood. This meant that the issue wasn’t in the physical movement of blood but rather the lack of an automated distribution system, so that’s what I shifted my focus to. 

After cold calling dozens of hospitals and standalone blood banks, one key insight that I was informed of was that all blood banks were governmentally regulated to update their daily inventory on the State Blood Transfusion Council portal. It was clear that this was exactly the portal I needed to get access to for the input data to train the final part of the Long Short-Term Memory Network. The man who answered the phone at the Sanjeevani Blood Bank in Pune explained that I would need permission from the STBC directly. A few phone calls later, I arranged a meeting with the Assistant Director of the SBTC at their head office in Mumbai, India.

During the meeting, I first walked through my research and the three main root causes of lack of adequate blood supply, no coordination between stakeholders and lack of safe blood transportation infrastructure for validation. Throughout the meeting, it was extremely helpful to have a single source of truth where I could validate my assumptions and data received from other sources since they were both experts in the field. After that, I walked them through the MVP of the algorithm and finally came to my ask. I explained how by using machine learning, specifically Multivariate Time Series Forecasting to predict the blood collection and utilization rates, the algorithm would be able to better estimate the blood banks' delta and therefore more accurately distribute the blood. This is because instead of simply looking at one instance in time, predicting the inventory for the upcoming week gives the algorithm several more data points. As we were discussing how large the blood network should be, the COO recommended that I focus on the city of Mumbai, prove that the algorithm works and then we can discuss the next steps. Therefore, they agreed to provide all the blood collection and utilization data for all the 54 blood banks in Mumbai for the previous 5 years!

Dr. Arun Thorat

Assistant Director of the Maharashtra State Blood Transfusion Council

Navneet Bali

Director of Narayana Health for Northern India

Dr. Hasan Abbas Zaheer

WHO Advisor on Blood Safety

Dr. Smita Hiras Sudke

Chief of Pathology Lab

Dr. Surpreet Mohanty

Blood Bank Director at the Indian Navy

Damodhar Sai

Projects Lead at Indian Red Cross Society

About Me

Aahaan Maini