Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Demand Forecasting for Platelet Usage: from Univariate Time Series to Multivariate Models

Maryam Motamedi, Jessica Dawson, Na Li, Douglas G. Down, Nancy M. Heddle

TL;DR

This study compares univariate and multivariate platelet demand forecasting models using CBS Hamilton data from 2010 to 2018. It shows that when data are limited, incorporating clinical predictors via methods like Lasso, Random Forest, or LSTM improves accuracy, while with large data volumes, ARIMA remains competitive. The rolling-origin evaluation reveals how training window length and retraining frequency influence forecast performance, highlighting a trade-off between model complexity, interpretability, and data availability. The findings inform both operational forecasting and inventory planning in the context of platelets’ short shelf life and variable demand, with practical implications for reducing wastage and urgent orders.

Abstract

Platelet products are both expensive and have very short shelf lives. As usage rates for platelets are highly variable, the effective management of platelet demand and supply is very important yet challenging. The primary goal of this paper is to present an efficient forecasting model for platelet demand at Canadian Blood Services (CBS). To accomplish this goal, four different demand forecasting methods, ARIMA (Auto Regressive Moving Average), Prophet, lasso regression (least absolute shrinkage and selection operator) and LSTM (Long Short-Term Memory) networks are utilized and evaluated. We use a large clinical dataset for a centralized blood distribution centre for four hospitals in Hamilton, Ontario, spanning from 2010 to 2018 and consisting of daily platelet transfusions along with information such as the product specifications, the recipients' characteristics, and the recipients' laboratory test results. This study is the first to utilize different methods from statistical time series models to data-driven regression and a machine learning technique for platelet transfusion using clinical predictors and with different amounts of data. We find that the multivariate approaches have the highest accuracy in general, however, if sufficient data are available, a simpler time series approach such as ARIMA appears to be sufficient. We also comment on the approach to choose clinical indicators (inputs) for the multivariate models.

Demand Forecasting for Platelet Usage: from Univariate Time Series to Multivariate Models

TL;DR

This study compares univariate and multivariate platelet demand forecasting models using CBS Hamilton data from 2010 to 2018. It shows that when data are limited, incorporating clinical predictors via methods like Lasso, Random Forest, or LSTM improves accuracy, while with large data volumes, ARIMA remains competitive. The rolling-origin evaluation reveals how training window length and retraining frequency influence forecast performance, highlighting a trade-off between model complexity, interpretability, and data availability. The findings inform both operational forecasting and inventory planning in the context of platelets’ short shelf life and variable demand, with practical implications for reducing wastage and urgent orders.

Abstract

Platelet products are both expensive and have very short shelf lives. As usage rates for platelets are highly variable, the effective management of platelet demand and supply is very important yet challenging. The primary goal of this paper is to present an efficient forecasting model for platelet demand at Canadian Blood Services (CBS). To accomplish this goal, four different demand forecasting methods, ARIMA (Auto Regressive Moving Average), Prophet, lasso regression (least absolute shrinkage and selection operator) and LSTM (Long Short-Term Memory) networks are utilized and evaluated. We use a large clinical dataset for a centralized blood distribution centre for four hospitals in Hamilton, Ontario, spanning from 2010 to 2018 and consisting of daily platelet transfusions along with information such as the product specifications, the recipients' characteristics, and the recipients' laboratory test results. This study is the first to utilize different methods from statistical time series models to data-driven regression and a machine learning technique for platelet transfusion using clinical predictors and with different amounts of data. We find that the multivariate approaches have the highest accuracy in general, however, if sufficient data are available, a simpler time series approach such as ARIMA appears to be sufficient. We also comment on the approach to choose clinical indicators (inputs) for the multivariate models.

Paper Structure

This paper contains 25 sections, 10 equations, 15 figures, 4 tables.

Table of Contents

  1. Introduction
  2. Literature Review
  3. Methods
  4. Problem Setting
  5. Data Description
  6. Exploratory Analysis for Trends, Seasonality and Holiday Patterns
  7. Demand Forecasting Models
  8. Univariate Models
  9. Multivariate Models
  10. Rolling Window Analysis
  11. Results
  12. Trends, Seasonality and Holiday Patterns
  13. Demand Forecasting Comparisons for Univariate Models
  14. Demand Forecasting Comparisons for Multivariate Models
  15. Selecting the predictors using Lasso Regression
  16. ...and 10 more sections

Figures (15)

  • Figure 1: CBS blood supply chain with one regional blood centre and multiple hospitals
  • Figure 2: Flowchart of the proposed system
  • Figure 3: Time series decomposition using STL method
  • Figure 4: Prophet model for exploring trends, holiday effects, weekly and yearly seasonality - Since these components are combined through a generalized additive model, the values of y-axes in the plots represent the quantity to be added to or substracted on each specific day
  • Figure 5: Mean daily units transfused
  • ...and 10 more figures