The application of time-series forecasting to quantify the deficit in colorectal 2-week wait referrals caused by the COVID19 Pandemic

Background: The COVID-19 pandemic has resulted in a significant reduction in the number of referrals to 2-week wait cancer pathways due to multiple factors including patient concerns and reduced primary care services. The aim of this study was to estimate the deficit in referrals on the colorectal cancer pathway to a large NHS hospital during the pandemic using time series forecasting. Methods: Weekly referral volume January 2015 to September 2020 was calculated from an electronic tracking database. Referrals during March 2020 were excluded. Data to February 2020 were used to train a forecasting model using Prophet, an open-source software library. Hyperparameter tuning with grid search and cross-validation were used to improve model accuracy. Seasonality and holidays were accounted for. The trained model was used to forecast referral volumes from April to September 2020. Results: 16704 referrals were received during study period. Referral numbers increased every year during the pre-pandemic period from 2106 in 2015 to 4049 in 2019, with 21% increase from 2017 to 2018 and 28% increase from 2018 to 2019. Forecasted referrals for April to September 2020 were 2427 (2113-2736). 1583 actual referrals were received, 844 (530-1153) fewer than forecasted but only 471 fewer than during the same period in 2019 (2054). Conclusion: Comparing pandemic referral volumes to historic data under-estimates the deficit caused by the pandemic. Forecast models are important in estimating burden of missed colorectal disease and in resource planning to deal with rebound increase in referrals of patients with potentially more advanced disease.