Related Experiment Video
Updated: Mar 21, 2026
Author Spotlight: Advancements in Multiplex Detection of Respiratory Viruses
Published on: November 10, 2023
Pandemic-associated disruption of seasonal mortality patterns in Lusaka, Zambia
Avulundiah Edwin Phiri1, Muleya Siakabeya2, Veronica Mtonga2
1Mathematics, The Copperbelt University School of Mathematics and Natural Sciences, Kitwe, Zambia edwin@aims.ac.za.
Introduction:
Climate-sensitive mortality in rapidly urbanising tropical Africa is poorly characterised, and how pandemics disrupt established seasonal patterns remains underexplored. We analysed long-term all-cause mortality in Lusaka, Zambia, to quantify pre-COVID-19 seasonality and examine pandemic-associated deviations.
Methods:
We conducted an observational, exploratory longitudinal time-series analysis of routinely collected monthly all-cause mortality data from Lusaka, Zambia (January 2013 to December 2023; n=180 276). Seasonal dynamics were assessed using classical decomposition and Kruskal-Wallis tests. Structural breaks were identified using Bai-Perron and cumulative sum control chart (CUSUM) analyses. A parsimonious Seasonal Autoregressive Integrated Moving Average (SARIMA) model, selected via Akaike information criterion-based automated procedures and validated using rolling-origin cross-validation, was used to benchmark pre-pandemic forecast performance. Analyses were based on aggregated mortality counts in a population with a high paediatric mortality burden.
Results:
Pre-pandemic mortality exhibited two seasonal peaks: the rainy season (November to March), 22% above the annual mean (95% CI 15% to 29%) and the cool-dry season (June to July), 18% above (95% CI 12% to 24%). Two structural breakpoints temporally aligned with pandemic onset (March 2020) and a subsequent transition toward more stable post-pandemic patterns (May 2022). During the pandemic period, cold-season mortality increased by 41% (95% CI 32% to 50%), while rainy-season mortality declined by 28% (95% CI -35% to -21%). Overall seasonality weakened after 2020 (p=0.14 vs pre-pandemic p=0.05). The SARIMA model showed good pre-pandemic accuracy (root mean square error (RMSE)=245; MAPE=9%) but underestimated mortality during pandemic surges.
Conclusions:
COVID-19 coincided with substantial disruption and attenuation of established seasonal mortality patterns in Lusaka. These observational findings highlight the value and limitations of routine mortality surveillance and forecasting for situational awareness and preparedness in rapidly urbanising, resource-constrained settings.
