https://orcid.org/0000-0003-0625-9454
Abstract
Storm Daniel, characterized as the most severe 21st century flood event in Greece, struck the Regions of Thessaly and Central Greece in early September 2023, resulting in 17 reported direct fatalities.
This study evaluated all-cause excess mortality in the prefectures of Fthiotida, Karditsa/Trikala and Magnesia affected by Storm Daniel. We compared weekly deaths during the first week and the 90 days following the flood, to expected deaths based on historical trends 2015–2019.
During the week of the flooding event, Fthiotida, Karditsa/Trikala and Magnesia prefectures experienced mortality increases of 57% (95%CI: 19%–95%), 40% (95%CI: 13%–66%) and 39% (95%CI: 7%–70%) over expected levels, respectively. A total of 54 (95% CI: 31–77) excess deaths were recorded during the week of Storm Daniel’s landfall, with the cumulative number of excess deaths reaching to 335 (95% CI: 250–420) over subsequent 90 days.
Conventional approaches that focus solely on immediate and direct flood fatalities underestimate the overall and actual health risks faced by flood-affected communities. Floods have complex and long-term health effects, necessitating continuous monitoring of the affected communities.
Introduction
Anthropogenic climate change is gradually increasing the frequency and intensity of extreme weather events (heatwaves, droughts, wildfires, inland, and coastal floods) and disasters that directly affect the health and health determinants of populations.1 Among these anthropogenic extreme weather disasters, floods are the most common ones in Europe and worldwide,2 resulting in remarkable economic and environmental damage,2 as well as short- and long-term adverse health effects for the affected populations.1
Storm Daniel, a tropical-like cyclone that affected several Mediterranean countries including Greece, started on 5 September 2023, and lasted for four days, causing heavy rainfall and subsequent flash flooding.3 The areas most severely affected in Greece were the Regions of Thessaly and Central Greece; it is estimated that during the Storm, the Peneus river basin, in which the greatest part of Thessaly falls into, received rainfall equivalent to 47% of its mean annual precipitation, leading to extreme flooding, and extensive infrastructure damage.3 A few days later on 27 September 2023, a second storm (Storm Elias) affected the same areas in Thessaly.
According to government and media sources, Storm Daniel caused seventeen deaths in Greece, all directly attributable to its physical forces, while no direct death was recorded during Storm Elias.4 In late September 2023, the European Center for Disease Prevention and Control warned that these fatalities did not represent the sole health risk for affected communities and that several factors could adversely affect the population health in the mid- and long-term, potentially resulting in an increase in all-cause mortality in flood-affected areas.4 The aim of this study was to calculate the all-cause excess mortality in the flood-affected regions in Greece, during the week of the landfall and the subsequent 90 days.
Methods
Excess deaths, defined as the difference between the actual number of deaths from all causes and the expected death counts had Storm Daniel not occurred,5 were calculated by comparing the observed weekly number of deaths following Storm Daniel with the expected death counts derived from historical mortality data. Τhe expected mortality rates were determined using a control period from 1 January 2015 to 31 December 2019, excluding the months from December 2016 to April 2017 due to increased mortality, as well as the period 2020–2022 due to the surge in deaths attributed to the COVID-19 pandemic (Fig. S1 and Table S1). Population estimates for 2015–2020 and 2021–2023 were derived from the Hellenic Statistical Authority’s Population Census of 2011 and 2021, respectively (https://www.statistics.gr/en/home/). Weekly death data were obtained from Eurostat (https://ec.europa.eu/eurostat/databrowser). All statistical analyses were conducted in R 4.2.2 using the excessmort R package.6 To estimate the percentage deviation from expected mortality we employed a quasi-Poisson Generalized Linear Model, which accounts for temporal trends in mortality, seasonality (e.g., higher mortality rates due to extreme temperatures during summer or seasonal influenza during winter), and natural random variation (Method Note S1).
Results
Storm Daniel predominantly affected the prefectures of Karditsa/Trikala and Magnesia in the Thessaly Region, and the prefecture of Fthiotida in Central Greece. Our analysis revealed that during the week of the landfall, all-cause mortality in Karditsa/Trikala was 40% (95% CI: 13%–66%) higher compared to the expected mortality based on 2015–19 mortality rates. Similarly, in Magnesia and Fthiotida, there was an increase of 39% (95% CI: 7%–70%) and 57% (95% CI: 19%–95%) over expected mortality, respectively (Fig. 1A). During the first week of the flooding, 21 (95% CI: 7–36) excess deaths were recorded in Karditsa/Trikala, 15 (95% CI: 3–28) in Magnesia, and 18 (95% CI: 6–30) in Fthiotida, amounting to 54 (95% CI: 31–77) overall in the three flood-affected prefectures (Fig. 1B).
(A) Percentage change over expected mortality in Karditsa/Trikala, Magnesia, and Fthiotida (September 2022 – December 2023). The blue lines and the blue ribbons correspond to the percent change over expected mortality and the corresponding 95% CI, respectively. The red dashed lines correspond to the landfall of Storm Daniel. The grey dots represent the observed weekly death counts. B) Cumulative excess deaths from September 5th, 2023, to December 3rd, 2023, in Karditsa/Trikala, Magnesia, and Fthiotida. The red points and red ribbons correspond to the estimated cumulative excess deaths and the corresponding 95% CI, respectively.
Cumulative excess deaths were also estimated from the day of the landfall on 5 September 2023, until 3 December 2023. During this period, 155 excess deaths (95% CI: 100–210) were recorded in Karditsa/Trikala, 78 (95% CI: 34–122) in Magnesia, and 102 (95% CI: 60–144) in Fthiotida (Fig. 1B). In the three Storm Daniel-stricken prefectures, a total of 335 excess deaths (95% CI: 250–420) was recorded during the 90 days following the storm event (Table S2). It is worth noting that during the period of interest, there was no evidence of a generalized increase in mortality rates in Greece, with the exception of four not affected by the Storm prefectures (Serres, Kastoria, East Attica, and Laconia/Messenia) located in four geographically dispersed Regions in Greece (Central and Western Macedonia, Attica and Peloponnese, respectively) (Fig. 2). In all these four areas the peaks in excess mortality preceded the landfall of Storm Daniel (Fig. S2), suggesting that the recorded excess deaths might be related to other local events (e.g., wildfire in East Attica during July 2023).
Cumulative excess deaths from 5 September 2023 to 3 December 2023, across the prefectures of Greece. The black points and gray ribbons correspond to the estimated cumulative excess deaths and the corresponding 95% CI, respectively. For the flood-affected areas, magnesia, Karditsa/Trikala, and Fthiotida, the red lines and the red ribbons correspond to the estimated cumulative excess deaths and the corresponding 95% CI, respectively.
Discussion
Our study shows a significant increase in all-cause mortality during the week of Storm Daniel’s landfall and the subsequent 90 days, compared to the expected mortality rates based on 2015–19 historical trends. The revealed discrepancy between the reported flood fatalities (17 deaths) and the excess deaths (335 over a period of three months) identified in our study has important public health implications.
Typically, flood events have immediate health effects (mainly due to drowning, trauma, hypothermia, or electrocution), mid-term (mainly due to water-borne and vector-borne diseases), and long-term health consequences (increased incidence of chronic conditions due to interruption of medication or difficulties in accessing healthcare) on flood-affected communities.7 Conventional approaches focusing only on the immediate flood fatalities, tend to underestimate the overall health risks faced by flood-affected communities. The discrepancy observed between the reported flood fatalities and the excess deaths recorded in our study, can be interpreted on the basis that the former capture the direct health consequences, while the latter capture the cumulative health consequences of Storm Daniel in Greece.
Our study has certain limitations. Excess deaths in Thessaly Region and Central Greece reported in our study might be the result of the combined effects of Storm Daniel and Storm Elias, even though no significant increase in mortality was recorded during the week of Storm Elias’ landfall. Similarly, although our model controls for seasonality (but not temperature per se) and no wildfires or heatwaves were reported in Thessaly and Central Greece during the period of the study (September until December 2023), other factors might have contributed to the observed increased mortality. Future data, such as age-specific and cause-specific mortality rates, will allow the identification of the main drivers of the excess deaths recorded in our study.
Hitherto, only three studies have investigated all-cause mortality or mortality due to natural causes following flood events in Europe, resulting in inconsistent findings.8–10 A recent global study covering 35 countries (12 of which European ones) showed that exposure to flooding was associated with increased risk for all-cause mortality for up to 60 days following a flooding event,11 a finding which is consistent with our study. In Greece, existing studies have focused solely on immediate flood fatalities, examining their temporal and spatial patterns.12 Our study is the first to investigate all-cause mortality following a flash flood in Greece.
Flood-related health impacts are complex and spread over time. Studies revealing the cumulative and long-term health effects of flooding, remind and reinforce the crucial importance of designing monitoring systems, such as epidemiological surveillance and mortality monitoring systems, to assess the health risks in real time, and address the root causes of emerging morbidity and mortality in flood-affected populations.4 This public health principle needs to be applied in all current and future anthropogenic extreme weather events and disasters.
Acknowledgments
We thank Professor Manolis Kogevinas for his constructive comments and editing that improved the final version of the manuscript.
Conflict of interest
The authors have no conflicts of interest to declare.
Funding
No funding has been received for this study.
Data availability
The data underlying this article are available in the article and its outline supplementary material.
Evdoxia Valavani, PhD Researcher
Vasilis Bellos, Assistant Professor
Fani Apostolidou-Kiouti, Research Assistant
Alexis Benos, Emeritus Professor
Elias Kondilis, Associate Professor
