Background
Climate change leads to extreme weather events that can have devastating effects on human society and the environment [
1]. Climate change models predict an almost exponential increase in atmospheric water-holding capacity, with increasing temperature and subsequent rise in atmospheric water content leading to an increase in extreme rainfall events in the Northern Hemisphere [
2]. As extreme rainfall events are expected to increase in frequency, intensity and duration, pluvial flooding events are also expected to increase in urban settings [
3]. Indeed, in the last few decades, countries like the Netherlands have observed an increasing trend in extreme rainfall events, causing recurrent pluvial flooding, especially in urban areas [
4].
Most urban sewage drainage systems in the Netherlands can only support an intensity of ~20 mm of rainfall per hour. Extreme rainfall events (usually > 30 mm rainfall/hour for > 1 h) may overwhelm this drainage capacity, leading to 10-50 cm of pluvial floodwater to accumulate on the surface [
3,
5]. Urban pluvial flooding often entails street flooding and/or flooding of combined sewerage system, where rainwater mixes with sewage water, thereby heavily contaminating floodwater with fecal material. In this way, floodwater becomes a possible vehicle of several pathogens such as noroviruses, enteroviruses, or
Campylobacter [
6], many of which are recognized causative agents of gastrointestinal and respiratory infections.
People will inevitably be exposed to this floodwater, especially in urban settings and when engaged in post-flooding cleaning or recreational activities (e.g. swimming, playing in water, etc.). It is also possible that people are accidentally exposed to floodwater when passing through it (e.g. walking, cycling, etc.) or because of splash exposure. Increased flooding events may therefore pose a threat for public health, particularly for the development of acute gastroenteritis (AGE) or acute respiratory infection (ARI) via ingestion and inhalation of, and dermal contact with, pathogens in floodwater as is shown by the study of De Man, et al. [
6]. However, little research is available on health risks associated with urban pluvial flooding.
With a focus on a high-income country like the Netherlands, the aim of this study was to quantify the AGE and ARI risks associated with exposure to pluvial floodwater, as well as to identify specific risk factors for AGE and ARI, in pluvial flood-ravaged urban areas.
Discussion
This study suggests an association between direct exposure to urban pluvial floodwater and the occurrence of both AGE and ARI in a high-income country like the Netherlands. Identified risk factors were contact with floodwater, such as skin contact with floodwater (for both AGE and ARI), post-flooding cleaning operations (for both AGE and ARI) and cycling through floodwater (for AGE).
This study included 1656 individual participants, resulting in a response rate of 21%. It is unknown how representative our sample is with regard to the target population (i.e. the people who experienced flooding), as we did not know the demographics of the people invited to participate in the study. Addressee-unknown invitations were sent to all house numbers of streets in which flooding had been reported. Therefore, it was impossible for us to know who the invitees were. It could be argued that representativeness of the sample could be assessed based on the demographics of the Dutch general population. However, this is not an optimal solution, because the ‘target population’, i.e. the people who experienced flooding, does not necessarily mirror the general population.
A response rate of 20-30% is commonly reported in this type of retrospective studies where self-reported health complaints are investigated [
17,
18], and some studies report even lower response rates [
19]. A low response rate could render these studies prone to, for example, selection bias [
18]. The type of selection bias that might have played a role in this study is self-selection bias, by which the group of participants who were exposed and diseased is overrepresented, as people who experienced flooding and health complaints are particularly motivated to complete and return the questionnaire [
18]. Another limitation that is inherent to retrospective studies that use self-reported data is recall bias, where people might have forgotten (mild) AGE and ARI episodes [
18]. On the other hand, an overestimation of the incidence might also have occurred, because of ‘telescoping’ (i.e. when people remember episodes as being more recent than they actually are) [
18]. Overall, retrospective studies with self-reported data like ours tend to produce incidence estimates that overestimate the true incidence [
20,
21]. A major advantage of these type of studies is that they allow for collection of data about ARI and AGE cases that are not reported to the General Practitioner (GP), i.e. cases that do not require medical attention.
Previous population-based studies on AGE in the Netherlands showed a baseline incidence of 0.95 episodes/person-year. In our study targeting pluvial flood-ravaged areas, the incidence of AGE was estimated at 1.69 episodes/person-year, which is almost twice as higher. This could suggest that flooding events increase AGE risk in the affected population also as compared to the baseline AGE incidence in the whole country [
19]. Likewise, the incidence of ARI derived from our study was 3.74 episodes/person-year, which is more than twice as higher than the incidence of influenza-like illness (ILI) in the general Dutch population (1.72 episodes/person-year) [
18]. Although ARI is not the same as ILI, they both include respiratory diseases and give an indication of the number of cases with respiratory disease in the country. However, direct comparison with the ARI/AGE incidence of the general population is hampered by the fact that we used a shorter recall period (2 vs. 4 weeks), which generally produces higher incidence estimates [
20,
21]. Moreover, our study might have been subject to reporting and selection bias, as is described above.
De Man, et al. [
3] performed a comparable study in the Netherlands at a much smaller scale (149 households) and lacked sufficient numbers of outcome events to use well-defined (standardized) AGE or ARI syndrome definitions. The larger scale of our study (699 households) allowed for estimates that are more precise and for the use of standardized definitions for both AGE and ARI. In agreement with De Man, et al. [
3], we found that participants exposed to pluvial floodwater were more likely to develop gastrointestinal and respiratory complaints (Table
2).
We also investigated differences in risk factors for AGE and ARI in children and adults, but the number of children enrolled in the study was rather low, so the analyses for this age category were underpowered (Table
2). This may be reason as to why the association between ARI and exposure to floodwater was positive but not statistically significant. Low statistical power did also not allow De Man, et al. [
3] to study differences in risk factors for AGE and ARI between adults and children. However, it is evident that children may display certain risk factors (e.g. playing in/with floodwater) more often than adults (e.g. post-cleaning operations). It was also shown by De Man, et al. [
6] that children are more likely to ingest floodwater compared to adults (1.7 ml vs. 0.016 ml), because they play in or around floodwater, and therefore have a higher risk of AGE (33% vs. 3.9%).
Sanitary sewer overflow (SSOs) events were shown to be associated with gastrointestinal illness (GI; emergency room (ER) visits with a primary diagnosis of GI) [
22]. However, specific risk factors leading to exposure and eventually GI were not identified. Furthermore, SSOs probably entail different risk factors (swimming, contaminated drinking water) compared to flooding of combined sewerage systems/street flooding as studied in this paper.
For ARI and AGE, skin contact was identified as a risk factor, which is probably a proxy for ingestion or inhalation of contaminated floodwater (Tables
3 and
4). This may happen, for example, when people do not wash their hands after floodwater contact or people splash aerosolized water particles in their face, while only reporting skin contact [
23]. This could possible also explain why for example ingesting droplets of floodwater was not identified as risk factor, as people only report the most evident risk factors, i.e. skin contact.
Post-flooding cleaning operations were associated with increased risk for both ARI and AGE (Tables
3 and
4), which was only associated with ARI in De Man, et al. [
3]. It could be that we were able to identify it as a risk factor for AGE due to our larger sample size. Our finding is supported by previous literature describing that cleaning leads to aerosolisation of contaminated water droplets and their inhalation/ingestion, explaining why it could be a risk factor for AGE [
23]. Similarly, cycling through floodwater was a significant exposure for AGE (Table
3), as water droplets could splash into their face and mouth. As reported in De Man, et al. [
3], these associations may reflect, to some extent, the primary transmission routes of the pathogens in question, i.e. inhalation for ARI and ingestion for AGE. Indeed, several are typically associated with flood-ravaged settings in developed countries. Sales-Ortells and Medema [
24] found, for example,
Campylobacter in all water plaza samples in which people recreate, leading to a risk of developing AGE for those people.
A limitation of this study is that there were no water samples obtained from pluvial floodwater as well as no faecal samples from the participants. Therefore, the causative agents remain unknown. However, pluvial floodwater in the Netherlands was shown to always be contaminated with faeces, as was demonstrated by the presence of
E. coli, intestinal enterococci, and enteropathogens such as enterovirus, norovirus and
Campylobacter in water samples of pluvial floodwater [
6].
Furthermore, there was no data on sex for 23% of the participants, primarily children, which was because the questionnaire only collected information regarding sex of the person filling in the questionnaire, but not for other household members. This explains most of the missings and is unlikely to affect our results because sex effects on AGE and ARI are less likely to be seen in children [
19,
25]. Moreover, we do not know whether participants used a measure tape to measure the height of the floodwater in cm to answer the question. Although the magnitude of exposure came out as a risk factor in the analysis for AGE and ARI, those data could potentially be erroneous as most of the collected data are self-reported estimates of the true height of the floodwater.
Chronic diseases, which the models were corrected for, may be effect modifiers of the association between floodwater exposure and AGE/ARI. This was checked by running the models with and without chronically diseased cases therein: results can be found in Additional file
4: Table S3. They show that chronic diseases were not significant effect modifiers. However, they were associated with the outcome, so the models were always corrected for underlying chronic disease.
Future prospective studies are needed to confirm the associations found in this study. An example of such a study could be a longitudinal study that monitors the rainfall pattern in a given area. As soon as an extreme rainfall event occurs (> 30 mm rainfall/hour for > 1 h), it should be checked whether flooding has also occurred in that specific area. A representative selection of the inhabitants of the flooded area (which have been actively enrolled at the beginning of the study, perhaps upon financial incentive) should use health diaries to record their symptoms and their exposure to floodwater. This would reduce both recall and selection bias. In order to further confirm causality, floodwater and patient faeces samples should also be taken to identify and characterize potential pathogens.
Due to climate change, extreme rainfall events will occur more frequently, leading to more people being exposed to pluvial floodwater in urban areas in countries like the Netherlands [
2]. As this study suggests, such events are likely to lead to increased risk of health effects. Generally, people in developed countries tend to not perceive the associated health risks. The Netherlands is a highly populated country (~400 people/km
2) [
3], so urban flooding will not pass easily unnoticed. Recently, awareness has been arisen, with Dutch residents demanding municipal services to improve drainage systems as to prevent pluvial flooding [
26]. In response, governmental authorities are promoting greenness in urban areas to facilitate natural drainage of the water in the soil [
26]. This study adds another perspective to this debate and shows that it is necessary to take proper care of water drainage systems/sewage systems in terms of their drainage capacity to mitigate health risks in urban areas.