Global Health Impacts of Floods: Epidemiologic Evidence

INTRODUCTION

Floods are the most common natural disaster in both developed and developing countries, and they are occasionally of devastating impact, as the floods in China in 1959 and Bangladesh in 1974 and the tsunami in Southeast Asia in December 2004 show (1). Their impacts on health vary between populations for reasons relating to population vulnerability and type of flood event (2–5). Under future climate change, altered patterns of precipitation and sea level rise are expected to increase the frequency and intensity of floods in many regions of the world (6). In this paper, we review the epidemiologic evidence of flood-related health impacts. The specific objectives were as follows:

1. to summarize and critically appraise evidence of published studies, covering flood events in all regions of the world, and

2. to identify knowledge gaps relevant to the reduction of public health impacts.

MATERIALS AND METHODS

We developed a search algorithm to identify the published literature concerning the health impacts of flood events. A search was made of the BIDS (Bath Information and Data Services, Bath, United Kingdom), CAB Abstracts (Commonwealth Agricultural Bureau International, Wallingford, Oxfordshire, United Kingdom), PsychInfo (American Psychological Association, Washington, DC), Embase (Elsevier B. V., Amsterdam, the Netherlands), and Medline/PubMed (National Library of Medicine, Bethesda, Maryland) reference databases using combinations of terms for flooding and selected health outcomes as terms in the title, keywords, or abstract (figure 1). Terms from Medical Subject Headings (MeSH; National Library of Medicine) were used where relevant. We excluded papers that addressed only population displacement, economic losses, and disruption of food supplies. The search found 3,833 references, of which 212 were identified as epidemiologic studies from review of the abstract and/or title. The scientific quality of these papers was assessed on a case-by-case basis. In drawing inferences, we made no exclusions from these 212, but we gave greatest weight to studies based on epidemiologic designs with controlled comparisons. We report and reference the main findings in this review.

Floods are classifiable according to cause (high rainfall, tidal extremes, structural failure) and nature (e.g., regularity, speed of onset, velocity and depth of water, spatial and temporal scale), but in this review we discuss impacts according to health outcome. The influence of flood characteristics on health impacts is discussed where appropriate.

RESULTS

Mortality

Deaths associated with flood disasters are reported in the EM-DAT disaster events database (Centre for Research on the Epidemiology of Disasters, Ecole de Santé Publique, Université Catholique de Louvain, Brussels, Belgium) (1) and also in two reinsurance company databases (www.munichre.com, www.swissre.com). These databases include little epidemiologic information (age, gender, cause), however. Flood-related mortality has been studied in both high- (7–17) and low-income (18–21) countries. The most readily identified flood deaths are those that occur acutely from drowning or trauma, such as being hit by objects in fast-flowing waters. The number of such deaths is determined by the characteristics of the flood, including its speed of onset (flash floods are more hazardous than slow-onset ones), depth, and extent (3). Many drownings occur when vehicles are swept away by floodwaters (12, 14, 15, 17). Evidence relating to flash floods in high-income countries suggests that most deaths are due to drowning and, particularly in the United States, are vehicle related (9, 22). Information on risk factors for flood-related death remains limited, but men appear more at risk than women (22). Those drowning in their own homes are largely the elderly.

Although the risk of deaths is most obviously increased during the period of flooding, in a controlled study of the 1969 floods in Bristol, United Kingdom, Bennet (8) reported a 50 percent increase in all-cause deaths in the flooded population in the year after the flood, most pronounced among those aged 45–64 years. Few other studies have examined such a delayed increase in deaths, but it was also reported by Lorraine (11) in relation to the 1953 storm surge flood of Canvey Island, United Kingdom, but not in two Australian studies (7, 10).

Inconclusive evidence for diarrheal deaths has been reported from several studies of floods in low-income countries. Surveillance data showed an apparent increase of mortality from diarrhea following the 1988 floods in Khartoum, Sudan, but a similar rise was also apparent in the same period (May–July) of the preceding year (20). Routine surveillance data and hospital admissions records similarly showed diarrhea to be the most frequent (27 percent) cause of death following the severe 1988 Bangladesh floods, but again the effect of the flood was not separately quantified from seasonal influences (19). Kunii et al. (21) conducted a cross-sectional survey after the 1998 floods in Bangladesh, and of 3,109 people within flood-affected households, seven (0.23 percent) died during (but not necessarily a consequence of) the flood, two from diarrhea, two from suspected heart attacks, and three from undetermined/unrecorded causes.

Injuries

Flood-related injuries may occur as individuals attempt to remove themselves, their family, or valued possessions from danger. There is also potential for injuries when people return to their homes and businesses and begin the clean-up operation (e.g., from unstable buildings and electrical power cables).

In a community survey (108 of 181 households completed the questionnaire) of the 1988 floods in Nîmes, France, 6 percent of surveyed households reported mild injuries (contusions, cuts, sprains) related to the flood (13). In Missouri after the Midwest floods of 1993, injuries were reported through the routine surveillance system. Between July 16 and September 3, 1993, a total of 524 flood-related conditions were reported, and of these 250 (48 percent) were injuries: sprains/strains (34 percent), lacerations (24 percent), “other injuries” (11 percent), and abrasions/contusions (11 percent) (23). Similar data were also reported from Iowa (24).

Surprisingly little information is available on the frequency of nonfatal flood injuries, as they are mostly not routinely reported or identified as flood related. Although the international EM-DAT database (http://www.em-dat.net/) records such injuries, these data are much less robust than are reports of deaths (D. Guha-Sapir, Department of Public Health and Epidemiology, Université Catholique de Louvain, personal communication, 2004).

Fecal-oral disease

In flood conditions, there is potential for increased fecal-oral transmission of disease, especially in areas where the population does not have access to clean water and sanitation. Published studies (case-control studies, cross-sectional surveys, outbreak investigations, analyses of routine data) have reported postflood increases in cholera (25, 26), cryptosporidiosis (27), nonspecific diarrhea (18–21, 28–31), poliomyelitis (32), rotavirus (33), and typhoid and paratyphoid (34) (table 1). Some of the reported relative risks associated with flooding are substantial. In Indonesia, for example, Vollaard et al. (34) found flooding of the home to increase paratyphoid fever, with an odds ratio of 4.52 (95 percent confidence interval (CI): 1.90, 10.73), and Katsumata et al. (27) found it to increase the risk of cryptosporidiosis, with an odds ratio of 3.08 (95 percent CI: 1.9, 4.9).

In high-income countries, the risk of diarrheal illness appears to be low, as shown by studies from the former Czechoslovakia (35), Norway (36), and the United States (23, 24). However, a survey of households affected by Tropical Storm Alison found that diarrhea was significantly associated with residing in a flooded home (odds ratio = 6.2, 95 percent CI: 1.4, 28.0) (37), and Wade et al. (38) also found that flooding of the house or yard was associated with gastrointestinal illness (relative risk (RR) = 2.36, 95 percent CI: 1.37, 4.07). In the United Kingdom, Reacher et al. (39) reported an increase in self-reported gastroenteritis (RR = 1.7, 95 percent CI: 0.9, 3.0) following the Lewes floods of 2001. Another US study (40) investigated an outbreak of oyster-related hepatitis A and, although it was not possible to determine the precise cause of the outbreak, the authors hypothesized that flooding of the Mississippi Valley and discharge of fecal materials in the oyster-growing areas may have been factors.

Vector-borne disease

The relation between flooding and vector-borne disease is complex. Many important infections are transmitted by mosquitoes, which breed in, or close to, stagnant or slow-moving water (puddles, ponds). Floodwaters can wash away breeding sites and, hence, lower mosquito-borne transmission (41). On the other hand, the collection of stagnant water due to the blocking of drains, especially in urban settings, can also be associated with increases in transmission, and there have been numerous such reports from Africa (20, 30, 42, 43), Asia (44, 45), and Latin America (46–50). The 1982 El Niño event, for example, caused extensive flooding in several countries in Latin America and apparently sharp increases in malaria (46, 47). The Mozambique floods of 2000 also appeared to have increased the number of malaria cases by a factor of 1.5–2 by comparison with 1999 and 2001 (30), although the statistics are difficult to interpret in light of the major population displacement that the flood caused. The reports of flood-related outbreaks in India (44, 45) do not provide particularly strong epidemiologic evidence.

There have also been reported increases in lymphatic filariasis (51) and arbovirus disease in Africa (43), Australia (52, 53), Europe (54–56), and the United States (57–59), although few provide epidemiologic data. Having had a flooded basement has been reported to be a risk factor for West Nile virus among apartment dwellers in Romania (54).

Rodent-borne disease

Diseases transmitted by rodents may also increase during heavy rainfall and flooding because of altered patterns of contact. Examples include Hantavirus Pulmonary Syndrome (60) and leptospirosis. There have been reports of flood-associated outbreaks of leptospirosis from a wide range of countries, including Argentina (61), Brazil (62–65), Cuba (66), India (67–70), Korea (71), Mexico (72), Nicaragua (73–75), Portugal (76), and Puerto Rico (77). In Salvador, Brazil, risk factors for leptospirosis included flooding of open sewers and streets during the rainy season (65). After a series of tropical storms in 1995, two health centers in western Nicaragua reported cases of a fever-like illness and some deaths from hemorrhagic manifestations and shock (73). Dengue and dengue hemorrhagic fever were initially suspected (74), but after a case-control study, Trevejo et al. (73) concluded that the most likely explanation was increased exposure to floodwaters contaminated by urine from animals infected with Leptospira species. Although several articles (73–75) implicate contact with floodwaters, specific analyses have not been presented.

Mental health

The World Health Organization recognizes that the mental health consequences of floods “have not been fully addressed by those in the field of disaster preparedness or service delivery,” although it is generally accepted that natural disasters, such as earthquakes, floods, and hurricanes, “take a heavy toll on the mental health of the people involved, most of whom live in developing countries, where [the] capacity to take care of these problems is extremely limited” (78, p. 43). Here, the main evidence relates to common mental disorder, posttraumatic stress syndrome, and suicide.

Common mental disorder (anxiety, depression)

Most studies on the effects of flooding on common mental disorders are from high- or middle-income countries, including Australia (7, 79), Poland (80, 81), the United Kingdom (8), and the United States (82–89), but there is also a study from Bangladesh (90).

Bennet's analysis of the 1968 Bristol floods found a significant increase (18 percent vs. 6 percent; p < 0.01) in the number of new psychiatric symptoms (considered to comprise anxiety, depression, irritability, and sleeplessness) reported by women from flooded compared with nonflooded areas, although there was no significant difference for men. These results broadly agree with the findings for the 1974 Brisbane floods (7), except that in Brisbane men were also affected. Those between 35 and 75 years of age suffered the greatest impacts (79).

Other evidence for impacts on common mental disorder comes from a controlled panel study of adults aged 55–74 years flooded in 1981 and again in 1984 (87, 91). Flood exposure was associated with significant increases in depression (p < 0.005) and anxiety (p < 0.0008) (and also physical symptoms), especially in those with higher levels of preflood depressive symptoms and in those from lower socioeconomic groups—a finding that Phifer et al. suggest supports Logue et al.'s 1981 assertion that “low-income people are more vulnerable to the adverse effects of a disaster” (91, p. 417).

In a longitudinal study, Ginexi et al. (89) were able to compare symptoms for depression in both the pre- and postflood periods, and they found that, among respondents with a preflood depression diagnosis, the odds of a postflood diagnosis increased significantly (odds ratio = 8.55, 95 percent CI: 5.54, 13.2). A more recent case-control study from the United Kingdom (39) found a fourfold increase in psychological distress among adults whose homes were flooded compared with those whose homes were not (RR = 4.1, 95 percent CI: 2.6, 6.4).

On the other hand, more equivocal evidence comes from two case-control studies of the mental health impacts (82, 83) of Tropical Storm Agnes, which caused extensive flooding in Pennsylvania in 1972. The first study, conducted 3 years postflood, focused on working-class males aged 25–65 years; the second, conducted 5 years after the event, focused on women aged 21 years or more. In both cases, respondents from flooded households reported more mental health symptoms than did nonflooded respondents, but differences were not statistically significant. The authors speculate that “the failure to find a stronger relationship … may, in part, be the result of the length of time which had elapsed since the disaster impact” (83, p. 239).

Comparatively few studies have examined mental health impacts on children, but an exception is the 1993 study by Durkin et al. (90) that found postflood changes in behavior and bedwetting among children aged 2–9 years. Before the flood, none of the 162 children were reported to be very aggressive; postflood, 16 children were found to be very aggressive toward others. Bedwetting increased from 16.8 percent before the flood to 40.4 percent after it. In the Netherlands, Becht et al. (92) interviewed 64 children and their parents (n = 30) 6 months postflood and found 15–20 percent of the children having moderate to severe stress symptoms. Other studies after the 1997 floods in Opole, Poland (81, 93), also suggest long-term negative effects on the well-being of children aged 11–14 years and 11–20 years, with increases in posttraumatic stress disorder (PTSD), depression, and dissatisfaction with life. Six months after Hurricane Floyd, similar findings were found by Russionello et al. (94) for children aged 9–12 years.

Posttraumatic stress disorder

PTSD “arises after a stressful event of an exceptionally threatening or catastrophic nature and is characterised by intrusive memories, avoidance of circumstances associated with the stressor, sleep disturbances, irritability and anger, lack of concentration and excessive vigilance [and the specific diagnosis of PTSD] has been questioned as being culture-specific, and may be overdiagnosed” (78, p. 43). Nonetheless, studies showing increases in PTSD following floods come from Europe (95, 96) and North America (97–99).

McMillen et al. (98), who interviewed those affected by the 1993 Midwest floods, found that 60 subjects (38 percent) met the criteria for postflood psychiatric disorder and 35 (22 percent) met the criteria for flood-related PTSD. However, the limitations, recognized by the authors, included retrospective data collection, self-selection of interviewees, self-reporting, and the absence of a control group. Similar limitations applied only to a study of 1997 flood victims in the Central Valley of northern California (99): 19 percent (24) of the 128 participants who completed the acute stress disorder questionnaire met the criteria for the disorder's diagnosis, and of the 73 participants who completed the 1-year follow-up, seven (10 percent) met the criteria for full PTSD. Studies of the 1996 flooding in the Saguenay/Lac St. Jean region of Quebec, Canada, also suggest substantial increases in emotional distress and PTSD among flooded respondents (97). Evidence from Puerto Rico and from work by Norris et al. (95) suggests that PTSD symptoms are influenced by the extent of flooding, culture, and age. The difficulties of interpretation are demonstrated in a study by Verger et al. (100), who examined the mental health impacts 5 years after the 1992 floods in Vaucluse, France. They concluded that the subjects' reports of their disaster-related experiences (significantly worse for women and subjects older than 35 years) “are by nature subjective … and not entirely reliable” (100, p. 440).

Suicide

Evidence about suicides in relation to flooding is very limited. One US analysis that was initially interpreted as showing evidence for increased suicides in the 4 years after natural disasters (101) was subsequently retracted, with the conclusion that “the new results … do not support the hypothesis that suicide rates increase” (102, p. 148). A paper from China (103) reports that suicide rates in the Yangtze Basin, an area affected by periodic flooding, are 40 percent higher than in the rest of the country, but there is no direct epidemiologic evidence to suggest that the difference is attributable to flooding.

Other health outcomes

In addition to the health outcomes detailed above, our review identified reports of other flood-related health impacts, including Acanthamoeba keratitis (104), epilepsy (105), leukemia, lymphoma, spontaneous abortion (106), melioidosis (107), effects of chemical contamination (108, 109), infection from soil helminths (110), and schistosomiasis (111–113). Most were isolated reports or provided only weak evidence of a possible flood link.

DISCUSSION

There is a surprisingly limited evidence base about the health effects of floods, particularly in relation to morbidity. This may in part be due to the difficulty of carrying out rigorous controlled epidemiologic studies of floods, especially in low-income countries. Evidence on public health interventions (e.g., the need for measures to reduce the spread of infectious disease, dealing with mental health impacts, targeting of vulnerable groups) appears particularly limited. We found no studies on the effectiveness of public health measures, including early warning systems. Nonetheless, the wide range of risks to health and well-being, both physical and mental, is understood, though there remains scientific uncertainty about the strength of association and public health burden for specific health effects. The immediate risks of trauma and death are generally clear, but it seems that longer-term impacts, specifically on mental well-being, are often underestimated and probably receive too little attention from public health authorities. These and the location-specific infectious disease risks require further study.

In terms of public health responses, some caution is required in drawing general lessons from a global literature, because floods vary greatly in their character and in the size and vulnerability of the populations they affect. The evidence is also dominated by studies of slow-onset floods in high-income countries that may have little relevance to flash floods and floods in low-income settings. Some floods are catastrophic and affect thousands of people who may have little capacity to protect themselves, as was the case in Mozambique in 2000 and Bangladesh in 2004. At the other end of the spectrum is the small-scale flood in a high-income country where emergency and support services are better able to cope with the immediate and longer-term effects.

Comparison of different flood events suggests that the risk of death is influenced by both the characteristics of the flood (e.g., its scale and duration, the suddenness of onset, the velocity and depth of water, the lack of warning) and of the population that it affects. Floods with the largest mortality impacts have occurred where infrastructure is poor and the population at risk has limited economic resources.

The formulation of policies to protect against flood-related health risks is limited by the paucity of evidence on epidemiologic risk factors and public health interventions. We identified the following knowledge gaps:

• the mental health impacts of flooding, especially the long-term impacts, and their principal causes, which have been inadequately researched even in high-income settings;

• the nature and magnitude of mortality risks in the period after flooding;

• quantification of the risks of infectious and vector-borne diseases following floods;

• the effectiveness of warning systems and public health measures in reducing flood-related health burdens;

• the health-related costs of flooding that are often given little weight in decisions about specific interventions; and

• quantification of the degree to which climate and land-use change will contribute to flood risk and associated health burdens in different settings.

Epidemiologic study of these questions presents particular challenges because of the unpredictability of the timing and location of floods; the necessity for retrospective designs, therefore, which may suffer from recall bias; and the difficulty of obtaining appropriate comparison groups. The availability of geographically coded routine data in some countries may offer opportunities for future research.

There is much that can be done to reduce health and other impacts through public education, emergency service planning, and the implementation of early warning systems (114). On the other hand, there are clear research needs to improve understanding of the health risks in different settings and of the social and cultural modifiers of those risks. Some of the deficiencies in evidence are apparent from the summary we provide in this report. There is also more to understand about the long-term consequences of flooding on health and about the mechanisms by which such consequences can best be prevented or alleviated.

This study was supported by a grant from the Tyndall Centre for Climate Change Research for the project “Health and Flood Risk: A Strategic Assessment of Adaptation Processes and Policies” (project code T3.31). P. W. is supported by a Public Health Career Scientist Award (National Health Service (NHS) Executive grant CCB/BS/PHCS031).

References