Introduction
The number of weather-related disasters has increased over the last 20 years. Nearly one-half of the disasters were floods, which killed 157,000 people and affected 2.3 billion. Storms were one-quarter of the disasters but were responsible for a larger proportion of fatalities, causing 220,000 deaths and affecting 660,000 people. 1 As climate change transforms the frequency and severity of weather events in the coming years, the impact of floods and storms is likely to worsen.
The health care response to a disaster should be guided by the known and expected needs of the affected population. It should be prepared for needs caused by the direct effect of the disaster and for the indirect outcomes of the initial impact. This can be achieved by knowing if and how the health of a population changes after a disaster. Response plans should use the best available evidence to understand the immediate, short-term, and long-term health outcomes after a disaster, and how ordinary health care needs are affected.
Despite the commonality of flood and storm disasters, few reviews cover the quantifiable changes in health that occur afterwards or compare the differences. Previous systematic reviews have combined floods and storms in the analysis,Reference Du, FitzGerald, Clark and Hou 2 , Reference Alderman, Turner and Tong 3 but storms have the additional hazard of strong winds. Other reviews have summarized the impact of flood or storm disasters on mortality and the causes, risk factors, or burden of specific diseases.Reference Tempark, Lueangarun, Chatproedprai and Wananukul 4 - Reference Goldman, Eggen, Golding and Murray 7 These approaches take a preventative public health perspective and are important for reducing a disaster’s health impact.
Floods and storms affect health in various ways.Reference Du, FitzGerald, Clark and Hou 2 , Reference Alderman, Turner and Tong 3 , Reference Goldman, Eggen, Golding and Murray 7 Contact with flood waters and high-speed winds during a storm, contact with debris, and evacuation and clean-up activities can cause injury. Floods and storms can destroy residential buildings, leading to displacement, overcrowding, increased exposure to animals and insects, and worsened living conditions. They also can damage infrastructure (such as sewage systems or electrical supply), the agricultural process, and health care facilities. This can indirectly lead to the transmission of infectious diseases, increased physiological stress, exacerbation of existing conditions, malnutrition, and lack of access to preventative and curative health care. Improperly using carbon-based fuels during electrical power outages can result in poisonings. However, in order to guide health care activities afterwards, it is essential to know to what extent the burden of disease varies during different phases after floods and storms.
The aim of this systematic review is to elucidate the human health problems following flood and storm disasters. The review will highlight the health problems and needs over time and the knowledge and evidence gaps discovered during the review.
Report
A search strategy was developed in collaboration with experienced librarians. A literature search using Medline (US National Library of Medicine, National Institutes of Health; Bethesda, Maryland USA); Cinahl (EBSCO Information Services; Ipswich, Massachusetts USA); Global Health (EBSCO Information Services; Ipswich, Massachusetts USA); Web of Science Core Collection (Thomson Reuters; New York, New York USA); Embase (Elsevier; Amsterdam, Netherlands); and PubMed (National Center for Biotechnology Information, National Institutes of Health; Bethesda, Maryland USA) was conducted in June 2015. The search used keywords for floods, storms, mortality, and morbidity (Supplement 1; available online only). Details of the protocol for the study were registered on the PROSPERO (Centre for Reviews and Dissemination, University of York; York, United Kingdom) systematic review register (ID: CRD42016033554).Reference Saulnier, Brolin Ribacke and von Schreeb 8
The primary inclusion criteria for the search were peer-reviewed journal articles on human health written in English. Floods or storms that occurred before 1980 were excluded. No geographical limitations were set. The full eligibility criteria are listed in Table 1.
A disaster was defined using the Centre for Research on the Epidemiology of Disasters’ (Brussels, Belgium) International Disaster Database guidelines 9 as an event that met at least one of the following criteria: 10 or more people reported killed; 100 or more people reported affected; or a declaration of a state of emergency or call for international assistance. Health problems were defined as states related to the physical condition that were harmful or unwelcome and could lead to ill health, morbidity, or mortality. Storm refers to an organized, rotating storm system that originates over water and reaches sustained wind speeds of at least 34 knots. 10 The generic terms for storms are hurricanes, typhoons, or cyclones. For this review, flood refers to a partial or complete inundation of normally dry land with water that was not the result of a storm.
Initial searches produced 15,447 articles. Articles were independently screened by two authors. Screening titles for eligibility left 914 articles. The remaining abstracts were critically reviewed and 148 studies were read in full; 35 were excluded (Figure 1). All articles were organized in an EndNote X7.7 library, version 17.5.0.9325 (Thomson Reuters; New York, New York USA). Data were extracted by a trained assistant into a Microsoft Excel spreadsheet, version 15.0.4841 (Microsoft Corporation; Redmond, Washington USA) created by the first author. The extracted data were double-checked for accuracy and completeness by the first author. Three authors of four studies were contacted by email to provide information that was not available in the published material.
Health problems were sorted into five groups: (1) injuries and poisonings; (2) infectious and parasitic diseases; (3) noncommunicable diseases (NCDs) and chronic illnesses; (4) contact with health services; and (5) other. Groupings were based on the International Classification of Diseases and Related Health Problems 10th Revision (ICD-10) 11 and the medical input of the last author and other qualified clinicians. For the purpose of this study, all reported diarrhea was grouped into the infectious and parasitic diseases category, based on the World Health Organization’s (WHO; Geneva, Switzerland) recommended surveillance standards for diarrhea, 12 unless the article specified another cause.
Due to the large variation in study design and outcomes, data were analyzed by narrative synthesis. Studies that included a control or comparison in their design, and indicated a change in health problems, were selected and presented as key articles for the analysis. Studies were individually assessed for quality. The review followed the PRISMA reporting guidelines.Reference Moher, Liberati, Tetzlaff, Altman and Group 13
Results
The final analysis included 113 studies. Twenty-three articles studied floods, 89 studied storms, and one article included both. The disasters occurred between 1985 and 2014, and the most frequently studied disaster was Hurricane Katrina (2005; Gulf Coast, USA) with 25 articles. There were 77 articles that studied disasters in high-income countries (72 of which were in North America), five were in upper middle-income countries, and 14 each in lower middle-income or low-income countries; three studies included more than one country or the income level was unavailable.
Fifty-two articles were descriptive or correlational studies and described the types and proportions of health problems or causes of mortality seen after a flood or storm.Reference Ahmed, Khan and Nisar 14 - Reference Ridenour, Cummings, Sinclair and Bixler 66 These descriptive results are summarized in Table 2 for health problems and in Table 3 by causes of death. Table 4 lists all health problems reported in the articles. Table 5 gives the number of articles that reported changes for each health problem grouping.
Abbreviation: NCD, noncommunicable disease.
a Household or community surveys.
b Number of cases.
c No frequency data available for blank cells.
Abbreviation: NCD, noncommunicable disease.
a Number of cases.
b No frequency data available for blank cells.
Abbreviation: NCD, noncommunicable disease.
Abbreviation: NCD, noncommunicable disease.
Injuries and Poisonings
Storms
Poisonings and injuries, especially wounds, were reported as a major source of morbidity after storms. Injuries regularly increased in the period immediately following storms: a study of all emergency care facilities (ECFs) on the island of Hawaii (USA) found an increase in the proportion of injury-related visits within two weeks after a hurricane compared to the five previous years (213 injuries/100,000 people/week increase to 1,461 injuries/100,000 people/week) and an increase in the relative risk (RR) of being injured (RR: 6.85; 95% confidence interval [CI], 5.98-7.87);Reference Hendrickson, Vogt, Goebert and Pon 67 and a study of 113 ECFs found a significant increase of 22.3% in the number of injury-related visits when compared to the same week in the previous year.Reference Miller, Kearney and Proescholdbell 68
Multiple studies reported increases in carbon monoxide (CO) and gasoline poisonings after storms; CO intoxications presenting at an ECF significantly increased from zero (0.0%) to 1.1% of all ECF visits (P=.015) in the three days after landfall.Reference Platz, Cooper, Silvestri and Siebert 69 Within two to four weeks after other storms, electronic surveillance identified a significant increase in CO poisoning,Reference Chen, Shawn and Connors 70 gasoline exposure was 18 to 283 times higher than the previous four years,Reference Kim, Takematsu, Biary, Williams, Hoffman and Smith 71 and calls to poison control centers for CO and gasoline exposure increased from baseline values.Reference Forrester 72 , Reference Forrester 73 A pediatric ECF recorded 13 cases of hydrocarbon and/or bleach poisoning within two weeks after a storm, compared to zero cases in the pre-storm control week and 26 cases for the previous year.Reference Quinn, Baker and Pratt 74
Within one week after storms, the proportion of visits for lacerations and corneal abrasions significantly increased at treatment facilities.Reference Platz, Cooper, Silvestri and Siebert 69 , Reference Alhinai 75 , 76 Within two weeks, the proportion of visits for open wounds increased.Reference Quinn, Baker and Pratt 74 The risk of contusions (RR: 3.29; 95% CI, 2.36-4.59) and open wounds (RR: 17.3; 95% CI, 13.0-23.0) also rose at ECFs in the two weeks following a storm compared to the two weeks prior.Reference Hendrickson, Vogt, Goebert and Pon 67 Within three weeks after storms, significant increases at treatment facilities in visits for infected wounds,Reference Sjoberg and Yearwood 77 lacerations,Reference Sheppa, Stevens, Philbrick and Canada 78 and puncture woundsReference Sheppa, Stevens, Philbrick and Canada 78 were all reported. No studies reported significant changes in wounds more than one month after landfall.
There were less data showing changes in orthopedic and other injuries. One study found a small but significant decrease (P<.001) in the percentage of orthopedic trauma visits during the two weeks after a storm,Reference Alhinai 75 while a second ECF-based study reported an increase in the risk of sprains (RR: 3.02; 95% CI, 2.28-3.99) and fractures (RR: 3.38; 95% CI, 2.14-5.32) compared to two weeks before the storm.Reference Hendrickson, Vogt, Goebert and Pon 67 Other reported changes included a significant decrease in visits for soft tissue wounds at a pediatric ECFReference Quinn, Baker and Pratt 74 and an increase in the risk of burns and head injuriesReference Hendrickson, Vogt, Goebert and Pon 67 in the first week after a hurricane, along with an unexpected number of cases of hypothermia within six weeks after a storm. 79
While animal bites and insect stings are causes of injuries, not outcomes, it should be noted that five studies reported significant increases at treatment facilities in injuries from bites and stings during the first four weeks after storms. 76 , Reference Sheppa, Stevens, Philbrick and Canada 78 - Reference Warner 81
Floods
No studies measured changes in injuries or poisonings after flooding.
Infectious and Parasitic Diseases
Storms
Gastrointestinal illness, leptospirosis, and skin or soft tissue infections generally increased after storms, both at the population and health facility levels. Some studies reported changes in respiratory infections, fevers, and other infectious diseases, but the given information remained inconclusive.
Within seven days, significant increases were seen in diarrhea or gastroenteritis in flooded households and in visits to treatment facilities.Reference Quinn, Baker and Pratt 74 , Reference Alhinai 75 , Reference Waring, DesVignes-Kendrick and Arafat 82 An analysis of national surveillance system data found an increased risk for shigellosis and other infectious diarrhea, peaking at five days post-cyclone (odds ratio [OR]: 3.56; 95% CI, 2.98-4.25).Reference Deng, Xun and Zhou 83 Within five weeks of landfall, the proportion of visits to treatment facilities for diarrhea remained significantly higher than pre-storm levels at an ECF,Reference Longmire, Burch and Broom 80 the RR for visiting an ECF for diarrhea was elevated (RR: 2.0; 95% CI, 1.4-2.8) compared to a same week in the previous year, 79 and a community survey found a slightly higher risk of diarrhea in two of four storm-affected areas (Area 1 OR: 1.6; 95% CI, 1.52-1.65 and Area 2 OR: 1.3; 95% CI, 1.21-1.32) compared to the previous two years.Reference Panda, Pati, Bhattacharya, Koley, Pahari and Nair 84 In the long-term, atypical and significant increases were seen in the incidence of acute diarrhea and dysentery in the eight months after a storm,Reference Myint, Kaewkungwal and Singhasivanon 85 and the average number of cases utilizing health services for intestinal infections doubled in the year after storm landfall (6.5 cases/month to 13.1 cases/month; P<.01).Reference Setzer and Domino 86 Alternatively, two surveillance system studies and one ECF study found no change in gastrointestinal illnesses in the post-storm period.Reference Sheppa, Stevens, Philbrick and Canada 78 , Reference Vilain, Pages and Combes 87 , Reference Greene, Wilson, Konty and Fine 88 Outbreaks of choleraReference Fredrick, Ponnaiah and Murhekar 89 and norovirusReference Palacio, Shah and Kilborn 90 were both identified within two weeks, acute watery diarrhea within six weeks,Reference Bhattacharjee, Bhattacharjee, Bal, Pal, Niyogi and Sarkar 91 and cholera again within three months.Reference Bhunia and Ghosh 92 The outbreaks all declined within one to four weeks after they began.
Leptospirosis outbreaks were identified within two weeks,Reference Vilain, Pages and Combes 87 six weeks, 55 and two monthsReference Trevejo, Rigau-Perez and Ashford 93 after storms. A study of patients who presented at treatment facilities with a dengue-negative fever in the first month after a hurricane found a RR of 4.4 (95% CI, 1.6-12.4) for having leptospirosis.Reference Sanders, Rigau-Perez and Smits 94
Significant increases in impetigo, conjunctivitis, and cellulitis at treatment facilities were reported in the first two weeks after storms,Reference Quinn, Baker and Pratt 74 , Reference Vilain, Pages and Combes 87 and the number of lower extremity cellulitis cases receiving treatment at ECFs increased during the first week after landfall (RR: 2.8; 95% CI, 2.0-2.4) compared to the control period.Reference Lin, Lin, Guo and Chen 95 However, one island-wide surveillance study of all treatment facilities found no increase in visits for skin infections in the week after a cyclone.Reference Vilain, Pages and Combes 87 Eighteen cases of wound-associated Vibrio illness, linked to a hurricane, also were identified two weeks after landfall.Reference Engelthaler, Lewis and Anderson 96
While one study found a significantly higher incidence rate of acute respiratory illness over the course of a cyclone (pre-storm: 4,041.91 cases/100,000 people; storm period: 7,279.7/100,000; post-storm: 4,661.59/100,000)Reference Myint, Kaewkungwal and Singhasivanon 85 and a second study saw a single, unexpected case of legionellosis in a severely affected area,Reference Greene, Wilson, Konty and Fine 88 two treatment facility-based studies and one surveillance study found no significant increases in respiratory infections.Reference Quinn, Baker and Pratt 74 , Reference Alhinai 75 , Reference Vilain, Pages and Combes 87
Information on other infectious diseases was limited. Visits to treatment facilities for unspecified febrile illness increased in the first week after a storm. 79 A study of West Nile neuroinvasive disease found an increased incidence (RR: 2.45; 95% CI, 1.77-3.477) in hurricane-affected areas in the three weeks post-storm compared to the three previous years.Reference Caillouet, Michaels, Xiong, Foppa and Wesson 97 Malaria was diagnosed in only 2.5% and dengue in 1.7% of patients presenting with fever at a treatment facility more than three months post-storm.Reference Beatty, Hunsperger and Long 98 A national surveillance system study reported a slight increase in malaria incidence over the course of a storm but no change in dengue fever.Reference Myint, Kaewkungwal and Singhasivanon 85
Floods
Outbreaks of both gastrointestinal illness and leptospirosis occurred after flooding, but there were little data that supported changes in other infectious illnesses.
Three studies identified diarrhea outbreaks: four flood-associated cholera epidemics after three separate floods, with a median onset time of 8.5 days post-flood;Reference Schwartz, Harris and Khan 99 an outbreak of cryptosporidiosis within four weeks;Reference Gertler, Durr and Renner 100 and 115 disease alerts for acute watery diarrhea and one for blood diarrhea by a surveillance system, indicating the number of diarrhea cases had surpassed a pre-defined threshold over the course of seven weeks from the end of a flood. 101 However, a time series analysis of diarrheal illness and flood water data from a health and demographic surveillance site found no support for a significant increase in diarrhea, including cholera, non-cholera, and rotavirus, in the three years following major flooding.Reference Milojevic, Armstrong and Hashizume 102 Two outbreaks of leptospirosis were confirmed within one monthReference Smith, Young, Wilson and Craig 103 and more than four weeksReference Dechet, Parsons and Rambaran 104 after flooding, while approximately 27% of children presenting to a treatment facility with acute fever of unknown origin were diagnosed with leptospirosis within the first month after flooding.Reference Pradutkanchana, Pradutkanchana, Kemapanmanus, Wuthipum and Silpapojakul 105
Eye infections were seen in a significantly larger proportion of people living in flooded areas within four weeks of a flood compared to those living in non-flooded areas.Reference Bich, Quang, Ha le, Hanh and Guha-Sapir 106 One time series study using health and demographic data found a moderate increase in acute respiratory infections (RR: 1.25; 95% CI, 1.06-1.47) within six months after a flood.Reference Milojevic, Armstrong and Hashizume 102 Two studies of patients presenting at treatment facilities with acute fever identified malaria in 22% of the patients,Reference McCarthy, Haberberger and Salib 107 dengue fever in 29.4%, and scrub typhus in 1.1%Reference Pradutkanchana, Pradutkanchana, Kemapanmanus, Wuthipum and Silpapojakul 105 within two to three months and one month post-flood, respectively, and no change was seen in the proportion of people diagnosed with dengue fever who lived in flooded areas four weeks after a flood.Reference Bich, Quang, Ha le, Hanh and Guha-Sapir 106
A surveillance system study of multiple diseases gave seven alerts for suspected measles, one alert for acute hemorrhagic fever syndrome, two alerts for suspected meningitis, and two alerts for suspected poliomyelitis, but no alerts for acute respiratory infection or unexplained fever in the seven weeks after a flood. 101
Noncommunicable Diseases and Chronic Illnesses
Storms
The articles in this review suggested an increase in cardiovascular disease and diabetes outcomes in the short- and long-term after storms, and nutritional outcomes manifested in the long-term after some storms.
While one island-wide surveillance study of all treatment facilities found no significant increase in visits for cardiac decompensation in the week after a cyclone,Reference Vilain, Pages and Combes 87 the proportion of cardiovascular complaints significantly increased at treatment facilities, especially in people 45 years and older, in the two weeks to one month after a hurricane.Reference Hendrickson, Vogt, Goebert and Pon 67 , Reference Cookson, Soetebier and Murray 108 The incidence of stroke increased by seven percent (attributable rate ratio [ARR]: 1.07; 95% CI, 1.03-1.11) and myocardial infarction increased by 22% (ARR: 1.22; 95% CI, 1.16-1.28) in areas highly impacted by a hurricane during the first year afterwards,Reference Swerdel, Janevic, Cosgrove and Kostis 109 and the percentage of admissions to hospitals for acute myocardial infarctions increased significantly within two years.Reference Gautam, Menachem, Srivastav, Delafontaine and Irimpen 110
Evaluations for diabetic foot increased significantly at an ECF within four weeks after a cyclone.Reference Sjoberg and Yearwood 77 A study of diabetic patients between six and 16 months after a hurricane found worsened levels of glycated hemoglobin (7.7% to 8.1%; P<.01) compared to measurements taken in the six months before the storm, with a significant linear trend over time.Reference Fonseca, Smith and Kuhadiya 111
The available information on changes in respiratory illness was contradictory. During the first week after a typhoon, a review of insurance claims from typhoon-affected areas found a small increase in patients seeking care for allergic rhinitis (rate ratio: 1.19; 95% CI, 1.15-1.23) but a decrease in patients seeking care for asthma (rate ratio: 0.90; 95% CI, 0.86-0.93);Reference Park, Moon and Ha 112 an island-wide treatment facility study found no significant increase in visits for asthma or chronic obstructive pulmonary disease in the week after a cyclone,Reference Vilain, Pages and Combes 87 although other treatment facilities saw significant increases in visits for asthma two weeksReference Hendrickson, Vogt, Goebert and Pon 67 and one month post-hurricane. 79 In the one to three months after a hurricane, children with chronic conditions were more likely to have worsened asthma (16.3% versus 1.9%; P<.01) than children without chronic conditions.Reference Rath, Donato and Duggan 113 However, no significant relationship was seen between exposure to mold and respiratory allergic response in patients at an allergy clinic in the year after a hurricane.Reference Rabito, Perry, Davis and Levetin 114
The remaining studies on NCDs and chronic illnesses saw significantly less height gained in the three months following a hurricane in nutritionally at-risk children under five years oldReference Simeon, Grantham-McGregor, Walker and Powell 115 and a significant increase in the stunting/underweight one year after.Reference Barrios, Stansbury, Palencia and Medina 116 A spike in renal-related hospital admissions was seen in the month after a hurricane (rate ratio: 2.53; 95% CI, 2.09-3.06) compared to pre-hurricane rates,Reference Howard, Zhang, Huang and Kutner 117 and a study of babies born with neural tube defects after a hurricane suggested a link between significantly decreased consumption of folate by mothers and destruction of food crops by a hurricane.Reference Duff and Cooper 118
Floods
Information on NCDs after floods was limited, but worsened hypertension and diabetes and long-term malnutrition outcomes were both seen. A household survey four weeks after a flood found a significant proportion of those living in flooded areas whose hypertension worsened since the flood (42.9% versus 20.3%; P<.05)Reference Bich, Quang, Ha le, Hanh and Guha-Sapir 106 and a significant increase in glycated hemoglobin (7.6% to 7.9%; P=.002) was seen among diabetic patients in the first year after exposure to flood compared to the 12 months before the event.Reference Ng, Thorpe, Walton, Atkin and Kilpatrick 119 A community survey of children living in villages that had flooded found they were more likely to be stunted and underweight than children from non-flooded villages (adjusted prevalence ratio: 1.86; 95% CI, 1.05-2.44) at 26 to 36 months post-flood.Reference Rodriguez-Llanes, Ranjan-Dash, Degomme, Mukhopadhyay and Guha-Sapir 120 One study found no significant difference in musculoskeletal symptoms from wading through flood waters in the three months after a flood.Reference Sihawong, Janwantanakul and Pensri 121
Contact with Health Services
Storms
Although articles frequently described an increase in contact with health services after storms, especially within the first week and for health maintenance reasons like prescription refills, data-based results were not often given. Of the reported results, an increase in ECF visits for oxygen, medication refills, dialysis, vaccination, or hemodialysis was seen in two studies in the first three daysReference Platz, Cooper, Silvestri and Siebert 69 and first week after a storm, 79 and also in “miscellaneous” visits for vaccinations and routine care to a primary health care facility in the week after a hurricane.Reference Alhinai 75
Floods
No studies measured changes in contact with health services after flooding.
Other Health Problems
Storms
Symptoms were frequently reported in the included studies, and no particular trends were seen. Nutritionally at-risk children under five years of age experienced nasal discharge or cough more often in the first one to three months after a hurricane than in the four months preceding the storm,Reference Simeon, Grantham-McGregor, Walker and Powell 115 and a linear increase (P<.05) was observed for lower and upper respiratory tract symptoms among people exposed to water-damaged homes.Reference Cummings, Cox-Ganser, Riggs, Edwards, Hobbs and Kreiss 122 However, the percent of children self-reporting a lower respiratory tract symptom at least once a week decreased in the eight to nine months after a hurricane and no data were found to support a change in upper respiratory tract symptoms or spirometry test values.Reference Rabito, Iqbal, Kiernan, Holt and Chew 123 Finally, a significant increase in visits for dermatitis at an ECF was seen within one week. 79 By the second week after storms, significant increases were seen in visits to ECFs for dermatologic conditionsReference Quinn, Baker and Pratt 74 and gastrointestinal complaintsReference Longmire, Burch and Broom 80 while significant decreases were seen for genitourinary complains and abdominal pain.Reference Quinn, Baker and Pratt 74 A similar significant decrease in visits for abdominal pain also was seen within four weeks after a cyclone.Reference Sjoberg and Yearwood 77
Floods
Only one paper reported a change in other health problems after flooding: there was an increase in the proportion of those living in flooded areas that were diagnosed with dermatitis within four weeks of a flood (96.5% versus 57.9%; P<.05) compared to those living in non-flooded areas.Reference Bich, Quang, Ha le, Hanh and Guha-Sapir 106
Mortality
Thirty-two studies reported information on mortality, and fifteen studied it exclusively. Deaths were most often presented as a summary of the cases and causes. Table 6 lists all given causes of death.
Abbreviation: NCD, noncommunicable disease.
Storms
Studies at the population level suggested that there was a minimal change in mortality following storms. Two studies using nation-wide and island-wide mortality surveillance records found no significant increases in mortality in the eight to 12 months after storms for the following causes: heart disease, stroke, cancer, respiratory disease, acute respiratory infections, tuberculosis, injury, diarrheal disease, vaccine-preventable illnesses, meningitis, hepatitis, rabies, malaria, dengue fever, or snake bites.Reference Myint, Kaewkungwal and Singhasivanon 85 , Reference Hendrickson and Vogt 124 Notably, the island-wide study did find a significant increase in the number of deaths due to diabetes (RR: 2.61; 95% CI, 1.44-4.74).Reference Hendrickson and Vogt 124
Local studies found more changes. A study of areas highly impacted by a hurricane found that the 30-day mortality rate for both myocardial infarctions (ARR: 1.31; 95% CI, 1.22-1.41) and other cardiovascular events (ARR: 1.22; 95% CI, 1.15-1.30) increased when compared to the five years before the storm; no significant changes in stroke-related mortality were seen.Reference Swerdel, Janevic, Cosgrove and Kostis 109 Over the course of a hurricane season, overall mortality was significantly elevated for two months after the first storm landfall when compared to non-hurricane years; heart, cancer, and accident-related deaths all significantly increased.Reference McKinney, Houser and Meyer-Arendt 125 A study of dialysis-dependent patients in the one-and-a-half years after a hurricane found no significant increase in their mortality risk.Reference Kutner, Muntner and Huang 126
Floods
A study of health and demographic surveillance site data found no support for an increase in all-cause mortality in the three years following a flood.Reference Milojevic, Armstrong and Hashizume 102
Discussion
This review builds on earlier reviews.Reference Du, FitzGerald, Clark and Hou 2 , Reference Alderman, Turner and Tong 3 , Reference Ahern, Kovats, Wilkinson, Few and Matthies 127 It included new articles, separated floods and storms for the analysis, and assessed the timing of health problems. The results of the review show that health does change differently after floods and after storms. The increases in both acute injuries and some infectious diseases, and the increase in treatment facility visits for NCDs and chronic illnesses seen after both floods and storms, is likely to significantly raise the need for both emergency and routine health care services. There was a remarkably small quantity of data on the health effects of floods alone, suggesting a certain degree of extrapolation for knowledge on floods from storm data.
Although outbreaks of gastrointestinal infections and leptospirosis occurred after floods and storms, no articles reported sustained epidemics of any infectious diseases. This may help in dispelling the myth that epidemics are an inevitable consequence of disasters.Reference Watson, Gayer and Connolly 128 The risk of infectious disease transmission is primarily related to a flood or storm’s indirect outcomes, as was seen in this review, where the reported outbreaks were all caused by contaminated water, overcrowded shelter, displacement, and poor sanitation. Public health interventions can be targeted to these outcomes in order to prevent outbreaks.
The increase in treatment for poisonings and wounds seen after storms but not floods is likely the result of strong winds and wind damage during storms. Some wounds were directly sustained during the impact phase of storm (eg, being struck by flying debris); the remaining wounds and the poisonings were indirectly caused in the weeks afterwards (eg, stepping on debris while cleaning up, and CO poisoning from incorrect use of a gasoline-powered generators). The variation in timing and cause means injuries and poisonings will require different medical needs and this should be accounted for during storm responses.
Floods and storms are indirectly responsible for the exacerbation, acute onset, and worsened management of chronic illnesses and NCDs. Disasters affect both individuals and health service delivery, which in turn affects the management and continuity of care for chronic diseases and NCDs. This is reflected in the results of most of the mortality studies and may be connected to the increases seen in people seeking contact with health services. The care needs of people living with chronic illnesses or NCDs and the expected burden of disease have to be clearly defined for the post-disaster period.
An examination of the available information shows that changes in some infectious diseases and NCDs are well-understood, albeit without conclusive information on the strength of association. But discrepancies were seen on whether certain health problems increased or decreased. Overall, the long-term effects of both floods and storms were not well-covered by the literature and changes have not been well-quantified.
A global overview on health after floods and storms has been presented in this review, but the direct and indirect effects of disasters are tied to the local context. Factors such as climate, income level, infrastructure, and residential security will affect the vulnerability and exposure of the population to different health outcomes, and their long-term medical needs. Since the majority of the studies used to generate the results of this review – 77 out of 113 articles – come from studies conducted in high-income settings, the results must be carefully applied to local settings. Health changes also likely differ with the severity of the disaster and the frequency of the population’s exposure to floods or storms. Populations that are regularly exposed to one disaster type may be less vulnerable or better adapted to dealing with the disaster’s direct and indirect effects than an unexposed population, although these gains can be reduced if the disaster is more severe than expected. An analysis of the contextual factors contributing to the health outcomes seen in this review would be a useful complement to these results.
The lack of conclusive results across the review articles means the usual approach to flood and storm responses may be suboptimal. This was specifically noted in two articles read during the review process, which both described medical teams being over-prepared for injuries and under-prepared for chronic or routine complaints after hurricanes.Reference Alson, Alexander, Leonard and Stringer 129 , Reference Jhung, Shehab and Rohr-Allegrini 130 Combining inconclusive data with limited information on the strength of association can intensify the effect. For instance, the numerous studies on injuries included in this review could be the result of a real association between storms and injuries or the result of an increase in the reporting or surveillance of injuries over the disaster period. If this review is to be used to inform a response, the limitations of the results must be considered.
Limitations
The variety in study designs, health problems, and quality of the studies led to a number of limitations. The major limitation is the overall quality of the included articles. Nearly one-half of the articles were descriptive and it was not possible to ascertain the relationship between the health problems and the disaster: how much illness was exacerbated by the disaster, how much was new, and how much followed the same pattern as before? In addition, a lack of systematic reporting and data collection in the studies meant that health problems lacked definitions, definitions varied between studies, and there was minimal information on the severity of the problems. As a result, the analysis was unable to compare and contrast information on the same health problems between studies or quantify the changes in health.
Second, more than one-half of the studies used health facility data, most frequently from single-site hospital or emergency care records, as a data source. This is likely to skew the results of the review towards the health outcomes of people who had access to health care after the flood or storm, and to an over-representation of health problems that require emergency or specialized care. It is likely that this difference is more pronounced in low-income settings. For instance, in Bangladesh, only 26% of study participants who developed health problems that were severe enough to require medical care after a flood visited a health care facility.Reference Kunii, Nakamura, Abdur and Wakai 64 Applying the results of facility-based studies to the population level, and outside of their original context, should be done prudently. An increase in research at the population level, such as cohort or registry-based studies in disaster-prone settings, would not only generate new, comprehensive data, but would help with the interpretation and application of facility-based results.
Third, articles were often not clear on the length of time between the disaster and the onset or acquisition of a health problem. The results have therefore been presented as a period within which the health problems occurred, and can be used as a starting point for more detailed research. Finally, the relationship between a number of health problems and disasters was limited to only one study or result and should not be considered indicative of a true relationship.
Conclusion
The review summarizes the current published knowledge on health effects of flood and storm disasters. It has shown an increase in some injuries, poisonings, infectious diseases, and chronic diseases or NCDs after floods and storms. However, there is a clear need for more information on the long-term changes in health, the strength of the relationships between floods or storms and health problems, and on changes in health after floods only. It has highlighted areas for consideration by responders, but stronger results are required before specific recommendations can be made.
Despite the difficulty of conducting research and collecting data in disaster settings, a move must be made towards more rigorous, systematic, and high-quality research. It is necessary to move on from descriptive studies if an improvement is expected in knowledge and outcomes. Decision making regarding type and quantity of medical responses should be done before a disaster occurs and should use the best available data that are based on research that is held to the same standards as research conducted outside disaster settings. Conducting high-quality, systematic research is one step towards improving the knowledge base and consequently, the quality and effectiveness of the medical response to storms and floods.
Acknowledgements
The authors would like to thank the librarians C. Gornitzki and L. Mathiesen at Karolinska Institutet University Library (Solna, Sweden) for their assistance in developing the search strategy, and F. von Schreeb for data extraction.
Supplementary Material
To view supplementary material for this article, please visit https:/doi.org/10.1017/S1049023X17006574