Study Selection

We identified 7742 abstracts and performed full-text review of 859 articles (Figure 1). Of the 859 articles, 25 articles met stated inclusion criteria with responses from a combined total of n = 616 657 people (Table 1).

Figure 1.

PRISMA flowchart of study selection. PRISMA flow chart of study selection for systematic review of the association between hurricane and flooding disasters and psychological distress. Contact with authors was not made due to the length of time since publication.

Table 1.

Studies included in review presented by study type, n = 25

NR, Not Reported; PHQ-9, Patient Health Questionnaire-9; PSQW, Penn State Worry Questionnaire; PCL-C, Post-traumatic Checklist-Civilian; DASS, Depression, Anxiety and Stress Scale; Zung SDS, Zung Self-Rating Depression Scale; SCL-90, Symptom Checklist; BDI, Beck Depression Inventory; CES-D, Center for Epidemiologic Studies Depression Scale; IES-R, Impact of Events Scale-Revised; RCMAS, Revised Children’s Manifest Anxiety Scale; CAPS, Clinician administered PTSD scale; SDQ, Strengths and Difficulties Questionnaire; SF-36, Short Form 36; RCADS, Revised Child Anxiety and Depression Scales, Child and Parent Version; PSS, Perceived Stress Scale; K6, Kessler Psychological Distress Scale; RCMAS, Revised Children’s Manifest Anxiety Scale; RCADS-MD, Revised Child Anxiety and Depression Scales, Major Depression; RCADS-GA, Revised Child Anxiety and Depression Scales, Generalized Anxiety

Study Characteristics

Eight case-control studies,Reference Cherry, De Vito and Calamia ^{26 ‒}Reference Walling, Tucker and Pfefferbaum ³³ 12 cohort studies, and 6 cross-sectional or panel studies contained a valid control group and an assessment of a mental health outcome. The most assessed outcome was depression (n = 13 studies), 8 studies assessed anxiety, and 7 studies measured PTS (Table 1).

More than half of the studies in the sample (n = 17) focused on the 2005 disaster Hurricane Katrina (Global identifier number [GLIDE] #TC-2005-000144-USA), 3 of which also included Hurricane Rita (GLIDE #TC-2005-000163-USA), which made landfall in the month following Hurricane Katrina, exposing many people to 2 storms. Four studies concentrated on other hurricanes or tropical storms (Hurricane Andrew [GLIDE #TC-1992-000002-USA], Hurricane Michael [GLIDE #TC-2018-000433-USA], Hurricane Maria [GLIDE #TC-2018-000433-USA] and Tropical Storm/Hurricane Agnes [GLIDE #TC-1972-000002-USA]). The remaining 4 studies focused on flooding in Baton Rouge, LA (GLIDE #FL-2016-000145-USA), Puerto Rico (1985), North Dakota (1997), and Iowa (FL-1993-000005-USA). Global identifiers were not available for the flooding in Puerto Rico in 1985 and North Dakota in 1997.

Quality Assessment

Seven studies were rated as low risk of bias/high quality, 3 of these were case-control studies (Table 1). Five studies received a medium quality rating, 3 of which were case-control studies. The remaining studies (n = 13) were rated as high risk of bias/low quality or unable to assess. Typical issues of quality assessment were lack of a reported response rate, lack of information about missing values, and questions about the representativeness of the cases in case-control studies. One study reporting attrition in the sample found that only 30% of the original sample was able to be located and assessed for follow up.Reference Costa, Weems and Pina ³⁴ Unfortunately, this study was one of the few that reported pre- and post-PTS symptoms assessments in a cohort. Pre-disaster assessments were performed in a range from less than 1 monthReference Vu and Vanlandingham ³⁵ to 2 years before disaster.Reference Kessler, Galea and Jones ³⁶ Post-disaster assessments were performed in a range from 1 monthReference Ginexi, Weihs and Simmens ^{28 ,} Reference Brown, Cherry and Marks ³⁷ to 5 years after disaster,Reference Melick and Logue ^{29 ,} Reference Abramson, Park and Stehling-Ariza ³⁸ and 2Reference Cherry, De Vito and Calamia ^{26 ,} Reference Zahran, Peek and Snodgrass ³⁹ studies failed to report the time of assessment relative to disaster.

Exposure Assessment

The primary method of exposure assessment was geographic (n = 14 studies), with descriptions ranging from state-level to residence within a 40-mile (64.4 km) radius of the storm path. Among these, 6 studies also assessed self-reported disaster exposure. Four studies recruited individuals displaced by flooding or storm damage, with some case-control studies recruiting participants directly from relocation camps. Seven studies utilized self-reported exposure measures, all incorporating standardized instruments for reporting of traumatic events experienced during or after a disaster, such as the Hurricane-Related Traumatic Experiences QuestionnaireReference La Greca, Silverman and Wasserstein ⁴⁰ or the Survey of Exposure to Hurricanes and their Aftermath.Reference Costa, Weems and Pina ³⁴

Depression

Out of 25 studies, 13 (52%) measured depression for a combined population of n = 4086 people. Study sample sizes ranged from 42 to 1735 participants. Studies assessing depression included 6 case-control designs, 5 cohort studies, and 2 cross-sectional mean comparison studies. Eight studies were excluded from the meta-analysis; studies were excluded when either the authors did not report both mean and standard deviation estimates, and/or the authors only reported the percent of people who met a pre-specified cut-point. For example, FerraroReference Ferraro, Morton and Knutson ⁴¹ reported that 8% and 9.5% of the sample were diagnosed with depression pre- and post-disaster, respectively. The most used instrument was the Center for Epidemiologic Studies Depression scale (CES-D), used by 4 studies. Ten studies (77%) reported increased depression scores in people who were exposed to disasters. Twelve studies included information about the length of time between the hurricane or flood and assessment; the minimum time was 1 month, the maximum time was 60 months (mean = 13 months, SD = 15 months). Regarding quality assessment, all studies that reported an increase in depression in people exposed to flooding or hurricanes were assessed as low or medium risk of bias.

Anxiety

Nine studies measured anxiety with a total of n = 1447 individuals and sample size ranging from 52 to 392 people. Six studies included anxiety assessments that met inclusion criteria for the meta-analysis. One study was excluded because the anxiety measure (the Perceived Stress Scale) was reported as prevalence of the sample with a score above a cutoff score of 7, with 20% of the cohort meeting this threshold 1 year before hurricane exposure; 31% met this criteria between 7-19 months after exposure.Reference Rhodes, Chan and Paxson ⁴² Another excluded study reported higher scores on the Symptom Checklist (SCL-90) Anxiety subscale in controls compared to Tropical Storm Agnes exposed cases (scores 17 vs. 15, respectively), but standard deviations were not reported.Reference Melick and Logue ²⁹ Also excluded was a study in which the authors reported the results of the Depression, Anxiety and Stress Scale (DASS-21) as a single score instead of reporting the anxiety subscale.Reference Fincham and May ⁴³ The authors reported a decrease in the single summary score indicating a decrease in depression, anxiety, and stress in the time period after exposure to Hurricane Michael. Fifty percent of studies assessing anxiety enrolled children older than 10 or adolescents. The mean age of participants ranged from 11.3 yearsReference Costa, Weems and Pina ^{34 ,} Reference Weems, Pina and Costa ⁴⁴ to 49.6 years,Reference Cherry, De Vito and Calamia ²⁶ and 1 study enrolled fourth through sixth grade children but did not include a mean age of the sample.Reference La Greca, Silverman and Wasserstein ⁴⁰ Four studies of 9 assessing anxiety reported increased anxiety in those exposed to disaster, 1 study reported no difference between the 2 groups, and 3 studies found lower anxiety scores in exposed relative to controls or pre-disaster comparisons. However, out of these 4 studies which reported an increase in symptoms, only 1 was assessed as having low risk of bias.Reference Vigil ³²

Post-Traumatic Stress

Eight studies measured PTS, 6 of which were included in the meta-analysis. Of the 2 studies not included in the meta-analysis, 1 reported an increase in PTS in unexposed compared with exposed persons, but did not report standard deviations along with mean scores,Reference Walling, Tucker and Pfefferbaum ³³ and the other study did not report mean scores from the PTSD Checklist for Civilians (PCL-C), but reported no difference in scores in people who were exposed compared to the unexposed.Reference McLeish and Del Ben ⁴⁵ Two cohort studies (both assessing people exposed to Hurricane Katrina less than 1 year after the hurricane) performed pre-disaster and post-disaster assessment of PTS, but the studies had divergent findings. One study found an increase in PTS symptoms in youths and their parentsReference Costa, Weems and Pina ³⁴ while another study found no increase in PTS in adolescents.Reference Weems, Pina and Costa ⁴⁴ Four case-control studies found an increase in PTS in the exposed compared to the unexposed.Reference Cherry, De Vito and Calamia ^{26 ,} Reference Tucker, Pfefferbaum and Khan ^{31 ‒}Reference Walling, Tucker and Pfefferbaum ³³ Only studies that conducted diagnostic interviewsReference Canino, Rubio-Stipec and Bravo ⁴⁶ specifically stated that the outcome of interest was related to hurricane or flood experience. Standardized questionnaires like the PCL-C do not specify an index event but instead ask about PTS symptoms in the previous 30 days.Reference Ruggiero, Ben and Scotti ⁴⁷

Global Mental Health Status and Other Outcomes

Several studies included global measures of mental health status, such as results from the Kessler Screening Scale for Psychological Distress (K6), the mental composite score from the SF-36, number of poor mental health days, and suicide behavior. None of these studies were included in the meta-analysis due to small numbers and inconsistent reporting. Two cohort studies found increased prevalence of psychological distress defined by K6 scores above pre-defined cut-points in the post-exposure groups.Reference Rhodes, Chan and Paxson ^{42 ,} Reference Arkin, Lowe and Poon ⁴⁸ One study used K6 scores pre- and post-Hurricane Katrina in 1849 people and found increased prevalence of severe mental illness post-hurricane (6.1% to 11.3%) and increased mild-moderate mental illness (15.7% to 31.2%).Reference Kessler, Galea and Jones ³⁶ This study was the only one to include suicide-related outcomes. The authors found that in people with serious mental illness, the prevalence of suicidal ideation decreased from 8.4% pre-Katrina to 0.7% post-Katrina, and suicide plans dropped from 3.6% to 0.4%, while no difference was found in the percentage of people reporting a suicide attempt, which the authors hypothesize could reflect protective factors activated by the hurricane.Reference Kessler, Galea and Jones ³⁶ Two studies reported an increase in the psychological SF-36 Mental Health Composite Score (MCS), indicating an increase in negative mental health symptoms, but increases were small and not statistically significant.Reference Vu and Vanlandingham ^{35 ,} Reference Brown, Cherry and Marks ³⁷ One case-control study observed a decrease in the MCS, indicating improved mental health in people unexposed to a hurricane.Reference Stanko ³⁰ Three studies used cross-sectional panel data to compare the number of poor mental health days or mental distress days, with 2 studies finding small increases in poor mental health days in geographic regions impacted by Hurricane Katrina.Reference Zahran, Peek and Snodgrass ^{39 ,} Reference An, Qiu and Xiang ⁴⁹ A study using statewide data from Louisiana found a decrease in the population reporting no mental distress days from 76% to 69% comparing the year prior to Hurricanes Katrina and Rita to the year after the storms.Reference Mukherjee, Canterberry and Yore ⁵⁰

Meta-Analysis of Depression, Anxiety, and PTS

Eleven of the 25 studies were included in the meta-analysis. Reasons for excluding studies were: (1) not including standard deviations along with means for groups;Reference Melick and Logue ^{29 ,} Reference Ferraro, Morton and Knutson ^{41 ,} Reference An, Qiu and Xiang ^{49 ,} Reference Hutchins and Norris ⁵¹ (2) reporting percentages representing prevalence;Reference Walling, Tucker and Pfefferbaum ^{33 ,} Reference Vu and Vanlandingham ^{35 ,} Reference Kessler, Galea and Jones ^{36 ,} Reference Abramson, Park and Stehling-Ariza ^{38 ,} Reference Rhodes, Chan and Paxson ^{42 ,} Reference Mukherjee, Canterberry and Yore ⁵⁰ and (3) reporting only a global mental health status score.Reference Brown, Cherry and Marks ^{37 ,} Reference Zahran, Peek and Snodgrass ³⁹ Overall, results of the meta-analysis suggest there was evidence of increased depression and PTS symptoms in people with disaster exposure, but no evidence for a change in anxiety symptoms. Results of the meta-analysis found depression scores were increased in those who were exposed to a disaster compared to the unexposed (k = 9; g = 0.28; 95% CI 0.04, 0.53). There was high heterogeneity in effect sizes in studies: I² = 88.5%, τ² = 0.17, Q_resid (8) = 33.9, P < 0.01 (Figure 2). The meta-analysis resulted in no meaningful change in anxiety scores associated with disaster exposure (k = 6; g = -0.05; 95% CI −0.30, 0.19). There was moderate heterogeneity in effect sizes across studies: I² = 69%, τ² = 0.06, Q_resid (5) = 15.73, P < 0.01 (Figure 3).

Figure 2.

Studies assessing depression symptoms (n = 9), grouped by study design.

Figure 3.

Studies assessing anxiety symptoms (n = 6), grouped by study design.

Regarding disaster exposure and PTS (Figure 4), exposure was associated with increased PTS (k = 6; g = 0.44; 95% CI 0.13, 0.76). There was high heterogeneity in effect sizes across studies of PTS symptoms: I² = 76%, τ² = 0.11, Q_resid (5) = 17.57, P = 0.0035.

Figure 4.

Studies assessing PTSD symptoms (n = 6), grouped by study design.

Sensitivity analyses

After excluding 1 study that assessed participants more than 12 months after exposure, the association with depression changed from g = 0.82 (95% CI 0.04, 0.53) to g = 0.15 (95% CI −0.085, 0.375). Repeating the meta-analysis for anxiety after excluding 1 study, altered the result from g = −0.05 (95% CI −0.30, 0.19) to g = −0.12; (95% CI −0.37, 0.12). Excluding 2 studies from the PTS meta-analysis, left 4 studies, changing the association with PTS symptoms from g = 0.44 (95% CI 0.13, 0.76) to g = 0.27; 95% CI −0.01, 0.54).