|Home | About | Journals | Submit | Contact Us | Français|
We examined patterns and correlates of speed of recovery of estimated posttraumatic stress disorder (PTSD) among people who developed PTSD in the wake of Hurricane Katrina.
A probability sample of pre-hurricane residents of areas affected by Hurricane Katrina was administered a telephone survey 7-19 months following the hurricane and again 24-27 months post-hurricane. The baseline survey assessed PTSD using a validated screening scale and assessed a number of hypothesized predictors of PTSD recovery that included socio-demographics, pre-hurricane history of psychopathology, hurricane-related stressors, social support, and social competence. Exposure to post-hurricane stressors and course of estimated PTSD were assessed in a follow-up interview.
An estimated 17.1% of respondents had a history of estimated hurricane-related PTSD at baseline and 29.2% by the follow-up survey. Of the respondents who developed estimated hurricane-related PTSD, 39.0% recovered by the time of the follow-up survey with a mean duration of 16.5 months. Predictors of slow recovery included exposure to a life-threatening situation, hurricane-related housing adversity, and high income. Other socio-demographics, history of psychopathology, social support, social competence, and post-hurricane stressors were unrelated to recovery from estimated PTSD.
The majority of adults who developed estimated PTSD after Hurricane Katrina did not recover within 18-27 months. Delayed onset was common. Findings document the importance of initial trauma exposure severity in predicting course of illness and suggest that pre- and post-trauma factors typically associated with course of estimated PTSD did not influence recovery following Hurricane Katrina.
Hurricane Katrina was one of the most devastating natural disasters in U.S. history, with more than 1000 people killed, 500,000 displaced, and more than $100 billion in damage. People in hurricane-affected areas were exposed many stressors including death of loved ones, risk of death, property loss, difficulty obtaining food and clothing, and exposure to violence after the storm. Prior research has reported strong associations between severity of disaster-related trauma and posttraumatic stress disorder (PTSD).[3,4] It is thus unsurprising that high rates of PTSD symptoms were found among adults exposed to Hurricane Katrina and were related to hurricane-related stressors.[2,5,6] Little is yet known, though, about course of PTSD after Katrina. In the one previous study, mean post-Katrina PTSD duration was approximately 600 days in a sample of Mississippi residents. This is similar to estimates studies of PTSD associated with a broader set of traumas.[7,8] Greater exposure to hurricane-related stressors and post-disaster stress were associated with estimated PTSD persistence, but social support and socio-demographic factors were not. However, that study excluded individuals living in New Orleans, where hurricane-related trauma exposure was greatest, and examined PTSD course only among individuals who developed symptoms within one month of the hurricane. The study was also retrospective. Prospective studies of PTSD find marked fluctuations in PTSD symptoms following disasters,[9,10] which might be obscured in a retrospective study.
Previous studies of PTSD course in other populations suggest that course varies substantially across individuals,[11,12] with 30-40% of cases recovering in the year following onset and another 30-40% exhibiting a chronic course.[7,8,13] Predictors of PTSD persistence include female gender, history of psychiatric disorder, and exposure to additional traumatic events.[13,14] Degree of combat exposure, racial minority status, and low social support have been found to predict chronicity of combat-related PTSD.[15,16]
Although course of PTSD following disasters has been found to be similar to that associated with other traumas, persistence varies across disasters. Certain features of Hurricane Katrina may have contributed to increased chronicity of PTSD: elevated stress exposure persisted after the storm due to forced relocation, difficulty obtaining housing and employment, and lack of access to basic necessities as well as community disruption and dissolution of support networks. Identification of predictors of chronic PTSD is important in targeting post-disaster treatment planning.
The current report uses longitudinal data to study patterns and predictors of PTSD persistence in the Hurricane Katrina Community Advisory Group (CAG), a representative sample of pre-hurricane residents of hurricane-affected areas who were interviewed on multiple occasions. We examined socio-demographics, pre-hurricane history of psychopathology, personality characteristics, hurricane-related stressors, and post-hurricane stressors and social resources as predictors of recovery.
We recruited English-speaking adults (≥18 years of age) using a multi-frame sample design that included random-digit-dial calls to households in the FEMA-defined disaster area and a selection of families that applied for assistance from the American Red Cross or FEMA. Overlap across frames was corrected with a weighting adjustment. The baseline CAG included three sub-samples. The first was surveyed January-March, 2006, 5-7 months post-hurricane (n=1,043, 41.9% cooperation rate). The second was interviewed April-June, 2006, 7-10 months post-hurricane (n=723, 33.1% cooperation rate). The third was interviewed December, 2006-April, 2007, 15-19 months post-hurricane, (n=1,322, 32.3% cooperation rate). The low cooperation rates resulted from difficulties tracking pre-hurricane residents of the disaster area, many of whom changed residences, and our requirement that respondents make a long-term commitment to complete multiple interviews. The current report focuses only on the second and third sub-samples, as complete information about onset and offset of our screening measure of PTSD was not available for the first sub-sample. All respondents provided informed consent. Recruit and consent procedures were approved by the Human Subjects Committee of Harvard Medical School.
A non-response survey was carried out to compare CAG participants with non-participants. The two groups were similar on socio-demographics, but non-participants had somewhat higher levels of hurricane-related stress exposure and psychological distress than participants. The data were weighted to adjust for these biases. Weights also were used to account for variation in within-household probabilities of selection and residual discrepancies between the CAG and the 2000 Census population on a range of socio-demographic and pre-hurricane housing variables. These sampling and weighting procedures are described in detail elsewhere.[2,19]
A follow-up interview was carried out with a sub-sample of respondents from the second and third sub-samples August-November, 2007 (24-27 months post-hurricane). All respondents estimated to have PTSD or moderate-serious psychological distress at baseline and a probability sub-sample of other respondents were selected for follow-up, resulting in 1,195 baseline respondents traced (including 16 deceased) and 901 interviewed, for a conditional response rate of 76.4% and an overall cooperation rate of 24.9%. This sample was weighted to adjust for under-sampling of baseline respondents without evidence of PTSD or psychological distress and for small differences between baseline and follow-up survey respondents on socio-demographic characteristics.
PTSD was assessed with the validated 12-item Trauma Screening Questionnaire (TSQ).[20,21] TSQ responses are described below as measures of “estimated” PTSD because only a subset of the DSM-IV symptoms of PTSD was assessed. Dimensional response options (never, less than once a week, about once a week, two to four days a week, and almost every day) were used to assess 30-day symptoms, resulting in a 0-48 total score. A clinical reappraisal study of 30 baseline respondents (20 probable cases and 10 randomly-selected others) to calibrate TSQ responses to DSM-IV diagnoses of PTSD used blinded clinical interviewers to assess current PTSD with the Structured Clinical Interview for DSM-IV. A TSQ cut-point of 20 was found in the weighted (for under-sampling of asymptomatic respondents) clinical reappraisal study to generate a case threshold that most closely approximated the SCID diagnostic threshold. Sensitivity (0.89), specificity (0.93), and area under the receiver operating characteristic curve (0.91) for this screen were all excellent in reproducing DSM-IV/SCID PTSD diagnoses. Respondents provided information on frequency of PTSD symptoms during the month their symptoms were worst, which we used to estimate PTSD prevalence since the hurricane. Respondents who screened positive indicated when they began experiencing symptoms at least once a week, which we defined as onset of estimated PTSD. The most recent month with symptoms was used to define PTSD recency.
Socio-demographics assessed included age, sex, race/ethnicity, education, pre- and post-hurricane family income, marital status, and health insurance. Age was coded 18-39, 40-59, and 65 and older. Race/ethnicity was coded Non-Hispanic White versus Other. Education was coded high school or less (0-12 years) versus at least some college (13+ years). Family income was coded low/low-average (less than or equal to the population median on the ratio of pre-tax income to number of family members) and high-high/average (greater than 1.0 on this ratio). Marital status was coded married, previously married, and never married. Health insurance was coded yes-no. We distinguished pre-hurricane residence: New Orleans Metropolitan Area (which experienced a flood) versus the remaining hurricane area. Analyses investigating the implications of using continuous rather than discretized measures of socio-demographics found no evidence of difference in significance. Results are consequently reported for the discretized measures, as these allow inspection of nonlinear associations.
Baseline respondents completed screening scales assessing lifetime history of mood, anxiety, and substance disorders, intermittent explosive disorder, and suicidality. These scales were adapted from the Family History Research Diagnostic Criteria interview[23,24] while suicidality items were from the Self-Injurious Thoughts and Behaviors Interview. A number of specifications of these measures were investigated in predicting estimated persistence of PTSD. The most parsimonious specification was a truncated count of number of pre-hurricane lifetime disorders (0, 1, and 2+).
Exposure to hurricane-related stressors was assessed at baseline. Ten categories generated from preliminary qualitative interviews were sufficiently common to be studied: life-threatening experiences (e.g., narrow escape from flood waters requiring emergency rescue), death of a loved one, victimization after the storm (e.g., robbery-assault), victimization of a loved one, physical illness/injury caused or exacerbated by the storm, extreme physical adversity (e.g., difficulty obtaining food-clothing), extreme psychological adversity (e.g., living in circumstances where the respondent had to use the toilet or change clothes without privacy), major property loss, income loss, and housing adversity (e.g., multiple moves).
Social network structure, perceived social support, and social competence were briefly assessed at baseline. Network structure was assessed with a question about number of friends who lived in the county/parish. Perceived social support was assessed with questions about number of people in the county/parish with whom the respondent could discuss private feelings without feeling embarrassed. These variables were coded dichotomously (none versus 1+) based on preliminary analysis showing these were optimal specifications. Social competence was assessed with a 12-item scale that asked respondents to rate themselves on a variety of abilities reflecting competence (e.g., staying calm in a crisis, getting along with people, acting responsibly). The scale had high internal consistency (Cronbach's alpha = .87) and was divided into tertiles for analysis, again based on preliminary analysis showing this specification to be optimal. A 0-3 summary measure of social resources summed these three variables (one point given for the top two tertiles of social competence).
Exposure to post-hurricane stressors was assessed at follow-up for events occurring in the 12 months prior to interview, including serious illness/injury of respondent or loved one, death of loved one, marital separation/divorce, break-up of close relationship, serious interpersonal problems, job loss, major financial crisis, legal problems, and property loss. Respondents also rated current hurricane-related difficulties in finances, employment, housing, transportation, interpersonal problems, physical and mental health, and neighborhood crime on a 5-point scale ranging from 1 (none) to 5 (extreme difficulty). We also assessed post-hurricane relocation, coded as living in either the same pre-hurricane area or in a different area.
Prevalence of PTSD at baseline, follow-up, and at any point since the hurricane was estimated along with the proportion of respondents with PTSD that had remitted by the follow-up. Speed of PTSD recovery was estimated by calculating a survival curve based on duration reports. Multivariate associations of predictors with PTSD recovery were estimated using discrete-time survival analysis with person-months the unit of analysis. As noted above, we examined a variety of strategies for coding predictor variables and estimating their associations with PTSD recovery. Coefficients and standard errors were exponentiated to create odds ratios (ORs) and 95% confidence intervals (CIs). Because the data were weighted, the Taylor series linearization method was used to calculate design-based significance tests. Statistical significance was evaluated using two-sided .05-level tests.
An estimated 29.2% of respondents (n=324) had hurricane-related PTSD at some time prior to the follow-up interview. (Table 1) Median time-to-onset of symptoms was one week (inter-quartile range: 1-30 days). Nearly half (41.2%) of respondents estimated to have PTSD had delayed onsets that began only after the baseline interview (and thus more than six months after the hurricane). Less than half (39.0%) of all estimated cases were resolved by the follow-up, and only about 10% were resolved within the first year following onset. Median time-to-recovery among those who recovered was 16.5 months, but time-to-recovery was greater than 27 months among all estimated cases. (Figure 1) The estimated point prevalence of PTSD was 17.1% at baseline and 17.8% at follow-up. Prevalence estimates were quite similar across the two sub-samples of respondents.
Associations of socio-demographics with estimates of PTSD recovery were examined in a multivariate survival model. Family income was inversely related to recovery; respondents with low/low-average income had the highest odds of recovery (OR=2.8,χ21=11.7, p<.001 compared to the high-average/high income). Other socio-demographics and prior history of psychopathology (χ22=0.3, p=.84) were unrelated to recovery. Associations did not differ for immediate versus delayed-onset cases.
We next examined associations of hurricane-related stressors with estimated PTSD recovery in multivariate survival models that controlled for socio-demographics and prior psychopathology. Two stressors were significant predictors of recovery: experiencing a life-threatening situation and housing adversity. (Table 2) None of the 14 respondents with estimated PTSD who experienced a life-threatening situation recovered, while the odds of recovery among those who experienced housing adversity was 0.4 (χ21 = 6.4, p=.012). None of the other hurricane-related stressors was related to recovery either in univariate or multivariate models. We examined univariate models to make sure the negative results were not due to multicollearity among these stressors. We also created a continuous variable representing number of hurricane-related stressors experienced exclusive of life-threatening situations and housing adversity. This composite measure was not significantly related to recovery (OR=0.9, χ21=0.6, p=.443). Associations between hurricane-related stressors and recovery did not differ for immediate versus delayed-onset cases.
Post-hurricane stressors were assessed only at follow-up and no information about timing of those stressors was collected, precluding firm conclusions regarding associations of post-hurricane stressors with estimates of PTSD recovery. We conducted an exploratory analysis, though, of these associations. Eighteen of the 21 stressors (85.7%) were negatively associated with estimated recovery in models that added one stressor at a time and controlled for socio-demographics, prior psychopathology, housing adversity, and exposure to a life-threatening situation. (Detailed results not reported but available on request.) Difficulties with physical health was the only stressor significantly associated with recovery (OR=0.5,χ21=5.0, p=.025).
Measures of social resources were added one at a time to the same base model used to examine post-hurricane stressors. Neither social network structure (OR=1.9, χ21=1.9, p=.165), perceived social support (OR=1.5, χ21=1.1, p=.299), social competence (χ22=1.4, p=.257), nor a summary measure of social resources was associated with estimated recovery (χ21=2.0, p=.155). None of the social resources interacted with hurricane-related stress to predict estimated recovery. The associations of social resources with estimated recovery did not differ for immediate versus delayed-onset cases.
Nearly 30% of adults exposed to Hurricane Katrina were estimated to have had PTSD at some point after the hurricane. As previously reported, this estimate is high relative to estimates of PTSD prevalence following other natural disasters in the U.S. The estimated prevalence of PTSD at baseline was slightly higher and at follow-up slightly lower than in previous reports of the CAG[2,19] because we excluded the first sub-sample of baseline respondents because of incomplete information on course. Among respondents estimated to have PTSD, approximately 40% had recovered by the time of follow-up. Only about 10% of cases were estimated to have had the disorder for one year or less. This pattern of recovery contrasts with epidemiological evidence on PTSD associated with a wider range of traumas showing that 30-40% of cases recover within the first year.[7,8] Average duration of estimated PTSD among those who recovered was sixteen-and-a-half months while median duration of estimated PTSD among all cases was more than 27 months. These findings suggest a more persistent course of PTSD associated with Hurricane Katrina than previous US disasters,[4,29,30] although caution is needed here because prior studies generally defined remission as the absence of full diagnostic at follow-up,[4,29,30] whereas we defined offset as having no significant PTSD symptoms. Of note, we found high rates of delayed-onset estimated PTSD, which may have resulted from prolonged exposure to traumatic events and stressors following the hurricane, although this possibility remains to be examined in future research.
Of the pre-hurricane characteristics examined here, only family income was associated with course of illness. The finding of a negative association between income and persistence is surprising given that hurricane damage was most pronounced in lower-income areas and resulted in higher rates of forced migration and stress exposure among individuals with fewer economic resources. Because greater vulnerability to stressful events among low-SES individuals is well-documented, this finding warrants further investigation in future research. However, such a result is not unprecedented, as some previous research suggests that unexpected traumas might be less able to shatter world views of fairness and safety among low-income compared to middle-income people due to the fact that low-income people are less likely to hold such views in the first place, leading to lower risk of psychopathology after some extreme stressors.[33,34] Also surprising was the lack of association between pre-hurricane psychopathology and PTSD course, which contradicts prior findings regarding predictors of recovery from PTSD associated with other stressors in community samples.[13,14]
Exposure to hurricane-related stressors was the only other factor associated with recovery. No one with estimated PTSD who experienced a life-threatening event recovered, consistent with evidence that exposure to life-threatening situations during natural disasters is associated with long-term elevations in psychiatric symptoms. Trauma severity has also been found to predict course of combat-related PTSD as well as PTSD after Hurricane Katrina in a cross-sectional survey of Mississippi residents. We also found lower odds of recovery among respondents with hurricane-related difficulties in housing. Housing-related stressors were often ongoing following the hurricane and may be markers of chronic strain associated with repeated relocation and poor quality housing.
Other post-hurricane stressors were unrelated to recovery, with the exception of perceived stress related to physical health problems. Previous studies have shown stress exposure following natural disasters to predict prolonged psychological distress,[30,36] PTSD onset,[37,38] and PTSD persistence after Katrina. Our failure to find such associations should be interpreted with caution because we did not have information about the timing of stressors relative to either onsets or offsets of estimated PTSD and therefore were unable to evaluate temporal sequencing. Other studies have been similarly limited.[6,30] Although our finding of few predictors of recovery may have been related to relatively better measurement of exposures that occurred during the hurricane than of post-hurricane stressors, these results are broadly consistent with prior research that finds numerous predictors of PTSD onset but few predictors of course.[16,28]
None of the social resources examined was associated with recovery. Prior research found that social support is associated with PTSD recovery among combat veterans,[15,16,38] and with low distress following natural disasters.[39,40] Personality characteristics reflecting competence also have been found to protect against PTSD and distress following extreme stressors.[38,41] The lack of association between these factors and recovery in our study mirrors a number of previous findings in the CAG that factors typically associated with psychopathology following trauma and disaster did not predict mental health after Hurricane Katrina.[19,42,43] Taken together, these findings suggest that the devastation caused by Hurricane Katrina was so severe and persistent that it overwhelmed the influence of factors that typically promote resilience and recovery.
Study findings should be interpreted in light of several important limitations. First, PTSD was estimated using a screening scale rather than diagnostic interview. Although the TSQ demonstrated excellent psychometric properties in a CAG clinical reappraisal study, screening scales are less precise than clinical interviews, potentially resulting in misclassification. Because misclassification was likely non-differential, it would have resulted in attenuated associations. Second, the CAG response rate was low and the sampling frame excluded individuals who were unreachable by telephone, which likely resulted in under-representation of those with the highest levels of stress exposure and mental illness. This likely resulted in conservative estimates of PTSD prevalence and inflated estimates of recovery. Third, assessment of hurricane-related and post-hurricane stressors was retrospective and subject to recall bias, although available evidence suggests that reports of acute traumatic events are reliable and largely free of recall bias.[44,45] Assessment of post-hurricane stressors might have been most susceptible to such biases. Findings related to post-hurricane stressors should therefore be interpreted with special caution. Fourth, some individuals estimated to have PTSD may have remitted and relapsed between surveys, which would not have been detected in our assessment. Finally, although the TSQ was only validated as a screen for current PTSD symptoms, we also used it to estimate PTSD retrospectively during the week when symptoms were the worst. It is unknown whether the psychometric properties of such a retrospective assessment are equivalent to those associated with screening for current symptoms.
We found a high estimated prevalence of PTSD among adults exposed to Hurricane Katrina and a more persistent course of illness than found in other US disaster-exposed populations. A host of factors associated with PTSD course in other studies were unrelated to recovery following Hurricane Katrina. High family income and exposure to hurricane-related stressors were the only factors associated with reduced odds of recovery. Although it is possible that discrepancies in our measurement of PTSD onset and offset compared to previous studies explain these differences, it might be that the findings are due to the magnitude of devastation caused by the hurricane outweighing the importance of other factors typically associated with PTSD course. Individuals exposed to life-threatening situations and housing-related stressors were at elevated risk for delayed recovery and therefore represent important targets for mental health intervention.
This study is supported by NIH Research Grants R01 MH070884-01A2 and R01 MH081832 from the US Department of Health and Human Services, National Institutes of Health (NIH), the Office of the Assistant Secretary of Planning and Evaluation, the Federal Emergency Management Agency, and the Administration for Children and Families.
Disclosure: In the past 36 months Dr. Kessler has been a consultant for Analysis Group, GlaxoSmithKline Inc., Kaiser Permanente, Merck & Co, Inc., Ortho-McNeil Janssen Scientific Affairs, Pfizer Inc., Sanofi-Aventis Groupe, Shire US Inc., SRA International, Inc., Takeda Global Research & Development, Transcept Pharmaceuticals Inc., Wellness and Prevention, Inc., and Wyeth-Ayerst; has served on advisory boards for Eli Lilly & Company, Mindsite, and Wyeth-Ayerst; and has had research support for his epidemiological studies from Analysis Group Inc., Bristol-Myers Squibb, Eli Lilly & Company, EPI-Q, Ortho-McNeil Janssen Scientific Affairs., Pfizer Inc., Sanofi-Aventis Groupe, and Shire US, Inc. He owns stock in Datastat, Inc. The remaining authors report nothing to disclose.