Table 1 Key methodological features of the 6 analyzed studies

Bonanno et al., 2008

SARS epidemic, Hong Kong 2003

Hospitalized SARS adult (≥ 18 years) survivors tracked by the Hong Kong Hospital Authority (N = 1,331). Total of 1,775 individuals infected by SARS in Hong Kong

A face-to-face longitudinal study, including 3 interviews at 6, 12, and 18 months after SARS-related hospitalization

n = 951 (6 months); n = 977 (12 months); n = 856 (18 months)

Latent class growth curve modeling (test the association of a trajectory with a set of predictors)

Approximated well Hong Kong's population characteristics, except by having a higher proportion of women (59.2%) compared to the 2001 census (51.7%). All analysis controlled by this factor

Lee et al., 2009

Hurricane Katrina, New Orleans (USA) 2005

African American Hurricane Katrina evacuees aged 18 or older living in New Orleans area but residing in Houston, Texas, in emergency shelters (N ≈ 8,000 evacuees)

A face-to-face cross-sectional survey, administered on a random sample of evacuees in emergency shelters located in Houston, Texas (Kaiser Washington Post Harvard Poll #2005 WPH020) within one month after the hurricane

n = 621, but analysis conducted only on 363 respondents with complete questionnaires (list wise deletion used)

Logistic regression and LISREL analysis (path diagram and path analysis)

No analyses to account for differences were reported

Johannesson et al. 2011

Tsunami, South East Asia 2005

All Swedish citizens registered at Swedish airports during the first weeks after the disaster and older than 16 (N = 10,501)

A longitudinal mail survey using exhaustive sampling 14 months and 36 months after disaster

n = 4,910 at 14 months (T1); n = 3,457 at 36 months (T2)

Analysis of resilient trajectories related to exposure levels and bereavement status (descriptive). Odds Ratios for the association of mental health and each risk factor in multivariate logistic regression analysis (adjusted by all covariates)

Likely, as no difference detected between respondents and non-respondents

Hobfoll et al., 2009

Terrorist attacks, Israel 2004-2005

All Jews and Arabs, 18 years of age of older, living in Israel (N = 4,503,785 according to 2004 census - total population, excluding an estimated 34% of the population younger than 18). Sampling frame selection based on telephone land lines

A face-to-face longitudinal survey on a random sample, including 2 interviews (August-September 2004, August-October 2005) coincident with the latter period of the Second Intifada

n = 1,613 (August-September 2004); n = 709 (August-October 2005)

Analysis of resilient trajectories associated with a set of risk factors using multivariate logistic regressions (adjusting by all covariates with p < 0.01 in bivariate analyses)

Likely, the sample represented the distribution in the Israeli population on gender, age, place of residence and voting behavior

Bonanno et al., 2007

9/11 terrorist attack, New York 2001

Adult (18 and older) citizens in New York City and contiguous geographic areas in New York State, New Jersey, and Lower Fairfield County in Connecticut (N ≈ 6,080,000, according to census 2000, and excluding 24% of the population younger than 18 years)

Random digit-dial household cross-sectional survey with questionnaires administered face-to-face

n = 2,752 approximately 6 months after September 11, 2001

Multivariate logistic regression. Final model selection taking a hierarchical approach (adjusted by all covariates)

Likely, the sample represented the distribution in this population on gender, age, and race

Hobfoll et al., 2012

Chronic exposure to political violence and social upheaval, Palestinian Authority 2007-2008

All citizens of the Palestinian Authority and East Jerusalem older than 18 years (2010 total population is estimated at N ≈ 4,400,000 by United Nations). Around half should be <18 years old giving a rough final figure of 2,200,000

A longitudinal survey including three waves of interview (September-October 2007, April-May 2008, October-November 2008). A stratified three-stage cluster random sampling strategy was used to select the participants. The questionnaire were administered face-to-face

n = 1,196 (initial sample) and n = 769 (analysed)

Multivariate simultaneous equation models (SEM). This model estimates the complex relationship among variables. This analysis also control for other modeled variables.

Unknown, as the authors did not have data to analyze distribution of non-response and similarly they did not have a detailed census to compare with