Objectives: Nonprobability sampling, commonly used in disaster research, can lead to incorrect estimates or limit the generalizability of results. We collected data through the Texas Flood Registry (TFR) and used raking and propensity score weighting to provide insight into the effect of Hurricane Harvey (hereinafter, Harvey) on Harris County, Texas. Methods: From April 2018 through October 2020, residents of areas affected by Harvey enrolled in the TFR completed a survey on their storm-related experiences (n = 20 653). Using logistic regression, we assessed the relationship between Harvey-related exposures and distress among Harris County residents (n = 12 279). We used raking to adjust the sample distribution to reflect demographic characteristics of Harris County and propensity scores to address confounding. Results: Of respondents, 56% and 43% reported home damage and income loss due to Harvey, respectively. From April 2018 through April 2020, respondents completed the Impact of Event Scale questionnaire (n = 10 631), with 23% reporting symptoms consistent with severe distress related to Harvey. The raking-adjusted odds ratio of greater Harvey-related distress was 6.21 (95% CI, 5.44-7.09) times higher among residents who had home damage than among those who did not and 2.92 (95% CI, 2.59-3.30) times higher among those who had economic loss than among those who did not. Conclusions: We found consistent associations between adverse storm experiences and Harvey-related distress across unweighted and weighted approaches. We recommend using raking to adjust a nonprobability sample to better reflect population demographic characteristics and obtain general trends of postdisaster exposures and outcomes. We recommend using propensity scores when outcomes may be related to unmeasured confounding.
Callender et al. (Sat,) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: