The Importance of Estimating Selection Bias on Prevalence Estimates Shortly After a Disaster
Introduction
In the aftermath of disasters, both short- and long-term health problems have been reported, such as feelings of anxiety or depression, severe sleeping difficulties, medically unexplained physical symptoms, and posttraumatic stress disorder 1, 2, 3, 4, 5. Representative prevalence estimates of these health problems are needed after disasters to adequately organize health care for those who need professional help. Policy makers and health care providers need to know whether, and to what extent, an increase in use of health care facilities can be expected (6).
In the wake of the chaos that follows a disaster, it often is very difficult to obtain a representative sample. In addition, apart from material losses and injuries, there are no objective criteria to define who is a victim and who is not, making it difficult to create a methodologically sound sample. Nearly half the health surveys after disasters do not report a participation rate (4), prompting one to question whether participants in those surveys are representative of all survivors. It is common knowledge that representativeness of health surveys can be affected by selection in demographic and health factors. Shortly after disasters, disaster-related experiences additionally might contribute to selective participation. How these experiences influence participation is unknown. One may speculate that survivors who were highly affected by the disaster or had high levels of postdisaster distress would be more motivated to participate in the health survey than survivors who were less affected. Conversely, highly exposed or distressed survivors could be less likely to respond because they do not want to be reminded of the stressful event.
Determining selective participation with regard to demographic-, health-, and disaster-related factors is important, but understanding whether that selection leads to actual bias in the outcome measures in a survey is essential for policy makers and health care providers. To our knowledge, none of the health studies after disasters presents information on the selection of adult survivors on the basis of health characteristics and disaster experiences, let alone the effect of the selection on outcome measures. Investigators of the few longitudinal studies that had predisaster data and described the impact of the disaster on postdisaster attrition rate did not comment on the representativeness of respondents 7, 8, 9.
To investigate selective participation and its potential bias on prevalence estimates of health problems in residents in a survey 3 weeks after a man-made disaster, we had the unique opportunity to combine these survey data with predisaster and postdisaster data retrieved from residents' general practitioners (GPs). The direction of the selective participation was studied with respect to demographics, disaster experience, use of the GP, and health problems presented to the GP. Furthermore, we used multiple imputation to examine the magnitude of selection bias on prevalence estimates of self-reported health problems.
Section snippets
Methods
On May 13, 2000, a firework storage facility exploded in a residential area of Enschede, The Netherlands. The explosion and subsequent fires killed 23 persons, more than 900 people were injured, and about 500 homes were destroyed. The Dutch government declared this a national disaster and the Ministry of Health, Welfare, and Sports supported the regional authorities and launched a larger health care program especially designed for the survivors (11).
Results
Participation rate in the health survey was 26% (N = 1171). Table 1 lists crude differences in participation rates for demographic variables. Compared with nonparticipants, survey participants were more likely to be women, aged between 25 and 44 or 45 and 64 years of age, live with a partner, be a single parent, and be of immigrant background. Participation rate was not affected by relocation caused by the disaster.
Survey participants consulted their GPs more often and had more than an average
Discussion
This study is unique in that we were able to estimate whether participants in a survey 3 weeks after a man-made disaster were representative of the entire cohort of affected residents on many aspects related to participation.
We observed a lower participation rate for men, young adults, and singles. This selection was observed previously in survivors of disasters 7, 10, 27. In contrast to other studies 9, 26, we observed a greater participation rate for immigrants. A possible reason is that even
References (29)
- et al.
Survey non-response in the Netherlands: Effects on prevalence estimates and associations
Ann Epidemiol
(2003) - et al.
Buffalo Creek survivors in the second decade: Stability of stress symptoms
Am J Orthopsychiatry
(1990) - et al.
60,000 Disaster victims speak: Part I. An empirical review of the empirical literature—1981–2001
Psychiatry
(2002) - et al.
60,000 Disaster victims speak: part II. Summary and implications of the disaster mental health research
Psychiatry
(2002) - et al.
Medically unexplained physical symptoms in the aftermath of disasters: A systematic review
Epidemiol Rev
(2005) - et al.
The epidemiology of post-traumatic stress disorder after disasters
Epidemiol Rev
(2005) - et al.
Post-disaster mental health problems and the utilization of mental health services: A four-year longitudinal comparative study
Adm Policy Ment Health & Ment Health Serv Res
(2006) - et al.
Natural disaster and depression: A prospective investigation of reactions to the 1993 Midwest Floods
Am J Community Psychol
(2000) - et al.
Psychological symptoms in older adults following natural disaster: Nature, timing, duration, and course
J Gerontol
(1989) - et al.
The impact of hurricane Andrew on deviant behavior among a multi-racial/ethnic sample of adolescents in Dade County, Florida: A longitudinal analysis
J Trauma Stress
(1997)
Postdisaster PTSD over four waves of a panel study of Mexico's 1999 flood
J Trauma Stress
Post-disaster health effects: Strategies for investigation and data-collection; experiences from the Firework disaster
J Epidemiol Community Health
Physical and mental health shortly after a disaster: First results from the Enschede firework disaster study
Eur J Public Health
Health problems of victims before and after disaster: A longitudinal study in general practice
Int J Epidemiol
Cited by (49)
Does attrition affect estimates of association: A longitudinal study
2019, Journal of Psychiatric ResearchThe long-term physical and psychological health impacts of flooding: A systematic mapping
2018, Science of the Total EnvironmentUsing mixtures of t densities to make inferences in the presence of missing data with a small number of multiply imputed data sets
2015, Computational Statistics and Data AnalysisSocial organizational stressors and post-disaster mental health disturbances: A longitudinal study
2014, Psychiatry ResearchAcculturation and post-disaster mental health problems among affected and non-affected immigrants: A comparative study
2012, Journal of Affective DisordersCitation Excerpt :The estimated response of all affected residents at the first wave was 33%. Compared to non-responders, responders were more likely to be women and affected immigrants (Grievink et al., 2006). The response from the first wave to the second wave for the affected immigrant group (N = 221) was 51%.
Stress sensitization following a disaster: A prospective study
2012, Psychological Medicine