Nonresponse to a population-based ... - Wiley Online Library

C 2009) Journal of Traumatic Stress, Vol. 22, No. 4, August 2009, pp. 324–328 (

BRIEF REPORT

Nonresponse to a Population-Based Postdisaster Postal Questionnaire Study Ajmal Hussain, Lars Weisaeth, and Trond Heir Norwegian Centre for Violence and Traumatic Stress Studies, Oslo, Norway We examined nonparticipation in a 2-year postdisaster mail survey of Norwegian tourists evacuated from countries affected by the 2004 tsunami. One hundred seventy-one persons out of a random sample of 330 nonparticipants were telephone interviewed concerning disaster exposure, current posttraumatic stress reactions, and reasons for not participating. Fewer nonparticipants than participants had been in a place directly affected by the tsunami. Nonparticipants reported less perceived threat of death and lower levels of posttraumatic stress reactions. Reasons for not participating were “lack of interest or time” (39.2%), “lack of relevant experiences” (32.2%), and “too personal or emotionally disturbing” (15.2%). Our findings suggest that postdisaster studies may be biased in the direction of more severe disaster exposure and pronounced posttraumatic stress reactions. Compared with public health studies, there is less knowledge about nonresponders in disaster research. Disaster research studies have often used indirect approaches such as sociodemographic information to evaluate sample representativeness. However, similarities in such information alone may mask real differences in other variables (Etter & Perneger, 1997; Hill, Roberts, Ewings, & Gunnell, 1997). Bias occurs if nonresponders differ from responders with respect to the characteristics being studied. In disaster research, posttraumatic stress symptoms are the most commonly studied outcomes (Foa, Stein, & McFarlane, 2006; Leon, 2004). Consequently, information about exposure and posttraumatic stress symptoms is important when investigating nonresponse. As far as we know, no study in the disaster literature has collected information by interviewing nonresponders. Information about exposure and stress response is available on individuals who are lost to follow-up in longitudinal studies, but does not describe those who decline to participate at baseline. Postdisaster surveys can be at a particular risk of nonresponse bias, because avoidance symptoms may cause unwillingness to respond (Weisaeth, 1989). However, in a previous tsunami study higher response was related to areas with extensive destruction and high death toll (Heir & Weisaeth, 2008). In the present study, we compared nonparticipants and participants from a postal survey of Norwegian tourists who experienced the 2004 tsunami. We wanted to explore whether these two groups differed regarding exposure and tsunami-related psycho-

logical distress. We also aimed to examine self-reported reasons for not participating.

METHOD Participants The 2004 tsunami in Southeast Asia was one of the largest tragedies in recent history, with 226,408 deceased (Centre for Research on the Epidemiology of Disasters, de Louvain, Belgium, personal communication, June 25, 2008). In the days following the disaster, all Norwegian tourists were repatriated from disaster-affected countries and registered by the police on arrival at the Norwegian National Airport. To investigate long-term mental health aspects, 2 years after the tsunami a questionnaire was mailed to all registered individuals over the age of 18 at the time of the disaster (N = 2,468). From the nonparticipating population (n = 1212), a random sample of 330 individuals, stratified for age and gender, was contacted for a telephone interview. Of these, 171 individuals accepted to take part in the interview, hereafter called “interviewed nonparticipants” (Figure 1). Mean age of interviewed nonparticipants was 43.1 years (SD = 13.3) compared to 42.9 years (SD = 13.0) among the rest of the nonresponder population (ns). Among the interviewed 56.7% were men, whereas among the rest of the nonresponder population 62.2% were men (ns).

The study was funded by The Research Council of Norway. Correspondence concerning this article should be addressed to: Ajmal Hussain, Norwegian Centre for Violence and Traumatic Stress Studies, Building 48, Kirkeveien 166, N-0407 Oslo, Norway. E-mail: [email protected]; [email protected]. C 2009 International Society for Traumatic Stress Studies. Published online in Wiley InterScience (www.interscience.wiley.com) DOI: 10.1002/jts.20431

324

Nonresponse in a Postal Survey

Evacuated from Southeast Asia N = 2468 Not traceable n = 76

Target population of the mail survey N = 2392

325

IES-R items entered as independent variables. Pearson’s correlation between the IES-5 and IES-R scores of the postal survey was satisfactory (r = .96, p < .001). The interviewees were presented 10 possible reasons for their nonparticipation, and asked about the importance of each of them and to pick the most important reason. The reasons were grouped into three categories: lack of interest or time, lack of relevant experiences, and items being too personal or emotionally disturbing.

Procedure Participants n = 1180 (49.3%)

Nonparticipants n = 1212 (50.7%) Prestudy reservation, n = 10 Randomized selection—stratified for age and gender

Contacted n = 330 Did not want to participate, n = 159

Interviewed n = 171

Figure 1. Flowchart of the tsunami studies.

Measures The interviewees were asked the same questions as the postal survey participants had been regarding disaster exposure: whether they stayed in a place hit by the tsunami, whether a close relative or friend was deceased in the tsunami, and to what degree they perceived the threat of death. Threat of death was the best predictor of health complaints in our study of Norwegian tourists 6 months following the 2004 tsunami (Heir & Weisaeth, 2008). It correlated highly with other items that measured danger, such as whether individuals were caught, touched, or chased by the waves (Spearman r = .67) or suffered physical injuries (r = .56). Symptoms of posttraumatic stress were measured using five items from the Impact of Event Scale-Revised (IES-R; .Weiss & Marmar, 1997), scored 0–4. These five items, trouble staying asleep, intrusive thoughts and associations (two items), avoidance of feelings, and watchfulness, hereafter called IES-5, accounted for the maximum explained variance of the IES-R in the postal survey population. The items were selected through linear regression with the sum score of the IES-R as the dependent variable and all 22

Information was gathered by computer-assisted telephone interviews using a structured questionnaire. An information letter had been sent prior to the call; 10 persons did not want to be contacted. A maximum of three calls were made in attempting to reach each person. The interviews were carried out by professional trained interviewers under the supervision of one of the authors (A.H.). All interviews were carried out in a 4-week timeframe approximately 3 months after the postal survey ended. The study was approved by the Regional Committee for Medical Research Ethics and by Norwegian Social Science Data Services.

RESULTS The age of interviewed nonparticipants (M = 43.1, SD = 13.3) did not differ significantly from the age of the postal survey participants (M = 45.0, SD = 12.9). There were also no significant differences between the groups regarding gender (Table 1). A lower proportion of nonparticipants had stayed in a place that was hit by the tsunami. Similarly, a lower proportion of nonparticipants reported a perceived threat of death; however, a higher proportion of them lost a close relative or family member in the disaster. The interviewed nonparticipants had a lower IES-5 mean score, 0.63 (SD = 0.77), than did participants, 1.07 (SD = 0.97), t(241) = 6.61, p < .001. Among those less exposed to the tsunami, postal survey nonparticipants and participants had similar IES-5 scores (Table 2). Among those more severely exposed, the nonparticipants had significantly lower IES-5 scores than participants. The most frequently reported reason for not participating was lack of interest or time, followed by lack of relevant experiences (Table 3). A minority, mostly women (21.6% vs. 10.3% men), χ 2 (1, N = 171) = 4.17, p < .05, found the study too personal or emotionally disturbing. Among those who reported great or overwhelming threat of death, this statement was more frequently reported as the main reason for nonparticipation (32%).

DISCUSSION Our findings demonstrate that nonparticipants can differ from participants with respect to the characteristics being studied despite similarities in age and gender. Nonparticipation was primarily

Journal of Traumatic Stress DOI 10.1002/jts. Published on behalf of the International Society for Traumatic Stress Studies.

326

Hussain, Weisaeth, and Heir

Table 1. Gender and Degree of Exposure in a Random Sample of Nonparticipants and the Total Population of Participants in a Disaster Study of Norwegian Tourists Who Experienced the 2004 Tsunami Nonparticipants (n = 170)

Gender Men Women Stayed in a place hit by the tsunami No Yes Unanswered Threat of death None Small/moderate Great/overwhelming Unanswered Loss of family members or close friends No Yes Unanswered ∗p

Participants (n = 1180)

n

%

n

%

97 74

56.7 43.3

581 599

49.2 50.8

83 88 0

48.5 51.5

308 701 89

26.1 66.4 7.5

99 36 34 2

58.6 21.3 20.1

339 422 315 104

28.7 35.8 26.7 8.8

149 22 0

87.1 12.9

1095 79 6

92.8 6.7 0.5

χ2

df

3.35

1

28.51∗∗

1

47.55∗∗

2

8.09∗

1

< .01. ∗∗ p < .001.

Table 2. Symptoms of Posttraumatic Stress (Mean of IES-5) in Disaster Exposure Subgroups Nonparticipants Exposure variables Did not stay in a place hit by the tsunami Stayed in a place hit by the tsunami No threat of death Small/moderate threat of death Great/overwhelming threat of death No loss of family members or close friends Loss of family members or close friends

Participants

n

M

SD

n

M

SD

t-test

81 83 97 36 31 144 20

0.51 0.76 0.48 0.61 1.14 0.52 1.43

0.76 0.76 0.73 0.57 0.90 0.70 0.81

301 781 335 419 315 1086 78

0.52 1.31 0.49 1.07 1.76 1.00 2.00

0.73 0.97 0.68 0.86 0.96 0.92 1.10

0.10 6.10∗∗∗ 0.06 4.46∗∗∗ 3.42∗∗ 7.41∗∗∗ 2.17∗

Note. IES-5 = Five items from the Impact of Event Scale-Revised. ∗ p < .05. ∗∗ p < .01. ∗∗∗ p < .001.

related to low exposure and lack of interest, whereas posttraumatic stress symptoms were positively associated with participation, even after adjusting for the level of exposure. This is in agreement with a firework disaster study, which extracted nonresponder data from electronic medical records (Grievink, van der Velden, Yzermans, Roorda, & Stellato, 2006). The most important reason for not participating in the postal survey was lack of interest or time. Similar reasons for nonresponse were also reported in a population study after an environmental disaster (Foster, Campbell, Crum, & Stove, 1995). We did not

find other postdisaster surveys that systematically investigated selfreported reasons for not participating. However, many of the reported reasons for not participating in our study are common and have previously been reported in studies of public health: “I did not have the drive or time to complete the questionnaire” (Ronmark, Lundqvist, Lundback, & Nystrom, 1999), “I usually do not reply to questionnaires” (Janzon, Hanson, Isacsson, Lindell, & Steen, 1986), and “The study was not of any use for me personally” (Bakke et al., 1990; Foster et al., 1995). Very few found our study too personal or were skeptical of the study, which means that,


327

Nonresponse in a Postal Survey

Table 3. Different Reasons Reported by Interviewed Nonparticipants for Not Taking Part in the Postdisaster Postal Survey Most important reason

Contributing reasonsa

Possible reasons for not participating

N

%

N

%

Too personal or emotionally disturbing I found the study too personal. I was skeptical of the study. The study reminded me too much of the disaster. Lack of relevant experiences I was not in the disaster area when the tsunami struck. I was not directly affected by the disaster. My experiences were of little importance to the study. Lack of interest or time The study was not of any use for me personally. I usually do not reply to questionnaires. I did not have the drive or time to complete the questionnaire. The questionnaire was too long. Not sure/other reasons

26

15.2

37

21.6

a Contributing

55

67

23

n

%

1 5 20

0.6 2.9 11.7

13 20 22

7.6 11.7 12.9

11 2 41 13

6.4 1.2 24.0 7.6

32.2

118

39.2

140

n

%

15 15 19

8.8 8.8 11.1

55 63 82

32.2 36.8 48.0

75 80 37 70

45.9 46.8 21.6 40.9

69.0

81.9

13.4

reasons are not mutually exclusive.

although not participating, the majority was not hostile towards the study itself. This finding is consistent with other studies of nonresponders (Nakash, Hutton, Lamb, Gates, & Fisher, 2008; Ronmark et al., 1999). The findings may be limited to brief natural disaster exposures. Long-term exposure to environmental disaster which creates somatic health concerns may produce entirely different responses to outreach of surveys (Guey, Bromet, Gluzman, Zakhozha, & Paniotto, 2008). One may assume that in situations where information is gathered shortly after the disaster, ongoing stress in daily life may play a more significant role as a reason for nonresponse. Many survivors may find participation overwhelming because of very high levels of psychological distress. However, this assumption was not supported in a study conducted shortly after a fireworks disaster (Grievink et al., 2006). Our findings ought to be interpreted with caution due to a moderate response rate. We do not know whether the interviewed nonparticipants were representative for the whole population of nonresponders regarding exposure or symptoms of posttraumatic stress. However, among those who refused to be interviewed when called, refusals were attributed to attitudinal reasons rather than emotional stress. Another limitation could be the use of IES-5, which may not have achieved exactly the same estimates as the original measure. However, the very high correlation validity of the IES-5 compared to the IES-R indicates that the demonstrated differences in symptom severity between groups are valid. The study of nonparticipants was carried out 3 months after the questionnaire study, which may complicate the comparison

of posttraumatic stress symptoms. However, it is unlikely that the great difference in symptom scores between the two groups is due to a time variable. Normally, the postdisaster symptom trajectory demonstrates a steep decline in symptom severity up to 12–15 months after the event (Breslau, 2001). As time passes, the remaining symptoms tend to be more persistent in nature (Briere & Elliott, 2000; Foa et al., 2006; Norris, Perilla, Riad, Kaniasty, & Lavizzo, 1999). To our knowledge, this is the first study where nonresponders to a postdisaster survey were directly contacted and asked about exposure, posttraumatic stress, and reasons for not responding. Overrepresentation of individuals with more disaster experiences or more posttraumatic stress may bias the results in postdisaster studies. This can lead to artificially high posttraumatic stress estimations. Data from other contexts and populations are necessary to have a more thorough understanding of this issue.

REFERENCES Bakke, P., Gulsvik, A., Lilleng, P., Overa, O., Hanoa, R., & Eide, G. E. (1990). Postal survey on airborne occupational exposure and respiratory disorders in Norway: Causes and consequences of non-response. Journal of Epidemiology and Community Health, 44, 316–320. Breslau, N. (2001). Outcomes of posttraumatic stress disorder. Journal of Clinical Psychiatry, 62(Suppl 17), 55–59. Briere, J., & Elliott, D. (2000). Prevalence, characteristics, and long-term sequelae of natural disaster exposure in the general population. Journal of Traumatic Stress, 13, 661–679. Etter, J. E., & Perneger, T. V. (1997). Analysis of non-response bias in a mailed health survey. Journal of Clinical Epidemiology, 50, 1123–1128.


328

Hussain, Weisaeth, and Heir

Foa, E. B., Stein, D. J., & McFarlane, A. C. (2006). Symptomatology and psychopathology of mental health problems after disaster. Journal of Clinical Psychiatry, 67(Suppl 2), 15–25. Foster, K., Campbell, D., Crum, J., & Stove, M. (1995). Non-response in a population study after an environmental disaster. Public Health, 109, 267– 273. Grievink, L., van der Velden, P. G., Yzermans, C. J., Roorda, J., & Stellato, R. K. (2006). The importance of estimating selection bias on prevalence estimates shortly after a disaster. Annals of Epidemiology, 16, 782–788. Guey, L. T., Bromet, E. J., Gluzman, S. F., Zakhozha, V., & Paniotto, V. (2008). Determinants of participation in a longitudinal two-stage study of the health consequences of the Chornobyl nuclear power plant accident. BMC Medical Research Methodology, 8, 27. Retrieved July 8, 2009, from http://www.biomedcentral.com/1471-2288/8/27 Heir, T., & Weisaeth, L. (2008). Acute disaster exposure and mental health complaints of Norwegian tsunami survivors 6 months post disaster. Psychiatry, 71, 266–276. Hill, A., Roberts, J., Ewings, P., & Gunnell, D. (1997). Non-response bias in a lifestyle survey. Journal of Public Health Medicine, 19, 203–207. Janzon, L., Hanson, B. S., Isacsson, S. O., Lindell, S. E., & Steen, B. (1986). Factors influencing participation in health surveys. Results from prospective population

study ‘Men born in 1914’ in Malmo, Sweden. Journal of Epidemiology and Community Health, 40, 174–177. Leon, G. R. (2004). Overview of the psychosocial impact of disasters. Prehospital and Disaster Medicine, 19, 4–9. Nakash, R. A., Hutton, J. L., Lamb, S. E., Gates, S., & Fisher, J. (2008). Response and non-response to postal questionnaire follow-up in a clinical trial—a qualitative study of the patient’s perspective. Journal of Evaluation in Clinical Practice, 14, 226–235. Norris, F. H., Perilla, J. L., Riad, J. K., Kaniasty, K., & Lavizzo, E. A. (1999). Stability and change in stress, resources, and psychological distress following natural disaster: Findings from Hurricane Andrew. Anxiety, Stress, and Coping, 12, 363–396. Ronmark, E., Lundqvist, A., Lundback, B., & Nystrom, L. (1999). Non-responders to a postal questionnaire on respiratory symptoms and diseases. European Journal of Epidemiology, 15, 293–299. Weisaeth, L. (1989). Importance of high response rates in traumatic stress research. Acta Psychiatrica Scandinavica, 355(Suppl), 131–137. Weiss, D. S., & Marmar, C. R. (1997). The Impact of Event Scale-Revised. In J. P. Wilson & T. M. Keane (Eds.), Assessing psychological trauma and PTSD (pp. 399–411). New York: Guilford Press.
