5.2 Comparison of Participants with Non-Participants

5.2  Comparison of participants with non-participants

Participation bias can occur if participants differ from non-participants on characteristics which are associated with the study dependent measures, such as health status. A complete examination of participation bias would require the collection of comprehensive and current health, demographics and deployment information for all non-participating Gulf War veterans and comparison group subjects. Whilst such data are not available for non-participants, we are able to conduct a number of comparisons of participants and non-participants using data which were collected at the time of the baseline study, to assess the extent to which participants are representative of the study groups from which they were drawn.

Table 15 shows the comparison of participants with non-participants in each study group, across a number of demographic, service and health-related variables using data from the baseline study. Participating Gulf War veterans were older and more likely to have been Officers in 1991 compared to veterans who did not participate. Participating Gulf War veterans had very slightly lower median SF-12 mental health scores (representing poorer mental health) than non-participating veterans, with this difference only just achieving statistical significance. There were no statistically significant differences observed between participating, and non-participating, Gulf War veterans in regard to their gender, service branch, whether or not they were still serving in the ADF at baseline, their SF-12 physical health score at the time of the baseline study, and whether or not they met CIDI criteria or not for PTSD, major depression or alcohol dependence in the 12 months prior to the baseline study. Participating comparison group members were older, more likely to have been in the Air Force and more likely to have been Officers in 1991, relative to comparison group members who did not participate. There were no statistically significant differences observed between participating, and non-participating, comparison group members across the remaining demographic and health variables included in the table.

The differences observed between participants and non-participants in the two groups may mean that the study slightly over-estimates or underestimates the true health of each group; i.e. the observed health results for each group may vary slightly from that which would have been observed if full participation has occurred. However, because a similar pattern of non-participation amongst the younger and lower ranked group members occurred in both study groups, it is unlikely that those two factors will notably affect the magnitude or direction of any differences in health observed between the Gulf War veteran and the comparison group.

The observed difference between participating and non-participating Gulf War veterans in median SF-12 mental health is so small that it is also unlikely to have an impact on the magnitude of differences in health observed between the participating groups. The increased participation rate by Air Force members in the comparison group may contribute to the study slightly overestimating the true health of the comparison group; and this highlights the need to statistically adjust for service type when comparing the health outcomes across the two groups.

Table 15 Comparison of participants with non-participants in the follow up study

/ Gulf War veterans / Comparison group /
Measure / Participants
(N=712)* / Non-participants
(N=719)† / Participants
(N=674)* / Non-participants
(N=899)† /
/ n (%) / n (%) / χ2 p-value / n (%) / n (%) / χ2 p-value /
Gender
male / 697 (50.0) / 702 (50.2) / 0.742 / 659 (43.0) / 874 (57.0) / 0.489
female / 15 (46.9) / 17 (53.1) / 15 (37.5) / 25 (62.5)
Age at follow up‡
< 40 / 34 (37.0) / 58 (63.0) / 19 (32.8) / 39 (67.2)
40 - 44 / 171 (39.8) / 259 (60.2) / 118 (28.9) / 290 (71.1)
45 - 49 / 208 (53.6) / 180 (46.4) / <0.001 / 185 (41.9) / 257 (58.1) / <0.001
50 - 54 / 172 (55.3) / 139 (44.7) / 189 (51.5) / 178 (48.5)
>= 55 / 127 (60.5) / 83 (39.5) / 163 (54.7) / 135 (45.3)
Service branch in 1991
Navy / 606 (49.3) / 624 (50.7) / 457 (40.5) / 671 (59.5)
Army / 51 (55.4) / 41 (44.6) / 0.515 / 75 (42.4) / 102 (57.6) / 0.001
Air Force / 55 (50.5) / 54 (49.5) / 142 (53.0) / 126 (47.0)
Rank in 1991
Officer / 157 (57.7) / 115 (42.3) / 205 (51.1) / 196 (48.9)
Other-ranks supervisory / 358 (52.3) / 327 (47.7) / <0.001 / 339 (45.8) / 401 (54.2) / <0.001
Other ranks-non supervisory / 196 (41.5) / 276 (58.5) / 130 (30.1) / 302 (69.9)
Still serving in the ADF at baseline
Yes / 291 (48.0) / 315 (52.0) / 0.260 / 277 (44.0) / 352 (56.0) / 0.436
No / 421 (51.0) / 404 (49.0) / 397 (42.1) / 547 (57.9)
CIDI diagnosis in 12 months preceding baseline§
PTSD
absent / 667 (50.7) / 650 (49.4) / 0.300 / 606 (44.1) / 768 (55.9) / 0.899
present / 31 (44.3) / 39 (55.7) / 10 (45.5) / 12 (54.6)
Major depression
absent / 644 (50.9) / 621 (49.1) / 0.161 / 585 (44.3) / 735 (55.7) / 0.547
present / 54 (44.3) / 68 (55.7) / 31 (40.8) / 45 (59.2)
Alcohol dependence/abuse
absent / 674 (50.8) / 653 (49.2) / 0.102 / 605 (44.4) / 758 (55.6) / 0.206
present / 24 (40.0) / 36 (60.0) / 11 (33.3) / 22 (66.7)
SF-12 at baseline / Median (IQR) / Median (IQR) / Wilcoxon test p-value / Median (IQR) / Median (IQR) / Wilcoxon test p-value
Physical Health Component / 52.4
(45.3-55.9) / 52.2
(44.4 - 55.6) / 0.611 / 53.4
(46.9-55.9) / 53.1
(46.0-55.9) / 0.345
Mental Health Component / 51.4
(41.1 – 56.5) / 50.6
(38.5 - 55.9) / 0.059 / 55.0
(47.6-57.8) / 53.7
(45.8-57.4) / 0.012

* Includes all those who completed the health questionnaire at follow up

† Excludes those who are known to be deceased
‡ Age was calculated as of 01 June 2011

§ Only calculated for those with baseline CIDI results

Australian Gulf War Veterans’ Follow Up Health Study: Technical Report 2015 Page 76