Chapter 2: Construction and Analysis of the Cohort
2.5 Scheme of Presentation
2.5.1
Literature Review
The heterogeneity of veterans’ studies, the predominance of conflict-specific50
studies, and the large number of recent studies on veterans’ mental health, dictated a targeted approach rather than a single generic literature review. Accordingly, a brief review of the literature relevant to the general background has been given in Chapter 1, with
supplementary material in Appendices 1 to 3, and a series of focussed reviews of the relevant literature is presented as part of the Introduction, Discussion or Commentary for
49
Also known as lex parsimoniae or Occam’s Razor, after the philosopher and theologian William of Ockham (c.1287-1347)
50
Scottish Veterans Health Study Chapter 2 – Contruction of Cohort 73
each condition discussed in Chapters 4-7. Studies were selected for inclusion on the basis of their relevance to the issue under discussion, using appropriate search strategies followed by review of the abstracts and, where indicated, reading of the full papers. Where a consensus between authors was clear, only one or more representative papers were selected; where there was lack of consensus amongst researchers, an adequate number of references was included to reflect the competing views. Recent publications were chosen where available and relevant; review papers and meta-analyses were cited if available.
2.5.2
Results
A systematic approach was taken whereby ‘core’ analyses of cumulative incidence and Cox proportional hazard ratios, comparing veterans and non-veterans, are presented for each diagnosis or outcome as defined at Section 2.2.3.3 and Appendix 6; overall, by sex, by age, by birth cohort51
and by length of service52
. For some rare conditions, small numbers precluded subgroup analysis by birth cohort and/or length of service; where this was the case, these analyses have been omitted. Where differences between veterans and non-veterans were found in the ‘core’ analyses, additional focussed analyses such as period of service were undertaken to clarify the findings where appropriate; these are presented within the relevant subsection, or as an additional subsection. Geographical distribution was examined for conditions where the number of cases was sufficient to yield meaningful results. Age at diagnosis was compared for conditions where it was considered important to determine whether a difference existed between veterans and non-veterans, for example mental health and substance abuse-related conditions and outcomes, and neurological conditions, but was omitted where it would not have added materially to the understanding of veterans’ health. Other than as described above, the full analysis is presented for each diagnosis or outcome deemed to be of a priori interest, whether or not differences were found between veterans and non-veterans. It was considered to be equally important to be able to demonstrate robustly that no difference
51
In 5-year groups, except for some rare conditions where 10-year groups were used. Where numbers were sufficiently large, a combined analysis of length of service and birth cohort (in two categories, pre-1960 and 1960 onwards) was also presented in order to clarify changing patterns over time.
52
By common lengths of military engagement, including classification as Early Service Leavers, who were further subdivided into those who left before completing training, and those who completed training but left before completing the minimum term of engagement. Comparison is made against all veterans.
Scottish Veterans Health Study Chapter 2 – Contruction of Cohort 74
could be shown to exist as it was to identify differences; some conditions were widely believed to be more prevalent in the veterans’ community but analysis of the data gave no cause for concern, whereas the identification of unexpected differences at subgroup level in other conditions has highlighted the need for further research.
Principal results have been presented graphically, and also in tabular form where there were sufficient numbers of cases and where the greater degree of granularity added clarity. Nelson-Aalen plots were used to compare cumulative hazards, thereby
demonstrating the probability of having experienced the outcome under investigation by a given age, for both veterans and non-veterans. This was considered to be a more appropriate form of presentation than the Kaplan-Meier plot, which shows the survivor function and hence the likelihood of not having experienced the outcome under
investigation. All Nelson-Aalen plots showing cumulative hazards by subgroup, and forest plots depicting hazard ratios by birth cohort53
, present univariate results unless otherwise stated or implicit from the context. For rare but potentially important conditions, where statistical power was limited, a graphical presentation has been used to highlight
apparent differences between veterans and non-veterans. Ethical issues in respect of the presentation of small numbers have been discussed at Section 2.2.6; graphical presentation alone was used where small numbers rendered the underlying data potentially disclosive.
A table of, or commentary on, co-morbidities has been included for those diagnoses and outcomes where the literature indicated that co-morbidities may have an important role, or where there was a likelihood of shared aetiological factors.
2.5.3
Context
A study of veterans’ health would be of limited value if it did not seek to place the findings into the wider context of military culture, ethos, operational activity or policy. Accordingly, a broad range of sources has been used to provide this contextual
background, including legislation, military policy documents, information obtained from
53
Error bars on forest plots are terminated with a short horizontal line; those which are unterminated have been truncated. Owing to software limitations, it was not possible to show truncated error bars with the convential arrowhead terminator.
Scottish Veterans Health Study Chapter 2 – Contruction of Cohort 75
the Ministry of Defence through specific Freedom of Information (FOI)54
inquiry, published and unpublished dissertations, military historical sources and soldiers’ and veterans’ own narrative55
. These sources have been used to inform the interpretation of the results and the implications for service delivery and military policy.
54
Freedom of Information Act 2000. Chap.36 55
Websites run by regimental associations and veterans’ groups have provided a rich source of narrative, providing important qualitative information on such issues as alcohol use.