Study design

A cross-sectional study design was used in which a random sample of adults from a family practice population completed a self-report questionnaire (Reference Mol, Dinant and Vilters-van MontfortMol et al, 2002). The family practice population in The Netherlands is representative of the general population, since practically every Dutch inhabitant is registered with a family practice.

Study population

The study recruited a random sample of the 67 500 patients registered with 12 practices and their 31 general practitioners. The practices participate in the Registration Network of Family Practices at the University of Maastricht in the province of Limburg. The population in this network resembles the population of The Netherlands as a whole in terms of age, gender, education, medical insurance and type of household (Reference Metsemakers, Höppener and KnottnerusMetsemakers et al, 1992).

The questionnaire was sent to 2997 patients randomly selected from the Registration Network, aged 20 years and over, with a covering letter inviting people with and without adverse experiences to participate. The questionnaires were completed between February and April 1997. Half of the questionnaires were returned (43% of respondents were men, mean age 50 years). The majority (85%) of the respondents lived with family or partner, 14% lived alone and 1% in other circumstances. A third (33%) had private health insurance and 67% had national health insurance; 44% had only primary education, 43% had completed secondary education and 13% had completed higher education. Except for insurance type, the demographic variables differed significantly (P<0.05) between respondents and non-respondents. The differences considered relevant were the following:

1. (a) gender (43% of the respondents were male v. 52% of the non-respondents);

2. (b) education (56% of respondents had secondary or higher education v. 44% of non-respondents);

3. (c) age (respondents were 2 years older on average, with an overrepresentation of those aged 60-70 years and under-representation of those aged 20-30 years).

Of 1498 respondents, 832 were eligible for the purpose of our study, the comparison of PTSD scores after traumatic v. life events. Reasons for exclusion were not having experienced any event, not having specified one's worst event or having chosen more than one worst event. Those whose worst event had happened in 1997 were also excluded, as some of these individuals could have been suffering from an acute stress disorder (symptoms of acute stress within 1 month of an event).

Measures

A questionnaire was developed covering demographic data and several health status correlates (use of medical care, drugs, alcohol and sedatives). This was followed by a checklist asking about personal experience - ‘ever in one's life’ - of accidents, burglaries, robberies, sexual or physical abuse (in childhood or adulthood), disaster or war, or chronic serious illness or sudden death of a loved one. The next questions were: ‘Are there other stressful events that you have not noted in this questionnaire up to now? If yes, please describe them below’ and ‘What was the worst event you experienced in your life? In which year did you experience this event?’ Next, the respondents were asked to fill out Part 3 of the Post-traumatic Stress Symptom Scale - Self-Report version (PSS-SR; Reference Foa, Riggs and DancuFoa et al, 1993), keeping this worst event in mind.

Part 3 of the PSS-SR consists of the 17 criteria on the three sub-scales for PTSD as listed in DSM-III-R or DSM-IV. There are five items on re-experiencing, seven on avoidance/numbing and five on arousal. The respondents were asked how often they had experienced each symptom in the past month (never, 0; a few times a month, 1; a few times a week, 2; a few times a day or continuously, 3). The first eight questions (five items on re-experiencing and three items on avoidance/numbing) explicitly ask about symptoms related to this worst event, e.g. ‘How often in the past month did you have bad dreams or nightmares about the event?’ The maximum score on Part 3 of the PSS-SR is 51. As the distribution of PTSD scores on the PSS-SR was skewed to the right, we performed a transformation: log₁₀(total PTSD score +1), referred to as ‘log PTSD score’. The transformed score was our main outcome variable; when relevant it was converted back to the original score (geometric mean).

Criteria for missing values on the PSS-SR were the following: one missing value was allowed on each of the five-item sub-scales, and two missing values on the seven-item sub-scale (B. E. Foa, personal communication, 1997). The score filled in for the missing value was the average of the respondent's values on that sub-scale.

The PSS-SR was originally validated in two American samples - sexually abused women and a group of people who had experienced various forms of trauma (Reference Foa, Riggs and DancuFoa et al, 1993). The reliability of the Dutch version of the checklist, studied in patients referred to ambulatory care for symptoms of post-traumatic stress (n=63), showed Cronbach's α=0.88 for the total score (further information available from the authors upon request). In a sample of 113 Dutch women who had experienced a miscarriage, α=0.8 (Reference Engelhard, van den Hout and ArntzEngelhard et al, 2001). In our sample α=0.92 for the total score; for the three sub-scales of re-experiencing, avoidance and arousal α values were 0.83, 0.83 and 0.78, respectively.

Analysis

Based on the examples given in DSM-IV, the following were classified as traumatic events: accidents, robbery, sudden death of a loved one, murder or suicide of a loved one, physical or sexual abuse as an adult or child, disaster, war, learning about trauma experienced by a loved one and witnessing violence. Burglary without confrontation with the burglar, relational problems, problems with study or work, chronic illness or non-sudden death of a loved one and serious illness (self) were classified as life events. When it was not clear whether a death was sudden or not, it was classified as non-sudden.

Fig. 1 Group allocation of respondents (PTSD, post-traumatic stress disorder).

The demographic characteristics and history of stressful events of the respondents in the traumatic events group and the life events group were compared, using t-tests and chi-squared tests for statistical significance. Log PTSD scores were calculated per type of event. Next, the mean log PTSD scores (total and three sub-scales) for the traumatic events group and life events group were calculated. This was followed by an analysis of covariance in which the mean log PTSD scores were again calculated, but were adjusted for differences between the two groups in terms of demographic factors and history of stressful events. All variables shown in Table 1 were included in the latter analysis.

Table 1 Demographic characteristics and history of stressful events: traumatic v. life events groups (n=832)

	Worst event was traumatic event (n=299)	Worst event was life event (n=533)
Demographic factors (%)
Age, years1
< 45	48	46
45-64	30	34
> 64	22	20
Gender1
Men	44	40
Country of birth1
Non-Dutch	7	4
Marital status1
Single	20	16
Married	67	68
Divorced/widowed	13	16
Living situation1
Living alone	15	14
With partner (and children)	76	77
One-parent family	3	3
Living with parents/family/home	7	7
Highest education1
Lower	41	35
Middle	28	32
Higher	32	32
Occupation1
Elementary	13	14
Lower	30	28
Secondary	33	34
Higher	19	20
Academic	5	5
History of stressful events
Confronted with other people's traumatic events and life events on a day-to-day basis1
Yes (%)	10	11
Abused as a child (physically or sexually)1
Yes (%)	7	5
Total number of traumatic events (PTSD-type)	0.8	0.8
excluding index event (mean n)2
Years since worst event2**	18	12

To see whether respondents from the traumatic events group would score significantly higher on certain items of the PSS-SR scale, and whether those from the life events group would score higher on other items, a non-parametric test (Mann-Whitney) was chosen (the distributions of the item scores were skewed to the right). Because of the large number of items (17), Bonferroni correction was done.

Main findings

To our knowledge, ours is the first study in an adult population on the propensity of a range of severe difficulties in life that do not fulfil the DSM-IV PTSD stressor criterion A1 to give rise to PTSD symptoms. Our findings show that people from the general population whose worst event is a life event, such as chronic illness, marital discord or unemployment, on average have more PTSD symptoms from this event than people whose worst event is traumatic, such as an accident or disaster. As this is a rather unexpected finding, we have tried to refute it in several ways. After stratifying for the only difference found in demographic and trauma characteristics between the two groups - time since the event - we found that the finding holds true for events that occurred at some time in the past 30 years. This finding perhaps implies that in the very long run the impact of a life event wears out, in terms of PTSD, whereas that of a traumatic event is more persistent.

One could argue that the average difference in total PTSD score between the two groups, at 0.7 on the PSS-SR, is not clinically relevant. Yet, although this difference is indeed small, the finding remains curious as one would expect the life events group to have fewer, not more, symptoms. Also, as was shown by comparing the distributions of the PTSD scores, the scores at the top end of the range after both types of events are comparable. This means that considerable suffering results from both types of events in terms of PTSD.

Another explanation for our finding was sought in the allocation of those whose worst event was a sudden or a non-sudden death of a loved one to either the life events group or the traumatic events group. The results, however, did not counter our first finding.

If no difference could be found between the two groups in terms of total PTSD score there could still be a difference on the item level; perhaps some of the 17 symptoms of post-traumatic stress typically occur after life events and others after traumatic events. Our analyses showed that this was not the case.

Limitations of our study

The reliability of the PSS-SR checklist was found to be good in selected populations and it was also found to be reasonably predictive of PTSD diagnosed by interview (Reference Foa, Riggs and DancuFoa et al, 1993; Reference Engelhard, van den Hout and ArntzEngelhard et al, 2001). However, use in open populations has not been documented. The reliability in our sample, expressed as Cronbach's α, was good. However, as the PSS-SR checklist is a written questionnaire, we were unable to check whether the respondents understood all the items correctly. For example, the question about ‘reliving the event, acting or feeling as if it were happening again’ might be interpreted as actively remembering the event, rather than as an intrusive memory, by those who have never experienced a traumatic event. Another limitation of this study is that we did not confirm the occurrence of the events and symptoms, but relied on self-report.

We excluded all respondents who had chosen two or more worst events because the variable ‘number of years since the worst event’ could not be calculated in this group. Besides, we would have been unable to include those who had a worst event in each category in the item analysis, because it would not have been possible to tell to which type of event (life or traumatic) they were referring when responding to a particular item.

There may be complicated relationships between life events, traumatic experiences and general psychological distress. To name a few, life events and daily hassles might increase general psychological distress, which in turn might trigger PTSD symptoms related to an earlier trauma; or, experiencing a trauma might increase general psychological distress, reducing the capacity to deal with other life stressors, increasing in turn the level of general psychological distress. Both experience of a trauma and general psychological distress, separately or in interaction, might increase the risk of experiencing a trauma and of developing PTSD symptoms after a trauma or a life event. Our study was not designed to unravel these complicated relationships. Nevertheless, our finding that life events give rise to symptoms similar to those caused by traumatic events indicates that further study of the interaction of these factors is needed.

Although a higher response than 50% would have been preferable, our response rate is not unusual, taking into account the taboo around some of the topics in the questionnaire (Reference Koss, Woodruff and KossKoss et al, 1991). Also, the questionnaire was quite lengthy and complicated: besides a section on PTSD, there were questions about general health, and a detailed section about care sought and received for a number of traumatic events experienced by the respondent.

Another point to be considered is whether the non-response was selective. The questionnaire contained a list of eight traumatic events and only two life events (burglary and illness of a loved one), followed by an open question about other events experienced. If a person has experienced one event only, a mild life event, the memory of that mild event might not be evoked by the open question, and respondents might therefore not feel the questionnaire pertains to them personally. The memory of a severe life event might be evoked more easily, triggered by reading the list of mostly traumatic events. Non-response among those who had experienced a mild life event (other than burglary or illness of a loved one) might have ensued, resulting in relatively severe life events being reported. Yet we also know that some people prefer to avoid painful memories, and therefore people with severe life events might refrain from responding, countering the former effect. This would also hold true for the traumatic events: those without the experience of a traumatic event might think the questionnaire was not meant for them; those with such an experience might not want to be reminded of it and therefore not respond. Altogether, we have insufficient reason to think that there was a selection bias. Enhancing the total response in further research, however, is important.

Further research

To have a better idea of how patients interpret the questions of the PTSD checklist, for further studies an interview could be considered, although the costs in a population study would be tremendous. Questions on life events experienced, as well as daily hassles, should be included. Considering recent findings about the overlap of symptoms between patients labelled as having a depressive disorder and those suffering from PTSD (even for intrusive memories), including a depression scale would be important (Reference Reynolds and BrewinReynolds & Brewin, 1999). To enhance the response, paying home visits is a good option, as was shown in a Dutch mental health study (Reference Bijl, van Zessen and RavelliBijl et al, 1997).

Personality traits such as IQ and neuroticism have been shown to affect the development of PTSD after certain traumatic and life events (Reference McNally and ShinMcNally & Shin, 1995; Reference Engelhard, van den Hout and KindtEngelhard et al, 2003). It would be interesting to study whether the same relationship can be found after traumatic and life events in general.

Implications

Our study adds to the evidence that PTSD is perhaps not specific to A1 criterion traumatic events, but that it can also arise after life events. To quote Bremner:

Should we now advise clinicians to ask about symptoms of post-traumatic stress after life events? Or should we reconsider the validity of the diagnosis of PTSD (Reference Kudler, Shalev, Yehuda and McFarlaneKudler, 2000)? It is too early for either of these actions, but our study does stress the importance of looking for more empirical evidence on the consequences of events other than typically traumatic ones, in terms of PTSD.