All patients were referred to our ICU from other German and European hospitals for the treatment of severe ARDS, and were admitted between 1991 and 2000 after transport by a specialized team. We started the study in 2002 and investigated only those patients who had been discharged from the ICU for more than one year because a diagnosis of chronic PTSD requires persistent symptoms. The patients were mailed six questionnaires that measured HRQoL, psychological disorders, perception of social support, and socio-demographic data. Patients were asked to recall the ICU stay and to answer each question promptly rather than after protracted consideration. If we received no reply from the patients, we attempted to contact them by phone three times. If at this stage we were still unable to contact the patient the case was counted as lost to follow-up. Data about age, sex, cause of ARDS, ICU length of stay, duration of mechanical ventilation, and admission scorings were extracted from the ICU database. To assess the severity of illness and ARDS, the Lung Injury Score and the Acute Physiology and Chronic Health Evaluation (APACHE) II scores were calculated and the length of ICU stay and duration of mechanical ventilation were shown 1011. To minimise potential selection bias resulting from a desire of some patients with PTSD to avoid recollection of their ICU stay as a symptom of PTSD, we evaluated characteristics of the non-participants in particular detail. A direct or collateral history of mental disease such as alcohol or drug abuse reported in direct or indirect anamnesis and lack of informed consent were exclusion criteria.

We used a questionnaire to obtain demographical data about professional life, family status, educational level, and the current living situation of our patients.

The Post-Traumatic Stress Syndrome 10-Questions Inventory (PTSS-10) was developed to diagnose PTSD within the framework of the Diagnostic and Statistical Manual of Psychiatric Disorders (DSM-III) and includes symptoms of hyperarousal 12. In the first part of PTSS-10, patients were asked about possible traumatic memories of their ICU stay and symptoms of severe illness such as pain, nightmares, anxiety, and respiratory distress. In part two, a ten-item self-report scale recorded the intensity of the following PTSD symptoms: sleep disturbance, nightmares, depression, hyperalertness, withdrawal, general irritability, frequent mood swings, guilt, avoidance of activities prompting the recall of possible traumatic events, and increased muscle tension. For each item a rating between 1 (never) and 7 (always) was possible, leading to a total score ranging from 10 to 70 points. A total of 70 points indicates severe PTSD symptoms. Using the German version of the two-part PTSS-10 to assess PTSD-related symptoms, Stoll and colleagues demonstrated criterion validity by receiver operating characteristic curve analysis and showed that a score of 35 points or more was a cutoff with a sensitivity of 77% and a specificity of 97.5% for the diagnosis of PTSD 13. The ability of PTSS-10 in comparison to structured clinical interviews (SKID) to indicate patients at risk of developing PTSD has also been evaluated in patients with ARDS 3. To investigate the hypothetical association between PTSD and HRQoL we divided the patients into two subgroups: 'high-scoring patients' with a PTSS-10 score of at least 35, and 'low-scoring patients' with a score of less than 35.

HRQoL was measured with the Medical Outcomes Study 36-Item Short Form (SF-36) 14. This questionnaire includes eight scales, each reflecting a different so-called 'dimension of quality of life'. Four dimensions reflect physical health (physical component summary), namely physical function, physical role function, bodily pain and general health, and the four other dimensions reflect mental health (mental component summary), namely vitality, social function, emotional function and mental health. The total score lies between 0 and 100, with higher values indicating a more favourable quality of life. Normative data are available for a German-speaking population 15. We matched healthy controls in terms of age and gender from the continuously updated norm database. SF-36 has previously been validated for critically ill patients 16.

The Symptom Checklist-90-R contains 90 items and is a brief multidimensional self-report inventory containing the following dimensions: somatization, obsessive-compulsive, interpersonal sensitivity, depression, anxiety, hostility, phobic anxiety, paranoid ideation, and psychoticism 17. These scales are summarized under three global indices: Global Severity Index, Positive Symptom Distress Index and Positive Symptom Total. For each of the 90 items a rating on a five-step Lickert scale between 0 (not at all) and 4 (extremely) is possible; data were presented with the use of T values (mean 50, SD 10). The Symptom Checklist-90-R was used to screen for psychological distress and multiple aspects of psychopathology in our patients with ARDS.

The Questionnaire for Social Support assessed the perception of emotional and instrumental social support and social integration 18, and a German version is available 19. This 22-item questionnaire measures the quality of social relationships and support. For each item a rating between 1 (low) and 5 (high) is possible. Scores for the three dimensions and a total score are calculated as means of summed scores. A rating of 5 points indicates the highest degree of social support.

We used SPSS Software (SPSS for Windows, version 10.0; SPSS Inc., Chicago, IL, USA) for statistical analysis. The alpha level was set to the conventional 5%. Multivariate analysis of variance using the F statistic was used to test group (more than two) comparisons. Multivariate analysis of variance (MANOVA) was performed for HRQoL using only 'group' ('high-scoring patients', 'low-scoring patients', healthy controls) as a subject factor. Analysis of variance was also used to check a possible influence of time on the perceived outcome parameters in this long-term follow-up. The period between discharge of ICU and study was therefore examined as a covariant factor. t tests were applied as a post-hoc analysis to evaluate differences between two groups. Proportions were tested with a χ² test or a likelihood-quotient χ² test.

At the time of this study, 18 patients (29%; 8 male, 10 female) were identified as being at increased risk for PTSD according to PTSS-10. Consequently we divided the entire study population into two subgroups: 'high-scoring patients' at increased risk of developing PTSD, and 'low-scoring patients'. PTSS-10 scores were significantly different between 'high-scoring patients' with increased risk for developing PTSD and 'low-scoring patients' (t = - 3.7; p < 0.0001; Table 1). Demographic data for all participating patients and the two subgroups are presented in Table 1. There were no significant differences between the subgroups in relation to age, gender, period between discharge from ICU and entry into study, duration of mechanical ventilation, cause of ARDS, the severity of ARDS as measured by means of lung injury score, or severity of illness by APACHE II score. Requirements for extracorporeal membrane oxygenation were also comparable between groups. In relation to length of stay (t = - 1.95; p < 0.056) and employment status (χ²(3) = 8.2; p < 0.084) we observed a trend towards a difference between groups. 'High-scoring patients' tended to be disabled more frequently and to stay longer on the ICU (Table 1).

A significant positive correlation between the number of traumatic memories and the severity of PTSD was revealed (Spearman r = 0.522; p < 0.0001; Table 2). In particular, a significant positive relationship between the experience of anxiety in the ICU and an increased risk of developing PTSD was demonstrated (χ²(1) = 7.59; p < 0.01; Table 3). 'High-scoring patients' at an increased risk of developing PTSD showed a tendency to recall experiences of pain more often. The whole patient group recalled nightmares or difficulties in breathing more frequently than anxiety or pain. Only experiences of anxiety differed significantly between the subgroups (Table 2).

HRQoL measured by SF-36 in all patients with ARDS investigated was significantly reduced in all dimensions, physical as well as mental, in comparison with age- and gender-matched healthy controls (Figure 2). Using MANOVA we detected a significant difference between 'high-scoring patients', 'low-scoring patients' and healthy controls, and verified a significant effect between subject factor 'group' in both main dimensions (physical and mental component summary) and in all subdimensions of HRQoL. Post-hoct tests revealed a significant difference between 'high-scoring patients' and 'low-scoring patients' in the mental component summary as well as in all subdimensions of mental health (Table 4). In contrast, 'low-scoring patients' were not different from healthy controls in the mental component summary (Figure 3) and the subdimension mental health (Table 4). With regard to the physical component summary, post-hoc tests revealed a significant difference between healthy controls, and 'high-scoring patients' and 'low-scoring patients' suffering from ARDS as well as in the subdimensions physical function, physical role function, bodily pain and general health (Table 4). In contrast, there was no significant difference in the physical component score (Figure 3), bodily pain and physical role function between 'high-scoring patients' and 'low-scoring patients' (Table 4).

Psychological problems measured by the Symptom Checklist-90-R were significantly more intense for 'high-scoring patients' than for 'low-scoring patients' in all dimensions (t values more than 1 SD over the mean for all scales (mean 50, SD 10); Table 5).

Perceived social support, measured by using the total score from the Questionnaire for Social Support, was significantly higher for 'low-scoring patients' than for the 'high-scoring patients' (t = 2.90; p < 0.01). Using the F-Sozu, we demonstrated a significantly higher subdimension score for perception of emotional support (t = 2.24; p < 0.05) and social integrity for 'low-scoring patients' (t = 3.53; p < 0.01; Table 6). The perceived social support correlated negatively with the value of the PTSD score (Pearson correlation r = - 0.31; p < 0.05; Figure 4).

Testing the period between discharge from ICU and study as a covariable, a MANOVA could not detect any influence on PTSD scores, severity of PTSD, distribution of percentages of patients suffering from recollections, psychological impairments, perceived social support and HRQoL with the exception of one subdimension of HRQoL: physical role function.

Psychiatric diagnosis of PTSD according to DSM-III-R criteria requires a triggering event which must be a catastrophic stressor outside the range of usual human experience. Furthermore, the stressor should be perceived as a traumatic event by nearly everyone. PTSD has a strong negative influence on QoL. This probably reflects the importance of recollection of anxiety in the development of PTSD. The lifetime prevalence for PTSD in western countries is reported with 8% within higher rates in females (10–12%) than males (5–6%) 21. However, in some populations, the prevalence of PTSD is considerably higher, for example in ICU-survivors (28%) 2.

To measure PTSD symptoms as a result of an ICU stay we focused solely on recollections of the ICU stay. In this context, Weinert and colleagues 1 investigated patients with ARDS using interviews; they considered disease-specific questionnaires in a 'focus group' to yield more information about an ICU stay as a result of group interactions than as a result of individual memory. Their patients reported the following to be associated with anxiety: noisy ICUs, hallucinations, paranoia, fear of disconnection from the ventilator, and guilt.

Disconnection from the ventilator is in many cases perceived as a life-threatening situation resulting in severe emotional stress 1. In another prospective clinical trail analysing patients 6 months after discharge from ICU, Granja and colleagues showed that in only 41% of the patients a memory of disconnection from the ventilator was associated with stress 22; 53% of the patients recalled tracheal tube suction, and of these 81% associated the procedure with stress. In our study 'difficulty breathing' was recalled by 68% of all patients, but the frequency of these experiences did not differ between 'high-scoring patients' and 'low-scoring patients'. It seems that it is not the procedure itself but rather the individual's experience of it that determines the development of psychological sequelae of intensive care treatment.

In our study 74% of all patients remembered 'nightmares', but their incidence was comparable between 'high-scoring patients' and 'low-scoring patients'. In the study of Granja and colleagues an unexpectedly low rate of 30% of all patients experienced 'nightmares', but when they did occur they had a tremendous effect on quality of life after discharge 22. In our opinion the subjective perception of nightmares as a fearful experience is the crucial factor in the development of PTSD after treatment on intensive care wards.

The duration of mechanical ventilation was not associated with the severity of PTSD symptoms in our study, suggesting that mechanical ventilation itself does not affect the development of PTSD. Kress and colleagues investigated psychological effects of daily interruption of sedation 23. Patients without daily interruption tended to recall awakening in ICU more often than those whose sedation was interrupted daily. Moreover, study patients with a daily interruption of sedation showed significantly fewer symptoms of PTSD. However, the patients did not differ in terms of HRQoL. A perception of the ICU situation that is close to reality improves the integration of treatment experiences into episodic memory, and it might prevent the formation of a memory of traumatization. Moreover, the study group of Kress and colleagues might have been too small to prove the influence of mechanical ventilation (5.6 days on average) on HRQoL. It is worth repeating that weaning strategies deploying early spontaneous breathing require appropriate strategies to avoid fear, anxiety and the feeling of helplessness.

Jones and colleagues described the influence of delusional memories (nightmares, dreams and hallucinations) on acute symptoms of PTSD 24. In their study, factual memories in particular protected against the development of acute symptoms of PTSD. Even though factual memories were sometimes unpleasant, they may have helped in coping with moments of unavoidable traumatic events. Our patients benefited from social support, notably in terms of social integrity. Even though Kapfhammer and colleagues found no correlation between social support and PTSD rate, they demonstrated a reduced social function in patients with PTSD, which in turn led to diminished social activities and communication 3. The compromise of social function in their patients may have partly induced an avoidance of social relationships, and to some extent it may have led patients to reject the social support offered to them. Correlation between social support and severity of PTSD symptoms in our patients suggested that emotional support and social integration acted as factors in preventing PTSD symptoms. Because membership of the 'high-scoring' group was related to disability pension, social assistance by family caregivers might be associated with a better social outcome in ARDS survivors.

With regard to psychosocial characteristics, it has been demonstrated that objective injury criteria are not correlated with the incidence of PTSD in trauma patients who are evaluated during the first 3 weeks after the accident, whereas pre-trauma variables such as gender and mental health, biographical risk and stressful life events associated with PTSD symptomatology are correlated 25. The association between high social support and fewer PTSD symptoms might reflect a better social and emotional state before the ICU trauma 25. Central factors in the development of active coping strategies and a stable mental health status to prevent traumatization are emotional support, empathy and helpful accepting behaviour of family caregivers during the life-threatening and traumatic ICU stay. Passive coping strategies, which are related to diminished social support, inhibit cognitive function and psychological recovery from a traumatic event. A meta-analysis identified a lack of social support after the trauma as one of the major risk factors for PTSD 26. In addition, family characteristics, for example family dysfunction or instability, seem to be a risk factor for the development of PTSD symptomatology 27. In contrast, high social support might imply good communication and a stable family network and might consequently constitute a protective factor against the development of PTSD.

There are several limitations to this study. The great variance of period after discharge in our study might be responsible for a lower feedback rate. Feedback rate and questionnaire results might have been biased by PTSD patients seeking to avoid memories of their ICU treatment. Structured interviews on discharge from the hospital might be more accurate and might encourage patients to participate in follow-up studies later in life. We chose chronic PTSD as an outcome because the diagnosis of PTSD requires a 6-month interval after the traumatic events. Fortunately, acute PTSD symptoms frequently resolve spontaneously 3. Selection bias as a result of the avoidance of memories of the trauma might be acceptable because several patients with numerous traumatic memories responded. We showed a positive relationship between the number of traumatic memories and the value of PTSS-10, confirming the results of previous studies 2. Numerous studies have demonstrated that both the physical and, in particular, the mental convalescence of patients with ARDS takes a long time after discharge from ICU 2028. In addition, non-participating patients had similar patient characteristics.

The period between discharge from ICU and entry into the study was not important for any of the investigated parameters with the exception of physical role function. This may serve as an indicator for the strength and representativeness of the study group. We wanted to investigate the long-term outcome of survivors in a specialized centre for the treatment of ARDS. The high rate of 'nightmares' in our study compared with the study of Granja and colleagues might be a result of the duration of analgosedation 22. Our critically ill patients had a tenfold longer length of ICU stay with a correspondingly longer duration of analgosedation in than their patients. Interestingly, the length of stay was correlated with the development of PTSD. Unfortunately, we did not score acute delirium. Further studies should measure patients' experience in ICU in greater detail. We assume that assessing traumatic experiences such as delirium or fear during the weaning phase and the use of a specialized score might be of further assistance in novel strategies for the treatment of delirium associated with mechanical ventilatory support and analgosedation.