The results of the study yielded a six-factor solution with the factors "Humiliation" "Physical adverse effects", "Separation", "Negative environment", "Fear", and "Coercion". These factors explained 54.5% of the total variance of intercorrelations.
Cronbach alpha ranged from 0.67 to 0.93, which can be interpreted as a high internal consistency of the single factors. The highest internal consistency reached "Humiliation" (0.93), followed by "Separation" (0.92), "Coercion" (0.83), "Negative environment" (0.78), "Physical adverse effects" (0.72) and least "Fear" (0.67).
Except for the high intercorrelation (0.84) between "Separation" and "Coercion" the subscales show for the most part low to moderate intercorrelations (0.22 - 0.64), indicating an adequate independence of the respective subscales
To determine the convergent validity of the "Coercion Experience Scale" the correlations between this questionnaire and a visual analogue scale assessing the global strain during the same index-intervention was used. The analysis of correlation yielded a highly significant result (r = .79, p < .001). However, there was no significant correlation between the "Coercion Experience Scale" and the Impact of Event-Scale. Probably, this fact can be attributed to adaptation to the traumatic impact of coercive interventions, because the IES-R interview was the only scale assessed one year after the index-intervention and only 3 patients could be diagnosed with PTSD after assessment with the IES-R. Discriminant validity was measured by patient satisfaction and correlated negatively with the "Coercion Experience Scale" (r = -.38, p < .001).
The correlation between the "Coercion Experience Scale" and a screening instrument for PTSD was high (r = .64, p < .001). Together with the result mentioned above this supports the conclusion of convergent validity. Furthermore, the defined cut-off value of the screening on PTSD was used to estimate a critical point of strain induced by the coercive intervention. The regression showed that a global score of more than 70 seemed to indicate a highly restrictive measure. This cut-off point has to be considered as preliminary and is only an estimate of traumatisation. Due to a very low prevalence of PTSD in the follow-up (n = 3) we had to waive analyses of predictive values.
The index-intervention was additionally observed by experienced staff-members. Their assessment of the assumed global strain experienced by the patient during the index-intervention (VAS) was the external point of reference. Concurrent validity was low and only by trend significant (r = .18, p = .09). The reasons for the lower assessment by staff-members may be the difficulties in perceiving the full extent of the very subjective suffering induced by coercive interventions in general. Staff-members seem to differentiate the extend of coercion related to the multitude of coercive interventions carried out by them. They may set a maximum of restrictiveness at a seldom occurring intervention during which they had to forcefully overpower a severely agitated, highly aggressive patient with the help of policemen not being able to prevent the patient from injuries while on the other hand patients may feel already heavily traumatised by the circumstance of being led to the seclusion room and being locked in.
The present study has several limitations. Firstly, the study was conducted in a hospital located in a rural area with a high socioeconomic standard in South Germany, which is not representative compared to more populated areas. In other facilities with a different practice applying coercive interventions strains and restrictive experiences might be somewhat different.
Secondly, it is possible that patients might have overreported the intensity of their experiences on both the restrictions and the VAS in order to emphasise the necessity to reduce coercive interventions.
Thirdly, the sample size with 102 analysed patients is rather small. The proportion between number of patients and items is inappropriate. However, according to the Kaiser-Meyer-Olkin criterion the data were suitable for this PAF. Furthermore, the six factor solution accounted for 54.5% of the total variance of intercorrelations which is a good result in consideration of the sample size.
Fourthly, it may be contradictive to validate this questionnaire with strict statistical methods and to rely on subjective instruments like patient satisfaction and IES-R. For example, patient satisfaction and perceived coercion may be not two theoretically completely unrelated constructs as demanded for discriminant validity. Discriminant validity may therefore be questionable. On the other hand, there is no better instrument for discriminant validity than patient satisfaction which is validated itself. This objection leads to the problem that there is no gold standard on this regard.
Fifthly, the same problem occurred concerning the reliability. As there is no gold standard in this respect to come into consideration the alternate forms method for measuring reliability was excluded. However, the single factors showed a high internal consistency, which may be an estimate for reliability.
From an ethical point of view the subject of this questionnaire is more than overdue to be examined more exactly, at the same time research in this field has to deal with subjective assessment of scientifically not exactly definable variables such as human rights. The appraisal of this questionnaire has to consider the relation to the psychiatric surrounding and the ethical complexity. Although further research is urgently needed, we assume that the CES is an important scale to fill in the gap between scientific research and ethical founded constructs in psychiatry.