Capítulo II Crónica de la Intervención
2.3 E L CASO DEL DOBLE : M ARIANA M ARÍA I NÉS
Specific problems with measurement are tied to the use of the scales used in this study. The PCL-R was the scale that provoked discussion between the two blind raters and the researcher. In particular item 3, pathological lying, item 7, shallow affect and item 8 lack of empathy were particularly difficult to score for almost every case. In the absence of clear verbal evidence these items rely on non-verbal cues for assessment. This information was not available due to the absence of video or audio records of interviews. When items could not be scored, they were omitted according to the instructions in the manual, then prorated in accordance with instruction to calculate final scores for all participants. However, future research using this scale should incorporate video or audio recording. Confidentiality issues are pertinent to this video recording. Audio or video data can be made available to courts where
research participants are the subjects of court proceedings. Specifically, this could be the case if participants were to speak about the circumstances of an alleged offence during interview. This was the reason for using written transcripts only for the purposes of this study.
The measure of mental state used for this study was undoubtedly crude and inadequate. To use diagnosis at admission excluded daily information regarding symptoms and did not allow conclusions to be drawn as to which symptoms have relationships with violence. Due to the retrospective nature of the project, the only available measure of mental state for the time that acts of violence took place was diagnosis at admission. Future studies using a prospective design would rectify this problem. Douglas et al. (1999) recommend a monitoring of symptomatology rather than diagnosis in order to accurately monitor mental state over time. Unfortunately, the variation in detail in the admission notes available for review made any meaningful recording of symptomatology at admission comparatively random. Future studies could usefully standardise the recording of symptoms on admission. Another useful measure could include symptomatology at the time of violent incidents. This would usefully explore the links between mental state and violent behaviour.
The measurement of historical, clinical and risk factors using the HCR-20 scale may have been improved if assessments of impulsivity or psychopathy at admission had been available. As these traits require a thorough assessment over time, it was not possible to include these factors in the HCR-20 scales completed. The scores were
pro-rated for these two items according to the instructions in the manual. Clinical implications of this problem are discussed in section 4.6.
The measurement of Risk factors for the R scale seems to have yielded low figures for reliability. The R scale achieved a Cronbach’s Alpha of .4095 following the deletion of item 5. An explanation for this might be the difficulty in anticipating Risk factors after discharge when patients were in hospital. Information available from admission summaries was generally focussed on current presentation and problematic behaviours in the community. Therefore, it could be suggested that this scale is difficult to score reliably early on in the course of admission.
The measurement of violence for this study took place in an environment where all therapeutic efforts were consistently mobilised to reduce violent incidents. Therefore as with many other studies of violent behaviour the rate of severe violence was low (Monahan and Steadman 1994). However, although the rates were low it is fair to say that the use of an inpatient setting holds the environment constant for all participants. Therefore, differential rates of violence are more likely to be due to internal characteristics and states than to external challenges or provocations (Belfrage et al. in press).
A second difficulty with the measure of violence for this study was the data collection from note review for the rating of the OAS scale. The scale was designed to be completed by nursing staff at the end of each shift for each participant under study. A concerted effort was made to recruit nurses to the project early on, but the
combination of nursing time constraints and the diluted responsibility that comes from being a large group meant that no records were available contemporaneously. The notes were all reviewed by the researcher and were therefore consistently rated using the same rationale and definitions. Score sheets were only completed for incidents not for each day during the first month of admission. Using notes means that the information used to complete the scale could have passed from the eyewitness to the person making the notes entry and then to the researcher reviewing the notes. The loss of detail through this process is likely to have been considerable. Possible solutions to this problem are discussed in section 4.5.
Another problem with measurement of violence was that due to low rates of violence it was not possible to split types of aggression into object directed, person directed and verbal. Larger numbers of participants and several sites may have increased the variation of violence recorded, particularly as on any one unit there is usually a limit to how many extremely violent people can be managed at one time. Therefore, different sites would have increased the variety of men at the high end of the violence scale allowing an examination of types of violence.
The measure of substance use used in this study only examines lifetime occurrence of use and not use around the time of admission. A more useful measure might have asked someone on admission what he or she had taken before coming into hospital. Analysis of urine on admission would have provided clear evidence of any substance use on admission. This analysis could have continued throughout the violence rating period.
4.4.3 Design
Possibly, the major improvement that could be made to the current design would be to predict violent behaviour prospectively rather than retrospectively. The advantage of this would be that measures would be current to the time at which violence takes place. The reason for choosing the first month of admission for the violence recording period was to ensure that the phase of admission was comparable for all participants. To predict violence during the month following recruitment would have reduced the number of participants by approximately a third, due to discharges occurring soon after recruitment to the research project. Given the time scale of the project there was not time to recruit people and then discard data because they had been discharged. Also to follow people up either in the community or at another hospital would have meant that the environment was different, not controlled and comparable across the sample. The hospital has an average length of stay of 17 months. Total bed capacity is 60 beds. To have carried out a prospective study, recruiting only new admissions and stick to the time scale of this project would have reduced the sample by approximately two thirds.