Descheemaeker and colleagues aimed to differentiate an anaclitic and an introjective cluster of “psychological disorders” (PD) in DSM-IV PD-characteristics. In a cross-sectional study (n = 48) they investigated the relationship between subscales of the SCID-II Personality Questionnaire and the Depressive Experiences Questionnaire. Multiple linear-regression analyses were used to control for the severity of the depression - and controlling for the introjective and anaclitic dimension , “respectively” the authors add. But that is somewhat confusing - the model is: 

PQ subscale = B0 + B1(introjective) + B2(anaclitic) + B3(severity) + error

What is controlling what in this model is a matter of interpreting the regression coefficients “respectively”. However, assumptions for linear-regression analyses were not met and the conclusion are overstated. 

Van Horn et al. investigated subgroups of 148 male (domestic) violent offenders. The study aimed to investigate whether the dual diagnose group, offenders with a substance-related disorders and co-occurring disorders (n = 50), differed from offenders with a substance-related disorder (n = 23), with one or several comorbid axis I/ axis II disorders (n = 47), and offenders without an axis I or axis II disorder (n = 28). The authors conclude that the double-diagnosis delinquents did not differ from other subgroups in terms of substance-related disorders or psychopathology, but did engage more often in recidivism. Van Horn et al. report that results of Cox regression analyses indicated that merely belonging to the double-diagnosis group increased the risk of violent recidivism by a factor of 5.21. However, the tables do not support such a conclusion.

Survival analyses showed significantly higher general (60%) and violent (45%) recidivism rates in the double-diagnosis group – in other subgroups the rates were 40% to 29% for general and 35% to 15% for violent recidivism. Thus maximum differences were 25% to 45% for general and violent recidivism respectively. Two Cox regression analyses were conducted, entering 6 predictors. No variables selection procedure was used. Consequently, the regression coefficients for subgroups were estimated based on few cases. The rule of thumb (N = 10 k / p) indicates a sample size of 200 to 400 for six predictors (k) and .30 to .15 as the smallest proportion of cases (p). So the number of cases needed is 1.3 to 2.6 times the study’s sample size. Which shows in some relatively large confidence intervals (e.g. 0.79 – 9.42 and 1.10 – 10.70).

Van Horn et al. report that both models predicted recidivism significantly better than the null-model (the overall difference a.k.a. intercept-only model). But that is not surprising given the number of predictors in the final model. However, the model for general recidivism has no predictors with p-values lower than .05 and the model for violent recidivism is difficult to interpret. Apparently, the risk for offenders without an axis I or axis II disorder (HR = 1.10) is not different from the double-diagnosis reference group.

So, it looks like a 25-45% difference does not hold up in a multiple regression analyses. Bottom line: readers cannot check the reported 5.21 increase of the risk of violent recidivism in the double-diagnosis group.

van Horn JE, Eisenberg MJ, van Kuik S, van Kinderen GM (2012). Psychopathology and recidivism among violent offenders with a dual diagnosis. A comparison with other subgroups of violent offenders. Tijdschrift voor psychiatrie, 54,6, 497-507


Van der Post and colleagues report on the links between opinions about prior psychiatric treatment and risk of civil detentions in Amsterdam, the Netherlands. The authors conclude that a history of previous involuntary admission is associated with low treatment satisfaction and that more satisfaction seems to reduce the risk of civil detention. This is another way of saying that the three- to fourfold increase in the risk of repeated compulsory admission is mediated by patients’ perspectives. Unfortunately, these conclusions are not supported by the data from the Amsterdam Study of Acute Psychiatry.

In an earlier study Penterman et al. (2009) found that using the ‘Checklist Risk Emergency Service’ helped to predict aggressive behaviour of patients contacted by outpatient psychiatric emergency teams. The replication study (Penterman et al. 2013) also showed that a visual-analogue aggression scale (range 0 to 100) and one or more dangerous persons in the patient’s social environment (yes or no) are useful ‘predictors’ that arrive at 92% correct classifications (91% in the previous study).

The Journal is to be praised to give room for papers concerning the results of a replication study. However, in this case a prediction model was replicated that has no predictive value. For how good is 92% correct classifications?

Kortrijk and colleagues investigated how treatment outcome was associated with demographic factors, clinical factors, and motivation for treatment. Their observational study comprised 637 assessments of 139 patients of severely mentally ill patients treated in assertive community treatment (ACT) teams. Psychosocial outcome was routinely assessed using the Health of the Nation Outcome Scales (HoNOS). Trends over time were analyzed using a mixed model with repeated measures; mean duration of follow-up was 27.4 months (SD = 5.4).
The HoNOS total score was modelled as a function of time (treatment duration) and patient-dependent covariates. To capture a curvilinear decline indicating a more rapid change in the early months, the analyses also included a square-root transformation of time. The authors conclude that in male patients psychosocial functioning improved significantly over time, but mainly in the first months of treatment. Later on, the level of functioning appeared to stabilize. However, Kortrijk et al. missed that the model indicates a dramatic increase of psychosocial problems after 6 months.

Model misteries

feed-image Follow new articles