The qualitative variables were expressed by their frequency distribution and the quantitative variables (normally distributed) by their mean±standard deviation (SD).

For the comparison between the quantitative variables of the 2 study groups, Student's t-test (or non-parametric Mann–Whitney U test) was used. In the case of qualitative variables, the comparison between groups was assessed using the χ2 test (or Fisher's exact test if more than 25% of the expected values were less than 5).

Subsequently, a logistic regression analysis was carried out with a forward stepwise selection method (Wald) considering the existence, or not, of a criminal record as a dependent variable; and as independent variables those that showed statistical significance in their univariate association with the main variable (criminal record). This technique creates a probabilistic model that allows for estimation of the risk (through odds ratio [OR] provided by the model) that involves the different values of independent variables on a dichotomous dependent variable (existence, or not, of a criminal record). In our case, the use of this technique identifies the factors that may be associated with a criminal record.

For all of these tests, the accepted level of significance was 5%. Data was processed and analysed using the SPSS statistical package version 15.0 for Windows (SPSS, Chicago, IL, USA).

This study was carried out with the approval of the Independent Ethics Committee (IEC) of the Hospital Clínico San Carlos de Madrid, fulfilling all the requirements of the Declaration of Helsinki and Spanish legislation on data protection.

In the descriptive analysis of the HI with SMI sample (No.=118) it was found that up to 24.6% (n=29) had some criminal record (at the time of inclusion in the programme), and, of these, 58.6% (n=17) had been to prison at least once.

The descriptive analysis of the group of HI with SMI with a “criminal record” is shown in Table 1. And the typology of the crimes for which they had been convicted is shown in Fig. 1.

Figure 1.

Crimes committed by homeless individuals with severe mental illness.

Note: Crimes of unknown nature are not included in the figure (n=10).

(0.22MB).

In the comparative analysis of both groups (Table 1), “with a criminal record” and “without a criminal record”, statistically significant differences were observed in: the existence of pending convictions, gender, years of evolution of homelessness, co-morbidity with poly-drug use, disease awareness and severity of psychotic symptoms. Whereas no significant differences were found for the rest of the variables analysed in the study.

The model obtained with the logistic regression showed a good fit (χ2 Hosmer–Lemeshow=0.324; gl=; p=0.956), correctly classifying 82.2% of the sample with a sensitivity of 97.4% and a specificity of 20.7% with a cut-off value of 0.5 in the regression equation. The 3 variables included in the model were: gender (χ2 Wald=8.27; gl=1; p=0.004); years of evolution of homelessness (χ2 Wald=.17; gl=1; p=0.023) and history of poly-drug use (χ2 Wald=8.46; gl=1; p=0.003).

The ORs for these variables showed that being male as opposed to female increased the risk of having a criminal record ninefold (OR=9.31; 95% CI=2–42.7). A duration of homelessness of more than a year increased the risk up to fivefold (OR=5.03; 95% CI=1.2–20.1). And co-morbidity with poly-drug use quadrupled said risk (OR=4.37; 95% CI=1.2–11.7).

Finally, it was observed that of the HI with SMI who had a criminal record, 58.6% (n=17) had been convicted for “serious crimes” (which corresponded to 14% of the total sample); as opposed to 41.4% (n=12) for “minor offences”.

In the comparison of both groups (“serious crimes” and “minor offences”), statistically significant differences were found in (Table 2): age (p=0.034; t=2.2); in the “serious crimes” group the mean age was 47.5±13.4 years, while in the “minor offences” group it was 36.4±12.5 years. And nationality (p=0.008; χ2=9.7); with 17.6% (n=3) foreigners in the “serious crimes” group as opposed to 75% (n=9) foreigners in the “minor offences” group. And, although it was not statistically significant, a significant trend was found (p=0.08; χ2=12.1) (Table 2) of a greater frequency of “crimes against life (injuries, murder, etc.)” (23.5%; n=4) in the “serious crimes” group; and a greater frequency of “crimes against property (theft, robbery, damage, vandalism, etc.” in the “minor offences” group (33.3%; n=4). Finally, no significant differences were found between the two groups for the rest of the variables: gender (p=0.218; χ2=1.5), place of origin (p=0.123; χ2=4.1), type of living environment before homelessness (p=0.125; χ2=7.3), years of evolution of homelessness (p=0.289; χ2=2.5), typology of homelessness (p=0.793; χ2=0.1), level of education (p=0.430; χ2=1.7), profession (p=0.374; χ2=1.9), pending convictions (p=0.286; χ2=1.1), victim of crime (p=0.393; χ2=1.7), co-morbidity with consumption of toxic substances (p=0.822; χ2=0.1), co-morbidity with poly-drug use (p=0.460; χ2=0.5), main diagnosis (p=0.904; χ2=1), years of evolution of mental illness (p=0.260; χ2=1.2), form of evolution of mental illness (p=0.876; χ2=0.1), previous contact with mental health (p=0.158; χ2=1.9), previous psychiatric hospitalisation (p=0.555; χ2=0.3), recognition of the degree of disability (p=0.141; χ2=2.1), disability (p=0.226; χ2=1.4), co-morbidity with severe infectious or somatic disease (p=0.394; χ2=5.1 and p=0.534; χ2=0.1; respectively), disease awareness (p=0.119; χ2=2.4), attitude towards pharmacological treatment (p=0.793; χ2=0.1) and severity of somatic, psychotic, depressive and anxiety symptoms (p=0.161; t=2.1; p=0.823; t=0.2; p=0.225; t=1.2 and p=0.179; t=1.3, respectively).

In this sense, in our sample, a duration of homelessness of greater than one year increased the risk of criminal behaviour up to fivefold. This finding is very relevant since, typically, there has been a tendency to pay less attention to contextual factors than to clinical risk factors in studies on the relationship between mental illness and criminal behaviour.3 While it is true that in a previous study conducted in the United States3,11 it has been suggested that in terms of participation in the criminal justice system (arrests, convictions and imprisonment), HI with SMI tend to be more similar to the general HI population than to the general population of individuals with SMI. And this may be due to the fact that when people with SMI have no home, and are therefore visible in the public space, they are more likely to attract the attention of the public and of the security forces and, therefore, to be arrested more frequently.18 However, all conclusions should be considered tentative given the scant available studies.3

Furthermore, in our sample, co-morbidity with poly-drug use quadrupled the risk of criminal behaviour. So, just as has been consistently replicated in all the previous scientific literature in this regard,3 poly-drug use is a differential factor between both groups (“with a criminal record” and “without a criminal record”). In fact, many studies on the general population of adults with SMI have found that substance abuse is the main cause of crime and not mental illness in itself.19–23 And, in this sense, it is possible that for the HI population with SMI, several causal pathways underlie the relationship between poly-drug use and the existence of a criminal record; such as, for example, that the disinhibition effect of drug and alcohol abuse may cause some HI with SMI to exhibit problematic behaviour on the street, thereby increasing the risk of criminal altercations.3,18

Another finding that our study draws attention to is that almost a quarter of the HI with SMI in the municipality of Madrid (up to 24.6%) had a criminal record; these rates are lower than those observed in other international studies with the same population (mainly in the USA), where rates range between 28.1 and 80.0%.3 And, although the lowest rates found in other studies conducted in the United States are close to those found in our study, it should be noted that, in the case of the lowest rate19 (28.1%), it only represents convictions for serious crimes, which in the study of our sample would be considerably lower – i.e. 14.4%. This observed difference was expected, and is explained by the fact that the previous studies published refer to cities in the US states (New York,10 California,7 Washington,8 Florida4 and Missouri5), where criminalisation of the homeless mentally ill comes from having been a socio-health and legal problem for a few decades.

With regard to gender, it was foreseeable that being male (compared to being female) markedly increased the risk of having a criminal record, as has been described in all the previous literature on this matter.24

From the psychopathological point of view, it was observed that the group of HI with SMI and “with a criminal record” presented a greater apparent severity of psychotic symptoms in relation to the group of HI with SMI “without a criminal record”. These findings would be in line with other studies on criminal behaviour, where the presence and intensity of psychotic symptoms has been related to a pattern of delinquency associated with situations of homelessness, mainly when it is linked to the consumption of toxic substances.25

This psychopathological context consequently explains that criminal behaviour is also associated with a lower awareness of illness; with none of the patients in the “criminal record” group having an adequate awareness of their illness. Although, paradoxically, 72.4% of them agreed to undergo pharmacological treatment, which is explained because a large part of them were complying with involuntary outpatient treatment as a safety measure.

Another difference found in the “criminal record” group was the coexistence of pending (or open) criminal justice cases; which would support the existence of the “revolving door” phenomenon. This consists of the mentally ill subjected to safety measures being recidivists in the commission of criminal acts (usually minor), so that they come and go (often almost unceasingly) to and from the penal/prison system.26 In this regard, in the relevant scientific literature it has also been found that the number of previous convictions is one of the most common predictors of recidivism among offenders with and without SMI.3

Finally, in the study on the type of crime, the analysis of the “minor offences” group, which involved just under half of the patients with a criminal record, can provide information about the phenomenon of criminalisation. It appears that an HI with SMI is more susceptible to being accused of a minor offence3 compared to the general population. And, accordingly, it appears that foreign nationality and youth is associated with more frequent minor offences, as has also been described in other studies.

This is a retrospective study with a sample selection based on quality of data from medical records, which means that these results cannot be easily generalised.

This retrospective nature implies that there may be an under-recording of data relative to criminal records, although the selection of the sample based on the completion of all the items of the medical history would partially limit this bias.